Photovoltaic and electron-voltaic properties of diffused and Schottky barrier GaAs diodes

Solid-Stare

Electronics

Pergamon

Press

1971. Vol.

14, pp. 529-540.

Printed

in Great

Britain

PHOTOVOLTAIC AND ELECTRON-VOLTAIC PROPERTIES OF DIFFUSED AND SCHOTTKY BARRIER GaAs DIODES* R. KALIBJIAN

and K. MAYEDA

Lawrence Radiation Laboratory, University of California, Livermore, California 94550, U.S.A. (Received

8 September

1970; in revisedform

14 December

1970)

Abstract-The photovoltaic and electron-voltaic properties of GaAs diodes have been investigated for an irradiance in the range of IO-” to 104~W/cm2. These diodes were Zn diffused mesa-type diodes and Schottky barrier diodes having barriers of Au, Cu, and Cu/SiO. At IO0 pW/cm” diffused diodes can have a photovoltaic power conversion efficiency as high as 18 per cent for an absorption edge excitation at 8500 A and an open circuit voltage of 0.7 V. To achieve high efficiencies at short wavelengths, controlled etching of the front window can yield 1I per cent at 7000 A. For shorter wavelengths yet, Schottky barrier diodes are used to give efficiencies of 7 per cent at their peak spectral response of 6500 A and an open circuit voltage of 0.4 V. The quantum efficiency for these diodes is about 50 per cent for the photovoltaic mode of operation. The power conversion efficiency in the electron-voltaic mode is about 20 per cent of that in the photovoltaic mode, because greater energy is required to produce an electron-hole pair with an electron source than with an infrared radiation source. The energy required to produce an electron-hole pair in GaAs is 6.8 eV as deduced from the electron-voltaic quantum efficiency vs. electron energy curve.

On a Ctudii les propri&Cs photovoltai’ques et Clectronvoltai’ques de diodes GaAs pour une radiation dans la gamme 1O-3 g IO”pW cm2. Ces diodes ttaient du type m&a-diffustes au Zn et du type barritre Schottky ayant des barribres en Au, Cu et Cu/SiO. A 100 pW/cm2 les diodes diffustes peuvent avoir une efficacitt? de conversion deOpuissance photovoltai’que atteignant 18 pour cent pour une excitation de rebord d’absorption 5 8500 A et une tension de circuit ouvert de 0,7 V. Pour obtenir de grandes efficacitts B ondes courtes, une gravure contr&Ce de la fen&tre avant peut donner I I pour cent 2 7000 A. Pour les ondes encore plus courtes les diodes de barritre Schottky sont utilisCes pour donner des efficacitks de 7 pour cent & leur rkponse spectrale de cr&te de 6500 A et une tension de circuit ouvert de 0,4 V. L’efficacitC quantique pour ces diodes est environ de 50 pour cent pour le mode de fonctionnement photovoltaique. L’efficacitt de conversion de puissance dans le mode tlectronvoltaique, est d’environ 20 pour cent de celle du mode photovoltdique, car plus d’Cnerg,ie est nCcessaire pour qu’un ilectron-trou s’accouple avec une source d’Clectrons qu’avec une source de radiation infra-rouge. L’tnergie nCcessaire pour qu’un Clectron-trou s’accouple dans le GaAs est de 63 eV tel qu-il est dtduit de la courbe d’efficacitk quantique Clectron-voltayque en fonction de I’tnergie de I’tlectron. Rbumk-

- Die Photospannung bei Licht- und Elektroneneinfall wurde fiir GaAs-Dioden im Bestrahlungsbereich von IO-” bis lo-” pW/cm’ untersucht. Die Dioden waren zinkdifindierte Mesadioden und Schottky-Dioden mit Au, Cu und Cu/SiO. Diffundierte Dioden kiinnen bei 100 pW/cmY einen Leistungswirkungsgrad haben bis zu 18% bei einer Anregung nahe der Absorptionskante bei 8500 A und einer keerlaufspannung von 0.7 V. Hohe Wirkungsgrade von 11% sind bei kurzen Wellenlangen von 7000 A zu erzielen, wenn man in die Diodenvorderseite kontrolliert ein Fenster Itzt. Bei noch kiirzeren Wellenlsnge? wurden Schottky-Dioden benutzt, die beim Maximum ihrer spektralen Empfindlichkeit von 6500 A einen Wirkungsgrad von 7% bei einer Leerlaufspannung von 0.4 V liefern. Der Quantenwirkungsgrad betrKgt beim Betrieb als Photoelement 50%. Der Wirkungsgrad fiir die Leistungsumsetzung betragt bei BeschuR mit Elektronen nur etwa 20% des Wertes bei Lichteinstrahlung. da bei Elektronenanregung eine grb;Bere Energie zur Erzeugung eines ElektronZusammenfassung

