J. Phys. Chem. 1994,98, 8765-8774
8765
Measurement of Barrier Heights of Semiconductor/Liquid Junctions Using a Transconductance Method: Evidence for Inversion at n-Si/CH30H-l,l'-Dimethylferrocene+'o Junctions Paul E. Laibinis,+Colby E. Stanton, and Nathan S. Lewis* Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California 91 125 Received: March 9, 1994; In Final Form: May 23, 1994"
Transconductance measurements have been used to characterize the space4harge regions of various n-Si/ liquid contacts. To perform these measurements, Si electrodes were photolithographically processed to introduce p+-contact areas into the surface of an n-type Si electrode. The electrical conductance between these p+ regions was then used to probe the minority carrier concentration in the near-surface region of the n-type Si. Unlike conventional differential capacitance or current-voltage measurements, these transconductance measurements can be performed under near-equilibrium conditions and can be performed in the presence of gaseous ambients or when the sample is in contact with ionically conducting electrolyte solutions. In contact with the electrolyte solutions, faradaic and solid-state conduction pathways were distinguished using ac impedance measurements. The impedance spectra provided clear evidence that contact with MezFc (l,l'-dimethylferrocene)+/O and Fc+/O redox couples in CH30H(l)-1.0 M LiC104 formed an inversion layer in the n-Si, but that CH3OH(l)-l.O M LiClO4-MeloFc+/O solutions did not yield an inversion layer. These observations are consistent with prior current-voltage measurements on these junctions. The barrier heights of the n-Si/CH30H-Me2Fc+/O and n-Si/CH30H-Fc+/O junctions were determined to be 1.01 and 1.e2 V, respectively. These measurements provide new insight into the photoelectrochemical behavior of S I / C H ~ O Hcontacts and provide an alternate method for characterizing the energetics of semiconductor/liquid contacts.
I. Introduction The barrier height, &, is a key parameter in defining the photoelectrochemical operation of a semiconductor/liquid junction (Figure l).I In particular, &, defines the available driving force for interfacial electron transfer and determines the strength of the electric field that is responsible for separating charges created by illumination of the semiconductor.1J Despite the fundamental importance of '#Jb, barrier heights have not yet been determined unambiguously for most semiconductor/liquid contacts. Most methods of barrier height measurement at semiconductor/liquid contacts involve significant uncertainties and/ or simplifying approximations.l+5 Specifically, surface states at the solid/liquid interface often confound conventional barrier height measurements, because such states generally contribute additional frequency-dependent capacitance and/or induce hysteresis in the steady-state current-voltage behavior.lpbll To address these issues, this paper describes an alternate method for measuring the barrier height of a semiconductor/liquid junction. In this method, the conductance of the semiconductor is measured parallel to the semiconductor/liquid contact (Figure 2).I2J3 The measurement can be performed under conditions of charge-transfer equilibrium at the solid/liquid contact and does not require extrapolation from severe reverse bias (as in capacitance-voltage measurements)lJ or from severe forward bias (as in current-voltage mea~urements).~,~ The method is inherently free from conductance effects of immobile carriers that are localized in surface states, because only mobile carriers in the near-surface region can contribute significantly to the conductance measured parallel to the solid/liquid interface. Through use of Poisson's equation, the near-surface conductance can be analytically related to the surface carrier concentration, and thus to the barrier height, of the semiconductor/liquid contact.14J5
* To whom correspondence should be addressed.
Present address: Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139. *Abstract published in Aduance ACS Abstracts. August 1 , 1994. f
0022-3654/94/2098-8765$04.50/0
[Semiconductor] Figure 1. Band diagram for a depleted n-type semiconductor/liquid junction. The rate of heterogeneous electron transfer from the semiconductor depends on various factors: the heterogeneous electron-transfer rate constant, k,,, (with units of cm4 s-I), the solution concentration of electron acceptors ([A]), the electron concentration in the bulk of the voltage" semiconductor (nb), and the potential difference-'built-in (Vb)-between the potential of theconduction band (QB/q) at the surface and in the bulk of the semiconductor. Ef is the Fermi level of the semiconductor and is in equilibrium with the electrochemical potential of the contacting solution (E(A/A-)). EVBis the position of the top of the valence band, I#Jb is the barrier height of the semiconductor/liquid junction, k is the Boltzmann constant, and T is temperature.
