вход по аккаунту



код для вставкиСкачать
Low-noise humidity controller for imaging water mediated processes in atomic force
I. Gaponenko, L. Gamperle, K. Herberg, S. C. Muller, and P. Paruch
Citation: Review of Scientific Instruments 87, 063709 (2016);
View online:
View Table of Contents:
Published by the American Institute of Physics
Articles you may be interested in
Towards reversible control of domain wall conduction in Pb(Zr0.2Ti0.8)O3 thin films
Applied Physics Letters 106, 162902 (2015); 10.1063/1.4918762
Ferroelectric domains in epitaxial PbxSr1-xTiO3 thin films investigated using X-ray diffraction and
piezoresponse force microscopy
APL Materials 4, 086105 (2016); 10.1063/1.4960621
A virtual instrument to standardise the calibration of atomic force microscope cantilevers
Review of Scientific Instruments 87, 093711 (2016); 10.1063/1.4962866
G-mode magnetic force microscopy: Separating magnetic and electrostatic interactions using big data
Applied Physics Letters 108, 193103 (2016); 10.1063/1.4948601
Interface modulated currents in periodically proton exchanged Mg doped lithium niobate
Journal of Applied Physics 119, 114103 (2016); 10.1063/1.4943934
Calibration of higher eigenmodes of cantilevers
Review of Scientific Instruments 87, 073705 (2016); 10.1063/1.4955122
Low-noise humidity controller for imaging water mediated processes
in atomic force microscopy
I. Gaponenko,a) L. Gamperle, K. Herberg,b) S. C. Muller, and P. Paruch
DQMP, University of Geneva, 24 Quai E. Ansermet, 1211 Geneva 4, Switzerland
(Received 25 April 2016; accepted 7 June 2016; published online 21 June 2016)
We demonstrate the construction of a novel low-noise continuous flow humidity controller and
its integration with a commercial variable-temperature atomic force microscope fluid cell, allowing precise control of humidity and temperature at the sample during nanoscale measurements.
Based on wet and dry gas mixing, the design allows a high mechanical stability to be achieved
by means of an ultrasonic atomiser for the generation of water-saturated gas, improving upon
previous bubbler-based architectures. Water content in the flow is measured both at the inflow and
outflow of the fluid cell, enabling the monitoring of water condensation and icing, and allowing
controlled variation of the sample temperature independently of the humidity. To benchmark the
performance of the controller, the results of detailed noise studies and time-based imaging of the
formation of ice layers on highly oriented pyrolytic graphite are shown. Published by AIP Publishing. []
In the field of materials science, the control of environmental conditions—such as humidity and temperature—
is essential for optimising the quality of measurements, and
thus for an improved understanding of the underlying physical
processes. The presence of water in air and its adsorption on
surfaces are a well known phenomenon, studied as much for its
fundamental interest1,2 as for its technological applications,3
where its effects on the adhesion, friction, and wear of solid
interfaces play a particularly significant role in nanotribology.
On silicon dioxide, a material of high industrial importance,
the evolution of the adsorbed water layer has been mapped out
in detail as a function of humidity,4,5 demonstrating a complex
behaviour with initial ice-like structure, followed by the
appearance of liquid-like water, leading to a non-trivial growth
of the layer thickness. Such studies have been determinant for
the development of new devices, such as sensors exploiting
the complex water interactions on their surfaces.6 However,
high precision humidity control is also crucial for studies in
other fields, such as at multiple observation scales in biology,
where the properties of organic samples can be greatly affected
by the presence of surface water or more broadly ambient
humidity. At the macroscopic scale of living organisms, for
example, the attachment ability of certain spiders is strongly
humidity dependent,7 while at the microscopic scale bacterial
cell morphological properties likewise change with humidity,8
and even at the nanoscale significant effects of surface water
have been shown in DNA—which at high humidities regains
mobility and exhibits conformational changes.9
The technique of choice for measurements of the nanoscale properties of materials and their interaction with surface
adsorbates is atomic force microscopy (AFM). By means of
a)Electronic mail:
b)Present address: LIFMET, École Polytechnique Fédérale de Lausanne,
1015 Lausanne, Switzerland.
a nanometrically sharp scanning probe tip, interactions such
as adhesion and friction have been studied as a function of
environmental conditions,3,10 showing distinct changes with
increasing ambient water content. In order to perform such
studies, the sample and scanning probe are usually isolated
in humidity-controlled cells, with constant relative humidity
(RH) mediated either by saturated saline solutions11,12 or by a
mixing flow of wet and dry gas.13–18 In the former case, only
a limited number of specific humidities is available due to
the dynamic evaporation of the saturated solution—requiring
changes in sample temperature to accommodate the missing
values. In the latter, although the humidity can be changed
continuously from essentially 0% RH to full saturation, the
usual bubbler design severely limits measurements of local
properties at the nanometer scale due to the mechanical noise
In this paper we describe a low-noise humidity control
apparatus tailored for scanning probe microscopy applications, where stability and low noise are essential. In order to
achieve this, whilst keeping the capability to vary the humidity
at the sample without changing its temperature or stopping the
data acquisition, we have developed a continuous flow mixing
humidifier using a water atomiser as the source of wet gas.
As a benchmark of the high performance of this design, we
show the noise characteristics of the humidifier integrated
with a commercial Asylum Research Cypher ES system and
demonstrate high resolution time-based imaging of ice growth
on highly oriented pyrolytic graphite (HOPG).
A. General design
The humidity controller design, shown in Fig. 1(a), is
based on a continuous flow mixing mechanism. The basic
idea behind it is the mixing of a dry and a wet gas flow, the
ratio of which is precisely determined for a target value of
87, 063709-1
Published by AIP Publishing.
Gaponenko et al.
FIG. 1. (a) Working principle of the humidity controller. A dry N2 flow
is split into two by means of a T-junction and injected into REG_DRY
and REG_WET regulators coupled to computer-controlled motors. The
REG_WET flow is saturated with H2O by means of an ultrasonic atomiser
and then recombined with the REG_DRY flow by means of a T-junction. The
combined flow then passes through a mixing chamber, MIX_INJ, where the
humidity and temperature are measured, and is then injected into the AFM
fluid cell. Finally, the output from the cell passes through a second mixing
chamber, MIX_EXH, for its humidity and temperature to be measured again
before it is exhausted out of the system. (b) Inside view of the humidity
controller showing the pneumatic cabling and the monitoring electronics.
humidity. In our system, dry N2 gas at a pressure of 4 bar
(≈60 psi) enters the apparatus and is split into two pathways
by means of a T-junction. The two resulting flows are regulated
by Porter VCD 1000 variable constant flow differential
controllers (REG_DRY and REG_WET) with interchangeable
flow elements—110 cc/min in our case, as we do not wish
to have too high a flow in the AFM—coupled to Trinamic
PD-1140 stepper motors for a precise computer control of
the mixing ratio. The REG_WET flow is water-saturated by
passing through the atomiser mechanism shown in Fig. 2 and
discussed below, and is then recombined with the REG_DRY
flow by means of the second T-junction. The recombined flow
is then injected into a passive mixing chamber (MIX_INJ)
designed to break the non-mixing laminar flows before
entering the AFM fluid cell. From here, the flow then travels
back into the humidity controller and passes through a second
passive mixing chamber (MIX_EXH) before exiting through
the exhaust. In both mixing chambers the temperature and
relative humidity of the flow are measured by means of
Sensirion SHT1519 sensors coupled to an Arduino Leonardo
board for data display and computer acquisition.
Because of the dual-sensor continuous flow design, the
controller can not only be easily adapted onto virtually
any enclosed volume, but the humidity can also be varied
programmatically without the need for a human interface.
Rev. Sci. Instrum. 87, 063709 (2016)
FIG. 2. (a) Schematic diagram of the atomiser unit, describing both the
dry/wet N2 exchange chamber and the electric wiring. The ultrasonic element power is controlled by means of the POWER_POT potentiometer and
can be switched on/off with the ATOM_SW switch. Active airflow cooling
composed of a heat-sink and two fans—FAN_IN and FAN_OUT—can be
switched on with the FAN_SW switch. The external pulse width modulation board can be plugged into the system with a 3.5 in. connector on
PWM_CONN and controls the PWM_RLY relay to minimise heating due
to the thermal dissipation of the piezoelectric bimorph. (b) Once turned on,
the ultrasonic element saturates the chamber with a dense, visible water mist.
(c) Outside view of the atomiser unit. The red pneumatic tubing brings the
dry N2 in and carries the H2O-saturated N2 out of the system.