*Work performed under the auspices Atomic Energy Commission.

of the U.S. 529

R. KALIBJIAN and K. MAYEDA

530

Loch-Paares erforderlich ist als bei Anregung mit infrarotem Licht. Aus dem Wirkungsgrad bei Elektronenanregung als Funktion der Elektronenenergie ergeben sich 6,8 eV fur die Energie zur Erzeugung eines Elektron-Loch-Paares.

A

4

P PO Pin P .cmax

AP R T

VI7 VB VI,

NOTATION constant related to generation and recombination of carriers diode area (cm*) 250 MlpZ’@ figure of merit for diode sensitivity fill factor electron energy (keV) electronic charge (C) photon energy (V) beam current (A) short-circuit current (A) numeric Boltzmann’s constant molecular weight subscript denoting n-side of diode junction I .2/ ( l-O.29 log,, Z) majority carrier concentration density (carriers/ cm3) intrinsic carrier concentration density (carriers/ cm3) subscript denotingp-side of diode junction quiescent carrier hole concentration density (holes/cm3) input power density (pW/cm*) maximum power (pW) excess hole concentration density (holes/cm3) range of electron penetration (A) temperature (“K) extrapolated dead voltage (V) beam potential (V) load voltage at the maximum power point of the I-V characteristic curve (V) open-circuit voltage of a photovoltaic diode (V) average atomic number optical absorption coefficient (cm-‘) electron absorption coefficient (cm-*) subscript denoting electron-voltaic operation photovoltaic conversion efficiency (%) photovoltaic quantum efficiency (%) e/AkT(

V-‘)

bulk density (g/c@ hole lifetime (set) energy required to produce an electron-hole (eV)

pair

1. INTRODUCTION

attention has been directed to photovoltaic diodes operated in a solar environment of 130 mW/cm2. However, some applications require generation of electrical power from low power sources. Two examples are (1) betavoltaic batteries using direct beta conversion or double conversion and (2) low level fluorescence detectors. In this paper we investigate the photovoltaic and CONSIDERABLE

electron-voltaic properties of diffused and Schottky barrier GaAs diodes for irradiance less than 10 mW/cm2 ( lo4 pW/cm2). This is an extension of our previous investigation of the photovoltaic properties of diffused GaAs diodes[ 11. 2. THEORY

2.1. Photovoltaic case According to Johnson[2], the open circuit voltage V,, of a photovoltaic diode can be written vo,

_

AH

,n

e

(I+I+ (AP/P~).~ (AP/P,)JJ 1

(1)

where ApIp is the ratio of excess hole concentration density to quiescent carrier hole concentration density, subscripts N and P denote the side of the junction being considered, e is the electronic charge, k is Boltzmann’s constant, T is temperature, and A is a constant related to generation and recombination of carriers. For moderate and low input power densities, (1) becomes V,, = Fin

[(z),v+l]

(2)

assuming that (AP/P,)~ 6 1. Consider a p+/n diode with thickness of the p+ region very small compared to the absorption depth of the incident exciting light. Radiant power will therefore be effectively absorbed in the n-region for production of electron-hole pairs.* Taking A = e/AkT, V,, becomes (3) Voc= + In ( pineAB~~ + 1) In (3), e h8s= 6.25 X lOI qqa7hnJE,hniZ where r,, is hole lifetime, (Yis the optical absorption constant, pin is input radiant power density, EP,, is photon energy, Q is the quantum efficiency (number of *For a IO per cent absorption in a Zn-doped p+ layer, the required thickness of the p+ layer is (I) about I pm at the absorption edge of 8500 A for (Y = IO3 cm-‘, and (2) about 0.1 pm at 7000 A for 01 = IO4cm-‘.

