close

Вход

Забыли?

вход по аккаунту

?

Variable frequency microwave processing of materials for microelectronic applications

код для вставкиСкачать
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI films
the text directly from the original or copy submitted. Thus, some thesis and
dissertation copies are in typewriter face, while others may be from any type of
computer printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted. Broken or indistinct print, colored or poor quality illustrations
and photographs, print bleedthrough, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and continuing
from left to right in equal sections with small overlaps.
ProQuest information and Learning
300 North Zeeb Road, Ann Arbor, Ml 48106-1346 USA
800-521-0600
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
NOTE TO USERS
Page(s) not included in the original manuscript
are unavailable from the author or university. The
manuscript was microfilmed as received.
203
This reproduction is the best copy available.
UMI*
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
VARIABLE FREQUENCY MICROWAVE PROCESSING OF
MATERIALS FOR MICROELECTRONIC APPLICATIONS
A Thesis
Presented to
The Academic Faculty
by
Ravindra V. Tanikella
In Partial Fulfillment
o f the Requirements for the Degree
Doctor o f Philosophy in Chemical Engineering
Georgia Institute o f Technology
February 2003
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number. 3085003
UMI*
UMI Microform 3085003
Copyright 2003 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
VARIABLE FREQUENCY MICROWAVE PROCESSING OF
MATERIALS FOR MICROELECTRONIC APPLICATIONS
Approved:
Paul A. Kohl, Co-Chair
SuenA. Bidstrup-Allen, Co-Chair
Dennis W. Hess
C. P. Wong
C7ary SjM ay
DATE APPROVED & / I v /o 3 >
ii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ACKNOWLEDGEMENTS
I would like to acknowledge several people who have supported and encouraged me
during my graduate studies.
First and foremost, I would like to express my sincere
gratitude to my advisors Dr. Paul Kohl and Dr. Sue Ann Bidstrup-Allen, for their
guidance and encouragement throughout my stay at Georgia Tech. Being part o f their
research group was an invaluable learning experience. They have always been willing to
share their wealth o f knowledge and expertise with me and I am extremely grateful for
their inspiration and support. I would also like to thank members o f my thesis reading
committee Dr. Dennis Hess, Dr. John Muzzy, Dr. Gary May and Dr. C.P. Wong for their
useful suggestions.
A special note o f appreciation goes to Taehyun Sung. It has been a pleasure to work
with him and I would like to thank him for his contribution to this work. Silvia Liong,
Zhuqing Zhang, and Krishna Tunga are also acknowledged for their input to this work. I
would also like to thank the past and present members o f the Kohl-Bidstrup-HessHenderson research groups in MiRC. Their friendship and support have made my stay at
Georgia Tech very enjoyable. I would especially like to thank Rahul Manepalli, Sairam
Agraharam, Agnes Padovani, Ankur Agrawal, Sam Park, Tazrien Kamal, Punit
Chiniwalla and Reena Agarwal. Their personal and professional contribution has been
invaluable. A special note o f thanks goes to Gary Spinner and the rest o f the MiRC
Cleanroom staff for all their assistance over the last few years.
iii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Finally, I would like to thank my family for their love and support. I owe all my
success in life to my parents who have always believed in me and encouraged me. I
would like to thank my sister Sunitha and brother Rajasekhar for all the love and
affection. I certainly would not have been able to accomplish this without their support
and encouragement.
iv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TABLE OF CONTENTS
LIST OF FIGURES______________________________________________________IX
LIST OF TABLES___________________________________________________ XVIII
SUMMARY____________________________________________________________XX
INTRODUCTION_________________________________________________________1
BACKGROUND_________________________________________________________12
2.1 THEORY OF MICROWAVE PROCESSING.......................................................... 13
2.1.1 Microwave systems and instrumentation............................................................ 14
2.1.2 Molecular mechanisms o f polarization...............................................................16
2.1.3 Macroscopic effects o f electromagnetic field on m aterials...............................19
2.1.4 Energy conversion in a microwave field.............................................................24
2.2 LITERATURE REVIEW.......................................................................................... 28
EXPERIMENTAL METHODS AND PROCEDURES________________________ 37
3.1 VARIABLE FREQUENCY MICROWAVE FURNACE: EXPERIMENTAL
SETUP...............................................................................................................................37
3.1.1 Temperature measurement and control in the VFM........................................ 40
3.2 CHEMICAL STRUCTURE AND PROPERTIES................................................... 44
V
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.2.1 Fourier Transform Infrared (FTIR) A nalysis.................................................... 44
3.3 OPTICAL PROPERTIES.......................................................................................... 46
3.3.1 In-plane and through-plane refractive indices and birefringence....................46
3.4 ELECTRICAL PROPERTIES.................................................................................. 49
3.4.1 Relative permittivity (&) & dielectric loss......................................................... 49
3.5 MECHANICAL PROPERTIES................................................................................ 51
3.5.1 Residual stress.....................................................................................................51
3.5.2 Young's modulus..................................................................................................54
3.6 THERMAL PROPERTIES........................................................................................ 56
3.6.1 Thermal stability..................................................................................................56
3.6.2 Pyrolytic mass spectrometry............................................................................... 57
3.6.3 Differential scanning calorimetry (DSC)............................................................58
3.6.4 Thermo mechanical analysis............................................................................... 59
3.7 PHYSICAL PROPERTIES....................................................................................... 60
3.7.1 M oisture uptake...................................................................................................60
RAPID CURING OF BENZOCYCLOBUTENE BY VARIABLE FREQUENCY
MICROWAVE PROCESSING____________________________________________ 62
4.1 BENZOCYCLOBUTENE (BCB-CYCLOTENE 3022-63).................................... 62
4.2 RESULTS...................................................................................................................65
4.2.1 FTIR studies on VFM curedfilm s....................................................................... 66
4.2.2 Comparison ofproperties o f thermal and VFM cured BCB film s.................... 76
4.3 DISCUSSION.............................................................................................................81
vi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
87
4.4 CONCLUSIONS
LOW TEMPERATURE RAPID CURING OF POLYIMIDES ON SILICON BY
VARIABLE FREQUENCY MICROWAVE PROCESSING___________________89
5.1 RESULTS................................................................................................................... 94
5.1.1 P I 2611................................................................................................................. 94
5.1.2 Negative tone photosensitive polyimides (HD 4000 andXP 7001):............... 107
5.1.3 Positive tone photosensitive dielectrics (CRC 8650, PWDC1000 and
PW 1200):.....................................................................................................................117
5.2 DISCUSSION........................................................................................................... 122
5.3 SUMMARY...............................................................................................................137
RAPID CURING OF POSITIVE TONE PHOTOSENSITIVE
POLYBENZOXAZOLE BASED DIELECTRIC RESIN BY VARIABLE
FREQUENCY MICROWAVE PROCESSING._____________________________ 139
6.1 POLYBENZOXAZOLEKSUMIRESIN EXCEL CRC 8650)...............................140
6.2 RESULTS..................................................................................................................143
6.3 DISCUSSION........................................................................................................... 162
6.4 SUMMARY AND CONCLUSIONS.......................................................................167
RAPID CURING OF POLYIMIDES ON ORGANIC SUBSTRATES BY
VARIABLE FREQUENCY MICROWAVE PROCESSING__________________168
7.1 RESULTS AND DISCUSSION............................................................................. 170
7.1.1 VFM heating characteristics o f organic substrates......................................... 170
vii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7.1.2 Thermal stability o f organic substrates............................................................173
7.1.3 Substrate selectivity o f VFM cured film s..........................................................186
7.1.4 Anhydride formation and other side reactions................................................ 204
7.1.5 Effect o f solvent evolution on imidization........................................................ 208
7.1.6 Solvent effect on VFM heating characteristics................................................ 211
7.2 CONCLUSIONS.................................................................................................... 220
SOLDER REFLOW AND ELECTRICALLY CONDUCTIVE ADHESIVE
CURING BY VARIABLE FREQUENCY MICROWAVE PROCESSING______222
8.1 RESULTS AND DISCUSSION.............................................................................. 222
8.1.1 Electrically conductive adhesive curing by VFM processing......................... 222
8.1.2 Solder reflow by VFM processing..................................................................... 235
8.2 SUMMARY.............................................................................................................. 250
SUMMARY AND FUTURE DIRECTIONS________________________________ 252
9.1 SUMMARY.............................................................................................................. 252
9.2 RECOMMENDATIONS FOR FUTURE W ORK................................................. 255
APPENDIX A __________________________________________________________ 261
viii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF FIGURES
Figure 1 Thermal cure cycle for polyimide PI 2611 in a conventional convection thermal
furnace..............................................................................................................................5
Figure 2 Electromagnetic spectrum and frequencies, adapted from Gardiol [28]..............14
Figure 3 Dielectric response o f a polymer........................................................................... 22
Figure 4 Variable Frequency Microwave Fumace-Microcure™ 2100...............................39
Figure 5 Schematic diagram o f the Metricon prism coupler............................................... 48
Figure 6 Detector intensity as a function o f angle o f incidence.......................................... 48
Figure 7 Equivalent circuit model for measuring the dielectric properties in thin films.. SO
Figure 8 Polymer thin-film test structure for tensile testing................................................56
Figure 9 Chemical structure o f BCB monomer...................................................................64
Figure 10 Thermal cure reaction in BCB [89]: Ring opening followed by Diels -Alder
addition reaction............................................................................................................ 64
Figure 11 The effect o f central frequency on the heating characteristics o f BCB on silicon
at a constant power o f 100 W....................................................................................... 66
ix
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 12 Comparison o f FTIR spectra o f soft-baked, VFM and thermally cured BCB
samples........................................................................................................................... 68
Figure 13 Comparison o f percent conversion o f films cured in the thermal oven, hot plate
and VFM furnace under identical conditions...............................................................70
Figure 14 FTIR spectra o f BCB films cured in the VFM furnace for different times at a
temperature of225°C..................................................................................................... 71
Figure IS Progress o f VFM cure reaction o f BCB with time at different temperatures
(from FTIR data)............................................................................................................ 73
Figure 16 Plots o f-L n (l-x ) vs time at different cure temperatures to determine the rate
constants......................................................................................................................... 74
Figure 17 Plot o f Ln(k) vs 1/T to determine the apparent activation energy Ea, for the cure
reaction........................................................................................................................... 75
Figure 18 Comparison o f in-plane index o f refraction o f BCB films cured in a furnace,
hot plate and VFM under different processing conditions.......................................... 76
Figure 19 Dielectric constant o f VFM and thermally cured BCB films processed under
different conditions........................................................................................................ 77
Figure 20 Residual stress o f VFM and thermally cured films as a function o f percent
conversion.......................................................................................................................79
x
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 21 FTIR spectra o f a) softbaked, b) standard thermally cured, and 3) lowtemperature VFM cured PI 2611 films......................................................................... 95
Figure 22 Effect o f cure temperature on the percent imidization o f PI 2611, HD 4000 and
XP 7001. Ramp rate: 15°C/min, Hold time: 5 min..................................................... 98
Figure 23 Effect o f cure temperature on the birefringence o f VFM cured PI 2611...........99
Figure 24 Effect o f ramp rate on the birefringence of VFM cured PI 2611 films............100
Figure 25 Effect o f ramp rate on mechanical properties o f VFM cured PI 2611 films.. 104
Figure 26 Comparison o f low temperature VFM and thermally cured samples with
standard thermally cured PI 2611................................................................................106
Figure 27 FTIR spectra o f a) softbaked, b) standard thermally cured, and 3) VFM cured
HD 4000 film................................................................................................................ 108
Figure 28 In-situ curing o f HD4000 and XP7001 in the DSC...........................................109
Figure 29 FT-IR spectra o f PWDC 1000 (a) after pre bake and before exposure, and (b)
after standard thermal cure......................................................................................... 118
Figure 30 Schematic representation o f a) cure reaction o f polybenzoxazole (PBO)
formation and b) DNQ reaction-Wolff rearrangement.............................................. 142
Figure 31 DSC scan o f a PBO sample under a constant ramp rate o f 3°C/min------------ 144
xi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 32 FTIR spectra o f a) softbaked b) standard thermal cured and c) VFM cured
films: A) 900-2000 cm'1 and B) 2000-4000 cm'1.......................................................146
Figure 33 FTIR spectra o f thermally cured PBO films showing an increasing absorbance
at 1054 cm'1with cure temperature.............................................................................150
Figure 34 Percent conversion achieved in thermally cured PBO films as a function of
cure temperature estimated from FTIR analysis with peak height at 1054 cm '1 and a)
1600 cm'1 and b) 963 cm'1 as an internal standard.....................................................151
Figure 35 Comparison o f percent conversion achieved in VFM and thermally cured PBO
films from FTIR analysis. (VFM 3-Step: 30°C to 150°C 5min, 30°C to 250°C 10
min, 10°C/min to 275°C 30min).................................................................................. 152
Figure 36 Photoreaction o f PBO precursor films: a) FTIR spectra b) UV-visible spectra.
.......................................................................................................................................155
Figure 37 TMA plot o f VFM cured PBO film cured at 320°C for 5 min. TMA ramp rate:
5°C/min. Load: 0.05 N.................................................................................................157
Figure 38 TGA scan o f a standard thermal cured PBO film..............................................160
Figure 39 Mass spectroscopy results o f PBO films ramped at 10°C/min to 450°C......... 161
Figure 40 VFM Heating rates o f different substrate materials at a constant power o f 200
W, central frequency o f 6.245 GHz, 10% bandwidth and O.lsec sweep time.........172
xii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 41 The effect o f microwave power on the heating rate o f blank FR4...................173
Figure 42 Dynamic TGA o f an FR4 board at ramp rates o f 3, 10 and 15°C/min in a
nitrogen atmosphere..................................................................................................... 175
Figure 43 Time-temperature dependence o f the thermal stability o f FR4 board under a
nitrogen atmosphere..................................................................................................... 176
Figure 44 Weight loss o f FR4, BT and CF-Epoxy boards under dynamic TGA at a ramp
rate o f 10°C/min in a nitrogen atmosphere................................................................. 178
Figure 45 Comparison o f heating rates o f FR4 board with and without Polyimide film at
400 W constant microwave power..............................................................................181
Figure 46 Infrared spectra o f PI2611 Films: a) Soft-baked, b) standard thermally cured on
silicon ramped at 3°C/min to 350°C and held for 1 hour at 350°C, c) VFM cured on
FR4 substrate, ramped at 15°C/min to 200°C and held for 5 min at 200°C, d)
thermally cured on FR4 substrate, ramped at 3°C/min to 200°C and held for 1 hour at
200°C............................................................................................................................ 183
Figure 47 Comparison o f the extent o f imidization o f PI26U films (estimated from
FTIR) cured on different substrates in the VFM oven at a constant ramp rate o f
15°C/min and 5 min hold time at 175 and 200°C.......................................................186
xiii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 48 Comparison o f the extent o f imidization o f PI2734 films (estimated from
FTIR) cured on different substrates in the VFM oven at a constant ramp rate o f
15°C/min and 5 min hold time at 175 and 200°C....................................................... 188
Figure 49 Thermo-mechanical analysis o f PI 2611 VFM cured at a ramp rate o f 15°C/min
to 200°C for 5 minutes. TMA conditions: Ramp rate: lO°C/min; Load: 0.05 N .... 195
Figure 50 Dynamic mechanical analysis o f PI 2611 VFM cured on FR4 substrate. VFM
cure conditions 15°C/min ramp to 200°C 5min hold. DMA test conditions: ramp
rate: 5°C/min. frequency: 1 Hz....................................................................................196
Figure 51 Weight loss o f VFM and thermally cured films under a TGA ramp o f 10°C/min
to 500°C. Cure conditions: 5 min at 200°C.................................................................198
Figure 52 Mass spectrometry results o f PI 2611 film VFM cured at 200°C for 5 min on an
FR4 substrate................................................................................................................201
Figure 53 Mass spectrum o f PI 2611 film VFM cured at 200°C for 5 min on an FR4
substrate........................................................................................................................202
Figure 54 Two reaction pathways for polyamic acid BPDA-PDA: a) polyimide formation
(desirable) b) anhydride formation (undesirable side reaction)................................ 205
Figure 55 FTIR spectra o f PI 2611: Anhydride formation................................................ 206
xiv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 56 Ramp rate dependence o f solvent decomplexation and imidization a) 5°C/min,
b) 10°C/min, and c) 15°C/min. (Inset: schematic representation o f solvent -amide
complexation mechanism............................................................................................210
Figure 57 Effect o f solvent content on the heating rates o f polyimide film coated on FR4
substrate subjected to a constant power o f 400 W (1): Temperature o f polyimide
film, (2) Temperature o f FR4 board (3) temperature o f FR4 without polyimide film.
...................................................................................................................................... 213
Figure 58 Comparison o f thermal stability o f PI 2611, HD 4000 and XP 7001 films cured
on BT substrate. Cure conditions: l5°C/min to 220°C, 5 min hold. TGA ramp rate:
10°C/min..................................................................................................................... 219
Figure 59 VFM heating o f ECA with 70% loading o f silver flakes (type A: < 10 pm) at a
constant microwave power o f 200 W.........................................................................225
Figure 60 Skin depth o f some common metals as a function o f frequency..................... 226
Figure 61 Effect o f filler loading on the VFM heating characteristics o f ECA at a constant
microwave power o f 200W.........................................................................................228
Figure 62 Effect o f particle size on the VFM heating characteristics o f EC As...............230
Figure 63 VFM heating rates o f ECAs with bimodal filler distribution.......................... 232
XV
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 64 Test structures for ECA cure and attachment: A) mask features and B) top and
bottom view o f silicon die attached to glass substrate by VFM curing o f ECA at 200
W for 2 min.................................................................................................................233
Figure 65 A Typical reflow profile used for Sn-Pb solder alloys in a convection reflow
oven............................................................................................................................. 236
Figure 66 Typical bumping process for electroplated solder............................................238
Figure 67 SEM micrographs o f electroplated solder bumps after resist strip and before
reflow........................................................................................................................... 240
Figure 68 A typical EDS spectrum of an electroplated Sn/Pb solder bump.................... 241
Figure 69 VFM temperature and power profile for solder reflow.................................... 243
Figure 70 SEM micrographs o f VFM reflowed solder bumps......................................... 244
Figure 71 ANSYS finite element modeling comparing the strain in A) convection oven
reflowed solder and B) VFM reflowed solder. (DNP: distance from neutral plane).
......................................................................................................................................249
Figure 72 Mass spectroscopy results o f polyimide HD 4000 ramped at 10°C/min to
450°C........................................................................................................................... 263
Figure 73 Mass spectroscopy results o f polyimide XP 7001 ramped at 10°C/min to
450°C........................................................................................................................... 264
xvi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 74 Mass spectrum of polyimide HD 4000 at probe temperature o f 450°C and
probable parent species (tetra-ethyleneglycol-dimethacrylate)............................... 265
Figure 75 Mass spectra of CRC 8650 film at probe temperature o f 400°C and probable
parent species (2,2’-methylenebis-phenol)................................................................ 266
Figure 76 Mass spectrum o f polyimide PW 1200 at probe temperature o f 452°C and
probable parent species (4 ,4 ’, 4”-ethylidynetris-phenol)......................................... 267
xvii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF TABLES
Table 1 Characteristic infrared absorbance peaks for BCB cure reaction and probable
peak assignments........................................................................................................... 67
Table 2 Comparison o f properties of BCB films cured by the standard thermal cure, 1 hr
at 250°C and VFM cure o f 5 min at 250°C...................................................................81
Table 3 Material processing conditions................................................................................ 93
Table 4 Polyimide characteristic IR peaks and probable assignments............................... 96
Table S Effect o f VFM cure conditions on the optical and thermal properties o f polyimide
PI 2611 films................................................................................................................ 102
Table 6 Effect o f cure conditions on the optical and thermal properties o f HD 4000.... 110
Table 7 Effect o f cure conditions on the optical and thermal properties o f XP 7001.... 111
Table 8 Effect o f cure conditions on the Young’s modulus o f VFM and thermally cured
HD 4000 and XP 7001 polyimide films..................................................................... 115
Table 9 Effect o f cure conditions on the residual stress o f VFM and thermally cured HD
4000 and XP 7001 polyimide films...........................................................................116
Table 10 Effect o f cure conditions on the percent reaction and thermal stability o f VFM
and thermally cured CRC 8650 films.................................................
xviii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
119
Table 11 Effect o f cure conditions on the thermal properties and percent imidization o f
PWDC1000 and PW 1200 polyimide films................................................................121
Table 12 Peak assignments o f PBO and the corresponding changes occurring during the
cure reaction................................................................................................................. 147
Table 13 Comparison o f properties o f VFM and thermally cured films cured at 320°C and
275°C............................................................................................................................ 159
Table 14 Extent o f imidization achieved in polyimide PI 2611 films cured on blank FR4
substrate by VFM and conventional thermal furnace under different cure conditions.
.......................................................................................................................................185
Table 15 Comparison o f electrical and optical properties o f VFM cured, standard
conventional oven cured and hotplate cured films.....................................................191
Table 16 Percent imidization and residual solvent in VFM cured PI 2611......................215
Table 17 Comparison o f properties o f PI 2611 films cured on FR4 and BT substrates. 217
Table 18 Shear strain in solder ball without underfill....................................................... 246
xix
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
SUMMARY
Polymer dielectrics such as polyimides are widely used in the microelectronics
industry for a variety o f applications such as interlevel dielectrics and stress buffer
passivation layers. Typically, low dielectric constant, good chemical resistance, excellent
mechanical strength and high thermal stability are essential requirements for polymers to
be used for these applications. In order to achieve the desired properties, these polymers
are subjected to cure cycles that involve high temperature treatments (~ 300 to 400°C)
and last up to several hours.
Rapid curing alternatives are necessary to increase
throughput and lower the fabrication cost o f these devices. There is also a need to reduce
the thermal budget in advanced electronic devices to mitigate the stress induced due to
the coefficient of thermal expansion (CTE) mismatch between the silicon substrate and
the packaging compounds.
Novel low-temperature selective processing methods are
required to address these issues.
In this study, variable frequency microwave (VFM) curing is investigated as a novel
rapid, low-temperature curing alternative to conventional thermal curing.
Several
commercially available polymer dielectrics with different backbone chemistries currently
o f interest in the microelectronics industry such as benzocyclobutenes (BCB), polyimides
and polybenzoxazoles were chosen for this study. The kinetics and mechanism o f the
cure reactions in microwave-processed films were investigated. The chemical changes
occurring in the film during the cure reaction were monitored by Fourier transform
infrared spectroscopy.
The optical, electrical and thermo-mechanical properties o f
XX
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
microwave-cured films were characterized and compared to those o f thermally cured
films to determine the effectiveness o f microwave curing. The thermal stability o f rapid,
low-temperature VFM cured films was studied by thermogravimetric analysis. Structureproperty relationships and their dependence on processing conditions were investigated.
The results from this study show that rapid VFM curing o f polymer dielectrics is
feasible giving properties comparable to traditional thermal curing in a much shorter cure
time. Studies on the kinetics o f VFM curing o f benzocyclobutene (BCB) showed a 30%
lower apparent activation energy as compared to thermal curing.
Low-temperature
(275°C) VFM curing o f rigid rod polyimide PI 2611 on silicon showed significant
improvement in thermal stability over low-temperature thermally cured films with a high
degree o f orientation and properties comparable to films cured thermally at 350°C in a
conventional thermal oven. Studies on photosensitive dielectrics showed that backbone
flexibility, structure, reaction temperature and photochemistry are critical factors that
determine the effectiveness o f low-temperature microwave processing. The feasibility o f
curing high performance dielectrics such as polyimides on organic substrates was
demonstrated. The solvent content was found to be the critical factor affecting both VFM
heating characteristics as well as the final film properties. VFM curing and attachment o f
electrically conductive adhesives and solder reflow were investigated.
Thermo­
mechanical modeling o f die-attachment showed the potential benefits o f low-temperature
VFM processing.
xxi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER I
INTRODUCTION
The demand for smaller, faster and low-cost electronic devices has driven the
microelectronics industry towards new materials and processing techniques.
Novel
organic, inorganic, and porous materials have come to replace the traditional dielectric
materials such as silicon dioxide and silicon nitride due to their low dielectric constant
(k), ease o f processing, good planarization, and superior electrical, mechanical and
chemical properties. These dielectric materials are used for several applications such as
inter-level dielectrics, passivation layers, encapsulants and underfills.
The interconnect delay and power consumption in electronic devices depend on the
dielectric constant o f the insulator between the interconnects.
This has driven the
semiconductor industry towards developing novel low-k materials. Polymers, spin-on
glasses and porous materials are potential candidates. SiLK™, an aromatic organic spinon material, has already been integrated in chip-level applications and porous versions
are being developed for future devices [1].
Traditionally, inorganic dielectrics such as silicon dioxide and silicon nitride have
been used as passivation layers for integrated circuits fabricated on silicon [2]. However,
there are several limitations with the use o f these materials with respect to depositing
thick, planarizing, stress-free and pinhole-free layers for passivation.
1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Further, the
coefficient o f thermal expansion (CTE) mismatch between the silicon die with the
inorganic passivation and the packaging compounds induces stress in the passivation
layer and results in cracking and delamination.
This can also drastically affect the
reliability o f the devices as moisture and ionic contaminants can permeate through such
cracks causing corrosion and eventually device failure. New polymeric based materials
have been developed for application as a ‘stress buffer passivation’ layer over the
primary silicon nitride passivation to address these issues and improve the device
reliability [3]. The ability to coat thick, defect-free, and planarizing layers with excellent
electrical, thermo-mechanical properties and chemical resistance makes polymers ideally
suited for these applications. Polymers can also serve as adequate alpha particle barriers
protecting the device from ‘soft-errors’.
Over the past several years, semiconductor technology has progressed to a point
where in addition to protecting the die and interconnecting signal and power to the
external circuit, the packaging has come to play an integral role in affecting the
performance o f the semiconductor device [4]. The increase in the number o f chip-topackage interconnections requires novel packaging schemes.
Direct chip attachment
using flip chip interconnections offers several benefits such as low cost, high I/O density,
and superior electrical performance.
In order to adapt traditional peripheral design
architectures meant for wire bonding to these new area array interconnect configurations,
the final metal bond pads need to be redistributed throughout the face o f the die. The
excellent insulating properties o f polymers make them well suited for application as
redistribution layers.
2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In order for polymers to be used for these applications, several key property
requirements need to be satisfied.
Low residual stress is desirable to prevent
delamination and stress induced cracking. A suitable modulus and CTE to minimize
stress and a high elongation to break are essential. Excellent chemical resistance to
organic and inorganic chemicals used in the subsequent processing steps such as
metallization and assembly is required. For inter-level dielectric applications, such as
those in multi-chip modules (MCMs) and redistribution layers, a low dielectric constant
is required as the signal propagation speed is inversely proportional to the square root o f
the dielectric constant o f the surrounding insulator [5]. For high-frequency devices, a
dielectric material with a low dielectric loss over a wide range o f frequencies is vital.
Excellent adhesion to various materials such as inorganics, organics and metals is
essential for long-term reliability. Finally, the polymers should be easy to process in
thick, defect-free layers.
Several polymeric materials are currently available commercially which satisfy some
o f the above requirements. Polyimide based materials have been extensively studied and
are the most widely used materials for device passivation layers and interlevel dielectric
applications in electronic packaging [6]. Several other novel chemistries are also being
considered due to their superior performance in certain aspects.
These include
benzocyclobutenes, polybenzoxazoles, polynorbomenes, polyarylene ethers, epoxies and
siloxanes [7, 8, 9,10, 11].
For example, benzocyclobutenes offer a lower dielectric
constant than polyimides, polybenzoxazoles show superior chemical resistance and
epoxies are desirable for their low cost and excellent adhesion.
3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
However, one o f the major issues with the use o f these materials is the time required
for processing. In order to attain the desired properties, most o f these dielectric materials
need to be processed (cured) at a high temperature to carry out chemical reactions and
drive off the casting solvent completely. These high temperature annealing steps can last
for several hours. For example, Figure I shows the thermal cure profile o f polyimide PI
2611, a commonly used polyimide. The cure cycle lasts up to 4 to 5 hrs in a traditional
thermal oven. Moreover, typical MCM-D structures have up to 6 to 7 layers o f inter­
level dielectrics. Thus, the lengthy processing times o f polymers pose a severe limitation
on the total cost and process time in the fabrication o f these devices.
Photosensitive polyimides (PSPIs) have been routinely used as the polymers o f
choice for high volume multi-level processes [12] as they simplify the processing by
eliminating an entire photolithography step, thereby reducing the manufacturing cost.
Traditionally, negative tone PSPIs have been used for these applications. More recently,
positive acting aqueous base-soluble PSPIs are being investigated as they offer distinct
processing and performance advantages over negative tone systems [13]. Positive tone
PSPIs allow the use o f the same reticle and developer previously used for non­
photosensitive polyimides, making them easy to integrate with traditional processes.
Further, the sidewalls profiles o f positive tone PSPI are better suited for subsequent
metallization steps. There is also interest in moving towards aqueous base developer
based systems due to environmental concerns with the use o f organic solvents typically
used for negative tone systems and the associated volatile organic compounds.
4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Thermal Cure Cycle for PI 2611
^ 4001
v
tfa 3002 200
-
H
0
1
3
2
4
5
6
T im e (h o u rs)
Figure 1 Thermal cure cycle for polyimide PI 2611 in a conventional convection thermal
furnace.
Longer cure times and higher cure temperatures are required for PSPIs as compared
to the non-photosensitive polyimides. The more stringent cure conditions are required to
evolve the high molecular weight alcohol and photoreaction byproducts from the PSPI
film [14].—Therefore, it is desirable to have rapid processing techniques, which give
comparable or better properties as compared to conventional thermal processing.
With advances in silicon technology, there is an increased interest in reducing the
thermal budget [IS].
High processing temperatures for the passivation layers are
undesirable as they result in increased thermo-mechanical stress and can degrade device
characteristics. This is especially important with the use o f low-k dielectrics as high
5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
temperature processing induces delamination and cracking o f the dielectric layer. High
temperatures are also not suitable for special substrates used in high-speed applications
such as GaAs. In such applications, cure temperatures less than 300°C are required [12,
16]. There have been several attempts to develop rapid and low-temperature curing
polymers for these applications; however, these materials exhibited inferior performance
in certain film properties [17,18]. Therefore, novel processing techniques are required
which will allow rapid and low-temperature processing o f these traditional high
performance materials and still give properties comparable to those obtained via the
traditional lengthy high temperature cure methods.
Advances in semiconductor manufacturing have increased the performance and
functionality o f electronic devices. The power and signal requirements in these advanced
high performance microelectronic devices have resulted in a significant increase in the
density o f interconnects between the chip and the substrate, which results in more
complex packaging [19]. High density interconnects and substrates are required to fully
utilize the rapid advances made in silicon technology.
At the same time, while the
transition to 300 mm wafers has reduced the cost per die, the increased complexity in
packaging has resulted in a tremendous increase in the packaging cost. Novel high
density packaging technologies should therefore also be cost effective.
Currently, most microvia build-up layers in high-density packages use epoxy-based
dielectrics and low cost organic substrates like FR4 [20]. Future requirements o f higher
interconnect density, greater thermal stability (to accommodate the transition to lead-free
solder) and improved thermo-mechanical stability necessitate the use o f high
6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
performance dielectrics with a lower dielectric constant such as polyimides. However, as
discussed earlier, polyimides require annealing treatments at high temperatures that are
well above the degradation temperature o f traditionally used low-cost epoxy-based
organic substrates. Novel low temperature processing techniques are required to apply
these materials for use as dielectric build-up layers on organic substrates.
Low-temperature selective processing is also o f interest in the microelectronics
industry for processing conductive adhesives for die attach and surface mount
applications. Electronic components such as resistors, capacitors, inductors, transistors,
and integrated circuits are typically surface mounted on the printed circuit board. Solderpaste is commonly used for surface mounted soldering o f components to the circuit
board. The solder upon heating melts or reflows and upon cooling, solidifies and bonds
the components permanently to the board. Conventionally, this process is done over a
temperature cycle typically in a belt furnace, and the entire package is subjected to the
full oven temperature. This can affect the properties o f the board and other components.
It is desirable to selectively reflow solder without affecting the substrate or other
components.
In order to address these issues, alternative ways o f processing these materials are
currently being investigated. Microwave processing is one such technique, which has
been shown to drastically reduce the processing time o f materials, hi this study, the
feasibility o f rapid low-temperature curing o f high performance dielectrics by variable
frequency microwave processing has been investigated.
7
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Microwave processing has unique characteristics that are not observed in
conventional thermal heating o f materials.
The differences include rapid heating,
selective heating o f materials through differential absorption, penetrating radiation,
controllable electric field distributions, and self-limiting reactions [21].
Microwave
processing overcomes some o f the limitations o f thermal processing such as penetration
depth and uniformity o f the temperature distribution [22].
There is evidence of
enhancement of processes due to microwaves alone, such as faster kinetics in polymers
and new reactions in synthetic chemistry.
There are major differences between
microwave and thermal processing. In the case o f conventional heating, the ambient is
maintained at the required process temperature.
On the other hand, in microwave
processing, heat is transferred within the material by microwave absorption.
How a
material heats depends critically on the dielectric properties o f the material, how the
properties change with temperature, and on the microwave field distribution in the cavity.
This provides an opportunity to selectively heat materials.
There have been a number o f reports where microwave processing resulted in better
properties than conventional thermal processing. Microwave processing has been shown
to be more efficient than convection heating in terms o f cost and energy, and higher
throughput. However, there are limitations to single frequency microwave processing o f
materials due to the formation of standing waves.
Moreover, thermal runaway and
uneven heating in single frequency systems are problematic.
Many o f these limitations can be overcome in a variable frequency microwave
furnace (VFMF), first developed at the Oak Ridge National Laboratories (ORNL). The
8
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
unique feature of VFM heating, as compared to conventional microwave heating, is the
ability to quickly and repeatedly step through a range o f frequencies (4096 frequencies
over a 1.15 GHz range every 0.1 seconds).
This stepping process provides a time-
averaged uniformity in the energy distribution throughout the cavity and thereby
eliminates the nonuniformities in temperature that occur in single frequency microwave
chambers [23]. The VFM technique also allows metals and conducting materials to be
placed in the microwave cavity. By cycling through thousands o f frequencies in less than
one second, the residence time o f any established wave pattern is on the order of
microseconds and problems with charge build up and arcing are eliminated [24].
This study focuses on the details o f VFM processing as a rapid, low-temperature
alternative for curing high-performance dielectric materials for microelectronic
applications with an emphasis on understanding 1) the effect o f microwave processing on
the kinetics and mechanisms o f the cure reaction and 2) the underlying structure-property
relationships in microwave-processed films and their dependence on the processing
conditions. The feasibility o f selective processing o f conductive adhesives and solder
reflow by VFM processing was also investigated.
In order to study the feasibility o f rapid low-temperature VFM curing o f polymer
dielectrics, several commercially available polymers were chosen for curing studies.
These include: Cyclotene benzocyclobutene (also known as BCB) from Dow Chemical
Company, PI 2611, HD 4000 and XP 7001 from HD Microsystems, CRC 8650 from
Sumitomo Bakelite Co., and PWDC 1000 and PW 1200 from Toray-Dow Coming Inc.
These dielectrics are representative o f the different chemistries o f interest in the
9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
microelectronics industry for the applications discussed above. Some o f the important
characteristics o f these polymers and their chemistry are discussed in the following
chapters.
For ease o f organization, this work is presented in nine different chapters. Chapter II
provides a brief background o f the theory behind microwave processing.
The
fundamentals o f microwave-material interactions and the different polarization
mechanisms induced in materials under an applied electromagnetic field are reviewed. A
literature review o f previous work in the field o f microwave processing o f materials is
also presented. Chapter m describes in detail the experimental set up used in this work
and the different experimental techniques used to characterize the chemical, optical,
electrical, mechanical, and thermal properties o f cured films.
The results from
microwave curing o f polymer dielectrics are discussed in Chapters IV through VII.
Chapter IV presents the details o f the reaction kinetics study on microwave-cured
benzocyclobutene and a comparison o f the chemical, electrical and mechanical properties
o f VFM and thermally cured BCB films. Chapter V discusses the results from studies on
rapid low-temperature VFM curing o f polyimides on silicon.
Structure-property
relationships and their dependence on VFM cure conditions are discussed. Chapter VI
details the studies on VFM curing o f CRC 86S0, a positive-tone polybenzoxazole
chemistry based dielectric resin. Results from rapid VFM curing o f polyimides on lowtemperature organic substrates are presented in Chapter VII. The factors affecting lowtemperature polyimide curing and the impact o f solvent on the heating characteristics and
resulting properties are discussed. The results from microwave processing o f silver filled
10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
epoxy adhesives and solder are discussed in Chapter VIII.
A brief summary and
recommendations for future work are provided in Chapter IX.
11
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER II
BACKGROUND
Since the first discovery of material heating by microwaves in the late 1940s, several
industrial applications have used microwaves for material processing. The early uses of
microwave heating have been in the food industry where microwave systems have been
employed for tempering, thawing, drying, pasteurization, and sterilization processes.
More recently, microwave processing has been utilized extensively in the ceramic,
rubber, and plastic industry. Rubber vulcanization was the first industrial application of
microwaves for polymer processing and continues to be the single most significant
commercial application for polymers [25]. Microwave systems have since been applied
to process several thermosetting and thermoplastic polymers and polymer composites.
Medical and biomedical application o f microwave technology is currently an active area
o f research. The advent o f VFM technology has opened new avenues for the application
o f microwaves in advanced material processing for die-attach and bonding in surface
mounting, optoelectronics and other applications in the electronic packaging industry
[26]
The use o f microwave energy in the processing o f materials offers some distinct
advantages over conventional heating. The advantages include rapid, volumetric and
selective heating. However, the application o f this technology in the microelectronics
12
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
industry is limited because o f the lack o f an understanding o f the interaction between
microwaves and different materials, and hence the lack o f control over the processes.
While recent developments in equipment manufacturing have allowed tools with
reasonable control over the required process parameters, little research has been focused
on understanding the mechanisms involved in microwave processing.
A thorough
understanding o f the underlying principles would broaden the scope o f application o f this
technology.
2.1 Theory o f microwave processing
Microwaves are electromagnetic waves in the frequency range 300 MHz to 300 GHz
(Figure 2) [27,28] of the electromagnetic spectrum, corresponding to wavelengths in the
range 1 m to 1 mm. This includes the region between the far infrared region and radio
waves. It is often categorized under three bands o f frequencies: the ultra high frequency
band (UHF) from 300 MHz to 3 GHz, the super high frequency band (SHF) from 3 GHz
to 30 GHz and the extremely high frequency band (EHF) from 30 GHz to 300 GHz. The
most commonly used frequency in household ovens and most industrial and scientific
applications is 2.45 GHz. More recently, microwave ovens with a broad frequency band
from 0.8 to 18 GHz have become available for material processing [29,30].
13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Frequency (Hz)
104
106
108
1010 1012 IO14 1016
M icrowaves
Radiowaves
106 104 102
:
SH F
10 cm
1022
X-rays .G am m a Rays
Infrared
:
Wavelength (cm)
3 GHz
UHF
1020
1 ,10 -2 10-4 io -6 io -8 io->° io -12
•
3001
1018
30,GHz
i
300 GHz
EH F
1 cm
1 mm
Figure 2 Electromagnetic spectrum and frequencies, adapted from Gardiol [28].
2.1.1 Microwave systems and instrumentation
References [27] and [31] provide a review of currently available microwave systems
and their characteristic features. Some important features are discussed in this section.
Typical microwave systems have three major components: a microwave source or
generator, a waveguide and an applicator.
The most commonly used microwave
generators include magnetrons, klystrons, and traveling wave tubes. Magnetrons are the
most widely used sources, practically in all domestic microwave appliances because o f
their compactness and low cost. Resonant cavities serving as tuned circuits determine the
output frequency o f the system. Most magnetrons operate at 2.45 GHz and larger tubes
14
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
are employed for 915 MHz industrial furnaces.
Klystrons, like magnetrons, also
modulate the movement o f electrons using resonant cavities. Klystrons are typically
employed in UHF-TV, satellite communications, and industrial heating. Klystrons are
well suited for high power applications due to the high gain achievable with these
sources.
Traveling wave tubes (TWTs) are linear-beam tubes that employ electron
beams to amplify a microwave signal. The operation o f TWTs differs from magnetrons
and klystrons in that no resonant cavities are used. The microwave signal in a TWT is
attenuated within a periodic nonresonant structure such as a helix, causing the signal to
interact with the electron beam. A magnetic field along the axis o f the tube focuses the
electron beam. Electrons emitted from an electron gun (cathode) are accelerated toward
the collector surrounded by a traveling wave at microwave frequency with a strong field
component in the path o f the electron beam. The microwave signal velocity is modulated
as the electron energy is transferred to the microwave signal on the helix.
The
nonresonant structure o f the TWTs allows a wide range o f microwave signals to be
amplified.
TWTs can operate between 0.5 and 18 GHz and are used in variety of
applications including transmission and heating. Variable frequency microwave heating
systems employ TWTs as the source due to the wide range o f available frequencies.
The function o f the waveguide is to transmit the microwave radiation generated from
a source to the oven or the material to be heated. Hollow metal tubes due to their high
reflectivity are used for high frequency operation to reduce transmission losses.
Rectangular cross-sections are most commonly used in waveguides typically with
quarter-wave dimensions.
