Sensors and Actuators B 106 (2005) 347–357
Electrical porous silicon chemical sensor for detection of
organic solvents
M. Archer
a,∗
, M. Christophersen, P.M. Fauchet
a,b
a
Department of Biomedical Engineering, Center for Future Health, University of Rochester,
601 Elmwood Avenue, Rochester, NY 14642, USA
b
Department of Electrical and Computer Engineering, Center for Future Health, University of Rochester,
601 Elmwood Avenue, Rochester, NY 14642, USA
Received 20 April 2004; received in revised form 17 August 2004; accepted 18 August 2004
Available online 25 September 2004
Abstract
A novel electrical sensor platform containing a porous silicon (PSi) layer on a crystalline silicon substrate has been developed in which the
electrical contacts are made exclusively on the backside of the substrate allowing complete exposure of the surface to the sensing molecules.
The PSi layers were 20 ␮m thick with an average pore diameter of 1␮m. Real-time measurements of capacitance (C) and conductance (G)
were performed and the response produced by the addition of different organic solvents was evaluated. The observed response is attributed to
the combined effect of achange in dielectric constant inside the porousmatrix and amodification in the depletionlayer width inthe crystalline
silicon structure. A space charge region modulation model was used to explain the effect induced by molecules of different dipole moments,
dielectric constants, polarizabilities and water solubilities.
© 2004 Elsevier B.V. All rights reserved.
Keywords: Macroporous silicon; Electrical sensors; Organic solvent detection; Chemical sensor
1. Introduction
Porous silicon(PSi) isproduced by electrochemicaldisso-
lution of crystalline silicon in a hydrofluoric acid based elec-
trolyte. The resulting structure consists of pores alternating
with crystalline silicon rods attached to a crystalline silicon
substrate. PSi is characterized by a large internal surface area

doi:10.1016/j.snb.2004.08.016
348 M. Archer et al. / Sensors and Actuators B 106 (2005) 347–357
tected DNA hybridization and the presence of ethanol via a
change in the conductance [14] of mesoporous silicon layers
(pore diameter 20–50 nm) in structures with two concentric
circular electrical contacts made on the backside of the crys-
talline silicon substrate. We explained our findings as due
to charge redistribution in the crystalline silicon substrate
induced by the changes in the PSi layer. In this paper, we
present an optimized design of our porous silicon sensor in
which the size of the device is greatly reduced, the geome-
try of the backside electrodes is modified and a porous layer
having a different morphology is used. Systematic experi-
ments with different organic solvents in the liquid phase and
realistic modeling of the structures are also presented.
Four different geometrieshave beenusedto measure elec-
trical properties of PSi. These include metal contacts evapo-
rated only onthe porous layer surface [15]orin a “sandwich”
configuration [8,9,16,17], the use of coplanar contacts on the
porous layer [10] and interdigitated electrodes using pnjunc-
tions surrounded by PSi [18,19]. In all cases the response
of the device depends on the characteristics of the electrical
contact with the PSi. In comparison, in our devices, the field
propagates from the crystalline silicon substrate to the PSi
layer. Since no electrical contact is made on the porous layer,
any influence of the contact barrier and chemical reactions
that may occur between the metal and the organic solvent is
eliminated. The observed response is therefore related only
to the presence of molecules inside the porous layer and their
interaction with the surface of the layer.