PROPERTIES OF DIFFUSED

AND SCHOTTKY

electron-hole pairs contributing to current/number of absorbed photons), n, is the majority carrier concentration density, and ni is the intrinsic carrier concentration density. The given form for (3) simplifies whenp,, ehBo+ 1 to

Voc=t In ,t&+Bo B, is therefore

numerically equal to V,, when pin is 1 pW/cm2, and can be considered a figure of merit for diode sensitivity. Diode conversion efficiency Q is defined as the ratio of maximum electrical power output PL,,, to api, where a is the area of the diode. PL,,, is the maximum power point obtained experimentally from the I-V characteristic curve and is related to short circuit current I,,, V,,, and fill factor c by the expression P,,,, = cI,,I/,,. The c is a figure of merit for the squareness of the Z-V characteristic curve, and I,, is related to p,,, by I,, = apinqq/ED,,. Therefore, 7, becomes

BARRIER GaAs DIODES

531

V. is an extrapolated dead voltage from the 779B vs. V, curve, and C#Jis the energy in eV required to

produce an electron-hole pair in the semiconductor. In the linear range of the qq8 vs. V, curve, 4 is independent of the electron beam energy and should correspond to the generation of carriers in the space-charge region of the diode (i.e. the electron beam has sufficient energy to penetrate into the space-charge region). The remaining symbols are similar to those of the photovoltaic operation. The subscript p pertains to electron-voltaic operation, and the terms CY~ (the electron absorption coefficient), A,, and cp are different from those in the photovoltaic case because of differences in the generation rates of the minority carriers. Of course, Thr ni, kT, and pin remain the same. The ratio of qCa to 77, derived from (5) (7), and (8) is (9)

2.2. Electron-voltaic case Irradiation of the diode with monoenergetic electrons is considered. The open-circuit voltage VOCoand the conversion efficiency qCp are similar to those of the photovoltaic case and are given by: 1 l&3 = - In (pin ehaBOa+ 1) b (where eAoBOp = 6.25 X 1Ol2~qpcx~~,pin/VBni2) and”

where V, is the electron beam potential the quantum efficiency defined by:

api,

(assuming that pin e*@Op 9 1). As we shall demonstrate later, K and eE,,J$ are numerics having experimental values less than unity. Therefore, (9) shows that a diode operates less efficiently in the electron-voltaic mode than it does in the photovoltaic mode. The reason for this is that creation of an electron-hole pair in the photovoltaic mode requires an energy equivalent to only one band-gap energy (eEph), whereas the electronvoltaic mode requires approximately (4 = 5eEPh for GaAs) for the same purpose.

and qQ4is 3. EXPERIMENTAL METHODS

3.1. Optical measurements Optical measurements using a grating monochrometor were described in our previous report [Il.

*

llc&L@mx=~

where

IBVB

)

where I,V, is the electron beam power. PLBmal= c,J,,V,, and Ises= r),l,, where Is is the beam current. Substitution in the above gives (7).

3.2. Electron beam measurements _1_. * 1 ciecrron oeam measurements were maae in a demountable vacuum system. An all-metal,

532

R. KALIBJIAN

bakeable, straight-through valve coupled the electron gun section to the demountable target section. The target holder assembly was readily removable from the holding pins in the demountable section for reloading on the bench. The assembly holding the target was mounted off-axis so that the photovoltaic effect caused by optical radiation from the cathode did not perturb the electron-voltaic measurement. Operating pressure with the loaded target was usually in the low range of the lop7 Torr scale. Pressure in the electron gun section, which was valved off from the demountable section, was in the 10eg Torr range. For the initial installation of the electron gun (type 35) the system was baked at 450°C and the oxide cathode was subsequently processed. Stable electron emission occurs when the system is recycled with a new target load if: (1) The accelerator electrode of the electron gun is RF fired during cathode processing. (2) The demountable section received a 125°C bake upon each reloading. (3) The filament power supply remains on constantly. To adjust the spot diameter of the electron beam to that of the diode (- 1 cm), a phosphor was deposited onto the collector of the target assembly. The beam was centered onto the collector aperture (diameter slightly less than that ot the diode) and the focus adjusted so that the aperture edge was barely luminescent. Focus was adjusted for each setting of IB and I/,+ The V, was varied from 1 to 3 kV for Pin ranging from 1 to 1000 pW/cm’. Care must be taken to define IB because of the secondary emission ratio of the target. In OUI investigation, the collector electrode was maintained at - 1SOV with respect to the target electrode, and IR was measured by an ammeter on the target electrode. This ensured that all secondary electrons from the target electrode were suppressed. Failure to suppress secondary emission could cause large errors. For example, if the collector were at the potential of the target electrode, apparent IR could be one-half that of the previously described suppressed current. Consequently, the calculated qca could be erroneously greater by a factor of two.