The specific semiconductor/liquid junctions of interest in this work are based on n-Si/CHsOH contacts. n-Si/CH30H-11 1'dimethylferrocene (MezFc)+/O contacts have been shown to provide efficient, stable photoelectrochemical cells.1+18 Surprisingly, nearly intrinsic Si samples (with n+-dopedback contacts), under high-level injection conditions, have displayed high-short circuit quantum yields in contact with CH30H-Me2Fc+10 and have proven to be highly efficient photoelectrodes.19 This behavior is not expected on the basis of simple photoelectrochemical theories,20?21because the lack of significant internal electric fields in a semiconductor under high-level injection should induce substantial charge recombination near the surface of the solid. Furthermore, the presence of a high concentration of donors and acceptors in the solution phase of the Si/CH3OH-MezFc+/O cell, 0 1994 American Chemical Society
8766 The Journal of Physical Chemistry, Vol. 98, No. 35, 1994
Laibinis et al.
iL 1
,. ,, ...._._ ......_._..._..._....... _..._ ........ , ... ~
*A\
1
--+q
Figure 3. Band diagram of an n-type semiconductor/liquid junction in which the near-surface region of the semiconductorcontains an inversion layer. In this situation, the semiconductor/liquidjunction device acts as if an "in situ" n-p homojunctiondevice were created in the semiconductor sample. Under these circumstances, interfacial recombination at the solid/liquid contact is much less severe than if the inversion layer were absent, and the solid/liquid junction photoelectrochemicalcell performs similarly to a diffused junction solid-state solar cell that is placed electrically in series with a regenerative electrochemical cell. A IK
Figure 2. Schematic illustrations of the (a) Mott-Schottky and (b) transconductancemethods for probing the semiconductor/liquidinterface. A primary difference in these methods is that the ac current crosses the semiconductor/liquid interface in a but not in b. (c) Equivalent circuit for the transconductance arrangement containing the following reduced elements: the impedancethrough the semiconductornear-surface region, Z,, represented as the near-surface resistance, R,, in parallel with the near-surface differential capacitance, Csc;the impedance through the bulk of the Si wafer, zb, represented as a resistance, Rb, in parallel with a capacitance, c b ; and the impedance through the electrolyte solution, Z,l,, which is represented as a double-layer capacitance, c d l , an ohmic series resistance through the solution, Rn,a charge-transfer resistance, R,, and a faradaic Warburg impedance, 2 .,
and the absence of significant electric fields in the solid to repel one carrier or the other from the solid/liquid junction,lg should allow significant recombination at the solid/liquid interface. One proposed explanation19 for the observed experimental behavior is the formation of an inversion layer at the Si/liquid contact (Figure 3). To obtain such a situation,the semiconductor/ liquid interface must have such a large initial contact potential difference that charge-transfer equilibrium between the solid and liquid phases requires ionization of Si lattice atoms as well as ionization of dopant impurity atoms.12-15*22*23 The thin minority carrier inversion layer would shield the electrons created in the semiconductor bulk from deleterious recombination effects at the solid/liquid contact. In fact, the recombination processes in
'
such a device would be characteristic of a prefabricated, solidstate, n+-i-p+ Si device in which neither metallurgical junction is a site of high recombination.24925 This situation is in accord with the experimental behavior observed for the Si/CH3OHMezFc+/Osystem19 and provided the motivation for performing the barrier height measurements described herein. The near-surface conductivity measurements presented in this work are designed to exploit the unique, diagnostic electrical characteristics that should be present in the surface region of an inverted semiconductor. If a depletion region is established, the conductance of the near-surface region will be lower than that of the bulk sample, due to the loss of *mobilemajority carriers near the solid/liquid contact. However, if an inversion layer is established, the formation of mobile minority charge carriers in the near-surface region should produce a significant increase in the electrical conductance of this region of the specimen. The conductance measurement can be made selective to minority carriers and thus can be diagnostic for the presence of an inversion region, if minority carrier contacts (i.e., p+-doped regions in our n-type samples) are used to probe the electrical properties of the near-surface region of the solid (Figure 2).26927 To date, near-surface "transconductance- measurements have had limited application to the characterization of semiconductor/ liquid junctions.12J3 Experiments using n-MoSe212 and p-Si electrodes13 have employed transconductance