close

Вход

Забыли?

вход по аккаунту

?

el%3A20010056

код для вставкиСкачать
spectral shape or the far-field characteristic of the emission have
been presented. The only argument for lasing (in contrast to spontaneous emission) provided is the dependence of the far infrared
output power on the injection current (Fig. 3 in [l]). It consists of
very few data points which might also be interpreted as a superlinear increase in the emission (without threshold), as has been
previously observed on different samples [5, 61.
Finally, the supposed far infrared laser spectrum (inset of Fig. 3
in [l]) is extremely similar to spontaneous emission spectra (Fig. 3
in [2]) reported by the same authors. This close similarity is even
more astonishing since the samples are equipped with different
waveguide designs, which are supposed to influence the spectral
shape of the emission, if it arises from the quantum dots.
In conclusion, the existence of laser emission corresponding to
transitions between discrete bound electron states could not be
evidenced by the results presented in [l].
0 IEE 2001
31 October 2000
Electronics Letters Online No: 20010055
DOI: 10.1049/el:20010055
A. Weber and M. Grundmann (Institut f u r Festkiirperphysik,
Technische Universitut Berlin, Hardenbergstr. 36, 0-10623 Berlin,
Germany)
E-mail: aweber@physik.tu-berlin.de
N.N.
Ledentsov
(A.F. Iofle
Physical-Technical
Institute,
Politekhnicheskaya 26, 194021, St. Petershurg, Russiu)
as described in our Letter [3]. This also addresses the transition
linewidth issue. We also observed that, in our device, the interband lasing threshold current was a factor of 1.6 lower than the
observed intersubband threshold current. The light-current characteristics of another device, showing a distinct threshold behaviour,
which is different from the superlinear behaviour of LEDs
observed by us [4] and Grundmann et al. [5], are shown in Fig. 1.
Unfortunately, device heating prevents us from making measurements at higher injection currents. We have also measured the
polarisation of the output and find it to be TE polarised as predicted by theory [2]. The output is broad and multimode as is
characteristic of interband QD lasers at high injection levels. It
may be noted that our waveguide design is not too different from
that of an interband QD laser, for reasons mentioned in [3]. A distributed feedback (DFB) device should exhibit more single-mode
behaviour.
I .o
M. Grundmann: Currently at: Universitat Leipzig, Institut fur
Experimentelle Physik 11, Linntstrasse 5, D-043 17 Leipzig, Germany
N.N. Ledentsov: Also with TU Berlin
References
1
KRISHNA, S , BHATTACHARYA, P., MCCANN, P.J., and NAMJOU, K.:
‘Room-temperature long-wavelength (h = 1 3 . 3 ~ )unipolar
quantum dot intersubband laser’, Electron. Lett., 2000, 36, pp.
1550-1551
2 KRISHNA, s., QASAIMEH, o., B H A ~ A C H A R Y A P.,
,
MCCANN, P.J., and
NAMJOU, K.: ‘Room-temperature far-infrared emission from a selforganized InGaAsiGaAs quantum-dot laser’, Appl. Phys. Lett.,
2000, 76, pp. 3355-3357
3 STIER, o., GRUNDMANN, M., and BIMBERG, D.: ‘Electronic and optical
properties of strained quantum dots modeled by 8-band k.p
theory’, Phys. Rev. B, 1999, 59, pp. 5688-5701
4 SAUVAGE, S., BOUCAUD, P., GERARD, J.-M., and THIERRY-MIEG, V.: ‘Inplane polarized intraband absorption in InAsiGaAs self-assembled
quantum dots’, Phys. Rev. B, 1998, 58, pp. 10562-10567
5 VOROB’EV, L E., FIRSOV, D.A., SHALYGIN, V.A., TULUPENKO, V.N.,
SHERNYAKOV, YU.M.,
LEDENTSOV, N.N.,
USTINOV, v.M.,
and
ALFEROV, zH.1.: ‘Spontaneous far-IR emission accompanying
transitions of charge carriers between levels of quantum dots’, J.
Exp. Theor. Phys. Lett., 1998, 67, pp. 215-279
6 GRUNDMANN, M.,
WEBER, A.,
GOEDE, K.,
USTINOV, v.M.,
ZHUKOV, A.E., LEDENTSOV, N.N., KOP’EV, P.s., and ALFEROV, ZH.I.:
‘Midinfrared emission from near-infrared quantum-dot lasers’,
Appl. Phys. Lett., 2000, 77, pp. 4-6
0.5
12.5
16.5
h, Ctm
1
0
0.5
’
“
”
‘
1
1
’
1 .o
injected current, A
1.5
