Photoconduction and trapping in sputtered tantalum oxide films

J. Phys. Chem. Solids

Pergamon

Press 1967. Vol. 28, pp. 279-289.

PHOTOCONDUCTION

Printed in Great Britain.

AND TRAPPING

TANTALUM

IN SPUTTERED

OXIDE FILMS*

F. G. ULLMAN~ The National Cash Register Company,

Dayton, Ohio 45409

(Receiwed 9 May 1966; in rewisedform 20 June 1966) Abstract-Electron trapping in tantalum oxide fihns, prepared by reactive sputtering onto polished, fused-silica substrates, has been studied by mearauements of photocurrents and thermallystimulated-currents. The results indicate the presence of a single set of trapa situated about 0.25 eV from the nearest band edge. The initial trap concentration depends on the tantalum sputtering rate and the oxygen partial pressure during film preparation. Heating in a vacuum increases the trap concentration; heating in air or several days exposure to air at room temperature reduces it. Increaeing the trap concentration increases both the steady-state current during optical excitation, and the time to rise to and decay from the steady state on initiation and removal of excitation, respectively. A model ia proposed in which the empty traps are oxygen vacancies occupied by one electron. Once filled by optical excitation, the traps are proposed to behave like donors in quasi-thermal equilibrium with the conduction band. The rate of conductivity decay, on removal of excitation, is time determined by the relative rates of retrapping and recombination which, in turn, depend on the relative concentrations and cross-sections of traps and recombination centers.

1. INTRODUCTION

2.

THE SUGGESTIONS of other investigators(lW3) that electron trapping in insulator films played a dominant role in the transport of electrons through metal-insulator-metal sandwiches nrovided the impetus for the trap studies de&bed in this paper. We have used the method of thermallystimulated-conductivity,(4J’) coupled with observations of photoconductive behavior, to study trapping in reactively-sputtered tantalum oxide fihns. Since our results indicate that the steadystate photoconductivity of fihns with the largest trap concentrations is dominated by the thermal equilibrium between trapped and free majoritycarriers, the conditions for this type of behavior are derived from a simple energy-band model in the next section. A similar model has been analyzed by B&n and VOGSL@.~)to explain the S-shaped photocurrent growth in cadmium sulfidetype photoconductors.

KINETICS

OF

PHGTOCONDUCTIVITY TRAPPING

AND

The energy-band model shown in Fig. 1 illustrates the processes of optical excitation, trap filling, trap emptying, and recombination, in a

CONDUCTON

BAND

4 1

1

-

F

-

RECOIIBNATION CENTERS. N i

VALENCE

BAN0

. - ELECTRON

* This work was supported by the Air Force Avionics Laboratory, U.S. Air Force, under Contract No. AF 33(657)-10459. t Now at Department of Electrical Engineering, The University of Nebraka, Lincoln, Nebraska 68508.

o-HOLE

FIQ. 1. Energy band model for photoconductive with electron traps. 279

insulator

280

F. G. ULLMAN

photoconductive insulator (negligible free chargecarrier concentrations in the dark). The following restrictions and assumptions are imposed. 1. The only optical transitions are across the band gap and each absorbed photon produces one electron-hole pair. 2. The holes are trapped so rapidly at recombination centers that the free hole concentration can be ignored relative to the free electron concentration. 3. The traps are located at a single depth below the conduction band. 4. The recombination centers are located at a single level above the valence band. 5. Optical transitions from the valence band to the traps, and recombination of trapped electrons with free holes, can be ignored. The kinetics are described by the terms defined below. trap concentration (cme3) recombination center concentration (cme3) N N' hole concentration in recombination centers (cmW3) ni trapped electron concentration (cmn3) n free electron concentration (cms3) Pl N,Sz exp(-~~kT) = rate per electron of thermal excitation of trapped electron into conduction band (see-l) PS Sv = p,/Nc exp( - E/kT) = rate per electron per unit volume of trapping an electron from the conduction band (cm-” set-l) P3 Sp = rate per electron per unit volume of recombination of free electron with hole in recombination center (cm-” see-r) effective density of states in conduction band (cmm3) trap cross-section (ems) recombination center cross-section for conduction electron (cm”) trap depth (eV) number of photons absorbed per unit volume per unit time (cm-” set-I) thermal velocity of conduction electron V (cm/set) Boltzmann constant (eVpK) k absolute temperature (“K) T The kinetic equations are

Ni

dn,/dt = -plnf +hn(N,

-4

(1)

and dnfdt = F-p3nNf -PANI -nJ +A% (2) where psN + = l/7 and 7 is a recombination lifetime. A steady-state free electron concentration will be reached when the rate of generation of free electrons equals their total recombination rate. This condition can be expressed as F+p,n,

= p3N~n~p~(N~ -n&z.