The cavity o f the oven or furnace where the material is
15
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
processed is considered the applicator as it transfers the electromagnetic energy from the
waveguide to the material to be processed. Nonuniformity and hot spots in single mode
applicators are often minimized by using mode stirrers and rotation devices. Single mode
applicators are best suited for small volumes due to localized field intensities in
applications such as glass fiber drawing and sintering. Multimode applicators are used
for larger and complex shaped materials. Multimode applicators powered with variable
frequency sources, such as the one used in this study (Chapter III) offer the advantage of
uniformity in electric field distribution throughout the cavity allowing processing of
advanced materials requiring greater uniformity [26].
A good understanding o f the microwave-materials interactions and the resulting
energy transfer at the molecular level is essential for successful implementation of
microwave processing. The fundamentals o f dielectric properties and the effects o f an
electromagnetic field are reviewed in the following sections.
2.1.2 Molecular mechanisms o f polarization
When an electromagnetic field interacts with a material four different types o f
polarization can occur in the material [32].
Under an applied electric field, the
displacement o f the negatively charged electron cloud relative to the positively charged
atomic nucleus results in induced dipole moments, which causes electronic polarization.
This type o f polarization occurs in all materials at very high frequencies corresponding to
the visible and the UV region o f the electromagnetic spectrum [33].
Similarly, the
displacement o f atomic nuclei or ions relative to other nuclei or ions results in atomic
16
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
polarization also referred to as ionic polarization. The movement o f heavy atoms or
nuclei is however sluggish as compared to the electrons and consequently the time
required for atomic polarization corresponds to the infrared region. The electronic and
atomic polarizations are often referred to as displacement deformations or distortion
polarizations.
These polarization mechanisms are together also termed ‘optical
polarizations’ [33]. They act so fast that the net polarization observed under an electric
field at microwave frequencies is in phase with the field, and as a result they do not
contribute to microwave absorption.
Asymmetric charge distribution between unlike atoms in a polar molecule gives rise
to permanent dipoles in materials. In the absence o f an electric field, these permanent
dipoles are randomly distributed with a zero net polarization. The introduction o f a static
electric field tends to align these permanent dipoles in the field direction resulting in a net
polarization called the orientation or dipole polarization. The extent to which this can be
accomplished depends on the relative mobility o f the dipole, the electric field strength
and the time for which the electric field is switched on relative to the relaxation time for
the dipole.
At microwave frequencies, the contribution due to induced or optical
polarization is negligible and orientation polarization by the electric field is the primary
mechanism for energy coupling at the molecular level for dielectric materials including
polymers.
When the applied field is removed, relaxation o f the polarization in the
dielectric material occurs which is referred to as dielectric relaxation.
Dielectric
relaxation in polymers is characterized by a distribution o f relaxation times and is
dependent on the structure and molecular arrangements in molecules.
17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
When an alternating electric field is applied, the dipoles tend to oscillate in phase
with the electric field thereby storing the energy o f the applied filed.
At higher
frequencies, the rotation o f the dipoles cannot follow the applied alternating field. The
polarization o f dipoles where translational motion is restricted or the polarization o f
molecules where rotational motion is restricted thus results in a lag between the applied
electric field and the polarization. This phase lag is often expressed as a phase difference
or loss angle 5. The dipole rotation thus increases the energy o f the molecule and is
accompanied by intermolecular friction (viscous transfer) resulting in dissipation and
heating.
In the case of composite materials, interfacial or space charge polarization is also an
important mechanism that contributes to heating. In electrically heterogeneous materials,
differences in the ease o f motion o f charge carriers between two phases exist. This
impedes the free motion o f such carriers at phase boundaries and results in a macroscopic
field distortion. The carriers build up at the interface causing polarization due to charge
separation under the action o f an applied field. This polarization phenomenon is also
referred to as Maxwell-Wagner polarization [32].
The Maxwell-Wagner effect is
dependent on the material properties, geometry, and frequency. The rate o f charging at
the interface between two phases under the applied field and hence the resulting
dissipation is proportional to the difference between the current densities in the two
phases.
The total polarizability ( a T) o f a material can thus be expressed as a sum o f the four
terms as shown in Equation 2.1:
18
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
= ae + at + ad + a s
aT
(2 . 1)
where a t is the electronic polarizability, a, is the ionic or atomic polarizability, a d is the
dipolar or orientational polarizability and ccs is the space charge or interfacial
polarizability.
2.1.3 Macroscopic effects o f electromagnetic field on materials
From a macroscopic perspective, the electrical behavior o f a material when subjected
to an electromagnetic field is characterized by the following parameters: the permittivity
(e) which describes the interaction with the electric field, the permeability (n) which
describes the interaction with the magnetic field and the conductivity ( ct) which is
characterizes the free-electron conductive properties. The constitutive equations relating
the field strength with the respective flux densities are as follows:
D
=
sE
(2.2)
B
=
HH
(2.3)
J
—
<
j E
(2.4)
19
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
where, D, B and J are the electric, magnetic and current flux densities respectively and E
and H are the electric and magnetic field strengths.
Under the influence o f an alternating electric field E, the current density J through a
dielectric is related to the complex dielectric constant (e*) of the material as:
(2.5)
J = i to e CoE
where i = (-1)I/2, co is the measurement frequency, and e<, is the permittivity o f free space
(8.85 x 10'14 F/cm). The complex dielectric constant can be separated into its real and
imaginary parts as
(2 .6)
e (co) = e (co)-i e (co)
where e is the relative permittivity and e' is the relative loss factor. The relative
permittivity and the loss factor are functions o f the measurement frequency, temperature
and structure o f the material. The relative permittivity is a measure o f the electrical
20
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
polarization or alignment o f the species o f a material under an electric field, and the loss
(e ) is a measure o f the energy required for molecular motion in the presence o f the
electric field. The relative loss factor (e’> is an effective loss factor accounting for the
individual loss mechanisms resulting from the four different polarization phenomena
discussed earlier. The delay between changes in field and changes in polarization is often
expressed as a phase difference or loss angle, S. The dissipation factor or loss tangent,
tan (5) is defined as the ratio o f the effective loss factor (e ) to the relative permittivity
(e). The relative loss factor can alternatively be represented as an equivalent or effective
conductivity representative o f all these mechanisms.
The variation o f polarizability and dielectric loss o f a polymer with frequency
referred to as dielectric response o f a polymer is shown in Figure 3.
21
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Frequency (Hz)
IO4
106
108
Radiowaves
IO6
IO4
IO10 1012 IO14 1016 1018 1020
| Micro-
IO2
1
1022
j Infrared |Y| Ultra- j X-rays | Gamma Rays
s
io-2 io-4 io-6
Wavelength (cm)
IO-8 IO-10 1 0 ‘2
Dipolar
Ionic
•8
N
•c
e3
O
a.
Electronic
10s
IO10 1012 10M
Frequency (Hz)
Figure 3 Dielectric response o f a polymer.
It can be seen from Figure 3 that the principal mechanism o f coupling microwave
radiation to polymer dielectrics is through dipole orientation by the electric field. The
efficiency o f coupling microwave energy into a material is dependent on a number o f
factors, including the dipole strength, the mobility o f the dipole and the mass o f the
dipole.
This suggests that small strong dipoles couple microwaves most efficiently.
Moreover, small molecules are more likely to couple energy into the system than
macromolecules as they obtain translational energy apart from rotational energy caused
22
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
by reorientation. Similarly, a dipole as part o f a side group will couple more strongly
than a dipole on the backbone which is less mobile.
Analogous to Equations 2.5 and 2.6, the losses resulting from the interaction o f a
material with a magnetic field can be expressed as a complex magnetic permeability
given by Equation 2.7.
(2.7)
where n B is the permeability o f free space (//„ = 47t X 10*7 H/m), /u is the relative
permeability and f i is the relative loss factor accounting for the relaxation and resonance
phenomenon under the effect of an alternating magnetic field. However, the effect of
magnetic loss factor to the total heating of materials in a microwave oven is often
neglected as most o f the materials used in microwave processing such as polymers,
ceramics, and many metals are typically magnetically transparent.
In the case o f conducting materials with free electrons, electronic conduction plays a
significant role in microwave heating. The resistive dissipation resulting from the motion
o f free electrons under an applied filed contributes to conduction losses, which can also
be expressed as an equivalent effective loss factor in Equation 2.6. These can be related
by Equation 2.8.
23
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
When the bulk conductivity o f the material is very large, the electric fields attenuate
rapidly due to skin effect. The skin depth (8,), defined as the distance into the sample at
which the electric field strength reaches e*1 o f its value, is dependent on the
electromagnetic properties o f the material and is given by Equation 2.9.
8 .
(2.9)
=
yjT tfOV
where / i s the applied frequency. When the skin depth o f a material is larger than the
dimension o f the material, this effect can be neglected.
2.1.4 Energy conversion in a microwave field
The dielectric properties o f materials coupled with the electromagnetic field intensity
and distribution determine the conversion o f electromagnetic energy into heat
The
electromagnetic wave propagation in a medium is governed by Maxwell's equations and
24
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the power transmitted to the medium can be derived [27, 34] from these equations by the
use o f the Poynting Vector. The Poynting vector S is given by the cross product E X H
and has the units o f surface power density (W/m2). The total power transmitted to an
arbitrary volume V bounded by the surface S is given by Equation 2.10, which represents
the real portion o f the Poynting power theorem:
P
=
Jff|£|V + ^wje'lEfdV + j( o $ M'\H\2dV
V
V
(2.i0)
V
The first integrand on the right hand side o f Equation 2.10 represents the power
dissipated in the medium (as heat) through electric conduction, while the second and third
term represent the power dissipation associated with electric and magnetic loss
respectively. In dielectric materials, the loss associated with magnetic permeability is
small and hence the third term o f Equation 2.10 can be neglected. The first two terms o f
Equation 2.10 can be combined by defining an equivalent dielectric loss for the electric
conduction loss (as shown in Equation 2.8) or by defining an equivalent conductivity for
the dielectric loss. Assuming uniform field distribution throughout the volume V, a
simplified expression for the power ( P ) absorbed per unit volume in a microwave field
can be obtained from Equation 2.10 as follows.
25
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
p
— 2n
2 s s'
** f
J E nms
where P is the power absorbed per unit volume,
(2.11)
is the root mean square electric
field strength, Gois the permittivity o f free space, e is the relative loss factor and / is the
frequency.
Equations 2.10 and 2.11 govern the microwave absorption characteristics o f
materials. From Equation 2.11, it can be seen that for a given frequency and electric
field, the amount o f power absorbed by a material per unit volume is a function o f its
dielectric properties, specifically the dielectric loss and how it changes with temperature.
Materials with high conductivity such as bulk metals limit the penetration depth o f
microwaves and serve as reflectors leading to negligible microwave absorption. On the
other hand, materials with a low dielectric loss such as alumina, quartz and
polytetrafluoroethylene have a large penetration depth and very little absorption occurs
within the material, as these materials are transparent to microwave energy. The energy
coupling by microwaves is most efficient in materials that have high dielectric loss and
moderate conductivity such as alcohols and liquid resins [34]. A good example o f a
microwave receptive material is water, which has a high dielectric loss at microwave
frequencies.
It is important to note that while Equation 2.11 is useful is understanding the
importance o f dielectric properties on microwave absorption, material processing in
26
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
practice is much more complex as several factors affect the actual heating process. These
factors are both electromagnetic field as well as material property related. The geometry,
dimensions, placement, homogeneity o f the sample, and the mode o f applied microwaves
are some factors that affect the energy coupling to the material. Further, the dielectric
properties o f the material are typically functions o f temperature and frequency.
Moreover, for reactive systems, the extent o f reaction and the ensuing changes in material
property (based on structural and phase changes) also affect the dynamics o f absorption.
For example, in many polymers and ceramics, microwave absorption dramatically
increases with temperature. Absorption characteristics can also vary significantly with
slight changes in the purity or the presence o f defects. The complexity of the materialmicrowave interactions emphasizes the need for extensive experimentation and
characterization o f material properties under a variety o f processing conditions and a
thorough understanding o f the theoretical background and the practical implications.
27
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.2 Literature Review
There has been considerable interest over the past few years in the application o f
microwave radiation as an alternative to thermal heating for material processing in
industrial applications owing to the potential advantages offered by microwave
processing which include reduced cost and increased throughput. Research and advances
in the field o f microwave processing in recent years is reported in a series o f symposia on
microwave processing o f materials by the Materials Research Society (MRS) [35, 36, 37,
38, 39] and the American Ceramic Society (ACS) [40] and reference [25]. Most of this
research has been carried out with fixed frequency systems o f 0.915 GHz or 2.45 GHz.
References [22,25,41,42, and 43] review some o f the recent literature on the microwave
processing of polymers and polymer composites.
Traditional areas o f microwave
research have focused on rapid processing capability as applied to fibers and fiberreinforced composites.
The most significant commercial application o f microwave
technology for polymer processing has been in the vulcanization o f extruded rubber for
the automotive and construction industry [25]. There is significant interest in applying
this technology to the processing o f high performance materials such as high-temperature
polymers and polymer-composites for semiconductor and electronic packaging
applications [44, 45].
Some o f the essential requirements for any new polymer
processing technology to be applicable in the microelectronic industry include: 1) a high
degree o f uniformity over large area (across a 300 mm diameter for wafer level
applications), 2) a high degree o f control and repeatability over several cycles, 3)
28
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
compatibility with metals and conducting materials as integrated circuits and electronic
packages are composed o f metal interconnections and other circuitry, and preferably
selective processing capability. Traditional single frequency microwave ovens are not
suitable for these applications due to issues such as non-uniformity (hot spots),
insufficient control over process variables (thermal runaway) and incompatibility o f
metals due to charging and arcing concerns. VFM technology addresses these concerns
and has been shown to be compatible with electronic device functionality [44]. As
discussed earlier, broadband VFM technology is in its infancy and there have been very
few studies [39, 46, 47] using this technology. There has been very little work done on
evaluating this technology, especially in the processing o f high-performance polymers
such as polyimides [46, 48, 49]. While VFM technology differs significantly in terms o f
potential advantages over traditional single frequency systems, many o f the underlying
process-effect dependencies in single frequency systems are similar and may be
applicable to VFM processing. A brief overview o f some o f the key observations from
previous studies on microwave curing on polymers will be presented in the following
sections.
To date, epoxies have been the most widely studied polymers using microwave
curing [22, 25,42, 43, 50]. A few studies on microwave curing o f polyimides have been
reported [45,46,48,49, 51, 52,64]. A common feature reported by most o f these studies
is a reduced processing time by microwave processing. Also, improved mechanical [53]
and structural properties [54] were reported for microwave-processed materials, which
could be a result o f better temperature distribution than conventional thermal processing.
29
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Improved properties in microwave-processed samples, such as higher tensile strength
[55], reduced internal and residual stresses [56], and higher glass transition temperature
[57], have also been reported. Similarly, shorter cure times and improved properties
upon VFM processing o f polyimide-based materials are also reported [48].
As mentioned earlier, there are a number o f reports in the literature that claim
acceleration in reaction kinetics upon microwave processing [58, 59,60]. A 10-20 fold
reduction in the time required to achieve full cine in thermosetting systems such as
bimaleimides and epoxy resin networks is reported [61].
Likewise, a 10-35 fold
acceleration in reaction rates in solution imidization studies was reported by Lewis et al.
[59]. This acceleration in kinetics, called the microwave effect, has been reported both in
small molecular reactions [62] and polymeric systems [63].
However, this issue is
controversial as there have also been some reports that do not show enhanced reaction
rates [64, 65, 66]. A retardation o f reaction kinetics has also been reported in some
instances [67]. A rational comparison o f these effects is not possible since each o f these
reports pertains to distinct materials under different experimental conditions and varied
methods o f temperature measurement.
For instance, work has been done on single
frequency microwave processing o f epoxy resins [68, 69, 70]. The reactivity o f epoxy
systems can vary more than an order o f magnitude, depending on the specific reactants
studied. Similar reports are also found in the literature on microwave processing o f
polyimides [64, 71].
This uncertainty is, in part, due to the different methods o f
temperature measurement employed in each o f these studies. Any report on kinetics o f
30
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
cure should provide a means o f accurately determining the temperature o f the sample
during microwave processing [discussed in section 3.1.1 o f Chapter III].
The reduction in processing times, acceleration in reaction kinetics and/or
improvement in properties o f microwave processed materials, often referred to as the
“microwave effects’, have been categorized under a) thermal and 2) non-thermal affects
[72]. The thermal effects are ar.ibuted to a more efficient mode o f energy transfer in
microwave-processed materials [73]. Energy transfer in conventional thermal ovens or
hotplates occurs through conduction, convection, and radiation modes o f heat transport
and is limited by the thermal conductivity o f the material. However, in the case of
microwave processing, heating occurs by dielectric dissipation resulting in uniform and
rapid global heating throughout the sample. This allows the material to heat faster as the
dissipation is volumetric in nature. Non-thermal microwave material interactions are
often explained as occurring due to localized heating effects at the molecular level.
Lewis el al. [71] proposed the existence o f a nonequilibrium nonuniform energy
distribution at the molecular level, which results in certain functional groups (such as the
dipoles) having a greater energy than the average energy o f the adjacent groups.
Temperature o f a medium is determined by the statistical average kinetic energy o f the
species within the medium.
This effect o f localized dissipation was estimated to
correspond to an increase in an effective temperature o f the reacting groups to be about
SO°C above the bulk temperature in solution imidization studies [71]. The microwave
energy is believed to directly couple to the reactive polar groups, which dissipate the
energy through the adjacent groups by the usual mechanisms. However, when the rate o f
31
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
energy absorption by the functional groups is greater than the rate o f energy transfer,
there exists a nonuniformity at the molecular level which drives reactions faster.
Furthermore, the existence o f this high local temperature is also likely to enhance
molecular agitation and improve the transport properties o f reactive species particularly
in reaction schemes where diffusion limitations are important.
Another implication o f the existence of a high local temperature is the possibility of
achieving reactions at a lower bulk temperature.
Fanslow et al. [74] reported the
enhancement o f chemical and physical reactions by microwave processing.
It is
proposed that the combined thermal and electrical excitation produced by the microwaves
have a synergistic effect in producing chemical reactions at a lower temperature than
would be required for a similar thermal treatment.
As discussed earlier, considering the potential cost, processing, and performance
advantages offered by microwave processing, there is interest in applying this technology
for microelectronic applications, especially with the advent o f VFM technology which
obviates most o f the problems encountered in traditional single frequency ovens. As
mentioned earlier, the actual chemistry and reactivity o f species, the microwave
applicator used and the processing conditions significantly affect the effectiveness of
microwave processing. However, there have been very few studies on VFM curing o f
currently used (or potential) electronic grade epoxy materials [44,46] and fewer on high
performance dielectrics such as polyimides. In this study, the feasibility o f rapid curing
o f high performance dielectric polymers o f interest such as benzocyclobutenes (BCB)
(Chapter IV) polyimides (Chapter V) and polybenzoxazoles (PBO) (Chapter VI) by VFM
32
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
processing was investigated. Cyclotene™ BCB from Dow Chemical Company is an
example o f a thermoset resin where polymerization is essentially a temperature driven
crosslinking reaction. The thermal cure kinetics o f BCB are well studied [75]. The
reaction kinetics o f VFM curing of BCB were studied (Chapter IV) to determine if
microwave processing enhanced the reaction rates o f BCB similar to other thermoset
resins. The effect o f VFM processing on the chemical, electrical and physical properties
was also investigated.
As discussed earlier, there is considerable interest in reducing the thermal budget for
processing thin film polymer dielectrics applied as passivation and redistribution layers
on silicon. Polyimides require high temperature processing for long cure times. While
the imidization reaction by itself occurs at lower temperatures, high temperature
treatments are essential to completely remove the casting solvent. It has been reported
that in many polymer/photo-polymer systems, complete evolution o f high molecular
weight alcohol and photoreaction byproducts requires an annealing step with exposure to
high temperatures for a long period of time [14]. Along with the photo-crosslinkable
polymer backbone, several additives are added to the polymer to achieve photodefinition.
These include the photosensitive monomer, sensitizers (which absorb the incident UV
energy), and initiators (which are activated by the excited sensitizers and initiate the
photoreaction in the polymer) [14].
The initiator-induced photoreaction or thermal
polymerization o f monomer typically introduces crosslinks in the polymer matrix. There
have been several reports o f microwave-induced acceleration in crosslinking reactions
[70]. However, there have been very few studies on VFM curing o f photosensitive
33
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
polyimides [49]. Moreover, the effect o f VFM processing on crosslinking reactions in
photosensitive polymers has not been well studied. Further, the influence o f polymer
backbone rigidity and photochemistry o f photosensitive polyimides on the effectiveness
o f microwave curing is not well understood. In this study, some o f these issues and their
dependence on processing conditions were investigated (Chapter V). It is anticipated that
the high local temperatures o f functional groups in microwave-processed films will assist
in preferential diffusion and evaporation o f solvent and photoproducts from the polymer
films at a lower bulk temperature as compared to thermal curing. The feasibility o f rapid
low-temperature curing o f different commercially available polymer dielectrics on silicon
to obtain properties comparable to traditional high temperature thermal cures was
investigated (Chapter V). The structure-property relationships and their dependence on
the processing conditions were also studied.
Although there is widespread acceptance o f the selective heating capability o f
microwaves, very few studies have demonstrated the feasibility o f achieving these effects
[44,46]. As discussed earlier, the use o f high performance dielectrics such as polyimides
on organic substrates is not feasible by conventional thermal curing due to the high cure
temperature o f the dielectric and the low thermal stability o f the organic substrate. In this
study, the feasibility o f low temperature selective heating o f high performance
polyimides on organic substrates by VFM processing was investigated (Chapter VII).
Factors affecting low temperature polyimide curing and the effect o f solvent on the
properties and thermal stability o f cured films were also studied.
34
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
VFM technology helps overcome the problem o f charging and arcing when
conductive materials are placed in the processing cavity. The microwave conductivity o f
bulk metals is much too high to permit their heating by microwaves; they simply reflect
the energy back into the cavity. On the other hand, finely divided metals can be heated
[76]. The increased surface area accentuates the role o f skin depth o f the metal. The skin
depth is the near-surface region (~1 to 100 pm), which absorbs a small amount o f
microwave energy owing to its finite conductivity.
In the case o f metal powders, a
significant fraction o f the material is in the near surface region and hence can play a
significant role in microwave absorption.
On a similar note, filled polymers show different heating characteristics than the neat
resin. Fillers such as fibers, carbon black, and conductive additives such as metal flakes,
spheres or needles with sizes ranging from 0.1 to 100 pm are often added to polymer
systems to improve the thermal, electrical, and mechanical properties. The presence o f
these fillers affects the electric field pattern in and around the composite material,
resulting in potentially different curing profiles with losses being substantially higher
than for the pure matrix resin. The resultant heating is dependent upon the electrical
resistivity o f the additives, and the temperature distribution is strongly dependent on the
distribution o f the fillers in the matrix. For instance, it is reported [77] that microwave
processing carbon fiber filled polymer showed a temperature profile that exhibits a
maximum at the surface, which results in the enhancement o f interfacial adhesion and
subsequently improves the fracture properties o f the composite material.
An
understanding o f these mechanisms would help in designing materials with the desired
35
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
properties for several applications. Applications in the microelectronics industry include,
solder reflow for flip-chip applications, ceramic or metal filled polymers for thermally
conductive polymer applications, silver filled epoxies as electrically conductive adhesives
(ECA) for direct chip attach applications. Very few studies on microwave processing o f
electrically conductive adhesives are reported [78]. In this study, rapid and selective
curing o f silver flake filled epoxy adhesives and solder reflow by VFM processing was
investigated (Chapter VIII).
Although there has been a lot o f work on single frequency microwave processing o f
materials [25], the exact mechanism o f energy coupling is still not clear.
VFM
technology is relatively new and there have been very few studies so far, especially on
high performance dielectric polymers. Hence, there is a need to study the effect o f VFM
processing o f advanced materials to understand the underlying principles and realize the
potential benefits o f this emerging technology. A thorough and complete investigation o f
chemical, mechanical and electrical properties o f VFM cured polymer films and an
understanding o f the structure-property relationships in a wide range o f VFM cured
materials is essential.
36
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER III
EXPERIMENTAL METHODS AND PROCEDURES
In this chapter, a detailed review o f the VFM experimental setup and the different
analytical techniques used to characterize VFM and thermal cured polymer films is
presented. A brief overview o f the temperature sensors used in the VFM furnace is also
provided. The chemical structure and changes occurring during the cure reaction were
monitored using Fourier Transform Infrared Spectroscopy (FTIR). The properties o f
interest include electrical and optical properties (i.e. permittivity, loss and refractive
indices and birefringence), mechanical properties (i.e. residual stress and Young’s
modulus), and thermal properties such as coefficient o f thermal expansion (CTE), glass
transition temperature, and thermal stability. A detailed description o f these techniques is
also presented in this chapter.
3.1 Variable Frequency Microwave Furnace: Experimental Setup
The features o f the VFM furnace (VFMF) used in this study are discussed in this
section. All the microwave cure studies were performed using Microcure 2100™ VFMF
from Lambda™ Technologies Inc. This system is similar to the one described elsewhere
37
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[79]. Some o f the important features o f the VFMF are summarized here. Figure 4 shows
a picture o f the microwave cavity (Dimensions: 14" H x 15" L x 19" D), the different
temperature sensors and the experimental set up.
Microcure 2100 is a multimode applicator powered by a traveling wave tube variable
frequency microwave source [26].
The VFMF allows a number o f controllable
parameters such as central frequency, bandwidth, sweep rate, power level and ramp rate.
Samples may be processed either at a fixed frequency ranging from 5.85 to 7.0 GHz or a
variable frequency sweep about the same range o f central frequency with band width
varying from 0-10% and variable sweep rates. The forward or applied input power to the
applicator can be varied to a maximum o f 500 W in 20W increments.
Certain
modifications were incorporated for better temperature control and operation.
One
important modification was the incorporation of a feed back control system to control the
temperature o f the sample to be processed. The control system adjusts the power levels
automatically to maintain the sample at the desired temperature. This allows very good
control o f ramp rates and final hold temperature o f the samples to be processed.
The VFMF also has the provision to maintain an inert atmosphere while processing.
The processing chamber can be pumped down using a mechanical pump and back-filled
with an inert gas such as nitrogen for processing materials in an oxygen-free atmosphere.
As shown in Figure 4, samples are placed on a quartz substrate supported by quartz
mounts. Quartz was used as the substrate (holder) due to its negligible microwave
absorption [76]. The contribution o f the substrate (holder) to the heating characteristics
o f the material to be processed can be therefore minimized. The VFMF provides three
38
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
different temperature control devices; 1) an infrared pyrometer, 2) a fiber optic probe and
3) a thermocouple. The fiber optic probe and thermocouple are held in contact with the
substrate using a high temperature Kapton™ tape and a metal clip insulated with the tape.
F ib e r opt
p ro b e
Figure 4 Variable Frequency Microwave Fumace-Microcure™ 2100.
39
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.1.1 Temperature measurement and control in the VFM
Temperature measurement and control in a microwave furnace is one o f the most
difficult yet important parameters in microwave processing. Maintaining good thermal
contact with the sample is critical when using contact temperature probes. Further, the
probes should produce a minimal pertubation o f the existing fields in the furnace. The
probes by themselves should not be significantly affected by the electromagnetic fileld.
Initial effort in this research was therefore devoted towards determining the appropriate
temperature measurement technique. The most commonly used devices for temperature
measurement in microwave processing are thermocouples [63,76], fiber optic probes [57,
80] and infrared pyrometers [45].
Each method has its own advantages and
disadvantages as discussed below.
The three temperature measurement devices investigated were 1) a fiber optic probe,
2) a thermocouple, and 3) an infrared pyrometer.
Fiber optic probe:
Fiber optic temperature sensors offer impressive versatility over a temperature
range -200 to 300°C with an accuracy o f ± 0.1 °C and a response time o f ~ 0.25 to 5
sec. They are widely used for temperature measurement and monitoring in a number of
applications that include high voltage applications, r f and microwave applications.
The most commonly used fiber optic temperature probes consist o f a teflon-coated
optical fiber with a semiconductor sensor and a high temperature dielectric at the tip.
These probes cannot be used beyond a temperature o f 300°C and good contact with the
sample is essential for accurate measurements. For the VFMF experiments, a Nortech™
40
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
NoEMI-TS® family fiber optic probe from Nortech Fibronic Inc. with a teflon cladding
and high temperature epoxy tip was used.
The fiber optic probe was used as the
temperature control device for the feedback control for some experiments and also as an
independent auxiliary reference device for the calibration o f the thermocouple and the
infrared pyrometer. It was found that the fiber optic probe was not affected by the
microwave field and could be used for temperatures less than 250°C when proper contact
with the sample was ensured.
Thermocouple:
Thermocouples are most often used for direct measurement o f sample temperature
and as temperature sensing devices for feedback control o f ramp rates and hold
temperature. As mentioned earlier, arcing might be an issue with very fine thermocouple
wires during microwave processing. Although arcing can be avoided in the VFMF, it is
desirable to have a sheath protecting the thermocouple wires. It has been reported [76]
that microwave field has no effect on the temperature measurements as made by a
thermocouple and that there are no discontinuities in temperature measurements when the
microwave power is turned on and off.
However, experiments in the VFMF with a K-type grounded thermocouple showed
that the thermocouple was reading temperatures higher than the actual temperature (as
shown by a reference fiber optic temperature probe) with microwaves as the source of
heat No such effect was observed when a hot plate was used as a heating source instead
o f the microwaves. This suggests that higher temperatures shown by the thermocouple
with microwaves as the source o f heat could have been due to the absorption o f
41
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
microwave energy by the metal sheath o f the thermocouple. Further, several experiments
showed that there could be concentration o f microwave field (energy) at the tip o f the
thermocouple. Such phenomenon is reported in literature [76]. Moreover, as with any
contact technique, good contact and repeatability are always issues.
It has been
suggested that coating the tip o f the thermocouple (sheath) with a dielectric barrier such
as calcium aluminum cement (Alundum), alumina or boron nitride helps in mitigating the
electric field concentration around the thermocouple tip [76].
Due to the aforesaid reasons, the thermocouple was not used as a temperature control
device.
Thermocouples may be used only with a suitable coating o f non-absorbing
material at the tip. Another approach suggests the incorporation o f a circular coil at the
thermocouple tip that prevents the flow o f current in the thermocouple in the presence of
the microwave field [31].
Infrared pyrometer:
To avoid the contact issues of thermocouples and fiber optic probes, non-contact
temperature measurement devices were investigated.
In this study, a Raytek™
Thermalert® T30 series infrared pyrometer was used. It consists o f an infrared detector
equipped with a laser sighting sensing head. In the current set-up used, the target spotdiameter was about 1" in diameter and the sample size used was always greater than the
spot diameter. The IR pyrometer provides temperature measurements with an accuracy
o f ± 2.5°C in the temperature range used.
The infrared pyrometer requires the emissivity o f the material whose temperature is
to be measured. The emissivity o f a given material is a function o f several factors most
42
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
important o f which are temperature, thickness, and surface roughness.
Hence the
calibration o f the pyrometer for emissivity at different temperatures for specific substratefilm combinations and thicknesses was needed. This was done using a fiber optic probe
and thermocouple as reference temperature devices and a hot plate as the source o f heat
The emissivity determined by this technique matched that determined using microwaves
as the heating source for the low temperature measurements. The emissivity o f the
different samples was determined and was input to the infrared pyrometer whenever it
was used as the temperature control device for automatic ramp rate control.
It was
observed that the infrared pyrometer allowed better control o f ramp rate and final hold
temperature than a fiber optic probe.
For most high temperature experiments, the
calibrated infrared pyrometer was used (preferred) as the primary temperature control
device while the calibrated fiber optic probe was used for temperatures o f up to 250°C.
43
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.2 Chemical structure and properties
3.2.1 Fourier Transform Infrared (FTIR) Analysis
Fourier transform infrared (FTIR) analysis can be used as an analytical tool for
monitoring the physical and chemical changes occurring in thin polymer films during the
cure reaction. The vibrational and rotational energy transitions o f molecules occur in the
frequency range corresponding to the mid-infrared region (4000 cm' 1 to 400 cm' 1 or 2.5
pm to 25 pm). This principle is used to obtain infrared absorption bands/spectra in an
FTIR spectrometer. The infrared absorbance can be used to detect the presence (or
absence) o f certain functional groups. Further, the changes in the infrared spectrum can
be correlated to the changes in the chemical structure o f a sample.
The infrared
absorbance (peak height) depends on the functionality, its concentration and the path
length o f the infrared beam. The intensity o f the peaks in the resulting spectrum can be
used as a semi-quantitative measure o f the concentration o f the species in the sample.
Typically, an internal reference or standard is used to account for any thickness
dependence of the absorption.
For transmission mode FTIR, the spectra o f all the samples were collected using a
Nicolet Magna IR 560 spectrometer.
Typically, the spectra were collected with a
resolution o f 4 cm'1 and averaged over 512 scans for each sample. Double polished
silicon wafers were used in some cases to improve the signal to noise ratio o f the
collected spectra. Single side polished silicon wafers with a suitable dopant level (to give
44
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
sufficient signal and reliable estimates o f peak height) were used in some cases. The
polymer cured film thickness was chosen such that a reliable estimate o f the peak heights
was obtained for a semi-quantitative estimate o f the extent o f reaction. For transmission
mode FTIR analysis, this thickness varied from ~ 4 to 6 pm for BCB (Chapter IV) and ~
2-3 pm for the polyimides (Chapters V and VI). The procedure followed for collecting
the FTIR spectra is presented below [81].
1. The FTIR spectrometer is purged with dry nitrogen for 10 to 15 minutes to eliminate
any absorbance from atmospheric contaminants such as moisture and carbon dioxide.
2. A double side polished (or a single-side polished) silicon wafer is pre-scanned, to
collect the background in transmission mode. .
3. The polymer solution is spin coated to obtain the desired cured film thickness.
4. The wafer is processed (exposed/cured) and the FTIR scans o f the sample (after softbake or after exposure or after cure) are collected using the previously collected scan
as the background with the same resolution and number o f scans as the background.
In some cases, attenuated total reflectance infrared spectroscopy (ATR-IR) was used
for the analysis o f the chemical structure o f cured films (Chapter V to VII). The spectra
were taken at an angle o f 45° in the infrared spectrometer fitted with a thunderdome ATR
attachment. The germanium crystal in the ATR unit allows collection o f spectra in the
wavenumber range 800 to 4000 cm*1. All the spectra were collected with a resolution o f
45
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4 cm'1 and averaged over 128 scans. The experimental procedure followed for collecting
the ATR-IR spectra is given below:
1. The ATR thunderdome setup is attached to the spectrometer
2. The spectrometer is purged with dry nitrogen for 10-15 min.
3. A background spectrum is collected before every sample. Here, the background
spectrum is taken in nitrogen ambient (without any wafer).
4. The sample (wafer/substrate with polymer film) is placed face down on top o f the
germanium crystal and the pressure head is tightened in order to obtain intimate
contact between the film and the crystal.
5. The spectrum of the sample is collected for the same number o f scans as the
background and the background spectrum is subtracted from the sample spectrum.
3.3 Optical properties
3.3.1
In-plane and through-plane refractive indices and birefringence
The in-plane and through-plane refractive indices o f the films were determined using
a Metricon thin-film prism coupler (Model 2100). The Metricon prism coupler contains a
prism o f a known refractive index (np). The film coated on a reflecting substrate is held
in close contact against the prism as shown in Figure 5 using a high-pressure plunger.
The prism coupler uses a Helium-Neon laser at 632.8 nm as the light source, and the
prism-film set-up is rotated on a circular table in order to achieve variable angles o f
46
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
incidence for the laser beam. At certain angles o f incidence, the energy o f the beam is
coupled into the film through tunneling o f the photons from the prism and into the film
establishing propagation modes. A photo detector records the intensity o f the reflected
beam as a function o f the incident angle. An example o f this detector intensity is shown
in Figure 6 below. As seen in Figure 6, the reflected intensity shows a series o f maxima
and minima (the minima correspond to the angles where energy o f the beam is coupled
into the film). The indices o f refraction and the thickness o f the film can be determined
once the angles at which the intensity minima (modes) occur are established [82]. A
minimum o f two modes is necessary in order to obtain values for both the index and
thickness independently.
In-plane and through plane refractive index values can be
obtained by changing the polarization state o f the incident laser beam.
The index
measure in the transverse electric (TE) mode is reported as the in-plane index o f
refraction (n ") and the index o f refraction measured in the transverse magnetic mode
(TM) is reported as the through-plane index o f refraction (n^-). The birefringence (An) is
determined from the difference between the in plane and the through plane index o f
refraction (An = n " - &l).
47
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
rPrism tip
Photo detector
Film index, n
Substrate B
n2 =index o f air; n ^ in d e x o f prism; n =index o f film
0 =film angle o f incidence; a =prism angle o f incidence
Figure 5 Schematic diagram o f the Metricon prism coupler.
p
u
C
0>
e
In c id e n c e a n g le
Figure 6 Detector intensity as a function o f angle o f incidence.
48
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.4 Electrical Properties
3.4.1 Relative permittivity (Sr) & dielectric loss
The relative permittivity (dielectric constant) and dielectric loss measurements were
performed using parallel plate capacitor structures (metal-insulator-metal) fabricated with
the polymer film as the dielectric between the two parallel plates (ASTM D 150-95 [83]).
The experimental procedure for the fabrication o f the parallel plate capacitor structures is
as follows.
1. Silicon wafers are oxidized in a tube furnace (~ 2500 A).
2. Approximately 2500 A o f Ti/Au/Ti or Al are deposited in a DC sputterer.
3. The polymer is spin coated and cured in a conventional thermal oven or in the VFM.
4. The top metal layer o f Al or Ti/Au /Ti ~ 2500 A is deposited.
5. The top metal layer is patterned with the capacitor mask (using a photolithography
step).
6. The metal Al or Ti/Au/Ti in the exposed areas is etched away using an aluminum
etchant (PAN etch) or a gold etchant (100 g o f KI and 65 g o f I2 in 1000 ml o f water).
Note: In some cases a shadow mask was used to sputter metal onto the polymer films for
fabrication o f the top electrode.
The diameter o f the circular apertures in the mask
determined the area o f the capacitor.
All the dielectric measurements were made using an HP 4236 LCR meter @ 10 kHz
on a Karl Suss Probe station. Probes were connected to the top and bottom electrodes,
sandwiching the dielectric to measure the capacitance and conductance o f the sample.
49
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The dielectric behavior o f a thin film polymer dielectric in a parallel plate capacitor
structure in the frequency range between 1(T* Hz to 100 MHz is typically modeled using
an equivalent circuit shown in Figure 7.
Figure 7 Equivalent circuit model for measuring the dielectric properties in thin films.
From the measured value o f the capacitance, area o f the top metal pattern, and the
thickness o f the film the relative permittivity may be calculated using Equation 3.1. The
dielectric loss in the sample can be estimated using the conductance o f the sample from
Equation 3.2.
Fringing effects are neglected in accordance with the assumptions of
reference [83] for estimation o f the dielectric constant.
C = £ °£ r A
(3.1)
SO
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
G = (o C tan 5
(3.2)
In the above equations 80 is the permittivity o f vacuum (8.85 xlO' 12 F/m), 6r is the
relative permittivity o f the medium, A is the area o f the electrode (measured using a
micrometer scale), t is the thickness o f the film (measured using the Metricon prism
coupler or the Alpha-step profilometer), oa=27if, where f is the measurement frequency, C
is the capacitance, tan 5 is the loss tangent and G is the conductance. All measurements
were typically performed at 50 % relative humidity and room temperature.
3.5 Mechanical Properties
3.5.1 Residual stress
The stress generated in a thin film polymer coated on a substrate as a result o f
solvent evaporation and curing is referred to as the residual stress. Residual stress results
from I) volume shrinkage on account o f solvent evaporation and evolution o f
condensation products during the cure process and 2) the CTE mismatch between the
polymer film and the silicon substrate. A majority o f the film shrinkage occurs in the
thickness direction as the film is constrained in the in-plane direction (by the substrate).
As a result, stress develops in the in-plane direction o f the film. Differences in the CTE
o f the polymer film and substrate cause different rates o f expansion and contraction
51
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
during temperature cycles (as in the case of cure reaction and cool down) further
increasing the stress in the film. High levels of residual stress can lead to delamination
and film cracking and are undesirable for the long-term reliability o f packages. When the
thickness o f the substrate (silicon: ~ 550 pm) is significantly greater than the thickness of
the polymer film (~ 5 to 10 pm), the residual stress in the film can be approximated to be
completely in the plane o f the film [84].
The stress resulting from the aforementioned processes induces warpage (change in
curvature) in the substrate. The change in the curvature o f the substrate can be correlated
to the residual stress built in the film. The residual stress in films cured on silicon due to
film shrinkage and CTE mismatch can thus be estimated by measuring the change in the
wafer curvature upon cure. The magnitude o f the residual stress in the film can be
estimated by measuring the radius of curvature o f the wafer before and after film
deposition and curing. The stress generated in a thin film is related to the change in the
radius o f curvature o f the wafer and the material properties o f the substrate by Equation
3.3.
The changes in wafer curvature were calculated by measuring the radius of
curvature o f the wafer using a He-Ne laser based Flexus stress analyzer (Model F2320).
Measurements were made on 3” and 4" <100> silicon wafers at room temperature. The
experimental procedure for estimation o f residual stress is described below.
1. A bare silicon wafer (3"or 4") is scanned in the Flexus to ensure that it is single mode
i.e., it has a single finite radius, and the initial radius o f curvature is measured.