and

K. MAYEDA 4. EXPERIMENTAL

PROCEDURE

4. I. Diffused diodes

The diffused diodes were mesa-type diodes made from n-type GaAs doped with Se having 3 X IO’” donors/cm3 and mobility of 4730 cm”/V-sec. The GaAs ingot was sliced into 0.05cm wafers and then sized into 1.25cm (and some as large as 2.5 cm) diameter disks with crystal orientation in the (111) direction. Saw cut damage was removed by lapping at least 0.005 cm of material on each side with No. 1900 grit compound. The wafers were etched approximately 0.0025 to 0.0050 cm per side with a warm standard etch solution of 3H,S04: H,O, : H20, followed by a high resistivity water rinse. The closed ampoule method[4] was used for the diffusion cycle. The wafers were placed in a quartz ampoule containing p-type Zn dopant with As added to prevent decomposition of the GaAs during the diffusion process. The quantities for both Zn dopant and As were 1 mg/cm’ of ampoule volume. The quartz ampoule was then sealed onto a vacuum system and baked at 110°C until the pressure had reduced into the lOm6Torr range, and subsequently tipped-off from the vacuum system. The ampoule was placed in the diffusion furnace at a temperature of 700°C. At the end of the diffusion period the ampoule was slowly pulled from the furnace so that the excess Zn and As condensed on the end of the tube away from the GaAs wafers. Diffusion times of 90 and 25 min were used, giving diffusion depths of 6, and 3 pm, respectively. The diffusion depth was determined by angle (13”) lapping and subsequent[5] etching with dilute HNO, to reveal the p-n junction. The ohmic contact to the n-side was formed[6] by an evaporated layer of Ag (heat treated at 300°C in vacuum) and followed by an evaporated layer of Ni. The p-side appeared to have good ohmic contact without specially prepared surfaces for our evaluation purposes. Performance characteristics did not improve for an evaporated ohmic contact [7] of 96-4% alloy of Ag-Mn. Electrical leads were readily soldered onto the ohmic contacts. To obtain highly efficient diodes it is necessary to optimally etch the p+ layer of the diode. For response at the absorption edge (8500 A), a short etch period is usually sufficient. For response at even shorter wave lengths (e.g. 6000 A), the etch process becomes very critical because of the very shallow pi layer required. For this reason, a dilute

PROPERTIES

OF DIFFUSED

AND SCHOTTKY

BARRIER

GaAs DIODES

concentration of 10 to 20 parts of Hz0 is used in the etchant solution. The average thickness of the shallow p+ layer can be determined by the chord-lap method [ 81 and is typically about O-1 pm; it appears to be quite non-uniform (*O.OS pm) due to the non-planar nature of the etching process. 4.2. Schottky barrier diodes The Schottky barrier diodes (- 1.25 cm in diameter) were made with the same (10’” donor/ cm”) n-type GaAs and had metal layers of either Au or Cu in thicknesses ranging from 100 to 200 A. Following deposition of the ohmic contact on the base, a dilute standard etch (10 parts of H,O) and water rinse were given to the window side of the diode. The etch was given before commencing evaporation of the metal for the Schottky barrier, an essential step in producing high efficiency diodes. Thickness of the evaporated metal layer was monitored with a crystal counter at a rate of approximately 3 A/set. The metal was evaporated from a tungsten boat at a pressure in the low range of 1Oe6Torr. Following the window evaporation, a thick metal (same as that used for the Schottky barrier) contact ring was evaporated onto the window. Diodes with an intervening 100 A SiO layer (for the purpose of controlling the surface potential) between the Cu layer and the wafer were also fabricated. The thin films were not checked for uniformity or for pinholes. Forward current-voltage characteristics were measured for the Au( 100 A) and Cu( 100 A) Schottky barrier diodes to compare their deviations from the ideal slope of the Schottky barrier as shown in Fig. 1. Additionally, photothreshold measurements (c vs. EPh) using a prism monochromator gave the barrier heights of 0.85 eV, 0.92 eV, and 0.84 eV, respectively, for the Au (100 A). Cu(100 A), and Cu(lO0 A)/SiO(lOO A) GaAs Schottky barrier diodes. Corresponding apparent barrier heights are 1.04 eV, 1.OOeV, and I.04 eV. as determined from capacitance measurements (the voltage intercept of the l/C’ vs. bias curve). 5. EXPERIMENTAL 5.1.