(3)

Solving for n, we find n = (F+PPM~N+

fMNf-ndl.

(4)

The free electron concentration consists, therefore, of two parts: an optically generated concentration and a thermally generated concentration. Equations (1) and (2) for this case, d~~dt = d~~idt = 0, reduce to n = PP&(%

- nJ9

(5)

and TV= Fjp3N+.

(6)

Clearly, equations (4-6) can only be compatible if P&F

= ~~(N~-~*)IP~N+= c, where c is a constant;

i.e. equations (1) and (2) are not independent in this case. If the excitation temperature is high enough for (N<-nt) N Nf,, then for c = paNi/p3Ni Q 1, the steady-state conduction electron concentration is predominantly optically generated. For larger trap concentrations, c = p~N~~p3Ni > 1, the steady-state conduction electron concentration will consist predominantly of electrons thermally excited from the traps and in quasithermal equilibrium with the traps. The magnitude of the steady-state conductivity in this case will increase with increasing trap concentration; i.e. the conductivity behavior will be that of a partially compensated n-type semiconductor once the traps are populated by optical excitation. From the definitions of the constants given earlier, we may write’ c = psN1/psNf = SNJS,N+.

(7)

Clearly, the magnitude of the effect of trapping on the photoconductivity is determined by both the trap cross-section and the trap concentration. The

PHOTOCONDUCTION

AND TRAPPING IN SPUTTERED TANTALUM

condition c $ 1 can be satisfied if S $22, even if the trap concentration N,, is much smaller than the concentration of holes in recombination centers, N+. Alternately, trapping can be important for Ni B N”, even if S =$ S,. 3. ANALYSIS OF ~L~~A~-

4XJRRENTs 3.1 Estimation of the impmtame of retrappitg Most analyses of thermally-stimulated trap emptying assume negligible retrapping,‘4*6) i.e. the second term on the right-hand side of equation (1) can be neglected. In this work, the dependence of TSG peak temperature on excitation time (and hence, trap population) has been studied to indicate the importance of retrapping. For equal retrapping and recombination rates (in our notation, pa = pa), GARLICK(~~) has shown that the they-stim~ated luminescence peak will shift to higher temperature with decreasing initial fractional trap population, ntjNi. When retrapping is much greater than recombination, the effect of excitation time can be seen from Bube’s equation for trap depthc4) as a function of TSG peak temperature, which is E = kTrn ln(N~e~l~~),

(81

and T,,, = TSG peak temperature, 0, = conductivity at the maximum, e=: electron charge, P= electron mobility, and the other quantities are as defined previously. Since N, = CJZ”~‘~(for a spherical energy band) where “a” is a proportionality constant, equation (8) can be written as E = KT,,, ln(aT~3’2~~~).

(9)

If we make the reasonable assumption that the free electron concentration at the maximum, n,,,, increases with increasing initial trap population, then for fixed trap depth, trap concentration, and heating rate, equation (9) suggests that the TSG peak should shift to higher temperature for longer excitation time. In summary, the TSG peak will shift to lower temperature for longer excitation time if retrapping is smaller than, or of the same ‘magnitude as recombination, and to higher temperature if retrapping is much greater than recombination.