2. The polymer film is spin coated on the scanned wafer and softbaked.
3. The radius o f curvature o f the softbaked film is measured.
52
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4. The polymer film is cured in a thermal oven or the VFM.
5. The wafer with the cured polymer film is scanned in the Flexus for its radius o f
curvature.
6. Film thickness is measured using a Metricon prism coupler or an Alpha-step
profilometer.
7. When the film thickness is much smaller than the thickness o f the substrate, the
residual stress in the polymer film can be estimated by using Equation 3.3.
In
Equation 3.3, £ / ( 1 - v) is the biaxial elastic modulus o f the substrate (1.805 x 1011
Pa for <100> silicon), h is the substrate thickness in meters, t is the film thickness
(meters), R is the differential radius o f curvature o f the substrate (meters)
(1//? = y /? 2 - V * . where R,=radius o f curvature o f bare substrate and R 2 is the
radius o f curvature o f the substrate with polymer film), and a is the film stress [85].
By convention, the tensile residual stresses in the polymer film were reported as
positive residual stress values in this study.
(33)
53
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.5.2 Young's modulus
The Young’s modulus o f cured polymer films was estimated from the stress strain
curves obtained from tensile testing o f thin, free-standing polymer films using Instron™
tensile tester (Model 5640). The Young's Modulus is calculated from the slope o f the
stress-strain curve in the initial low strain region. The elongation to break and tensile
strength are considered to be the maximum values the polymer film withstood prior to
failure. For films cured on silicon, the free films used for tensile testing were typically 5
to 10 pm thick. Rectangular strips o f the polyimide films (40-60 mm in length and 5-8
mm wide) were cut with the sample still on silicon substrate and lifted off using a "lift
off' technique, the details of which are described below.
1. The polymer film is spin coated directly on a bare silicon wafer.
2. The polymer film is cured on wafer (thermal or VFM).
3. About 5 to 10 mm wide strips are cut in polymer film while still on the silicon
substrate.
4. The film thickness is measured by an Alpha step profilometer.
5. The silicon wafer (with the cut polymer film strips) is immersed in hydrofluoric acid
(HF) for 2 to 3 minutes. HF etches the native oxide on the silicon wafer and thereby
lifts off the polymer strips.
6. The polymer strips are then dried in a vacuum oven for 12 to 18 hrs to remove
moisture absorbed during HF etching and subsequent rinsing in deionized water.
54
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7. The polymer strips are taped over each end with a masking tape to facilitate better
gripping o f the film by the Instron pneumatic grips, leaving ~ 40 to 60 mm between
the inside edges o f the two pieces o f tape (refer Figure 8). Note: Samples with
smaller dimensions (~ 25 mm X 5 mm) were used for tensile testing o f films cured on
organic substrates due to the difficulty in sample preparation.
8. The exact film dimensions (length and width) o f each test structure are measured.
9. The samples are then allowed to equilibrate to the humidity and temperature o f the
ambient before actual testing.
10. The samples are pulled at a constant strain rate, typically S mm/min until failure
occurred. A 100 N static load cell was used in the Instron tensile testing apparatus.
Four to five test strips were tested for each sample.
11. The Young’s modulus is calculated from the low strain region o f each stress / strain
curve. The elongation to break (ETB) and tensile strength are considered to be the
maximum values the test structure withstood prior to failure.
It was not possible to obtain reliable estimates o f ETB and tensile strength values as
defects o f nicks produced in the films while making the test structures resulted in fracture
modes other than brittle, tensile fracture thereby significantly lowering the measured
ETB.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Polymer Film
SO mm
Figure 8 Polymer thin-film test structure for tensile testing.
3.6 Thermal Properties
3.6.1 Thermal stability
Thermal stability o f cured polymer films and organic substrates was determined by
thermo-gravimetric analysis (TGA) using a Seiko TG/DTA 320. The weight loss o f
cured polymer samples was measured as a function o f temperature along with an inert
reference pan.
Samples o f cured polyimide films were lifted off silicon or organic
substrates by a lift-off technique described earlier. About S to 10 mg o f dried samples
were placed in TGA pans and ramped to the desired temperature. Samples were analyzed
in both an ‘isothermal’ mode and a ‘dynamic’ mode. The experimental procedure is
outlined below.
1. About S to 10 mg o f the sample is placed in an aluminum (or platinum) sample pan.
56
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2. The TGA chamber is purged with nitrogen at a flow rate o f ~ 100 sccm/min.
3. The TGA is ramped according to the desired temperature schedule.
For isothermal runs:
a. The sample is ramped to the desired final temperature at 80°C/min.
b. The sample temperature is held at the final temperature for approximately
60 minutes.
For dynamic runs:
a. The sample is ramped at a rate o f 10°C/min from room temperature to ~
550°C (~ 650°C in the case o f polyimide PI 2611).
4. The data is reported as percent mass remaining versus temperature.
The polymer samples show a certain amount o f weight loss at low temperatures.
This is attributed primarily to desorption o f moisture absorbed in the sample. The sample
mass is normalized accounting for this mass change. The degradation temperature in the
polymer is the temperature at which a significant weight loss o f the polymer is observed.
Thermal stability is reported in different ways: a) weight loss during the ramp to 500°C b)
temperature for 1 % weight loss (T/%) and c) temperature for S % weight loss (Ts%).
3.6.2 Pyrolytic mass spectrometry
Pyrolytic mass spectrometry o f cured films was performed using a VG Instruments
Model 70-SE spectrometer in order to identify the nature o f the species evolving from the
cured films as observed from the TGA studies. Polymer free films prepared by the lift­
off technique described earlier were placed in a capillary tube and subjected to probetemperature analysis in the mass spectrometer in the electron ionization (E.I) mode where
57
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the tube is heated at lO°C/min to 425°C. Only positive ions produced by ionization were
detected in this analysis. The ionizer potential was set to ~ 70 eV and the system pressure during the analysis is maintained below lO-6 torr. The evolved species with a
mass to charge ratio in the range 10 to 700 amu were detected by the mass spectrometer.
3.6.3 Differential scanning calorimetry (DSC)
DSC studies were performed on different polymers to study the curing behavior and
evolution o f solvent or photoproducts. In a differential scanning calorimeter, the heat
flux required to keep the sample pan and the reference pan at the same temperature can
be accurately measured.
Therefore, the different energy changes and transitions
occurring in the polymer during the temperature scan can be studied from the DSC heat
flow measurements.
The glass transition temperature (Tg) is the temperature above
which significant segmental mobility can be achieved in a polymer.
The Tg can be
observed in a DSC study by an enotiothermic shift in the base line curve indicating the
increased heat capacity for the polymer. As received samples o f polymer solutions (~10
to 20 mg) were placed in a DSC pan for curing studies. Free films were prepared as
described previously for other test methods by the lift-off technique to estimate the Tg of
the cured films. All the DSC data was collected using a Seiko DSC220. A description o f
the experimental procedure followed is given below.
1. About 10-20 mg o f polymer sample is placed in an aluminum sample pan.
2. The DSC chamber is purged with -100 sccm/min o f dry nitrogen.
3. The baseline is then allowed to stabilize and the sample is ramped in the DSC
according to the desired cure/temperature schedule.
58
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4. The different reactions occurring in the polymer are observed as exothermic or
endothermic peaks in the DSC. The glass transition temperature may be identified as
the temperature at which a shift (at the beginning o f the shift or the midpoint) occurs
in the baseline o f the DSC plot.
The glass transition temperature for most polyimides is often close to the
degradation temperature and therefore can be difficult to determine through calorimetry.
Although calorimetry data was collected for some samples no distinct transitions could be
inferred from the results.
3.6.4 Thermo mechanical analysis.
The CTE and Tg o f cured polymer films were determined using a thermo­
mechanical analyzer (TMA). The TMA measures the dimension change o f a thin film
sample as a function o f temperature under a controlled atmosphere.
The change in
dimension can be correlated to the CTE o f the polymer film and the temperature at which
an inflection in the expansion profile occurs may be considered as representative o f the
Tg o f the material. All the TMA studies were conducted using a TMA Model 2940 by
TA Instruments. Free films prepared by the lift-off technique were used for this purpose.
The typical sample size used was 25 mm L x 5 mm W. The polymer sample was
subjected to a static load o f ~ 0.05 N and the TMA furnace was ramped at a rate o f
5°C/min to a final temperature o f 400 to 450°C. The Tg and CTE o f the polymer film in
the range 50 to 200°C are determined from the TMA data.
59
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.7 Physical Properties
3.7.1 Moisture uptake
A Quartz Crystal Nanobalance (QCN) was used to measure the moisture uptake in
polymer films: The QCN is based on the principle that the characteristic frequency
changes in a quartz crystal resonator can be correlated to the mass changes that occur in
material directly attached to it. The frequency change is related to the mass change by
the following equation [86].
(4.5)
where Af is the change in the frequency for a change in mass Am, fo is the resonant
frequency o f the unloaded quartz crystal resonator, pQ is the density o f the quartz crystal
and jiQ is the elastic modulus o f the resonator. The quartz crystal can be calibrated
initially through controlled electrodeposition o f any metal (Cu for example) and relating
the mass change (estimated independently through a potentiostat) to the frequency
change.
hi order to use the QCN for estimating moisture absorption, a thin layer o f the
polymer is spun onto the quartz crystal (diameter o f the quartz crystal disc is 14 mm).
The mass o f the polymer deposited on the crystal can be estimated by measuring the
60
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
change in the resonance frequency of the crystal.
Further, the mass change in the
polymer when subjected to different humidity environments can be tracked to give an
estimate o f the amount of moisture absorbed by the sample.
61
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER IV
RAPID CURING OF BENZOCYCLOBUTENE BY VARIABLE FREQUENCY
MICROWAVE PROCESSING
In this chapter, results from VFM curing of Dow Chemical Cyclotene™ 3022
benzocyclobutene (BCB) are presented. The chemical changes occurring during the cure
reaction and the reaction kinetics were studied by Fourier transform infrared (FTIR)
spectroscopy.
The electrical, optical, mechanical and chemical properties o f VFM
processed films were characterized and compared with thermally processed films to
determine the effectiveness o f microwave processing.
4.1 Benzocyclobutene (BCB-Cvclotene 3022-63)
Benzocyclobutenes are a family o f thermoset resins with a low dielectric constant
(2.65) and good planarizability (> 90 %) and are used as inter-level dielectric materials in
a variety o f electronic packaging applications [87, 88]. The resin, as received, is partially
polymerized (B-staged) and dissolved in mesitylene (solvent).
The structure o f the
monomer unit is shown in Figure 9. The polymerization (cure) o f BCB proceeds through
a two step process: a thermally driven ring opening m echanism followed by a Diels Alder
reaction resulting in a crosslinked polymer matrix [89].
The two-step reaction is
62
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
schematically represented in Figure 10. The degree o f cure in the polymer is a function
o f the temperature and the time at that temperature. The final cure temperature used for
curing this polymer can be varied between 200°C and 350°C.
The standard
recommended thermal cure process is a 1 hr cure at 250°C [90]. The effect o f VFM
processing on the physical and chemical properties o f BCB was investigated under
different processing conditions and compared to the standard thermal cure.
63
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
a § U M j3 n
ch3
ch3
Figure 9 Chemical structure o f BCB monomer.
h3
ch3
ch3
ch3
[Q l
ch3
ch3
Polymer
Figure 10 Thermal cure reaction in BCB [89]: Ring opening followed by Diels -Alder
addition reaction.
64
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2 Results
The effect o f VFM processing on the physical and chemical properties o f BCB was
investigated under different processing conditions and compared to the standard thermal
cure. The cure temperature was varied between 175°C and 250°C, and the cure time was
varied from 5 to 45 minutes. The effect o f VFM parameters such as central frequency,
bandwidth, sweep rate and ramp rate was investigated.
To determine the influence o f central frequency on the cure characteristics, films
were cured on silicon substrate at different central frequencies between 5.8 GHz to 7.0
GHz with a narrow bandwidth at constant power. The heating characteristics o f BCB
films cured on silicon substrates at a constant power o f 100 W and central frequencies
6.0, 6.4 and 6.S GHz and a narrow bandwidth o f 0.1 GHz are shown in Figure 11. The
gray line shows the heating rate at a central frequency o f 6.425 GHz and full bandwidth
o f 1.15 GHz.
As expected, the rate o f heating increases with increasing central
frequency. Moreover, the ultimate temperature reached by the samples also increases
with increasing central frequency.
65
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
250
200
£
150
s
I lOO W Constant power H
0 1
0
1
1
1
1
100
200
300
400
1-----------
500
600
tim e, sec
■ 6.0+/-0.1 GHz
_
A 6.4 +/- 0.1 GHz
• 6.8+/-0.1 GHz
6.425 +/- 1.1 GHz
Figure 11 The effect o f central frequency on the heating characteristics o f BCB on silicon
at a constant power o f 100 W.
4.2.1 FTLR studies on VFM cured films
Fl'lR analysis has been shown [75] to be an effective analytical tool in following the
reaction progress o f BCB. Some o f the characteristic infrared absorbance peaks are given
in Table 1.
66
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table I Characteristic infrared absorbance peaks for BCB cure reaction and probable
peak assignments.
1500
Tetrahydronapthalene
1253
Rocking motion of CH
methyl groups
1194
BCB ring
1050
Si-O
985
Vinyl groups
1700-1800
C *0 Oxidation Peaks
The infrared spectra o f an uncured BCB film, a thermally cured film and a VFM
cured film are shown in Figure 12. As seen from this figure, there are distinct changes
that occur in the spectrum upon cure. These changes include: increase in intensity o f the
peak at 1500 cm*1, representative o f the tetrahydronapthalene group being formed during
cure [75], and decrease in intensity o f peak at 1475 cm*1, which corresponds to reacting
BCB group. The extent o f cure (x) may be estimated (Equation 4.1) from the ratio o f the
peak height at 1500 cm'1o f the sample spectrum to the same peak in a control sample.
67
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Extent o f cure
=
W
V
(„.!)
1500 7 ^ 1 2 5 3 /F a B C w c
2. 0 -
Absorbance
VFMcot
1. 5 -
0. 5 -
0.0
1600
1400
1200
1000
800
Wavenumbers (cm-1)
Figure 12 Comparison o f FTIR spectra o f soft-baked, VFM and thermally cured BCB
samples.
68
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
After a one-hour cure at 300°C in a conventional thermal furnace, the absorbance at
1475 cm '1 completely disappears. This sample is considered fully cured, and the peak
height at 1500 cm '1 o f this sample is taken as the control. Both the absorbencies are
normalized to the absorbance at 1253 cm'1, which corresponds to the rocking mode o f the
methyl groups attached to the silicon atoms and remains unaffected by the polymerization
and hence serves as an internal reference to account for thickness differences between
samples.
From Figure 12, it may be seen that the spectra o f VFM and thermally cured films,
are essentially identical with no significant differences indicating that the chemical
structure of VFM cured films is similar to conventional thermally cured films. This
analysis cannot account for any possible side reaction or oxidation during cure.
However, there were no absorption peaks in the LR spectra in the range 1700-1800 cm'1,
indicating the absence o f any oxidation during cure.
FT1R spectra were collected in both transmission and attenuated total reflection
(ATR) mode. Transmission FTIR is a bulk sampling technique while the ATR mode
samples only the top surface (< 1 pm) o f the cured films. Similar trends in extent o f cure
with cure conditions were observed by both these techniques indicating that the chemical
structure o f the cured films was identical at the surface and in the bulk o f the film. This
shows the uniformity o f VFM processing within the film thickness.
BCB films cured under identical conditions in the VFM furnace, a conventional
thermal furnace and on a hot plate were compared to determine the effectiveness o f VFM
processing. Figure 13 shows the percent conversion (as estimated by FTIR) o f BCB
69
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
films cured by these three methods at 175, 200 and 225°C for 30 minutes. All the
samples were ramped to the final cure temperature at a ramp rate o f 30°C per minute. For
each cure condition, comparable or higher conversion was achieved by VFM processing.
100
80
s
e
£
v
5
£
60
40
20
0
175°C 30 nun
200°C 30 min
225°C 30 min
□ Thermal Qven________ 9 Hot Plate________ B YEM___
Figure 13 Comparison o f percent conversion o f films cured in the thermal oven, hot plate
and VFM furnace under identical conditions.
A cure kinetics study was performed to determine if VFM processing enhanced the
reaction kinetics o f BCB. For this study, BCB films were cured for different times at
final cure temperatures ranging from 175°C to 250°C. Samples were processed at a
70
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
central frequency o f 6.425 GHz with full bandwidth o f 1.15 GHz and a sweep time o f 0.1
second. All samples were ramped at 30°C per minute. Figure 14 shows the infrared
spectra o f BCB films cured in the VFM furnace at 225°C for different hold times.
1400
1300
1000
800
‘Waramben (cm-1)
Figure 14 FTIR spectra o f BCB films cured in the VFM furnace for different times at a
temperature of225°C.
The progress o f the cure reaction can be studied by monitoring the absorbance at
1500 cm'1. It can be seen that the absorbance at 1500 cm '1 (indicative o f the formation o f
the tetrahydronaphthalene functional group that forms during polymerization), and hence
the extent o f cure increases with increasing cure time.
71
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The extent o f cure data (from FTIR analysis) o f the VFM cured films for different
processing conditions are summarized in Figure IS. From Figure IS, it is observed that
the extent o f cure increases with both temperature and cure time.
At any given
temperature, the rate o f reaction levels off which suggests vitrification o f the polymer
matrix. This phenomenon is commonly seen in thermally cured thermosetting polymer
systems [91].
From Figure IS, it is also observed that a 5-minute cure at 240°C and a 15-minute
cure at 22S°C give the same extent o f cure as the prescribed standard cure o f one hour at
250°C. (Note: A one-hour cure at 250°C gives a conversion o f about 0.97 relative to the
same control full cure sample). This shows the efficacy o f VFM processing in driving
reactions to completion. Moreover, comparable extent o f cure can be achieved at shorter
cure times at any given temperature and at lower temperatures for the same cure time.
Figure 16 shows plots o f -Ln (1-x) as a function o f cure time for different
temperatures. The cure times used for this analysis correspond to the time before the
onset o f vitrification (as may be inferred from Figure IS). The linearity o f these plots
indicates that the reaction follows first order kinetics. The slope o f each o f these plots
gives the kinetic rate constant, k at that temperature.
72
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1.0
Extent of cure (x)
0.9
0.8
0.7
0.0
0.5
0.4
5
0
10
20
15
25
30
35
40
T im e, m in
♦
175°C
□ 200°C
A 210°C
«
225°C
•
240°C
Figure 15 Progress o f VFM cure reaction o f BCB with time at different temperatures
(from FTIR data).
73
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.5
4.0
3.5
3.0
2.5
2.0
mi
i
1.5
-♦
1.0
0.5
0.0
0
5
10
20
15
25
30
35
40
T im e, m in
♦ 175°C
□ 200°C
A 210°C
■225°C
•240°C
Figure 16 Plots o f-L n (1-x) vs time at different cure temperatures to determine the rate
constants.
74
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
From the rate constants so derived, Ln(yfc) vs 1/T is plotted in Figure 17. This plot
shows an Arrhenius-type relationship between temperature and rate constants. The slope
o f this plot yields an apparent activation energy Ea, o f 25.7 +/- 4.4 kCal/mol which is
about 30% lower than the activation energy reported for thermally cured samples [89,
92].
0.0
-0.5
-
1.0
E. = 25.7 +/- 4.4 kCal/mol
-1.5
~
- 2.0
* -2.5
-3.0
-3.5
-4.0
-4.5
0.0019
0.0020
0.0020
0.0021
0.0021
0.0022
0.0022
1/T, per °K
Figure 17 Plot o f Ln(k) vs 1/T to determine the apparent activation energy E,, for the cure
reaction.
75
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2.2
Comparison o f properties o f thermal and VFM cured BCB films
The optical and electrical properties o f VFM were characterized and compared to
thermally cured films. Figure 18 compares the index o f refraction (at a wavelength o f
632.8 nm) o f BCB samples cured by different methods: hot plate, thermal furnace and the
VFM furnace under similar conditions. The index o f refraction decreases from 1.S9S for
a soft baked sample to about 1.555 for a fully cured sample. It is observed that at each o f
these conditions, the VFM cured samples show comparable or lower index of refraction
than the thermally cured films. This is consistent with the higher extent of cure attained in
these samples as compared to the thermally cured samples.
1.60
ee
s
&
e
M
w
•a
B
150
175
200
225
Temperature, *C
VFM cure - 5 min
Hot plate cure - 30 min
250
275
Furnace cure - 30 min
VFM cure - 30 min
Figure 18 Comparison o f in-plane index o f refraction o f BCB films cured in a furnace,
hot plate and VFM under different processing conditions.
76
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Similar trends are also seen on comparing the electrical properties o f films processed
in the VFM furnace and the thermal furnace. Figure 19 compares the dielectric constant
o f BCB films processed under different conditions by both these methods. The dielectric
constant decreases with increasing extent of cure and VFM cured films show comparable
dielectric constant to thermally cured films for all the cure conditions studied. The
dielectric constant of (VFM processed) folly cured BCB films was 2.69 and the loss
tangent was 0.0011.
3.40
jj
2.40
175*C
200*C
IVFM cure - 30 min
225*C
250*C
IThermal cure - 30 min
Figure 19 Dielectric constant o f VFM and thermally cured BCB films processed under
different conditions.
77
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The residual stress within the polymer films cured on silicon results from the CTE
(coefficient o f thermal expansion) mismatch between BCB and silicon. The magnitude
o f the residual stress has been shown to be dependent on the extent o f cure [93, 94]. The
residual stress of a soft baked film and a cured film at its cine temperature (maximum
temperature) is nearly zero.
Upon cooling to room temperature, the residual stress
increases to ->33 MPa (tensile) for a fully cured film. Figure 20 compares the room
temperature residual stress o f BCB films cured in the VFM and thermal furnace at
different temperatures as a function o f percent conversion (degree of cure). At each
temperature, VFM samples were cured for 5 min whereas the thermal samples were cured
for much longer times to reach about the same conversion. It may be seen that the
residual stress in films depends on the extent of cure and the cure method. For any
partially cured condition, VFM processed films show significantly lower residual stress
than the corresponding thermally cured films.
Further, there is a significant difference in the rate o f increase o f stress in the VFM
cured films as compared to the thermally cured films. As seen from Figure 20, the
residual stress o f thermally cured films increases linearly with percentage conversion and
gradually levels o f at high conversions. However, VFM cured films show very low stress
until about 70% conversion and thereafter a final residual stress close to that o f the
thermally cured films.
78
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
40
Residual Stress, MPa
35
30
25
20
15
10
5
0
40
50
60
Thermal Cure
O 175°C ■ 200°C
•
70
80
% C onversion
— —
90
100
VFM Cure
210°C A 225°C ♦ 250°C
Figure 20 Residual stress o f VFM and thermally cured films as a function o f percent
conversion.
79
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The effect o f VFM processing on film properties was studied by comparing the
properties o f VFM cured films with thermally cured films. Table 2 compares the optical,
electrical, mechanical and chemical properties o f VFM and thermally cured films. Table
1 indicates that a 5-minute VFM cure at 250°C gives properties comparable to films
cured at 250°C for one hour in a conventional thermal oven. The in-plane and out of
plane index o f refraction measurements o f cured films were performed using a
Metricon™ prism coupler. The optical birefringence o f VFM cured BCB films is as low
as 0.003 indicating the isotropy o f the cured films. The dielectric properties o f VFM
cured films are comparable to thermally cured films. VFM cured films have a moisture
uptake o f < 0.2% by weight at 85% relative humidity as determined by quartz crystal
microbalance measurements. VFM cured films show thermal stability comparable to
thermally cured films with < 1% weight loss after 1 hr at 330°C.
80
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 2 Comparison o f properties of BCB films cured by the standard thermal cure, 1 hr
at 250°C and VFM cure o f 5 min at 250°C.
Standard
Thermal Cure
VFM Cure
I hr at 250 °C
5 min at 250°C
1.55
1.55
0.002
0.003
Dielectric constant (at 10 kHz)
2.69
2.70
Loss tangent (at 10 kHz)
0.0009
0.0011
Thermal stability
(wt. loss after I hr at 330 *C)
<1%
<1%
Moisture uptake
(wt% absorbed at 85% RH)
0.194
0.197
Residual Stress ( MPa)
30
33
Property
In-plane index of Refraction
(at 632.8 nm)
Birefringence
4.3 Discussion
Microwave processing o f polymers has been shown to be an efficient, rapid
processing technique offering potential advantages over conventional thermal processing
(Chapter II).
Microwave processing differs from thermal processing in that heat is
produced within the material by dielectric loss mechanisms unlike thermal processing
where heat is transferred from the ambient.
81
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 11 shows that the heating rate and the ultimate temperature reached by films
cured at a constant power increase with increasing central frequency. The amount o f
microwave energy absorbed by a material at a given power depends on the applied
frequency and the dielectric behavior o f the material (Equation 2.11). The dielectric loss
primarily determines the relative electromagnetic dissipation or the rate o f conversion of
electrical energy into thermal energy. The dielectric loss o f a material varies with both
temperature and frequency. This depends on the charge distribution within bonds, chain
conformations, bulk morphology and the statistical thermal motion o f the polar groups in
the material.
Previous studies [95] on high frequency dielectric characterization of
dielectric materials indicate that the dielectric loss o f BCB increases with frequency in
the range 2.4 to 8.1 GHz. Hence, with increasing central frequency we would expect to
see greater absorption and dissipation o f microwave energy, which results in higher
heating rates and higher ultimate temperature.
Previous studies with hot plate [96] and infrared radiation [97] demonstrated the
feasibility o f rapid thermal curing o f BCB dielectric polymer. No significant differences
were observed in the properties of cured films such as residual stress and adhesion. No
enhancement in chemical reaction kinetics was reported by the absorption o f infrared
radiation. From Figure 12, the FTIR spectra o f VFM and thermally cured films are
identical. This suggests that the chemical structure o f VFM cured BCB films is similar to
that o f thermally cured films within the sensitivity o f FTIR. Furthermore, the absence o f
any distinct peaks in the entire range studied other than those seen in the thermally cured
films indicates that a ring opening mechanism followed by a Diels-Alder crosslinking
82
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
reaction is the primary reaction mechanism for BCB curing by VFM processing with no
significant alternative reaction pathways.
Microwave processing o f thermosetting resins such as epoxies, polyesters and
polyurethanes has been studied [25] and many promising property improvements along
with a reduction in cure time were shown. Thermoset resin systems are characterized by
gelation and vitrification phenomena. For BCB, gelation has been shown [75] to have
negligible effect on reaction rate indicating that local mobilities are virtually unaffected.
Vitrification however has a significant impact on the reaction rate. Vitrification occurs
when the glass transition temperature (Tg) equals the cure temperature. Loss o f free
volume associated with a glass transition reduces mobility locally.
Prior to vitrification, the crosslinking reaction is kinetically or chemically controlled
and after vitrification (i.e. once the polymer is in its glassy state), mobility is limited and
the rate o f reaction slows down and tends to become diffusion controlled.
This
phenomenon is very common in most thermosetting systems like epoxies. In the case o f
microwave processing, the dielectric loss in the microwave frequency range is primarily
due to dipolar polarization. Before vitrification, the reaction rates are high due to high
dipolar mobility, which can have translational apart from rotational contributions. After
vitrification, the mobility of the dipoles is rather constrained, and the mechanism o f loss
has primarily electronic and vibrational contributions. Figure 15 clearly shows a drop in
reaction rate after a certain time and conversion at a given temperature indicating
vitrification.
83
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Both chemical reaction controlled kinetics and diffusion controlled kinetics, are
favored by higher temperatures. It is possible that even though the bulk temperature of
the material is lower, the local temperature resulting from localized loss mechanisms
(relaxation o f polarization) is higher. This would result in reaction rates that are higher
than those expected at the bulk temperature. A higher conversion or shorter cure time
would result for any given processing temperature. Alternately, one would require a
lower processing temperature for the same cure time. The apparent activation energy for
the cure reaction by VFM processing, -25 kCal/mol, is about 30% lower than the
reported value o f 36 kcal/mole, which is not as significant an enhancement in kinetics as
other systems studied [48]. This is not surprising as microwave induced acceleration of
reaction kinetics is known to be significant in slower-reacting systems and the magnitude
o f the observed effect is greater at lower temperatures. Similar results o f enhancement in
reaction kinetics by microwave processing have been reported by a number o f previous
studies [68] on thermoset systems.
A proposed mechanism for enhanced kinetics
suggests [51] a nonequilibrium, nonuniform energy distribution on the molecular level,
which results in certain dipoles having a greater energy than the average energy of
adjacent dipoles. This increased energy was shown to correspond to a 50°C increase in
effective temperature for solution imidization studies.
A 10-minute VFM cure at 250°C gave the same conversion as the reference/control
full cure sample. At this temperature, for conventional thermal curing, vitrification is
known to occur in about five minutes giving a conversion o f about 90% and in the next
55 min o f the prescribed standard cure, reaction advances to only about 95%. Microwave
84
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
processing thus shows a significant improvement in the post-vitrification reaction rates.
Alternately, it is likely that due to high local temperatures, the transition to the glassy
phase occurs at a higher percent conversion for any given bulk temperature over a shorter
time. As a result o f higher reaction rates, the reaction proceeds farther to completion
before vitrification for microwave processing as compared to thermal processing.
Residual stress in thin-film dielectrics arises due to CTE mismatch between the films
and the substrate. The polymer film is at nearly zero stress at its cure temperature and
when it cools, it contracts more than the silicon substrate due to its higher CTE. This
contributes to tensile stress within the film.
The origin o f the lower residual stress of partially cured VFM cured films is not
known. Several factors may contribute to the lower residual stress in VFM cured films.
First, the residual solvent in VFM and thermally cured films may be different. The short
cure times o f VFM processed films (especially at low temperatures), may result in higher
residual solvent as compared to thermally cured films. The resulting solvent-induced
plasticization could lead to a lower effective modulus and hence lower residual stress.
However, this is probably not a significant factor since the index o f refraction and
dielectric constant o f VFM and thermally cured films, at the same conversion, are similar.
Moreover, VFM cured samples did not show a significant change in residual stress even
after vacuum treatment for 24 hrs to remove any residual solvent.
Second, there may be an apparent lower CTE mismatch penalty,
hi a thermal
furnace, the bulk o f the film and the substrate are at the same temperature as the furnace
temperature. On the other hand, in VFM processing, the bulk temperature o f the film, the
85
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
substrate and the local temperatures could be different. The higher local temperatures at
the reaction sites could result in enhanced crosslinking reaction as the mobility o f the
reacting species increases with increasing temperature. However, if the bulk o f the film
was at a lower temperature, the associated CTE mismatch penalty is due to the bulk
temperature rather than the local temperature. This could result in a lower residual stress.
Indeed, if the modulus o f films cured by both the techniques is the same, the lower
residual stress corroborates the hypothesis o f high local temperatures.
Residual stress is set in a film, upon cooling from the highest Tg that has been
reached [97] independent o f the maximum temperature excursion the film experiences in
reaching this degree o f cure. If curing occurs at a temperature greater than the Tg, the
film becomes plastic and stress is reset at the Tg corresponding to the higher conversion
reached. However, if the cure temperature is always below the Tg (i.e., reaction proceeds
in the solid or glassy state) the stress will not increase. Hence, if the Tg o f microwave
cured films were to be higher, enhanced post-vitrification reaction rates could lead to
lower residual stress. From the results, films cured in the VFM at a temperature as low as
210°C for 30 minutes give a conversion o f -95% and a residual stress o f about 22 MPa
which is lower than the residual stress for thermally cured films o f the same conversion.
Further, it would take a longer time to reach the same conversion in a thermal oven at that
temperature.
Lastly, possible differences in the intrinsic modulus and CTE o f the cured films
could contribute to differences in residual stress. Although IR spectra did not show any
significant differences in chemical structure, subtle differences in bonding or crosslinked
86
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
network structure and the nature o f crosslinks could exist, which can affect the bulk
properties o f films. A lower intrinsic modulus and/or CTE results in lower residual
stress. It was not possible to verify this hypothesis as measurement o f modulus and CTE
o f BCB films is very challenging.
Free thin films o f BCB are brittle and heating
characteristics on any other substrate could be significantly different.
Hardness and
modulus measurements from indentation techniques could validate this hypothesis.
4.4 Conclusions
Variable frequency microwave processing o f benzocyclobutene was investigated.
Results from this study show that VFM processing o f BCB is feasible. The chemical
structure o f VFM cured films is the same as thermally processed films. Ring opening
followed by Diels-Alder crosslinking is the primary reaction mechanism for VFM curing
o f BCB.
Study o f reaction kinetics shows that the cure reaction follows first order
kinetics before vitrification. The rate constants show an Arrhenius-type relationship with
temperature with an apparent activation energy o f 25.7 +/- 4.4 kCal/mol, which is about
30% lower than the reported thermal activation energy.
The optical, electrical,
mechanical and chemical properties o f VFM cured films were characterized and
compared with thermally cured films to determine the effectiveness o f VFM processing.
VFM cured films showed comparable or improved properties than thermally cured films.
Processing improvements such as shorter cure time or lower processing temperature can
be achieved. The residual stress o f VFM cured films was lower than thermally cured
films especially for partially cured films. Property measurements such as Tg, elastic
87
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
modulus and CTE could explain the lower residual stress in partially cured (VFM) BCB
films.
88
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER V
LOW TEMPERATURE RAPID CURING OF POLYIMIDES ON SILICON BY
VARIABLE FREQUENCY MICROWAVE PROCESSING
In this chapter, the results from studies on rapid low-temperature VFM curing of
polyimides on silicon are presented in detail. The effect o f low-temperature VFM curing
on the chemical, mechanical and thermal properties o f polyimide films was investigated.
Fourier transform infrared analysis was used to monitor the chemical changes in the
VFM cured films. Thermal stability o f cured films was studied by thermo gravimetric
analysis (TGA) and mass spectroscopy (MS) and was used as a metric to determine the
effectiveness o f low-temperature VFM curing. Structure-property relationships and their
dependence on VFM processing conditions were also investigated.
Polyimides have found a variety o f applications in the microelectronics industry such
as interlayer dielectrics, stress buffer passivation layers and alpha particle barriers.
Photosensitive polyimides are o f particular interest as they offer additional processing
and cost advantages. High processing temperatures can increase the thermo-mechanical
stress, and in some applications, degrade the device characteristics or other materials
present. In order to reduce the thermal budget, novel low-temperature curing methods
such as microwave processing are o f interest. In this study, microwave curing was used
to cure traditional polyimides on silicon substrates at reduced temperatures as compared
to traditional thermal curing. The objective was to investigate the feasibility o f VFM
89
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
curing o f polyimides at a lower temperature to obtain properties similar to a traditional
high-temperature (350°C) thermal furnace cure. The targeted maximum cure temperature
for curing on silicon was 275°C.
Several commercially available polyimides with
different backbone structures and chemistries and a polybenzoxazole based dielectric
were evaluated in this study: PI 2611, HD 4000 and XP 7001 from HD Microsystems,
CRC 8650 from Sumitomo Bakelite Company, PWDC 1000 and PW 1200 from Toray Dow Coming Inc.
PI 2611 is a non-photosensitive rigid-rod polyimide based on
biphenyltetracarboxylic acid dianhydride/p-phenylenediamine (BPDA-PDA) chemistry
and yields a low residual stress upon traditional thermal curing in a furnace as a
consequence o f a high degree o f in-plane orientation [98, 99, 100]. HD 4000 and XP
7001 are negative-tone photosensitive polyimides based on an acrylate monomer
crosslinking chemistry, with a semi-rigid rod and flexible backbone structure
respectively. The manufacturer recommended thermal cure (referred to as ‘standard
thermal cure* hereafter) for all three polyimides is a one hour cure at 350°C ramped at
3°C/min in a conventional thermal oven [101].
CRC 8650 is a polybenzoxazole chemistry based positive tone dielectric resin from
Sumitomo Bakelite Co. with a diazonaphthoquinone (DNQ) based photochemistry (VFM
curing o f CRC 8650 is discussed in detail in Chapter VI). PWDC 1000 and PW 1200 are
positive tone polyimides manufactured by Toray Industries, Inc. Japan, under the trade
nam e Photoneece®. These are DNQ-based photosensitive polyimides with a novel
(proprietary) crosslinking chemistry [102] aimed at improving the mechanical properties
and increasing the glass transition temperature o f the cured films. PW 1200 has a lower
90
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
cure temperature (Refer Table 3) and has better adhesion strength as compared to PWDC
1000.
The effectiveness o f low-temperature VFM processing was determined by
comparing the properties o f cured films with those o f standard thermal cured films.
Polyimide film quality was evaluated based on the degree o f imidization, residual
solvent, residual stress, birefringence and the Young’s modulus.
The effect o f cure
temperature, hold time, ramp rate, and intermediate holds was studied in order to identify
an optimum VFM low-temperature curing condition.
Some optical, thermal, and
mechanical properties and their dependence on structure and processing conditions are
also discussed.
For these curing studies, thin films o f polyimide (S-20 pm) were spin cast onto 4inch diameter <100> silicon wafers. A Raytek Therm alert T30 series infrared pyrometer
was used for temperature measurement.
The infrared pyrometer was calibrated for
emissivity o f the sample using a hot plate. The spin-coating, softbake, exposure dose
(post-exposure bake if applicable) and standard manufacturer recommended cure
conditions for all the materials are summarized in Table 3.
The manufacturer
recommended cure schedule outlined in Table 3 will be referred to as the 'standard
thermal cure’ condition for each material. The negative tone polyimides were blanket
exposed before curing and the positive tone dielectrics were cured unexposed.
The degree o f imidization o f the cured films was monitored by Fourier transform
infrared spectroscopy (FTIR) analysis. Infrared spectra were collected in the attenuated
total reflection (ATR) mode using a Nicolet Magna-IR Fourier transform infrared
spectrometer. All spectra were recorded at a resolution o f 4 cm '1 and averaged over 128
91
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
scans. The degree of imidization was calculated according to following equation where
A is the absorbance and the subscript notes the peak assignment. The reference
absorbencies are determined from a film for which 100% imidization is assumed
(standard thermal cure).
% Imidization = ---------------------------------------ring I
x
100
nfmmet )s*.Th*nmsl
92
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(5.1)
Table 3 Material processing conditions.
er-Trade name
PI 2611
HD 4000
XP 7001
Manufacturer
Cure Conditions
HDMS
Softbake: 2 min at 120*C
ure: 3°C/min to 350°C 1 hr
HDMS
Softbake: 2 min at 8 S°C, 2 min at 95*C
Exposure: 200mJ/cm2 (I-line)
Post exposure bake: 7(PC 1 min
Cure: 3°C/min to 350°C, lhr
HDMS
Softbake: 2 min at 90°C, 2 min at 100°C
Exposure: 400mJ/cm2 (I-line)
Post exposure bake: none
Cure: 3°C/min to 3S0°C, lhr
CRC 8650
Softbake: 4 min at 125°C
Sumitomo Bakelite Exposure: 1ISOmJ/cm2 (g-line)
Post exposure bake: None
Co.
Cure: 3*C/min to 150°C 30min,
320°C 30mm
PWDC 1000
Softbake: 3 min at 120®C
Exposure: 675mJ/cm2 (I-line)
Toray Industries Inc. Post exposure bake: Nooe
Cure: 3°C/min to 170°C 30min,
2S0*C 30min,
320°C 60 min
Softbake: 3 min at 120°C
Exposure: 275mJ/cm2 (I-line)
PW 1200
Toray Industries Inc. Past exposure bake: None
Cure: 3°C/min to 140°C 30min,
350*C 60mm
' HDMS: Hitachi Dupont Microsystems
All softbake times on hot plate
93
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.1 Results
The feasibility o f VFM curing o f six different polymer dielectrics at a low
temperature on silicon substrates to obtain properties similar to a traditional hightemperature (350°C) thermal furnace cure was investigated.
The impact o f VFM
processing on the chemical and mechanical properties and the thermal stability o f the
cured films was studied. For ease o f organization, the results and discussion will be
presented in the following order. The results from the non photosensitive rigid-rod
polyimide, PI 2611 will be presented followed by those from VFM curing o f negative
tone photosensitive polyimides HD 4000 and XP 7001. Finally, the results from VFM
curing o f positive tone photosensitive dielectrics, CRC 8650, PWDC 1000 and PW 1200
will be discussed.
5.1.1 PI 2611
Figure 21 compares the FTIR spectra o f a) a soft baked film, b) a standard thermal
cured film and c) a low-temperature (275°C) VFM cured film o f PI 2611. The changes in
the FTIR absorbance after the cure reaction and the corresponding peak assignments are
summarized in Table 4. Notable changes in the spectra (comparing spectra (a) and (b))
are the appearance o f the different imide ring absorbencies including the -C-N-C- stretch
which has an IR absorbance at 1359 cm '1. As seen from these spectra, there are no
significant differences in the FTIR absorbance o f the low-temperature VFM cured films
(spectrum c) and the standard thermal cured films (spectrum b) suggesting that the
94
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
chemical structure o f VFM cured films is identical to that o f conventional thermal cured
films.
This shows that a similar condensation reaction for ring closure (or imide
Absorbance
formation) with no detectable side reactions occurs during VFM curing.
(b)
/V
2000
1500
1000
Wavennmbers (cm-1)
Figure 21 FTIR spectra o f a) softbaked, b) standard thermally cured, and 3) lowtemperature VFM cured PI 2611 films.