porous silicon. The oxide is hydrophilic enough to allow the
infiltration of water soluble molecules without the need of
Fig. 1. Scanning electron microscopy (SEM) cross sectional view of a
macroporous silicon layer produced form p-type silicon (ρ ∼ 10–20 cm)
with an organic electrolyte. The bright areas correspond to c-Si rods and the
dark areas to pores propagating from the surface parallel to the c-Si rods.
a thicker thermally grown oxide. After oxidation the porous
layers wererinsedwithdeionizedwater andethanol anddried
under a stream of nitrogen. The oxide on the backside of the
crystalline siliconsubstrate was strippedwith a15% HFsolu-
tion (7:1, water: 49 wt.% HF) prior to the contact placement.
The wafers were cleaved into sections of 4 × 7mm and two
coplanar electrical contacts were placed 700␮m apart on the
crystalline silicon substrate. In our approach, the PSi surface
is completely exposed to the sensing molecules and no metal
contacts are made to it, avoiding the introduction of foreign
materials into the porous matrix. Fig. 2 shows a schematic
cross-sectional view of our device and images of the front
and backsides showing the electrical contacts and the actual
dimensions. In order to avoid any of the solvents tested from
reaching the backside contacts, the sensors were fixed on a
glass slide, which ensures a horizontal surface for a uniform
distribution of the solvent on the porous layer and protects
the backside of the device.
2.2. Measurement setup
Real-time capacitance (C) and conductance (G)
measurements were performed with an inductance–
capacitance–resistance (LCR) multifrequency meter. The
measurement parameters (frequency and bias voltage) and
the data acquisition and storage were controlled with a

representation of the measurement setup is shown in
Fig. 4.
Fig. 4. Schematic of the measurement setup. An inductance–capacitance–
resistance(LCR)metercontrolledbyLabView
TM
isusedtomeasure the real-
time changes of capacitance (C) and conductance (G). The measurements
are performed under a low humidity ambient.
2.3. Measurements with organic solvents
To evaluate the response of our sensor to organic solvents,
we exposed thedevice to moleculeswithdifferent dipolemo-
ments, polarizabilities and dielectric constants. The solvents
were separatedin two groups,polar andnon-polar molecules.
Water was analyzed independently. The characteristics of the
solvents used are shown in Table 1 [25] Four characteristics
were selected to understand the response:
• Dielectric constant: related tothe electric field distribution
inside the porous layer.
• Dipole moment:relatedto the local fields on the surface of
the porous layer.
• Polarizability: related to the orientation of the molecule
with respect to the porous layer surface.
• Bond character: related to water solubility (polar
molecules are more soluble in water than non polar
molecules).
Prior to the addition the sensor was allowed to stabilize
for at least 20 min under the temperature and humidity con-
ditions indicated in Section 2.2. Individual experiments were
performed on different layers by adding 10␮l of solvent. We
have investigated the sensitivity of the sensor to volumes in

with those in the literature we have developed a space charge
region modulation (SCRM) model. We will first describe our
sensor as a field effect device and derive an equivalent elec-
trical circuit. The principles ofdetectionand the assumptions
made in the analysis will then be discussed. Finally, the cor-
relation of the electrical response with the electrical charac-
teristics of the molecules will be presented in Section 4.
In a field effect transistor (FET) current flows through a
channel when a voltage is applied between the source and
drain terminals. The conductance in the channel, which is
directly proportional to its dimensions and the number of
carriers can be modulated by changing either of these two
variables. This modulation is done by applying an electric
field through a metal gate terminalwhichcan be placed at the
same plane between the source and drain (e.g., MESFET) or
parallel to them (e.g., JFET). When FETs areusedaselectro-
chemical transducers the metal gate is substituted by an elec-
trolyte or a synthetic selective membrane and modulation of
the conductance in the channel results from thechangeinpo-
tential at the semiconductor surface when chemical species
are present. In our device the gate electrode is substituted by
the porous layer and the channel is the c-Si substrate. Al-
though the experiments are carried out under a controlled
humidity ambient, the presence on the porous layer surface
of at least a monolayer of water is unavoidable. This initial
condition of the porous layer surface renders it with a larger
hydrophilic character, which influences its adsorption prop-
erties. When a molecule is infiltrated in the porous layer its
interaction with the surface will change the field distribution
in the c-Si rods. The porous layer then becomes a charged

