~D-R192 963 PIOTOLUMZNESCENT PROPERTIES OF P-GRAS ELECTRODES 1RELATED TO THE "PHOTOCUR.. (U) WISCONSIN UNIV-MRDISONDEPT OF CHEMISTRY P 9 JOHNSON ET AL. 88 JUL 87

UNCLASSIFIED UiIS/DC/TR-87/4 N@iS14-85-K-8631 F/G 9/5 NL

ONE

L "-' ,v" - w- . - ' w w -mill - wl- .-l " l - w.: ... - -,-ra -

II'U

| &-I .t 4

OFFICE OF NAVAL RESEARCH

Contract N00014-85-K-0631 l FILE CtR&T Code 4134003

Technical Report No. UWIS/DC/TR-87/4

Photoluminescent Properties of p-GaAs Electrodes Relatedto the "Photocurrent Anomaly": Determination of SurfaceElectron-Capture Velocities and Depletion Widths in

Photoelectrochemical Cells

by

Phelps B. Johnson,.Christopher S. McMillan, Arthur B.Ellis, and William S. Hobson

CV Prepared for Publication in the [ 1 on For

Journal of Applied Physics >"Uainr) n cd

University of Wisconsin JX4tiicn tioroDepartment of Chemistryi Madison, Wisconsin 53706 By~Distribution/

Availability CodesJuly 8, 1987 -- Availand/or

Dist Special

40--

Reproduction in whole or in part is permitted for anypurpose of the United States Government

This document has been approved for public release andsale; its distribution is unlimited.

DTIC*To whom all correspondence should be addressed : J U L 211987

E11"

Unclassified A p qSECUTY CLASSWICATION OF THIS PAGE

REPORT DOCUMENTATION PAGEIs. REPORT SECURITY CLASSIFICATION lb. RESTRICTIVE MARKINGS

NA NAra. SECURITY CLASSIFICATION AUTHORITY 3. DISTRIBUTION/ AVAILABILITY OF REPORT

NA Distribution Unlimited; Approved2b. DECIASS1ICAT1ON I DOWNGRADING SCHEDULE for Public Release

NA4. PERFORMING ORGANIZATION REPORT NUMBER(S) S. MONITORING ORGANIZATION REPORT NUMBER(S)

UWIS/DC/TR-87/4 NA

6a. NAME OF PERFORMING ORGANIZATION 6b. OFFICE SYMBOL 7a. NAME OF MONITORING ORGANIZATIONChemistry Department (/ aplcble)

University of Wisconsin-Madiso OR

SC. ADDRESS (Cty, Stet, and ZAP Code) 7b. ADDRESS (City, State, and ZIP Code)1101 University Avenue 800 N. Quincy St.Madison, WI 53706 Arlington, VA 22217

ft. NAME OF FUNDING ISPONSORING 8b. OFFICE SYMBOL 9. PROCUREMENT INSTRUMENT IDENTIFICATION NUMBERORGANIZATION (if applicable)

NA Contract N00014-85-K-0631Sc ADDRESS (Cty, State, and ZIP Code) 10. SOURCE OF FUNDING NUMBERS

PROGRAM PROJECT TASK R&T WORK UNITELEMENT NO. NO. NO. Code ACCESSION NO

4134003

11. TITLE (Include Security Clauification)Photoluminescent Properties of p-GaAs Electrodes Related to the "Photocurrent Anomaly":Determination of Surface Electron-Capture Velocities and Depletion Widths in Photoelectro-

2. PERSONAL AUTHOR(S)

13a. TYPE OF REPORT 13b. TIME COVERED 14. DATE OF REPORT (Yer, Month, Day) t5. PAGE COUNTTechnical FROM TO 6/20/87

16. SUPPLEMENTARY NOTATIONPrepared for publication in the Journal o pplied Physics

17. COSATI CODES 18. S UBJeC.TftS (Continue on reverse if necesso and idenify by block number)'LD GROUP SUB-GROUP 9W ; photoluminescence; dead ayer model; depletion

width, surface electronrcapture velocity; photoelectro-chemical cells I, , G'.. 4 )e" r

19. AB ACT (Continue on reverse if necessary and identify by block number)Steady-state photoluminescence (PL) measurements have been used to determine depletion

widths W and surface electron-capture velocities S for p-GaAs electrodes in aqueous acidicelectrolytes. Electrodes subjected to slow cycling of applied potential (10 mV/a) exhibita hysteresis in PL intensity in the "photocurrent anomaly" (PA) potential regime, character-ized by negligible photocurrent at voltages up to .5 V negative of the flatband potentialO0.1 V vs. SCE). A marked hysteresis in S values s observed in the Ppotential regimeof -0.1 to -0.4 V vs. SCE. Estimated S values arel high, approaching 17.lcm/s, but decreasesignificantly during anodic-going scans. Variationk in W between -0.1 4d -0.8 V vs. SCEindicate that strong Fermi-level pinning does not olktain. Electrode ope ation under a pulsedpotential program (-0.05 to -1.00 V vs. SCR) caused transient reductio in S of approxi-mately an order of magnitude.f ,' , "7, ,,

e#

S"0. DISTRIBUTION/IAVAILASILITY OF ABSTRACT 21 ABSTRACT SECURITY CLASSIFICATION[ UNCLASSIFIEDJNLIMITED 0- SAME AS RPT Q]OTIC USERS Distribution Unlimited