measurements to establish the presence of an inversion layer at the semiconductor/ liquid interface. These experiments employed two electrical leads that were positioned close to the surface of the semiconductor/ liquid junction, with changes in conductance between these solution-phasecontacts monitored upon addition of a redox species to the solution. In somecases, significant increasesin conductance were observed, and such measurements were consistent with the presence of an inversion layer.12.13 However, the conductances obtained using this method included unquantified contributions from the solution and a poorly-defined contact-sample geometry and were thus not amenable to a rigorously quantitative analysis of either the barrier height or the inversion layer thickness. The Si/CH30H-Me2Fc+/O system has many advantages for such transconductancemethods. The processing of silicon is well- . established so that minority carrier contacts can be lithographically defined, implanted, and fabricated into a Si photoelectrode in a straightforward fashion.28 The contact depth and contact separation can therefore be known and controlled so that a quantitative interpretation of the transconductance data is possible. The functional form of the relationship between the channel conductance and the surface potential is well-known for Si through reference to the extensive database on Si-based solidstate field-effect devices.14J9 Furthermore, Si/CH30H-Me2-
Barrier Heights of Semiconductor/Liquid Junctions Fc+/O contacts have been shown to exhibit "ideal" solid/liquid junction b e h a v i ~ r , ~ J ~ and J ~ , ~the O electrochemical potential of the solution phase, E(A/A-), in CH30H(l) can be readily manipulated by choice of the substituents on the ferricenium/ ferrocene (Fc+/Fc) core. Thus, redox couples with positive electrochemical potentials, such as Me2Fc+/O and Fc+/O, are expected to produce an inversion layer and should lead to significant increases in the transconductance of the n-Si sample, while redox couples with more negative E(A/A-) values, such as MeloFc+/O, should produce a depleted spacesharge layer and should not induce a significant mobile minority carrier concentration in the n-Si ~ample.~J3J~JO This behavior provides a convenient test for the sample, and for the method, without requiring a significant change in the chemical structure of the solid or liquid phases that form the electrical contact. 11. Method of Characterization
Three parallel conductance pathways are present in a device of the type represented schematically in Figure 2b. A conductance pathway with impedance Z b is present through the bulk of the semiconductor crystal, another pathway with impedance Z,, is present through the near-surface region, and a third pathway with impedance Zsolnis present through the solution. A simplified equivalent circuit for this device is represented by the three reduced impedance elements in Figure 2c. Our goal is to separate these impedances so that the desired resistance, the real part of Z,, can be determined experimentally. To obtain an expression for the impedance behavior of the complete device, each component will be discussed individually below and then will be placed in the equivalent circuit of Figure 2c. Zb is relatively unimportant in our configuration, because the use of minority carrier contacts ensures that the impedance through the bulk of the crystal is so large that it will not influence the propertiesof the device when in contact with our electrolytes.25 Current flow through the bulk of the crystal would require charge movement through two high-impedance contacts (Le., through each p+-n junction) that are connected in series. Before use in our experiments, each p+-bulkjunction was independently verified, using current density vs voltage (J-V) measurements, to be highly rectifying to charge motion. Thus, the thickness of the sample and any possible conductance shunts through the bulk crystal are unimportant in our experimental work. Another pathway for conductance between the p+-contacts involves charge flow through the electrolyte, Zsoln. In order to ensure that any inversion region could extend to contact each of the doped p+-regions, our device configuration was designed to permit a small region of each p+-area to contact the electrolyte solution. During the impedance measurement, this sample geometry therefore necessarily permits some unwanted faradaic conductance between the oppositely-biased p+ solid/liquid contacts. A similar unwanted conductance path would be produced if regions of the channel adjacent to the p+-contacts facilitated a low-impedance faradaic shunting path of thechannel through a pathway involving the electrolyte solution. Both of these pathways for Zsolnare characterized by an electrochemical cell i m p e d a n ~ econsisting ,~~ of a Warburg-type faradaic impedance at low frequencies and a double-layerdominated nonfaradaic electrochemical impedance at high frequencies. The impedance through an electrochemical cell has a known, characteristic functional form that depends both on the measurement frequency ( w ) and on the concentration of electroactive species.31 The inclusion of terms for the solution resistance, Rn, and the charge-transfer resistance, Rct,in this circuit path yields eqs 1a,b as the expressions for the real ( Z s o l n , ~ e ) and imaginary (Zsoln,lm) parts of Zsoln.31 In eqs 1 and 2, DO and DR are the diffusion coefficients for the oxidized and reduced forms of the redox species, respectively, CO* and CR* are the bulk concentrations of the oxidized and reduced forms of the