tric constant of the porous layer. The change in the space
charge region will depend on the sign of the charge “felt”
by the c-Si rod and its magnitude, which is also influenced
by the orientation of the molecule with respect to the sur-
face. If the effect is that of a positive charge then the width
of the space charge region increases, the opposite being true
for negative charges. Water soluble molecules tend to as-
sociate closer to the surface therefore inducing a stronger
effect.
3.2. Simulations and equivalent circuit
In order to evaluate the field distribution in our device
we performed a simulation of the electric field using com-
mercially available software (Ansoft, Maxwell 2-D). Fig. 5
M. Archer et al. / Sensors and Actuators B 106 (2005) 347–357 351
Fig. 5. Simulation of the electric field distribution in the cross section of a c-Si substrate with two parallel electrodes. An ac voltage of 1 V peak to peak at
100 kHz excitation frequency and a dc voltage of 0V were used as experimental parameters.
shows the calculated electric field distribution for 15 cm
p-type Si with two parallel contacts placed on the back. The
applied ac voltage is 1 V peak to peak at 100 kHz excita-
tion frequency and the dc voltage is 0 V. The field intensity
is larger near the contacts and gradually decreases towards
the top c-Si surface. This field distribution is very similar to
the one reported by Ramos et al. [27] for dielectrophoresis
applications with a similar electrode geometry. The current
density distribution is also such that its magnitude is larger at
the interface with the electrodes close to the gap, and grad-
ually reduces towards the surface of the c-Si substrate. The
simulation shows that the electric field can reach the top of
the c-Si where the PSi layer is located. As mentioned before
the response of our device is based on changes in the porous

tions were used in the model:
• The interface states at the metal–semiconductor junction
donotaffecttheresponse.Ajunction capacitanceispresent
at each electrical contact and its value remains constant at
the given excitation frequency [24].
• The porous layer is modeled as a composite material made
of alternating pores (or void space) and crystalline silicon
rods with bulk silicon properties.
• The electric field (E) and the current density (J) reach the
porous layer.
• The infiltration of the porous layer with materials of dif-
ferent physical properties produces a change in the width
of the depletion region of the c-Si rods and a change in
dielectric constant of the void space.
• The PSi layer can be considered as a charged layer in con-
tact with thec-Si substrate. Changes in thecharge distribu-
tion within the porous layer extend into the c-Si substrate.
352 M. Archer et al. / Sensors and Actuators B 106 (2005) 347–357
Fig. 7. Schematic section of the sensor showing a PSi layer composed of
void space or pores and c-Si rods, and the c-Si substrate. Each element in
the porous layer is modeled as a parallel capacitor–resistor element, C
pore
|| R
pore
for the pores and C
rod
|| R
rod
for the c-Si rods. The c-Si substrate is
represented by G

junc
(C
junct
). The equivalent impedance is
given by:
Z
eq
=
[(Z
subs
)(Z
pore
+ Z
rod
)]
Z
subs
+ Z
pore
+ Z
rod
+ 2Z
junct
(1)
In addition, C
rod
∝ 1/W
d
, C
pore

we measure a higher reference conductance between 0.1 and
0.2 mS. This change in magnitude is due to the presence of
the c-Si substrate and the metal–semiconductor contact. The
model can therefore be further simplified by neglecting R
rod
and R
pore
and considering only C
pore
and C
rod
in the porous
layer impedance. It is also worthy to clarify that a Schot-
tky barrier is considered at the metal–semiconductor contact.
This is based on calculations of the difference in the work
functions between silver contact and the low-doped p-type
c-Si. This assumption is difficult to address experimentally
since the measured conductance includes the contribution of
the c-Si substrate and not only that of the junction.
When the PSi layer is infiltrated with charged molecules
the electricaldouble layer changes.Charge redistributionand
changes in the dielectric constant take place. Fig. 8 shows a
schematic of the effect produced by positive charge on the
surface and the simplified electrical equivalent circuit. The
molecule’s charge on the surface changes the width of the
space charge region (W
d
) and the majority carrier distribu-
tion underneath it, which in turn influences C
rod