Moreover, with the delocalised sensors, the temperature of
the probed sample can be varied without adverse effects
on the measurement and control of the humidity. This
modus operandi can therefore give additional insight into
the processes happening in the system under measurement—
mechanisms such as condensation and ice formation can be
investigated at a constant flow rate and water content of the
incoming gas, which can be calculated from the humidity
measured at MIX_INJ, with changes to the water content
monitored by a drop in the exhaust humidity at MIX_EXH.
B. Atomiser
For a mixing type humidifier, a steady supply of gas
fully saturated in H2O is necessary. In previous designs, this
requirement was usually fulfilled by the use of a water bubbler,
in which dry gas is saturated as it passes through liquid
H2O as bubbles.12–18 We initially tested such a system using
Supelco 64712-U/64834-U threaded midget bubblers as the
source of H2O-saturated gas.20 However, our tests showed
that although the required humidities were reachable, the
bubbler was found to induce unacceptable levels of mechanical
noise at the cantilever, resulting in periodic nanometer-scale
jumps in the topography as shown in Fig. 4(e) and discussed
below. To counteract this problem, a wet sponge filter was also
tested20 but due to its obstructions of the gas flow it could not
Gaponenko et al.
provide for fast enough stabilisation, although it is a viable
cost-effective alternative for non-time-critical measurements
where a fast response is not required. Thus, an atomiser design
was favoured, based on the 8 TDK NB-80E-01-H atomiser
unit. In this configuration, a piezoelectric bimorph is excited
at ultrasonic frequencies (2.35-2.6 MHz), generating a stream
of microdroplets above the water surface, fully saturating the
surrounding gas with H2O.
The design is shown in Fig. 2(a), with the ultrasonic
atomiser unit represented in orange. The latter produces a
dense water mist, as shown in Fig. 2(b), which saturates the
input dry N2 flux before it leaves the chamber, as schematically
shown by the blue exit tube. A splash guard is present in
order to avoid drops of water blocking the tubing. The density
of the water mist is controlled by means of a potentiometer
(POWER_POT), and the generation of mist can be switched
on or off mechanically (ATOM_SW). In order to avoid the
heating of the water in the chamber due to the thermal
dissipation of the piezoelectric bimorph, two countermeasures
were implemented. First, an optional active cooling system
composed of a heat sink and fans (FAN_IN and FAN_OUT)
can be switched on (FAN_SW). This provides partial thermal
dissipation in conjunction with a passive aluminium heat sink
increasing the thermal exchange between the atomiser unit
and the airflow. Second, because of the relatively low N2
flux and therefore small volume that needs to be maintained
at saturation, continuous atomising is unnecessary and even
counterproductive due to heat dissipation. Thus, a simple
pulse width modulation controller was implemented with an
Arduino Leonardo board in order to modulate the on/off state
of the atomisation by means of a relay (PWM_RLY) to achieve
the desired output.
C. Integration and characterisation
For tests as well as actual measurements, the controller
was integrated into a commercial Peltier cooler/heater fluid
cell mounted on an Asylum Research Cypher ES atomic force
microscope. This setup is capable of delivering temperatures
between −8 ◦C and +120 ◦C (with a software controlled closed
loop feedback allowing for a reliable and stable sample
temperature between 0 ◦C and 100 ◦C). By virtue of this setup,
we could harness the capability of independent control over
both temperature and humidity at the sample.
First, noise measurements were performed in order to
characterise the effect of the humidity controller on the
mechanical stability of the atomic force microscope. A thin
film of monocrystalline epitaxial PbTiO321 was imaged under
four different conditions: no flow, dry gas flow, atomiser
wet flow, and bubbler wet flow.22 The results, shown in
Figs. 3(a)-3(d), demonstrate the mechanical stability of the
atomiser compared to the classical bubbler design. Images
taken with the atomiser in operation (Fig. 3(c)) show identical
topography to those performed with no flow on dry gas, with a
subnanometer resolution allowing unit cell high terraces (∼4 Å
high and atomically flat) in the film to be clearly distinguished.