OXIDE

FILMS

281

3.2 ~~~i~ of trap deptk and ~~~~‘~ Under the assumptions of no retrapping and that the rate of change of free-carrier concentration is small compared to the rate of trap emptying, the initial trap population is directly proportional to the area under the TSG curve. The proportionality constant contains the product of freecarrier mobility and lifetime and experimental variables. Our Hall effect measurements have been inconclusiveCQ’although they indicated a mobility of the order of low3 cm2/Vsec or less in some of the films. This lack of mobility data prevented the determination of either the initial trap pop~ations or the free-carrier lifetimes. The trap depth was obtained by the method of GARLICK and GIEXWN.(~~)This technique has proved successful(5.13) and is influenced less by retrapping effects than other methods. 4. DEXRIPTION OF SAlWLES AND 7ALTECDNIQUES 4.1 Descr$tion of samples Four groups of films were studied. All were deposited onto polished, fused-silica substrates.* These films are amorphous but were identified by electron diffraction to be /3TaaOs after recrystallization at 400”G.(8) Pertinent preparation conditions and film characteristics are shown in Table 1.t Optical transmission spectra for the Group I films showed a slowly increasing absorption with decreasing wavelength, beginning at about 8000 A and rising more sharply near 3500 A. A similar spectrum for a film sputtered at 4000 V and a spectrum of one of the Group II films have been reported by BLASINGAME and YOUNG.(~)The optical transmission spectra of films from Groups III and IV are similar to those of Group II films. Typical examples are shown in Fig. 2. The film thicknesses shown in Table 1 were calculated from the interference peaks in the transmission spectra and published values for the dispersion of the refractive index of tantalum oxide films.‘14) * Other substrates, such as glass microscope slides, effects and exhibited interfering photoconductive deny-sedated-cu~en~. There were no measurable photocurrents or thermally-stimulated-currents in the fused-silica substrates. t Groups I and II were prepared in and by the Air Force Avionics Laboratory, Wright-Patterson Air Force Base, Ohio. Groups III and IV were prepared in these laboratories.

F.

282

G. ULLMAN

Tai% 1. Description of films

Group

Sputtering atmosphere

sputtering voltage

I

Argon-Nitrogen-Residual Oxygen Argon-Nitrogen-Residual Oxygen Argon-Air Argon-Oxygen

1.500 loo0 1200 875

I:: IV

Film

Film

thiCkneS9

COl0r

(II>

Brown Transparent Transparent Transparent

Preparation conditions

moo-2000 7650 715 4750

Reference 8 Reference 8 Reference 9 Reference 9

5 percent, even at the fastest heating rates of about l”C/sec. Currents were measured with a Keitbley 410 picoammeter and both current and temperature were plotted on an X-Y recorder. Since no phot~urren~ were ever observed with excitation wavelengths greater than 3OOOA, all excitations were performed with the predominantly 2537 A radiation from a low pressure, mercury vapor, germicidal lamp (General Electric G8T5). Assuming the light source to be monochromatic with a wavelength of 2537 A, the incident intensity, as measured by a thermopile, was l-6 x lOI photons/cm2 sec. The photocurrent in all cases was independent of the polarity of the applied voltage and varied linearly with applied voltage over the measurable range; for films with the largest conductivity, this range was O-l-218 V, the largest applied voltage used in these studies. Prior to excitation, the sample was always outgassed at least once to eliminate surface leakage, I by heating to about 200°C. A second heating was then performed to determine if there was any measurable conductivity in the unexcited sample since this would have to be subtracted to obtain the true TSC curve. Only Group I films had measurable conductivity in the unexcited sample; current-temperature curves for these films were obtained both before and after the TSC measurement to insure the reproducibility of the current that had to be subtracted.

4.2 Expehental details After sputtering, the samples were electroded by vacuum deposition of aluminum or gold on top of the film with an electrode width of 1-Ocm and an electrode gap of 0.1 cm. Measurements on films from Groups 1 and II were made in a mechanicallypumped vacuum at a pressure of about 0.04 torr. Messurements on f&ns from Groups III and IV were made in an ion-pumped vacuum at pressures of about 7 x lo-* torr. (This pressure rose during the TSC measurement, because of unavoidable outgassing, to near 1.0~ 10q4 torr at the highest measurement temperature, about 300°C.) ZOOV d.c. were applied in most of the measurements; the few exceptions are noted in the text and figures. Temperature measurements were made with a copper-constantan thermocouple attached to the sample support; the sample temperature was estimated to lag the thermocouple by less than 100,

80 -

60 -

40 -

20-

5. EXPERIMENTAL

FIG. 2. Optical transmission spectra of tantahun oxide films. Dashed line-Group III fdm. Solid lintiroup IV film.