along with the change of each variable.
M. Archer et al. / Sensors and Actuators B 106 (2005) 347–357 353
Fig. 9. Real-time capacitance (C) and conductance (G) measurements of individual porous silicon sensors upon exposure to (a) chloroform, (b) acetone, (c)
ethanol, and (d) acetonitrile. In each case 10 ␮l of the solvent was added at a time indicated by the black arrow.
change in capacitance (%C) and conductance (%G) with
respect to reference. A negative (positive) value corresponds
to a reduction (increase) below (above) the reference value.
Fig. 10 shows that the results of Table 2 are correlated with
the dielectric constant.
According to the SCRM model infiltration of the pores
with a different material changes C
pore
. We need to explain
why chloroform,ethanol andacetone producea negative shift
with respect to reference. If the only variable involved in the
response was C
pore
then as the solvent penetrates the pores a
reduction in the capacitance would never be observed. This
suggests that a mechanism other than pore filling needs to
be considered and that the other physical properties (water
solubility and dipole moment) of the molecules affect C
rod
,
specifically via the electrical double layer as described ear-
lier. The four polarsolvents evaluated in this partofthestudy
possess different degrees of water solubility and a dielectric
constant larger than three. The water solubility influences
the adsorption on the surface and the value of the dielectric
constant the type of polarizability. For a dielectric constant

rod
C
pore
C
pore
+ C
rod
(2)
Fig. 10. Measured change in capacitance (%C) and conductance (%G)
with respectto thereference value as a function of the dielectric constant for
chloroform, acetone, ethanol and acetonitrile.
354 M. Archer et al. / Sensors and Actuators B 106 (2005) 347–357
If C
rod
 C
pore
then C
PSi
≈ C
rod
and, since C
rod
decreases
as a result of the increase in W
d
, then the change will be
negative (−C
PSi
). If C
rod

 C
pore
. The mod-
ulation of the space charge region width in the c-Si rods is
similar to what has been reported for silicon nanowires [29]
and is strongly dependent on the surface charge.
Fig. 11presentsasimplesimulationofthe modelvariables
over time for molecules acting as a positive and a negative
charge on the surface. Both the time scale and the values
shown are arbitrary values for illustrative purposes only. In
these simulationswe havefurther assumedthat theequivalent
capacitance (C
eq
) is given by the porous silicon capacitance
(C
PSi
) in series with a fixed junction capacitance (C
junct
)at
each contact and that the equivalent conductance (G
eq
)is
given by the substrate conductance (G
subs
). The conductance
is considered to be directly proportional to the number of
carriers in the c-Si substrate (N
A
) as well as the width of the
channel (a). Starting with a dry device (air in the pores) three

irreversible modification of the oxidized surface takes place.
Similar considerations apply for acetonitrile, which has not
been studied on oxidized surfaces but can be chemically ad-
sorbed on clean silicon surfaces [31]. The relevance of the
oxide properties on the sensitivity of PSi has been widely in-
vestigated by Sailor and coworkers [13]. It is also interesting
to notice that acetonitrile and chloroform have the highest
ionization potential of the four solvents tested (12.194 and
11.37 eV, respectively). The energy of chemisorption is the
difference between the work function of the semiconductor
and the ionization energy of the molecule [32]. If chemisorp-
tion is taking place then changes in the surface charge influ-
ence the response of the device. When the samples initially
exposedtoacetonitrileweresubsequentlyexposedtoethanol,
no response was observed which confirms that exposure to
acetonitrile produced a permanent surface modification. This
wasfurther confirmedwhenthe sensorswerereexposedto the
solventfor severaltimes allowingtherecovery ofthe baseline
between additions. For ethanol and acetone, in which the re-
sponse was reversible, the variation of the maximum change
in capacitance (%C) andconductance(%G) was no more
than ±1% with a maximum shift in the baseline of 5% over
a 20min timeline. For acetonitrile and chloroform, the re-
sponse is not reversible therefore the same sensor could not
be tested but the reproducibility of the response in different
sensors was good (±5%).
4.2. Influence of the channel width (a) in the response
It was mentioned before that the value of G
subs
depended