In contrast, as can be seen in Fig. 3(d), the pressure waves
generated by the bubbler can be seen as periodic ≈5 nm jumps
in the measured topography. To confirm this observation, the
Rev. Sci. Instrum. 87, 063709 (2016)
FIG. 3. Topography of an epitaxial PbTiO3 thin film as imaged by low
amplitude AM-AFM during (a) no flow, (b) dry gas flow, (c) atomiser wet
flow, and (d) bubbler wet flow. As can be seen, the bubbler induces periodic
≈5 nm jumps in the topography. (e) The spectral response of the vertical
deflection of the fully retracted AFM tip, measured by laser reflection onto a
quadrant photodetector, demonstrates the periodic nature of the mechanical
noise due to bubbler-induced pressure waves, which is absent in identical
measurements carried out on the closed fluid cell with no flow, during dry
gas flow, as well as a wet flow generated by the atomiser.
vertical deflection signal of the AFM photodetector was then
acquired as a function of time, with the tip retracted 50 µm
from the surface. The resulting spectral response is shown in
Fig. 3(e). As highlighted on the spectrum (red), the bubbler
indeed generates a periodic noise depending on the frequency
at which bubbles are expelled from the nozzle, whereas no
visible difference can be seen between the spectra produced
when the measurement is performed with the atomiser, with
dry flow, or with no flow through the fluid cell.
D. Control and measurement
The measurement of humidity in the injection chamber
MIX_INJ and the exhaust chamber MIX_EXH is performed by Sensirion SHT15 sensors interfaced to an Arduino
Leonardo board. The sensors are polled once per second, and
the resulting humidities and temperatures are displayed on
the onboard LCD screen for manual operation. Ideally, the
Arduino is connected to a computer and outputs the data
acquired from the sensors in addition to its other functions.
Thus, the humidities and temperatures can be saved, plotted,
and used for active closed loop control.
Gaponenko et al.
FIG. 4. Relative humidity and temperature measured in the injection and
exhaust mixing chambers as a function of time during a six day experiment.
As can be seen from the time axis, the humidities were cycled five times up to
∼85% RH during the measurement. The small changes in temperature reflect
the variations in ambient conditions in the laboratory. The increase in the
baseline humidity as a function of time can be explained by the incomplete
desorption of water in the tubing. Finally, the difference in levels between
the humidities measured in the injection and exhaust chambers is due to the
significant system load, composed of not only the fluid cell but also over two
meters of tubing.
The control of the humidity itself is done by adjusting
the two high precision 14-turn flow regulators REG_DRY
and REG_WET, allowing for a good finesse in the control
of dry and wet gas flow between 0 and 110 ml/min each—a
range which can be changed by replacing the flow elements
in the regulators to adjust for the system size and required
finesse.23 Although they can be adjusted manually, one of
the main advantages of the controller design is that the two
motors are coupled to the regulators in order to allow for
automatic control. The motor control is performed by a set
of LabVIEW scripts, which range from simple open-loop
control to closed-loop regulation of humidity, depending on
Rev. Sci. Instrum. 87, 063709 (2016)
the desired degree of precision and responsiveness. Moreover,
it is also possible to move the two motors simultaneously
in opposite directions in order to maintain the constant
110 ml/min flow, if that is an important parameter for a
particular measurement.
An example of open-loop operation is shown in Fig. 4,
with the relative humidities in black/orange and temperatures
in red/blue plotted as a function of time for both the injection
and exhaust chambers, respectively. These measurements
demonstrate that the humidity controller can be operated for
durations of several days with continuously cycled humidity.
In this experimental run, we observed an increase in the
baseline level of relative humidity at the end of each sweep
when the REG_WET is fully closed. This is a result of
the incomplete desorption of water in the tubing and the
mixing chambers during the descending humidity ramps.
Over multiple cycles, this additional water in the system
increased the baseline humidity to 40%–50%—but this effect
can be counteracted by allowing the system to stabilise
longer during the return to low humidities, or by the use of
higher flow elements to improve desorption rates. Finally, the
difference between the injected and exhausted humidities can
be explained by the large load on the system composed of the
atomic force microscope fluid cell as well as over two meters
of tubing.