The Journal of Physical Chemistry, Vol. 98, No. 35, 1994 8767
redox species, cdl is the double-layer differential capacitance, R is the molar gas constant, T i s the temperature, n is the number of electrons transferred in the reaction converting one molecule of 0 to one molecule of R, F is the Faraday constant, A is the area of the relevant electrode, Rctis the electrochemical charge transfer resistance, and RQis the series resistance of the cell. At low frequencies, the electrochemical impedance is dominated by the faradaic, Warburg impedance. As w 0, eqs 1a,b reduce to
-
Under these conditions, a Nyquist plot (Zsoln,lm vs will have a slope of unity.31 The faradaic impedance can therefore be experimentally identified through its 4 5 O phase angle 9, relative to theapplied acvoltage and through itscharacteristicdependences on w, CO*,and CR*. At high frequencies, the nonfaradaic impedance elements dominate the expressions in eqs la,b, and the impedances reduce to
A Nyquist plot of Zsoln,~m vs Zsoln,~c at high frequencies therefore should result in a semicircle centered at Zsoln,Re = Rn + Rct/2, Zsoln,l,,, = 0, with a radius of Rct/2.31The very high frequency limit of this expression tends toward the values Zsoln,Re
= RQ
(5a)
These last expressions make sense because at sufficiently high frequencies, the double-layer capacitance, in series with the ohmic resistance of the cell, should dominate the electrochemical impedance. Because Cd is dependent on the supporting electrolyte concentration, both frequency limits of the electrochemical impedance can therefore be manipulated, and identified, through changes in the supporting electrolyte concentration, measurement frequency, and redox couple concentration. We now consider the magnitude and frequency dependence expected for Zsc,It is convenient to represent this pathway by a resistance and a capacitance in parallel (eq 6).
The Journal of Physical Chemistry, Vol. 98, No. 35, 1994
8768
Laibinis et al. Q,
= [2kTc,]1t2ND[
(
+ -qvbi kT
1)
+
Combining the various parallel elements into one equivalent circuit, the total impedance is expressed by eq 7. As mentioned
I + - + -1
1-=
=b
'tot
'sc
1
(7)
'soh
earlier, the impedance term due to the bulk semiconductor, Zb, can be considered to be negligible; therefore, the measured impedance will primarily result from the parallel impedances of the semiconductor spacexharge region and of the solution. At high frequencies, the measured impedance (Ztot)is described by eq 8. Thus, Z,,, will be dominated by C,,, which is not directly related to I#+,.
At low frequencies, either Zsc(=R,,) or Z,,I, will dominate the measured impedance. Equation 9 is obtained from substituting the low-frequency expressions for Z,,I, (eq 3) and Z,, (eq 6) into eq I . 1 R,, Ztot
1
1 -+ -=
R,
(9)
+ R,, + uw-1/2- iuw-'I2 - 2ia2cd,
Two scenarios for the behavior of Z,,, a t low frequencies are then envisioned: if ]Zscl> IZsolnl,(Le., if R,, is large), then Z,,, will display the characteristic Warburg behavior (Le., 4 45O as w 0), and JZ,,,Jwill increase with decreasing frequency (eq 3). Alternatively, if lZw,,,1> JZ,(at low frequencies, theimpedance will asymptote with decreasing frequencies to a value (R,,) that is purely resistive (i.e., = 0" as w 0) and will become independent of frequency (eq 9). These differences can be used to determine qualitatively whether a particular redox species has produced an inversion layer in the semiconductor near-surface region. The impedance of the semiconductor near-surface region depends on the mobile charge carrier concentration. If the surface is in depletion, the carrier concentration in the near-surface region will be lower than in the b ~ l k . The ~ ~channel , ~ ~ will then be more resistive than the bulk of the solid, and at low frequency Z,, will be larger than zb. However, in an inversion layer, the presence of significant mobile minority carriers will lead to a rather low resistance in the ~ h a n n e l . I ~ ~Furthermore, ~6~2~ the capacitance of the near-surface region in either depletion or inversion conditions will be relatively independent of frequency (for w < 1 MHz investigated herein).14 Thus, if an inversion region were present, Z, would be expected to be rather small, independent of frequency, and primarily resistive in nature, a t the low frequencies of interest in this work (Le,, Z,, = Z,,,R~= R,, and Zsc,lm= 0). The measured resistance of the near-surface layer of an inverted semiconductor can then be used further to quantify the barrier height of the junction. In an inversion region, the resistance across the channel is directly related to the number of minority carriers in the n-Si by eq l0,where L is the separation between
