Metal on GaAs: from Schottky barriers to ohmic contacts

Metal on GaAs: from Schottky barriers to ohmic contacts

Applied Sat facu Science 56-58 (19921 335-340 North-HeSland surface science M e t a l o n G a A s : from S c h o t t k y barriers to o h m i c c o n...

243KB Sizes 0 Downloads 7 Views

Applied Sat facu Science 56-58 (19921 335-340 North-HeSland

surface science

M e t a l o n G a A s : from S c h o t t k y barriers to o h m i c c o n t a c t s s. Perfetti, R. Carluceio Dipurlimento di ~sica, Unit'L'r~it~'l "'Lu Stlpienza'" Pie A. Mum 2, 00185 Roma. Italy

C. Lanzieri, A. Cetronio ALENIA, Reran. Ittdy

and

C. Coluzza IPA-Dt~artemellt de Physique. PHB.F.cuhh,n.~, 1015 Lalt~dmu; Switzerland and Dil~artimcnlo di Ft.rtt'a, Unh'er~it~'l "'1~1Sapit'nza '; I~& A. MonJ 2, 1~185 Rmna, Italy Received 311 May 1991; accepted for publicalion 12 July 1991

Ohmic contacts on GaAs obtained hy using different ledlnological procedures, metal evuporatlun an chemically etched samples fECES) and in-situ metal sputtering on argon pre-bombarded samples (SANS1 have been inves6galed. In bmh cases, samples post-annealed at different temperatures have been studied. The kinetics of formation of ohmic ctmtacls in an Ni-Ge-Au/GaAs heterostructure has been determined by internal phatoemission tlPEI measurements at 77 K. whose results have been compared with those of deep level Iransienl spectroscopy (DLTS). and with C-I: and I - V lechniques. The ohmic behavior of the contact has been related to localized levels at the interface and to the band offset values, Barrier height values (Oh) of 11.81 cV for ECES and 0.91 eV for SAILS have been estimated hy IPE measurements. I~th values decrease to ~ 0.75 eV after annealing at 3OO°(?.. A feature al 137 eV, delected on top of the pholnemission background of untreated samples, quenches out in ECES once these are annealed at lemperatures higher than 31hl°C. while it increases in SABS. Furlbermore. the GaAs gap in the depletion region ,;brinks after a temperature Ireatmenl. Finally. the resistivity tff these helerojunctions, as determlned hy transmission line model (TLM) measurements on bnlh lyp¢ of s;imples after an annealing c'jcle, are 5P% lower in ECES than in SANS. This result agrees with DLTS data. which give evidence of two donor levels 0.12 and g,2 e v below the GaAs conduclion band. The density of these levels is higher in SAnS than in ECES for annealing lemperatures lower than 25[)°C~ prohably because of the surface damage induced by the sputtering process.

1. I n t r o d u c t i o n C o n t a c t s play a n increasi,tgly i m p o r t a n t role in solid s t a t e e l e c t r o n i c m i c r o - d e v i c e s , w h o s e final d i m e n s i o n is s o m e t i m e s l i m i t e d by lack o f stability o f c o n t a c t s a n d i n t e r c o n t , e c t l o n s . O h m i c c,t m tacts o n n - t y p e O a A s a r e c o m m o n l y realized by alloying o f a N i - G e - A u m i x t n r e lit l c m p c r a t u r e s b e t w e e n 360 a n d 5 0 0 ° C. A h h o u g h a g r e a t d e a l o f w o r k h a s b e e n d o n e (m t h e electrical a n d

m e t a l l u r g i c a l p r o p e r t i e s o f t h i s system, r e c e n t l y subject o f a review [ 1 - 3 ] , t h e basic m e c h a n i s m o f c o n d u c t i o n in this syslcm h a s n o t yet b e e n established. A m o n g t h e several type o f t h e o r e t i c a l models which have been proposed, two seem In receive t h e w i d e s t c o n s e n s u s , i n t h e first o n e , G e diffuses i n t o t h e G a A s f r o m t h e m e l t e d A u - G e e u t e c t i c l e a d i n g to t h e f o r m a t i o n o f a s u r f a c e n * - G a A s layer. E l e c t r o n s m a y t h e n q u a n t u m m e chanically tunnel through the thin surface poten-

0169-4332/92/$05.1hl ~:, 1002 - Elsevier Science Publishers B.V. All righl.~ reserved