95
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 4 Polyimide characteristic IR peaks and probable assignments.
Peak Waveaumber (cm '*)
Probable assignment
1776
Symmetric C = 0 stretch
1720
Asymmetric C = 0 stretch
1680
Solvent NMP
1660
Amide II
1600-1606
1516
1359/1370
736
1054
Aromatic C=C stretching
Comments
Coupled stretch o f the fivemembered imide ring
Same as above
Internal reference
1,4 C 6H4 substituted
benzene
Ring breathing vibration
associated with aromatic
diamine; intensity does not
change with cure hence used
as internal reference
C-N-C stretch o f imide ring
Peak generally used for
calculating extent o f cure.
A bsent in uncured film and
increases as imidization
proceeds
C-N-C bending
Imide IV bending out o f
plane
-C-O-C stretching o f
benzoxazole ring
Increases with cure reaction
96
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The percent imidization achieved in cured films was estimated by FTIR analysis
using Equation S.l. The peak heights corresponding to the standard thermal cure were
taken as 100% imidized. For PI 2611, the absorbance at 1359 cm '1, which corresponds to
the -C-N-C- stretch o f the imide ring, was used to identify ring closure. The absorbance
o f the -C-H stretch o f the aromatic diamine, which occurs at 1516 cm '1, remained
unchanged during curing and was used as the reference peak. The peak at 1359 cm*1 was
chosen for estimation o f imide ring formation as it is reported to be independent o f side
reactions and dichroic effects [103].
Figure 22 shows the percent imidization o f PI 2611 films cured in the VFM. All the
samples were ramped at 15°C/min up to the desired temperature (as indicated on the xaxis o f Figure 22) and held for 5 min. As seen in Figure 22, almost complete imidization
at temperatures over 225°C with a 5 min hold time at the final temperature can be
achieved by VFM curing o f PI 2611. At high percent conversion, orientation effects and
subtle changes in the baseline can alter the estimates o f imidization. Hence, estimates o f
percent imidization by FTIR analysis are semi-quantitative in nature.
97
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
150
IM
1
*
5.
• PI2611
■ HD4000
AXP7001
-I__________ L_
150
175
200
225
250
275
300
325
350
Temperature, *C
Figure 22 Effect o f cure temperature on the percent imidization o f PI 2611, HD 4000 and
XP 7001. Ramp rate: 15°C/min, Hold time: 5 min.
Figure 23 shows the effect o f temperature on the birefringence o f VFM cured films.
Each o f these films was ramped at 15°C/min to the final temperature and held for S min.
As seen in Figure 23, the birefringence o f the cured films increased with cure temperature
for temperatures less than 275°C (the error bars account for the data scatter over an
average o f five measurements at each condition).
98
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0.25
ea 0.15
175
200
225
250
275
300
325
350
375
Temperature, °C
Figure 23 Effect o f cure temperature on the birefringence o f VFM cured PI 2611.
Figure 24 shows the effect o f ramp rate on the birefringence o f PI 2611 films VFM
cured at 27S°C for 5 min each. The birefringence decreases only slightly for films cured
at 27S°C (from 0.2195 to 0.2122). PI 2611 films rapid cured thermally on a hotplate at
350°C without a ramp showed two distinct modes in the Metricon (high and low
refractive indices), possibly due to increase in crystalline order in the film.
The
birefringence estimated form the low index (amorphous) region ranged between 0.1254
and 0.1623.
The reduced orientation in the rapid thermal cured film increased the
residual stress in the film to ~ 33 MPa. This effect was also reported [79] previously
where a high residual stress o f 35 MPa was obtained for rapid hot plate cured films.
99
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
From Figure 24, the birefringence o f films cured under all the four conditions is still
significantly higher than that for rapid thermal cured films.
0.24
0.22
w
w
s
V
DC
0.20
1
0.18
u
S
0.16
275°C
0.14
0.12
20
40
60
80
Ram p rate, °C/m in
Figure 24 Effect o f ramp rate on the birefringence o f VFM cured PI 2611 films.
The imidization in the films cured at a final cure temperature o f 275°C was about
100% and was independent o f the ramp rate (between IS to 60°C/min). Also, as seen
from Table 5, the birefringence slightly increased with cure time at 27S°C and
significantly with the introduction o f an intermediate hold at 200°C. Thermo gravimetric
100
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
analysis was performed to estimate the amount o f residual volatiles in the cured films.
Table 5 shows the dependence o f percent weight loss as a function o f VFM cure
conditions for PI 2611 films. For TGA analysis, each o f these cured films was ramped at
a constant rate o f 10°C/min to SOO°C (650°C in some cases) in a nitrogen atmosphere.
The total weight loss during this temperature ramp as well as the T/% and Ts% are shown
in Table 5.
From Table S (comparing C, D, J, and K), it can be seen that for the same ramp rate
and cure time, the percent weight loss in VFM cured films decreased with increasing cure
temperature.
From Table 5, it can also be seen that for the same cure time and
temperature, the TGA results are independent o f ramp rate (conditions D, F, G and H)
suggesting negligible differences in solvent content. Further, an intermediate hold at a
temperature o f 200°C (L) and longer cure times (E) reduced the solvent content and
increased the temperature stability o f cured films.
101
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 5 Effect o f VFM cure conditions on the optical and thermal properties o f polyimide
PI 2611 films.
Cure Schedule
W eight loss
Birefringence
@ 500°C
An —njE * nTM
T 1%(°C)
T5./.(°C )
%
A
3°C/min, 350°C, 60 min
0.23 IS
|
< 1.0
B
3°C/min, 275°C, 60 min
0.2037
j
c
15°C/min, 250°C, 5 min
D
3.0
529.13
340.83__
588.39
581.78
0.1908
4.0
262.42
575.76
15°C/min, 275°C, 5 min
0.2195
3.2
286.90
585.06
E
15°C/min, 275°C, 10 min
0.2213
22
306.38
592.73
F
30°C/min, 275°C, 5 min
0.2175
3.3
278.42
583.02
G
45°C/min, 275°C, 5 min
0.2128
3.2
279.52
582.20
H
60°C/min, 275°C, 5 min
0.2122
3.0
277.88
J
15°C/min, 300“C, 5 min
0.2053
2.5
-
584.13
-
K
I5°C/min, 350°C, 5 min
0.2095
< 1.0
-
-
L
15°C/min, 200°C, 5 min
...............275°C 10 min
0.2440
< 1.0
519.4
584.56
102
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The residual stress and elastic modulus o f PI 2611 films cured at 275°C by VFM for
5 minutes as a function o f ramp rate are shown in Figure 25. As seen in Figure 25, both
the residual stress and Young’s modulus o f the cured films show a sim ilar trend with
increasing ramp rate. The residual stress and modulus o f PI 2611 are not affected by
ramp rate with a final cure at 275°C except for very high ramp rates (60°C/min). This
could possibly be due to changes in the microstructure or morphology o f the cured films
with increasing ramp rate. Previously, it was shown that rapid thermally cured PI 2611
films had a high residual stress (~ 35 MPa) [79, 104]. It is important to note that the
residual stress o f VFM cured PI 2611 films is much lower owing to the high degree o f in­
plane orientation. Moreover, increasing the cure time and introducing an intermediate
hold in the VFM cure cycle at 200°C further lowered the residual stress with about the
same Young’s modulus. This is labeled ‘two step’ in Figure 25 and corresponds to cure
condition L o f Table 5.
103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Residual Stress, MPa
12
10
Cm
O
8
6
i09
4
"Ofi
£
2
0
0
0
10
20
30
40
50
60
Ramp rate, °C/min
Residual Stress
Elastic Modulus
O Two Step: Residual Stress
□ Two Step: Elastic Modulus
Figure 25 Effect o f ramp rate on mechanical properties o f VFM cured PI 2611 films.
104
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TGA studies show that a temperature o f 27S°C or higher is required to achieve low
residual volatiles.
Moderate ramp rates and adequate cure times were essential for
achieving high orientation o f PI 2611 and preserving the low stress advantage provided
by traditional slow thermal cured films. Based on these observations, an optimum low
temperature VFM cure condition (L) for PI 2611 was chosen.
Comparison o f the
properties o f VFM cured PI 2611 under these conditions (L) shows that properties similar
to the standard thermal cure (350°C) and much better than a low temperature (27S°C)
thermal cure can be achieved in a much shorter cure time via a low temperature (27S°C)
VFM cure [105]. For example, Figure 26 compares the thermal stability o f PI 2611 films
cured under three different conditions: a) the standard thermal cure o f 1 hr at 350°C, b)
low-temperature (275°C) VFM cure (L) and c) low-temperature thermal cure (1 hr at
275°C). As seen in Figure 26, the low-temperature thermal cured film begins to lose
weight early showing a Ti% at 340.8°C whereas the low-temperature VFM cured film
shows a Ti% o f 519.4°C which is comparable to the standard thermal cured film (Ti% o f
529.1°C).
105
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
% Mass remaining
100
90
50
150
250
350
450
Temperature, °C
550
650
!------ Std. Thermal Cure ( 350*0 o Low Temp VFM ( 275*) » Low Temp Thermal ( 275° Q
Figure 26 Comparison o f low temperature VFM and thermally cured samples with
standard thermally cured PI 2611.
106
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.1.2 Negative tone photosensitive polyimides (HD 4000 and XP 7001):
Figure 27 compares the FTIR spectra o f a) a soft baked film, b) a standard thermal
cured film and c) a low-temperature (275°C) VFM cured film o f HD 4000 respectively.
Notable changes in the spectra (comparing spectra (a) and (b)) are the appearance o f the
imide ring absorbencies The -C-N-C- stretch o f the imide ring has an IR absorbance at
~1370 cm '1 in HD 4000. As seen from Figure 27, there are no significant differences in
the FTIR spectra (b) and (c) suggesting that the chemical structure o f low-temperature
VFM cured films is identical to that o f conventional thermal cured films. Similar FTIR
studies on XP 7001 showed that the chemical structure o f low-temperature VFM cured
films was similar to that o f standard thermal cured films.
In the case o f HD 4000 and XP 7001, for the estimation o f the percent imidization
achieved in cured films, the absorbance at 1370 cm '1 was chosen to identify imide ring
formation and the peak absorbencies at 1008 cm '1 and 1608 cm '1 were used as internal
references. Figure 22 compares the percent imidization o f PI 2611, HD 4000 and XP
7001 films cured in the VFM. As seen in Figure 22, all three polyimides show similar
imidization curves as a function o f temperature and achieve almost complete imidization
at 225°C with a 5 min hold time at the final temperature by VFM processing.
107
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
/\
V
A
2000
1500
«
V ajb/w
.
^
1000
Wavennmbers (cm-1)
Figure 27 FTIR spectra o f a) softbaked, b) standard thermally cured, and 3) VFM cured
HD 4000 film.
The heat (energy) changes during the standard thermal curing o f polyimides HD
4000 and XP 7001 were followed by DSC as shown in Figure 28. A peak in the DSC
corresponding to the evolution o f solvent and other volatiles occurs at a temperature of
about 165°C. The endothermic peak centered around 185°C is attributed mainly to the
imidization reaction. This imidization reaction occurs until ~240°C. This is consistent
with the FTIR analysis and estimates o f imidization, which show that imidization is
completed at temperatures above 225°C. Unlike the non-photosensitive polyimide, PI
2611, HD 4000 and XP 7001 exhibit an additional high temperature endothermic peak, as
108
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
shown in Figure 28. This endotherm can be attributed to the reaction by-products o f the
photo-package constituting HD 4000 and XP 7001 that continue to evolve at high
temperatures.
2000
400
1000
350
300
|
250
«
-1000
200
g -2000
150
-3000
100
*— HD4000
——XP7001
-C- —Curing Profile
-4000
-5000
0
50
100
150
200
Tim e (m in)
Figure 28 In-situ curing o f HD4000 and XP7001 in the DSC.
Table 6 and Table 7 show the optical and thermal properties o f HD 4000 and XP
7001 respectively. Several key observations can be made from Table 6 and 7.
109
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 6 Effect o f cure conditions on the optical and thermal properties o f HD 4000.
Cure Schedule
IlTE
»TM
An
TS%(°C)
Therm al
A
3°C/min, 350°C 60min
1.6611
1.6428
0.0183
404
B
3°C/min, 275°C 60min
1.6222
1.6112
0.0110
343
VFM
C
15°C/min, 350°C 5min
1.6663
1.6489
0.0174
402
D
15°C/min, 275°C 5min
1.6313
1.6212
0.0101
334
F.
15°C/min, 275°C ISmin
1.6328
1.6195
0.0133
339
G
15°C/min, 275°C 30min
1.6375
1.6260
0.0115
353
H
15°C/min, 200°C lOmin
275°C 30min
1.6291
1.6175
0.0116
353
J
15°C/min, 200°C 20min
275°C 30min
1.6235
1.6139
0.0096
353
110
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
iite
O tm
An
O
n
Cure Schedule
$
Table 7 Effect o f cure conditions on the optical and thermal properties o f XP 7001.
Thermal
A
3°C/min, 350°C 60min
1.6621
1.6481
0.0131
398
B
3°C/min, 275°C 60min
1.6287
1.6201
0.0086
363
VFM
C
15°C/min, 350°C 30min
1.6396
1.6314
0.0082
385
D
15°C/min, 275°C 5min
1.6325
1.6261
0.0064
354
1.6333
1.6273
0.0069
374
E
15°C/min, 275°C ISmin
F
15°C/min, 275°C 30min
1.6357
1.6276
0.0081
373
G
15°C/min, 200°C lOmin
275°C ISmin
1.6299
1.6153
0.0075
370
I ll
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The photosensitive polyimides HD 4000 and XP 7001 typically show a lower
birefringence (suggesting a lower degree o f anisotropy and/or orientation) as compared to
PI 2611, which is highly anisotropic. As noted earlier, HD 4000 has a semi-rigid and XP
7001 has a flexible backbone, whereas PI 2611 has a rigid-rod backbone chemistry.
Accordingly, the birefringence o f flexible XP 7001 is lower than that o f semi-rigid HD
4000 which is in-tum lower than that o f PI 2611. Further, the birefringence o f both HD
4000 and XP 7001 increased with increasing curing temperature, as was the case with PI
2611.
As seen from Table 6 and Table 7, a short time VFM cure at 350°C (B) gives similar
thermal properties as compared to films cured at 350°C for lh r with a slow 3°C/min ramp
rate (A) in a conventional thermal oven. At a lower cure temperature o f 275°C, VFM
curing shows better thermal properties and chain alignment than conventional thermal
curing in both HD 4000 (Table 6 B and G) and XP 7001 (Table 7 B and E).
From Table 6 and 7, it can be seen that for low-temperature cures, XP 7001 films
show a higher T$% than HD 4000. Also, the Ts% o f both polymers increases with cure
time at 275°C indicating that the photoproducts from HD 4000 and XP 7001 continue to
evolve at 27S°C (or higher), consistent with the DSC endotherm noted earlier.
Furthermore, it can be seen that it takes a longer time for HD 4000 than for XP 7001 to
achieve the maximum Ts% possible at 275°C by VFM processing suggesting that removal
o f photoproducts is easier in XP 7001 than in HD 4000. The effect o f an intermediate
hold at 200°C in VFM curing on the thermal and optical properties was also investigated.
As seen from Table 6 and Table 7, unlike PI 2611, an intermediate hold at 200°C did not
112
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
significantly change the birefringence or the T$% in either HD 4000 (Table 6 G ->H ->J)
or XP 7001 (Table 7 F-» G).
Table 8 shows the effect o f cure conditions on the Young’s modulus o f VFM cured
HD 4000 and XP 7001 polyimide films (the standard deviation on these measurements
was in the range 0.01 to 0.20 GPa).
For films cured for 5 min at the final cure
temperature, the Young’s modulus o f HD 4000 shows a greater dependence on cure
temperature than XP 7001. The Young’s modulus o f HD 4000 increases with cure
temperature until 27S°C and then remains constant. The Young’s modulus o f XP 7001 is
relatively independent o f cure temperature. Table 8 also compares the Young’s modulus
o f VFM cured HD 4000 and XP 7001 films to the thermally cured films. The Young’s
modulus o f HD 4000 decreases with cure time at 275°C whereas that o f XP 7001 is
relatively constant. This can be attributed to differences in the nature and amount o f
photoproducts and backbone flexibility.
The effect o f VFM cure conditions on the room temperature residual stress o f HD
4000 and XP 7001 is shown in Table 9. It can be seen that at every cure condition, the
flexible backbone XP 7001 has a lower residual stress than the semi-rigid backbone HD
4000. However, the residual stress is much higher than the rigid-rod like PI 2611. From
Table 9, it is also seen that the residual stress in the standard therm ally cured (A) HD
4000 films is slightly higher than that for XP 7001 while the residual stress o f the low
temperature (275°C) VFM cured HD 4000 is much higher than that o f XP 7001. The
effect o f intermediate hold tim e at 200°C on the residual stress o f HD 4000 and XP 7001
113
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
is also shown in Table 9 (I -> J ->K). The introduction o f an intermediate hold decreased
the residual stress for both the negative tone photosensitive polyimides.
114
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 8 Effect o f cure conditions on the Young’s modulus o f VFM and thermally cured
HD 4000 and XP 7001 polyimide films.
Young's Modulus (GPa)
XP 7001
HD 4000
Cure Schedule
3 C/min, 350*XJ, 60 min
3 C/rain, 275°C, 60 min
c
15°C/min, 225°C, 5 min
2.98
-
D
15°C/min, 250“C, 5 min
3.11
2.29
E
15°C/min, 275°C, 5 min
3.37
2.17
F
15°C/min, 350°C, 5 min
3.22
2.35
G
15°C/min, 275°C, 15 min
3.18
2.28
H
15°C/min, 275°C, 30 min
3.03
2.36
I
15°C/min, 200DC, 10 min
275°C, 30 min
3.00
-
J
15°C/min, 200°C, 10 min
275°C, 15 min
-
2.44
115
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 9 Effect o f cure conditions on the residual stress o f VFM and thermally cured HD
4000 and XP 7001 polyimide films.
Residual Stress (MPa)
XP 7001
HD 4000
Cure Schedule
3 C/min, 350°C, 60 mm
3 C/min, 275#C, 60 min
c
l5°C/min, 225°C, 5 min
25.9
22.4
D
15°C/min, 250°C, 5 min
26.6
23.2
E
15°C/min, 275°C, 5 min
28.0
23.2
F
15°C/min, 350°C, 5 min
29.2
26.1
G
l5°C/min, 275#C, 15 min
29.2
23.2
H
15°C/min, 275°C, 30 min
29.2
24.4
I
15°C/mm, 200°C, 10 min
275°C, 30 min
26.5
-
J
15°C/min, 200°C, 20 min
275°C, 30 min
24.9
-
K
15°C/min, 200°C, 10 min
275°C, 15 min
-
20.1
116
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.1.3 Positive tone photosensitive dielectrics (CRC 8650, PWDC1000 and PW1200):
A few studies on the effect o f rapid low-temperature VFM curing on the properties
o f positive tone photosensitive polyimides were performed with emphasis on the thermal
stability o f cured films. For CRC 8650, the extent o f conversion (% ring closure) was
estimated as a ratio o f the peak height o f the absorbance o f the benzoxazole ring, which
occurs at wavenumber o f 1054 cm*1 in the sample film, to that o f the standard thermal
cured film.
Both peaks were normalized to the peak height o f the absorbance at
wavenumber 1600 cm*1, which corresponds to the -C=C- stretching vibrations o f the
aromatic ring. FTIR studies on CRC 8650 showed that significant ring closure could be
achieved by VFM processing in a short process time at temperatures as low as 275°C [to
be discussed in Chapter VI].
The FTIR spectra o f PWDC 1000 a) after prebake and b) standard thermal cure are
shown in Figure 29. Notable peaks in Figure 29, are the imide peak at 1370 cm*1, which
increased as the cure reaction proceeded and the amide absorbance at 1657 cm*1, which
decreased upon cure. The low temperature curable PW 1200 also showed similar FT-IR
behavior. The aromatic ring stretch at 1606 cm*1 was chosen as the internal reference.
FTIR studies showed that complete imidization o f PWDC 1000 was possible by VFM
curing only at temperatures > 275°C whereas complete imidization was observed in the
case o f PW 1200 after 30 min at 275°C.
117
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1606
0.030 -j
1370
u 0.020
cm
§S 0 0 1 5 -j
<
1
0.010
]
0005 i
—i22Q0
2000
1800
1800
1400
1200
WOO
800
W a w iu a n t a t (cm -1)
Figure 29 FT-IR spectra o f PWDC 1000 (a) after pre bake and before exposure, and (b)
after standard thermal cure.
As discussed earlier, FTIR studies on CRC 6850 indicated that complete conversion
o f hydroxy-amide groups to the benzoxazole rings could be achieved only at temperature
o f > 275°C by VFM processing. These results suggest that apart from solvent removal,
the evolution o f photoproducts from CRC 6850 films is critical not only to improve the
thermal stability o f cured films but also to achieve complete reaction. Accordingly,
thermal stability studies were performed only on the films showing close to complete
reaction. The results are tabulated in Table 10. It can be seen that a slow-ramp one-hour
thermal cure at 275°C (B) produces only 78% curing whereas the three-step VFM cure
(D) gives almost complete ring closure at 275°C. Further, the three-step VFM cure at
275°C (D) gives a higher Ts% than a 1 hr thermal cure at 275°C.
118
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 10 Effect o f cure conditions on the percent reaction and thermal stability o f VFM
and thermally cured CRC 8650 films.
T1%(°C)
T5%(°C)
3°C/min, 150°C, 30 min
................ 320°C, 30 m in
360.7
432.6
3°C/min, 150°C, 30 min
................ 275°C, 60 m in
326.6
400.4
15°C/min, 150°C, 5 min
...................320°C, 5 m in
351.9
421.9
30°C/min, 150°C, 5 min
30°C/min, 250°C, 10 min
15°C/min, 275°C, 30 m in
350.6
419.4
Cure Schedule
% Reaction
The thermal stability and percent imidization o f PWDC 1000 and PW1200 as a
function o f cure method and cure conditions are shown in Table 11. It can be seen that
for high temperature cures, (320°C for PWDC 1000: conditions A & E and 350°C for PW
1200: conditions F & K), VFM processing gives complete imidization and the same or
better thermal stability for a much shorter cure time (a faster ramp rate and a single step
as compared to the standard thermal cure). As was the case with the negative tone
polyimides, the thermal stability improved with cure temperature and time at cure
temperature.
119
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
However, at a lower cure temperature (27S°C), films cured in the VFM furnace were
not completely imidized and did not show any improvement in thermal stability as
compared to thermally cured films. For example, in the case o f PWDC 1000, even a 60
min VFM cure at 275°C (D) shows the same T$% as a thermal cure at 275°C (B). For PW
1200, VFM curing gave marginal improvement (G & I) in both degree o f imidization and
thermal stability than thermal curing.
The effect o f an intermediate hold on the
effectiveness o f VFM curing was also investigated for PWDC 1000. No significant
increase in the percent imidization or the thermal stability was observed with
intermediate holds at 170°C and 250°C.
Longer hold times at an intermediate
temperature (> 10 min at 170°C and 250°C) slightly reduced both the percent imidization
and the thermal stability.
120
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 11 Effect o f cure conditions on the thermal properties and percent imidization o f
PWDC1000 and PW 1200 polyimide films.
C ure
Method
PMme 1000
A
Cure conditions
T is (°C)
r ,* (° c )
% Imidization
Thermal
3/170(30)/250(30)/320(60)
351
400
100
B
Thermal
3/170(30)/250(30)/275(60)
317
360
94
C
VFM
15/275(30)
292
346
85
D
VFM
15/275(60)
300
356
89
E
VFM
15/320(30)
354
414
100
PW 1200
F
Thermal
3/140(30)/320(60)
385
450
100
G
Thermal
3/140(30)/275(60)
307
375
94
H
VFM
15/275(30)
305
373
99
I
VFM
15/275(60)
311
383
99
J
VFM
15/295(60)
342
412
100
K
VFM
15/320(30)
387
459
100
121
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.2 Discussion
Microwave processing o f polymers has been shown to be an efficient, rapid
processing technique offering potential advantages over conventional thermal processing.
Microwave processing differs from thermal processing in that heat is transferred within
the bulk o f the material via dielectric loss mechanisms unlike thermal processing where
heat is transferred via conduction. It has been established by earlier studies that VFM
curing o f polyimides gives similar or improved properties in a shorter cure time as
compared to conventional thermal curing at the same temperature [79]. The efforts o f
this study are focused on determining the feasibility o f achieving similar properties using
VFM curing at a reduced temperature as compared to thermal curing.
From Figure 21, and Figure 27, it can be seen that the FTIR spectra o f the standard
thermal cured films (b) and the low-temperature VFM cured films (c) are essentially
identical with no significant differences in the range 1000 to 4000 cm'1. This shows that
the chemical structure o f films cured by either method is the same (within the sensitivity
o f the IR spectrometer). From Figure 22, it can be seen that for all the three polyimides
investigated, the imidization reaction reaches completion for temperatures above 225°C
even for a 5 min cure in the VFM oven. This is consistent with other reports in literature
which report that the imidization reaction by itself begins at 160°C and completes at a
temperature o f ~ 240°C. The acceleration in reaction kinetics by microwave curing can
be understood by the presence o f a local temperature (of the dipoles or the reacting
species), which is higher than the bulk temperature,
hi other words, the high local
122
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
temperature can help drive the reaction faster or achieve the same conversion at a lower
bulk temperature.
As mentioned earlier, PI 2611 is based on BPDA-PDA chemistry, which is a rigidrod type backbone. This material has a low stress when cured on silicon by the slow
standard thermal cure [106, 107, 108, 109]. The in-plane orientation, morphology, film
properties and stress depend on the processing conditions used. High temperature curing
using a slow ramp rate promotes a high degree o f in-plane chain orientation, which
results in a CTE match with silicon yielding zero residual stress. Rapid curing with high
ramp rates in a thermal oven does not allow enough time for orientation to occur and
leads to a high residual stress. Previous studies demonstrated that low residual stress
could be achieved by rapid curing o f PI 2611 by VFM processing at high temperatures
(3S0°C or greater) [48]. High temperature cures are typically used to improve orientation
and thermal stability o f cured films.
From Figure 23, the birefringence o f VFM cured films increased with cure
temperature from 22S°C to 350“C. (It is important to note that only films cured above
22S°C showed complete imidization.) These results are similar to literature reports on
thermally cured films using a slow heating rate [106].
Increased in-plane molecular
orientation with temperature in the range 250 to 350°C is responsible for increase in
birefringence.
At higher temperatures (> 350°C), the increase in crystalline order
affected the birefringence and increased the stress. An increase in the ramp rate o f VFM
cured samples reduces the birefringence (Figure 24) due to reduced chain orientation, bi­
plane chain orientation in the film depends on a) the intrinsic chain mobility, which is a
123
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
function o f backbone rigidity, extent o f imidization and amount o f residual solvent, and
b) the processing conditions used especially the ramp rate and temperature. The ramp
rate used influences the ‘effective glass transition temperature’ o f the film [106]. When
the film temperature is above its Tg (effective) during the cure process, as in the case o f
rapid thermal heating, it allows molecular relaxation to occur which leads to loss o f in­
plane orientation. Single frequency microwave cured BPDA-PDA films are reported [45]
to show loss o f in-plane orientation at a ramp rate o f 45°C/min while VFM processed
films showed [79] high in-plane orientation even at ramp rates as high as 60°C/min. Cure
time has only a secondary effect on birefringence.
Increase in packing density or
densification o f the film due to solvent less increases the index o f refraction. However,
as discussed earlier, ramp rate is more critical in determining the in-plane orientation.
The introduction o f an intermediate hold at 200°C also increases the birefringence. This
could be due to an increase in the effective glass transition temperature due to solvent
loss at 200°C, which causes curing to occur below the effective Tg thus promoting in­
plane orientation, which could also reduce the in-plane CTE and hence the residual stress.
Rapid imidization and solvent evolution by microwave processing could significantly
alter the rate o f change o f the effective Tg.
This could produce morphologies and
properties different from rapid thermal processing.
For example, conditions that
constantly enable curing below the effective Tg could result in increased in-plane
orientation as compared to rapid thermal heating and hence result in low residual stress.
From Table 5, it may be seen that for a given ramp rate and cure time, the thermal
stability o f cured films increases with increasing cure temperature suggesting increased
124
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
imidization and solvent evaporation during the cure process. This volume shrinkage of
films due to imidization and solvent evaporation occurs preferentially in the thickness
direction as the spin-cast film is constrained in the in-plane direction by the substrate
[106]. The relatively rigid backbone o f BPDA-PDA and the increased stiffness due to
imide ring formation constrain mobility and restrict the molecule from relaxing this
orientation resulting in increased anisotropy which explains the increase in birefringence
(Figure 23) with cure temperature.
The TGA results show negligible differences in
solvent content for films cured with different ramp rates (Table 3 D, F-H) indicating that
lowering of birefringence with ramp rate is primarily due to reduced orientation rather
than a solvent effect
Also, longer cure times did not significantly impact the
birefringence (orientation) and slightly increased the T$% for cure films. Increased cure
times remove any residual solvent from the films. Rapid imidization could also hinder
solvent evaporation as it becomes more difficult for solvent to diffuse through a highly
imidized and oriented film. Intermediate holds (Table 3 L) help alleviate this problem by
promoting solvent evaporation at a lower temperature before complete imidization
occurs. However, longer hold times at intermediate temperatures could potentially limit
the imidization reaction due to the loss o f solvent-induced plasticization, which otherwise
promotes imidization [110].
Figure 25 shows the mechanical properties o f VFM cured films as a function o f
processing conditions. As seen in Figure 25, the Young's modulus o f the VFM cured
films is high and comparable to standard thermally cured films (-8.66 GPa) indicating a
high degree o f imidization and orientation (as confirmed from Figure 22 and Table 5).
125
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The Young’s modulus is nearly independent o f ramp rate except for very high ramp rates.
This is consistent with the birefringence and TGA data presented earlier. The slight
reduction in chain orientation with ramp rate is compensated by the possible increase in
crystalline order in micro domains at very high ramp rates [106], which could also
increase the modulus o f cured films. The Young’s modulus o f the two-step VFM cure (L
o f Table 3) showed the same Young’s modulus as the standard thermally cured film.
Residual stress built in thin polymer films cured over silicon is due to a) the extrinsic
or thermal stress and b) the intrinsic stress o f the polymer film. The thermal stress in turn
depends on the in-plane CTE mismatch between the silicon substrate and the polymeric
film and the temperature excursion (cooling below the cure temperature or Tg). The
intrinsic stress in the polymeric film occurs due to volume shrinkage, which is in turn
dependent on the structure and properties o f the polymer [111].
The processing
conditions used also significantly affect the evolution o f the intrinsic stress in the films
[104].
Figure 25 shows that the residual stress follows the same trend as the Young’s
modulus.
Studies have shown that an increase in molecular orientation results in a
decrease in the in-plane CTE o f cured films [111].
With increasing ramp rate the
birefringence o f VFM cured films decreased slightly (Figure 24). Since TGA results
showed negligible differences in solvent content with ramp rate, the increase in residual
stress at very high ramp rates could be attributed to an increase in the modulus o f the
film. Indeed, previous studies have shown that heating a film rapidly above its Tg results
in an increase in stress [106]. The residual stress for all the VFM cure conditions studied
126
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
is still lower than that for rapid hotplate cured films, which showed a residual stress as
high as 35 MPa for PI 2611 [104]. This may be attributed to a) a lower thermal stress
associated with a low temperature cure, b) a higher degree o f orientation in VFM cured
films than the rapid hotplate cured films which leads to a lower in-plane CTE o f the film
and c) differences in crystalline order. Furthermore, as discussed earlier, microwave
processed films could have an effective Tg different from a rapid thermal cured film
during the cure process due to the existence o f high local temperatures, which promote
rapid imidization as well as solvent evaporation. The further decrease in the residual
stress with the introduction o f intermediate temperature hold for the two-step process is
also due to increased molecular orientation (from Table 5).
Significant improvement in the thermal performance in a much shorter cure time at
low temperatures (275°C) over thermal curing can be achieved by VFM curing (Table 5
and Figure 26). In the VFM cure process, highly effective coupling o f microwaves by
localized absorption o f the highly polar NMP solvent and the relatively flexible and polar
polyamic acid groups promotes rapid imidization during the initial curing stages and at
higher temperatures energy absorption any residual solvent promotes evaporation from
the imidized film. The localized absorption leads to higher local temperatures and more
effective removal o f solvent as compared to a low temperature cure process where the
entire film is at the same bulk temperature.
HD 4000 and XP 7001 are negative tone photosensitive polyimides based on
acrylate photochemistry [112]. Typical negative tone photosensitive polymers undergo a
cross linking reaction upon exposure to UV-light, which makes the exposed regions
127
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
insoluble in developer solution. During the cure reaction these crosslinks break leading
to imide ring formation and subsequently leave the polyimide film. The photoactive
species may be included as part o f the backbone or as a monomer additive [14]. As
shown in the DSC scan in Figure 28, both negative tone photosensitive polyimides
showed an endothermic peak at high temperature not seen in the non-photosensitive PI
2611. The acrylate monomers, which act as the crosslinking agents for pattern definition,
are known to polymerize both photolytically and during the cure reaction [113, 114]. The
monomer by itself and the by-product o f this polymerization reaction tend to be high
molecular weight species making it difficult to remove from the polyimide film. These
high molecular weight species evolve at higher temperatures and are shown as the high
temperature endotherm in the DSC scan.
Several studies have shown a correlation between backbone flexibility and in-plane
chain orientation [99]. Rigid rod backbones tend to show greater in-plane orientation.
The birefringence results from Table 6 and Table 7 are consistent with this trend in that
both the negative tone polyimides show markedly lower birefringence as compared to PI
2611. Processing conditions did not show any consistent trends in the optical properties
o f HD4000 and XP7001 films.
From Table 6 and Table 7, it can be seen that HD 4000 shows a lower T$% than XP
7001 for any given cure condition. This can be explained from a chemistry and structure
perspective.
Both the negative tone polyimides achieve complete imidization at
temperatures over 22S°C. When the imidization reaction is complete and the solvent
evolves from the film during the cure reaction, the glass transition temperature (Tg) o f the
128
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
polyimide is high causing reduced chain mobility, which makes solvent or photoproduct
evolution difficult (in the glassy state). Any trapped solvent or photoproduct can act as a
plasticizer and also reduce the T$%. Under such conditions, the rigidity o f the polymer
backbone plays an important role for evaporation o f the relatively high molecular weight
photoproducts. A relatively flexible XP 7001 would have greater chain mobility and
enable the photoproduct to be removed more efficiently than the semi-rigid HD 4000.
Further, the Tg, is dependent on the structure and molecular weight o f the polyimide. It
has been reported that the standard thermal cured XP 7001 has a lower glass transition
temperature than HD 4000 [105]. In general, curing above the Tg facilitates photoproduct
removal as the increased chain mobility in the rubbery state eases the removal o f
photoproducts and raises the Ts%. Indeed, the Tg o f XP 7001 films cured in the VFM at
275°C was lower than that of HD 4000 films VFM cured at the same temperature [105].
Hence, the greater backbone flexibility and lower Tg o f XP 7001 films explain the higher
T$% for XP 7001 films. Furthermore, it is also reported [105] that XP 7001 has a lower
molecular weight (lower boiling point) acrylate monomer (which also gives a lower
molecular weight polymerized species) as compared to HD 4000 making it easier to
remove the photoproducts, particularly so for VFM cured films. This was also verified
by mass spectroscopy studies (Appendix A). The T5% o f low-temperature VFM cured
HD 4000 and XP 7001 are greater than the corresponding low temperature thermally
cured samples due to the high local temperature (as discussed earlier) which enhances the
mobility as well as diffusion and evolution o f the photoproducts, especially the lower
molecular weight species.
129
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The Young’s modulus o f VFM cured HD 4000 increases with cure temperature until
~ 275°C (Table 8). This could be due to a) decrease in the amount o f residual solvent and
photoproducts with increasing cure temperature or 2) increase in the polymerization o f
the acrylate monomers which could act as physical crosslinks. The Ts% o f a 5 min VFM
cure at 27S°C (Table 6 D) is lower than that o f a low temperature one-hour cure HD 4000
film (Table 6 B). However, the modulus o f the S min VFM cured film is higher (Table 8)
suggesting that the polymerization o f the acrylate monomer and the resulting physical
crosslinking is the more dominant effect. It is possible that the thermal polymerization
reaction o f the acrylate monomer (crosslinker) and/or its UV-reaction products is
enhanced due to the high local temperature effect for VFM cured samples. However, the
modulus o f XP 7001 does not change significantly with cure temperature.
This is
expected because as discussed earlier, the lower molecular weight acrylate monomer,
backbone flexibility and low Tg o f XP 7001 make it easier to remove the monomer and
related photoproducts from the film. The small concentrations o f any residual products
(lower molecular weight as compared to HD 4000) are not high enough to cause a
significant change in the modulus o f the predominantly flexible polyimide.
The increase in cure time at 275°C shows a reduction in the modulus o f HD 4000
films (Table 8). This is because long cure times at high temperatures, (275°C or higher)
could disintegrate/break the polymerized acrylate species, which subsequently leave the
polymer film resulting in reduced crosslinking effect. Indeed mass spectroscopy studies
showed the evolution o f photoproduct related species at temperatures above 275°C.
Furthermore, an increase in 7jx o f HD 4000 with cure time as seen from Table 6
130
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(D->E->F) also validates this observation. No significant change in modulus occurs
with cure time at 275°C for XP 7001 as the acrylate polymerization is not as significant in
XP 7001 as discussed earlier and evolution o f such species just increases the Ts%.
The residual stress o f VFM cured photosensitive polyimides is higher than that of
rigid-rod PI 2611. As discussed earlier, the low residual stress in PI 2611 is attributed to
the high degree o f in-plane orientation (as evidenced by the birefringence results from
Table S) achieved during the cure reaction. The in-plane orientation of both HD 4000
and XP 7001 is much lower than PI 2611 (from birefringence values o f Table 6and Table
7). Hence these films show a higher residual stress. Further, as the cure temperature
increases, the extrinsic stress increases due to increased CTE mismatch between the film
and the substrate and the intrinsic stress also increases as the residual solvent and
photoproducts leave the film (shrinkage). This explains the increase in residual stress o f
both HD 4000 and XP 7001 with increasing cure temperature.
The residual stress in low-temperature (27S°C) VFM cured HD 4000 is significantly
higher than XP 7001 for a S min cure (Table 9). This could possibly be due to the higher
modulus (as a result o f increased polymerization as discussed earlier) o f the short cure
time HD 4000 films. Increasing the cure time increases the residual stress for both HD
4000 and XP 7001. This may be attributed to increased intrinsic stress as a result of
increased solvent or photoproducts evolution, which is evident with the increase in Ts%
with cure time for both polyimides (from Table 6 and Table 7). Also, in the case o f HD
4000, the increase in residual stress is not as significant because the increase in intrinsic
stress is countered by the reduction in modulus with cure time (Table 8).
131
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The
introduction o f an intermediate hold reduced the residual stress o f both HD 4000 and XP
7001. Increased solvent evaporation at the intermediate hold temperature before reaching
the final cure temperature is the likely cause for reduced residual stress as it affects the
effective glass transition temperature and molecular mobility during the cure process.
The residual stress with the intermediate hold is however lower than that o f the standard
thermally cured films (A of Table 6 and Table 7), which could be a result o f both a lower
modulus and/or higher residual volatiles (photoproducts) which serve to plasticize the
cured films. Further, the lower temperature excursion associated with a low-temperature
cure also reduces the extrinsic stress.
VFM curing o f positive tone dielectrics was studied using polybenzoxazole
chemistry based CRC 8650 and polyimide chemistry based Photoneece™ systems. The
photosensitivity in all three materials is based on DNQ chemistry wherein the addition of
10 to 20% by weight o f DNQ inhibits the base solubility o f the unexposed resin while
drastically increasing the solubility o f the exposed regions.
The DNQ may be
incorporated as an additive or a part o f the main chain. The actual chemistry o f these
materials is proprietary but typical DNQ based systems typically consist o f 3-4 DNQ
groups as part o f a multifunctional phenol or benzophenone type photoactive compounds
[115]. Depending on the actual structure, molecular weight, and thermal stability o f the
photo package the cure conditions are typically set to minimize any residual volatiles.
FTIR and DSC studies o f CRC 8650 showed (to be discussed in Chapter VI) that the
ring closure by condensation as well as photoproduct evolution occurred simultaneously
unlike the negative tone polyimides where practically complete imidization occurred well
132
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
before significant photoproducts evolution. In the case o f CRC 8650, (unlike the HDMS
polyimides), significant ring closure or cyclization reaction occurs at temperatures greater
than 275°C thermally.
Consequently, higher temperatures and longer cure times are
essential for achieving high percent reaction as compared to the negative tone
polyimides. Further, the thermal stability (as characterized by TGA) o f films completely
reacted was higher than that o f films partially cured, suggesting that the crosslinking of
photoactive species to the backbone may also be inhibiting ring closure. A three-step low
temperature VFM cure with intermediate holds at 150°C and 250°C improved the thermal
stability o f the cured film well above that o f a one-hour cure at 275°C in a conventional
thermal oven.
Moreover, short cure times at higher temperatures gave comparable
thermal properties to standard thermal cured films. Low temperature VFM curing o f
CRC 8650 showed a greater improvement in thermal stability over a low temperature
thermally cured film as compared to the negative tone polyimides.