RESULTS AND DISCUSSION

Dl@used diodes-photovoltaic

operation

Experimentally for the diffused diodes, V,, agrees with the predicted logarithmic behavior of (3). In Fig. 2, VO, is shown for a diode with B, = 0.3 1. with absorption edge light (8500 A) excitation,

Fig. 1. Forward

I-V characteristics barrier diode.

of Schottky

p,,- lLWW/Crn2 Fig. 2. V,, for diffused diode at 8500 A.

and pin less than 20 pW/cm2. The departure from logarithmic operation at the 0.01 pW/cm2 range can be explained[3] as degradation from the loading condition of the voltmeter. The threshold sensitivity for logarithmic response is in the 0.001 pW/cm2 range.

534

R. KALIBJIAN and K. MAYEDA

The nC, V,,, and V, (load voltage at the maximum power point of the Z-I/ characteristic curve) are shown for another diode in Fig. 3, as reproduced from our previous report [ 11. The exciting wavelengths are 8500 A and 7000 A, with corresponding B0 of 044 and 0.41, respectively. Performance data at 7000 A are for the same diode, which has been optimally etched to obtain maximum Q. Since a photovoltaic diode is generally a shallow junction device, recombination processes at the surface play a very important part in diode efficiency. A diode which has not received an etch after diffusion has a rather large A (= 2.5). With the first etch, A decreases to about 2 then continues to decrease very slightly even when the diode is slightly over-etched. However, excessive overetching produces a saturation curve as shown in Fig. 4. The saturation characteristic can be explained by referring to the expression for V,,. When the diode is undergoing optimization, junction depth becomes shallower. For an over-etched surface,

L7oooA

1

4-

I,

roQo4VL

o-

“I*’

c

-0.2

I 100

10

IOOP

pm- rWcn/ Fig. 3. ne for diffused diode.

__----

,

I

100

10 p;,

-

/ ,000

ww/cm2

Fig. 4. V,, for diffused diode etch-optimization.

_--

the p-region is perhaps slightly doped. Because of this, the approximation (AP/~~)~ Q 1 (assumed in deriving (2)) no longer holds. The reason for saturation of V,,, vs. pin can be ascertained trom (1); the denominator of the argument of the logarithmic function increases with increasing pin. TO support the hypothesis that (Ap/p,,)p becomes significant, we can make point probe tests of sheet resistivity on the p-layer during the optimization process. And, in fact, we do find regions where the ohmic contact changes to rectifying contact. Evidently, we are dealing with a p-layer which, at best, is a compensated region that is slightly p-doped. For this reason, we make the assumption (in section 2.1) that power absorption in the p-region is negligible. We also infer that there are ‘windows’ of n-region on the surface of the diode, as deduced from the nonuniform thickness of the p+ layer investigated with the chord-lap method previously discussed. Therefore, it becomes possible for the electron-voltaic efficiencies (both quantum and conversion) to extend below 1 keV. The effects of varying etch time are shown in the spectral response curves of Fig. 5. Generally, etching the p+ window improves the efficiency because the electron-hole pairs are generated more closely within a diffusion length to the p-n junction. Minority carriers are thereby collected more efficiently. Maximum response occurs for a 60-set etch at about 8500 A. Subsequent etching reduces the maximum responses but increases the short wavelength response. Optimum etching is quite critical for improving the short wavelength response. In some diodes, we have obtained nC of 11% at 7000 A and 200 pW/cm2. A major problem is determining proper termination time for the etch process. In general, we have observed that efficiency is destroyed by any slight amount of additional etching. Note (Fig. 5) the catastrophic decrease in nC at 7000 A when etch time is increased from 68 to 69 sec. This perhaps can be attributed to a smaller active area of the diode, as the non-planar etching process (as observed from chord-lapping) may produce islands of p+ regions which are no longer connected to one another, thereby reducing r)c. The decrease in 77, at the peak response (8500 A) from over-etching is primarily caused by decreases in the V,,, I,,, and c, as shown in Fig. 6. The initial

PROPERTIES

OF DIFFUSED

AND SCHOTTKY

20

535

BARRIER GaAs DIODES

7OOOi

50set

etch

7

Fig. 6. V,,, I,,, and c vs. relative time for diffused diode etch-optimization.

3^ P I

-0.4

::

8000 Wavelength

Fig. 5. Spectral

-

A

response for diised 200 pW/cm2.

Y

diode at

increase in r), at the short wavelengths (7000 A) from etch-optimization is caused by increases in VoC,I,,, and c. Near the maximum r),, V,,, and c decrease, and I,, increases (for increasing efficiency). Further etching, of course, decreases all three parameters.