22a "NAME OF RESPONSIBLE INDIVIDUAL 22b. TELEPHONE (Include Area Code) 22€ OFFICE SYMBOLArthur B. Ellis 608-262-04.21

DO FORM 1473, 84 MAR 63 APR edition may be used until exhausted. SECURITY CLASSIFICATION OF THIS PAGEAll OtheIr edtios are obsollete

Unclassi fied

p}

4

I. INMhO)UCTION

Photoluminewence (L) techniques offer unique advantages for

the study of ph otoelectrochemlcal cells WWC's). The PL intensity of

smicondutors is sensitive to thi-depletim width W in the

semiconductor and to the rate of minority carrier loss at the

surface 8. A simple dead-layer model (DLMO has been employed to

relate changes in the steady-state PL intensity of a variety of

semclnductor electrodes to Changes in W.2 - 6 Both pulsed 7 and

steoady-state8 ,9 PL intensity measurements have been interpreted

in terms of changes in S. Recently, we have shown that a PL

model developed by Mettler o is capable of simutaneously assessing

values of W and 8 as a function of electrode potential. The model

was applied to n-GaAs electrodes in PEC's employing aqueous

telluride electrolyte.11

The non-idemlities in the photoeetrhmica behavior of

p-type m-v smionductors have bernm the subject of several

studie.12- 1 Of particular interest Is the apotourent anomaly"

(PA), a 0.4 to 0.6 V separation of the photcurrent onset from the

flatband potential.14 As a consemuec of this overpotential, the

onet of cathodic currents leading to hydrogen evolution Is cls to

that of conventional metal electrode and much of the theoretical

efficiency for converting optical to chemical energy Is lost.

For EC's employing p-GaAs or p-GaP electrodes in H2804

electrolyte, several theories have been advanced to account for the

observed PA. In one study, a thermally-limited value of S was a

scontributor to the effect. 14 Another study attributed the

PA to increases in 8 at potentials near the fiat-band potential Vfp,

5due to the amrrespanding higher c ncentration Of majority Carriers

at the srace-13 Also releant to an understanding at the PA is

Fermi-level pinning which has bexen proposed for aqueous p-CAAspcs.12U7 Transient, PL effects have been ascribed to a shift in

band edge in the PA potential regime.6

We a spot herein an the steady-stte and transient PL

properties of a p-GaAs--based PBC employing an aqueous acidi'

eletroyt. Our results reveal that over the potential regime

corrzesponding to the PA. values of 8 show cosiderable hysteresis;

the values are high, ma- mll approximiating the thermal limit.

Furthermore, significant variations in W observed between -0.1 and

-0.8 V vs. SCE indicate that the electrode is not strongly pinned.

Operating the electrode under a pulsed potential program cause a

transient reduction in 8 al approximately an order of magnitude.

UI. THEORY

A simple deed-layer model (UDM has. been used to estimate

changes in W from FL quenching data: The model assumes that

electron-hole pairs formed within a distance on the order of W do

not radiatively recambine. Thus, a chang In dead-layer thickness

AD as a function of applied potential is axpreAe by Eq. (1).

AD (-1/a') ln(PL1/PL2), I

where a, u(a +ap) is the sum at the absorptivities of the

semiconductor for the excitation and emission wavelengths,

ae.pectively; and P!.1 and PL2 are PL intensities at two arbitrary

potentials tar which the semiconductor is in depletion. If one of the

6electrode potentials is Vb then the total dead-layer thickness,

appro aig the depietion width, may be obtained. The DLM is

valid under conditions of high or constant virtual surface

recombination velocity 8.2

Use of AD values as a lower limit for W in Eq. (2) yields an

approxmat lower limit far the Schottky barrier height VB at the

semiconductor-electrolyte interface. In this expression, e is the

dielectric constant of the semiconductor (12.9 for GaAs18), e0 is the

perrnlttlvity

W . (2tte0V B/qNj A "/ [2]

of vacuum; q is the electronic charge; and NA is the acceptor

concentration.