This may be
attributed to a) the polybenzoxazole reaction chemistry favoring effective removal of
photoproducts by enhanced microwave heating (high temperature ring closure), and b)
differences in the effectiveness o f microwave coupling between the DNQ based photo
package in CRC 8650 as compared to the acrylate based monomer and/or the
polymerized photo package in the negative tone polyimides.
Furthermore, the low
temperature thermally cured film is only 78% cured, which causes additional weight loss
in the TGA due to the evolution o f condensation byproducts. This is evident by the much
lower Tik in the low temperature thermally cured film.
133
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Similar results were seen in the case o f the PWDC 1000 and PW 1200 where the
films cured at lower cure temperatures (VFM and thermal) showed both incomplete
reaction as well as low thermal stability.
Imidization begins well before the
photoproducts begin to evolve (from FTIR and DSC results). However, as mentioned
earlier, complete imidization occurs only at temperatures above 275°C for PWDC 1000
and 30 minutes at 275°C for PW 1200.
Low temperature VFM curing showed no
improvement over the corresponding thermal cures.
Removal o f photoproducts and
imidization in these materials is much harder to achieve as compared to CRC 86S0. As
with CRC 8650, crosslinking o f the photoproducts to the backbone could hinder the
imidization reaction. MS studies also indicate that PW 1200 has a higher molecular
weight PAC as compared to CRC 8650 (Appendix A). Furthermore, VFM enhanced
crosslinking reaction o f the additional crosslinkers added to the system could also prevent
effective photoproduct removal from the cured films. Increased time at intermediate hold
temperatures however seemed to reduce the % imidization possibly due to reduced
mobility at the final cure temperature.
These results suggest that certain polyimide systems or chemistries may be better
suited for low temperature VFM processing than others.
In the case o f the HDMS
polyimides, the solvent used in all the three formulations, n-methyl pyrolidone (NMP),
complexes with the precursors (acid or ester or amide groups) and one o f the first steps of
the cure reaction involves decomplexation o f the solvent Decomplexation o f the solvent
is essential as the liberated solvent helps plasticize the high molecular weight polyamic
acid (or ester) and promotes the imidization reaction. However, too rapid an imidization
134
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
hinders solvent evolution from the film during the subsequent stages and can trap the
solvent (and/or photoproducts) within the film, which results in a low 7s%. In the case of
microwave processing, the existence o f the high local temperature can change the
kinetics o f solvent decomplexation and imidization. Furthermore, since heating occurs
primarily due to dielectric loss (the solvent being is highly lossy), one would expect the
solvent to be preferentially removed from the film by VFM curing, which could result in
higher Ts%temperatures. This is evident from the results on low temperature VFM curing
o f PI 2611 which showed that significant improvement in the thermal properties over
conventional low temperature thermal cured films can be attained by VFM processing
films at a lower cure temperature.
However, the same is not true with the other polymer systems studied. With the
negative tone polyimides, HD 4000 and XP 7001, it is easy to remove the solvent but the
photo package removal is inherently more difficult because o f the higher molecular
weight giving a lower vapor pressure. In general, removal o f the photo package requires
greater thermal exposure. Moreover, photo package can also polymerize with itself upon
UV exposure and during the cure step. It is these high molecular weight species that are
more difficult to remove from the film. Backbone flexibility facilitates solvent and
photoproduct evolution especially for VFM processing where greater side group and
backbone mobility assist in increased dissipation o f electromagnetic energy as heat [22].
However, it is also likely that VFM processing could enhance the propensity for self­
polymerization in the residual products.
Alternatively, the already photolytically
polymerized species could make it increasingly difficult to remove such high molecular
135
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
weight species especially at lower cure temperatures. The mechanical properties o f VFM
cured films, which showed increased modulus for low-temperature, short time VFM
curing (also verified by Itabashi et al. [105]), are also consistent with this argument.
Indeed, MS results also show the evolution o f high molecular weight species even at
temperatures o f 320°C. This suggests that films cured below these temperatures exhibit
the high molecular weight moieties that remain trapped inside the film. This explains the
lower Tra for the negative tone photosensitive polyimides.
Even though the photo
package by itself absorbs energy locally, the thermal and/or microwave induced
polymerization necessitates that the cure temperature be greater than 275°C for effective
removal o f photo products. Indeed, at temperatures greater than 275°C, VFM curing
gives the same thermal stability as the standard thermally cured films in a shorter cure
time.
The effectiveness o f removal o f the DNQ bearing photoactive compounds (and the
associated photoproducts) depends on the decomposition temperature o f the DNQ itself
(and the products) and the chemical and physical structure o f the polymer matrix. The
thermal reactions o f DNQs have been studied, which include crosslinking o f the DNQ
species to the backbone o f the base resin forming ester-type species [116]. In the case of
CRC 8650, one possible crosslinking reaction (between the DNQ moiety and the
backbone benzoxazole resin) during the cure reaction could be between the ketene
(formed by high temperature baking o f DNQ) and the hydroxyl group o f the polyhydroxyamide precursor. This reaction could be detrimental to the cure process on two
accounts: a) crosslinking to the backbone makes it even more harder for the removal o f
136
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the photoproducts and b) crosslinking also prevents the ring-closing reaction (i.e.
formation o f the benzoxazole ring).
Thus, ring closure and crosslinking could be
competing reactions during the cure process. The results from Table 11 show that it is
much harder to remove the DNQ and the resulting photoproducts in PWDC 1000 and PW
1200 as compared to CRC 8650. Complete imidization o f PWDC 1000 and PW 1200 is
also harder to achieve as compared to non-photosensitive PI 2611 and the negative tone
polyimides. In the case o f these materials, the situation is compounded because they
have about 5 to 10% by weight o f additional crosslinking agents (as mentioned earlier).
VFM processing could favor rapid crosslinking o f these groups, which prevents effective
removal o f the photo compounds and results in cured films with trapped photoproducts,
which reduces the Ts% value. Higher processing temperatures are required to completely
remove such trapped species and indeed films cured at higher temperatures, begin to
show improved Ts% even compared to conventional high temperature furnace processed
films.
5.3 Summary
As the processing requirements on materials become more stringent owing to
reduced thermal budgets in advanced electronic devices, novel low temperature
processing methods need to be explored which enable the use o f traditional materials in
future generation devices. Reducing the cure temperature o f polyimide coatings for use
as passivation layers on integrated circuits is o f particular interest Low temperature
VFM curing o f different commercially available polyimides was investigated.
137
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The
objective o f this study was to investigate the feasibility o f using VFM curing as a rapid
low-temperature curing technique to obtain properties comparable to traditional high
temperature thermal cures. The targeted cure temperature was 275°C. The structureproperty relationships and their dependence on the processing conditions were studied.
The thermal stability o f cured films was evaluated as a performance metric to determine
the effectiveness o f low-temperature microwave processing.
The results show that
certain chemistries are more suited than others for low-temperature microwave
processing. Non-photosensitive rigid rod backbone based PI 2611 showed significant
improvement in the thermal stability over low temperature thermally cured films and
properties comparable to standard high temperature cured films.
The negative tone
polyimides did not show a marked improvement in performance over low-temperature
thermal curing possibly due to microwave enhanced polymerization o f the photopackage.
In the case o f positive tone dielectrics, CRC 8650 showed slightly better thermal
performance than low-temperature thermally cured samples for long cure times at 275°C
while PWDC 1000 and PW 1200 did not show any significant improvement over lowtemperature thermal cures. The results from this study identified some o f the key VFM
processing issues and how they affect the cured film properties. Structure-property
relations and some general trends on the impact o f processing conditions on evolution of
mechanical and thermal properties were discussed. The results indicate that the backbone
flexibility, structure and actual photochemistry o f the photopackages, and the competition
between imidization and potential crosslinking reactions are important factors
determining the effectiveness o f low temperature microwave processing.
138
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER VI
RAPID CURING OF POSITIVE TONE PHOTOSENSITIVE POLYBENZOXAZOLE
BASED DIELECTRIC RESIN BY VARIABLE FREQUENCY MICROWAVE
PROCESSING
This chapter contains a detailed discussion on rapid VFM curing o f CRC 8650, a
positive tone polybenzoxazole based dielectric resin. The feasibility o f low temperature
curing o f this resin on silicon was also studied. The chemical changes occurring during
the cure reaction were monitored using Fourier transform infrared spectroscopy. The
percent ring closure achieved for different processing conditions is estimated from FTIR
analysis. The effect o f rapid low-temperature VFM curing on the optical, electrical, and
thermo-mechanical properties o f cured films was also investigated. The thermal stability
o f cured films was studied by thermo gravimetric analysis (TGA) and mass spectrometry
(MS).
Traditionally, negative tone polyimides with organic solvent-based developers are
used for semiconductor passivation and redistribution applications due to their high
thermal stability and excellent mechanical properties [12]. Positive tone photosensitive
polymers however offer distinct processing and performance advantages over negative
tone systems [13]. There is also interest in moving towards aqueous base developer
based systems due to environmental concerns. Polybenzoxazoles have similar thermal
139
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
stability and mechanical strength as polyimides and offer improved dielectric
performance and chemical resistance [8].
In order to achieve the desired properties, materials such as polyimides and
polybenzoxazoles require a curing step, which typically involves a high temperature
treatment (300 to 400°C) for long periods o f time lasting several hours. Further, low CTE
dielectrics and low temperature processing techniques are essential to minimize the stress
arising due to the coefficient o f thermal expansion (CTE) mismatch between the silicon
die and the packaging compounds. In this study, rapid VFM curing o f a positive tone
photosensitive polybenzoxazole chemistry based dielectric resin, Sumiresin Excel CRC
8650 from Sumitomo Bakelite Co., was investigated.
6.1 Polvbenzoxazole-fSumiresin Excel CRC 8650)
The actual chemistry o f CRC 8650 used in this study is proprietary and not known.
The as-received resin comes as 25 to 40% by weight solids with y-butyrolactone (GBL)
as the solvent [117].
Typically, the polybenzoxazole cure reaction involves a ring-
closing condensation (also referred to as cyclization or cyclo-dehydration) o f a polyhydroxy-amide (PHA) precursor obtained from polycondensation o f bis-o-aminophenols
and a dicarboxylic acid chloride or anhydride [13, 118, 119, 120]. The cure reaction and
the photoreaction in PBO are schematically represented in Figure 30. The photoreaction
in PBO is based on diazonaphthoquinone (DNQ) chemistry. The DNQ moiety is usually
included as part o f the main chain or as part o f a photoactive compound (PAC) typically a
multi-functional phenol or a benzophenone [115].
The DNQ acts as a dissolution
140
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
inhibitor o f the base resin in developer solution in the unexposed regions. Upon UV
exposure the DNQ undergoes W olff rearrangement forming indene carboxylic acid,
which causes the base solubility (Figure 30).
The manufacturer recommended cure
schedule for CRC 8650 (standard thermal cure) is a 30 min hold at 150°C followed by an
additional 30 min hold at 320°C.
The effect o f VFM processing conditions on the
properties o f PBO films was characterized and compared to thermally cured films.
141
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
dicarboxylic acid/anhydride
bis-o-aminophenol
\
H
H O
^•Q C C K X
OC
poly-hydroxy-amide (PHA)
Polybenzoxazole (PBO)
(a)
COOH
hv
(b)
Figure 30 Schematic represeiitation o f a) cure reaction o f polybenzoxazole (PBO)
formation and b) DNQ reaction-Wolff rearrangement.
142
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6.2 Results
Figure 31 shows a differential scanning calorimetry (DSC) scan o f a PBO sample
under a constant ramp rate o f 3°C/min to 400°C in a nitrogen atmosphere. As seen in
Figure 31, two prominent endothermic peaks are observed during the cure cycle. The
initial peak starting at ~ 150°C corresponds to the evolution o f solvent and photoproducts
from the DNQ based PAC. The broad endotherm at higher temperatures is due to solvent
evaporation and evolution o f residual higher molecular weight photoproducts. The peak
at ~ 27S°C is associated with an increase in the benzoxazole ring closing reaction, which
is completed at a temperature o f ~ 320°C. The DSC scan shows that a temperature
greater than 275°C is essential to achieve significant ring-closing reaction in PBO films
cured thermally.
These results are typical o f DNQ based PACs [121] and
polybenzoxazoles [120,122,123,124].
143
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
400
350
-500
300
-1000
s
S
250
-1500
200
-2000
150
100
-2500
50
-3000
Tim e, m in
Figure 31 DSC scan o f a PBO sample under a constant ramp rate o f 3°C/min.
The chemical changes occurring in the PBO films during the cure process were
monitored using FT1R analysis. As mentioned earlier, the actual chemistry o f Sumiresin
Excel CRC 8650 is proprietary.
However, the formation o f the benzoxazole ring-
structure from the poly-hydroxy-amide (PHA) precursor can be followed by infrared
spectroscopy. The infrared spectra o f a) softbaked (unexposed), b) standard thermally
cured and c) VFM cured PBO films are shown in Figure 32. The changes in the Fl'lR
spectrum can be correlated to the chemical or structural changes occurring within the
144
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
film. Table 12 summarizes some o f the peak assignments [119, 120, 122, 123, 125, 126,
127,128, 129, 130,131]. As seen in Figure 32, comparing the spectra (a) and (b) several
distinct changes occur during the curing process.
The softbaked film shows the
characteristic absorption peaks o f the poly-hydroxy-amide precursor which include: a
broad absorption in the range 3000 to 3500 cm'1 corresponding to the -N -H and -O-H
stretch, a carbonyl ( - 0 0 ) stretch at 1650 cm'1and an -N-H bending mode vibration at
wavenumber 1527 cm*1.
As seen from spectrum (b) o f Figure 32, all of these
absorbencies disappear completely after the standard thermal cure.
The formation of the benzoxazole ring during the cure reaction is characterized by
the appearance o f a distinct peak at wavenumber 1054 cm'1 representative o f the -C-O-C
stretch o f the benzoxazole ring in the standard thermally cured film. This peak is absent
in the softbaked film and increases (Figure 33) with the cyclization reaction by
condensation o f water as shown in Figure 30. The benzoxazole ring also shows an
absorbance at 1479 cm'1 and 1616 cm*1 (-C=N- stretch). Both these peaks however
interfere with the -C=C absorption bands o f the aromatic ring at 1500 and 1600 cm*1
respectively. Further, from Figure 32, (comparing (b) and (c)) the FTIR spectra o f the
standard thermal cured film and the VFM cured film are essentially identical with no
significant differences in the range 400 to 4000 cm*1. This shows that the chemical
structure o f VFM cured PBO is similar to that o f the standard thermal cured film.
145
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1054
1600
Absorbance
963
1650
2000
1527
1500
1000
Wavenurabers (cm*1)
(A)
<
4000
3000
Wavenumbers (cm*1)
2000
(B)
Figure 32 FTIR spectra o f a) softbaked b) standard thermal cured and c) VFM cured
films: A) 900-2000 cm '1 and B) 2000-4000 cm'1.
146
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 12 Peak assignments o f PBO and the corresponding changes occurring during the
cure reaction.
W av enu m bers P ro b a b le P eak A ssignm ent
C h an g es w ith c u re
cm*1
3000-3500
-N-H stree thing
3000-3400
-O-H stretching
1650
amide -0 = 0 stretching
1616
benzoxazole ring
-C=Nstretching
1479
benzoxazole ring
1054
963
-C-N-
benzoxazole ring -C-O-C
stretching
deformation o f aromatic -C-H
-C-F bending
1527
-N-H bending
1600
-C=C- from aromatic group
( |) Disappears completely
after cure
(j) Disappears completely
after cure
(i) Disappears completely
after cure
( t ) Increases with cure:
Indicative of cyclization
(t) Increases with cure:
Indicative of cyclization
(f) Increases with cure:
Indicative of cyclization
Remains unchanged during
curing: Internal standard
(4) Disappears completely
after cure
Internal standard
147
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The percent conversion (cyclization or percent ring closure) achieved in cured PBO
films was estimated by FTIR analysis using Equation 6.1.
% Conversion =
(a
1054 / A bucnul Standard JSample
I
^
jq q
(6.1)
Internal Standard / s j j . Thermal
The absorbance at 1054 cm*1 was considered representative o f the benzoxazole ring
formation.
The peak height corresponding to the standard thermal cured film was
considered fully cured (100 % conversion). The peak at 1054 cm*1was normalized by an
internal standard.
The absorbencies at 1600 cm*1 and 963 cm*1 (-C-H deformation/
bending o f the aromatic ring/ -C-F bending) were relatively unaffected by the cure
process (remained constant) and were used as the internal standard. All other peaks were
either affected by the cure reaction or interfered with the nearby peaks for e.g., with the
broad -C-F stretching peaks between 1000-1400 cm '1.
The estimates o f percent
conversion in the cured films based on either internal standard were in good agreement
and within 1 % o f each other (as discussed below).
Figure 33 shows the FTIR spectra o f PBO films cured in a conventional thermal
oven in the temperature range between 225 and 350°C. Each film was ramped at 3°C/min
and held at that temperature for one hour. As seen in Figure 33, the absorbance at
wavenumber 1054 cm '1 increases with the cure temperature indicating increasing ring
148
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
closure while the absorbance at 963 cm '1 remains constant.
The results from the
estimation of percent conversion (from Equation 6.1) in the thermally cured films are
shown in Figure 34. As seen in Figure 34, negligible reaction occurs at temperatures less
than 275°C, and complete conversion is possible only at a temperature o f 320°C or
greater. These results are consistent with the DSC results discussed earlier, which show
that significant reaction occurs only at temperatures greater than 275°C. Furthermore, as
seen in Figure 34, the percent conversion o f PBO estimated using Equation 6.1 using
both 963 cm'1 and 1600 cm'1 as the internal standard was in good agreement and within 1
% o f each other (within the sensitivity o f the IR spectrometer).
FTIR analysis thus
provided a semi-quantitative estimate o f the percent conversion (effectiveness o f the
curing) in the PBO films as a function o f cure conditions.
149
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1054
8
e01
JO
hm
8
Increasing
Temperature
.o
963
<
1000
1100
Wavenumbers (cm-1)
Figure 33 FTIR spectra o f thermally cured PBO films showing an increasing absorbance
at 1054 cm '1with cure temperature.
150
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
120
a 100
e
y»
oj
>
s
e
80
U
nP
60
o00
40
0s
eu
■
1054 /1600
- 0 - 1054/963
20
175
200
225
250
275
300
325
350
375
Temperature, °C
Figure 34 Percent conversion achieved in thermally cured PBO films as a function of
cure temperature estimated from FTLK. analysis with peak height at 1054 cm'1 and a)
1600 cm'1 and b) 963 cm '1 as an internal standard.
VFM curing o f PBO under various processing conditions was performed to
investigate the feasibility o f rapid and low-temperature curing. The percent conversion
achieved under different VFM cure conditions from FTIR results is shown in Figure 35
and compared with the thermally cured samples at the corresponding conditions.
151
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
120
s
e
100
G0
u
©
>
e
©
U
N®
ofifi
fiu
175
Temperature, °C
♦
Thermal, 3°C/min 1 hr
■ VFM, 15°C/min 5 min
□ Thermal, 3°C/min 30 min
• VFM, 15°C/min 15min
Thermal, 15°C/min 1 hr
a
VFM, 15°C/min 30 min
A VFM, 3 step 30 min
Figure 35 Comparison o f percent conversion achieved in VFM and thermally cured PBO
films from FTIR analysis. (VFM 3-Step: 30"C to 150°C 5min, 30°C to 250"C 10 min,
10"C/min to 275°C 30min).
152
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Several key observations can be made from Figure 35. The solid squares in Figure
35 represent VFM cured PBO films each ramped at 15°C/min to a given temperature and
held at that temperature for 5 min. As seen in Figure 35, similar to the thermally cured
films, the percent conversion achieved in VFM cured films increases with increasing cure
temperature.
Complete conversion is achieved even after a 5 min VFM cure at a
temperature o f 300°C or higher. Increasing the cure time at lower temperatures (250°C
and 275°C) for the VFM cured samples shows a significant increase in the percent
conversion achieved (solid circles and solid triangles o f Figure 35). Also, a much higher
percent conversion can be achieved by a 30 min VFM cure at 250°C as compared to a
slow ramp 1 hr long cure in the thermal oven at the same cure temperature. Also, as
expected increasing the ramp rate or decreasing the cure time at 275°C in the thermal
oven reduces the percent conversion. A three-step VFM cure (30°C/min to 150°C 5min,
30°C/min to 250°C 10 min, 10°C/min to 275°C 30min) gives a significantly higher
percent conversion than a one-hour thermal cure at 275°C. These results are consistent
with other studies, which have shown acceleration in cure kinetics by VFM processing
[22, 79] and demonstrate the effectiveness o f VFM curing as a rapid curing alternative to
conventional thermal curing.
The photoreaction in the PBO films was also monitored using FTIR and UV-visible
spectroscopy. Figure 36 (a) and Figure 36 (b) show the FTIR and UV-visible spectra
respectively o f PBO precuror (PHA) films before and after exposure. As discussed
earlier, PBO is based on a DNQ photochemistry. The unexposed PBO film shows a
distinct IR absorbance in the region 2000 to 2220 cm*1; 2118 cm*1 attributed to the -C-N
153
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
stretch [115, 132, 133] and 2164 cm'l (-N=N stretch) [133] o f the diazo group o f the
DNQ moiety. This absorbance is seen in the softbaked and unexposed film (Figure 36)
and disappears when fully exposed to UV-radiation. The UV-visible spectra o f PBO
films as a function o f increasing UV exposure dose (g-line) are also shown in Figure 36.
As seen in Figure 36 (b), the characteristic UV absorbance band in the wavelength range
350 to 450 nm decreases with increasing dose. Further, the fully bleached film (~ 1150
mJ/cm2 for a 25 pm thick film) has a negligible absorbance in range 350 to 450 nm,
which enables patterning o f relatively thick films. The PAC used in PBO is also heat
sensitive. Even without any UV exposure the fully cured films did not show an IR
absorbance in the range 2000 to 2220 cm'1 indicating the disappearance o f the -N=N
moiety.
Increasing bake temperatures reduced the IR and UV absorbance o f the
photosensitive species.
The FTIR absorbance o f the -N=N moiety disappeared
completely after a 10 min bake at 150°C. High temperature bakes are known to have
similar reaction mechanisms as those induced photolytically [134].
154
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
g-line UV Exposed
2200
2000
Wavenumbers (cm 1)
1800
(a)
5.0
4.5
i Increasing
i exposure dose
4.0
3.5
it
it
3.0
2.5
u
eCfl
A 2.0
< 1.5
1.0
0.5
0.0
300
400
350
450
500
Wavelength, nm
(b)
Figure 36 Photoreactioii o f PBO precursor films: a) FTIR spectra b) UV-visible spectra.
155
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
To investigate the effectiveness o f rapid VFM curing, the properties o f PBO films
cured at 320°C and 275°C were characterized and compared to thermally cured films. The
cure conditions used and the properties measured are summarized in Table 13.
As seen from Table 13, PBO films cured under all four conditions show a low
birefringence indicating that the films are essentially isotropic in nature.
Further, no
significant differences are observed in the index o f refraction and birefringence o f films
cured under the four conditions.
However, significant differences in the electrical
properties between the standard thermal cured films (A) and low temperature cured films
(B and D) can be observed. The standard thermally cured film has a dielectric constant
o f 2.97 measured at 10 kHz, whereas the low temperature thermal and VFM cured films
show a dielectric constant o f 3.24 and 3.29 respectively. Similarly, the loss tangents o f
the low temperature thermal and VFM cured films measured at 10 kHz are significantly
higher than the standard thermal cured films. Residual solvent and photoproducts are
believed to be responsible for the increased dielectric constant and dielectric loss in the
low temperature cured films.
The thermo-mechanical properties o f cured films are also shown in Table 13. As
explained in the experimental section, residual stress in cured films was estimated from
wafer curvature measurements and the Young’s modulus was determined from Instron
pull tests. As seen in Table 13, the residual stress in a standard thermal cured film was ~
32 MPa. No significant difference in the Young’s modulus o f films was observed under
the four cure conditions studied. The CTE and Tg o f cured films were determined by
TMA measurements. Figure 37 shows a typical TMA plot o f a cured PBO film (C)
1S6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ramped at 5°C/min in the TMA. As seen from Table 13, the Tg o f low temperature cured
films was lower than the Tg for standard thermal cured films (condition A o f Table 13).
Further, the CTE o f the low temperature thermally cured film (condition B o f Table 13) is
higher than the standard thermally cured film.
3500
3000
i*
v
BA
e
9
JB
u
s
e
M
e
v
S
O
2500
2000
1500
1000
CTE = 57.96 ppm^C
500
200. 10*
50.42*C
0
50
100
150
200
250
300
350
400
Temperature, °C
Figure 37 TMA plot o f VFM cured PBO film cured at 320°C for 5 min. TMA ramp rate:
5°C/min. Load: 0.05 N.
157
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Thermo gravimetric analysis (TGA) was performed on VFM and thermally cured
PBO films to evaluate the effect o f rapid low-temperature VFM curing on the thermal
stability o f cured films. About 5 to 10 mg samples o f the cured films were ramped to a
temperature o f SS0°C at 10°C/min in a nitrogen atmosphere. Figure 38 shows the TGA
scan o f a standard thermal cured PBO film ramped at 10°C/min to 550°C in a nitrogen
atmosphere. As seen in the Figure 38, there is no significant weight loss in the cured film
at temperatures below 300°C. However, at higher temperatures there is a gradual drop in
the mass initially (300 to 4S0°C), followed by and a drastic drop in mass at temperatures
greater than 500“C. As seen from Figure 38 and Table 13, the standard thermal cured
film shows a Ts% o f ~ 433°C. Low-temperature thermal curing significantly (B) lowered
the Ts% o f cured film. The thermal stability o f low-temperature VFM cured films (D) is
however greater than the low-temperature thermally cured films (B).
In order to identify the species responsible for weight loss in the TGA, mass
spectroscopy (MS) studies were conducted on the cured films. Figure 39 shows the total
ion-trace and that o f the primary species detected from a PBO film cured at 250°C. As
seen in Figure 39, at temperatures greater than 250°C, the primary species evolved
correspond to molecular weights 200 and 292 and their fragments, which could be
attributed to the photo-package associated with the PBO film.
1S8
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
i
A: 3*C/min to l50X:30min + 3,C/minlo320,C30min
C: !5*C/min to I50T Smin+ IS'C/min to320T 5 min
B: 3*C/min to l5(fC 30 min + 3*'C7min to 275*C60 min
D: 3(TC/min to 150*C S min + 3(fC/min to 25<TC 10min + KTC/min to27S*C 30min
Table 13 Comparison of properties of VFM and thermally cured films cured at 320°C and 27S°C.
&
159
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
on
s
•s
mi
et
g
90
OC
%
85
*
80
cs
S
50 100 150 200 250 300 350 400 450 500 550
Tem perature, °C
Figure 38 TGA scan o f a standard thermal cured PBO film.
160
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
IONTRACE RagBNpScan Number. Max.Scan=2W5#1:00:21.14.
Total lonCLrrent, M».lrt =10.40648.
100-g
SO
03
4
133 241 298 458 582 699 775 074 974 1006
15^10661765 18661987
S l S
S
f
(
I I V | M I 11 1 I | I I 11 I I I | I I I | I I 1 | I I I 11 I I 11 1 11 I I 11 I I I
aoo
10:00
20:00
lan-200(Nom). Max.N .=1.54432.
3Q00
40.00
100|
I ff * .1173
000 775 866 976 W 4
N N
^
N,
5Q00
1:0Q00
1373
1271
soi
0
1472
11001179 /
X
J56 6 166517831001 1957
111 1111 111 1111 111 11 11 111 11 11 111 JI 11 111 IJ 11 111 11 111 11 11 111 11 11 11111 11111 11 11 111 1111 111 1111 111 1111 III 111V| Tl l,|TnyTTT|T"n"|'l 111
—
000
L
*
10:00
.
20:00
275-C
505
1:0000
5Q00
3000
1270
1567
\y \ A
10441173
\
N^
15® 178418781966
V IZ
/
S
0= ’| III (III |MIfll l|IM |l IIIII f | III |l II |l M| 11 r| m piTI'm |l 11 |l ll| 111| rTi|» m|I11|I 11|I ti 1111| 111111111111*11p 1i-|n I[ »-n | II 11
000
10.00
20.00
3000
4Q00
5000
Retortion Time
Figure 39 Mass spectroscopy results of PBO films ramped at 10°C/min to 450°C.
1:0000
6.3 Discussion
Polybenzoxazole based dielectric resins are known for their exceptional mechanical
properties, high thermal stability and good chemical resistance. Photosensitive dielectrics
offer additional processing advantages. Typical conventional thermal cure methods for
polybenzoxazoles involve high temperature treatments for extended periods o f time. This
results in high thermo-mechanical stress when applied to semiconductor applications.
Microwave processing has been shown to provide significant reduction in the processing
times o f several thin films polymers [22].
Rapid VFM curing o f a positive tone
photosensitive PBO resin was investigated in this study. The effectiveness o f rapid VFM
curing was studied by comparing the properties o f rapid VFM cured films with
conventional thermal cured films.
As discussed earlier, polybenzoxazole formation occurs by cyclo-dehydration (or
condensation) o f a poly-hydroxy amide precursor. As seen in Figure 31, the ring closure
is characterized by a broad reaction endotherm beginning at ~ 220°C in the DSC scan.
The reaction peaks at ~275°C and is completed at ~ 320°C. Unlike polyimides where the
imide-ring closure is completed at ~240°C, the benzoxazole ring formation typically
occurs at much higher temperatures.
Several studies have reported the PBO ring
cyclization reaction temperatures in the range 240 to 350°C [122, 123, 124, 126].
Significant ring closure is possible only at high temperatures or very long times at low
temperatures [119]. Increased chain mobility and high local temperatures resulting from
localized relaxation o f polarization in microwave-processed films could enhance the
162
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
reaction rates at relatively lower temperatures. Indeed, as seen from Figure 35, FTIR
results show significantly higher percent conversion in a shorter cure time by microwave
processing, especially at low processing temperatures (250°C and 275°).
This is
consistent with observations by other researchers that microwave processing
improvements are more pronounced in slower reacting systems and particularly at lower
temperatures [25].
The optical properties o f cured PBO films show a low birefringence indicating that
the cured films do not have any preferential molecular orientation unlike rigid-rod type
backbone polyimides [99].
Accordingly, the properties o f PBO did not show any
significant dependence on ramp rates used during the cure process. The characterization
o f the electrical properties o f cured PBO films showed that low temperature cured films
had a higher dielectric constant and a higher dielectric loss. This may be attributed to
residual solvent and photoproducts.
It is reported that the evolution o f residual
photoproducts o f similar DNQ based compounds continues up to temperatures ~ 350°C
[121]. MS studies also showed the evolution o f high molecular weight species and their
fragments possibly related to the photoproducts o f the DNQ based PAC.
These
observations are consistent with other reports in literature that report increased relative
permittivity and loss tangent for rapid cured photosensitive dielectric films [49].
However, it is important to note that in this study, much slower ramp rates (15°C/min as
compared to 60°C/min), and longer cure times were used, which promote solvent and
photoproduct evolution during the cure reaction. As a result, the dielectric constant and
163
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
loss tangent o f the VFM high temperature cured films (C) are only moderately higher
than the standard thermally cured films (A).
TMA studies show that lowering o f cure temperature also lowered the Tg of the
cured PBO films. This may be attributed to a) incomplete reaction (ring closure) b)
residual solvent or photoproducts which could act as plasticizers and c) differences in the
packing density and associated free volume as a result o f low temperature cure. These
factors also explain the higher CTE o f low temperature thermally cured film (B).
Moreover, as seen from Table 13, the lower temperature thermally cured film is only
78% cured and shows a lower T/% and Ts% than the low-temperature VFM cured film (D)
during the TGA ramp.
No significant difference in the Young's modulus o f PBO films was observed under
the cured conditions studied. As discussed earlier, the low birefringence o f PBO films
suggests that they are isotropic in nature.
Hence, unlike rigid rod polyimides, no
significant orientation effects on modulus or their dependence on process conditions are
expected in the PBO films. Residual stress in thin films coated on silicon occurs due to
1) intrinsic stress built in the films due to volume shrinkage on account o f solvent and
reaction byproducts evolution and 2) thermal or extrinsic stress due to CTE mismatch
between the PBO film and silicon substrate. The residual stress in low temperature
thermally cured film (conditions B Table 13) is lower as compared to the standard
thermal cure (D). This is due to the lower extrinsic stress as a result o f the reduced
thermal excursion (275°C as compared to 320°Q and plasticization due to residual
photoproducts. However, the residual stress in VFM cured films at 320°C and 275°C is
164
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
slightly higher than that o f films cured thermally at the corresponding temperature. The
residual stress in the cured films depends on several factors. Differences in processing
methodology can significantly impact the evolution o f intrinsic stress in the film
depending on the inherent chain rigidity [106, 108]. For example, the rate o f solvent and
photoproduct evolution relative to the rate o f ring-closing reaction could be different in
VFM cured films as compared to thermally cured films. A decrease in the in-plane
orientation (increase in in-plane CTE) due to rapid heating could increase the residual
stress. This is the case with rigid rod type polyimides. However, as discussed earlier,
birefringence measurements did not show significant orientation effects in PBO films.
Rapid heating above Tg also causes differences in film morphology, packing and,
crystalline order and these factors impact the intrinsic stress and hence the residual stress
o f the cured films [106].
As seen from Table 13, the thermal stability of rapid low temperature cured films is
lower than the standard thermal cured films. While almost complete conversion could be
achieved by both rapid (C) and low temperature VFM curing (D), the cured films still
had trapped residuals that affected the thermal stability o f the films. No significant
weight loss was seen below temperatures o f 300°C indicating that negligible amounts o f
solvent (y-butyrolactone) and low molecular weight species were trapped in the cured
films. Weight loss in TGA for the low-temperature thermal cured films (B) could also be
due to the evolution o f ring-closure condensation products (water). This is evident by the
much lower Ti% for the low temperature thermally cured film, which is only 78% cured.
At higher temperatures, the films begin to lose mass and this may be attributed to the
165
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
residual photoproducts. These may result from the initial DNQ based PAC or its reaction
byproducts.
The PAC is usually fimctionalized to improve the lithographic and
processing performance and is typically a higher molecular weight species [115, 118] as
compared to the solvent, making it difficult to remove from the film during the cure
process. Further, reaction byproducts o f the PAC may also be responsible for TGA
weight loss.
The thermal reactions o f DNQ based PACs are well studied [115, 134, 135]. It is
well known that these species are both light and heat sensitive. The PAC typically has
~2-3 DNQ moieties [115, 118], and their reaction with neighboring backbone chains
forming ester-type species (which also affects the percent conversion achieved) could
make the PAC act as a crosslinking agent [135]. These crosslinks are reasonably stable
even up to temperatures o f — 200°C [136] (or higher based on the actual PAC
composition). This could affect the percent conversion achieved as well as the thermal
stability o f the PBO films especially at low cure temperatures. High temperatures and
long cure times are required to break the crosslinks (if any) and evaporate the
photoproducts.
Indeed MS studies show high molecular weight species possibly
associated with the photoproducts evolving at temperatures higher than 250°C (Figure
39). For example, the MS spectra of the parent species with amu 200 and its fragments
compared well to a 2,2 ’-methylenebis-phenol structure (C13H1202: MW-200)
(Appendix A), which is typical o f PACs used in DNQ based systems [115] as discussed
earlier. Low-temperature VFM cured films (D) showed significant improvement in the
thermal stability as compared to low-temperature thermal cured films (B). However, the
166
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
evolution of the photoproducts (high molecular weight species) is eventually diffusion
limited and requires a high bulk temperature and longer cure times.
6.4 Summary and Conclusions
Rapid curing o f a positive tone photosensitive polybenzoxazole dielectric resin was
investigated by VFM processing. The feasibility o f low temperature curing on silicon
was also studied. The chemical changes occurring during the cure reaction and the effect
on processing conditions on the percent conversion achieved in cured films were
monitored by FTIR spectroscopy. FTIR studies show that the chemical structure o f VFM
cured PBO films is identical to standard thermal cured films. These studies also show
that rapid curing o f PBO by VFM processing is feasible and significantly higher
conversion can be achieved by microwave processing at low temperatures as compared to
conventional thermal curing. The effectiveness of rapid VFM curing was studied by
characterizing the optical, electrical and thermo-mechanical properties o f VFM cured
films with thermally cured films. The thermal stability o f cured films was investigated
by TGA and MS studies. The results show that while higher percent conversion and
thermal stability than thermal curing can be achieved by VFM processing at lower
temperatures, complete removal o f photopackage related residual products requires
slower ramp rates and longer cure times.
167
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER VII
RAPID CURING OF POLYIMIDES ON ORGANIC SUBSTRATES BY VARIABLE
FREQUENCY MICROWAVE PROCESSING
In this chapter, results from the studies on rapid VFM curing o f polyimides on low
temperature organic substrates are presented. Four different organic substrates; I) un­
metallized FR4 (FR4), 2) copper laminated FR4 (Cu-FR4), 3) carbon-fiber epoxy
composite (CF-epoxy), and 4) copper laminated BT-epoxy laminate boards (BT) were
studied.
Rapid curing o f polyimide PI 2611 on these substrates was investigated
extensively. A few studies on three other negative tone photosensitive polyimides, PI
2734, a polyamic ester based photosensitive polyimide based on BTDA-ODA chemistry,
HD 4000 and XP 7001 (proprietary chemistry) were also performed. The VFM heating
characteristics and thermal stability o f all the substrates were characterized. FTIR studies
were performed to detect differences in the chemical structure between polyimide films
cured at low temperatures on organic substrates and conventional furnace cured films.
The optical, electrical and thermo-mechanical properties o f the polyimide films cured on
organic substrates were characterized and compared with thermal furnace cured films.
Factors affecting low temperature curing o f polyimides were studied.
168
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The effect o f
solvent on the VFM heating characteristics, properties and thermal stability o f cured
films was also investigated.
The demand for higher speed and greater functionality in low power electronic
devices necessitates the use o f high performance dielectric materials, which allow a
larger number o f I/Os and provide for better thermal management. Polyimides are often
used in a number o f semiconductor and electronic packaging applications due to their
excellent electrical, chemical and thermo-mechanical properties [87]. Photo-definable
polyimides are attractive because they require fewer processing steps [12]. In order to
achieve high mechanical strength and thermal stability, most polyimide systems require
high temperature (~350 to 400°C) processing for several hours [101]. FR-4 epoxy fiber­
glass is a common substrate used for most printed circuit board applications [20] because
o f its dimensional stability, good adhesion, large area processability and low cost.
However, the low thermal stability of epoxy-based boards (Tg ~ 130°C) makes it
impossible for use with high performance dielectrics like polyimides in applications like
micro-via build-up layers. This requires the use o f more expensive, high temperature
boards (substrates) such as polyimide-glass substrates. Thus, in order to cure polyimides
on low temperature organic substrates, new low-temperature processing techniques need
to be explored. In this study, variable frequency microwave (VFM) processing has been
investigated as a rapid, low-temperature processing technique to cure polyimides on lowcost organic substrates.
169
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7.1 Results and Discussion
7.1.1 VFM heating characteristics o f organic substrates
The VFM heating characteristics and thermal stability o f different substrates were
studied in order to determine the power-time-temperature window for VFM processing.
The VFM heating characteristics were investigated to understand the substrate
contribution to heating. The heating rates o f blank FR4, Cu-FR4, CF-epoxy composite
board, and silicon with no polymer film, at a constant power o f 200 W, central frequency
o f 6.425 GHz, 0.15 GHz bandwidth and 0.1 sec sweep time are compared in Figure 40.
Under these conditions, silicon has the highest heating rate o f all the substrates studied.
Blank FR4 and BT boards have a negligible heating rate, while Cu-FR.4, Cu-BT and the
CF-epoxy board show considerable heating rates. For substrates with negligible heating
rates, VFM processing thus provides an opportunity to selectively heat films with a
greater dielectric loss to a much higher temperature without substantially heating the
substrates.
From Equation 2.11, the amount o f microwave energy absorbed by a material at a
constant field and applied frequency is directly proportional to the dielectric loss in the
material. The primary mechanism of dielectric loss in a homogeneous dielectric material
at microwave frequencies is orientational or dipolar polarization. Interfacial polarization
may be significant in electrically heterogeneous materials. In heterogeneous materials,
the difference in the ease o f charge transfer across the phase boundaries results in charge
170
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
build up at the interfaces and leads to polarization in an applied electric field. This is
influenced by the permittivities and the conductivities o f the phases present.
As
discussed in the background section [Chapter II], differences in the dielectric permittivity
(e) and conductivity (a) at the interface between two phases for e.g., at the glass/carbon
fiber-epoxy interface, could lead to Maxwell-Wagner effects, causing charge build up
and dissipation. Hence, in the case o f blank FRA, Cu-FR4, CF-epoxy and BT substrates,
interfacial polarization and the resulting loss also contribute significantly to the total
heating. The microwave heating o f silicon is known to be significant and a function o f
the dopant concentration [79].
171
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
200
180
160
\
w
140
£ 120
g 100
& 80
i
H
60
40
20
0
0
20
40
60
80
100
120
140
160
180
time, sec
♦ FR4
Cu-FR4
a CF-Epoxy
° Silicon
Figure 40 VFM Heating rates o f different substrate materials at a constant power o f 200
W, central frequency o f 6.245 GHz, 10% bandwidth and O.lsec sweep time.