33fi

S. Perf*'lti ¢1at, / Metal rm Go~4s

tial barrier [4,5]. The second type of model postulates the existence of a graded N i - A u - G e / G a A s hcterojunction, where the structural disorder gives rise to electronic states in the forbidden gap of OaAs. Tunneling of electrons from these resonance states into the metal provides an ohmic contact. A model of electron tunneling through a metal-semiconductor gives p~O~exp{a~h/N~/2) [6,7], where pc is the specific contact resistance, ~n the interface barrier height, Nd the donor concentration, and

a = ~ x t0 10 c m - 3 / 2

eW- i

Howcvor this model cannot be applied to the case of an alloyed contact, because d~h as well as the effect of Ge, An, or Ni iutcrdiffusion on Na can be hardly determined in this ease. The purpose of this work is therefore to study how thc Ni-Og-Au/n-GaAs interface properties change as a function of the heat treatments at temperatures below the Au-Oe cutectic and of the GaAs surface initial conditions. Internal photocmission (IPE) measurements [8], I-V and C-V [9] characteristics have been obtained at room temperature and "17 K. Defect states induced at the interface by the contact, have been detected by DLTS [10] and by IPE measurements. Finally, all data have been correlated with R~ as deduced by the transmission line model ( T L M ) i l l ] .

2. Experimental Ohmic contacts on L E e (liquid encapsulated Chocralzky) GaAs(110) snbstrates have been obtained by two differentprocedures. In the first one, I ~ 0 A of Au, 250 A o f G¢ and 500 ~ of Ni have been evaporated, in that order, onto a chemically etched (NH 4OH : H 202 : H 20) GaAs surface (evaporation on chemically etched sampies, ECES). In the second one, the same metallic layers have been sputtered at 1 keV onto a GaAs surface previously, in-situ cleaned for 3 rain by GaAs surface Ar sputtering at 500 eV {sputtering on argon pre-bombardment samples, SABS). High vacuum conditions ( < 10-7 Tort) were maintained during these processes in both cases. A conducting GaAsn + layer was obtained by 2SSi+ ion-implanting at 120 keV and at a dose of 5 × 10 n: cm -2 before metal deposition; the

GaAs surface was then capped with Si3N4 arLd annealed at 820°C for 20 min to remove the implantation damage and activate the dopants. Temperature and time are the two common variables of heat treatments. In this work time was kept constant for all samples, while two temperatures, 250 ° C and 300 ° C, where used. The thermal process was achieved by rapid thermal annealing (RTA). This technique, based on lamp irradiation induced heating, permits a real control of temperature time-dependence in the process of contact formation. Details of the IPE technique set up are deseribcd elsewhere [8]. In this experiment, the light beam whose energies range between 0.5 and 1.6 ¢V, travels through the GaAs substrat¢ bcfore reaching the semicondoctor-metal interface. A current through the interfaca is measured when photons induce electron transitions from the metal Fermi energy to the GaAs conduction band, allowing a direct determination of barrier height. A current signal may also be produced by electronic transltlons from localized gap states to the conduction band, or from the GaAs valence band to localized gap states [8]. In this case, in fact, the photoexcited electron-hole pairs may be separated by the junction field before they recombine. However the carriers trapped at the localized gap states can be neutralized by carriers injected from the other side of the junction, thus giving a steady-state photocurrent, only if these states are near the interface. Thcrefore, the detection of steady-state photocurrent may provide evidence of near interface states. Using DLTS we studied Ihc deep levels at the interface. DLTS is a high-frequency capacitance transient thermal scanning toothed that can give information about deep levels in semiconductors. Typically in a DLTS spectrum traps are displayed by positive or negative peaks on a flat baseline: the intensity of the peak is related to the trap concentration, the sign indicates its closeness to the conduction or valence band and the position in temperature is a unique function of traps tbermal emission properties. Finally, applying the TLM model, we were able to determine the contact resistance in our samples. This experimental technique consists of

S. Perfetli er aL / Metal on G a A s

depositing a metal grid pattern having a n u m b e r ( N > 4) of opportunely spaced contacts on the surface of the sample. A current is forced between the two extreme contacts and the induced voltage drops are measured between the other contacts. By a direct extrapolation of data it is possible to determine the value of R~, the contact resistance normalized to the contact length.

337

3. R esul ts

In unannealed samples a barrier height {~b) of 0.91 e V for S.~ BS, and of 0.81 e V for ECES, has been estimatud from IPE (see insets in fig. l a and fig. lb). These results are in good agreement with data previously reported [12-14]. The difference in ~kb could be explained both in terms of oxygen

200 t

160i~ 12o g

%

C



rna~,r~i~ytee {3 77K -.025V

! 40 0

~ 1

,

77K

• *

77~ .75V 77K OV I

..sv

I

0.6

0.8

I

1,2

1.4

1.6

Pheton Energy leVI 200

b

"'7

== 120

80

[ 40

0

ii 12 ~o~ E ~ t~vl

o~

I



77K 0V

x o •

77K -.25V 77K -.SV 77K -.75V

'