RESULTS

5.1 Measurements on Group If;lms The room-temperature currents in Group I films before excitation were typically of the order of 1 x lo-lo A at 45 V; consequently, photocurrents less than 1 x lo-l1 A could not be detected

PHOTOCONDUCTION

AND

TRAPPING

IN

by our straightforward d.c. measurement technique. There was little or no measurable photocurrent immediately following initiation of U.V. excitation. However, a photocurrent could be detected after several minutes and this current continued to grow, increasing by a factor of about 250 in 48 hr. The decay was correspondingly slow when the excitation was removed. TSC measurements were then made either by quenching to liquid nitrogen temperature before heating or by heating from room -temperature. The roomtemperature current on heating from liquid nitrogen temperature was the same as the roomtemperature current before cooling and there were

1 10

SPUTTEREL,

TANTALUM

OXIDE

FILMS

283

significantly greater than after 17 hr of excitation. Therefore, the increased peak height and shift of the peak to higher temperature of the 17 hr curve relative to the 40 hr curve is a result of the larger heating rate, as expected for TSC measurements. On the other hand, the 2-S hr curve peaks at a lower temperature even though it was measured with the fastest heating rate of the three cases. This is a result of a smaller trap population for the shorter excitation time and illustrates quite vividly that calculations of trap depth based on TSC peak temperature alone cannot be reliable for these measurements. 5.2 iI!kWrMts o?t Groujl IIpns There was no measurable conductivity in Group 11 films without excitation. Photons with rise and decay times less than the time constant of about 1 set of the picoammeter were observed. Typical TSC measurements for applied voltages and excitation times as indicated, and a heating rate of 0*6°Kjsec, are shown in Fig. 4. 80

B x

P Ltl ?i

3ca

400 TEMPERATURE

100

(%I

FIU. 3. Thennally-&rnulated-current curves for a Group I f&n-applied voltage-45 volts. Excitation time and heating rate: (a) 24 hr, 0~75”K/eec, (b) 40 hr, 0*16*Kjeec. (c) 17 hr, 0*33*K/sec.

no sign&ant diierences between the TSC curves obtained by the two methods, for the same excitation times. Typical TSC curves for one film with 45 V applied are shown in Pig. 3. From other measurements, it was shown that the trap ponulation after 40hr of excitation was not

FIG. 4. ~~y-a~~~-~nt cumes for a Group II &n--he&g rate--W~K/sec. (a) 2 hr excitstkm, TSC at 217 V (b) 17 h.r excitation, TSC at 135 V.

F.

284

G.

ULLMAN

(The rising portion at high temperature resulted from leakage currents through a surface contaminant which was eliminated in subsequent measurements.) Comparison with Fig. 3 suggests that the trap concentrations in these films were roughly four orders of magnitude smsller than in the Group I films if the free-carrier lifetimes in the two groups were comparable. This point will be discussed later. 5.3 Measurements on Group III films No photocurrents were ever detected in Group III films, probably because they were so thin, and a miniium of 17 hr of excitation was required to obtain a detectable TSC. Otherwise, the TSC curves for Group III films were similar to those obtained for Group II films, peaking near 400°K and exhibiting a slight peak shift to higher temperature with longer excitation time. 5.4 Measurements on Group IVjilms These films were initially like Group III films. However, heating in vacuum produced a steadily increasing TSC after each heat cycle with concomitant changes in photoconductive behavior

from initially fast and insensitive to slow and sensitive. Measurements on a typical sample are summarized in Table 2, below, and the TSC curves obtained after two, four, six, and eight heating cycles, and 17.75 hr of excitation for each, are shown in Fig. 5. Group IV samples, once sensitized, revert back to their initial state if heated in air or exposed to air at room temperature for several days; short exposures, of the order of hours, have no measurable effect. They can then be recycled back to the sensitive state by further vacuum heating. The effect of vacuum heating was also checked on one Group III film. An increase in peak height of about one order of magnitude was obtained after seven heating cycles but there was still no measurable photocurrent. 5.5 Dependence of trap-Jlling on excitation temperature The slow decay of the persistent conductivity observed in the films with larger trap concentrations is evidence of a slow but significant emptying of traps at room temperature. In fact, order of magnitude calculations indicate that the maximum