The effect o f microwave power on the heating characteristics o f the substrates was
also studied. Figure 41 shows the influence o f applied power on the heating rate o f blank
FR4 board subjected to a constant power o f 100,200, 300,400 and 500 W. As expected,
the rate o f heating and the ultimate temperature reached increase with applied power.
The maximum temperature reached by the top surface o f the blank FR4 board (no
nitrogen) does not exceed 90°C even after 3 minutes at 500 W.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
VFM heating rates of FR4 board
.........
80
100
time, sec
OlOOW
a 200W
□ 300W
180
• 400W
XSOOW
Figure 41 The effect of microwave power on the heating rate o f blank FR4.
Similar results were observed with the other substrates (Cu-FR4, CF-epoxy, and
silicon) with significantly higher heating rates and ultimate temperature. Furthermore,
the effect o f higher power was more pronounced with higher absorbing substrates.
7.1.2 Thermal stability o f organic substrates
The thermal stability o f the different organic substrates i.e. the time-temperature
dependence o f the degradation/decomposition was determined by thermogravimetric
analysis (TGA). Both dynamic and isothermal TGA runs were performed. Figure 42
173
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
shows the weight loss o f an FR4 board ramped to 450°C at ramp rates o f 3, 10 and
15°C/min under a nitrogen atmosphere. Typically 5 to IS mg samples were used for
these studies.
As seen in Figure 42, in the temperature range S0°C to 200°C, negligible weight loss
occurs (weight loss if any may be attributed to absorbed moisture) under all the ramp
rates studied. However, at higher temperatures, the board begins to lose mass rapidly
(substantially, ~ 20 to 25% by weight) beyond a critical degradation temperature. This
degradation temperature depends on the ramp rate used with slower ramp rates typically
giving a lower degradation temperature. This ramp rate dependence is largely due to the
thermal conductivity limitations o f the organic material and the time it takes the sample
temperature to equilibrate with the TGA pan temperature at any given ramp rate. The
presence o f oxygen during the TGA run also seemed to slightly affect the degradation
temperature. Most o f the TGA studies were conducted under a nitrogen ambient (~30
sccm/min). Studies conducted under an air ambient showed slightly reduced degradation
temperatures.
174
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
70
♦ 3°C/min
■ 10°C/min
a 15°C/min
60
50
1
1
1
1
1
1------------
100
150
200
250
300
350
400
Temperature, °C
Figure 42 Dynamic TGA of an FR4 board at ramp rates o f 3,10 and lS°C/min in a
nitrogen atmosphere.
Alternately, time at temperature may also be used as a thermal stability metric. This
assumes critical importance as dielectric curing often involves hold times at different
temperatures for various reasons such as promoting the cure reaction or assisting solvent
(and photoproduct in the case o f photosensitive dielectrics) evolution during the curing
process. Isothermal TGA studies were conducted on the organic substrates to study the
time-dependence o f degradation at different temperatures.
For isothermal studies,
samples were ramped at a high ramp rate (> 80°C/min) to the desired temperature and the
175
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
weight loss with time was monitored. Figure 43 shows the effect o f temperature on the
time for degradation of the FR4 board.
100
Ofl
«
S
cc
«e
ct
£
95
220°C
230°C
240°C
250°C
90
85
80
75
10
20
30
40
50
60
Tim e, min
Figure 43 Time-temperature dependence o f the thermal stability o f FR4 board under a
nitrogen atmosphere.
As seen in Figure 43, the FR4 sample is reasonably stable even after 60 min at 220°C
and 230°C. However, at 240°C and higher, the film undergoes degradation with the time
for significant degradation decreasing with temperature. For example, the sample begins
176
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
to lose mass significantly after 40 min at 240°C and after only 20 min at 250°C. BT and
CF-epoxy organic substrates showed greater thermal stability than an FR4 substrate.
However, similar trends in thermal stability (dependence on ramp rates and time at
temperature) were seen with these substrates also.
Figure 44 compares the thermal
stability (percent weight loss with temperature) o f a) FR4, b) BT and c) CF-epoxy
samples under a dynamic TGA ramp o f 10°C/min (all the samples were purged in a
nitrogen atmosphere).
As seen in Figure 44, all the three organic substrates show negligible weight loss
during the ramp until a temperature o f about 200°C. The FR4 substrate rapidly loses
mass at a temperature o f 288°C whereas BT and CF-epoxy substrates begin to lose mass
around 318°C. The BT substrate shows a degradation behavior similar to FR4 and the
weight loss at 318°C is sudden and significant whereas, the CF-epoxy substrate shows a
more gradual weight loss. In general, the thermal stability o f the organic substrates
studied was in the order FR4< B T ^ CF-epoxy.
177
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
DD
95
90
5
6
tn
t0»
A
85
£
80
75
70
50
100
o FR4
150
200 250 300 350
Temperature, °C
A BT
400
450
CF-Epoxy
Figure 44 Weight loss of FR4, BT and CF-Epoxy boards under dynamic TGA at a ramp
rate o f 10°C/min in a nitrogen atmosphere.
As seen from Figure 42 and Figure 43, the organic substrates undergo catastrophic
failure once they reach the degradation temperature. That is, above a certain threshold
temperature, there is a significant weight loss in a very short period o f time. Initial
weight loss in the substrates could be attributed to absorbed moisture and possible
evaporation o f any residual solvent from the casting process.
The glass transition
temperature (Tg) o f the FR4 substrate used in these studies was ~ 140°C. The board is
178
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
dimensionally unstable if held for long periods o f time at temperatures greater than its Tg
due to increased polymer chain mobility above Tg. The glass/carbon fiber reinforcements
hold the layer in place in the plane o f the board (x-y). As a result, high temperature
exclusions (above Tg) cause the board to bubble up in the z-direction. Delamination o f
the inner layers/ loss o f adhesion between successive layers o f glass-reinforced epoxy
could precede the degradation of the board.
The thermal characterization o f organic substrates helped identify some key trends
with regard to time-temperature dependence of degradation o f the different organic
substrates. It is important to note however, that the thermal stability/behavior o f the
organic substrates could be significantly different when heated in a microwave field.
Unlike thermal heating, where the inner layers are heated through conduction from the
outer layers (which are exposed to the high temperature ambient), heat generation is more
uniform in the case o f microwave heating.
However, the existence o f interfacial
polarization and differences in thermal conductivities at the interfaces could result in
differences in heat distribution leading to thermal gradients. Moreover, delamination
effects could be exacerbated (could occur at a lower bulk temperature) due to microwave
enhanced chain mobility. Furthermore, it was also found that the processing of organic
substrates in the VFM was sensitive to the presence o f pinholes and other defects in the
organic substrate (due to field focusing at such defects). To eliminate any dependence on
absorbed moisture, all samples were allowed to equilibrate to the ambient conditions
before performing any further VFM studies.
179
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In order to study the feasibility o f curing polyimide films on these substrates, two
different approaches were followed: I) constant power processing, where the films were
subjected to a constant microwave power for a certain time and 2) time-at-temperature
approach where samples were ramped at a constant temperature ramp rate to a final cure
temperature and held at that temperature for a certain time. Figure 45 compares the
heating rates o f an FR4 board with and without a polyimide film at a constant microwave
power o f 400 W. As seen in Figure 45, the heating rates are significantly different with
and without the polyimide film. The temperature was measured in both cases using a
fiber optic probe taped to the top surface. The blank FR4 substrate heats to less than
80°C after 3 min at 400 W (also seen from Figure 41) whereas with the polyimide film
coated it reaches 190°C in less than 3 minutes.
180
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
200
180
160
U
Q 140
jf 120
s
2 ioo
I*
80
H
60
40
20
0
0
20
40
60
80
100
120
140
160
180
__________________________time, sec__________________________
|
♦ b lu k FR4
A with Polyimide Filin
Figure 45 Comparison o f heating rates o f FR4 board with and without Polyimide film at
400 W constant microwave power.
This shows that the polyimide film can be selectively heated on the FR4 substrate.
The high dielectric loss o f the polyimide film in its uncured state and the solvent in the as
spin-cast film contribute to the high heating rate initially. As the solvent evolves from
the film and the dielectric loss o f the film decreases as imidization proceeds, the
temperature o f the film reaches a steady value.
Differences in the processing methodology can affect the nature o f the material
response in a microwave field as noted by Lewis et al [51]. A pulsed power control o f
temperature is expected to produce a different response as compared to a constant power
181
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
case depending upon the different relaxation mechanisms in the polymeric material.
Furthermore, high power processing often results in a high initial heating rate leading to
bubbling or voiding in thin films due to rapid evolution o f solvent from the film. In this
current study, ramp rates as high as 60°C/min could be achieved without any significant
bubbling or voiding o f films in the thickness range 10 to 20 pm. However, preliminary
studies were directed at comparing the effectiveness o f VFM processing to traditional
thermal curing and hence a time-at-temperature approach was used to cure the polyimide
films. Factors contributing to the microwave absorption and the resulting relative heating
rates will be discussed in a later section
Figure 46 shows the IR spectra o f a) a softbaked PI 2611 film, b) a standard
thermally cured film on silicon, and a film cured by (c) VFM and (d) thermally on blank
FR4 board. Film c) was ramped at 3°C per minute to 200°C and held for one hour at that
temperature and film d) was ramped at 1S°C per minute to 200°C and held for S min. It
can be seen from Figure 46 that there are no significant differences between the IR
spectra o f films cured on either substrates at similar conditions (spectra (c) and (d))
suggesting that the chemical structure o f these films is identical. Similar results were
found from transmission mode IR analysis on free films peeled o f the substrates
suggesting that the surface and bulk chemical structure o f the films was identical.
182
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
a
n
jd
u
e
w
-O
<
1500
2000
1000
Wavenumber cm *1
Figure 46 Infrared spectra o f PI2611 Films: a) Soft-baked, b) standard thermally cured on
silicon ramped at 3°C/min to 350°C and held for 1 hour at 350°C, c) VFM cured on FR4
substrate, ramped at 15°C/min to 200°C and held for S min at 200°C, d) thermally cured
on FR4 substrate, ramped at 3°C/min to 200°C and held for 1 hour at 200°C.
However, subtle differences between the standard thermal cure and the lowtemperature cures (both thermal and VFM) such as a weak absorption at 18S0 cm*1 were
observed and were further investigated.
The extent o f imidization or ring closure achieved in the cured films was determined
by F I Ik analysis with the standard thermal cured film as the reference. The extent o f
im idization was estimated by using the following equation (as described in Chapter 5).
183
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Degree o f Imidization
=
------
(7.1)
'■•'*1359 ' " IS16/R eference
where, A is the absorbance (peak height) and the subscript indicates the wavenumber
corresponding to the peak.
The intensity o f the-C-N-C- stretch o f the imide ring,
measured by the 1359 cm*1 peak in the FTIR spectrum, can be correlated to the extent of
imidization in the film. The absorbance o f the -C-H stretch o f the aromatic diamine,
which occurs at 1516 cm*1, remained unchanged during curing and was used as the
internal standard to normalize the peak at 1359 cm*1.
The reference peaks heights
correspond to a film cured by the standard thermal cure described above for which 100%
imidization is assumed.
Table 14 shows the extent o f imidization (as calculated from Equation 7.1) achieved
in polyimide PI 2611 films cured in a conventional thermal furnace and the VFM furnace
at different processing conditions. It can be seen from Table 14 that a higher extent o f
imidization can be achieved by VFM processing for much shorter cure times as compared
to conventional thermal furnace cure. For example, a 4 hour thermal furnace cure at
175°C gives 50% imidization, while a 5 minute VFM cure at 175°C gives an extent o f
imidization o f 92%. Further, a 5 min VFM cure at 200°C gives almost 100% imidization
without degradation o f the epoxy board. Only 73% imidization is achieved in a film
cured for an hour in a conventional thermal furnace at 200°C. PI 2611 films ramped in a
184
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
conventional thermal oven at 3°C/min and cured for an hour at 250°C reach 100%
imidization but the epoxy board was decomposed. This is evident from Figure 43, which
shows that the FR4 substrate begins to lose mass (degrade) even after 20 minutes at
250°C.
Traditionally, it has not been possible to cure high performance dielectric
materials like polyimides on low cost low-temperature epoxy substrates.
Complete
imidization at low temperatures can be achieved on organic substrates without
degradation o f the substrate by VFM processing.
Table 14 Extent o f imidization achieved in polyimide PI 2611 films cured on blank FR4
substrate by VFM and conventional thermal furnace under different cure conditions.
Cure Method Ramp Rate I Temperature I Hold Time % Imidization
°C/min
°C
min
Thermal Cure
3
3
3
3
175
175
200
250
60
240
60
60
31
50
73
100*
VFM
15
15
175
200
5
5
92
100
♦Board decomposed
185
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7.1.3 Substrate selectivity o f VFM cured films
Figure 47 shows the extent o f imidization of polyimide PI 2611 films cured in the
VFM furnace on different substrates at temperatures o f 17S and 200°C as estimated from
ER analysis using Equation 7.1. Each o f these films was ramped at 15°C to the final
temperature o f 175 or 200°C and held for 5 min. It can be seen from Figure 47 that as
expected, the extent o f imidization increases with increasing cure temperature for all the
substrates.
PI 2611
FR4
C u -F R 4
C F -E p oxy
Silicon
Figure 47 Comparison o f the extent o f imidization o f PI2611 films (estimated from
FTIR) cured on different substrates in the VFM oven at a constant ramp rate o f 15°C/min
and 5 min hold time at 175 and 200°C.
186
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
However, the unique feature o f VFM processing is the selectivity to the substrate.
As seen in Figure 47, the extent o f imidization achieved in the polyimide films processed
for the same time and temperature varied depending on the substrate on which the films
were coated. No such dependence on the type o f substrate was observed for conventional
thermal curing in a tube furnace. Intuitively, one would expect faster heating substrates
to enable more rapid curing. However, it can be seen that slower heating substrates give a
higher extent o f imidization for the same processing conditions.
Almost 100%
imidization was achieved in films cured on FR4 and Cu-FR4 substrates after 5 in at
200°C while less than 100% imidization was achieved on both CF-epoxy and silicon
(faster heating) substrates. Similar trends were also observed with another polyimide PI
2734. As seen in Figure 48, slower heating substrates such as blank FR4 and Cu-FR4
give a higher extent o f imidization as compared to CF-epoxy or silicon.
187
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
P I 2734
100
e 80
!*»
-O
a
M
40
^
20
0
FR4
Cu-FR4
CF-Epoxy
Silicon
Figure 48 Comparison o f the extent o f imidization o f PI2734 films (estimated from
FTIR) cured on different substrates in the VFM oven at a constant ramp rate of 15°C/min
and 5 min hold time at 175 and 200°C.
The differences in the extent o f imidization o f VFM cured films depending on the
substrate on which the film is coated arise due to several reasons. The heating rate or
temperature reached in an applied field depends 1) on the amount o f heat generated and
2) the amount o f heat removed from the film via conduction, convection and radiation
modes o f heat transport The amount o f power absorbed per unit volume o f material in
an electromagnetic field at a constant power and applied frequency is a function o f its
dielectric properties, particularly the dielectric loss and how it changes with temperature.
For example, the dielectric loss decreases as solvent (high loss) leaves the system with
188
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
increase in temperature. When ramping the material under a set thermal profile, the
effectiveness o f coupling microwave energy is also determined the reflected power which
is a function of the material properties and surface characteristics.
At a molecular scale the amount o f thermal energy dissipated in a dielectric film
under an applied field depends on several factors [22] including the functional groups
present (dipoles) their mobility, the backbone flexibility, and more importantly the
presence o f solvent (discussed in the following section). The amount of heat removed
from the film depends upon the thermal characteristics o f the dielectric material (heat
capacity, thermal conductivity (k) etc. and their temperature dependence) and the
temperature of the ambient (determines the radiation loss from the film).
Typical
dielectrics, like epoxies and polyimides, have k values in the range 0.2 to 0.3 W/m-K.
When the dielectric film is coated onto a substrate, the microwave absorption
characteristics and the thermal properties o f the substrate also need to be considered. The
thermal characteristics o f the substrate also affect the distribution o f thermal energy in the
material stack. For example, a highly conducting substrate (such as silicon (k = ISO
W/m-K) alumina (k = 21 W/m-K) and CF-epoxy (k can be [137] as high as 270 W/m-K
depending upon the carbon fiber loading)) removes heat from the dielectric film more
efficiently than a relatively low thermal conductivity substrate like polyimide (k = 0.2 to
0.3 W/m-K), FR4 and BT (k = 0.3 to 0.4 W/m-K). This results in a larger thermal
gradient along the thickness o f a non-conducting substrate. Indeed, this effect has been
previously illustrated [138] in the form o f a simulation o f thermal profiles based on
material properties in a material stack with different substrate materials under a
189
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
microwave field using PHYSICA, an electromagnetic-thermal modeling simulation
software. Results form this study show that for the same volumetric heat generation rate
in a microwave field, the temperature distribution is significantly dependent on the
thermal and dielectric properties o f the substrate and consequently the effectiveness of
microwave processing also depends on these factors.
The energy needed to drive the imidization reaction in the polyimide film is
delivered from two sources: I) the heat due to dielectric loss within the polyimide film
itself and 2) heat from energy absorbed within the substrate. For substrates with a high
(dielectric) loss, a significant amount o f heat is transferred from the substrate and thus the
energy absorbed within the polyimide film itself to reach a certain temperature is lower.
For slower heating substrates however, the energy required for imide-ring closure comes
primarily from the dielectric loss o f the polyimide precursor film and its solvent. This
requires an even higher applied power as the solvent leaves and imidization occurs,
which increases the energy absorbed and dissipated by the polyimide film (from Equation
2.11). The microwave ‘acceleration’ effect is thus more pronounced in systems where
the dielectric loss in the film o f interest contributes primarily to the heating rather than
the substrate. When the dielectric loss in the substrate significantly contributes to the
heating, the process would compare closer to a thermal hot-plate cure. An ideal and
efficient microwave processing scheme would be one in which the material properties are
such that the dielectric loss o f the uncured film contributes the most to generated heat and
the thermal loss from the film to the ambient and the substrate are minimized. VFM
processing thus provides for both rapid and selective heating.
190
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In order to determine the effectiveness o f the VFM curing, the optical, electrical and
thermo-mechanical properties o f PI 2611 films cured on FR4 boards were compared with
those o f standard thermal furnace cured films on silicon. The optical and electrical
properties are summarized in Table IS.
Table IS Comparison o f electrical and optical properties o f VFM cured, standard
conventional oven cured and hotplate cured films.
Cure Method
Std. Thermal Cure on
Silicon: 3*C/min
350*C- 1 hour
VFM Cure on
FR4:
15’C/min
200*C-15 min
Hot plate Cure on
Silicon: 200*C-15 min
Refractive index
Birefringence
Dielectric Constant
“*y
n*
An
k* = n*2
s*
1.8384
1.6258
0.2126
2.6433
2.99
1.6890
1.6279
0.0611
2.6499
3.29
1.6909
1.6366
0.0543
2.6785
N/A
The high temperature standard thermal cure described previously, results in a high
degree o f in-plane orientation o f the polymer chains, which makes the cured film highly
anisotropic [99]. Standard thermal cured (highly oriented) PI 2611 has a high “In-plane”
dielectric constant (4.0) and index o f refraction (1.83) as compared to through-plane
values o f 2.9 and 1.62 [139]. As seen from Table IS, films cured on FR4 substrates show
191
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
a much lower index o f refraction and birefringence as compared to the standard thermal
cured films.
The measured through-plane dielectric constant o f fully imidized films
cured on the FR4 substrate is 3.29 (+/- 0.09), which is higher than that of the standard
thermally cured film. This may be a result of: 1) a decrease in the degree o f in-plane
orientation, which may also contribute to the increased through plane dielectric constant
and 2) residual solvent in the film which can lead to a higher loss and dielectric constant.
The high frequency dielectric constant o f the film, which is relatively independent o f
the dipole orientation in the film may be estimated as k = n2.
The through-plane
dielectric constant is approximated from the through-plane index o f refraction.
The
dielectric constant for all the samples was measured at a frequency o f 10 kHz where
dipole contributions to dielectric constant are significant.
From Table IS, the high
frequency through-plane dielectric constant (estimated as k = n2) for both the standard
thermally cured film and the VFM cured film (both fully imidized as inferred from FTIR)
is the same (2.64). However, the measured low-frequency dielectric constant is higher
for the VFM cured film. The measured in-plane index o f refraction o f the VFM cured
film is lower than the standard thermally cured film leading to a lower birefringence,
indicating reduced orientation in the plane o f the film.
This implies greater dipole
contribution to the low frequency through-plane dielectric constant, which is consistent
with the higher measured through plane dielectric constant o f 3.26. Reduced in-plane
orientation should also reduce the dipole contribution to the low frequency in-plane
dielectric constant.
192
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A decrease in the in-plane orientation could occur due to several reasons. As seen in
earlier studies, in-plane orientation in VFM cured films increases with cure temperature
between 200 and 350°C. Hence, PI 2611 films VFM cured on FR4 at 200°C are expected
to have a lower birefringence than the high temperature standard thermal cured films. It
was also shown that increasing the ramp rate also reduces the in-plane orientation o f the
cured films. The higher ramp rate used in the VFM cured films (15°C/min) as compared
to the standard thermal cured film (3°C/min) may also have contributed to a decrease in
the in-plane orientation. Further, the films cured on FR4 were typically thicker (10 to 20
pm) than the films cured in silicon (< 10 pm). Several studies [111] have shown that the
anisotropy in rigid-rod backbone type polyimides is thickness dependent. The in-plane
orientation and hence the birefringence reduces with increasing film thickness for several
reasons which include a) increased solvent evaporation times in thicker films which
allows greater chain relaxation and b) reduced substrate surface energy effects with
increasing thickness [111]. Furthermore, the birefringence and the index o f refraction o f
the films VFM cured on FR4 substrate were measured by lifting off free-films from the
substrate. Previous studies [111] have shown that cured films undergo a relaxation o f the
thermal or extrinsic stress induced during curing when lifted o ff the substrate on which
they were cured while retaining the intrinsic stress. This could also lead to a reduction in
the birefringence o f the films.
Another factor that can influence the dielectric constant o f cured films is the
presence o f residual solvent.
As discussed earlier, residual solvent allows greater
molecular relaxation to occur resulting in reduced orientation. The residual solvent in the
193
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
VFM cured films could be higher as compared to standard thermally cured films
considering the lower cure temperature and shorter cure time. Residual solvent may also
lead to a higher dielectric loss due to increased dipolar contribution to dielectric loss.
Residual solvent can also have an adverse impact on the mechanical properties o f cured
films and is a potential reliability concern.
The thermo-mechanical properties o f VFM cured PI 2611 on FR4 substrates were
characterized by a thermo-mechanical analyzer (TMA) and dynamic mechanical analyzer
(DMA).
The TMA results o f a PI 2611 film VFM cured on an FR4 board at a
temperature o f 200°C for 5 min ramped at 15°C/min are shown in Figure 49. Figure 49
shows the change in film dimension as a function o f temperature during two ramp cycles
(labeled run #1 and run #2). A 25 mm x 5 mm sample o f the film was ramped at a rate o f
10°C/min to 450°C over two cycles under a constant load o f 0.05 N. As seen in Figure
49, the film undergoes a transition at a temperature T, or Tg o f 220°C. The sudden
decrease in film dimension in the range 220 to 350°C is believed to occur due to film
shrinkage as a result o f residual solvent evolution during the TMA ramp cycle. The
coefficient o f thermal expansion (CTE) o f the film in the range 50°C to 150°C calculated
from Figure 49 was 12 ppm/°C. During the second ramp cycle in the TMA no distinct
transition was observed and the CTE in the range 50°C to 150°C was 12 ppm/°C. The
CTE o f PI 2611 films VFM cured on FR4 substrate (estimated at 12 ppm/°C from the
TMA results) is much higher than that o f standard thermal cured films (~ 3ppm/°C). The
CTE o f rigid rod type polyimides is known to increase with a decrease in the in-plane
orientation. The CTE estimates from TMA results are consistent with the optical property
194
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
measurements, which showed a reduced birefringence (in-plane orientation) in films
VFM cured on FR4 (Table 15). The lowering o f Tg o f VFM cured PI 2611 on FR4 could
be a result o f solvent-induced plasticization.
200
150
Run #2
i 100
&
e
«
.c
a - 1 2 ppm/"C
e
em
•m
m
220°C
ea> -50
E
Run # 1
s .100
-150
0
100
200
300
400
Tem perature, °C
Figure 49 Thermo-mechanical analysis o f PI 2611 VFM cured at a ramp rate of 15°C/min
to 200°C for 5 minutes. TMA conditions: Ramp rate: 10°C/min; Load: 0.05 N
Figure 50 shows the DMA results o f a PI 2611 film ramped at 5°C/min to 450°C at
test frequency o f 1 Hz. The storage modulus is plotted on the y-axis on the left and the
195
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
loss modulus and tan delta are plotted on the y-axis on the right as a function o f
temperature.
250
4.5
ea
a.
U
4.0
200
3.5
3.0
e
5
2.0
6
68
k.
1.5
e
t/2
150
2.5
a
100 *5
1.0
0.5
0.0
0
50
100 150 200 250 300 350 400 450
T e m p e ra tu re , °C
Figure SO Dynamic mechanical analysis o f PI 2611 VFM cured on FR4 substrate. VFM
cure conditions 15°C/min ramp to 200°C 5min hold. DMA test conditions: ramp rate:
5°C/min. frequency: 1 Hz.
As seen in Figure SO, the film exhibits a storage modulus o f - 4 GPa and shows a
peak in loss modulus and tangent delta at a temperature o f ~ 287°C which corresponds to
196
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the Tg o f the film. However, the data also shows fluctuations in the temperature range
200°C to 2S0°C. This may be attributed to the presence o f residual solvent. Further, a
higher Tg from the DMA results as compared to the TMA results could be due
evaporation o f the residual solvent during the slow ramp in the DMA.
Mechanical
testing using Instron pull tests showed a Young's modulus of - 4.52 GPa which is
significantly lower than a standard thermal cured film (~ 8.66 GPa). A decrease in the
Young's modulus o f VFM cured films may be attributed to decreased in-plane orientation
and solvent induced plasticization.
The CTE of VFM cured PI 2611 is significantly lower than commonly used epoxy
based build-up dielectrics (CTE - 40-80 ppm/°C) which could result in reduced CTEmismatch induced thermo-mechanical stress. The Tg is also higher than that of the FR4
itself (-140°C) and high-end epoxy dielectrics (~180-190°C).
However, the thermo­
mechanical properties o f VFM cured PI 2611 on FR4 are significantly different from
those o f the standard thermal cured films. The optical, electrical and thermo-mechanical
characterization o f films show the presence o f residual solvent in the cured films leading
to degradation in cured film properties. Thermogravimetric analysis o f cured samples
can identify possible differences in solvent content o f cured films.
Figure 51 compares the weight loss o f a VFM and a thermally cured PI 2611 film
ramped at 10°C/min to 500°C in the TGA. Both films were cured at 200°C for 5 min. As
seen in Figure 51, both the VFM and thermally cured film show significant weight loss
with temperature, - 10 % for the VFM cured film and -23 % for the thermally cured
film.
197
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
VFM
Thermal
75
70
65
60
55
50
50
150
250
350
450
T em p eratu re,°C
Figure 51 Weight loss o f VFM and thermally cured films under a TGA ramp o f 10°C/min
to 500°C. Cure conditions: 5 min at 200°C.
The weight loss in the cured polyimide films under the TGA run could be due to
several reasons. Weight loss can occur due to the evolution o f absorbed moisture as
polyimide films are known to absorb moisture and partially cured films and films with
residual solvent in particular can absorb a higher amount o f moisture as they serve as
potential hydrogen bonding sites. The weight loss can occur as a result o f evolution o f
condensation products that are a) trapped within the film and could not be removed
198
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
during the cure reaction and b) liberated during imidization o f any unreacted acid groups.
Weight loss can also occur due to the evolution o f any residual solvent in the cured film
or due to other high molecular weight species resulting from any side reactions. It can
also be seen from Figure SI, that the thermally cured film begins to lose weight at a lower
temperature than the VFM cured film. As seen from Table 15, even a one hour cure at
200°C in a thermal oven gives an imidization o f only 72 %. Hence, in the case o f the
thermally cured film, weight loss could also be due to the evolution o f condensation by­
products. The lower weight loss in the VFM cured film shows the effectiveness o f VFM
processing in solvent removal. However, a 10 % weight loss is significantly higher than
that o f a standard thermal cured film (< 1 % as discussed earlier in Chapter V).
Mass spectrometry (MS) studies were performed to investigate the nature o f the
products evolved during the TGA ramp studies. A ramp rate o f 10°C/min, similar to that
used in the TGA run, was used in the constant ramp direct probe-temperature MS. Figure
52 and Figure 53 show the MS results o f a PI 2611 film (cured at 200°C for 5 min on an
FR4 substrate) ramped at 10°C/min in mass spectrometer. From Figure 52, it can be seen
that the total ion trace follows the same trend as the ion trace from species with amu 99,
which corresponds to the molecular weight o f the solvent NMP (C5H9NO). Also from
Figure 53, all the significant species evolved are solvent related [140] (Mol. wts. 99, 71,
56, 44). Further, the solvent related moieties begin to evolve at a temperature o f 220°C.
This confirms that weight loss in the TGA and the degradation in the electrical and
thermo-mechanical properties discussed earlier are indeed residual solvent related.
199
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
These results show that while VFM curing o f PI 2611 on FR4 substrate at 200°C
showed significant improvement over thermal curing at 200°C, the properties are not
comparable to the standard thermal cured films.
The films are almost completely
imidized but the residual solvent in cured films is as high as 10 % by weight and
drastically affects the electrical and thermo-mechanical properties. Moreover, residual
solvent is not desirable from a processing and reliability standpoint. Residual solvent can
cause blistering in metallization and film shrinkage during subsequent thermal steps
could affect the dimensional stability o f assembled packages. Further efforts were thus
focused on understanding the critical factors affecting low temperature polyimide curing
and the effect o f processing conditions in reducing the residual solvent.
200
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ION TRACE. Flaggir*g=Scan Number. Max.Scan=1500#44:18.06.
100 Total Ion Current. Max.Int.=59.3853.
3
749
5 0 -i
101
280
i "
i
"i
1318
561
348 450
1409
"-»
rT-ji »i »; i i i i j i i i i | i i i i j » i i i | r »» »| »i i »|
0
200
400
k>n»99 (Norn). Max.Int.=8.402.
600
800
1000
1200
1400
1126
277
NMP
N
O>
p r
I" 1
0
r
;
I
I
I
I
|
I
200
371
461
i
j
1 1
125 274
i
i
[
i
^
i
1 1 1 1
i
> i
I
600
40
Ion -113 (Nom). Max.Int.
100 -
?
1409
i
i
i
i
I
i
i
I
800
i
[
I
t t
>j
i
»
1000
» » I
I
I
I
I
j
I
I
1200
I
»
I
1400
.1584.
1316 1414
367 4 ,9
40<|
I
1
r j
r i i
600
800
_______Scan Number
I | 1 1 I I | I I I I | I I I I | I I I I | I I I 1 |
1000
1200
1400
200°C
Probe Temp. Mass Spec at 10°C/mb (RT - 450°C)
Figure 52 Mass spectrometry results of PI 2611 film VFM cured at 200°C for 5 min on an FR4 substrate.
2 o i.
(L O)
in
K m
202
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 53 Mass spectrum of PI 2611 film VFM cured at 200°C for 5 min on an FR4 substrate.
CO ■.
NOTE TO USERS
Page(s) not included in the original manuscript
are unavailable from the author or university. The
manuscript was microfilmed as received.
203
This reproduction is the best copy available.
UMf
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The following were identified as critical issues to consider for low temperature
polyimide curing: a) side reactions and their impact on cured film properties b) impact o f
processing conditions on the solvent evolution and imidization and c) the effect o f
solvent on VFM curing characteristics. The following sections will discuss these issues
and how they impact polyimide curing on organic substrates.
7.1.4 Anhydride formation and other side reactions
Several studies in literature have reported the existence o f competing side reactions
(especially for low temperature cures) during the imidization process [141, 142, 143,
144] including those that result in the formation o f cyclic and linear anhydrides,
isoimides and intermolecular links. The formation o f an anhydride or an anhydride-imide
system is especially of concern as it can lead to a reduction in the molecular weight o f the
cured polymer, which drastically affects the mechanical strength o f the material. The
mechanism o f formation o f an anhydride and/or imide-anhydride is represented in Figure
54.
During the cure reaction, the carbonyl carbon can undergo a nucleophilic attack
[143] either from 1) the nitrogen o f the amide group, resulting in the formation o f the
desired imide linkage or 2) the oxygen from the acid-hydroxyl group resulting in the
formation o f an imide-anhydride. The formation o f the imide-anhydride is a reversible
reaction and favors the formation o f the acid at high temperatures (> 270°C). The
formation o f an anhydride may be monitored by FTIR spectroscopy. The symmetric C = 0 stretch o f the cyclic anhydride has an infrared absorbance at - 1850 cm'1 [141,142,
204
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
144]. Figure 55 compares the FTIR spectra o f low temperature VFM and thermal cured
films with the standard thermal cured film.
<V
C n O
o
y D
H .N - 0 - N H ,
o
PDA
BPDA
Polyamic acid
Polyimide
Jm
Polyimide anhydride
Figure 54 Two reaction pathways for polyamic acid BPDA-PDA: a) polyimide formation
(desirable) b) anhydride formation (undesirable side reaction).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Therm al 350°C
V
w
s
a
€e
w
JB
<
2000
VFM 200°C
Therm al 200°C
1500
1000
Wavenumbers cm'1
Figure 55 FTIR spectra o f PI 2611: Anhydride formation.
As seen in Figure 55, a folly cured polyimide film (VFM or standard thermal cure at
350°C) does not show any absorbance at 1850 cm '1. However, for the low-temperature
thermal and VFM cured polyimide films, a weak absorbance can be seen at a
wavenumber o f ~1850 c m 1 which is attributed to the -C = 0 stretching o f the cyclic
anhydride. Films were cured at different temperatures ranging from 175°C to 350°C in a
conventional tube furnace and in the VFM furnace. It was found that the absorbance at
1850cm'1 initially increases, peaks at ~ 220°C and then gradually decreases before
completely disappearing at - 270°C consistent with similar studies on polyimides [142].
Similar trends with anhydride formation were observed with the low temperature VFM
cures except that the relative concentration o f anhydride to imide (as estimated from the
206
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
IR absorbance peaks with the same internal reference) formed at each temperature was
lower than the corresponding thermally cured sample. This observation corroborates the
theory o f the existence o f a high local temperature, as such a high local temperature
would result in any anhydride formed to be converted back into an imide. For films
cured on organic substrates at 200°C, the extent o f imidization obtained was higher for
VFM cured films as compared to films cured thermally at the same temperature while the
concentration o f the anhydride was much lower. The percent imidization achieved by
VFM curing after 5 min at 200°C is close to 100% although there is evidence of
anhydride formation. However, as discussed earlier, subtle differences in the estimates of
imidization at such high % imidization are often beyond the resolution/sensitivity limits
o f the FTIR. The imidization reaction is not complete (reaches ~ 73%) even after 1 hr at
200°C in the thermal oven. Incomplete reaction in the thermally cured sample is also
evidenced (spectrum (d) o f Figure 46,) by the existence o f a weak absorbance (shoulder)
at 1680 cm'1 corresponding to the carbonyl (-C=0) stretch o f the amide group o f the
polyamic acid precursor. No evidence o f other side reactions such as isoimde [14S] (C=0: 1795-1820 cm'1) or inter-chain links (-C = 0 :1670 cm'1) was observed.
Whether or not the chain scission due to the formation o f an anhydride results in a
significant reduction in molecular weight depends on the location o f the anhydride
species formed. From Figure 54, for cases where the anhydride formation occurs at the
chain ends or for m « n, given the concentration o f such species formed, no significant
change in molecular weight and hence the mechanical properties is expected.
A
reduction in the molecular weight occurs only via reaction schemes where m * n/2.
207
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Indeed, mechanical testing by Instron pull tests showed a low Young’s modulus o f 4.52
GPa for PI 2611 films cured on FR4 substrate, which is significantly lower than the
standard thermal cured film (~ 8.66 GPa). However, considering the small amount o f
anhydride-type species formed in VFM cured films, their impact on the mechanical
properties is considered negligible. Reduced in-plane orientation and residual solvent
induced plasticization are likely to have a more dominant effect on the mechanical
properties of PI 2611 films cured on FR4 substrate.
7.1.5 Effect o f solvent evolution on imidization
Imidization o f a polyamic acid to a polyimide is a complex process involving several
critical steps.
The casting solvent used plays a very critical role in the curing of
polyimide resins [141, 146].
The structure, morphology and the final cured film
properties are affected by the imidization kinetics, side reactions and the solventprecursor interactions.
All o f these factors are in-tum affected significantly by the
processing conditions.
The so 1vent-precursor (polyamic acid or polyamic ester)
interactions are well studied and reported [143, 146, 147]. The as-received resin PI 2611
is available as a 13 to 14 wt.% solids mixture in n-methyl pyrrolidone (NMP) solvent.
The polar aprotic solvent (NMP) shows a strong tendency to complex/hydrogen bond to
both the acid as well as the amide group (schematically illustrated in Figure 56) o f the
polyamic acid (PA) or ester precursor. The different stages o f the cure reaction include
initial decomplexation o f the solvent from these groups (which helps plasticize the
precursor polyamic acid film and facilitates the cyclic ring-closing reaction which is
208
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
highly hindered by entropic limitations) followed by imidization and finally solvent
evolution from the film.
For conventional thermal curing, solvent decomplexation is essential for significant
imidization to occur and is key to lowering the imidization temperature. Figure 56 shows
the DSC scans o f as-received PI 2611 samples ramped at a rate o f 5, 10 and 15°C/min to
400°C. As seen in the Figure 56, the heat flow scans show a distinct dependence on the
ramp rate used. At very low ramp rates (curve (a) 5°C/min) the solvent decomplexation
(and evaporation) and imidization reaction occur simultaneously showing a single
endothermic peak between 50°C and 200°C. Increasing ramp rates (curves b and c)
increase the peak decomplexation temperature and also the peak imidization (and
completion) temperature. The ramp rate used in the curing step thus determines the
relative rates o f solvent decomplexation and imidization.
209
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0
-20
-40
i ______
-80
C—OH
-100
0
300
200
100
400
Tem perature, °C
a )~
5'C/min
b) ~
10'C/min
c ) — 15°C/min
Figure 56 Ramp rate dependence o f solvent decomplexation and imidization a) 5°C/min,
b) 10°C/min, and c) 15"C/min. (Inset: schematic representation o f solvent -amide
complexation mechanism.
High ramp rates cause faster evolution o f NMP from the film, and the reduced
plasticization effect leads to delayed imidization. Hence, in conventional thermal curing,
low ramp rates are typically used to increase the solvent-induced plasticization and to
promote imidization. Further, even though the imidization reaction by itself is completed
[45] at relatively low temperatures (~ 240°C) at low ramp rates, the films are cured at a
210
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
much higher final cure temperature of 3S0°C to drive off the residual solvent from the
system.
For microwave-processed films, the existence o f the high local temperatures changes
the kinetics o f solvent decomplexation and imidization. Rapid decomplexation o f the
solvent at a low bulk temperature (low applied power) increases the plasticization effect
promoting rapid imidization.
At higher temperatures (higher applied power), the
imidization is completed and solvent evolution takes place.
However, too rapid
imidization could possibly trap solvent as it becomes increasingly difficult for solvent to
diffuse out o f a highly rigid imidized structure. In Chapter V it was shown that solvent
can be preferentially removed from the film by VFM curing due to selective microwave
absorption by the solvent. When processing on organic substrates, the impact o f heating
rates on the thermal behavior of the substrate should also be factored in. An optimum
VFM ramp rate should not only effectively balance solvent decomplexation,
plasticization and imidization but also ensure the substrate integrity.
7.1.6 Solvent effect on VFM heating characteristics
As seen in Figure 45, the heating rates o f the FR4 board with and without the
polyimide film coated on it are significantly different. It is important to understand the
effect o f solvent content on detennining the effective heating rate. Several samples with
varying solvent content were coated on blank FR4 substrates and the relative heating
rates o f the polyimide film and the top surface o f the FR4 substrate were characterized.
The temperature was measured in all cases using a fiber-optic probe. For each o f these
studies, the sample size and film thickness were kept constant.
211
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The heating
characteristics at a constant microwave power o f 400 W and the set up used for these
measurements are shown in Figure 57.
The curves denoted (1) correspond to the
polyimide film temperature; curves denoted (2) correspond to the FR4 temperature
outside the polyimide film (masked during the coating process)
As seen in Figure 57 (b), the as-received polyimide film coated on FR4 substrate
(before softbake) shows a significant absorption o f microwave energy initially (also seen
in Figure 45) leading to a very high ramp rate. The temperature rise in the FR4 board is
higher with the polyimide film coated on it as compared to that without the film (curve
labeled 3 in Figure 57 (b)) i.e., although the volumetric heat generation rate for the same
sample size is the same, the heating rate with the polyimide film is higher due to
increased heat transfer (conduction) from the polyimide film. It was not possible to
measure the temperature o f the FR4 bulk or at either interfaces due to the lack o f a nonintrusive temperature sensor. However, it is likely that the temperature o f the FR4 at the
polyimide interface is much higher than represented by curve (2) (as predicted by Glinski
et al [138]). Figure 57 (c) shows that the sofibaked film heats to a lower temperature
(120°C) than the as-received film (190°C) for the same time at 400 W. The temperature
rise o f FR4 also drops concurrently. Figure 57 (d) shows the heating rate o f a film
softbaked and dried in a vacuum oven for 24 hours at room temperature. A further
decrease in the heating rate o f both the polyimide film and FR4 is observed. In this case,
the temperature rise comes mostly due to heating in the FR4 substrate. The dielectric loss
in the uncured polyamic acid (without a significant amount o f solvent) is not high enough
to cause temperature rise in the film.