In order to assess the practicality of the humidity
controller for real atomic force microscopy measurements,
low-amplitude AM-AFM was performed on a flake of highly
oriented pyrolytic graphite (HOPG). The flake was first pasted
with silver paint onto a standard metallic sample holder. It
was then inserted in the AFM fluid cell, micromechanically
cleaved in a pure nitrogen flow, and heated for several minutes
to 120 ◦C at ∼5% RH in order to partially remove the remaining
surface water. An initial increase in exhaust humidity at
MIX_EXH was observed during this process, followed by a
FIG. 5. Selected images from continuous measurement of ice layer formation on HOPG, illustrating the growth of ice layers. Each image is 3 × 3 µm2. The top
row is the topography, the middle row is the phase, and the bottom row is the amplitude signal. Ice layers can be seen growing laterally on the surface, with the
most visible contrast in the phase images. The thickness of this ice is ≈0.5 nm, as measured from the topographic images.
Gaponenko et al.
return to lower values, indicating the efficient water desorption
from the graphite. Then, after a short initial stabilisation
period, the atmosphere in the cell was almost fully saturated
with water to ∼85% RH. The temperature was then lowered
from 22 ◦C to −7 ◦C at constant input humidity measured at
MIX_INJ. A drop in the exhaust humidity at MIX_EXH was
observed, characteristic of condensation processes. When the
temperature at the sample had stabilised, continuous amplitude
modulation AFM measurements were performed on a clean
area of the sample.24
The formation of an ice layer was observed in the form
of thin nanoscale layers growing in size as a function of time.
The dynamics of the process were imaged by continuous
high speed AFM scans acquired for an area of 3 × 3 µm2,
with a speed of 256 s/scan. The growth of the ice layer,
shown in Fig. 5, is primarily lateral, and is concentrated
around nucleation sites distributed randomly in the scan.
The thickness of the ice structures formed is measured to
be ∼0.5 nm. Moreover, the growth seems to be constrained
by the topographic features, with propagation stopping at
graphitic step edges. These results are in accordance with the
hydrophobic nature of freshly cleaved HOPG, and with the
previous studies of icing on this material,25 where comparable
thickness layered growth was observed with similar edgelimited dynamics.
We note that throughout these measurements we were able
to maintain a high subnanometer resolution, with extremely
low mechanical nose, as can be seen in both the topography and
amplitude images. Such performance was not possible with the
bubbler as the source of saturated gas, where the topography
signal in similar measurement was unexploitable—leaving
only the phase contrast in order to distinguish the growth
of the ice layers.20
In this paper we have described a novel continuous
flow humidifier design for noise-critical applications, with
a particular focus on atomic force microscopy. We have
shown that with the motorised control, relative humidity
can be varied continuously over the full range with good
finesse and stability. Moreover, we have introduced a dual
sensor design which allows the possibility of monitoring
more complex water-mediated processes such as condensation
and icing by comparing water content of flows injected
into and exhausted from the experimental cell. Comparative
tests of noise measurements with this atomiser-based design
have demonstrated its clear advantage over the mechanically
far more noisy bubbler-based system, allowing an imaging
resolution comparable to that of a closed cell, or a cell
under dry flow only. Finally, we were able to observe the
formation of ice on HOPG at low temperatures during
continuous wet gas flow, showing extremely low noise
levels and high stability of the system over time-dependent
Our system paves the way towards further nanoscale
studies on water-sensitive materials, such as those in the
semiconductors industry, or towards the investigation of
Rev. Sci. Instrum. 87, 063709 (2016)
biological processes under various environmental conditions. For the latter, the computer controlled design allows
for continuous measurements as a function of humidity
without adding potentially stressful external perturbations
to the system under measurement as done previously.8,9
Finally, the flexibility of our design does not limit it to
the combination of nitrogen and water, and different gases
and liquids could be used in order to study their interaction
with materials—for example, to probe in real time the
response of bacterial growth dynamics to the surrounding
This work was supported by the Swiss National Science
Foundation under Division II Grant No. 200021-153174. The
authors also thank A. Verdaguer and A. Gruverman for helpful
discussions. K.H. and L.G. acknowledge the MaNEP student
internship programme.
Verdaguer, G. M. Sacha, H. Bluhm, and M. Salmeron, Chem. Rev. 106,
1478 (2006).
2G. Rubasinghege and V. H. Grassian, Chem. Commun. 49, 3071 (2013).
3A. J. Barthel, A. Al-Azizi, N. D. Surdyka, and S. H. Kim, Langmuir 30, 2977
4D. B. Asay and S. H. Kim, J. Phys. Chem. B 109, 16760 (2005).
5A. Verdaguer, C. Weis, G. Oncins, G. Ketteler, H. Bluhm, and M. Salmeron,
Langmuir 23, 9699 (2007).