Table 2. Summary of measurements on a Group IV jilm Total heat cycles

Initial* photocurrent

Final photocurrent

(A)

(A)

17.75

3 x10-i’

9x10-=

<

1 set

405

Twice to 200°C

2

Figure 5 Curve a

17.75

2x10-”

8x10-‘1

50% in several set

417

once to 200°C and then once to 300°C

4

Figure 5 Curve b

17.75

1 x10-“1

2x10-10

25% in a few min

425

Once to 200°C and then once to 3OO”C

6

FigWe Curve c

47.25

2x10-1’

5x10-‘0

< lO%in afewmin

429

once to 200°C

7

Figure 8 Curve a

17.75

1 x10-=

5.1 x 10 -I0

Very slow

445

Once to 200°C

8

Figure 5 Curve d

Decay behavior

TSC peak temperature (OK)

Vacuum heating prior to TSC measurement

Excitation time (hr)

TSC curve

* There is a smah photoemission leakage which is strongly dependent on temperature and pressure. Since the initial phouxurren ts were sometimes meaaumd before the aample had completely cooled to room temperature and before the ayatem pressure had completely returned to its equilibrium value, the variations in initiaI photocurrent and the apparent decay during the first excitation are probably a result of variations in this leakage current.

PHOTOCONDUCTION

AND

TRAPPING

IN

achievable fractional trap concentration, n,/Nr, at room temperature, is less than O-01. One would expect that excitation at lower temperatures would permit the traps to be completely filled and in a much shorter time than the long time required to

10-12

SPUTTERED

TANTALUM

OXIDE

FILMS

285

inability to fill certain traps at low temperature has been observed in CdS.‘11*16) 5.6 Determination of trap depth Plots of the logarithm of the current in the initial portion of the TSC curve against reciprocal temperature for Groups I and II are shown in Fig. 6 and for Groups III and IV in Fig. 7. The trap depth determined from the curves in Fig. 6 is 0*29+0*02 eV, and from the curves in Fig. 7, 0*22+ O-01 eV. The curves in Fig. 6 were taken with a different vacuum chamber and sample mount from the one used to obtain the curves in Fig. 7. The lag of thermocouple temperature behind sample temperature at the lower temperatures was greater in the latter case which could easily account for this apparent difference in trap

cTEMPERATURE

(*Kl

FIG. 5. Effect of vacuum heating on thermally-stimulated-currents in a Group IV film-Heating rateO*S”K/sec. , (a) after two heating cycles (b) after four heating cycles (c) after six heating cycles (d) after eight heating cycles.

reach a steady-state at room temperature. However, this is not the case. If the film is excited at room temperature and then cooled before the TSC measurement, the same TSC curve is obtained as on heating from room temperature. If the excitation is performed at about -lOO”C!, a TSC peak is observed but its magnitude is only about five percent of those obtained for room-temperature excitation for the same length of time. A similar

FIG. 6. Logarithm of initial portion of thermallystimulated-current curves for Group I and Group II films vs. reciprocal temperature. Excitation time : (a) 2.5 hr (b) 17hr (c) 20 hr (d) .,2hr; +,16hrand64hr :;j:: (g) 17hr.

F.

286

G.

ULLMAN

depth. For purposes of discussion, we will assume a value of 0.25 eV for the trap depth. Of importance are (l), the good agreement in trap depth between the various samples which had widely different phot~o~du~ve .bebavior and trap concentrations, (21, the apparent constancy of trap depth calculated from measurements on the same sample for the different excitation times indicated in Fig. 6, and {3), the absence of any dependence of trap depth on the temperature range of the initial portion of the TSC curve (i.e. cooled before heating or heated from room temperature). 6. DISCUSSION OF RESULT8