Despite its strengths, the DLMs inability to accommodate

significant changes in 8 or to yield absolute depletion widths

restricts its applicability. A more powerful methodology was

developed by Mettler for the analysis of P1. fromGaAs in air.10 His

treatment, encompassing both the dead-layer formalism and the

influence of minority carrier loss at the surface upon PL intensity,

results in Eel. (3). In this equation, IL is the observed

IL K exp [-(ae.4up)W] 0 eLn

( *eLn) 2 1*(3

.~(Sr:GO(*L:~) (*e+GP)Ln

PL intensity divided by the excitation intensity, corrected for

7

reflectve lkes; x is a constant containing the internal quantum

efflicW ad g@a - trical factors Involved in light collection; Ln is

the electron diffusion length; and Or. is the reduced surface

elStv n-apture velocity. The reduced velocity is a 1--nsimles:

Waaut related to the more comm only determined surface

eectron-capture velocity 8 by Sq. (4), where 'Ini s the electron

lifetime. Our us of the term

Sr -3('TA~n)(4)

osurface electron-capture velocitya rather than the conventional

term asurtace recombination velocityu used by Mettler reflects the

fact that additional pathways for surface recombnation involving

interfacial charge transfer are available in a PEC.

The derivation of Eq. (3) assumes that the optical Pen-etration

depthi (OPID) Is significantly less than the minority carrier diffusion

length, i.e., £e'1 -(Ln. Given an estimate Of Ln. a Plot Ot IL vs- £e 1

(IL-OPD plots) can be fit to Eq. (3) to obtain both W and Sr. The

value of w obtained can be used to appromimt the Schottky

bailJer height VB with Eq. (2) (vide supra).

Data acquired at different potentials is analymed by using the

solution of Eq. (3) (shape fit) obtained at one potential V1 as a

refeence A rati R formed With the VI data and data obtained at

another potential V2 , is fit to Eq. (5) (ratio tits). As expected. Eq. (5)

IL(VO ep[(1)p((l)WV)] ( IV,Q fL(V2) I x (eo)WV)(2) ]V2

(5)

where (IV- Sr(V) + - 1n

(Sr(V)+1)(SpLui+1) (ae+up)Ln

U.A 9

reduces to Sqg. () iftSr is relatively large (OSn I and a6 Ln) or

lip Anu-t at the ayplied potential? our use of ratios (Q-OPD plots)

rather than iniida IL-OPD plots leads to smaller standard

deviaion n W and S.

Use of thesn equations requires absorptivities a., reflectivitles,

and an estimate of Ln. The absorptlvlties used for p-GaAs were

thoae employed in the study at n-GaAs electrodes.11 An

absoptlvtyof 9 z 10 cm-4 was used for the emitted light,

monioredat 865 rim.19 For data acquired in solution, literature

retectvltes 0 were corrected for the refractive index of the

Literature values for Ln in mnelt-grown or diffused p-GaAs:Zn

varY1-23, but reasonable estimates for use in Eqs. (3) and (5) are 5

pmn for p - 3.1 z 10 17 cm73 , 4 Mr for p a I x 1018 m3, and 3 Ian

for p - 1.8 x 1019 cm 3 . The qualitative aspects of this analysis are

unaffected by this paaee, although the data cannot be fit if

very small ('1 Wn) values Ot Ln are used. Larger values Of L

yield larger calculated W and Sr. values. Values of W are relatively

ftnsensitive to the value Of Ln chosen; however, calculated values at

rwe nearly propor-tional to the Ln value employed.

M. EKPElvMN

The p-GaAs samples used in this study were melt-grown

samples obtained from Lse Diode ILaboratories (Cd doped, (111), 3.1 x

1017 cm-3 ), Atomnergic Chels, Inc. (Zn doped, (111), 1.0 z 1018

cm-3 ), and IMorgan Semicnuco, Inc. (Zn doped, 1/20 off (100), 1.8

z 1019 cm73 ). The (111)-riented samples were receved with a

pad IBI-ace (As ftoe). and this surface was used in our

ezperiment.. The orientation of the (111) samples was determne

9by chemical etching and optical microscopy.2 4 Ohmic contacts to the

crystals (- O.S x O.S x 0.1 mm3) were made by soldering a HgtnZn

eutect to the back of the crysta. Crystals were then mounted as

described eewl'ere.2 5 Before use in a 1:C, electrodes were etched

in a room temperature 3:1:1 H2-0 4 :H 2 (30X):H20 solution, followed

by rinsing with M140H and triply-dlstille (3D) water.

The electrolyte was prepared frm H2804 (Malllnckrodt

analytical reagent grade or Alfa Products Ultrapure gave similar

results) and VSCN (Baker, reagent grade). An solutions were

deyenated with N2 for at least 30 mi. befe use, bubbled and

blanketed with N2 during use, and magnetically stirred.