6H. Bi, K. Yin, X. Xie, J. Ji, S. Wan, L. Sun, M. Terrones, and M. S.
Dresselhaus, Sci. Rep. 3, 2714 (2013).
7J. O. Wolff and S. N. Gorb, Proc. R. Soc. B 279, 139 (2012).
8H. Nikiyan, A. Vasilchenko, and D. Deryabin, Int. J. Microbiol. 2010,
9D. J. Billingsley, J. Kirkham, W. A. Bonass, and N. H. Thomson, Phys.
Chem. Chem. Phys. 12, 14727 (2010).
10L. Chen, X. Gu, M. J. Fasolka, J. W. Martin, and T. Nguyen, Langmuir 25,
3494 (2009).
11A. Wexler and S. Hasegawa, J. Res. Natl. Bur. Stand. 53, 19 (1954).
12O. Nakamura, I. Ogino, and T. Kodama, Rev. Sci. Instrum. 50, 1313
13J. M. Maxwell and M. G. Huson, Rev. Sci. Instrum. 73, 3520
14J. Lievonen, K. Ranttila, and M. Ahlskog, Rev. Sci. Instrum. 78, 043703
15A. Jackson, K. M. Beers, X. C. Chen, A. Hexemer, J. A. Pople, J. B. Kerr,
and N. P. Balsara, Rev. Sci. Instrum. 84, 075114 (2013).
16P. W. Majsztrik, A. B. Bocarsly, and J. B. Benziger, Rev. Sci. Instrum. 78,
103904 (2007).
17D. S. de la Cruz, J. G. Denis, M. D. Griffith, D. R. King, P. A. Heiney, and
K. I. Winey, Rev. Sci. Instrum. 83, 025112 (2012).
18O. Stukalov, C. A. Murray, A. Jacina, and J. R. Dutcher, Rev. Sci. Instrum.
77, 033704 (2006).
19The Sensirion SHT15 is a surface mountable combined capacitive
relative humidity and bandgap temperature sensor. The humidities can
be measured between 0% and 100% RH with an absolute accuracy of
±2%. Temperatures can be measured beween −40 ◦C and 123.8 ◦C, with a
precision of ±0.3 ◦C.
20See supplementary material at for a
detailed discussion on passive humidification devices, the bill of materials
and HOPG icing imaged with a bubbler mechanism.
21Sample provided by Christian Weymann at the Department of Quantum
Matter Physics, University of Geneva. The growth was performed by
off-axis RF magnetron sputtering with the following conditions: LaNiO3
back electrode 510 ◦C 180 mTorr 10:35 O/Ar 50 W, PbTiO3 (target with
1060 W).
22Measurements were performed with a ASYELEC-01 tip (42.4 nm/V
sensitivity, 1.81 N/m spring constant) in amplitude modulation AFM with
a free amplitude of 100 mV and a setpoint of 60 mV, corresponding to
Gaponenko et al.
calibrated amplitudes of 4.2 nm and 2.5 nm, respectively. The 1 µm x
1 µm area was scanned at three lines per second, with a 512 × 512 pixel
23Available flow elements for the Parker VCD-1000 variable constant
differential flow controller span full scale flow rates from 5 to 1500 sccm,
a range that can be adapted to the volume of the environmental cell under
24Measurements were performed with a ASYELEC-01 tip (42.4 nm/V
sensitivity, 1.81 N/m spring constant) in AM-AFM with a free amplitude
Rev. Sci. Instrum. 87, 063709 (2016)
of 100 mV and a setpoint of 50 mV, corresponding to calibrated amplitudes
of 4.2 nm and 2.1 nm, respectively. The 3 µm × 3 µm area was scanned
at one line per second, with a 256 × 256 pixel resolution.
25Y. Zheng, C. Su, J. Lu, and K. P. Loh, Angew. Chem., Int. Ed. 52, 8708
26U. Herbert, S. Rossaint, M. Khanna, and J. Kreyenschmidt, Poult. Sci. 92,
1348 (2013).
27J. L. Etchells, T. A. Bell, R. N. Costilow, C. E. Hood, and T. E. Anderson,
Appl. Microbiol. 26, 943 (1973).
Без категории
Размер файла
7 278 Кб
Пожаловаться на содержимое документа