6.1 Choice of the model Much of the ensuing discussion is based on the caption that the observed increases in

‘O-@ 3

l----L?

~~~y-st~~t~-~ndu~vi~ result from increases in trap concentration rather than increases in majority carrier lifetime. Although all of the experimental results that have been described are consistent with the mode1 presented in Section 2, which is similar to the model used by B&R and VOGEL@)to explain the slow S-shaped photoconductivity growth in cadmium sulfide-type photoconductors, the same type of photo~rrent growth is predicted by the now generahy accepted Ros+Bube photoconductor madelw in which minority-carrier trapping by sensitizing centers controls the majority-carrier lifetime and hence the magnitude and speed of response of the photocurrent. The photocurrent growth we have observed is also S-shaped, as shown in Fig. 9. (This growth curve does not show clearly the approach to the steady-state; however, other measurements on the same sample have shown that a steady-state is approached in four to five hundred

d

t

I

\

I

\

I-

RECPROCAL

TEMPERATURE, 1000/T

(*K)-’

FIO. 7. Logarithm of initial portion of thermallystimulated-current curves for Group III and Group IV @ms vs. reciprocal temperature (a) Group III f&n afk 7 heating cycles (b) Group IV film curve b, Fig. 4 (c) Group IV fihn curve c, Fig. 5 (d) Group IV film curve di Fig. 5.

FIG. 8. Dependence of ~e~y-a~~t~-cu~ent maximum on excitation time for Group IV film (a) 47.25 hr excitation after 7 heating cycles (b) 17.75 hr excitation after 8 heating cycles (curve d, Fig. 5).

PHOTOCONDUCTION

AND TRAPPING IN SPUTTERED

houk with a steady-state current about twice the maximum value shown in Fig. 9.) Since this behavior is qualitatively similar to that predicted by the Rose-Bube model, the question of whether or not our other observations can also be explained by this model, demands an answer.

TANTALUM

OXIDE FILMS

287

temperatures of about - 100°C reduces the conductivity by several orders of magnitude; on reheating, with or without excitation, the conductivity returns to the same value at room temperature as before cooling. Smce this conductivity is not optically generated, it must derive from the

l - F44OTOCURREW RISE

OECAY

0-m

IO

20

30

Y)

40

so

m

so

TIME ll4OuRS~

FIG. 9. Rise and decay of photocurrent in group IV film after several vacuum heating cycles. Our rejection of the sensitizing center model is based primarily on two sets of observations. First, with excitation time held constant, the shifts in TSC peak temperature accompanying the increase in TSC with each vacuum heating, indicate an increasing trap concentration and a corresponding increase in importance of retrapping relative to recombiiation. A lifetime increase with no change in trap concentration would not produce such a TSC peak shift. Further, the Group I 6lms in which the TSC were some two orders of magnitude larger than in any other films, showed the opposite dependence of TSC peak on excitation time, as expected for a trap-dominated conductivity. These excitation time effects are described later. Second, the effects of cooling on the conductivity of “sensitized’ samples strongly suggest a free electron concentration controlled by thermal equilibrium between conduction electrons and a set of donor levels (assuming n-type photoconductivity). The conductivity in these samples decays so slowly that the behavior on cooling is the same whether the light is on or off. Cooling to

trap levels which have been filled optically and are emptied only by heating or long periods (weeks) of decay at room temperature. Based on these observations, we conclude that the observed increases in TSC and accompanying changes in photoconductive behavior result primarily from increases in trap concentration. 6.2 Chemical identity of the traps The dependence of trap concentration on preparation conditions and subsequent vacuum or air heating demonstrates quite clearly that the traps detected in this investigation are associated with an excess of tantahun. Similar observations have been made by others. HARTMANN has observed that the conductivity of tantahun oxide was decreased by heating in hydrogen and the activation energy for conductivity also decreased to a value of about O-3 eV. KoF~TAD(~‘)hasstudied the conductivity of tantahun oxide in the temperature range 877-1380°C as a function of oxygen partial pressure. He fmds that tantal~ oxide is p-type at oxygen partial pressures greater than about lOWa