Luvotag (L-V) and current-voltage (l-V) curves

were obtained by cycling between -0.05 V and -0.65 V vs. SC at 10

mV/s using a standard three-electrode n tic setup and

6 .. m -' equipmnt described previously.2 5 Front-surface PL

was monltored at the unrt band m8axium o 665 rum. The

experimental apparatus is as described previously, 11 ept for the

use of a -- m-- z -~o assembly. Ilumination was

accomplished with a Photn Technologies Inc. Model 01-ISO

high-intensity mumination system, which includes a Model O-ISOXI

150 W Xe lamp, an elliptical mirror, and a Model 01-001 0.25-m

moochromator. This excitation system permits computer control

o0 the wavelength scan rate and entrance sit, the latter providing

control of the system' output intensity. The output was kept

roughly photon-matched and the uncorrected measured bandpass of

the system was typicaly -10 nrn over the 450-690 nm spectral

range used. The intensity at the sample surface with 570 nm

excitation was -1.S mW/cm 2 . A Melles-Griot hot mirror (03 MHG

007) was placed at the exit slit of the excitation monochromator to

10reduce near-intrared Inefer ence (2, -700 nm) with the PL signal. A

polarizin filter was also used to prevent spectral variations in the

excitation luito.

The program, for the pulsed-potential experiments (toggling

between -0.05 V (0.5 9) and -1.0 V vs. BCE (0.5 s)) was provided by

an EG&G PARC Model 175 Universal Pogramer and Model 173/176

Moentiostat/Current, Follower. PL intensity for the pulse

eperiments was obtained with an EGRG ORTEC Model 9349 Log/Lin

Ratmetrused to monitor the amplifter/discrlrnlnator, and a

Uinsels LY 18100 X(T)-Y recorder operated in time base mode. The

system response time was estimated at '0.2 s.

A difficulty in Characterizing the p-GaAs PL is osdby its

temporal instability during photoelectrochernical operation. As

described in the text, addition of KSCN to the electrolyte improves

the lang-term stability of L-V behavior. Reproducible IL-OPD curves

were typically obtained by cycling the electrode between -0.05 V

and -0.65 V with excitation at 450 nm for at least 45 min. prior to

the acustion of data for the plots; by acquiring the data over 30

rim intervals rather than the 10 nm intervals used in air or at

open circuit; and by perfomin ratio fits (Q-OPI) Plots, Eq. (5))

relative to data obtained at -0.8 V on the cathodic-going scan, a

potential for which IL-0131)curves ezhibited the least change with

timne The generated IL-OIM and Q-OPI) curves were fit using a

nonlinear least-square curve-fitting program.

MI. RESULTS

A. PL properties in a PEC

Typical luInececeotage (L-V) and current-voltage (l-V)

curves for a p-GaAs electrode (p.- 3.1 z 1017 cm-3) in aqueous H2804

11

electrolyte are shown in Fig. 1. The PL intensity, shown over

one-anl-a-half cycles, exhibits a substantial enhancment between

the first and second cycles as well as considerable hysteresis

between anodic- and cathodic-going scans. The potential regime

most affected is from -- 0.4 to 0.0 V vs. SM. As shown in the

bottom panel of Fig. 1, this is just the regime corresponding to the

htcrt anomaly" (PA): negligible Faradaic current flows

between the estimated flatband potential, under illumination, of 0.1

V vs. SCE14 and -0.4 V. We will show that the PL effects derive

largely from changes in the reduced surface electron-capture

velocity S r over the PA potential regime, although some

contribution from a change in W cannot be excluded.

With continued cycling between -0.05 V and -0.85 V, the PL

settles into the patterns shown in Fig. 2. These plots were obtained

in the presence of KSCN, which markedly improved the long-term

stability of the data; the salt is reported to aid the rev rsbility of

Ga plating2 6 and to assist the anodic dissolution of Ga from GaAs

surface. 2 7 The figure emphasizes the greater relative hysteresis at

shorter excitation wavelengths, implying that surface chemistry

causes the hysteresis (vide infra).

Electrode ratflectance was montored over several cycles to

discern whether the PL changes Were attributable to this

pa rameter. No changes were observed within the error ( 2X

relative) of the expertment. The variations in PL thus result from

changes in W, 8r, or both.

An analysis using Mettler's methodology provides a means for

obtaining Sr and W as a function of applied potential. Typical

IL-OPD plots for a p-GaAs electrode at -0.1 V (anodic-going) and -0.8

V (cathodic-going) vs. SCE are shown in Fig. 3. Good fit* are

.I

12obtained to Eq. (3) for both A ratio fit ("OPD plot) using Eq. (S) for

the data at -0.1 V was somewhat les satisfactory, although it

resulted in similar values for W and Sr.

The discrepancy observed with the ratio fit is likely due to

errors of several percent in the relative spectral response of the

apparatus and/or in the absorptivities and reflectivittes of the

semiconductor. Occaulonyall small negative values for W are

obtained, but they are within experimental error of zero.