-0.

OF--7’-=-

I

___+-------. “110

I

I

I,

I 100

I

/

,000"

Fig. 7. Electron-voltaic and photovoltaic properties of diffised diode.

nqp becomes linear and can be written as (8). V,, is the extrapolated dead voltage of approx5.2. Difused diodes-electron-voltaic operation imately 1.9 kV for the diffused diode. The slope of the quantum efficiency curve in the linear range Electron-voltaic measurements were made with diffused GaAs diodes. Experimentally, the VoCa of Fig. 8 gives an average energy of 4 = 6.8 eV to produce an electron-hole pair in GaAs. and qca are logarithmic functions of pin (Fig. 7). We now determine the numeric K appearing in This generally agrees with (6) and (7). ( 10) as follows: The experimental 7pp vs. V, (Fig. 8) is measured (1) Experimentally, c is about the same for both For high beam energies, as s3 = (I,, -[,)/I,.

R. KALIBJIAN and K. MAYEDA

536

electron-voltaic and photovoltaic operation. Therefore, co/c = 1. (2) The experimental r], = 0.46. From Fig. 8, at 3 kV for the diffused diode, (1 - V,,/V,) = 0.38; therefore, (1 - V,,/V,), Q = 0.83. (3) From Fig. 7, where both photovoltaic and electron-voltaic data are given, 0.7 < I/& I’,, 4 0.9 for pi,, in the 1 to 1000 pW/cm2 range. 40(

A:

3oc

/3

-----0

Schottky

barrier

diode

B: Cu( lOO,&)/GaA,nl

Au( lOOA)/GaAr

Schottky

barrier

diode

C:

Au( 2008,)/GaAr

Schottky

barrier

diode

D:

p+!‘n

GaAr

diffused

vCp and voCp vs. V, (Fig. 9) for the diffused diode. Below 2.5 kV, r~,.~decreases slowly with respect to decreasing beam energy, similar to the decrease in qQOin Fig. 8. We will show that the calculated range of 2.5-keV electrons in GaAs is approximately equal to the measured thickness of the pi region. According to Feldman [9], the range-energy expression is of the form:

diode

m 7 -m _u Y- 2oc m LT c

100

R = hE”

-25 2

1 VB -

3

kV

(11)

2

VW/cm

0.2,L---;___’ 3 VB

Fig. 8. nquvs. beam potential.

-kV (0)

The numeric K has a value 0.6 < K < 0.7 in the pi,, range of 1 to 1000 pW/cm* for the diffused diodes. In the photovoltaic case, electrons can be excited from the valence band to the conduction band with energy equivalent to only eE,,, whereas the electron-voitaic case requires 4.7 times the band-gap energy (+/eE,, = 68/1.45 for GaAs) to excite electrons into the conduction band. Evaluation of (9) gives the ratio ~~/r), = 0.213 K. The calculated points for qCp are shown by the dashed curve in Fig. 7 and are based upon the estimate given by (9). Some of the discrepancy between the calculated points and the measured qCa curve can perhaps be attributed to the inaccuracies involved in the estimate of c from the measured Z-V characteristic curves, and to our neglection of reflection losses at the diode surface. In addition, we show

VB

-kV (b)

Fig. 9. Electron-voltaic operation of diffused diode. (a) VOCrr vs. beam potential. (b) vCs vs. beam potential.

PROPERTIES

OF DIFFUSED

AND SCHOTTKY

where E is the energy of the electrons in keV, n is a constant given by n = 1*2/( 1 - 0.29 log,, Z), and b is a constant related to the material by: b = 250MlpZ”” A M is the molecular weight of the material, p is the bulk density, and Z is the average atomic number. Substituting the proper parameters for GaAs (M = 144, p = 5.3 g/cm3, and A = 32) gives n = 2.51 and b = 88.5. For 2.5keV electrons, the range in GaAs becomes R = 885 A. The diodes were optimally etched to obtain high During the etch process (after efficiencies. diffusion) the surface of the wafer acquires a rough finish. It is highly probable that the p-region occurs only on the peaks of the hill-and-valley type of surface and that the 885 A thus represents a thickness equivalent to the p-region peaks averaged over the entire area of the diode. The average thickness of the p-region is found to be about 1000 A when measured by the chord-lap method which compares favorably with the theoretical range calculated with [ 111.