288

F. G. ULLMAN

atm and n-type below this pressure. He concludes from these studies that the conductivity of tantalum oxide is controlled by oxygen vacancies and oxygen interstitials, the former giving rise to the n-type conductivity at lower oxygen pressures and the latter to the ptype conductivity at higher oxygen pressures. Following KOFSTAD,(~‘) we believe a plausible argument can be given for the identification of the 0.25 eV trap level as an oxygen vacancy containing one electron. CIJRIE(~*) has derived an expression for the trap depth associated with a divalent anion vacancy containing one electron, in an ionic lattice. (The trap depth is the binding energy for the second electron which such a vacancy can accommodate.) Curie’s formula is

6.4 Depemhce of the TSC peak tem.ature on excitation time For very strong retrapping, as in Group I films, the TSC peak is expected to shift to higher temperature with increasing trap population. This effect can be seen in Fig. 3 where the TSC following 2.5 hr of excitation, peaks at lower temperature than the TSC curves plotted after 17 and 40 hr of excitation, even though the 2.5 hr curve was plotted with the fastest heating rate. Based on the integrated areas under these curves, the trap population after 2.5 hr of excitation was roughly two orders of magnitude smsller than after 17 and 40 hr of excitation, for which the trap populations were the same within about ten percent. On the other hand, when retrapping and reE = (RI/16&,) +(1/1=X (10) combination are of comparable importance as in Group IV films after vacuum heating, an inwhere creasing TSC peak temperature with decreasing R = 13.5 eV, fractional trap population should be expected. K = static dielectric constant, This effect can be seen in Fig. 5. The excitation K, = optical dielectric constant, time was held constant for all four curves, so the and concentration of filled traps, n,, is the same in each K, = effective dielectric constant given by case, at least to a first approximation. Since the vacuum heating continually increased the trap l/Kc = (l/K) +(5/16)[(l/Ke) -W1y)l. (11) concentration, N,, the fractional trap population, For K = 27(lg) and K, = 5 (refractive index n,/NI, continually decreased. The concomitant squared’l*)), we find E = 0.23 eV, ingood agreeincrease in TSC peak temperature is apparent in ment with the trap depth of about 0.25 eV Fig. 5. determined in the present investigation. However, one cannot conclude a priori, as we From the calculation above and the work of have above, that with varying trap concentration KOFSTAJP’) and in the absence of experimental a fixed excitation time still gives a fixed filled-trap evidence to the contrary, we conclude, tentatively, concentration. To check this point, a 47.25 hr that the traps we have observed are oxygen excitation was performed following the measurement of curve c in Fig. 5. The subsequent TSC vacancies containing one electron. curve is plotted in Fig. 8 along with the next measurement, curve din Fig. 5. The shift to higher 6.3 Energy distribution of the traps In the foregoing, we have treated the trap dis- temperature of the latter in spite of the shorter tribution as being monoenergetic. The very slow excitation time of 17.75 hr indicates that the shifts photocurrent rise and decay times prohibit any in TSC peak temperature, in this case, are mainly as predicted by GARLICK.(~~) extensive study of the trap distribution by photocurrent decay analysis(fl) in a reasonable time. However, the single TSC peak obtained for all 6.5 Trap-jilling at low temperature BUBE et aZ.(ll) and NICHOLAS and WOODS,(~*~~) films, the absence of any dependence of trap depth have reported a trap in CdS which cannot be filled on excitation time, and the fact that the same trap depth is obtained from TSC curves started at at low temperature and which empties by monoroom temperature as from those started at low molecular kinetics in spite of a relatively large cross-section. BUBEet al., suggest that this trap is a temperature following room temperature excitacenter surrounded by a Coulomb repulsive-bartion, all indicate that the traps are at a single depth rier; in this model, the capture cross-section rather than distributed in energy.