Values of W and 8r, extracted from the curves of Fig. 3 and

from Q-OPD curves at other potentials, are compiled in Table I and

plotted in Fig. 4. A comparison of the values for W and Sr for the

two scan directions indicates that the PL intensity in the PA regime

is affected by changes in both parameters: On the cathodic-going

scan, W increases modestly by -100 A, but Sr increases by a factor

of two between --0.1 and -0.4 V; similar values are obtained in the

PA regime on the anodic-ong scan, although corresponding values

for Sr are consistently lower. The dismrepancy in Sr for the two

scan directions can account for the hysteresis observed, although we

cannot exclude some contribution from W, due to the uncertainty

in our measurements. It is noteworthy that the hysteresis in PL

intensity and in 8 r occurs in the region of negligible photocurrent

where conventional surface recombination and not Faradaic charge

transfer is the dominant form of minority carrier relaxation at the

interface. The change in 8r implicates changes in the chemical

nature of the surface. The magnitude of 8r is also of interest. Our

maximum values of nearly 100 correspond to values for 8, using

Eq. (4) and literature electron lifetimes (-10 - 9 s)21,23 and diffusion

length* (1-5 ^m), that approximate the thermal limit of -107 cm/s.

In principle, values a W permit an assessment of the manner

13in which applied potential is partitioned across the

semiconductor-electrolyte interface. Although such an analysis

must be tempered by the uncertainty in W, values at W are

smaller than would be expected based on the flatband potential of

-0.1 V vs. SCe reported for illuminated p-Gas. 14 Assuming ideal

behavior, Eq. (2) predicts that W will change most rapidly with

potential at potentials near Vfb. Data in Table I exhibit the

opposite trend, indicating that partitioning of applied potential

occurs at the more positive potentials sampled. Beyond the PA

potential regime, as the electrode potential passes fro -0.4 to -0.8

V, W increases from 80 to 360 A; this cotresonds to a 0.3 V

increase in Schottky barrier height for a 0.4-V increment of applied

potential.

Data obtained an more conductive samples (p -1.0 x 1018 and t.8

x 1019 cm - 3 ) exhibit the same trends in W and Sr but generally

indicate larger barriers and barrier height changes with applied

potential. As a result of the linear relationship between carrier

concentration and barrier height (Eq. (2)), erroz in W increase with

p. Thus, although the calculated barriers may exceed those

theoretically possible for GaAs (i.e., VB E.), perfectly reasonable

barriers exist within the uncertainties of the measurement. In this

context it Is worth examining results found with the simple

dead-layer model (DLIV). Table I shows that values for AD from Eq.

(1) are considerably larger than the cding values for W

obtained using Mettler's treatment. In many cases the AD values

correspond to physically unreasonable changes in barrier height, a

consequence of interpreting the PL changes exclusively in terms of

changes in the width of the space-charge region.

L l~d IMMK

14B. Transient PL effects

As shown in Fg. 1, an initial cycling of the p-As electrode

produces a s tial enhancement in PL intensity in the PA

potential regime. A related observation was made by Uosaki, et al:

trarient PL enhan-cments were observed upon pulsing p-GaAs

electrodes from -1.0 V to rest potentials positive of or equal to -0.35

V vs. Ag/Aga in 0.5 M H2 804 electrolyte.6 We have examined the

transient PL enhancments and demonstrate below that they are

acoompanied by a significant reduction in Sr .

Brief operation of a freshly etched electrode cathodic of --0.4 V

vs. SCE yields a transient increase in PL intensity when the

electrode is subsequently taken out of circuit or brought to a

potential near the open-circuit voltage (OCV) of 0.1 to 0.2 V. Much

of the PL enhancement decays within seconds of the positive-going

pulse, although some enhancement persists for minutes. The

magnitude of the PL increase varies with the extent of etching and

with the duration and potential of the cathodic excursion; the PL

transient increased in intensity with more cathodic potentials up to

-1.0 V.

Repetitive pulsing between a cathodic potential (-1.00 V for 0.5

s) and a potential slightly negative of the OCV (-0.05 V for 0.5 s)

produces a maximum enhancement in PL intensity after several

minutes, as shown in Fig. 5. This maximum intensity was

sufficiently stable to permit acquisition of IL-OPD plots. Ratio fits

were then obtained relative to data obtained at the OCV prior to

commencing the pulse program. Representative data are shown in

Fig. 6.

Analysis of the Fig. 6 data reveals that 8 r declines by an order

of magnitude, from 90 at 0CV before the pulse program to -6 at

15-0.05 V after the pulse program; values for W were within

exprmental aerrort ofinro in both measurements . During the

exlpairment, the PL intensity at -1.00 V remains constant to within

l0X while the Pt. intensity at -0.05 V increases roughly six-told.