531

BARRIER GaAs DIODES

Pin > 10 pW/cm”. Saturation of the V,, curve is not caused by increased surface resistance. This was ascertained when we intentionally lowered surface resistance by evaporating metal bars onto the surface; no change in slope occurred. The n, for the Schottky barrier diodes is no longer a strictly logarithmic function of pin as was the case for the diffused diode. For example, the photovoltaic performance data for the Au/GaAs, Cu/GaAs, and Cu/SiO/GaAs diodes are shown in Figs. 11, 12. and 13. respectively. Curves for c and Q, are also shown; neither of these curves is constant with respect to pin as was

I

x

5.3. Schottky barrier diodes-photovoltaicoperation The V,, of a A~(100 A)/GaAs Schottky barrier diode, excited with the 6328 A line of the He-Ne laser is shown in Fig. 10. Also shown for comparison is the V,, for a diffused GaAs diode. NO saturation effects of V,, occur for the diffused diode for power densities up to lo4 pW/cm’, whereas the Schottky barrier diode begins to show saturation effects at 10 pW/cm2. From the slopes of the semi-log graph of the curves in Fig. 10, the diffused diode has an A = 2, and the Schottky diode has an A = 1.5 for pin < 10 pW/cm’, and an A = 1 for

Fig. 10. V,, for diffuse$ and Schottky barrier diodes at 6328 A (He-Ne laser).

0. I

I

I

IO Pin

-

100

pw ‘ml2

Fig.

11. Photovoltaic properties of Au/GaAs Schottky barrier diode at 6500 A..

Fig.

12. Photovoltaic properties of Cu/GaAs Schottky barrier diode at 6750 A.

I

R. KALIBJIAN

and K. MAYEDA

I

I

I 100

I

0.7-

0.6-

0.5-

0.4-

0.3-

0.21

1 10 pw

-

,,VUcm2

Fig. 13. Photovoltaic properties of Cu/SiO/GaAs Schottky barrier diode at 6750 A.

the case[ l] with the diffused diodes. The calculated data points for qc as determined from (5) are indicated by x in Figs. 11, 12, and 13 and can be compared to the experimentally determined nr from P Lmax/apin. The close agreement of these data points serves as a self-consistent check on our measurements. It is interesting to compare Figs. 12 and 13 because the intervening layer of 100 8, SiO for the Cu diode increases 7, from 6 per cent to about 8.5 per cent at the 100 pW/cm* level. According to (5) and Figs. 12 and 13, the higher Q for the Cu/SiO diode as compared to that of the Cu diode primarily results from the enhanced values of c and V,,,since the value of q, is about the same for both types of diodes. The spectral response of the Schottky barrier diodes is shown in Fig. 14. The peak response for nc occurs at about 6500-6750 A, whereas the peak for r), occurs at about 6000 A. The c and V,,are almost constant in the 5000 to 8000 A spectral

o.7+kq Cut lOOA)/SiOf

lOOi)/GaAr

c

CdlOO~,/GoA, 0.6-

0” I :-

> e 0 ”

:: ”

0.5rCuC

lOOi)/SiO(

cc

lOOi)/Gak

v oc -

0.3 5000

Au, IOOA

j,/GaAr

1

I

,000

6000

Wovelength

-

.&

(b)

Fig.

14. Spectral

response

of

Schottky barrier diode at 200 pW/cmZ. (a) nc and vs. (b) c and V,,.

8000

PROPERTIES

OF DIFFUSED

AND SCHOTTKY

BARRIER GaAs DIODES

539

range. As was done for the curves of Figs. 11, 12, and 13, the data points marked x (calculated from (5)) compare favorably to the r], determined experimentally from PLmJflPin. 5.4. Schottky operation

barrier

Performance data CuIGaAs Schottky Figs. 15 and 16. The larger than that for primarily because of

diodes-electron-voltaic

for the Au/GaAs and the barrier diodes are shown in qcs for the AulGaAs diode is the CulGaAs diode at 3 kV the higher qQoof the AulGaAs

I 2

0.4

V

B

I 3

-kV

(0) 3

I 0.1’

1

L25 tLw/;m2

I 1000 pW/cm

1

I

2 VB -

3 kV

(0)

5-

3 VB -

kV

b)

Fig. 16. Electron-voltaic operation of Au/GaAs Schottky barrier diode. (a) Cocavs. beam potential. (b) qea vs. beam potential.

o-

5-

/ * O1

2 VB -

kV

(b)

Fig. 15. Electron-voltaic operation of Cu/GaAs . . Schottky bamer diode. (a) VW0vs. beam potential. (b) T,~ vs. beam potential.