13

J

1.2 1.4 1.6 I Photon Energy [eV] Fig. I. Internal phntoemission cu,'v¢i for melal evaporation on chemically-etched ~araples ECES (fig. la) and in-situ metal sputterinf on argon pre-bombarded samples SABS (fig. Ib), before thermal annealing, for different bias. The insert gives delails oil the determination of ~h. 0.6

0.8

s, Perfelli et aL / Mrlat ~z GaAs

present at the GaAs surface before metal deposition in the case of chemical etching procedure (wet method) or in terms of a different change in the interface stoichiometry for the two procedures. In parilc'e!ar the !a!*,er ~.nlanation assumes that the wet method results into an As excess at the GaAs surface, while the sputtering procedure induces Ga excess [15], thus resulting into a higher value of d~h as reported in the literatore [14,15]. Furthermore it is well known that oxYgen at the interface results in a Fermilevel pinning at about 0.75 eV [14] above the GaAs valence band edge. This effect seems to be more consistent with the different preparation of semiconductor surfaces: the sputtering method results in a oxygen free surface, giving the higher 6b. The barrier heights obtained by IPE were compared with those obtained by I - V and C - V characteristics (see table 11. The lack of agreement points toward an important role of interdiffusion as well as of recombination at the interface. In particular scarce abruptness of the interface could result in misleading I - V and C - V analysis. In fig. 1 are also rcportcd IPE curves for different bias values: as expected, the GaAs band-to-band peak at ~ 1.47 eV as well as all photocurrent signals increase with the increase of reverse bias. In addition to the internal photoemission threshold and to the band-to-bend peak, a threshold at 1.37 eV is also evident whose strength is higher in ECES with respect to SABS. Table I

Comparison ~f barrier heights obtained by IPE and I - V and C-V analysis Temperature

~b(I-V) teV)

,~htC-V) teV}

~bh(IPEI {eV)

E C E S NrJ = 5 × I 0 1 7 c m

RT

0.50

0.6O

250oC 300"C

0.56 0.45

-

350 ~ C

0.40

-

0.81 0.~0 0.76

S A B S N D ,~ 3 X I O I T cra - s

RT

250°C 300°C 3S0°C

0.55 0.65

0.81

0.91 0.01 0.74

0.0 t. . . . . . . . . . . . . . . . . . . . . . . . . . .

-2.0i ~

~

-4.0

T

;

,,

i -IO.D 70

90

II0 130 150 170 190 210 23D 250 270 T(K}

Fig. 2. DLTS spectrum in ECES and SAGS.

This fcatura has been explained in terms of a transition from the GaAs valence band to a bulk GaAs defect state [8], Near the interface this slate, normally filled in n-type material, is emptied by band bending. The intel;sity of DLTS measurements in the same samples (see fig. 2) gives evidence of two levels, 120 and 200 meV from the GaAs conduction band. Their intensity is higher in SAGS, which could be explained in terms of the disorder on the GaAs surface induced by the sputtering procedure. This result docs not seem to agree with the data previously reported by lliadis [16], where it was suggested that the disorder at the interface could induce a barrier height reduction. However it is worth noting that in that case the barrier height was determined by the indirect method of I - V charaeteristics and that the disorder was due to alloying. lu ECES, the barrier height increases up to 0.90 eV after an annealing at 250 ° C, decreasing afterwards m 0.76 eV for an annealing at 300°C. This could be explained by a surplus of the native oxide at the interface followed by interdiffusion effects. Instead in SABS (oxygen free), the dJb remains constant for the first annealing and decreases to 0.74 eV after tile highest temperature annealin~ (see table 11. 1PE curves after different temperature treatments are reported in fig. 3a and fig. 3b, For both procedures, the collection efficiency of the diodes is reduced by metal diffusion into the GaAs substrate. In the insert, the same curves are shown after normalization to the band-to-band peak in order to correct for the

339

S. Perfeuict aL / Metal ol) GaAs

ment with the observation of an equal value of 4~b in these two samples, once one assumes that the threshold at 1.37 cV is due to Ga vacancies. The trend in the reduction of that signal is in good agreement with DLTS data showing that, in ECES samples, the features at 120 and 200 meV are basically cancelled after annealing, while the

difference in collection efficiency. The photocurrent signal for photon energies less than the GaAs gap is strongly reduced by the interdiffusiun in wet processed samples, while it is increased in the sputtered ones. Moreover, the spectrum of 250°C annealed ECES coincides with that of SABS without annealing, in agree-

2O

j

16

~=

a

"

n o ec,n e a l i n g

8.0 4.0 0.0

0.6

o.g

I 1.2 Photon Energy [oV]

1.4

1.6

10.0

6.04.0 I °°°!~ L

I~ /

/

25oOcannealing

\

E

2.0 i 0.0 ~

0.8

-

I 1.2 1.4 1.6 Photon Energy [eVl Fig,3. IPE measurementsin EC£S(fig.3a) and $AB$(fig.3b) before and after differentthermal annealing.The samecarvesare reporled in the inserlafter band-to-band normalization, 0.6