This indicates: that W and Sr at -1 .00 V (-N80 A and 90, res peactively,

for the experiment at Fig. 6) ame negligibly affected by the

P. rip rtment.

IV. DSUSO

Our Pt. data indicate that sigificant channges in Sr oocur in the

photcurentanomaly (PA) regime of p-GaAs-based PUW epoyn

aqueous H2804 electrolyte. These changes are evident in both

long-term cyclic voltamnmetric scans and in pulsed potentiail

exrperiments. The PL data prompt conm ertofa the surface

chemistry that could alter the number and/or location of surface

states mediating recomrbination. Two possiblities are chemistry

associated with hydrogen evolution arid i imdcw oax

prnpc e. The approxim ate net- potential of the Pt. transient,

behavior and ot the hysteresis in L-V plots (-0.4 V vs. 8CM

corresponds to the onset of cathodic current tor hydrogen

evolution2 6 , suggesting that surface coverage by hydrogen atoms

may influence P.

Alternatively, changes in Sr. could involve the preferential

reductive removal ot surface As. Woodall, et. al. have recently

used photochemnical methods to temapor arily remove surface states

frm n-.. and p-aAs29 The change was attributed to the

phoachrnialremoval at As anid/or A8203 and their associated

srflace states, with possible inhibition of the surface degradation by

a gallium oxide layer; the possibility at an oxide-free surface was

16not ruled out. Our results may reflect analogous variations in the

densty of As-related surface states. In this scenario, the cathodic

portio the cycle could be Interpreted as allowing preFerntial

removal at surface As, and the anodic portion as creating a

transient protective gallium oxide.

Turning to the values ot W obtained, we note that they are in

ar.rd with existing capacitance studies,14,30 which indicate that

SFemi-leve pinning (fixed VB and hence W) is not a

prevalent mchanism for p-GaAs In acidic aqueous electrolyte. A

200-mV positive band edge shift under illumination was inferred by

Kelly and Mummng from capacitance measurements as the

electrode potti nears V. Uosak et. al. explained their PL

tran nts, observed following a positive-oing potential pulse, in

terms o such a shift. The band edge shift was attributed to

surface chemistry in that study.6 Our results are not Isistent

with sorme prti n of applied potential. They also indicate that

changes in 8, derived fron surface chemistry, play a crucial role in

transient PL behavior.

A thermally limited value Of 8r in the PA regime had been

estmated previously by Kelly and Merrnlng. 14 Although we find

that Or varies -Agnificantly in the PA regime, our maximum values

do appoach the thermal limit ot -o 7 cm/s. Particularly intriguing

is the or=d-tof-magnitude reduction in Sr found by pulsing the

electrode across the PA potential regime. If the reduction i S is

due to the removal or shifting of surface states, then it is possible

that surface treatment during a pulsing program could preserve

the effect. Whether this could facilitate interfacial charge transfer

is hard to predict. Irnovements to date in p-type m-V

phoourrent properties have come from using redo. couples with

17

better kinetic than the hydrogan evolution reaction1 4 or by

enhancin the hydrogen evolution kinetics by meta 1 3 or metal ion

treetmuegt.15 ,16

V. CON1CLUSIONS

The PL from p-GaAs photocathodes employed in aqueous

sulfuric acid electrolyte has been used to moxnitor the depletion

width W and surface electron-capture velocity 8 while the electrode

$erve as the photocathode in an operating PEC. In the region of

the 011oto-current anomaly (between -0.1 and -0.4 V vs. SCZ),

hysteresis in the PL intensity correlates with hysteresis in values of

8. Value* of 8 are near the thermal limit. Repetitive pulsing of the

electrode potential between -0.05 and -1.00V vs. SC! yields an

order-of-magnitude reduction in 8. Variations in W between -0.1

and -0.6 V vs. SC! indicate that the electrode is not strongly

Fermit-leiel pinned, but uncertainties in W preclude a more detailed

analysis of the pattoigof applied potential across the

p-GaAs-electrolyte interface.

ACKNOWLEDGUMMIT

The authors thank the Office ot Naval Research anid the

University of WIsconsin-MadIson, University-Industry Research

Progrm for support of this research. We thank Laura Becke of

LAse Diodes, Inc. and Alan Hueliman of MIT for Hall

measuremnents.

18

1. I.E. Roldogagwarth and U. R. Sita, J. App. Phys.,SS, 5357 (1982) and retwea therein.

2. W. S. Hobson end A. B. Eis, J. AppI. Phys., 54,5956 (963).

3. R. Guruthra. M.Tamknklwlcz, and R.P.811brstain, J.Api. Phys., 54 , 6787 (1983)

4. W. 8. "Hbuam P. B. Johnson, A. B. EIs, and R. M.Diotld, Appl. Phys. Lett., 4S, ISO (194).

s. P. B. Jchns A. B. Ellis, R. M. ]iefeld, and D. 8.Ghnley, Appl. Phys. Lett., 47, 877 (198S).

6. 8. Kaneko, K. Uoaki, and H. Kita, Chem, Latt., 1951,(1986).