212
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
i a#
175
Polyimide
£
Fiber optic
125
— ri
F*4
i nu»
Quartz
3a
M
9#
Time, sec
12a
15#
ISO
Temperature,
a) Measurement set-up
150
15a
120
120
€»
e
SL
90
<a
—n
FE4
30
—
pi
$
o
120
150
ISO
Time, sec
30
M
90
120
ISO
ISO
Time, sec
d) Softbaked and dried
c) Softbaked
Figure 57 Effect o f solvent content on the heating rates o f polyimide film coated on FR4
substrate subjected to a constant power o f 400 W (1): Temperature o f polyimide film, (2)
Temperature o f FR4 board (3) temperature o f FR4 without polyimide film.
213
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
These results suggest that the casting solvent contributes significantly to microwave
heating and depending on the processing conditions used, the dielectric loss in the film or
the substrate contribute significantly to the heating. The extent o f imidization achieved
under these conditions was also studied.
Table 16 shows the percent imidization
achieved and residual solvent (approximately estimated by measuring mass using a
microbalance) as a function o f processing conditions. The as-received polyamic acid
comes as 13.7 % by weight solids in NMP. The solvent content in the film was varied by
vacuum drying. The percent imidization was estimated by FTIR analysis. The weight
percent solvent in the film (based on total mass o f film) was estimated from the measured
sample mass also accounting for 2 moles o f water per repeat unit o f polyamic acid lost
during condensation.
As seen in Figure 57, and Table 16, a vacuum dried polyimide film does not heat
significantly (even at an applied power o f 500 W, the maximum forward power on the
system). The measured surface temperature o f the film reaches less than 125°C after 15
min at 500 W. However, the percent imidization is significantly high (88.5%) for the
measured low temperature o f 125°C.
A further 10 min at 500 W advances the
imidization in the film to ~ 90%, while the maximum temperature reached at the end of
10 min was - 120°C. However, a similar vacuum dried sample cured in a thermal oven
for one hour at 150°C results in less than 10% imidization. The significantly higher
extent o f imidization achieved in low temperature microwave processed films shows the
effectiveness o f local heating in enhancing the imide-ring closure.
In conventional
thermal processing, the lack o f plasticization due to reduced solvent content hinders the
214
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
imidization reaction especially at low temperature due to reduced mobility. The high
local temperatures in VFM processed films enhance the mobility and diffusion o f the
residual solvent promoting imidization.
Table 16 Percent imidization and residual solvent in VFM cured PI 2611.
ID
Description
-
W0
PA + Solvent
0
-
w,
W0 + 24 hrs in
Vacuum oven at 100°C
0
19.6
VFM
W2
W, + IS min at 500W
(T < I25°C)
88.5
17.9
VFM
W3
W2 + 10 min at 500 W
(T < 120°C)
90.2
14.6
Thermal
W4
W, + 60 min at 150°C
ramped at 3°C/min
«10%
% Imidization Solvent W t%
|PA: Polyamic add
However, as seen from Table 16, the residual solvent in the cured films was still very
high. Longer times at 500 W tended to degrade the substrate due to increased loss in the
substrate. The evolution o f solvent could be limited by the maximum available forward
power. Short times at higher powers could potentially overcome this limitation. Similar
solvents studies also showed that greater the initial solvent content in the film, greater the
215
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
dissipation in the film and larger the temperature gradient across the substrate. High
initial powers were however detrimental on two accounts: 1) high initial ramp rates
caused bubbling o f the film due to rapid evaporation o f solvent and 2) early evaporation
o f solvent also slightly reduced the extent o f imidization for a given time and usually
resulted in increased substrate contribution to heating due to insufficient microwave
absorption within the dielectric film. This was also the case with slow ramp rates in
VFM processed films, where solvent evaporation caused increased dissipation in the
substrate. Furthermore, results from these studies showed that only a maximum cure
temperature of200°C (~ 15 min at 200°) could be used with FR4 substrate, which was not
adequate to achieve acceptable residual solvent and cured film properties. The low Tg o f
FR4 allowed
for greater mobility and delamination
decomposition/degradation
o f the
substrate.
Although
failure occurred before
results
from
thermal
characterization o f the substrates showed a higher thermal stability for FR4, low Tg and
increased dissipation due to the substrate-polyimide film interaction in the VFM limited
the maximum cure temperature to 200°C. Solvent evaporation from the highly rigid
imidized polyimide film was thus eventually limited by the maximum bulk substrate
temperature. Hence, BT substrates with a higher Tg and thermal stability (Figure 44)
were used to further decrease the residual solvent and improve the properties o f cured
films.
Indeed, PI 2611 films cured on BT substrates showed significant improvement in
properties over films cured on FR4 substrates. An approach similar to FR4 substrates
was followed in determining the optimum cure condition for curing PI 2611 on BT
216
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
substrates.
It was found that BT substrates were relatively stable when ramped to
temperatures as high as 260°C at 15°C/min. However, even a 30 sec hold at 260°C
caused significant delamination in the BT substrates. TGA results showed less than 4 %
by weight residual solvent in PI 2611 films cured on BT substrate at 220°C for Smin. To
decrease the residual solvent further, a cure schedule with a rapid ramp to 260°C followed
by a low-temperature hold at 200°C was developed and films cured under this condition
showed moderate improvement in the TGA results. The properties o f PI 2611 film cured
under this condition are summarized in Table 17 and compared to films cured on FR4.
As seen in Table 17, PI 2611 films cured on BT substrates show improved properties
primarily due to reduced solvent content.
Table 17 Comparison o f properties o f PI 2611 films cured on FR4 and BT substrates.
PI 2611 on FR4
PI 2611 on BT
15°C/min / 200°C / 5
min
15°C/min / 260°C
.....220°C / 5 min
Birefringence
0.0611*
0.1661
Tg, °C (TMA)
220
273
CTE, ppm/°C
12
6.8
Storage modulus, GPa
4
N/A
Young's Modulus, GPa
4.52
5.44
% Wt. loss (10°C/min
ramp to 400°C in TGA)
10.3
3.6
Cure Condition
Property
* IS min cure
217
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Similar results were found with photosensitive polyimides HD 4000 and XP 7001.
The photosensitive polyimides showed a greater weight loss in the TGA when cured
under the same cure conditions largely due to the high molecular weight acrylate-ester
based photo package, consistent with the results seen in Chapter V on low-temperature
cure studies on silicon. Figure 58 compares the thermal stability o f PI 2611, HD 4000
and XP 7001 films. Each of these films was VFM cured on BT substrates for 220°C for 5
min. As seen in Figure 58, as expected XP 7001 shows a greater T$% than HD 4000,
consistent with earlier observations that a lower Tg and a more flexible backbone
facilitates photoproduct evolution form the film during the cure process.
From these results it is evident that low temperature VFM curing o f polyimides on
organic substrates is feasible allowing processes otherwise not possible by conventional
thermal processing.
Residual solvent was found to be the most significant factor
affecting VFM heating characteristics and also the resulting final film properties.
Significant improvement in the properties o f cured films was observed as compared to
low-temperature thermally processed films. The lower solvent content in the microwavecured films as compared to films thermally cured at the same temperature is due to the
preferential microwave energy absorption by the high loss solvent and the polyamic acid
favoring both imidization kinetics as well as solvent evolution from the film. However,
properties as good as high temperature standard thermal cured films could not be
achieved. This is attributed to the higher residual solvent in the VFM cured films as
compared to standard thermal cured films. Although local heating in the VFM enhances
218
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
solvent evaporation as compared to low-temperature thermal curing, complete removal o f
the solvent is eventually limited substrate heating and by the effective bulk temperature.
100
M
W
(a): PI 2611
(b): HD 4000
(c): XP 7001
50
100
150
200
250
300
350
400
Temperature, °C
Figure 58 Comparison o f thermal stability o f PI 2611, HD 4000 and XP 7001 films cured
on BT substrate. Cure conditions: l5°C/min to 220°C, 5 min hold. TGA ramp rate:
10°C/min.
219
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7.2 Conclusions
The feasibility o f curing high performance polyimides on low-cost organic substrates
was investigated.
For the first time, rapid curing o f polyimides on low-temperature
substrates without degradation of the epoxy has been demonstrated.
VFM heating
characteristics and thermal stability o f FR4, copper-laminated FR4, CF-epoxy composite
and BT substrates were studied. Polyimides, PI 2611, PI2734, HD 4000 and XP 7001
were cured on these organic substrates by VFM processing. Infrared studies show that
there is no significant difference in the chemical structure o f VFM films cured on organic
substrates as compared to films cured in a conventional convective furnace (at the same
temperature) and nearly complete imidization can be achieved by curing the polyimide
films for much shorter cure times and at lower temperatures.
The critical factors
affecting low temperature polyimide curing were identified and studied. Residual solvent
was found to be the most significant factor affecting VFM heating characteristics and
also the resulting cured film properties. Significant improvement in the properties o f
cured films was observed as compared to low temperature thermally processed films.
TGA data showed high residual solvent in films VFM cured on FR4 substrates. BT
substrates allowed polyimide films to be cured at higher temperatures and accordingly
showed reduced residual solvent as compared to films cured on FR4 substrates. The
optical, electrical and thermo-mechanical properties o f the cured films were characterized
and compared to thermally cured films. These studies showed that although high local
220
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
temperatures enhance solvent evaporation, the effective bulk temperature eventually
limits complete removal o f solvent from the films.
221
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER VIII
SOLDER REFLOW AND ELECTRICALLY CONDUCTIVE ADHESIVE CURING
BY VARIABLE FREQUENCY MICROWAVE PROCESSING
This chapter contains results from solder reflow and VFM curing o f electrically
conductive adhesives (ECAs). In this study, the feasibility o f selective heating o f ECAs
in a microwave field was investigated. A limited number o f ECA samples were studied
with an emphasis on developing a qualitative understanding on the factors contributing to
heating o f the ECA in the microwave field, the effect o f filler size, loading and
distribution.
The feasibility of achieving selective heating and die-attachment using
VFM curing o f ECAs was investigated. Reflow o f solder paste and electroplated solder
by VFM processing was also studied.
8.1 Results and Discussion
8.1.1 Electrically conductive adhesive curing by VFM processing
The details o f the actual ECA formulation are provided elsewhere [148]. In this
study, the ECA composition is reported as a percent loading o f silver flakes in an epoxy
formulation. A typical epoxy formulation consisted o f 5 to 6% by weight o f catalyst.
The base epoxy resin used in the formulation was Epon 862 from Shell Chemical
222
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Company, which is based on diglycidyl ether o f bisphenol-F (DGEBF) chemistry. An
imidazole
derivative,
2E4MZ-CN
(l-cyanoethyl-2-ethyl-4-methylimidazole)
from
Shikoku Chemicals was used as the catalyst. Silver flakes o f three different sizes were
obtained form Degussa Corporation with average particle size (dp) as follows. A: dp < 6
pm (1.9 to 5.5 pm), B: dp < 2 pm (0.8 to 2.0 pm) and C: dp < 0.8 pm (0.5 to 0.8 pm). In
all the samples studied, the same epoxy formulation was used only differing in the type
and percent filler loading.
In order to study the microwave heating characteristics o f the ECAs, ~ I mil thick
samples were stenciled onto a relatively microwave-inert (non-absorbing) substrates.
Typically, an alumina substrate was used. Samples o f uniform size and thickness were
used for all these studies. Temperature measurements were made using a calibrated
infrared pyrometer. A central frequency o f 6.425 GHz was used for all these studies with
a full bandwidth o f 1.15 GHz swept within a 0.1 sec time frame. Figure 59 shows the
rise in temperature o f an ECA sample with 70% filler loading (type A) compared to that
o f a blank alumina substrate without the ECA, at a constant microwave power o f 200 W.
It is seen that the ECA can be heated rapidly in the VFM.
No arcing or charging
problems were encountered during this process due to frequency sweeping. Also, as seen
in Figure 59, the ECA heats at a much faster rate than the alumina substrate. Further, the
maximum temperature reached by alumina after 3 min is much lower than the ECA
sample i.e., the ECA can be selectively heated over an alumina substrate by microwave
heating. This is expected, as alumina is known to have an extremely low dielectric loss at
microwave frequencies [31, 149]. In the case o f the ECA, microwave absorption comes
223
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
from the dielectric loss in the uncured B-staged epoxy resin, the hardener/catalyst and
also due to dissipation in the metallic silver flakes.
Thick blanket metals reflect most o f the microwave energy due to their large
conductivity. However, thin metal inclusions can absorb energy due to their interaction
with the applied microwave field forming alternating currents at the surface.
This
phenomenon is particularly important when the dimensions o f the metal particle are o f
the order o f the skin depth o f the material. The skin depth (5), the distance by which the
electric field amplitude falls to e~l of its value at the material surface, is governed by the
conductivity o f the material and is given by Equation 8.1.
S
=
- i- --
-----
(8.1)
where / is the applied frequency, a is the conductivity o f the material and p is the
magnetic permeability.
224
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
160
140
120
100
80
60
£
40
20
0
0
30
60
-Alumina Substrate
90
T im e, sec
120
150
180
•ECA 70% filler loading (Type A)
Figure 59 VFM heating o f ECA with 70% loading o f silver flakes (type A: < 10 pm) at a
constant microwave power o f 200 W.
The skin depth of some commonly used metals in microelectronics calculated from
Equation 2 is plotted as a function o f frequency in Figure 60. The skin depth o f most
metals is in the range 100 nm to 100 pm in the frequency range 10 MHz to 1 THz, and
decreases with increasing frequency.
The skin effect becomes significant when the
dimensions of the particles are o f the order o f the skin depth. Also, from Figure 60, it can
be seen that silver has a skin depth o f ~ 1 pm at a frequency o f 6 GHz, the microwave
frequency used for the processing in this study.
225
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
—♦ " Cu
---A- ••Au
Sn
Pb
0.1
0.01
0.1
1
10
Frequency, G Hz
“ *»100
1000
Figure 60 Skin depth o f some common metals as a function o f frequency.
The effectiveness o f microwave coupling to metallic particles is also dependent on
the shape, size and geometry o f the particle since the effective conductivity o f the
metallic particles is dependent on these factors. Further, the surface area available also
strongly influences the microwave absorption characteristics [76]. For dimensions o f the
order o f the skin depth, alternating currents are set up along the whole length o f the metal
particle leading to localized dissipation. When the filler loading is so high as to form a
continuous path for transfer o f charge carriers (percolation threshold), eddy currents may
226
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
be induced in the metallic fillers. The microwave energy thus absorbed at the surface in
either case generates heat due to resistive dissipation (Joule heating). Analogous to the
power absorbed/dissipated by dielectric materials in an applied field, the power dissipated
per unit volume ( P ) by a conducting material may be expressed as shown in Equation
8 .2 .
P = *
where
( 8.2 )
is the root mean square electric field strength, and <r* is the effective
conductivity o f the material. As with the dielectric materials, the resulting temperature
profile is dependent on the percent loading, distribution and thermal characteristics o f the
filler-resin matrix [150].
Figure 61 shows the effect o f percent filler loading on the VFM heating
characteristics o f ECAs at a constant microwave power o f 200 W. An increase in filler
loading reduced the VFM heating rate and the final temperature reached. Similar trends
were seen for different filler types, that is, reduced heating rates with increasing filler
loading. This trend was also observed by Liu et al. [78] on studies with commercially
available ECAs with increasing amounts o f silver added. A direct comparison o f the
results is not rational as differences in ECA composition and sample size can
significantly change the amount o f power absorbed and the resulting heat dissipated.
227
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Nevertheless, both these studies illustrate the same effect o f increasing filler loading on
VFM heating characteristics. The decrease in the heating rate and final temperature
reached by the ECAs could be due to 1) a decrease in the amount o f microwave energy
absorbed and dissipated and 2) an increase in the amount o f thermal energy transported
away from the sample.
200
150
100
S
£
50
0
0
10
■70%
20
80%
30
Time, sec
-85%
40
50
60
-O- Blank Alumina Substrate
Figure 61 Effect o f filler loading on the VFM heating characteristics o f ECA at a constant
microwave power o f 200W.
228
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
An increase in the filler loading amounts to a corresponding reduction in the epoxy
concentration and if the epoxy base resin contributes significantly to the heating (as
postulated by Liu et al [78] for filler loading in the range 70 to 85%), the final
temperature reached would be expected to decrease with increasing filler loading. This
was also evident comparing the heating rates o f a cured and an uncured sample. For
example, the final temperature reached by a cured ECA sample with 80% filler loading
was 30°C lower than the final temperature reached by an uncured sample o f the size with
the same filler type and loading when exposed to the same microwave power for the
same time.
An increase in the filler concentration also increases the effective thermal
conductivity o f the ECA sample, which increases the heat transferred to the alumina
substrate resulting in a lower bulk sample temperature for the same applied power.
Moreover, an increase in the filler concentration could reduce the microwave absorption
efficiency due to increased reflection o f microwaves especially above the percolation
threshold as the skin effect become relatively insignificant for larger dimensions as
discussed earlier. Indeed, this effect was also seen by studying the effect o f particle size
on the microwave heating characteristics o f ECA samples with the same filler loading.
Figure 62 compares the heating rates o f ECA formulated with the same filler loading but
with different particle sizes (A < 6 pm, B < 2 pm, and C < 0.8 pm). As seen from Figure
62, the heating rates are higher for filler (C) with dimensions o f the order o f skin depth (~
I pm for silver at 6 GHz).
229
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
90
A A *
A A
80
70
sf
s
2m\ 60
****
♦
A
AA A A
A *
A
♦
w
a.
E 50
i t ; * : - - '
♦ A (< 6 um)
40
■ B (< 2 am)
30
* C ( < 0 i am)
20
30
60
90
120
150
180
Time, sec
Figure 62 Effect o f particle size on the VFM heating characteristics o f ECAs.
Bimodal filler distributions are often used to improve the effective conductivity of
the ECAs. The VFM heating characteristics o f these bimodal distributions were also
studied. Figure 63 shows the effect o f microwave power on the heating rates o f three
different ECA formulations with the same percent loading (80%). For example ECA ‘AB’ has 80% by weight o f silver flakes with equal percentage by weight (1:1) o f A and B
dispersed in the epoxy. As seen in Figure 63, the ECA A-C has the highest VFM heating
rate. Bulk resistivity measurements showed that ECA sample B-C was the most
conductive and sample A-C was the least conductive. Typically, the ECA sample with
the higher electrical conductivity also has the higher thermal conductivity. While, the
230
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
energy dissipation in the microwave field increases with conductivity, greater heat
removal from the ECA sample leads to a lower temperature.
Since the microwave
dissipation o f the epoxy is more dominant, for all the samples in Figure 63 with the
sample percent filler loading, the sample with the highest thermal conductivity will reach
the lowest maximum temperature for a given power. The thermal profiles in Figure 63
thus correlate well with the conductivity o f the samples. Indeed, increasing the percent
filler loading to 85% for the three formulations shown in Figure 63 further lowered the
maximum temperature reached by the ECA with the most significant reduction being
observed for ECA A-C.
231
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
140
120
44*“ “
C
0
30
60
90
120
: < OJJ fun
150
180
Time, sec
I
♦ a -b
■ B-C
* A-C
Figure 63 VFM heating rates o f ECAs with bimodal filler distribution.
The standard thermal cure for all these formulation was a 30 min cure at 150°C.
Rapid curing o f these ECAs could be achieved by a 5 min hold at 150°C.
The
effectiveness o f curing was verified by a low stable resistance in test structures.
Depending on the specific ECA formulation, an appropriate forward microwave power
was utilized. In some cases, samples with a high filler loading (^85% ) and large filler
size (A and B) could not be heated to temperatures high enough to cure the ECA within
the maximum available forward power o f 500 W. The dependence o f heating rate on
applied field is evident from Equations 2.11 and 8.1. Central frequency, bandwidth and
sweep rate did not show any significant difference in the heating rates.
232
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The feasibility o f die-attachment using an ECA by VFM processing was also
investigated. For studying die-attachment, a coarse pitch stencil mask was used to apply
the ECA onto pads defined on silicon and on a glass substrate. A simple daisy chain
attachment test structure (shown in Figure 64) was designed for pads on silicon and the
glass substrate. About 1 mil thick ECA was dispensed on the glass substrate. The silicon
die was manually aligned with the substrate pads (with the ECA) and the ECA was then
cured in the VFM. A constant VFM power o f 200 W for ~ 2 min was used for the dieattachment.
Substrate side
Top V iew
Silicon die side
Bottom V iew
(A)
(B)
Figure 64 Test structures for ECA cure and attachment: A) mask features and B) top and
bottom view o f silicon die attached to glass substrate by VFM curing o f ECA at 200 W
for 2 min.
233
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In the VFM, the glass substrate is relatively non-absorbing and heating comes
primarily from microwave absorption of the silicon die and the ECA. The ECA heats
faster (to a higher temperature) at a lower power due to increased heat flux from the
absorption in the silicon die. The thermal characteristics o f the substrate thus determine
how much o f the generated heat is removed away from the ECA. Actual temperature
measurements at the die-ECA or the ECA-substrate interface are not feasible due to the
lack o f a non-intrusive temperature sensor. However, electromagnetic finite element
modeling has shown [138] that the substrate dielectric and thermal characteristics can
significantly influence whether or not the ECA can be cured at a given time and
microwave power. Under the conditions investigated in this study, the ECA reached a
high enough temperature to achieve complete curing and successful attachment o f the die
to the substrate.
However, during this ECA cine, the measured maximum bulk
temperature o f the glass substrate away from the die was ~ 120°C. This demonstrates the
feasibility o f achieving selective heating by VFM processing for ECA curing and
attachment o f components.
The results from these studies show that rapid selective heating o f ECAs can be
achieved by VFM processing without arcing. Some o f the important factors affecting the
microwave heating rates o f conductive adhesives were identified.
The transient
temperature profile o f the ECA is determined by a balance o f the rate o f heat generation
and dissipation to the surroundings. The heat generation rate is governed by the power
dissipated in the ECA, which is a function o f the dielectric loss characteristics o f the
epoxy and the conductivity o f the silver flakes and their dependence on frequency.
234
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Increasing the filler loading reduced the microwave heating rate o f the adhesive. The
effectiveness o f coupling microwave energy to ECAs was more efficient when the size o f
the metal inclusions in the adhesive were o f the order o f the skin depth o f the material.
8.1.2 Solder reflow by VFM processing
The ability to selectively and rapidly process materials by microwave processing
offers some distinct advantages over conventional thermal processing in the packaging
and assembly o f electronic devices. The feasibility o f solder-reflow and attachment from
solder paste and electroplated solder by VFM processing was investigated. Processing o f
solder paste is similar in principle to the processing o f ECAs except that the solder paste
does not have any major organic constituent, i.e., it has about 90% metal content. Solder
paste is a mixture o f pre-alloyed solder particles formulated with a number o f other
constituents for stability and processability. Typical solder constituents include a flux (to
facilitate oxide cleaning and promote wetting), solvents, diluents and activators.
A
commercially available near eutectic (63% Sn and 37% Pb) solder paste from Indium Co.
was used to investigate the feasibility o f reflow and attachment o f solder paste.
The reflow profile for a typical bumping or attachment process in a conventional
thermal reflow oven is shown in Figure 65.
235
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
300
Preheat
250
. Flux
f activatioa/
Soak
Reflow
100
200
Cooling
2
2
150
la
at
a.
S
100
£
0
50
150
250
300
350
Time, sec
Figure 65 A Typical reflow profile used for Sn-Pb solder alloys in a convection reflow
oven.
The samples/components pass through four different zones the time in each zone
being set by the belt speed.
It involves a pre-heat zone, where the components are
ramped at an initial ramp rate o f 2 to 3°C to a dwell/soak zone (~ 140 to 160°C) for about
30 to 60 seconds where flux activation begins. The components are then ramped to a
temperature above the liquidus temperature (183°C) in the reflow zone and then allowed
to cool. A reflow profile similar to the standard convection reflow oven was developed
for the VFM process.
236
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
For solder paste reflow and attachment, a coarse stencil mask was used to apply the
paste onto an array o f gold pads defined on silicon by photolithography. A constant
VFM power o f 200 W was applied for~ 3 to 4 min.
Under these conditions, the
microwave absorption o f silicon and the solder paste was high enough to cause the solder
paste to reach the liquidus temperature and reflow to occur. A similar dependence on
applied forward power as with the ECAs was observed with the solder paste. However,
very high initial applied power led to solder splatter and this effect also seemed to depend
on the effectiveness o f stencil printing.
For studying the solder reflow o f fine pitch features for bumping and attachment,
electroplating technique was used to form solder bumps.
A typical solder bumping
process is shown in Figure 66. In this study, a fluoboric acid based solder plating bath
was developed [151] to electroplate near eutectic solder on metal pads defined on silicon.
The bath composition was as follows: (500 mL water, 140 mL fluoboric acid, 15 gm
PbO, 70 mL Sn(BF4)2, ~ 10 mL triton dissolved in 5 mL H2O and ~ 0.13 gm
Phenolphthalein in 5 mL ethanol). Alternatively, an organic acid based commercially
available solder plating solution, Techni Solder Matte NF 820 HS from Technic Inc., was
also used. As seen in Figure 66, a plating seed layer and solder mask resist are required
before electroplating.
A Ti-Cu-Ti seed layer (150“A-2500"A-150°A) was sputter
deposited and a 16 to 30 pm thick resist layer was used as a solder mask. A positive
resist AZ 4620 as well as a negative resist NR9-8000 were used for this purpose and both
were found to show good etch/dissolution resistance in the plating bath.
Pads for
electroplating were defined by photolithography o f the resist layer. The mask consisted
237
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
o f 60 |im solder pad features with a 330 pm pitch on a I cm2 area die. The titanium layer
in the open areas was etched away to expose the copper layer on the pads before plating.
A current density o f - 15 mA/cm2 was used for electroplating. Uniform plating across a
4-inch wafer could be achieved with both the plating solutions.
l.D ie pad
4.Electroplate solder
2.Plate seed layer
(UBM)
3.Pattern thick
resist/solder mask
5. Strip resist
Etch seed layer
6. Reflow solder
Figure 66 Typical bumping process for electroplated solder.
Scanning electron microscope (SEM) micrographs o f the solder bumps after
electroplating and stripping the solder mask resist (step S in Figure 66) are shown in
Figure 67. The ‘mushroom’ like structures result from plating above the thickness o f the
solder mask and are desirable for achieving good bump height after reflow. The liquidus
238
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
temperature o f Sn-Pb alloys is dependent on the alloy composition.
The reflow
temperature in the VFM was set based on the composition o f the plated solder bumps.
The composition o f the electroplated solder bumps was analyzed using a Noran energy
dispersive spectrometer (EDS). A typical EDS spectrum o f a plated solder bump is
shown in Figure 68.
239
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
240
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Sn
6000
5000
4000
3000
Pb
Pb
2000-
1000
keV
E 3 T T O W T T M I al'i I] I
Sn
Pb
I La
Ma
I 827.9 I 2.63 I 70.82 I 80.9
I 295.3 I
1.57 I 29.18 I
19.1
Figure 68 A typical EDS spectrum o f an electroplated Sn/Pb solder bump.
241
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The plating seed layer was then etched away and a no clean flux H208-X4-E from
Indium Co. was dispensed on the plated bumps prior to reflow in the VFM. The powertime-temperature window to reach the required liquidus temperature for a given die size
were determined and a reflow profile based on the composition o f plated solder was
developed for the VFM process. In this case, VFM heating comes primarily from the
silicon die. The set temperature and the actual measured temperature profile and the
forward microwave power used during this process are shown in Figure 69. Uniform
reflow o f solder without any significant bridging or other defects could be achieved in the
VFM using this reflow profile. Figure 70 shows SEM micrographs o f VFM reflowed
solder bumps. The collar surrounding the base o f the solder bumps after VFM reflow
was from the no-clean flux used in this study and could be removed completely in
organic solvents. As seen in Figure 69, the maximum power used during this process did
not exceed 300 W. From earlier results on the VFM heating characteristics o f organic
substrates, it was shown that under these conditions (power and time), the bulk
temperature o f the organic substrate is less than 90°C. Glinski et al. also demonstrated
this effect, by simulating the microwave induced dissipation and the resulting heat
transfer in a silicon die-attachment process on an FR4 substrate using PHYSICA, an
electromagnetic simulation software [138]. This selective low temperature attachment
process can significantly reduce the strain on the solder bumps thereby improving the
yield during the assembly/attachment process.
242
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
500
300
Cooling
400
200
300
150
200
100
100
0
30
60
90
120 150 180
Tim e, sec
210
Set temperature, °C ——“Measured temperature, °C
240
270
300
Forward Power, W
Figure 69 VFM temperature and power profile for solder reflow.
243
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Power, W
Temperature, °C
250
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission
Flip chip on organic substrate technology offers several benefits over other
packaging technologies due to its lower cost, high I/O density and superior electrical
performance. However, the difference in the CTE o f the chip carrier (FR4 substrate ~ 17
ppm/°C) and the silicon die (CTE ~ 3 ppm /°C) results in poor joint reliability. An
underfill is typically used in current flip-chip devices to reduce the CTE-mismatch
induced thermo-mechanical strain between the silicon die, solder and the organic
substrate and increase the fatigue life o f such devices. Higher I/O counts (> 10,000:
require long underfill dispense times) and increased requirements on the electrical
performance in future devices prohibit the use o f underfill, leading to direct flip chip on
board attachment with compliant interconnects [152].
The assembly yield o f such
structures can be enhanced by a low temperature attachment process.
The shear strain on the outermost C4 solder ball in a flip chip device without an
underfill is given by Equation 8.2 [20].
Shear strain m th no underfill
=
DNP * ^ C r£ ^
X-A7^ ) - ■■
Height o f Bump
AT^)]
(8.2)
where A7^ ,^ is the temperature excursion o f the chip carrier or the substrate and A7s is
the temperature excursion o f the silicon die and DNP is the distance from the center o f
245
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the die or the distance from neutral point.
In a traditional convection reflow oven
assembly process both the silicon die and the substrate undergo the same thermal
excursion. However, as demonstrated earlier, in the case o f VFM reflow, the bulk o f the
substrate undergoes a relatively lower temperature excursion.
Table 18 lists the strain under different bump configurations. As seen from Table 1,
the trend in reducing the solder pitch and bump height (and the transition to lead-free
solder which increases the AT in future devices A->D) increases the strain on the solder
bump. This effect is further exacerbated by higher values o f A T . There are two ways to
address this issue: 1) using a low CTE substrate such as alumina (E) or 2) a low
temperature process (F). Organic substrates are more desirable as they are lighter and
allow higher packing densities at a lower cost. The low temperature VFM reflow process
developed thus provides a unique way o f achieving lower strain in the solder during the
attachment on an organic substrate.
Table 18 Shear strain in solder ball without underfill.
Shear Strain
without underfill
CTE Carrier
Bump height
AT Chip
AT Carrier
A
B
C
D
E
F
20
40
64
129
37
40
17
140
200
200
17
70
200
200
17
50
230
230
17
25
230
230
7
25
230
230
17
25
230
100
246
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The effect of low-temperature VFM processing on the strain in solder bumps for
chip on board attachment by eutectic solder was also verified by finite element analysis
using ANSYS software [153]. Some of the model parameters chosen include: die-size: 1
cm square; bump diameter. 100 mm, bump pitch = 350 mm, bump height = 75 mm).
Figure 71 shows a schematic o f the test structure modeled and the strain contours o f
the solder bumps under two reflow conditions: A) Traditional convection reflow and B)
VFM reflow process described earlier. The FR4 substrate was modeled as consisting of
two layers. In case (A) both the layers were modeled as being at the same stress relief
temperature as the solder 183°C. In the case o f VFM processing, during the solder reflow
cycle such as the one shown in Figure 69, the microwave forward power required (for the
silicon die and the solder to reach the reflow temperature) does not exceed 200 W until 3
minutes. A forward power in the range 200 to 500 W is further required for an additional
half-a minute. From Figure 41, it is evident that the temperature o f a blank FR4 substrate
does not exceed 90°C even after 3 minutes at 500 W. However, heat flux from the silicon
die and solder could increase the temperature o f the FR4 substrate resulting in a
temperature gradient across the substrate thickness. The temperature distribution in the
FR4 substrate is determined, among other things, by the relative die-to-substrate size.
For simplicity o f analysis, in the case o f VFM reflow, (Figure 71, B), the top one-third of
the FR4, which was in close proximity to the solder and silicon, was modeled as being
close to the solder temperature o f 183°C and the remaining two-thirds was assumed to be
at a bulk temperature o f 120°C, the maximum temperature reached by the bulk o f the
substrate in the VFM under these conditions. As seen from Figure 71, the strain in the
247
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
solder bump increases with increasing distance from the neutral plane (DNP). This is
shown by the increasing deformation o f the bump and the strain shown by the scale
below the contours. Further, the total Von Mises strain in case B (the VFM solder
reflow), was ~ 0.172811 which is lower than that in case A (conventional thermal reflow
oven), which showed a strain o f ~ 0.181729. The reduced strain on the solder bumps
could potentially improve the assembly (attachment) yield o f such structures.
The
improvements in strain become more significant for smaller bump height and pad pitches
and for higher reflow temperatures as may be required for lead-free solder. VFM solder
reflow could result in potential cost, processing and performance advantages over current
convective reflow oven processes.
The ability to selectively achieve localized low
temperature attachment could be o f significant advantage especially for packages with
temperature sensitive circuitry.
248
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
.Silicon die
f
t. t
|C ,4t.
X X.,*' * X
I
FR4
Solder bump
Figure 71 ANSYS finite element modeling comparing the strain in A) convection oven
reflowed solder and B) VFM reflowed solder. (DNP: distance from neutral plane).
249
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
8.2 Summary
In this study, processing of metals and metal-filled composites in a VFM oven were
studied. VFM curing and attachment o f ECAs and Sn-Pb solder reflow was investigated.
The results demonstrate that rapid selective heating o f ECAs can be achieved by VFM
processing without any arcing issues encountered in conventional microwave ovens.
Some o f the important factors affecting the microwave heating rates o f conductive
adhesives were identified. The transient temperature profile o f the ECA is determined by
a balance o f the rate of heat generation and dissipation to the surroundings. The heat
generation rate is governed by the power dissipated in the ECA, which is a function of
the dielectric loss characteristics o f the epoxy and the conductivity o f the silver flakes and
their dependence on frequency.
Increasing the filler loading reduced the microwave
heating rate o f the adhesive. The coupling o f microwave energy to ECAs was more
effective when the size o f the conductive inclusions in the adhesive were o f the order o f
the skin depth o f the material. VFM solder reflow from solder paste and electroplated
solder was also investigated.
The power-time-temperature window to achieve the
required liquidus temperature for a given solder composition and die size were
determined. The results show that uniform solder reflow without any bridging or other
defects could be achieved by VFM processing. The impact o f low temperature, selective
reflow and attachment was modeled by finite element analysis using ANSYS. The
250
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
results show that VFM reflowed solder attachment results in a lower strain than
conventional reflow oven.
251
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER IX
SUMMARY AND FUTURE DIRECTIONS
9.1 Summary
In this study, VFM curing was investigated as a novel rapid, low-temperature curing
alternative to conventional thermal curing.
dielectrics
with
microelectronics
different
industry
backbone
such
as
Several commercially available polymer
chemistries
currently
benzocyclobutenes
o f interest
(BCB),
in
the
polyimides
and
polybenzoxazoles were chosen for this study. The results show that rapid curing o f these
dielectrics by VFM processing is feasible. A brief summary o f the results from this work
are presented in the following sections.
The thermal cure kinetics of Cyclotene BCB are well studied and can be easily
monitored by Fourier transform infrared spectroscopy. Hence this material was chosen
for VFM cure kinetics studies for a direct comparison to thermal curing. The chemical
structure o f VFM cured films was found to be similar to thermally processed films. Ring
opening followed by Diels-Alder crosslinking is the primary reaction mechanism for
VFM curing o f BCB. The results showed that the VFM cure reaction follows first order
kinetics. The rate constants showed an Arrhenius-type relationship with temperature and
252
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
an apparent activation energy o f 25.7 +/- 4.4 kCal/mol, which is about 30% lower than
the reported activation energy for conventional thermal curing. The optical, electrical,
mechanical and chemical properties o f VFM cured films were characterized and
compared with thermally cured films to determine the effectiveness o f VFM processing.
VFM cured films showed comparable or improved properties than thermally cured films.
The residual stress o f partially cured films (by VFM) was lower than that o f thermally
cured films.
To investigate the feasibility o f low-temperature curing o f polymer dielectrics on
silicon, five different commercially available polyimides with different backbone
chemistries were studied. The targeted cure temperature was 275°C. Structure-property
relationships and some general trends on the impact o f processing conditions on the
evolution o f mechanical and thermal properties were discussed. The thermal stability of
cured films was evaluated as a performance metric to determine the effectiveness o f lowtemperature microwave processing. The results showed that certain chemistries are more
suited than others for low-temperature microwave processing. Non-photosensitive rigid
rod backbone based PI 2611 showed a significant improvement in the thermal stability
over low-temperature thermally cured films and properties comparable to standard high
temperature cured films.
The negative tone polyimides did not show a marked
improvement in performance over low-temperature thermal curing possibly due to
microwave enhanced polymerization o f the photopackage. The positive tone polyimides
PWDC 1000 and PW 1200 did not show any significant improvement over lowtemperature thermal cures. The results from this study identified some o f the key VFM
253
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
processing issues and how they affect the cured film properties. The results indicate that
the backbone flexibility, structure and actual photochemistry o f the photopackages, and
the competition between imidization and potential crosslinking reactions are important
factors determining the effectiveness o f low-temperature microwave processing.
Rapid curing o f a positive tone photosensitive polybenzoxazole dielectric resin was
investigated by VFM processing. FTIR studies showed that the chemical structure of
VFM cured PBO films is identical to standard thermal cured films. These studies also
showed that rapid curing of PBO by VFM processing is feasible and significantly higher
conversion can be achieved by microwave processing at low temperatures as compared to
conventional thermal curing. The results showed that while higher percent conversion
and thermal stability than thermal curing can be achieved by VFM processing at lower
temperatures, complete removal o f photopackage related residual products requires
slower ramp rates and longer cure times.
Selective heating by microwave processing was studied by investigating the
feasibility of curing high performance polyimides on low-cost organic substrates. For the
first time, rapid curing o f polyimides on low-temperature substrates without degradation
o f the epoxy was demonstrated. Residual solvent was found to be the most significant
factor affecting VFM heating characteristics and also the resulting cured film properties.
Significant improvement in the properties o f VFM cured films was observed as compared
to low-temperature thermally processed films. TGA data however showed high residual
solvent in films VFM cured on FR4 substrates as compared to conventional high
temperature cured films.
BT substrates allowed a higher curing temperature and
254
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
consequently, polyimide films cured on these substrates showed reduced residual solvent.
The optical, electrical and thermo-mechanical properties o f the cured films were
characterized and compared to thermally cured films.
These studies showed that
although high local temperatures enhance solvent evaporation as compared to lowtemperature thermal curing, the complete removal o f solvent from the films is eventually
limited by the effective bulk temperature.
VFM curing and attachment o f ECAs and Sn-Pb solder reflow was investigated.
The results demonstrate that rapid selective heating o f ECAs can be achieved by VFM
processing without arcing issues encountered in conventional microwave ovens.
Increasing the filler loading reduced the microwave heating rate o f the adhesive. The
coupling o f microwave energy to ECAs was more effective when the size o f the
conductive inclusions in the adhesive were o f the order o f the skin depth o f the material.
Uniform solder reflow by VFM processing without any bridging or other defects was
demonstrated.
The impact o f low temperature, selective reflow and attachment was
modeled by finite element analysis using ANSYS.
The results showed that VFM
reflowed solder attachment results in a lower strain than conventional reflow oven.
9.2 Recommendations for future work
Microwave processing o f materials offers several advantages over conventional
thermal processing. The advent o f variable frequency microwave systems has opened
new areas o f applications for microwave material processing such as the microelectronics
255
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
industry. A good understanding o f the underlying principles and processes is essential to
realize the full potential o f this technology. Insights developed from this study will aid in
achieving a thorough understanding o f the structure-property relationships in microwavecured polymers in general, and polymer dielectrics for microelectronic applications in
particular.
Results from this study will also help design novel polymeric materials
specifically for microwave processing that can fully utilize the benefits o f selective
microwave-material interactions.
While significant improvements in the electric field uniformity and energy
distribution can be achieved in the VFM, the efficiency o f energy coupling is relatively
low.
The VFM furnace used in this study provides a method (known as ‘cavity
characterization’) to identify the frequencies at which microwave reflection is high.
These frequencies can be excluded during the cure cycle, in order to improve coupling
efficiency. Further studies are required to evaluate the effectiveness o f this technique in
improving the microwave processing efficiency.