Fig. 1 Light-current characteristics of quantum dot
infrared emission
The error bars in the measured data are also shown
Inset: Polarisation dependence of output light
TE
_ - _ - TM
T = 285K, I = 770mA, L = 0.8mm, ridge = SOP
To summarise, we believe we observe intersubband gain and
dominant stimulated emission in these devices with a distinct
threshold. The output power is, of course, woefully low, since in
the device we are converting a large population of interband photons to very few intersubband photons.
28 November 2000
Electronics Letters Online No: 20010056
DOI: 10.IO49/el:200I0056
S. Krishna and P. Bhattacharya (Department of Electrical Engineering
and Computer Science, University of Michigan, Ann Arbor, M I 481092122, USA)
P.J. McCann and K. Namjou (Laboratory for Electronic Properties of
Materials, Vniversity of Oklahoma, Norman, OK 73019-1023, USA)
0 IEE 2001
Reply to Comment on ‘Room-temperature
long-wavelength (h= 13.3pm) unipolar
quantum dot intersubband laser‘
S. Krishna, P. Bhattacharya, P.J. McCann a n d
K. Namjou
References
The comments made by Weber et al. [l] are welcomed. Obviously,
our Letter and the comment are in an important emerging area of
application of self-organised quantum dots (QDs). Since our original report, we have carried out further work which makes us
believe that we observe dominant stimulated emission and gain in
these devices, with a distinct threshold and polarisation of the output.
First, the emission at 1 3 . 3 p , in all probability, originates from
electron transitions from higher quantum dot excited levels to the
ground state, separated by -90-100meV. Theoretical calculations
have confirmed the presence of these excited states [2]. The important point to note is that the carrier dynamics between the barrier/
wetting layer states and the ground state in these dots is the same
WEBER, A., GRUNDMANN, M., and LEDENSTOV, N.N.: ‘C0n”nt:
Room-temperature long-wavelength (h = 1 3 . 3 ~ )unipolar
quantum dot intersubband laser’, Electron. Lett., 2001, 37, (2), pp.
96-97
2 JIANG, H.-T., and SINGH, J.: ‘Self-assembled semiconductor
structures: electronic and optoelectronic properties’, IEEE J.
Quant. Elect., 1998, 34, pp. 1188-1196
3 KRISHNA, s., BHATTACHARYA, P., MCCANN, P.J., and NAMJOU, K.:
‘Room-temperature long-wavelength (h = 1 3 . 3 ~ )unipolar
quantum dot intersubband laser’, Electron. Lett., 2000, 36, pp.
1550-1 55 1
4 KRISHNA, s., QASAIMEH, o., BHATTACHARYA, P., MCCANN, P.J., and
NAMJOU, K.: ‘Room-temperature far infrared emission from selforganized InGaAsiGaAs quantum dot laser’, Appl. Phys. Lett.,
2000,76, pp. 3355-3357
ELECTRONICS LETTERS
No. 2
18th January 2001
Vol. 37
1
97
5
GRUNDMANN, M.,
WEBER, A.,
GOEDE, K.,
ZHUKOV, A.E., LEDENSTOV, N.N., KOP’EV, P.S., and
USTINOV, V M.,
ALFEROV, ZH.1.:
‘Mid-infrared emission from near-infrared quantum dot lasers’,
=gk
’Yk
Appl. Phys. Lett., 2000, 77, pp. 4 6
Cwkj.93
(g
w+t
j
+ w3o + wka)
I C = 1,..., c
CAD of rectangular-waveguideH-plane
circuits by segmentation, finite elements
and artificial neural networks
J.M. Cid and J. Zapata
where x,is the ith input, yk is the Mh output, gk(.) and g,{.) are
activation functions (typically: sigmoid, tanh, ...), M is the number
of neurons in the hidden layer and the IVS are adjustable parameters called weights. The use of the MLPs is based on the universal
approximation theorem, using which it can be deduced that an
MLP with only two layers of weights is capable of modelling virtually any real function to any desired degree of accuracy if the
number of neurons in the hidden layer is large enough.
Rectangular waveguide H-plane circuits are efficiently analysed by
the segmentation finte element method. Artificial neural
networks, the computation times of which are very fast, are used
to generalise the responses to different configurations. The
analytic model obtained in this way is used in optimisation
programs for microwave circuit design. Application of the method
to a three cavity fdter is presented.