7. M. Evenar, 8. Gottedeld. Z. Harian. D. Hupper, and8.W. Fddber J. Phys. Chem., B6 , 6213 (1984)

8. K.H. Beckmann and R. emmnrrdg . lectrochem.Soc., 116 , 368 (969)

9. A. Etchaberry, J. Gautror, and J. L. Sculart, AppI.Phys. Lett., 46, 744 (1985).

10. K. Mettler, Appi. Phys., 12, 75 (1977)

11. A. A. Burk, Jr., P. B. Johnson, W. 8. Hobson, andA. B. Eflis, 4. AppI. Phys., 59, 1621 0986).

12. F.-R. F. Fan and A. J. Bard, J. Am. Chmn. Sac.,102, 3677 (1980).

13. Y. Nakato, 8. Tanornura, and H. Tubamura, Bar.Dunaungin-. Phys. Chemo., 60, 1289 (1976)

14. J. J. Kelly and R. IMernL 4. Klectrochem.Soc., 129, 730 (1982).

15. K. Uoeakl and H. Kita, Chem. Lett. 953 (1984).

16. M. P. Dhre-Edwards, A. Hamuett, and J. B.Goodenogh, J. Eletronma. Chem.. 119, 109 (1901).

1917. J. OM. BDaka and S. U. M. Xhan, Appl. Phys.

Lett.. 42, 124 (1983)

18. G. E. Stiman, C. M. Waife, and J. 0. Drmnock inmi"cductar and S,"intb, R. X. Willardson

and A. C. Beer, Ift., Academa: New York, 1977, Vol.12, p. 169.

19. M. D. Sturge, Phys. Rev., 127, 768 (1962).

20. B. 0. Scraphin and H. E. Bennett, in Ssmic-nductorsand Semd--nnta," R. K. Willardson and A. C. Beer,Eds., Acmdwnlc New York, 1967, Vol. 3, p.499.

21. J. Vflms and W. E. Spker, J. Appl. Phys., 36, 2815(1965).

22. L. JastruewkI, H.C. Gatos, and J. Lagowvik, J. Appl.Phys., 48, 1730 (1977).

23. L.W. Aukarman, M.F. MiUle, and M. McColl, J.Appl. Phys. 36, 685 (1967).

24. D. F. Kywr and M. F. Mille, J. lectraem. 8oc.,111, 1102 (1964).

25. B. R. Karat and A. B. EllIs, J. Am. Chem. Soc.,102, 96 (1980).

26. E. D. Moarhead, J. Am. Chum. Soc., 87, 2503(1965).

27. F. W. Ostermaer, Jr. and P. A. Kol, App!. Phys.Lett., 39, 76 0980.

28. H. Gerwiar, N. Muller, and 0. Haa,j. Eluectruaal. Chem., 119, 41 (1981), andreeens therein.

29. 8. D. Offwy, J. M. Woodall, A. C. Warren, P. D.KIrchner, T. 1. Chappell, and G. D. Pettit, Appl.Phys. Lett., 48, 475 (1986)

30. W. H. Laflore, R. L. Van Melrhaegh, and F. Cardon,Surface Oiemce, 59, 401 (1976).

20Thkil 1. Electrode Parameters of a p-CGaAs-based pECa

volts W AD volts W 40vs.sczrb (JJC .Jd ~. vs.SCzb (A)C (Ayd ~

(-)-0.1 -10 200 42 (+)-0.1 0 0 32

(-) -0.2 20 330 51 W)- 0.2 -20 70 37

(-) -0.3 30 400 59 (+) -0.3 30 260 46

-- 0.4 80 510 66 (W)-0.4 110 450 61

(-) -0.5 180 610 70 (W)-0.5 210 630 73

C-) -0.6 220 710 78 W)- 0.6 260 710 77

C-) -0.7 310 750 73 (W)-0.7 310 760 79

0--0.8 359 810 72 (4)- 0.8 340 820 84

0CVf 50 - 78

a Properties derived hran the FL af a p-GaAs electrode, (111)

As-rich face, with p - 3.1 x 1017 cm3. The FEC conuisted of a

thir-electrd potentiastatlc: setup and a 0.5 M 112804/0.1 M

KBCN electrolyte. Table entries are extracted fruit Q-OPD plot*

like that shown in Fig. 3. Magnetic stirring, a N2 blanket, and

slow N2 bubbling were used In all PEC exlpei ments. Dold-faced,

numbers are data acquired at the potential to which ratio fits

(Eq. (5)) are refte rn ed. Agr-eemnent between shape fits (Et. (3))

and the ratio results tabulated here (Eq. (5)) was gonerally

good. The Influence of uncertainties in Ln an the table entries

is discussed in the text.b potential at the p-GvaAs electrode; the symbol precedng the

voltage indicates a cathodic-going (-) or anodlc-coing (+) scan.

c Depletion width at the indicated potential, obtained by fitting

Q-=P curves to Eq. (5); at the referenc potential of (-) -0.8 V

21vs. wC, w is obtainbed by fiting the k,-OPD curve to Eq. (S).