PHOTOCONDUCTION

AND TRAPPING

IN SPUTTERED

decreases exponentially with temperature, thus accounting for the inefficient trap-filling at low temperature. The traps in our films also cannot be filled as efficiently at low temperature as at room temperature. Although this can be explained by the repulsive-center model, the possibility that the efficiency of trap-filling may be controlled by the rate of thermal filling of the recombination centers from the valence band, should not be ignored.

TANTALUM

OXIDE

FILMS

289

Ack~~~edg~ts-me author is indebted to Messrs. J. M. BLASINGAME and C. R. YOUNGof the Air Force

Avionics Laboratory for the preparationof Group I and Group II films, and to Messrs.J. R. JANN~NG and

L. E. BLAMPLY,JR. for preparation of Group III and Group IV fihns, respectively. The author also acknowIedges the assistance in these studies at various times, of Messrs. W. E. SANDERS,C. A. SIMMONS, and S. A. DRESI(IN.

6. SUMMARY OF RESULTS AND CONCLUSIONS

Reactively-spu~ered tantalum oxide films have been shown to contain a single set of traps with a depth of about 0.25 eV. Evidence has been presented favoring the identification of these traps as oxygen vacancies occupied by a single electron. A majority-carrier trapping model similar to one originally proposed by BOER and VOCEL@) is shown to be consistent with all of the experimental results. This model predicts fast, insensitive photoconductivity for low trap concentrations and slow, sensitive photoconductivity for large trap concentrations. The full range of this variation has been observed by studying films with trap concentrations ranging over about six orders of magnitude. High trap concentrations were achieved by film preparation at high tantalum sputtering rates and low oxygen partial pressures, and by vacuum heating of films with initially low trap concentrations. A unique property of these traps, that they cannot be filled at low temperature, has also been reported for a trap level in CdS;01*15) possible explanations for this effect are a Coulombrepulsive barrier surrounding the trapoX) or thermal filling of recombination centers from the valence band. Although the traps are clearly associated with a deficiency of oxygen, the extent to which photoadsorption and desorption effects are involved in the photoconduction and trapfilling processes is unknown. Such adsorption phenomena have been suggested to be responsible for some of the conductive properties of zinc oxide (e.g. Ref. 20) as well as cadmium sulfidetype photoconductorstz1*22) and consequently, might contribute significantly to the properties of tantalum oxide, as well.

33, 2993 (1962). 1. GBPPERTD, V.,~~ 2. MeAn C. A., Pkys. Rm. 128, 2088 (1962). C. E., Proc. Nat. Aerospace Conf, 3. DRUMHBL~ZR (1963). 4. BUBE R. I-I., Pkotoconductiwity of Solids, Chap. 9. John Wiley, New York (1960). 5. Nrcno~hs K. H. and WOODSJ., Br. J. Appl. Phys. 15, 783 (1964). K. W. and VOGBL H., Ann. Phytik. 17,10 (1955). 7. BUB~ R. H., ikid, pp. 28.5-287. J. M. and YOUNG C. R., Technical 8. BLASINGAMIZ 6. B&R

Report AFAL-TR-65-191, Air Force Avionics Laboratory, Wright-Patterson Air Force Base, Ohio (1965). 9. ULLMAN F. G. and DRESS

10.

S. A., FinaI Report AFAL-TR-66-77, Air Force Avionics Laboratory (1966). GARLICK G. F. J., Luminescent Materials, Chap. II, PP. 37-8. ClarendonPress. Oxford (1949). BUBE R. H., DUSSEL G, A., Ho C. and MILLER L. D., J. Appl. Phys. 37, 21 (1966). GARLICKG. F. J. and GIBSON A. F.. Proc. Phvs. Sac. 60, 574 (i948). BIJBE R. H., J. Appl:Phys. 35,586 (1964). YOUNG I.... Anodic Oxide Films. Ghan. 7. P. 81. ._I

11. 12. 13. 14.

Academic Press (1961). _ _ 15. WOODSJ. and NICHOLASK. II., Br. J. Appl. Phys. 15, 1361 (1964). 16. HARTM~ W., 2. Phys. 102,709 (1936). 17. KOFSTA~P., J. Electrochem. Sot. 109, 776 (1962). 18. CURIE D., Luminescence in Crvstals, Chap. VI, D. 166. M&huen and John Wiley (1963).* ._ Ibid, Ref. 14, Chap. 15, p. 186. E: MILLER P. H., Proceedings of the Conference on Photoconductivity, Atlantic City, Nov. 4-6, 1954 (Editor R. G. BRECKFZNRIDGE et al.), p. 287. John Wiley, New York. 21. MARK P., J. Pkys. Ckem. Solids 25, 911 (1964);

ibid 26, 959 (1965). 22. SHEAR H., HILTON E. A. and BUBE R. H., J.

Electrochem. Sot. 112, 997 (1965).