-
-
$t
-
R = L/pLpQnW
(10)
the p+-doped regions (=0.10 cm), pp is the mobility of holes in a Si inversion layer (=lo0 cm2 V-I ~ - 1 ) , 3 ~ . 3 3Qn is the density of minority carriers within the inversion layer of the semiconductor, and Wis the width of the channel between the p+-doped contact regions (=0.84cm). The minority carrier density is approximately given by the difference between the total charge density removed from the semiconductor, Q,, and that required to reach the point of inversion, Qb, eqs 11-13.14 In these equations, e, is the
permittivity of Si (=l.O5 X 10-I2 F cm-I), q is the elementary charge (11.60 X C), Vbi is the difference between the conduction band edge at the surface and in the bulk of the semiconductor, vi is the voltage drop at the surface required for theonsetofinversion (=(E,/q)-2V,;14seeFigure l),and NAand NDare the acceptor and donor concentrations in the semiconductor, respectively. If the near-surface conductance could be measured experimentally, it would therefore permit a direct calculation of qVbi and thus of &-,. Wedescribesuch measurements in the subsequent sections of this report. 111. Experimental Section
1. Fabricationof p+-n-p+ Devices. Figure 4 displays schematic diagrams of a fabricated chip in a top view and in a cross section. To obtain these devices, a wafer of (100)-oriented n-Si (100-mm diameter, primegrade, phosphorus-doped, 525-pm thickness; from Silicon Sense Inc., Nashua, NH) was first cut into =1.5 cm X 1.5 cm pieces. The resistivity of this material was reported to be between 30 and 60 fl cm by the supplier, with a measured value of 40 fl cm using a four-point probe in our laboratory. Each piece was processed individually to yield one device on the polished face of a chip, with a typical run consisting of four chips processed in parallel. Figure 5 depicts the processing steps, described in detail below, that were used to produce the desired device configuration. The n-Si chips were first cleaned by sequential immersion in water (20 s), acetone (60 s), water (20 s), buffered H F (3 min), and water (20 s). An oxide layer (-4000 A) was then grown on the surface by a 30-min exposure to wet 02 in a tube furnace operating at 1100 OC. A thinlayer ofpositive photoresist (Shipley Microposit 1400-31) was applied by spin coating for 30 s at 4000 rpm and was subsequently soft baked by illumination with IR lamps for 5 min. To generate the first desired pattern, the oxidized and photoresist-coated chip was exposed for 1 min to a 100-W Hg lamp UV source (Ultraviolet Products, San Gabriel, CA) through a glass mask. The mask was a 2.5 in. X 2.5 in. high-resolution emulsion glass plate (Imtec, Sunnyvale, CA) that was prepared using standard black and white photographic and slide processing techniques. The mask contained the three patterns for the fabrication process, and each included alignment markers for obtaining registry among the three patterns. After exposure, the sample was immersed in Microposit CD30 Developer (Shipley) for 30 s, immersed in water for 15 s, and blown dry in a stream of Nz. The photoresist was then hard baked in a 120 OC oven for 5 min. A sequential immersion of the chip in water (20 s), buffered HF (3 min), water (20 s), acetone (60 s), water (20 s), RCA solution (5:l:l H20:NH40H:30% HzOz(aq) at 80 "C for 5 min), and water (20 s) produced two parallel openings (0.25 X 0.84 cm2) in the oxide that were separated from each other along their longer dimension by 0.10 cm. These exposed regions were doped p-type at 1000 OC in a quartz tube under flowing N2(g). The quartz tube contained the Si chips and =1.3 cm X 1.3 cm pieces of activated boron nitride (Carborundum PDS BN-925) that faced the polished surfaces of the n-Si chips. The BN sources
Barrier Heights of Semiconductor/Liquid Junctions a) Top yiew
The Journal of Physical Chemistry, Vol. 98, No. 35, 1994 8769
I
n-Si
1
Field Oxidation
I/p+-Si Region t
1 I
Y
p+-Si
I
I'
Si02
n-Si
1) Masking 2) HF etch
H
1 mm
b) Side view Boron Doping
1 1 Back contact
1 mm
Figure 4. Top and side schematic illustrations of the p+-n-p+ device. (a) A fabricated device. The four rectangles inside the A1 areas represent the only regions of direct contact between the A1 pads and the p+-regions. In other areas, the A1 points and the n-Si were separated by a layer of Si02 and had no direct contact. The dashed lines mark where the fabricated chip was subsequentlyscored and the edges were then removed prior to mounting the electrode. (b) A mounted electrode. The vertical dimension is highly magnified relative to the horizontal dimension and is not drawn to scale. The exposed oxide was etched before use.