Introduction: In the last few years, the application of segmentation
methods, as well as the growth in computer capability, have made
the finite element method (FEM) faster and more flexible [l].
These advances suggest the introduction of the FEM in optimisation programs for microwave circuit design, which is the objective
of this work. The design time is also reduced by applying artificial
neural networks (ANNs) which model the response of the complete circuit starting from a small set of responses obtained by
FEM.
Theory: The optimisation of a microwave circuit is a problem
where the differences between the circuit response and the desired
response are minimised by adjusting the free parameters of the
system. As a consequence, precise knowledge of the circuit
response is needed to achieve a good design.
Thus, the segmentation finite element (SFE) method presented
in [l] is a powerful full-wave method of analysis which allows the
response of arbitrarily shaped 3D microwave passive circuits to be
obtained very efficiently. This method divides the circuit under
study into small regions connected by waveguide segments of arbitrary cross-section, which are analysed separately. In each region,
the generalised scattering matrix (GSM) is computed by FEM and
the response of the complete circuit subsequently obtained by connecting all the GSMs through the respective waveguides. To apply
this idea to H-plane rectangular waveguide circuits, a simplified
version of the method in [l] is implemented by forcing the derivatives with respect to the waveguide height to be zero. This allows
the problem to be solved in a 2D space.
In general microwave circuit design, the cost function that must
be minimised is extremely complex and has many local minima.
The algorithms which manage this type of function are based on
simulated annealing techniques or evolutionaq strategies, which
solve a problem at the expense of calling the cost function excessively. The attainment of the circuit responses by means of a
numeric method such as the SFE make the resolution of the problem too slow. To solve this matter, an analytic model based on
ANNs is introduced. Using this model and maintaining high precision and flexibility, the circuit response is obtained much quicker
than with SFE, and therefore the design time is reduced.
In the analytic model proposed in this work, the segmentation
concept is applied the complete circuit is divided (just as in the
SFE method) and the behaviour of each section is modelled separately. In this way, the modelling of the circuit response is simplified drastically because the number of free parameters is less in a
section than in the complete circuit, and the complexity of an
approximating function increases exponentially with the dimension of the input space (the so-called curse of dimensionality).
Moreover, the electromagnetic response of a small region is softer
and thus easier to model.
In each region of the microwave circuit, the GSM obtained
starting from the analysis with the FEM is approximated applying
multi-layer perceptrons (MLPs). An MLP is an ANN that can be
seen as an application of RDin RC.The expression of an MLP
with two layers of weights can be written [2]:
98
1377/11
Fig. 1 Geometry and segmentation of rectangular waveguide H-plane
three-cavity filter
Dimensions (in cm) are: A = 1.905; T = 0.2; R = 0.15
L I , L2, H I and H2 are free parameters of system
901
30
t
01
8
0.2
0.4
0.6
0.8
Hi, cm
1.0
1.2
’
y-14
1-17
1.4
Fig. 2 Variation of iris response against H, at fixed frequency
0
computed by FEM
computed by MLP
I
-60’
1 1 .o
11.5
12.0
12.5
frequency, GHz
13.0
13.5
)377/31
Fig. 3 Filter response for optimised dimensions (in cm) of free parume-
ters
L1 = 1.0607; L2 = 1.1918; H I = 0.9168; H2 = 0.5910
computed by SFE method
0
computed by analytic model
ELECTRONICS LETTERS
18th January 2001
Vol. 37
No. 2
Документ
Категория
Без категории
Просмотров
3
Размер файла
258 Кб
Теги
3a20010056
1/--страниц
Пожаловаться на содержимое документа