Furthermore, microwave dielectric
characterization studies are necessary to understand the loss behavior o f materials as a
function o f frequency and temperature. Microwave dielectric measurements are often
complicated and require methods such as ‘Cavity Perturbation’ and the use o f T-resonator
structures. These measurements are also affected by sample form and geometry. These
studies in conjunction with molecular modeling approaches could bridge the gap in
understanding the exact relationship between molecular polarization mechanisms and
macroscopic dielectric behavior.
256
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Several improvements in the VFM system are essential for greater control over
material processing issues. First and foremost, the lack o f a suitable non-intrusive noncontact sensor limits the ability to accurately measure the temperature o f materials being
processed. Future efforts should focus on developing novel temperature sensors for use
in a microwave field. Two approaches to address this issue are: 1) multi-wavelength
pyrometry and 2) acoustic temperature sensors. Multi-wavelength pyrometers preclude
the requirement o f emissivity calibration and hence could potentially avoid any
inaccuracies in temperature measurement and allow
simultaneous temperature
measurements for different materials. Similarly, non-contact versions o f acoustic sensors
such as the in-situ ultrasonic sensor developed by Khuri-Yakub et al. [154] are currently
being investigated for their suitability in a microwave environment.
Successful
implementation o f these temperature sensors will allow reliable and accurate material
characterization and control o f thermal processing of materials.
The Microwave 2100 VFM furnace used in this study allows a maximum forward
power o f 500 W and a minimum increment o f 20 W. Solder reflow and ECA studies
showed that certain structures and formulations were sensitive to applied power. A 20 W
increment in applied power resulted in a significant difference in the resulting
temperature profile. Accurate control o f such processes, especially those involving finepitch features, requires a tighter control on the applied power.
increments in applied power are essential.
That is, smaller
A significant improvement in the
understanding and control o f processes in the VFM can be achieved by incorporating insitu characterization techniques.
For example, in-situ FTIR analysis could provide
257
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
valuable insights into the cure mechanisms by allowing in-line monitoring o f the changes
occurring in polymeric films during the cure reaction in the VFM. The dependence o f the
various processes occurring during the cure reaction on the microwave processing
conditions can be studied in-situ and controlled as required. Similarly, differences in the
evolution o f the residual stress in thin films cured over silicon were observed from
studies on BCB and PI 2611.
In-situ measurement o f wafer curvature by the
incorporation o f in-situ interferometric techniques in the VFM could help understand
these effects.
While results from the current study identified some o f the key VFM processing
parameters that affect the cured film properties, optimization o f the process conditions
based on a design o f experiments methodology could further improve the properties of
interest. Process models based on neural networks and genetic algorithms are currently
being developed to predict the output response as well as optimize the VFM curing
process. Predictive control algorithms together with in-situ characterization techniques
that detect changes in material dielectric and frequency response during the cure process
will allow real-time control o f material processing in the VFM.
Investigation and
implementation o f such advanced control strategies could significantly improve material
processing in the VFM.
The VFM cure studies on commercial dielectric polymers helped identify the
important structure-property relationships o f microwave processed films. These results
could assist in the design o f novel polymeric systems specifically for microwave
processing. For example, it was found in the case o f photosensitive polyimides that
258
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
microwave processing could possibly be favoring the crosslinking or polymerization of
the photoproducts.
Photoactive compounds (PACs) that can selectively absorb
microwave radiation and transform into smaller molecular weight products by
degradation or other reactions need to be investigated. The impact o f incorporating such
PACs on the lithographic performance o f the PSPIs also needs to be studied. Further
studies are required to understand the synergistic effects as well as the contributions due
to individual components o f the photosensitive dielectric system.
Synthesis and
formulation of such microwave-specific compounds allows an additional degree o f
control on the microwave processability o f these materials.
The feasibility o f rapid and selective heating o f materials by VFM processing has
been demonstrated in this study.
This can be applied to several applications in the
microelectronics industry. Examples o f these applications include: 1) fabrication o f airgap structures by decomposition o f sacrificial polymers embedded in organic or inorganic
materials for micro-fluidic and micro-systems applications and 2) formation o f porous
dielectric materials by decomposition o f sacrificial polymer either blended or reacted
with a base dielectric material (template technique) for use as inter-level dielectric
materials or as optical waveguides. Preliminary studies using VFM processing have
demonstrated the feasibility o f significantly reducing the process time (10 to 20 minutes
as compared to 4 to 5 hours) required for fabrication o f such structures.
Further,
designing sacrificial polymers, which preferentially absorb microwave energy as
compared to the base resin, can be selectively decomposed in temperature sensitive
applications.
259
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
To understand the limitations o f selective processing, a multilevel build-up structure,
which incorporates all the different materials and circuits required for achieving various
digital, rf and optical functionalities in advanced substrates, such as those in embedded
resistors, capacitors and inductors, should be fabricated using VFM curing for the build­
up dielectric layers.
260
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
APPENDIX A
The thermal stability o f low-temperature VFM cured films was determined by
thermogravimetric analysis (TGA) (Chapter V). Mass spectrometry (MS) studies were
conducted to determine the nature o f the residual volatile species evolving in the TGA.
The experimental procedure for these studies was outlined in sections 3.6.1 and 3.6.2 of
Chapter III. The results from the MS analysis of the cured films are discussed in the
following section.
Figure 72 and Figure 73 show the total ion trace and that o f the primary species
evolving from the negative tone polyimides, HD 4000 and XP 7001 films in the mass
spectrometer. Both the films were cured at 250°C. As seen from Figure 72 and Figure
73, at temperatures greater than 275°C, the primary species evolving from the cured films
correspond to mass to charge ratio 113, 281 and 309 and their fragments for HD 4000
and 45 and 213 for XP 7001 respectively. These high molecular weight species and their
fragments may be attributed to the ester-based crosslinking monomer and the sensitizers,
which are part o f the photopackage added to these PSPIs, and other photoproducts. For
example, Figure 74 shows the mass spectra o f HD 4000 films, which shows the different
fragments evolving from the film. These fragments correlate well with the mass spectra
o f ‘tetraethylene glycol dimethacrylate’ (TEGDMA), or 2-propenoic acid 2-methyl-,
oxybis (2,l-ethanediyloxy-2, 1-ethanediyl) ester (C16H2607: MW-330), which is
reported to be the crosslinking monomer in HD 4000. Similarly, higher molecular weight
261
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
species may result from the polymerized photoproducts. It is also reported [105] that the
crosslinking monomer in XP 7001 is a lower molecular weight species than that in HD
4000.
From Figure 73, it is evident that the molecular weight and the relative
concentration o f the species evolving form XP 7001 films are lower as compared to HD
4000.
Similar MS studies were also performed on CRC 8650 and the positive tone
polyimide PW 1200.
Figure 75 and Figure 76 show the mass spectra of the fragments evolving at high
temperatures (> 275°C) from CRC 8650 and PW 1200 films respectively, each cured at
250°C. The structures o f the probable parent species (from NIST database) are also
shown. As discussed in Chapter VI, for CRC 8650, the MS fragments compared well to
a 2,2’-methylenebis-phenol structure (C13H1202: MW-200). Similarly, the fragments
from PW 1200 could be associated with the parent species, 4, 4’, 4”-ethylidynetrisphenol (C20H1803:MW-306).
These compounds are typical o f DNQ bearing
photoactive compounds used in positive tone systems [115]. These results together with
the TGA analysis show that the actual photochemistry, the structure and molecular
weight o f the photoactive compounds and their thermal stability significantly affects the
effectiveness o f low-temperature VFM curing.
262
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
IONTRACE. Flagging=Scan Number. Max.Scan=2342#1;09:0a81.
Totd IonCurort. Mat.W=96.92575.
Sols
V
213 283 439 504 665 772 877 991
NN
si N
N N
^
1196131SH
66017701877 2077 2090 2210
h r [ l i r j MHi i i [ r n | T r i | i i i | i i i | i i i | l ti j i i i | f r i p i i | m | i i i j i i n iiTpm ] i n [ T i i | i M| i i i [ i i n i i i | m | i i i | m | i i i | i i i | i i i | i i i | i i i | i i i | i i i | i i i
0:00
moo
20:00
1Q0 Ion- 113 (Nam). Max.W.=10.4832.
3000
40:00
SO:00
1767 1877 2078 2090 2210
267 416
416 467
467 660
660 772
772 876
876 991
991
501 89 200
200 267
f
W
Ni ^
s'» > ''r N
|n H i n [ 11 n n i| i
1:00:00
n r p r r p r q i n 11111111 | n i | i u | i H | i i r jT i n m | m | n i | m j i n | n i | m | i n | n r j i i i | m | n i | i n | m | i i n n i | i m n n n i|
0.00
10:00
2000
3000
Ion-281 (Norn). Max.H.=0.83072.
100
£ s o l 8 217 330 439 526 663 771 881, ^ ^ 2
40.00
1
5
SO00
1:0000
13301461 1S53 16S618021906 20772148 2262
^
S S
S
< < \ (
(
| T f q T I I | r u p i l | n i j l l i ; MI | l l l | l l l | l l l | l l » | » l » | l l 1| i T i | i i i | » i i | i > i | i i i | i i i | i i i | i i i | i M| i i i | TTi | i i i | M i | i M | i l 7jrTi1IY?jl l T[I H | I M | l l l | l
0.00
moo
2000
3000
4000
SOOO
1:0000
Ion- 309 (Nor). Max.W.=0.96744.
100
so 82 159 224 347 502 644 767 872 982 11!® 797''S!161328148615551863180719122094 2109 2211
\ ( S' /
\
\ W
N f S
/ /
\ S S'
0 11H 11H I i i | 11H n i | f I H I n [ m 11 n 111111 n ) 11H I I H I I n r r r j r T - q i >n n i | i n | i 11| i i i | i n ~ |'i i i |
| TTTpTTyTTT^TTTJTTTJTTTJTTTJTTTJTTTJTTrj
X/V
OOO
1000
20.00
3000
4000
Retortion lime
5000
1:0000
Figure 72 Mass spectroscopy results of polyimide HD 4000 ramped at 10°C/min to 4S0°C.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ION TRACE. Flagging=Scan Number. Max.Scan=2102#1:02:0266.
Total Ion Current. Max.lnt.=18.40539.
719 1819 1919 2021
1311 1355 1?
SO
105 209 374 457 565 701 800 881 1009115®
f'
S'
^
^
N
N
>
tll|lll|lll|lll|lll|lll|lll|lll|lll|lll|lll|lll|lll|lll|lll|
0:00
10:00
20:00
Ion - 45 (Nom). Max.Int.=2.22976.
| l I I 1 1 I I | l I I 1 1 I I 1111| II l | « 111 I I 111 I I 11 • • |
40:00
30:00
50:00
100 q
1305 1337 151
£
1:00:00
722 1819 1919
\ /
P
7 ^
2022
50 •:
692 782 890 1009 1108
ro
IH111111m11111II111 n | m1111111111ii 111111•h 111111 ■11ii t j
*
0:00
10:00
20:00
Ion - 213 (Nom). Max.lnt.=0.05104.
30:00
100 -a
40:00
50:00
1:00:00
10QL1111 1290i33S 15diia4i4i1ft«al1729 1874 1938 ^
7
2055
SO
?
p m
379 477
0
0:00
I
moo
I ..
I
20:00
I
30:00
Retention Time
... I ... I
40:00
I
50:00
. b .
. I I . I 111 . 1 1 . 4 , 1
1:00:00
Figure 73 Mass spectroscopy results of polyimidc XP 7001 ramped at 10°C/min to 4S0°C.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
SCAN GRAPH.
Scan 1500#44:17.11. ■*", El. POS. Probe T =451.
100-j
113.1
69
90-
80-
70-
60-
41.1
40-
30-
20-
10-
200.2
143.1
100
150
200
250
Low Resolution M/z
300
350
400
450
Figure 74 Mass spectrum of polyimide HD 4000 at probe temperature o f 4S0°C and probable parent species (tetraethyleneglycol-dimethacrylate).
©
on
o
©
CM
w,
(A
Ul
IA
CM
S'
8CM
O
cn
oIA
8
—w
u
K CO
O
n
©
©
oCM
O
266
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 75 Mass spectra of CRC 8650 film at probe temperature of400°C and probable parent species (2,2’-methylenebisphenol).
oIA
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
SCAN GRAPH.
1Q0 Scan 1693049:59.73.
El. POS. Probe T =452.
291.1
90
80
7094.1
H ®-®----- ^ - ° H
50
212.1
44
40-
§
30
115
20-
306.2
149.1
469.2
241.2
390.1
200
409.2
250
300
Low Resolution M/z
Figure 76 Mass spectrum of polyimide PW 1200 at probe temperature of 4S2°C and probable parent species (4 ,4 ’, 4"ethylidynetris-phenol).
REFERENCES
[1] Technology news, Solid State Technology, Vol. 43, No.5, pp. 42, May 2000.
[2]
P. K. Mukeiji and G. Demet "Polyimides for electronic applications," Polymers
for Advanced Technologies," Vol. 4, pp. 309-315,1992.
[3] S.Hall, C. Schuckert, “Single Mask W afer Overcoat Process Using
Photodefinable Polyimide”, Solid State Technology, pp. 95-96, October 1999.
[4] John Baliga, “High density packaging: the next Interconnect Challenge”,
Semiconductor International, pp. 91-100, Feb. 2000.
[5] J. D. Meindl " Low power microelectronics: retrospect and prospect."
Proceedings o f the IEEE. Vol. 83, No. 4, pp.619-635, 1995.
[6]
M. K. Ghosh, K. L. Mittal, " Polyimides: Fundamentals and Applications,"
Marcel Dekker, New York, 1996.
[7] P. Singer, ” Low k Dielectrics: The search continues," Semiconductor
International, pp. 88-96, May 1996.
[8] T .Banba, E. Takeuchi, A. Tokoh and T. Takeda, “Positive Working
Photosensitive Polymers for Semiconductor Applications”, proceedings, 21st Electronic
Components and Technology Conference, pp 564- 567,1991.
268
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[9] G.Schmid, R. Sezi, K. Lowack, & W. Radlik, “A novel Oxazole based low-k
dielectric addresses copper damascence needs”, Semiconductor FabTech, 12th Edition pp
231-235,2000.
[10] R. R. Tummala, E. Rymaszewski and A. G. Klopfenstein, " Microelectronics
Packaging Handbook, part II: Semiconductor Packaging," Chapman & Hall, 1997.
[11] G. Messner, I. Turlik, J. W. Balde, P. E. Garrou, "Thin Film Multichip
Modules," A technical monograph o f the International society for hybrid
microelectronics,” Reston, VA, 1993
[12] Kazuyuki Horie, and Takashi Yamashita, “Photosensitive Polyimides:
Fundamental and Applications”, Technomic Publishing Co. Inc. 1995.
[13] Hellmut Ahne, Roland Rubner, Recai Sezi, “Photopattemable insulating
Materials”, Applied Surface Science, Vol. 106, pp. 311-315, 1996.
[14] O. Rohde, M. Riediker, A. Schaf&er, and J. Bateman, " High resolution, high
photospeed polyimide for thick film applications, "Solid State Technology,” pp. 109-112,
Sept 1986.
[15] R. Sharangpani, K. C. Cherukuri and R. Singh, “Low thermal budget processing
o f organic dielectrics”, IEEE Trans. Electron. Dev., 43 (7), pp. 1168-1170, July, 1996.
[16] Guray Tas, Priya Mukundan, Tim A. Johnson, and Selezion Hambir, “Opto­
electronic thin-film Characterization”, Sei. Intl. Vol. 24. No. 13, pp 81, Nov 2001.
[17] Masayuki OBA, “Effect o f curing accelerators on thermal imidization o f
polyamic acids at low temperature,” Journal o f Polym. Sci. A: Polym. Chem., vol. 34, pp.
651-658,1996.
269
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[18] Ji-Heung Kira and J. A. Moore, “A low-temperature route to polyimides”,
Macromolecules, Vol. 26, pp. 3510-3513, 1993.
[19] “Flip chip Technology on Organic Pin Grid Array Packages”, Mring J Li, Bob
Sankman, Hamid Azimi et al., Intel Technology journal, Q3 2000.
[20] Rao R. Tummala, “Fundamentals o f Microsystems Packaging”, McGraw-Hill
Companies, Inc., 2001.
[21] D.F. Stein et al., Microwave Processing, an emerging industrial technology,
Materials Research Society Symposium Proceedings, Vol. 347, pp 3-7, 1994.
[22] David A. Lewis, Microwave processing o f polymers-an Overview, Materials
Research Society Symposium Proceedings, Vol. 269, pp 21-29, 1992.
[23] Robert Lauf et al., “2 to 18 GHz Broadband Microwave Heating Systems”,
Microwave Journal, pp 24, November 1993.
[24] B. Panchapakesan et al., “Variable Frequency Microwave”, Advanced
Packaging, p.60, September / October 1997.
[25] “Microwave Processing o f Materials”, National Academy Press, Washington,
USA, 1994.
[26] Zak Fathi, D. A. Tucker, W.A Lewis, and J. B. Wei, “Industrial applications o f
VFM energy in materials processing”, Materials Research Society Symposium
Proceedings, Vol. 430, pp 21-28,1996.
[27] A.C. Metexas and R. J. Meredith “Industrial Microwave Heating”, Peter
Peregruins Ltd., London, UK, 1983.
270
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[28] Fred E. Gardiol, “Introduction v* Microwaves”, Artech House Inc. NY, 1984.
[29] Lambda Technologies Inc., Research Triangle Park, Morrisville, NC.
[30] Superwav Technologies Inc., Santaclara, CA.
[31] Craig Saltiel and Ashim K. Dutta, “Heat and Mass Transfer in Microwave
Processing”, Advances in Heat Transfer, Vol. 33, pp. 94, 1999.
[32] Von Hippel, “Dielectrics and Waves” , Wiley, NY, 1954.
[33] M.Chen, E J. Siochi, T.C. Ward, and J. E. McGrath, “Basic Ideas o f Microwave
Processing o f Polymers”, Polymer Engineering and Science, Vol. 33, No. 17, pp. 10921109,1993.
[34] E.T. Thostenson, and T.W. Chou, “Microwave processing: fundamentals and
applications, Composites, Part A 30 pp. 1055-1071,1999.
[35] Sutton W.H., Brooks M. H., Chbinsky I. J., “Microwave Processing o f
Materials”, Materials Research Society Symposium Proceedings, Vol. 124, Pittsburgh:
Materials Research Society, 1988.
[36] Snyder W. B. Jr., Sutton W. H., Iskander M.F., Johnson D. L., “Microwave
Processing o f Materials IT , Materials Research Society Symposium Proceedings, Vol.
189, Pittsburgh: Materials Research Society, 1991.
[37] Beatty R.L. Iskander M. F., Sutton, W.H., “Microwave Processing o f Materials
IE”, Materials Research Society Symposium Proceedings, Vol. 269, Pittsburgh: Materials
Research Society, 1992.
271
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[38] Iskander M i7., Lauf R.J., Sutton W.H., “Microwave Processing o f Materials
IV”, Materials Research Society Symposium Proceedings, Vol. 347, Pittsburgh:
Materials Research Society, 1994.
[39] Iskander M. F., Kiggans J.O., Bolomey J.C., “Microwave Processing o f
Materials V”, Materials Research Society Symposium Proceedings, Vol. 430, Pittsburgh:
Materials Research Society, 1996.
[40]”Microwaves: theory and application in material processing”, ceramic
transactions, Vol.(s) 21, 36,59, Westerville, OH: American Ceramic Society, 1991-1995.
[41] D E Lewis and J.M Shaw, “Recent developments in Microwave Processing o f
Polymers”, Materials Research Society bulletin 18 (11), pp 27-40 (1993).
[42] Jovan Mijovic and Wijaya J., “ Review o f Cure o f Polymers and Composites by
Microwave Energy’, Polymer Composites, Vol. 11, No. 3, pp. 184-191,1990.
[43] Teibing Wang and Johan Liu, “ A review o f Microwave curing o f Polymeric
materials”, Journal o f Electronics Manufacturing, Vol 10, No. 3, pp. 181-189,2000.
[44] Iftickhar Ahmad, Zak Fathi, Mark Konarski, Denise Tucker, and Erin Yaeger,
“A look at Variable Frequency Microwave Curing”, Circuits Assembly, pp. 54-63, July
1998.
[45] David A. Lewis, Susan J . Lamarie, and Thomas Nunes, “Microwave
Processing o f Polyimide Thin Films”, Materials Research Society Symposium
Proceedings, Vol. 347, pp. 681-689,1991.
[46] Xiaomei Fang, and Daniel Scola, “Investigation o f Microwave Energy to Cure
Carbon Fiber Reinforced Phenylethynl-Terminated Polyimide Composites, PETI-5/IM7,
J. Poly. Sci. Part A: Poly. Chem. Vol. 37, pp. 4616-4628,1999.
272
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[47] Wei J. B., Zak Fathi, B. Pan, and Su Liang Chan, “Use o f Microwave
technology for rapid-cure of Chip-on-Board glob top encapsulants’, proceeding s o f 3rd
EPTC, pp. 181-185,2000.
[48] Kim Farnsworth, Manepalli R.H., Bidstrup-Allen S. A., and Paul A. Kohl,
“Variable frequency microwave curing o f 3,3',4,4-biphenyltetracarboxylic acid
dianhydride/P-phenylenediamine (BPDA/PPD)”, International Journal o f microcircuits
and Electronic Packaging, Vol. 23, No. 2, pp. 162-171,2000.
[49] Kim Farnsworth, Manepalli R.H., Bidstrup-Allen S. A., and Paul A. Kohl,
“Variable frequency microwave curing o f Photosensitive Polyimides”, IEEE transactions
on Components and Packaging Technology, Vol. o f microcircuits and Electronic
Packaging, Vol. 24, No. 3, pp. 474-481, Sep 2001.
[50] Catherine Jordan, Jocelyne Galy, and Jean-Pierre pascault, “Comparison o f
Microwave and thermal cure o f an Epoxy-Amine /M atrix”, Poly. Eng. Sci. 35, 3, 233,
1995.
[51]
D.A. Lewis, J.D. Summers, T.C. Ward and J.E. McGrath, “Accelerated
imidization Reactions using Microwave Radiation”, J. Poly. Sci. Part A. Poly. Chem.
Vol. 30, pp. 1647-1653,1993.
[52] In-Situ real Time Study o f Crosslinking Kinetics in Thermal and Microwave
fields,Mijovie et al Polymers for Advanced technologies, 9,231, 1998.
[53] Delmotte et al., Materials Research Society Symposium Proceedings, Vol.430,
pp 607,1996.
[54]
441,1996.
Hirota et al., Materials Research Society Symposium Proceedings, vol.430, pp
273
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[55] J.W ei, Proceedings o f the ACS Division o f Polymeric Materials, Science and
Engineering, 72,10,1995.
[56] Zak Fathi et al., Materials Research Society Symposium Proceedings, Vol. 445,
pp 125,1997.
[57] J. Wei et al. Polymer Engineering and Science, 33,17, pp. 1132, 1993.
[58] Albert Gourdenne , Proceedings o f the second international Conference on
Reactive Processing o f Polymers, pp 23, 1982.
[59] D.A.Lewis et al., Polymer preprints, Vol. 28,2, 30,1987.
[60] Jianghua Wei et al., Polymer Engineering and Science, 35,6,461,1995.
[61] J.C. Hedrick, D. A. Lewis, T.C. Ward, J. E. McGrath, Materials Research
Society Symposium Proceedings, Vol. 189, pp. 421-429, 1988.
[62] Hanlin Zhang et al., Materials Research Society Symposium Proceedings, Vol.
430, pp. 447,1996
[63] Xiaomei Fang et al., Journal o f Polymer science, Part A, Vol. 38,2526,2000.
[64] Jovan Mijovic et al., Polymers for Advanced technologies, 9,231,1998.
[65] Joan Mijovic et al., Macromolecules, 25,986,1992.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[66] Lee Peerry et al., Materials Research Society Symposium Proceedings, Vol.430,
pp. 391,1996.
[67] J. Mijovic et al., Macromolecules, vol. 23, pp 3671, 1992.
[68] Alazard et al., Materials Research Society Symposium Proceedings, vol.430, pp
593, 1996.
[69]
1988.
Gourdenne, Polymeric Materials Science and Engineering, vol. 59, pp 375,
[70] Eva Mamard et al., Macromolecules, 25, pp 2243,1992.
[71] D.A. Lewis et al., Journal o f Polymer Science, 29, pp 174, 1988.
[72]
J. Jacob, L.L Chia, Boey F.Y.C, “thermal and non-thermal interaction o f
microwave radiation with materials”, Material Science, Vol. 30 (21), pp. 5321-5327,
1995.
[73] J. C. Hedrick, D A . Lewis, T.C. Ward, and T.E. McGrath, “Electromagnetic
processing o f polymers:I. Basic concepts and Design o f Macromolecules”, Material
Research Society, proceedings, Vol. 189, pp. 421 1991.
[74] Glenn E. Fanslow, “Microwave Enhancement o f Chemical and physical
Reactions”, Materials Research Society proceedings, Vol. 189, pp. 43-48,1991.
[75] Ted Stokich et al., “Real-time FT-IR studies o f the reaction kinetics for the
polymerization o f divinyl siloxane bis-benzocyclobutane monomers” M aterial Science o f
High Temperature polymers Symposium, Proceedings, Material Research Society, Vol.
227, pp 103-114,1991.
275
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[76] Mark A Janney et al., Microwave processing o f Ceramics, Materials Research
Society Symposium Proceedings, Vol. 269, pp. 173, 1992.
[77] R.K Agarwal et al, Journal o f Adhesion, Vol. 29, pp. 63-79, 1989.
[78] Teibing Wang, Ying Fu, Matthias Becker, Johan Liu, “Heating and reliability
characteristics o f Electrically conductive adhesives Using Variable Frequency
Microwave cure”, 3rd EPTC 2000, pp. 373- 377.
[79] Kimberley Farnsworth, Ph.D. dissertation, Georgia Institute o f Technology,
1999.
[80] D.G. Rogers et al., High Performance Polymers, 10, pp 341,1998.
[81] C. A. Pryde, "IR studies o f polyimide I. Effects o f chemical and physical
changes during cure, ", Journal o f Polymer Science: part A: Polymer Chemistry, vol.27,
pp. 711-724, 1989.
[37]
R. Ulrich and R. Torge, "Measurement o f thin film parameters with a prism
coupler," Applied Optics, Vol.12. No.12, pp. 2901-2908, Dec. 1973.
[83]”Standard test methods for AC loss characteristics and permittivity o f solid
insulating materials," ASTM standard D150-95.
[84] P. A. Flinn, "Principles and applications o f wafer curvature techniques for stress
measurements in thin films," Materials Research Society Proceedings, Vol. 130, pp. 4151, Nov. 28-30, 1998.
[85] Flexus Incorporated, thin films stress measurement system, operational manual,
Model F2320, KLA Tencor, San Jose, CA.
276
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[86]
K. K Kanazawa, O. R. Melroy, “The Quartz Resonator: Electrochemical
applications,” IBM Journal o f Research and Development, vol. 37, pp. 157-171, No. 2,
March 1993.
[87]
Philip Garrou, “Polymer Dielectrics for Multi-chip module Packaging”,
Proceedings o f IEEE, Vol. 80, No. 1, pp 1942-1954, Dec., 1992.
[88]
2000 .
Philip Garrou, Semiconductor International, Vol. 23, No. 12, pp 119-128, Oct.,
[89] BCB Technical Data Sheets, Dow Chemical Company, Midland, MI, 48642.
[90]
Cyclotene™ processing guidelines and data sheets, Dow Chemical Company,
Midland, MI 48642.
[91] Gillham J.K, Polym. Eng. Sci. Vol. 26 (20), pp 1429-1433,1986.
[92] Bair H.E and C.A. Pryde, Proceedings, Soc. Plastic Engineers, ANTEC
Montreal, pp 1550-1553, 1991.
[93] Townsend P.H., Stokich T.M., and Huber B.S, “Mechanical properties o f BCB
films on silicon substrates”, Material Research Society Symposium Proceedings, pp 215223, Vol. 226,1991.
[94] Heistand R.H et al., “Advances in MCM fabrication with BCB dielectric”, The
International Journal o f Microcircuits and Electronic Packaging, Vol. 15, Number 4, pp
183-192,1992.
277
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[95] Kirk Laursen et al., “Examination o f High Frequency Dielectric Properties o f
Thin Film Polymers using an In-situ Resonant technique”, Multi-Chip Module
Conference, Proceedings. IEEE, pp 229-231, 1995.
[96] Thomas C. Hodge et al., “Rapid Thermal Curing o f Polymer Interlayer
Dielectrics”, The International Journal o f Microcircuits and Electronic Packaging, Vol.
17, Number 1, pp 10-19,1994.
[97] Philip E. Garrou et al., “Rapid thermal curing o f BCB dielectric”, IEEE
transactions on Components, Hybrids and Manufacturing Technology, Vol. 16, No. 1, pp
46-52, Feb. 1993.
[98] E Molis, "Infrared characterization o f molecular orientation in polyimide films,”
Polyimides: Materials, Chemistry and Characterization, pp. 659-672, Elsevier Science
Publishers, B. V., Amsterdam, 1989.
[99] M. Ree, K. Kim, S. H. Woo and H. Chang,” Structure, chain orientation and
properties in thin films o f aromatic polyimides with chain rigidities," Journal o f Applied
Physics, pp. 6333-6345, 1997.
[100] H. Nomura and M. Asano, ” M olecular orientation in polyimide films having
rodlike molecular skeleton on silicon substrate," Japanese Journal o f Applied Physics,
Vol. 32, pp. 3933-3937, Part 1, No.9A, 1993.
[101] "Pyralin polyimide coatings for electronics, PI 2610 Series-Product
information and process guidelines,” July 1998, HD Microsystems, Wilmington, DE
19880.
[102]
Tomoyuki Yuba, Mitsuhito Suwa, Yoji Fujita, Masao Tomikawa and Gentaro
Ohbayashi, “A Novel Positive Working Photosensitive Polyimide for W afer Level CSP-
278
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Packages, Proceedings from the tenth meeting o f ‘The Symposium o f Polymers in
Microelectronics’, Wilimington, Delaware, May 9-11,2002.
[103] C. A. Pryde, "FTER measurements o f polyimides. n. Factors affecting
quantitative measurement, " Journal o f Polymer Science, part A: Polymer Chemistry,
V ol.27,pp.711-724,1989.
[104] Rahul N. Manepalli, PhD dissertation, Georgia Institute o f Technology, May
2000.
[105] Toshiaki Itabashi et al., private communication, HD M icrosystems, Japan.
[106] J. C. Cobum and M. T. Pottiger, "The influence o f processing conditions on
structure, properties and stress development in spin coated polyimide films," Proceedings
o f the Fourth International Conference on Polyimides, pp. 360-374, Oct. 30-Nov. 1,
1991.
[107] Shin’ya Morino and Kazuyuki Horie, “In-Plane Orientation in Poly(amic
acid)s during Thermal Imidization, Polymer Journal, Vol. 31, No. 9, 1999.
[108] Hyunsoo Chung, Yungil joe, Haksoo Han, ‘The Effect o f Curing History on
the Residual Stress Behavior o f Polyimide thin Films”, Journal o f Applied Polymer
Science, Vol. 74, pp. 3287-3298, 1999.
[109] Hyunsoo Chung, Jonghwae lee, Wonbong Jang, Youggun Shul, Haksoo Han,
“Stress Behaviors and Thermal Properties o f Polyimide Thin Films Depending on the
Different Curing Process”, Journal o f Polymer Science: Part B: Polymer Physics, Vol. 38,
2879-2890,2000.
[110]
Radko Osredkar, “A Study o f The Imidization Kinetics o f Polyimide Resin
Curing”, Microelectronics Reliability, Vol. 28, No. 4, pp. 599-603,1988.
279
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[111] John C. Cobum, Michael C. Pottiger, Susan C. Noe, and Stephen D. Senturia,
“Stress in Polyimide Coatings”, Journal o f Polymer Science: Polymer Physics, Vol. 32,
pp. 1271-1283,1994.
[112] Material safety data sheets, 08/10/00, HD Microsystems, Wilmington, DE
19880
[113] “Polymers in Electronics”, American Chemical Society Symposium Series,
242, pp. 325-344,1984.
[114] Khudyakov, I.V.; Legg, J.C.; Purvis, M.B.; Overton, B.J., “Kinetics o f
photopolymerization o f acrylates studied with a cure monitor”, Proceedings o f the SPIE The International Society for Optical Engineering, Vol. 3848, pp. 151-156, 1999.
[115] Ralph Dammel, “Diazonaphthoquinone-based
Engineering Press, pp. 13-15, 1993.
Resists”,
SPIE
optical
[116] M. Koshiba, M. Murata, M. Matsui, and Y. Harita, “Thermally Induced and
Base Catalysed Reactions o f Naphthoquinone Diazides”, Proc. SPIE, 920, pp. 364-371,
1988.
[117] Material Safety Data Sheets, Sumtitomo Bakelite Co.MSDS-CRC-095,03/02.
[118] Banba et al., US Patent # 5,449,584.
[119] Tae-Kyun Kim, kil-Yeong Choi, Kwang-Sup Lee, Dong-Woo park, and Moon
Young Jin, “Thermal conversion o f t-butyloxycarbonylyoxy attached polyamides to
polybenzoxazoles”, polymer Bulletin, Vol. 44, pp. 55-62,2000.
280
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[120] Guey-Sheng liou “New rigid-rod poly(benzoxazole-imide)a containing chlorosubstituted p-phenylene units in the main chain”, J. Poly. Sci. part A: Poly. Chem. Vol.
37, pp. 4151-4158,1999.
[121] Hiroshi Seino, Osama Haba, Mitsuru Ueda, Amane Mochizuki,
“Photosensitive polyimide-precursor based on polyisoimide: dimensionally stable
polyimide with a low dielectric constant”, Polymer Vol. 40, pp. 551-558, 1999.
[122] Sheng-Huei Hsiao, Ming-Hsiang He, “Synthesis and properties o f Novel
Cardo Aromatic poly(ether-benzoxazole)s”, J. Poly. Sci: Part A: Poly. Chem. Vol. 39, pp.
4014-4021,2001.
[123] Sheng-Huei Hsiao, Jiun-Hsiang Chiou, “Aromatic polybenzoxazoles
containing ether-sulphone linkages”, J. Poly. Sci: Part A: Poly. Chem. Vol. 39, pp. 22622270, 2001.
[124] Jin-Hae Chang, Kwang min Park, and Ihn Chong Lee, “Thermal
cyclodehydration o f aromatic precursor polymers to polybenzoxazole and polyimide;
their thermal and mechanical properties:, polymer Bulletin, Vol. 44, pp. 63-69,2000.
[125] Gordon L Tullos, Jason M.Powers, Stacy J. Jeskey, and Ion J. Mathias,
‘Therm al conversion o f hydroxy-containing imides to benzoxazoles: polymer and model
compound study”, Macromolecules, Vol. 32, pp. 3598-3612, 1999.
[126] Ji-heung Kim and Jae Kwan lee, “Hydroxy-substituted polyenaminonitrile as a
soluble precursor for rigid-rod polybenzoxazole”, Bull. Korean Chem.. Soc. Vol. 22, No.
9, pp. 99-1004,2001.
[127] Jerry Workman, Jr. “Handbook o f Organic Compounds-Methods and
Interpretations”, Academic Press, 2001.
281
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[128] Jin Hae Chang, kwang min park, Soo-min lee, and Jae Buem Oh, “Two-step
thennal conversion from poly(amic acid) to polybenzoxazole via polyimide: their thermal
and mechanical properties”, J. Poly. Sci: Part B. Poly. Phy. Vol. 38, pp. 2537-2545,2000.
[129] Gordon 1. tullos and Lon j. Mathias, “Unexpected thermal conversion of
hydroxy-containing polyimides to polybenzoxazoles”, Polymer 40, pp. 3463- 3468,1999.
[130] Ki Hong park, Masa-Aki Kakimoto, and Yoshio Imai, “Synthesis and
properties o f new soluble aromatic polybenzoxazoles from 4,4’-diamino-3-3’dihydroxytriphenylamines and aromatic dicarboxylic acids”, J. Poly. Sci., Part A: Poly.
Chem. Vol. 36, pp. 1987-1994,1998.
[131] Tsuguo Yamaoka, Nobuko Nakajima, and Ken’ichi Koseki, “A study o f novel
heat resistant polymers: preparation o f photosensitive fluorinated polybenzoxazole
precursors and physical properties o f polybenzoxazoles derived from the precursors”, J.
Poly. Sci. Part A: Poly. Chem.Vol. 28, pp. 2517-2532, 1990.
[132] ChunShing Chen, Jen-Jiang Lee, and Wei Wu, “Fourier transform infrared
analysis o f resist process and its applications”, SPIE, Vol. 1086, pp. 331-339, 1989.
[133] Nigel P. Hacker, and Nicholas J. turro, “Photochemical generationof ketenes.
An ultra-violet and infra-red spectroscopic study o f 1-and 2-diazonaphthalenones at
77K”, Tetrahedron letters, Vol. 23, No. 17, pp. 1771-1774,1982.
[134] Marc T. Aronhime, Chava Gal, Shoshana Silady, and John J. Grunwald,
‘Improved thermal stability o f high resolution positive photoresists via elevated softbake
temperature”, SPIE, Vol. 1086, pp. 97-104,1989.
[135] Mitsunobu Koshiba, Makato Murata, Mariko Matsui, and Yoshiyuki Harita,
“Thermally induced and base catalyzed reactions o f Naphthaquinone diazides”, SPIE
Vol. 920, pp. 364-371,1988.
282
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[136] Andrew Blankey et al., “ Novolacs with alternating p-creosl and
polyhydroxyphenyl monomer units: developer cationic effects and thermal properties’,
SPIE Vol. 920, pp. 339-348, 1988.
[137] MMCC, Electronics Cooling, January 1999.
[138] G.P. Glinski and C. Baiely, “Microwave cure o f conductive adhesives for flipchip and microsystems applications”, IEEE 2002 Inter Society Conference on thermal
phenomena, pp. 847- 853.
[139] Alina Deutsch, et al., “ Measurement o f dielectric anisotropy o f BPDA-PDA
polyimide in multilayer thin-film packages”, IEEE transactions on Components,
Packaging, and Manufacturing Technology-Part B, Vol. 17, No. 4, Nov. 1994.
[140] http://webbook.nist.gov/cgi/cbook Web book: National Institute o f Standards
and Testing.
[141] FTIR Studies o f Polyimides: Thermal Curing, R.W. Snyder, B Thompson et
al., Macromolecules, 22, pp. 4166-4172,1989.
[142] D.E. Fjare et al, “IR studies o f polyamic acid curing: Solvent effects”,
Advances in Polyimide Science and Technology, edited by C. Feger et al., Technomic
publishing C. Inc. pp. 326-335, 1991, and references therein.
[143] M. J. Brekner and Claudius Feger, “Curing studies o f a polyimide precursor”,
J. Poly. Sci.: Part A: Vol. 25, pp. 2005-2020,1987.
[144]C. A. Pyrde, “IR studies o f Polyimides I: Effects o f Chemical and Physical
Changes During Cure”, J. Poly. Sci.: Part A: Vol. 27, pp. 711-724, 1989. and
references:4,9 ,1 0 ,8 ,3 ,1 1 therein.
283
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
[145]
W.S. Li et al., “FTIR Study o f Imidization Process o f Photosensitive
Polyimide PMDA/ODA”, Applied Spectroscopy, Vol. 52, No. 7, pp. 985-989,1998.
[146]
Tzu-Chien, J Hsu, and Zu-Ling Liu, “Solvent Effect on the Curing o f
Polyimide Resins”, J.App. Poly. Sci. Vol. 46, pp. 1821-1833,1992.
[147]
2485, 1987.
M. J. Brekner and Claudius Feger, J. Poly. Sci.: Part A: Vol. 25, pp. 2479-
[148] Daoqiang Lu, Ph.D. dissertation, Georgia Institute o f Technology, 2000.
[149] R.A wood et al., IEEE Electronics Letters Vol. 30, No. 25, pp 2120-2122.
[150]Teibing Wang, Ying Fu, Matthias Becker, and Johan Liu, “Microwave Cure o f
Metal Filled Electrically conductive adhesive”, IEEE, 51st ECTC, pp. 593-597,2001.
[151] Process developed at the Focus Research Center, Georgia Institute o f
Technology.
[152] J.D. Meindl, J.A. Davis, P. Zarakesh-Ha, C.S. Patel, K.P. Martin, and Paul A.
Kohl, “Interconnect opportunities for gigascale integration”, IBM J. Res. & Dev., Vol.
46, No. 2/3 March/May 2002.
[153] Based on a model developed by Krishna J. Tunga, Georgia Institute o f
Technology, ME 2003.
[154] “In-situ ultrasonic measurement o f photoresist glass transition temperature”,
Susan L. Morton, F. Levent Degertekin, and B. T. Khuri-Yakub, Applied Physics Letters,
Vol. 72. No. 19, pp 2457-2459,1999.
284
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
VITA
Ravindra Tanikella was bom in Hyderabad, India on December 13, 1976.
He
graduated with his Bachelor o f Technology (B. Tech) degree in chemical engineering from
Osmania University, India, in 1998.
He began his graduate studies in the school o f
chemical engineering at the Georgia Institute o f Technology in September 1998.
He
received his Ph.D. in chemical engineering in May 2003. He is currently employed as a
senior packaging engineer in the Assembly Technology Development division of Intel
Corporation at Chandler, Arizona.
28S
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Документ
Категория
Без категории
Просмотров
0
Размер файла
10 235 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа