Accepted Manuscript

Title: Fabrication and testing of nanoporous Si3 N4 optical

filters for gas sensing applications

Author: <ce:author id="aut0005" biographyid="vt0005"> R.

Vernhes<ce:author id="aut0010" biographyid="vt0010"> J.E.
Klemberg-Sapieha<ce:author id="aut0015"
biographyid="vt0015"> L. Martinu

PII: S0925-4005(13)00497-8
DOI: http://dx.doi.org/doi:10.1016/j.snb.2013.04.064
Reference: SNB 15430

To appear in: Sensors and Actuators B

Received date: 11-2-2013

Revised date: 9-4-2013
Accepted date: 15-4-2013

Please cite this article as: R. Vernhes, J.E. Klemberg-Sapieha, L. Martinu, Fabrication
and testing of nanoporous Si3 N4 optical filters for gas sensing applications, Sensors &
Actuators: B. Chemical (2013), http://dx.doi.org/10.1016/j.snb.2013.04.064

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
*Manuscript

Fabrication and testing of nanoporous Si3N4 optical

filters for gas sensing applications

t
ip
R. Vernhes, J.E. Klemberg-Sapieha, and L. Martinu ∗

cr
Regroupement québécois sur les matériaux de pointe (RQMP) and Department of
Engineering Physics, Polytechnique Montréal, P. O. Box 6079, Station Centre-Ville,
Montréal, Québec, H3C 3A7, Canada

us
an
Abstract

We demonstrate the concept of nanoporous multilayer optical filters as optical gas sensors,
based on plasma-enhanced chemical vapor deposited Si3 N4 films with controlled poros-
M
ity. In situ real-time spectroscopic ellipsometry, associated with the Bruggeman effective
medium approximation is employed for characterizing the film microstructure, particularly
the total volume fraction of porosity ( fv ), and the fraction of open porosity ( fo ). In order to
do so, the effective refractive index (neff ) of the as-grown films is evaluated both in vacuo
d

and when exposed to ethanol vapor. We find that fv = 34% and fo = 27% for porous films,
te

while fv = 13% and fo = 2% for dense films. Additionally, adsorption isotherms of ethanol
are performed and the resulting curves are interpreted using the Kelvin equation as well
as the Dubinin-Radushkevich model, allowing us to estimate the average micropore size
p

(L ≈ 2 nm) and the micropore/mesopore volumes (46%/54% of fo ). Infrared ellipsome-

try is then applied for the characterization of the film chemical structure and, particularly,
ce

for the study of heat treatment-induced chemical stabilization. Finally, porous/dense Si3 N4
quarterwave stacks are implemented in a Fabry-Perot filter that is tested as a prototype sen-
sor in a simple configuration. We observe high sensitivity to model vapors such as ethanol
(10 ppm) and ammonia (28 ppm), as well as fast response times (≈ 10 s), good repro-
Ac

ducibility, and complete signal regeneration after heat treatment.

Key words: Gas sensors, Optical sensors, Silicon compounds, Porous materials,
Porosimetry, Plasma deposition, Ellipsometry

∗ Contact author: Ludvik Martinu

Email address: Ludvik.Martinu@polymtl.ca
Tel.: 514-340-4099
Fax: 514-340-3218

Preprint submitted to Elsevier 9 April 2013 Page 1 of 29

 1 Introduction

During the past decade, constant improvements in the control and fabrication of
nanostructured materials have stimulated the development of a multitude of (bio)-
chemical sensors with increased sensitivity and selectivity, as well as miniaturized
5 dimensions [1]. Among the various classes and concepts, optical sensors are of

t
particular interest because the use of light to probe analytes and to carry information

ip
offers many advantages [2]; in particular, these are:

cr
- the ability to operate in conductive or spark-sensitive environments without dis-
turbance [3], or to work without any contact with the probed medium [4];
10 - the immunity to electromagnetic interference, cosmic rays or radioactivity [5];

us
- the possibility to combine the sensors with optical fibers in order to create an
extended sensor network or to perform in vivo investigations of blood or tissues
with minimal invasion [6].

15

an
Furthermore, using porous materials in optical sensors allows one to benefit from
large surface areas in small volumes, hence considerably increasing the sensitiv-
ity of the devices. Some research groups proposed to implement electrochemically
M
etched porous silicon in multilayer optical interference filters, in which the low
and high refractive indices (nL , nH ) are obtained by varying the amount of porosity
along the thickness [7–11]. However, such a method presents important drawbacks,
d

20 particularly the difficulty to control and monitor the fabrication process, the chem-
ical instability of porous silicon over time as well as the strong absorption in the
te

visible spectrum.

Vacuum deposition processes offer more flexibility as they allow one to deposit
p

many types of materials onto different substrates [12]. In addition, precise control
ce

25 of film thickness can be achieved by using in situ real time monitoring techniques
[13], leading to the fabrication of high quality multilayer filters with specific optical
characteristics. An interesting example of vacuum-fabricated porous materials is
optical films prepared by glancing-angle deposition (GLAD), namely porous TiO2 -
Ac

based rugate filters [14,15]. However, this deposition technique depends on the
30 complex rotation of the substrate relative to the vapor flux, an approach that may
render the scale up rather complicated.

In our group, we have developed a method based on plasma-enhanced chemical

vapor deposition (PECVD) [12,16] in order to control the film microstructure.
We use a dual-mode microwave/radiofrequency (MW/RF) plasma approach, where
35 the substrates are placed on the RF-powered electrode while being exposed to the
microwave discharge. This technique allows us to sensitively control the plasma-
surface interactions during the film growth by selectively adjusting the flux and
energy of the incoming ions and other chemical species, leading to films with a
tunable microstructure [16–18]. More specifically, we have shown that (i) MW

2 Page 2 of 29
40 plasmas present high ion fluxes (φi ) with low ion energies (E i < 10 eV), favor-
ing the formation of porous microstructures, in agreement with the structure zone
model [19,20]; (ii) RF plasmas exhibit lower φi and high E i (from 10 to 1000 eV),
giving rise to dense films. Intermediate porosity can be achieved by mixing the two
modes with an appropriate control of MW power and RF-induced substrate bias
45 (VB ).

t
Using this method, we demonstrated the fabrication of porous/dense (nL = 1.6,

ip
nH = 2.0 at 550 nm) Si3 N4 optical filters, based on both multilayer (quarterwave
stacks with abrupt interfaces) or rugate (continuous refractive index depth profiles)

cr
designs [21,22]. Coatings were stoichiometric, showing low absorption and low
50 scattering in the visible spectrum as well as good mechanical stability. Furthermore,
this low-temperature PECVD process is compatible with sensitive substrates such

us
as plastics [12,23].

In order to be implemented in optical sensing devices, porous materials, such as

an
nanoporous Si3 N4 , have to meet specific requirements. In particular, the pores have
55 to be interconnected, so that the analyte can easily penetrate across the whole thick-
ness of the filter and thus modify its optical characteristics. Moreover, porous films
have to be chemically stable and show no drift of their optical properties with time.
M
These aspects will be discussed in the first part of this study, devoted to the char-
acterization of the microstructure and chemistry of porous Si3 N4 films. The second
60 part will focus on the fabrication of a Fabry-Perot (FP) porous/dense optical filter
and its testing as an all-optical gas sensor prototype.
d
te

2 Experimental methodology
p
ce

Films were prepared in a turbomolecularly pumped vacuum chamber equipped with

microwave (MW, 2.45 GHz) and radiofrequency (RF, 13.56 MHz) plasma sources
65 (Fig. 1) [21]. Substrates were mounted onto the RF electrode-substrate holder us-
ing a load lock system. The MW power was applied through a fused silica window
Ac

to which a MW slotted waveguide applicator was attached. The substrate self-bias

voltage (VB ) developing on the RF electrode and the MW power (PMW ) were ad-
justed using a LabView interface that independently controlled the RF (Advanced
70 Energy) and MW (Hüttinger, pulsed at 60 Hz with a duty cycle of 10%) generators.

Silicon wafers were used as substrates for films devoted to ellipsometric charac-
terization, while microscope glass slides, washed with soap and rinsed with dis-
tilled water, were employed for the deposition of optical filters. Nitrogen and silane
were used as working gases for the fabrication of amorphous hydrogenated Si3 N4
75 coatings, with partial pressures of 22.5 and 5 mTorr, respectively, and a total work-
ing pressure of 40 mTorr. The system was operated in pure MW plasma (PMW =
100 W) to deposit porous films, and in dual-mode MW/RF plasma (PMW = 100 W,

3 Page 3 of 29
 VB = −300 V) to obtain dense films.

Although all the films discussed in this study present some level of porosity, we call
80 “dense films” those with low porosity and high refractive index (nH ), to distinguish
them from highly porous films possessing a low refractive index (nL ).

t
Film growth and vapor sorption experiments were monitored in situ by real-time

ip
spectroscopic ellipsometry (RTSE, M2000, J. A. Woollam Co., Inc.), which was
used to determine the thickness (d) and the effective refractive index (neff ) of indi-
vidual layers and stacks. Spectra were acquired at an incidence angle of 64◦ within

cr
a wavelength range of 246-1650 nm. The optical model consisted of the silicon
substrate, the stack of porous and dense layers, and the top surface roughness layer.

us
The neff dispersion of porous and dense layers was modeled by the Cauchy formula

Bn C n
neff (λ) = An + + , (1)

an
λ2 λ4

where An , Bn and Cn are fitting parameters and λ is the wavelength. Extinction coef-
M
ficient (keff ) of Si3 N4 films was below the sensitivity of ellipsometric measurements
in the visible range (keff ≤ 10−4 ).

Composite materials such as porous films (Si3 N4 /void) can be modeled using the
d

Bruggeman Effective Medium Approximation (BEMA)

te

1 − n2eff n2b − n2eff

fv + (1 − fv ) = 0, (2)
1 + 2n2eff n2b + 2n2eff
p
ce

85 where fv is the volume fraction of voids and nb is the refractive index of bulk silicon
nitride [24]. Eq. 2 allowed us to calculate fv from neff , considering that nb = 1.90,
which is the highest value obtained in our RF deposition system for hydrogenated
Ac

dense and stoichiometric Si3 N4 films. The fact that nb is lower than the standard
2.02 value generally considered for thermal CVD Si3 N4 is explained by the pres-
90 ence N–H bonds (see section 3.2) that broaden the gap and decrease the refractive
index of SiN x :H films [25,26]. Eq. 2 was also used to model the neff of the surface
roughness layer, by fixing fv = 50% and by setting nb = neff of the underlying
Si3 N4 layer.

Adsorption isotherms of ethanol were obtained by ellipsometric porosimetry [27]

in order to evaluate the pore characteristics of Si3 N4 films, such as the fraction of
open porosity ( fo ), the pore size distribution (PSD) and the average pore size (L). fo
was calculated by assuming that when the saturation pressure of ethanol is reached,
open pores in the film are completely filled with ethanol (n = 1.36 at 550 nm at

4 Page 4 of 29
 room temperature). Therefore, by adding a third term to Eq. 2, we get:

1.362 − n2eff 1 − n2eff

fo + ( fv − fo )
1.362 + 2n2eff 1 + 2n2eff
n2b − n2eff
+ (1 − fv ) = 0. (3)
n2b + 2n2eff

t
Additionally, considering that neff varies across a limited range during adsorption

ip
isotherms (typical variation of 0.1 for porous films), Eq. 3 can be very well approx-
imated by a linear relationship:

cr
neff = A + B fo, (4)

us
where A = neff at pr = 0, and B = 0.376 in the case that silicon nitride is the
95 adsorbent and ethanol is the adsorbate.

an
The PSD was calculated using Eq. 3 combined to the Kelvin equation:
−2γVL 1
log(pr ) = , (5)
RT rm
M
where pr is the relative pressure of ethanol, γ and VL are the surface tension and
molar volume of ethanol, R and T are the ideal gas constant and temperature, and
rm is the radius of curvature of the meniscus for a given pr at equilibrium [28]. rm
d

is related to the real pore size by

te

rm = r − ta , (6)

where r is the pore radius and ta is the thickness of the initial adsorbed film before
p

capillary condensation occurs.

ce

The Dubinin-Radushkevich (DR) model [28] was used in order to evaluate the av-
erage micropore size (L). The DR model assumes a Gaussian distribution of the
pore size, related to the degree of pore filling V/V0:
Ac

 A 2 
 !
V
= exp −  . (7)
V0 E

Here, V is the adsorbed volume, V0 is the total volume of open pores, A is the
adsorption potential and E is the adsorption energy. A is derived from the theory
of adsorption developed by Polanyi [29], and it can be linked to pr by

A = RT log(1/pr ), (8)

In Eq. 7, E depends on the nature of the adsorbing vapor, and it is defined by

E = βE0 , (9)

5 Page 5 of 29
 where β is the affinity characteristic of the adsorbate, and E0 is the energy charac-
teristic of the adsorbent. Dubinin chose benzene as the standard, and thus β = 1 for
benzene, while β = 0.61 for ethanol. Finally, E0 can be linked to the average pore
size using the Dubinin-Stoeckli (DS) equation [30]:

L/2 (nm) = K/E0 (kJ mol−1 ), (10)

t
with 12 ≤ K ≤ 14 kJ. By combining Eqs. 7, 8, 9 and 10, one obtains:

ip
!2
RT L
log(V/V0 ) = − log2 (1/pr ). (11)
2βK

cr
According to Eq. 11, the slope of log(V/V0 ) versus log2 (1/pr ) allows one to deter-

us
mine the L value.

100 The effect of heat treatment on the chemical characteristics of porous Si3 N4 films
was studied by infrared spectroscopic ellipsometry (IR-VASE, J. A. Woollam Co.,

an
Inc.). Prior to spectra acquisition, samples were mounted in an environmental cham-
ber compatible with the ellipsometer, which enabled us to perform measurements
under vacuum or in controlled atmosphere. In addition, the chamber was equipped
with a heating stage (20 to 300◦ C) and with KBr windows transparent in the in-
M
105

frared range. Ellipsometric spectra were acquired at an incidence angle of 60◦ over
a wavenumber range of 300-7000 cm−1 , and the spectral resolution was 16 cm−1 .
d

The configuration of the above chamber also allowed us to carry out direct optical
transmission measurements through the sample filters and thus test their perfor-
te

110 mance as gas sensors (Fig. 2). Such measurements were performed either spectro-
scopically using a white light source and a spectrometer (USB2000, Ocean Optics),
or at a single wavelength using a laser diode emitting at 655.2 nm and a photode-
p

tector. Sensitivity of the model optical filter-sensor was tested for N2 , NH3 , ethanol
ce

and air.
Ac

115 3 Results and discussion

3.1 Microstructural characteristics of porous Si3 N4 films

Our first objective was to determine the plasma parameters that would lead to Si3 N4
films with two distinct microstructures, namely different proportions of porosity
and thus different neff values that could eventually be used as nL and nH in optical
120 multilayer filters. In addition, “dense” films used as nH layers should be porous
enough to allow the diffusion of gases across the whole coating thickness. Applying
the conditions described in section 2, the RTSE measurements performed in vacuo
resulted in neff = 1.58 for porous films and neff = 1.78 for dense films at 550 nm.

6 Page 6 of 29
 Using Eq. 2 with nb = 1.90, we evaluated fv to 34% and 13% for porous and dense
125 films, respectively.

We deposited a three layer stack consisting of a porous layer (d = 85 nm) sand-

wiched between two dense layers (d = 102 and 62 nm) in order to verify the con-
nectivity of the pores and to measure the void fraction of open porosity ( fo ) in the
films. neff was evaluated for each as-deposited layer of the stack in vacuo, as well

t
as after admission of ethanol in the vacuum chamber at the saturation vapor pres-

ip
130

sure (p ≈ 48 Torr at 22◦ C). The results summarized in Table 1 show that neff is
higher in ethanol than in vacuo for all three layers. This confirms that sorption of

cr
ethanol occurred in the films, and that the pores are connected from top to bottom of
the 3-layer system. The neff increase was significantly higher for the porous central
135 layer (1.582 to 1.686) due to its higher porosity. Using Eq. 3 with three components

us
(Si3 N4 , void, ethanol), we were able to determine fo (see Table 1). One can observe
that most of the pores in the porous layer are connected (27% over 34%), while fo is
very small for the dense films (3% over 10%). We also noticed that the thickness of

an
the entire coating stack had increased by 1.6% when saturation was reached, prob-
140 ably due to both adsorption of ethanol at the coating surface and possibly swelling
of the films [31].
M
The fact that the two bottom and top dense layers exhibit different neff in vacuo does
not necessarily mean that the deposition process is not reproducible, but rather that
the nature of the underlying substrate (silicon or porous film) has a strong impact
d

145 on the microstructure of the growing film. This aspect was studied in detail in a
previous work, and various phenomena occurring during the growth of a dense film
te

on a porous one were discussed there [22].

In order to obtain more information about the characteristics of the pores, we de-
p

termined adsorption/desorption isotherms by measuring neff of a porous film versus

ce

150 the relative vapor pressure (pr ) of ethanol (see Fig. 3). According to the literature
[28], the adsorption curve presents the characteristics of both microporous (type I
isotherm, pores with a size below 2 nm) and mesoporous (type IV isotherm, pore
with a size between 2 and 50 nm) materials: (i) for 0 ≤ pr ≤ 0.1, the abrupt in-
Ac

crease of neff is caused by filling of the micropores due to their high attracting
155 potential, and by covering the mesopore surface with a monolayer of adsorbate; (ii)
for 0.1 ≤ pr ≤ 0.7, slow neff increase is due to the progressive multilayer coverage
of the mesopores; (iii) for pr ≥ 0.7, capillary condensation occurs, which leads to
the filling of the mesopores and to the rapid increase of neff . The fact that the desorp-
tion curves are shifted upwards and do not return to the starting neff values indicates
160 that chemisorption occurred. This also explains why the hysteresis loop resulting
from capillary condensation at pr ≥ 0.7 is hardly visible. The presence of meso-
pores agrees well with prior SEM micrographs (not presented here) that showed a
columnar microstructure with a column width of tens of nanometers [22].

Fig. 4 shows the pore size distribution (PSD) calculated from the sorption curve in

7 Page 7 of 29
165 Fig. 3 using the Bruggeman approximation and the Kelvin equation (Eqs. 3 and 5).
The sorption curve was preferred to the desorption one since in the latter case, mi-
cropores are not completely emptied at pr = 0, and therefore some information
about the finer pores could be missing. Note also that the PSD was plotted in loga-
rithmic scale along the x axis in order to better see the features in the microporous
170 range. Following the pore size definition provided in the aforementioned paragraph,
one finds that the proportion of micropores (62%) is larger than that of mesopores

t
(38%). The PSD in the micropore range presents a narrow distribution with a max-

ip
imum at rm = 3 Å, while the PSD in the mesoporous range is broad and featureless.
Unfortunately, we cannot rely on the PSD to get a precise value of the average size

cr
175 of micropores (L), firstly because rm represents an underestimated value of the real
pore size (see Eq. 6), and secondly because the Kelvin equation is no more valid in
this pore size range [28].

us
Consequently, the Dubinin-Radushkevich (DR) model, which was specifically de-
veloped for microporous materials, was rather employed (Eq. 11) and L could be

an
180 evaluated from the plot of log(V/V0) versus log2 (1/pr ) as shown in Fig. 5. The
fact that we obtain a linear relation for log2 (1/pr ) > 2 proves that the DR model
is applicable to our microporous Si3 N4 films. However, the model is not valid for
pressures close to saturation, when capillary condensation in mesopores dominates.
M
Using Eq. 11 with 12 ≤ K ≤ 14 kJ, one obtains L values between 1.5 and 1.7 nm
185 from the slope of the line in Fig. 5. In addition, the intercept of the line with the y
axis indicates the volume proportion of micropores/mesopores (46%/54%) relative
d

to fo . We think that these values are more accurate than the estimation obtained
using the Kelvin equation (i.e. 62%/38%) because the former relies on the shape of
te

the isotherm, which is linked to the adsorption mechanism and therefore depends
190 on the pore size, while the latter is based on a rigid and conventional definition of
the size limit between micropores and mesopores.
p
ce

3.2 Chemical characteristics of porous Si3 N4 films

Ac

In addition to microstructural analysis, we also determined the chemical character-

istics of Si3 N4 films, as well as their behavior during heat treatment. In particular,
195 evaluation of the chemical stability of the coatings is key to the performance of
optical sensors. As an example, we present in Fig. 6 the optical constants (n and k)
of a porous film with fv = 34% measured by IR-VASE under different conditions
controlled in the environmental chamber described in section 2.

Firstly, the porous film was characterized in air immediately after deposition. As ex-
200 pected, the k spectrum exhibits absorption peaks related to the chemical bonds char-
acteristic of PECVD SiN x :H films [32,33]: two broad bands at 490 and 925 cm−1
due to Si–N symmetric and asymmetric stretching modes, and two less intense
bands at 1170 and 3330 cm−1 related to N–H bending and stretching modes, re-

8 Page 8 of 29
 spectively. The two small peaks at 1540 and 1640 cm−1 are assigned to NH2 and
205 H2 O scissors bending modes, respectively, the latter one pointing to water adsorbed
in the pores. Let’s note that the H2 O stretching mode that should arise at 3300 cm−1
is entirely convoluted with the N–H stretching mode. In addition, a weak shoulder
corresponding to the Si–O stretching mode can be seen around 1040 cm−1 , which
indicates that film oxidation started right after deposition.

t
ip
210 After film exposure to ambient air for 30 min, the chamber was pumped down,

cr
ellipsometric spectra were acquired, and n and k spectra were evaluated (Fig. 6,
second spectrum from bottom). The k spectrum reveals that partial oxidation con-
tinued in vacuo, as indicated by the increase of the Si–O peak at 1040 cm−1 and by

us
the rise of a small shoulder around 830 cm−1 that is attributed to the Si2 O bending
215 mode. This is accompanied by a decrease of the peaks at 3330 cm−1 and 1640 cm−1
that are attributed to water desorption from the pores. However, some water is likely
to remain chemisorbed on the surface of the pores and to contribute to the film ox-

an
idation process, as documented by the results from Fig. 3. In addition, an overall
n decrease is observed in Fig. 6, which is explained by both water desorption and
220 oxidation of the film.
M
In the next step, n and k spectra were acquired at 200◦ C in vacuo. Further increase of
d

the Si–O peaks at 830 and 1040 cm−1 in the k spectrum is attributable to continued
thermally activated oxidation. The peak at 1640 cm−1 disappeared, indicating that
te

water entirely desorbed from the pores. This also means that the remaining peak at
225 3330 cm−1 represents the sole proportion of N–H bonds in the film. Again, the de-
p

crease of n is due to desorption of adsorbates and further oxidation. No significant

swelling or shrinkage was observed during heating, indicating that no microstruc-
ce

tural variation occurred.

Ac

Finally, the sample was cooled down and ellipsometric spectra were acquired at
230 ambient temperature in vacuo. One can see that the Si–O peaks remained constant,
proving that heat treatment is an effective way to cure and stabilize the porous film
chemistry. The peak at 3330 cm−1 increased and the one at 1640 cm−1 reappeared,
revealing that some water adsorbed again in the film. In addition, a small peak
observed at 2950 cm−1 is related to the presence of hydrocarbons in the pores. Both
235 water and hydrocarbon adsorption is indicative of possible contamination of the
vacuum chamber from the mechanical pump. This artefact clearly demonstrates
that porous films with high internal surface can be used as active materials in order
to detect vapor traces. The next section proposes a way to enhance the optical effect
of adsorption in porous Si3 N4 by fabricating an interference optical filter with a
240 periodic porous/dense multilayer structure.

9 Page 9 of 29
 3.3 Fabrication and testing of porous Fabry-Perot filter sensors

After a detailed characterization of the microstructure and of the chemistry of

porous Si3 N4 films, we implemented the porous layers in Fabry-Perot multilayer
filters in order to evaluate their performance as optical gas sensors.

t
We designed and fabricated a 43 layer FP filter ([HL]10 HLLH [LH]10 ) using porous

ip
245

and dense Si3 N4 as nL and nH materials. Dense Si3 N4 was employed as the initial
and final layers in order to increase adhesion of the coating to the substrate, as well

cr
as to improve its wear resistance. The central cavity was made of porous Si3 N4 to
enhance the optical variations caused by the vapor or gas adsorption in the pores.
250 The thickness of the quarterwave and halfwave layers was calculated to obtain a

us
central transmission peak at 660 nm. During deposition, the layer thickness was
controlled by time using calibrated deposition rates.

an
After the process was completed, the filter was inserted in the environmental cham-
ber described in section 2, and its optical transmission spectrum was acquired in
255 vacuo (see Fig. 7). The wavelength of the narrow pass band was evaluated at 665
nm, close to the predicted value (660 nm). In addition, the amplitude and the width
M
of the band, as well as the shape of the overall spectrum correspond very well to the
design, indicating that the deposition rates and the refractive indices were constant
during the filter fabrication process. This example clearly demonstrates the possi-
d

260 bility of realizing high-quality complex optical filters from porous materials. It is
also certain that a better precision of the position of the transmission band could be
te

obtained by performing standard optical thickness monitoring.

In the next step, the FP filter was annealed in 300 Torr of O2 at 300◦ C, and transmis-
p

sion spectra were acquired every 10 sec in order to monitor the process. This heat
ce

265 treatment presents two advantages: first it stabilizes the chemistry of the porous
coating by removing the weakest bonds (e.g. N–H bonds) and by replacing them
with stronger Si–O bonds as documented in section 3.1; second, oxidation causes
a decrease of neff , and hence a decrease of the filter optical thickness, leading to a
Ac

blue shift of the entire transmission spectrum. Using this method, it is possible to
270 tune the wavelength of the central peak to that of a monochromatic light source,
such as a laser diode (655.2 nm) in our case.

Fig. 8 shows the evolution of the central peak position at various stages of the
annealing process: (i) prior to annealing, a transmission spectrum was acquired with
the FP filter in air at 20◦ C, and the central peak was evaluated at 675 nm; (ii) the
275 chamber was then pumped down and transmission measurements revealed that the
peak position had shifted to 665 nm, which can be related to water desorption; (iii)
the FP filter was then heated at 300◦ C in vacuo for 15 min, which shifted the peak
further to 662 nm. As documented in section 3.1, this phenomenon is attributed
to the presence of residual water in the chamber and in the pores, which causes

10 Page 10 of 29
280 oxidation of the coating; (iv) O2 was admitted into the chamber (300 Torr) in order
to continue the oxidation and thus to tune the peak position to the wavelength of
the laser diode. Since the weakest chemical bonds had already been removed, the
oxidation process was much slower than for steps i–iv, and it took about 3 h to attain
656.5 nm; (v) After pumping and cooling the chamber, we obtained a peak position
285 at 655.9 nm, due to thermal contraction of the coating. We have intentionally not
adjusted the peak maximum exactly at the diode wavelength in order to obtain a

t
more linear response of the sensor. Finally, one should note that the changes of

ip
the transmission amplitude are artefacts caused by the instability of the white light
source intensity, and thus do not represent significant modifications of the FP filter.

cr
290 After heat treatment, the FP filter was ready to be used for gas sensing. The white
source and the spectrometer were replaced by a laser diode and a silicon detector

us
(see section 2), and the intensity variations of the light passing through the filter
were measured in various environments. Fig. 9 shows the response of the sensor
when sequentially exposed to 200 mTorr of ethanol, 1 atm of air and 300 mTorr of

an
295 NH3 , in alternation with vacuum. The sensor signal dropped within a few seconds
as soon as the gases or vapors were introduced into the chamber, and it then stabi-
lized between 30 s to 1 min, depending on the nature of the adsorbate. This drop is
clearly caused by the adsorption of molecules on the pore walls as the vapor pene-
M
trates through the coating thickness and increases neff . In the case of air, the sensor
300 was essentially sensitive to water vapor (40% of relative humidity in laboratory),
and the effect of N2 and O2 could be neglected.
d

The sensor response was strong for all three vapors, but it was more pronounced
te

for ethanol and air than for NH3 . This effect cannot be explained by a difference
of refractive index of the adsorbates since they are very similar to each other (see
Table 2). The cause rather lies in the difference of vapor pressure (Table 2), lower
p

305

vapor pressure leading to a greater amount of adsorbate in the pores. For the same
ce

reason, the signal intensity was almost completely restored for ethanol and NH3 at
the end of the first cycle, while it appeared to be lower in the case of air, as water
presents a lower vapor pressure. The sensor showed a good reproducibility for all
310 vapors along the five cycles. From these results, we conclude that our device meets
Ac

three important requirements for a sensor: fast response, high sensitivity and good
reproducibility.

In order to test the sensitivity limit of the sensor as well as to evaluate its per-
formance in a more realistic environment, we exposed the FP filter to various
315 amounts of ethanol and NH3 diluted in 300 Torr of nitrogen (Fig. 10). By com-
paring Figs. 10b and 9c, we observe that dilution lowers the sensor sensitivity: for
instance, it takes 850 mTorr of NH3 diluted in N2 (2840 ppm) to obtain the same
level of intensity as with 300 mTorr of pure NH3 . This is related to the fact that
a fraction of the adsorption sites on the pore walls are occupied by N2 molecules
320 rather than NH3 ones, with very low effect on the sensor response (see the first two
cycles in Fig. 10).

11 Page 11 of 29
 In addition, by comparing the steepness of the slopes immediately after the gas ad-
mission, we conclude that dilution slows down the signal response: with 426 mTorr
of NH3 diluted in N2 (1420 ppm) (Fig. 10b), it took around 150 s for the signal to
325 stabilize, while less than 60 s was necessary in 300 mTorr of pure NH3 (Fig. 9c).
Hence, the time ratio normalized to the pressure ratio (150/60 × 426/300) gives a
response time 3.55 times lower for diluted NH3 than for pure NH3 . This behavior
can be related to the much higher total pressure in the case of diluted NH3 (300 Torr

t
versus 300 mTorr), leading to a decrease of the mean free path of the molecules that

ip
330 slows down the diffusion of NH3 through the coating, thus lowering the response
time of the sensor. Regarding the sensitivity limits, a clear signal variation can be

cr
observed for 10 ppm of ethanol (cycle 4 in Fig. 10a) and 28 ppm of NH3 (cycle 4
in Fig. 10b) compared to pure nitrogen. Thus the sensor remains more sensitive to
ethanol than to NH3 .

us
335 In the final part of this study, we demonstrate a possibility to regenerate the sensor.
Indeed, we have shown in Fig. 9 that gases with low vapor pressures do not entirely

an
desorb from the FP filter after pumping, leading to a drift of the sensor signal. In
Fig. 11, we show that a rapid heating of the filter can empty the pores by desorbing
the remaining adsorbate. The following steps were performed in order to prove
340 this concept: (1) the sensor was exposed to air until the signal stabilized to a low
M
intensity level; (2) air was pumped for 5 min leading to partial desorption of water
and thus a rise of the signal intensity; (3) we started heating the filter in vacuo until
the temperature reached 300◦ C; (4) we kept the temperature constant at 300◦ C for
d
20 min and observed the desorption process, resulting in an increase of intensity;
345 (5) the filter was cooled down to ambient temperature in vacuo.
te

The decrease of intensity at the beginning of step 3 (see Fig. 11) and the presence
of small interference fringes can be associated with thermal expansion of the fil-
p

ter and the glass substrate. Similarly, the slow intensity increase in step 5 and the
occurrence of larger interference fringes are due to thermal contraction of the sam-
ce

350 ple. After treatment, the light intensity was slightly higher than the original value
(1.6 mW), proving the sensor regeneration. The temperature and heating time can
possibly be optimized in order to improve the reproducibility. Additionally, it is
Ac

expected that the sensor implementation in a small integrated device would consid-
erably reduce the heating and cooling times.

355 4 Summary

We studied the behavior of plasma-deposited single-material nanoporous Si3 N4 op-

tical filters with respect to gas and vapor sorption. Porous (neff = 1.58) and dense
films (neff = 1.78) were found, respectively, to contain a total void volume frac-
tion fv = 34% and fv = 10% with an accessible volume fraction of fo = 27%
360 and fo = 3%. Based on adsorption isotherms of ethanol in the porous films, we

12 Page 12 of 29
 estimated the average pore size to be about 2 nm and the proportion of microp-
ores/mesopores to be 46%/54% of fo . Porous films were investigated by IR spec-
troscopic ellipsometry and we showed that heat treatment is an efficient way to
cure and chemically stabilize the Si3 N4 matrix. In order to enhance the optical ef-
365 fect of the adsorbate present in the pores, we fabricated porous/dense quarterwave
Fabry-Perot filters as a prototype gas sensor. After annealing the filter for chemical
stabilization and fine tuning the pass band position, the filter exhibited high sen-

t
sitivity to various gases and vapors such as ethanol (10 ppm), NH3 (28 ppm) and

ip
air containing humidity. Moreover, the sensor response was fast (a few seconds for
370 pure gases), with good reproducibility, and it could easily be regenerated by heat

cr
treatment.

The main advantages of this approach are: (i) precise control of the fabrication

us
process, allowing one to deposit high-quality optical coatings with well-defined
characteristics; (ii) possibility to deposit such porous/dense filters on many types
375 of substrates, such as plastics or optical fibers; (iii) simplicity of the concept, which

an
makes it easy to integrate and scale up at low cost (inexpensive laser diode and sili-
con photodetector). An important aspect of the behavior of such all-optical sensors,
namely their selectivity, will be addressed in future work.
M
Acknowledgments
d
te

380 The authors wish to thank Mr. Francis Turcot for expert technical assistance. This
work was supported by the Natural Sciences and Engineering Research Council of
Canada (NSERC).
p
ce

References
Ac

[1] R. W. Bogue, Nanotechnology: What are the prospects for sensors?, Sensor Review
385 24 (2004) 253–260.

[2] J. Shi, Y. Zhu, X. Zhang, W. R. G. Baeyens, M. García-Campaña, Recent

developments in nanomaterial optical sensors, TRAC Trends Anal. Chem. 23 (2004)
351–360.

[3] K. E. Romo-Medrano, S. N. Khotiaintsev, V. Garcia-Garduno, Optical-fibre sensor

390 system for monitoring the performance of the gas-propellant centrifuge separator of a
spacecraft, Meas. Sci. Technol. 15 (2004) 1568–1575.

[4] R. W. Bogue, Non-contacting optical sensors for real-time water quality monitoring,
Sensor Review 23 (2003) 24–25.

13 Page 13 of 29
 [5] P. Ferdinand, S. Magne, O. Roy, V. D. Marty, S. Rougeault, M. Bugaud, Optical fiber
395 sensors for the nuclear environment, in: S. Martellucci, A. N. Chester, A. G. Mignani
(Eds.), Optical Sensors and Microsystems: New Concepts, Materials, Technologies,
Kluwer Academic/Plenum Publishers, New York, 2000, pp. 205–226.

[6] A. Mendez, Optical fiber sees growth as medical sensors, Laser Focus World 47 (2011)
91–94.

t
400 [7] A. Kovacs, P. Jonnalagadda, X. Y. Meng, U. Mescheder, Characterization of porous

ip
silicon based optical sensor system for biosensor applications, IEEE, 2009, pp. 21–26.

[8] L. D. Stefano, I. Rendina, L. Moretti, S. Tundo, A. M. Rossi, Smart optical sensors

cr
for chemical substances based on porous silicon technology, Appl. Opt. 43 (2004)
167–172.

us
405 [9] V. Mulloni, Z. Gaburro, L. Pavesi, Porous silicon microcavities as optical and electrical
chemical sensors, Phys. Status Solidi A 182 (2000) 479–484.

[10] J. Volk, J. Balázs, A. L. Tóth, I. Bársony, Porous silicon multilayers for sensing by

an
tuneable ir-transmission filtering, Sens. Actuators B 100 (2004) 163–167.

[11] Y. Y. Li, F. Cunin, J. R. Link, T. Gao, R. E. Betts, S. H. Reiver, V. Chin, S. N. Bathia,

410 M. J. Sailor, Polymer replicas of photonic porous silicon for sensing and drug delivery
M
applications, Science 299 (2003) 2045–2047.

[12] L. Martinu, D. Poitras, Plasma deposition of optical films and coatings: A review, J.
Vac. Sci. Technol. A 18 (2000) 2619–2645.
d

[13] H. A. Macleod, Turning value monitoring of narrow-band all-dielectric thin-film

te

415 optical filters, Opt. Acta 19 (1972) 1–28.

[14] D.-H. Chang, Y. J. Park, C. K. Hwangbo, Optical humidity sensor using a narrow
band-pass filter prepared by glancing angle deposition, J. Korean Phys. Soc. 53 (2008)
p

2700–2704.
ce

[15] J. J. Steele, Nanostructure engineering in porous columnar thin films: recent advances,
420 J. Mater. Sci., Mater. Electron. 18 (2007) 367–379.

[16] L. Martinu, J. E. Klemberg-Sapieha, O. M. Küttel, A. Raveh, M. R. Wertheimer,

Ac

Critical ion energy and ion flux in the growth of films by plasma-enhanced chemical-
vapor deposition, J. Vac. Sci. Technol. A 12 (1994) 1360–1364.

[17] D. Poitras, P. Leroux, J. E. Klemberg-Sapieha, S. Gujrathi, L. Martinu,

425 Characterization of homogeneous and inhomogeneous Si-based optical coatings
deposited in dual-frequency plasma, Opt. Eng. 35 (1996) 2693–2699.

[18] O. Zabeida, A. Hallil, M. R. Wertheimer, L. Martinu, Time-resolved measurements

of ion energy distributions in dual mode pulsed-microwave/radio frequency plasma, J.
Appl. Phys. 88 (2000) 635–642.

430 [19] J. A. Thornton, Influence of apparatus geometry and deposition conditions on the
structure and topography of thick sputtered coatings, J. Vac. Sci. Technol. 11 (1974)
666–670.

14 Page 14 of 29
 [20] R. Messier, A. P. Giri, R. A. Roy, Revised structure zone model for thin film physical
structure, J. Vac. Sci. Technol. A 2 (1984) 500–503.

435 [21] R. Vernhes, O. Zabeida, J. E. Klemberg-Sapieha, L. Martinu, Single-material

inhomogeneous optical filters based on microstructural gradients in plasma-deposited
silicon nitride, Appl. Opt. 43 (2004) 97–103.

[22] R. Vernhes, A. Amassian, J. E. Klemberg-Sapieha, L. Martinu, Rf plasma treatment of

t
porous SiN x :H films for the fabrication of porous-dense multilayer optical filters with

ip
440 minimized interfaces, J. Appl. Phys. 99 (2006) 114315–1–12.

[23] A. Bergeron, J. E. Klemberg-Sapieha, L. Martinu, Structure of the interfacial region

cr
between polycarbonate and plasma-deposited SiN1.3 and SiO2 optical coatings studied
by ellipsometry, J. Vac. Sci. Technol. A 16 (1998) 3227–3234.

us
[24] D. E. Aspnes, Optical properties of thin films, Thin Solid Films 89 (1982) 249–262.

445 [25] J. Robertson, Electronic structure of silicon nitride, Philos. Mag. B 63 (1991) 47–77.

an
[26] J.-W. Lee, R. Ryoo, M. S. Jhon, K.-I. Cho, Bond density and physicochemical
properties of a hydrogenated silicon nitride film, J. Phys. Chem. Solids 56 (1995)
293–299.
M
[27] M. R. Baklanov, K. P. Mogilnikov, V. G. Polovinkin, F. N. Dultsev, Determination of
450 pore size distribution in thin films by ellipsometric porosimetry, J. Vac. Sci. Technol.
B 18 (2000) 1385–1391.

[28] S. J. Gregg, K. S. W. Sing, Adsorption, surface area and porosity, Academic Press Inc.,
d

New York, 1982.

te

[29] M. Polanyi, Adsorption and capillary condensation, Phys. Z. Sowjetunion 4 (1933)

455 144–155.
p

[30] F. N. Dultsev, Investigation of the microporous structure of porous layers using

ellipsometric adsorption porometry, Thin Solid Films 458 (2004) 137–142.
ce

[31] C. H. Amberg, R. McIntosh, A study of adsorption hysteresis by means of length

changes of a rod of porous glass, Can. J. Chem. 30 (1952) 1012–1032.
Ac

460 [32] G. Lucovsky, J. Yang, S. S. Chao, J. E. Tyler, W. Czubatyj, Nitrogen-bonding

environments in glow-discharge-deposited a-Si:H films, Phys. Rev. B 28 (1983) 3234–
3240.

[33] E. Bustarret, M. Bensouda, M. C. Habrard, J. C. Bruyère, Configurational statistics in

a-Si x Ny Hz alloys: a quantitative bonding analysis, Phys. Rev. B 38 (1988) 8171–8184.

465 Biographies

Richard Vernhes completed his M.Sc. and Ph.D. degrees in physics at Polytech-
nique Montreal, Canada in 2006, and received his B.Sc in materials science from

15 Page 15 of 29
 Polytech Paris-Sud, France in 1999. He has been Research Associate in the FCSEL-
LaRFIS team at Polytechnique Montreal since 2008. His current research involves
470 the application of nanostructured materials for sensors and smart coatings, as well
as the development of new thin film architectures for optical and decorative appli-
cations.

Jolanta E. Klemberg-Sapieha received her Master’s and Ph.D. degrees in chem-

t
ical engineering and in materials science from the Technical University of Lodz,

ip
475 Poland. She is Associate Professor in the Department of Engineering Physics, Poly-
technique Montreal, Canada. She is the Associate Director of the Functional Coat-

cr
ing and Surface Engineering Laboratory (FCSEL) and Director of the Labora-
tory for Optical and Tribomechanical Metrology of the Thin-Film Research Center
(GCM). Her main research activity is the science and technology of thin films, sur-

us
480 faces and interfaces, in particular new fabrication processes, nanostructured materi-
als, and tribomechanical properties of coating systems aimed at aerospace, biomed-
ical, and optical applications.

an
Ludvik Martinu received his Master’s and Ph.D. degrees in physics from Charles
University, Prague, Czech Republic. He is a Professor at Polytechnique Montreal,
485 Canada, and former Head of the Department of Engineering Physics. He is the
M
founder and Director of the Functional Coating and Surface Engineering Labora-
tory (FCSEL), and presently Associate Director of the Thin-Film Research Center
on the Campus of the Université de Montréal. His main research interest is the de-
velopment of plasma-based processes and of nanostructured materials and thin-film
d

490 systems for optics, photonics, aerospace, biomedical, energy and other applications.
p te
ce
Ac

16 Page 16 of 29
Tables

Table 1
Refractive indices before and after ethanol sorption, and corresponding total void volume
fraction and open void volume fraction for a dense/porous/dense layer stack.

neff at 550 nm neff at 550 nm fv fo

t
Stack in vacuo in ethanol (%) (%)

ip
Bottom layer 1.806 1.816 10 3

cr
Central layer 1.582 1.686 34 27
Top layer 1.785 1.791 13 2

us
Table 2
Physical properties of ethanol, water and ammonia at room temperature.

an
Adsorbate n at 550 nm Vapor pressure (Torr)

Ethanol 1.361 50
M
Water 1.333 20
Ammonia 1.355 7000
d
p te
ce
Ac

17 Page 17 of 29
 Figure captions

Fig. 1 Schematic of the MW/RF dual-mode plasma deposition system.

Fig. 2 Schematic of the setup for treating and testing the optical filters.
495 Fig. 3 Refractive index variations of a porous film versus the ethanol relative pres-
sure during adsorption/desorption experiments. fo is calculated from Eq. 4 and
corresponds to the volume fraction of open porosity that is progressively filled

t
by ethanol as the relative pressure is increased.

ip
Fig. 4 Pore size distribution calculated from the Kelvin equation.
500 Fig. 5 Plot of log(V/V0) against log2 (1/pr ) according to the Dubinin-Radushkevich

cr
model in order to estimate the average pore size.
Fig. 6 Optical constants n and k of a porous film at different steps during the heat
treatment: in air, 20◦ C (#); in vacuo, 20◦ C (); in vacuo, 200◦ C (△); in vacuo,

us
20◦ C (♦). The k spectra are shifted upward for clarity.
505 Fig. 7 Optical transmission of a Fabry-Perot filter in vacuo: experimental data (—
–); design (– –).

an
Fig. 8 Wavelength shift of the Fabry-Perot filter transmission at different steps dur-
ing heat treatment: (1) in air, 20◦ C; (2) in vacuo, 20◦ C; (3) in vacuo, 300◦ C ; (4)
in 300 Torr of O2 , 300◦ C; (5) in vacuo, 20◦ C. Vertical line shows the laser diode
wavelength.
M
510

Fig. 9 Sensor response to various pure gases within five cycles: a) 200 mTorr
ethanol; b) 1 atm air; c) 300 mTorr NH3 .
Fig. 10 Sensor response to ethanol (a) and NH3 (b) diluted in various proportions
d

in 300 Torr of nitrogen.

515 Fig. 11 Sensor response during the regeneration process. Horizontal line shows the
te

original intensity before air exposure.

p
ce
Ac

18 Page 18 of 29
 t
ip
cr
RF power

us
SiH4
N2

an RF electrode
M
to pumps
d

sample
p te

MW power
ce

fused silica window

MW applicator
Ac

Fig. 1. Schematic of the MW/RF dual-mode plasma deposition system.