ductance (larger – %G) as the channel becomes narrower
(thicker PSi layers) up to 80 ␮m. This supports our assump-
tion that G
subs
is affected by the channel width (a). It is
likelythatthe maximumresponse occurswithin thefirst 50or
60 ␮m PSi thickness and after this depth has been infiltrated
the sensitivity of W
d
to any further charge compensation is
reduced. At this point the effect of W
d
and the carrier dis-
tribution may be defining the limits of sensitivity rather than
the geometry changes.
4.3. Response to water
It has been demonstrated that water increases the conduc-
tance [8,33] and capacitance [15,34] of PSi. This is the basis
for PSi humidity sensors. A change in dielectric constant,
dipole moment and possible chemisorption or physioadsorp-
tion on the surface of porous silicon has been proposed to
explain the response. An additional characteristic is the in-
trinsic dipolemoment of thewater moleculethat confers pure
orientation polarizability.
Fig. 12. Real-time capacitance (C) and conductance (G) measurements of the porous silicon sensor upon exposure to water. (a) The first part of the response
with a coupled behavior in capacitance (C) and conductance (G) is observed over the first minutes of the response. (b) Over a large period of time the two
responses are different.
TheexperimentalresultsshowninFig. 12canbeexplained
with the SCRM model and the factors previously mentioned.
We identified two phases intheresponse. The first one shown

pore
changes in such way that C
rod
 C
PORE
, making C
PSi
≈ C
pore
and producing a positive signal (+C
PSi
).
4.4. Response to non-polar molecules
Schechter and coworkers [8] reported an enhancement of
PSi conductivity withexposure tomoleculeswith zero dipole
moment. They suggested that the conductivity enhancement
could be related to other factors aside from the dipole mo-
ment. To investigate this possibility we performed experi-
ments using benzene (µ =0D) and toluene (µ = 0.43D).
Both exhibit a very low water solubility and given the value
of their dielectric constant (ε = 2.27 for benzene, ε = 2.38 for
toluene) the polarization of these molecules is purely elec-
tronic. The characteristic response in capacitance and con-
ductance along with the measured change in these variables
is shown in Fig. 13.
Since we do no have an independent way to probe the
influence of the field on the molecule inside the pores the
356 M. Archer et al. / Sensors and Actuators B 106 (2005) 347–357
Fig. 13. Real-time capacitance (C) and conductance (G) measurements of porous silicon sensors upon exposure to (a) benzene and (b) toluene. In each case
10 ␮l of the solvent was added as indicated by the black arrow.

of the device. Reversibility of toluene and benzene has also
been reported in as anodized luminescent PSi [12,30].
5. Conclusions
The large surface area of porous silicon and the sensitiv-
ity of its surface to charge molecules make it an ideal can-
didate in sensor development. We have evaluated the use of
a new electrical sensing device based on macroporous sili-
con (pore diameter 1-2 ␮m) layers in which the contacts are
made on the backside of the substrate. This approach allows
complete exposure of the surface without the presence of
metallic contacts on the surface. The sensitivity of this de-
vice is not only related to the dipole moment and the dielec-
tric constant of the molecules but also to their interaction
with the surface and the alignment of their dipole. Molecules
with different electrical and chemicalcharacteristicsproduce
a change in magnitude and sign in the capacitance and con-
ductance. To explain our results, we proposed a space charge
region modulation (SCRM) model that considers the effect
of changes in the dielectric constant of the porous silicon
matrix along with the interaction of different molecules with
the surface. The simulations performed consider the simul-
taneous change in dielectric constant and charge distribution
induced bymoleculeswithdifferent propertiesand theresults
obtained are in accordance with our experimental results.
In this paper, we demonstrated the use of our device as
a chemical sensor capable of producing a different response
upon exposure to water, ethanol, acetone, chloroform, ace-
tonitrile, benzene and toluene. The sensitivity to charged
molecules can however extend the use of these devices to
biological applications. We have also demonstrated the use