were activated by exposure to dry 02(g) for 30 min at 950 "C and were then stabilized for at least 30 min under a flow of N2(g) at 950 OC prior to use. After removal from the boron doping furnace, the chip was immersed sequentially in water (20 s), a boron glass etch solution (1:lOO:lOO HF:HN03:H20 for 20 s) and water (20 s) and was then blown dry in flowing Nz(g). To drive the dopant atoms further into the n-Si, the chip was then placed in a gate oxide furnace. In this furnace, the chip was exposed at 1100 OC to a flow of dry Oz(g) for 30 min and then to a flow of N2(g) for 10 min. The chip was removed from the furnace and allowed to cool to room temperature. A lithographic procedure of masking, baking, and etching (identical to that described above) was then used to open two separated 0.14 cm X 0.28 cm areas in the gate oxide over each of the B-doped regions. After this process was complete, a layer of photoresist was again spin-coated onto the sample and subsequently soft baked. The sample was then exposed to the UV source through a mask, developed, and blown dry. A thin film of A1 ( ~ 1 0 0 0A) was evaporated onto the sample from a resistively heated tungsten filament in a diffusion-pumpedvacuum chamber operating at =l X Torr. Ultrasonic cleaning of the Al-coated chip in acetone produced four A1 pads ( ~ 0 . 1 1cm2 each) that contacted the B-doped regions through the opened regions in the oxide. As a final step, the chip was annealed at 200 "C for 5 min to improve electrical contact between the A1 pads and the B-doped wells. Before transforming the devices into semiconductor electrodes, the fabricated area of each chip was isolated from any edge effects of the processing by removing ~ 0 . cm 2 of material from each of the four edges. Ohmic contacts were then made to the samples by scratching a Ga:In eutectic into the back (unpolished) side of the Si chip using a diamond-tipped stylus. A1 wires were electrically contacted to each of the four front A1 pads and to the back contact using Ag print (GC Electronics). The chip and the five points of contact were encapsulated in white epoxy (Dexter
Si02
Drive-in/ Oxidation
7
Si02
I
I
n-Si
1
1) Masking 2) HF etch
f)
1
Aluminum deposition
AI
1) Remove edges 2) Connect leads I
n
-
n-Si
I
I
.
n
I
.
I
1
I
Measure Impedance spectrum
"Isi
~I
HFetch
t
epoxy
-i 1 mm H
I
Figure 5. Schematic illustration of the fabrication of the p+-n-p+ device. The vertical dimension is highly magnified relative to the horizontal dimension.
Corp.) so that only the portion of the n-Si chip between the two p+-doped wells (-0.84 cm by 0.10 cm) and a minimum of the p+-wells were visible. None of the A1 pads were left exposed.
Laibinis et al.
8770 The Journal of Physical Chemistry, Vol. 98, No. 35, 1994
A working device was one that had ohmic responses (with resistances on the order of 0.1-10 kR) for leads connected across a common B-doped region, rectifying responses between the front leads and the back contact and high resistances (>1 MQ) across the p+-n-p+ connection. The resistivity across the p+-n-p+ connection was a function of the incident light intensity, and this light-sensitive behavior provided evidence that the prepared devices sensed changes in the carrier concentrations in the channel (in this case, changes due to the photogeneration of electron-hole pairs in the n-Si). To obtain a working device that met the above operational criteria, it was often necessary to cycle the leads connected across a common B-doped region between =:f5 V with a potentiostat until an ohmic response with a resistance of