diode (Fig. 8). In general, PO/oca of the Au/GaAs Schottky barrier diode is less than that of the diffused diode. However, qep is somewhat better than that of the diffused diode because of the higher nti for a beam energy greater than 1 keV. Also, from the npp curve of Fig. 8, the 4 required to produce an electron-hole pair is about 6.8 eV for the Schottky barrier diodes (same as that of the diffused diode), as may be expected since $I should be independent of the type of diode structure. Zt is interesting to point out that according to Feldman[9] the range of 3*4-keV electrons in Au

R. KALIBJIAN and K. MAYEDA

540

is about 100 A. Since qQa is linear with respect to V, beyond 2 keV in Fig. 8 for the Au( 100 &GaAs curve, the beam electrons appear to have penetrated into the space-charge region of the diode. Although no study was made of the metal film structure on GaAs, we tentatively suggest that the Au film is relatively porous (similar to island formations at nucleation sites with interconnecting links) as compared to the Cu film because of the 0.8-keV shift in the qqp vs. VB curves between the Au and Cu diodes. Additionally, the qcp for the Au/GaAs and Cu/GaAs Schottky barrier diodes is estimated from (9) by using the parameters from Figs. 11 and 12.‘” The results are shown by the dashed curves in Figs. 15 and 16 for pi,, of 100 pW/cm2, for the linear range of the respective curves for Au and Cu in Fig. 8, the numeric K has a value 0.6 < K < 1 .O. The estimate given by (9) is within 15 per cent of the experimental curve for Schottky barrier diodes. 6. CONCLUSIONS Experimental confirm the measurements theoretical prediction that the qrq is less than qc by the ratio K(eE,,,/$). This is consistent with the

fact that greater energy is required to produce an electron-hole pair with an electron source than with an infrared radiation source. The energy required to produce an electron-hole pair in GaAs is $J = 6.8 eV. The threshold energy for detection

.kExperimentally c,/c = I. From Fig. 8. V0(Au/GaAs) = 1.0 kV and P’,,(Cu/GaAs) = 1.7 kV. From Figs. 1 I and 12, r), is obtained to calculate (I - VO/VB)/vq Also, from Figs. 11, 12, 15, and 16, Vocd/VOris obtained. Therefore. K can be calculated from (10) for a given T/B,and thereby Q~/s, can be estimated for E,,,, (Cu) = 1.83 eV and E,,,, (Au) = 1.90 eV. and for 6 = 6.8 eV.

in the electron-voltaic mode is less than 1 keV for these diodes. For the diffused diodes at 100 kW/cmz, the Q can be as high as 18 per cent at the absorption edge excitation (8500 A). Short wavelength response can be enhanced with controlled etching on the front window to 7, as high as 11% at 7000 A. For extending the response to even shorter wavelengths, Schottky barrier diodes become attractive since the peak response can occur at about 6500 A with qlc of 6 to 7 per cent at 100 pW/cm’. The q. for both diffused and Schottky barrier diodes is about 50 per cent for the peak spectral photovoltaic response. The V,, at 100 pW/cm2 can be as high as 0.7 V for the diffused diodes and 0.4 V for the Schottky barrier diodes. The V,, and qc are logarithmic functions of pin for the diffused diodes, whereas saturation effects for the Schottky barrier diodes (presumably caused by surface traps) occur in the 10 pW/cm2 range for both photovoltaic and electron-voltaic modes of operation. These diodes (diffused and Schottky barrier) have a threshold sensitivity (V,,.) in the 0.001 pW/cm’ range. REFERENCES 1. R. Kalibjian and K. Mayeda, Solid-St. Electron. 12, 823 (1969). 2. E. 0. Johnson, RCA Rev. 18,556 (1957). 3. R. Kalibjian and H. Huebel, IEEE Trans. IECI-17, 173 (1970). 4. S. W. Ing and W. Davern, .I. elecfrochem. Sot. 111, 120 (1964). 5. J. H. Yeh and A. E. Blakeslee. J. electrochem. Sot. 110, 1018 (1963). 6. W. A. Schmidt, J. electrochem. Sot. 113, 860 (1966). 7. C. J. Nuese and J. J. Gannon,J. elecfrochem. Sot. 115, 327 (1968). 8. B. McDonald and A. Goetzberger, J. electrochem. Sot. 109, 141 (1962). 9. C. Feldman. Phys. Rev. 117,455 (1960).