19 Page 19 of 29
 t
ip
cr
us
À Vacuum chamber Ä Pumping
Á Heatable sample holder Å Light source

an
 Sample Æ Light detector
à Gas inlet Ç Computer

À
M
à =⇒

Á
d

Æ
Å
Ç
te

Ä =⇒
p
ce

Fig. 2. Schematic of the setup for treating and testing the optical filters.
Ac

20 Page 20 of 29
 t
ip
cr
30
1.68

us
D 25
1.66
20

an
1.64 C
15

fo (%)
neff

1.62
M
B 10
1.60 A→B: filling of micropores and adsorbate
monolayer formation in mesopores; 5
B→C: adsorbate multilayer formation in
d

1.58 mesopores;
C→D: capillary condensation of adsorbate 0
te

A in mesopores;
1.56

0.0 0.2 0.4 0.6 0.8 1.0

p

pr
ce

Fig. 3. Refractive index variations of a porous film versus the ethanol relative pressure
during adsorption/desorption experiments. fo is calculated from Eq. 4 and corresponds to
the volume fraction of open porosity that is progressively filled by ethanol as the relative
Ac

pressure is increased.

21 Page 21 of 29
 t
ip
cr
0.60
Micropores Mesopores Macropores

us
Area=62% Area=38% Area=0%
0.50

0.40

an
dfo /dr (nm−1 )

M
0.30
d

0.20
Data
te

Fit
0.10
p
ce

0.00 −1
10 100 101 102
rm (nm)
Ac

Fig. 4. Pore size distribution calculated from the Kelvin equation.

22 Page 22 of 29
 t
ip
cr
0
log(V /V0 ) = −0.06 × log2 (1/pr ) − 0.78

us
−1

an
log(V /V0 )

−2
M
−3
d
te

−4
p

−5
ce

0 10 20 30 40 50 60
2
log (1/pr )
Ac

Fig. 5. Plot of log(V/V0 ) against log2 (1/pr ) according to the Dubinin-Radushkevich model
in order to estimate the average pore size.

23 Page 23 of 29
 t
3.0 1.60

ip
in air, 20 ◦ C;
in vacuo, 20 ◦ C;

cr
2.5 in vacuo, 200 ◦ C; 1.55
in vacuo, 20 ◦ C;

us
n 2.0 1.50 n

1.5 1.45

1.0 an 1.40
M
N–H+O–H
Si–O

Si–O
Si–N

Si–N

N–H

O–H
N–H

C–H

0.4
d
2.5
0.3
te

2.0
k k
1.5 0.2
p

1.0
ce

0.1
0.5

0.0 0.0
Ac

600 900 1200 2000 3000 4000

Wavenumber (cm−1 )
Fig. 6. Optical constants n and k of a porous film at different steps during the heat treatment:
in air, 20◦ C (#); in vacuo, 20◦ C (); in vacuo, 200◦ C (△); in vacuo, 20◦ C (♦). The k spectra
are shifted upward for clarity.

24 Page 24 of 29
 t
ip
cr
Experimental data

us
100 Design

an
Transmission (%)

75 M
100
50
Transmission (%)

80
60
d

40
25
te

20
0
650 660 670 680
Wavelength (nm)
p

0
400 500 600 700 800
ce

Wavelength (nm)

Fig. 7. Optical transmission of a Fabry-Perot filter in vacuo: experimental data (—–); design
Ac

(– –).

25 Page 25 of 29
 t
ip
cr
À: in air, 20 ◦ C;
Á: in vacuo, 20 ◦ C;

us
100 Â: in vacuo, 300 ◦ C;
Ã: in 300 Torr of O2 , 300 ◦ C;
Ä: in vacuo, 20 ◦ C; Á À
Transmission (%)

75 Ä
Ã
Â
an
M
50
d

25
p te

0
640 645 650 655 660 665 670 675 680
ce

Wavelength (nm)

Fig. 8. Wavelength shift of the Fabry-Perot filter transmission at different steps during heat
Ac

treatment: (1) in air, 20◦ C; (2) in vacuo, 20◦ C; (3) in vacuo, 300◦ C ; (4) in 300 Torr of O2 ,
300◦ C; (5) in vacuo, 20◦ C. Vertical line shows the laser diode wavelength.

26 Page 26 of 29
 2.0
200 mTorr of ethanol

Intensity (mW)
1.5

1.0

t
ip
0.5

cr
0.0
0 15 30 45 60 75 90 105 120

us
Time (min)
2.0
1 atm of air

an
Intensity (mW)

1.5 M
1.0

0.5
d
te

0.0
0 10 20 30 40 50 60
Time (min)
p

2.0
ce

300 mTorr of ammonia

Intensity (mW)

1.5
Ac

1.0

0.5

0.0
0 10 20 30 40 50 60 70 80 90
Time (min)

Fig. 9. Sensor response to various pure gases within five cycles: a) 200 mTorr ethanol; b)
1 atm air; c) 300 mTorr NH3 .

27 Page 27 of 29
 t
1.6

ip
À Á Â Ã Ä
1.4 Å
Intensity (mW)

1.2 Æ

cr
À: 300 Torr N2
1.0 Á: 300 Torr N2
Â: 2 ppm of C2 H5 OH in À Ç
0.8

us
Ã: 10 ppm of C2 H5 OH in À
0.6 Ä: 20 ppm of C2 H5 OH in À
0.4 Å: 40 ppm of C2 H5 OH in À

an
Æ: 100 ppm of C2 H5 OH in À
0.2 Ç: 200 ppm of C2 H5 OH in À
0.0
0 15 30 45 60 75 90 105 120
M
Time (min)
d

1.6 À Á Â Ã Ä
Å
Intensity (mW)

te

À: 300 Torr N2
1.4
Á: 300 Torr N2
Â: 14 ppm of NH3 in À
p

1.2 Ã: 28 ppm of NH3 in À

Ä: 70 ppm of NH3 in À
ce

Æ
Å: 142 ppm of NH3 in À
1.0 Æ: 1420 ppm of NH3 in À
Ç: 2840 ppm of NH3 in À Ç
Ac

0.8
0 15 30 45 60 75 90 105 120
Time (min)

Fig. 10. Sensor response to ethanol (a) and NH3 (b) diluted in various proportions in
300 Torr of nitrogen.

28 Page 28 of 29
 t
ip
cr
À: in 1 atm air, 20 ◦ C

us
Á: in vacuo
2.0 Â: in vacuo, heating to 300 ◦ C
Ã: in vacuo, 300 ◦ C
Ä: in vacuo, cooling down to 20 ◦ C

an
Intensity (mW)

1.5
Ä
M
Ã
1.0
Â
d

Á
te

0.5
p

À
0.0
ce

0 15 30 45 60 75 90 105 120
Time (min)

Fig. 11. Sensor response during the regeneration process. Horizontal line shows the original
Ac

intensity before air exposure.

29 Page 29 of 29