The stantlard, deviation at the nonlinear least-squares fit for

tte reference value was ! 110 A. The relative error between

two values on the table ts typucally _*20460 A., with the larger

uncertainiIn the regbon of hysteresis. Negative values

reflect the uncertainty in the d-smnta of W.

d Change in dedlyrthickness relative to the value for()

-0.1 V vs. SM!. The values iste are those calculated for 570-nm

excitation.

*0 Reduced surface electron-capture velocity at the indicated

potential, obtained by fitting Q-OPD curves to Eq. ()k at the

reference potential of (-) -0.8 V vs. SC!. Sr was calculated using

Eq. (3). The standard deviation of the nonlinear least,-Wpure

fat for the reference value was ! St. The relative error between

two values in the table is typically !"-.

f The value obtained at o~pen circuit (0.17 V vs. OW! prio to

cycling the eletrode's voltage.

22

Figure 1. Relative photocurrnt (bottom panel) and PL intensity(top panel) as a function of electrode potential for ap-GaAs-based PEC employing 0.5 M H2804 electrolyte; PLintensity was monitored at Anm, 865 ram. The electrode (p3.1 z 1017 -cm3) was eicted i-ith -1.5 mW/cm 2 of 450-nm light.The first and second cat scans of PL intensity arelabeled A and B, respectively; the photocurrent Is only shownfor the initial scan. The curves were swept simultaneously at10 mV/s.

Figure 2. Photolutmnescence intensity as a function of electrodepotential for three emcitation wavelengths, 450, 570, and 690 nm.Data were obtained for a p-GaAs electrode (p,3.x1 7 cm"3 ) in0.5 M H12804/0.1 M KSCN electrolyte after cycling continuously for4 h between -0.05 and -0.85 V vs. SC! at 10 mV/s. The three-anels do not have the sa m vertical scale.

Figure 3. PhotolI -- / -- : Intensity vs. optical penetrationdepth (IL-OP)) curves for a p-GaAs electrode (p - 3.1 x 1017 cm 3)

in 0.5 M H2804/0.1 M KSCN electrolyte at two electrodepotent/als, -0.1 V (squares; acquired for an anodic-going can)and -0.8 V vs. SM (circles; acquired for a cathodic-going scan).Values of 8r and W eztracted at these potentials are given inTable I. The solid line passing through data at -0.8 V and thedashed line at -0.1 V are best fits to Eq. (3). The solid linepassing through data at -0.1 V Is the best fit to Eq. (5)

Figure 4. Calculated values of 8r and W as a function otelectrode potential. Values from the cathodic-going scans aredenoted by squares and those om anodic-going scans by

rcie. ma ete calculated for -0.6 V on the cathodic-goingscan serve as the reference values. The error bars for thereference values reflect the uncertainty in the absolute valuesand the other error bars reflect the uncertainty in the relativevalues.

23

Figure 5. Growth in P1. Intensity caused by electrode pulsing.The p-GaAs electrode (p a 3.1 x 1017 c7 3 ) was operated in 0.5

M "40&1 M KSCN electrolyte with 450-nm excitation. P1.

intensity for tinm preceding t -30 Is that of a freshly etchedelectrode at open circuit; atear t - 0, the data are the endPoints of the recorder Pen's deflection, obtained while pulsingthe electrode potential between -0.05 V vs. SCE (0.5 a), the top

Curve, and -1.00 V vs, SCE (0.5 s), the lower curve.

Figure 6. Photol m-nsec (P1. intensity vs. optical

pen rtiondepth (IL-OPI) curves) for a p-Ga.M electrode (p a 3.1x 1017 cmn3 ) in 0.5 M H2S04 electrolyte befor being placed incircuit (circles) and at -0.05 V vs. SCE after being repetitivelypulsed, as described in Fig. (5), until the P1. ehnmnt hadsaturated (squares). Values of Sr. and W extracted at thesepotentials are presented in the text. The solid lines are thebest fit to Eel. (3) for the open-circuit data and the best fit toEq. (5) for the -0.05 V data.

<~ B0

=L

0 4-

' 40s

.0.8 -0 04 0.20.

POTENTIAL, VOLTS vs. SCE

450 flim

570 nm

690 am

*A GA

POM-L VOT VLSC

* U W *~ U

q

EU

* '0* (~4w~*.b

* x

z0

* zw

'S..

In 000'id 9Aav'iH~J j

------ i-I-

So..

01

20

a

OA. .00. 2 0.0

POETIL VOT VS SCE

RELATIVE PL INTENSITY0 -4o Cl'

0* I

I

* I

)I

S.I aII

2*

S

**g~J

II

IaI

*aI

IA - - I - - a

RELATIVE PL MNENSITY0 0

o

0

zti

00

0-0

0%0

W*l IL

Maw*