34O

S. Perfenl el al. / Metal on

corresponding signals are always present, although reduced, in S A n S samples up to the 3 O 0 ° C treatment. In the same figures are indicated the G a A s gap at the interface as deduced by the extrapolation of the high photon energy cutoff [8]. The shrinkage induced by annealing is evident. This feature is different for the two samples: in particular the reduction is smaller in the SABS sample and it is already detectable at 2 5 0 ° C annealing te 'lperature, while in the E C E S sample it is necessary to reach 3 0 0 ° C to obtain the shrinkage. These results further support the presence of oxygen at the interface in the wet procedure. The oxide layer prevents interdiffusion until an annealing temperature of 250 ° C is achieved: at this point E C E S samples become similar to S A n S ones in terms of Ckh and the density of states inside the gap, still remaining the difference in surface damage induced by ion bombardment in S A B S samples. Such larger damage could induce recombination as well as traps at the interface explaining the different values of electrical resistivity of the contacts as measured by the T L M methods after the final thermal process of melting (450 ° C for 30 s in forming gas atmosphere). In effect we obtain a R e = l . ] 0 x 10 - t ~ . m m for the E C E S and R L = 1 .7 4 × 10 - j ~ q ' m m for the SANS.

4. Conclusions Using IPE measurements on N i - G e - A u / GaAs, we demonstrate a clear correlation between the semiconductor s u r f ~ e treatments and the final ohmic characteristics, in particular the pre-sputtering procedure induces more localized levels and disorder at the interface than the wet one, which cannot be completely recovered by thermal annealing. We al3o found that annealing at temperatures below A u - G e eotectie ( 3 6 0 ° C ) reduced the harrier height of about 150 m e V as well as the value o f the G a A s gap in the depletion region. Furthermore the G a A s energy gap at the depletion region, seems to be reduced by interdiffu-

GaAs

sion of the metals into the semiconductor. For the moment it is not clear if such effect is due to G e - G a A s alloy formation or if it is induced by. the presence of an amorphous phase at the interface. All these results indicate that best and more stable alloyed contacts are achieved by modifying the chemistry and electronic properties at the interface. O n the other hand, the formation or the increment of interracial states could increase the resonant tunneling increasing the current through the interface, but it also could increase the recombination rate at the interface: this resuits in worse final ohmic characteristics.

References [I] G.Y. Robinson. in: Physics and Cberaistp/ of III-V Compound Semiconductor Interfaces, Ed, C.W, WiImsen (Plenum, New York, 1985)p. 73. [2] CJ. Palrmstrornand D.V. Morgan. in: Gallium Arsenlde Materials, Devlces and Circuits. Eds. MJ. Howes and D.V. Morgan (Wil~y, New York, 1985) p. 195. [3] N. Uraslau, Mater. Res, Symp. Prec. 18 (1983) 393. [4] N. nraslau. J.B. Gunn and J.L Staples, Solid State Electron. IO (IQ67) 381.

[5) V.L Ridcout, Solid State Electron, 18 (1975)541. [6) F.A. Kruger, G. Dierner and H.A. Klasens, Phys. Roy. 103 (1959) 279. {7) C.Y. Chang,y.IC Fang and S.M. Sze, Solid State Ell.e-

m~. 14 (1971) 541. [Sl C. Coluzza, A, Neslia, A. Uennouna, M. Capizzi, R, Carluccio, A. Frova and P,C. Srlvasta','a,Appl, Surf. Sci. 55-58 (1992) 733. [9] S.M. Sze, Physics of Semiconductor Devices (Wiley, New Yozk, 1981). [10] D.V. Long, J. Appl. Phys. 45 (1974) 3023: G,L. Miller, DN. Long and LC. Kirnerling, Annu. Rev. Mater. SOL7 (1977) 37. [11] H.H, Burger, Solid State Electron. 15 (1972) 143. [12] W.G. Spitzer and CA. Mead. J. Appl. Phys, 34 (1963) 3Wol. [13] 3.R. W:ddrop. 3. Vac. Sci. Technol. n 2 (1984) 445; Appl. Phys. Lea. 44 (1984) 1002. [14] W.E. Spicer, g. Liliental-Weber, E. Weber, N. Newman, T. Kcndelewicz, R. Coo, C. McCauts, P, Mahowald, K. Miyano and I. Lindau, £ Vac. Sci. Teebeol, B 6 (1988) 1245. [15] Y.X. Wang and P.H. Holloway,J. Vac. Set Technol. E 2 (1984) 613. [16] A. niadis, $. Vac. Sci. Technol. B S (1987) 1340.