[9] I. Schechter, M. Ben-Chorin, A. Kux, Gas sensing properties of
porous silicon, Anal. Chem. 67 (1995) 3727–3732.
[10] S. Green, P. Kathirgamanathan, Effect of oxygen on the surface
conductance of porous silicon: towards room temperature sensor ap-
plications, Mater. Lett. 52 (2002) 106–113.
[11] J. Rehm, G.L. McLendon, P.M. Fauchet, Conduction and valence
band edges of porous silicon determined by electron transfer, J. Am.
Chem. Soc. 118 (1996) 4490–4491.
[12] T.A. Schmedake, F. Cunin, J.R. Link, M.J. Sailor, Standoff detection
of chemicals using porous silicon “Smart Dust” particles, Adv. Mater.
14 (2002) 1270–1272.
[13] T. Gao, J. Gao, M.J. Sailor, Tuning the response and stability of
thin film mesoporous silicon vapor sensors by surface modification,
Langmuir 18 (2002) 9953–9957.
[14] M. Archer, P.M. Fauchet, Electrical sensing of DNA hybridization
in porous silicon layers, Phys. Stat. Sol. (a) 198 (2003) 503–507.
[15] R.C. Anderson, R.S. Muller, C.W. Tobias, Investigation of porous
silicon for vapor sensing, Sens. Actuat. A A21/A23 (1990) 835–839.
[16] A. Motohashi, M. Ruike, M. Kawakami, H. Aoyagi, A. Kinoshita,
A. Satou, Identification of water molecules in low humidity and pos-
sibility of quantitative gas analysis using porous silicon gas sensor,
Jpn. J. Appl. Phys., Part 1 35 (1996) 4253–4256.
[17] A. Motohashi, M. Kawakami, H. Aoyagi, A. Kinoshita, A. Satou,
Gas identification by a single gas sensor using porous silicon as the
sensitive material, Jpn. J. Appl. Phys., Part 1 34 (1995) 5840–5843.
[18] K. Watanabe, T. Okada, I. Choe, Y. Sato, Organic vapor sensitivity
in a porous silicon device, Sens. Actuat. B 33 (1996) 194–197.
[19] S J. Kim, S H. Lee, C J. Lee, Organic vapour sensing by current
response of porous silicon layer, J. Phys. D: Appl. Phys. 34 (2002)
3505–3509.

Institute of Electrical Engineers, London, UK, 1997.
[31] F. Tao, Z.H. Wang, M.H. Quiao, Q. Liu, W.S. Sim, G.Q. Xu, Cova-
lent attachment of acetonitrile on Si(1 0 0) through Si
C and Si
N
linkages, J. Chem. Phys. 115 (2001) 8563–8569.
[32] A. Many, Y. Goldstein, N.B. Grover, Semiconductor Surfaces, Wiley,
New York, 1965.
[33] F. Moeller, M. Ben-Chorin, F. Koch, Post-treatment effects on elec-
trical conduction in porous silicon, Thin Solid Films 255 (1995)
16–19.
[34] Z.M. Rittersma, A. Splinter, A. Boedecker, W. Benecke, A novel
surface-micromachined capacitive porous silicon humidity sensor,
Sens. Actuat. B 68 (2002) 210–217.
[35] S J. Kim, J Y. Park, S H. Lee, S H.J. Yi, Humidity sensors using
porous silicon with mesa structure, Phys. D: Appl. Phys. 33 (2000)
1781–1784.
[36] M. Archer, M. Christophersen, P.M. Fauchet, Macroporous silicon
electrical sensor for DNA hybridization detection, Biomedical Mi-
crodevices 6 (3) (2004) 203–224.
Biographies
Marie Archer received her Masters Degree in Biomedical Engineering
from the University of Rochester in May 2003. The topic of her Ph.D.
research was the characterizat
ion of electrical porous silicon based sensors
for their use in biodetection. She carried out her research at the Center
for Future Health under the supervision of Professor Philippe Fauchet
and received her Ph.D. degree in May 2004.
Marc Christophersen received his Masters Degree in Engineering in
November 1998 from the Christian-Albrechts University in Kiel, Ger-

Engineering, and a member of the Materials Research Society.