- Email: [email protected]
Formation mechanism of Schottky barriers on MBE-grown GaAs surfaces subjected to various treatments
Applied Surface Science 56-58(1992) 317-324 Norlh-Hngand
appeal
surface ~3ience
Formation mechanism of Schottky barriers on MBE-grown GaAs surfaces subjected to various treatments Hideki Hasegawa, Hirotatsu Ishii and Ken-ichi K0yanagi Deparrme/ar of Elcclrical En~kJeering and Reseereh Cenler for Interface Qumtgml Electronics, Hokkaido Umrersiry,. Sapporo 060, Japan Received fi May 1991:accepted for puhllcaUon 12 July 1991
AI. MS and Au Scholtk'j barriers were forme,~ on MBE-grown GaAs(100) surfaces subjected lo various treatments including cbamical etching, ion etching, sulfur treatmcnl and insertion of an uhralhin $i interface control layer (Si-ICL), They were characterized by the X-ray plmmelectron spectroscopy (XPS), current-voltage (l-V) and capacilance-vuhagc (C-V) techniques, The XPS lesults indicated the presence el an interfacial layer (ILl in each case. Ahhougb Ihc bchavinr is far from Ihe ideal Schouky limil, the barrier height showed dependences on the properties of ILs and the metal work-ttmction. Based on the disorder-lnduced saP state (DIGS) model, a theory including Ihe effect or an insulator-like or sumiconductor-lik¢ IL was developed. Tba theory explains the observed behavior rca~mably well. showing that the detailed nature of the IL is the key feature for the understanding and control of Scholtky barriers.
I. Introduction In spite of a long history of research which has led to numerous models [1-4], the formation mechanism of the Sghottky barriers has not been amply clarified yet. Recently, artificial changes of Schottky barrier heights (SBH) over a wide range have been achieved by sulfur treatments [5-7], insertion of Si interlayers [8,9], ctc., whose mechanism, however, still is not clear. T h e present study was carried out to identify the dominant mechanism determining the SBH. AI, M g and An Schottky barrier were formed on MBE grown (100) G a A s surfaces subjected to various treatments including chemical etching, ion etching, sulfur treatment and insertion of an ultrathin Si inter layer or interface control layer (ICL) [10]. T h e X-ray photoelectron spectroscopy (XPS), current-voltage ( l - V ) and capacitancevoltage ( C - V ) techniques were used for characterization. The X P S measurements indicated the presence of interracial layers (ILl with various compositions in all cases including the ease of the bar-
tier on the clean M B E GaAs. $BH were found to depend on the properties of II..s and the metal work-function. A phenomenological theory of Schottky barrier formation based on the unified disorder-induced gap state ( D I G S ) model [4,11] was then L.eveloped where the presence o f insulator-like or semiconductor-like IL is taken into aCCount, it is concluded that the observed behavior of 8 B H can he explained reasonably well by the theory and that the detailed nature of the IL is the key feature for understanding and control of Schottky barriers.
2. Experimental The experiment was done in a system having the MBE, deposition, XPS, and electrical characterization chambers, all connected by a U H V transfer chamber. Clean G a A s surfaces were prepared by the conventional M B E technique. The surface exhibited a clean As-stabilized (2 × 4) reconstruction pattern at the growth temperature on the substrate (Tsuh) of 580°C.
0169.4332/92/$05.00 © 1992 - Elsevier Science Publishers B.V. All rights reserved
318
I~ Hulegawael aL / FomuJli~mnu~'hm#smof Schoaky tm~iers m~ MBE,gnmw GtMs
The chemical etching was done in air by an etehant of NH~OH : H.,O z : H zO = 8 : 1.4 : 296. Ar ÷ ion etching was done at room temperature with ion acceleration voltage of 500 V and the ion dose of 1014 cm -2, The sulfur-treatment (S-treatmeat) [5-7] was dope by immersing the GaAs sample into an (NH4)2S IS: 6%) solution for several hours at room temperature after chemical etching. Extra sulfur on surface was removed by heating the surface at 150°g2 in the vacuum chamber. The altrathin Si ICL [10] was formed at T,ur, = 25D°C in the MBE chamber just ;,filer the GaAs growth using the Si Knudsen-ceE ,source and the background As~ pressure of I x 10 -x Tom For Schottk'y barrier formation, AI was evaporated from the AI K-¢¢1| in the MBE chamber whose background pressurewas below 1O-~ Torr, and AI, An and Mg were evaporated using W-coils in the deposition chamber whose backgruund pressure was below 10 -7 T o m The evaporation rate of AI and Au was about I ,g,/s. and Ihat of Mg was about 10 ,g./s. The XPg measurements were carried out using a Perkin-Elmer PHI-SI0(} spectrometer. The A I K a radiation (1486~6 eV) was used for the excitation. The surface and interface Fermi level positions (Ev,] were measured by XPS, I - V and C - V measurements. To obtain EFi from XPS data, the previously reported relationships [12,13] of Evi = E(;~3d(in G a A s ) - 1 8 . 8 0 eV and Eri=E/~,~d(in OaAs) - 40.79 eV were used. For the analysis of I - V data, the thermionie emission theory [14] was used taking account of the image force lowering [14]. The values of the Richardson constants used in this study were 8.2 A c m - - " K z for n-OaAs and 48 A c m -" K -z for p-GaAs, respectively. C - V data was analyzed using the ( l / C - ' ) - V plots.
3. Experimental results 3.1. X P S analysis o f blterfaee stnwture
The structure of the Schnttks, barrier interface was investigated by the in situ XPS analysis. The spectra of the Ga2p~/,_ and the AI2p core levels
Ga2pa/a
Al2p
]
(a) AI(TA)/MBE g r o w n GaAs
|
0~pa
~ps
( b ) A l ( l S . ~ ) / e h e m i e a l l y eLehed GaAsl ~,
C~aAS
~,c , m G~
~2°3
rneta~c AJ
GaAs meta~c ~ ~ _ S r~et~c
,to) A l ( a ~ ) / S - t r e a t e d
I5~
(d) Alta,~)/Si-ICL/GiAs met~,c t4-~s
......
1125
/
~etatr+z
. . . . . .
1120 1115 80 75 70 BINDING ENERGY (eV)
65
Fig. I. XPSspcclraof Ga2p~/-andAI2pcorelevehfromAI S¢lsatky interfaces formud on [a) an MBE clcan GaAs surface, (b) a chemicallyetched GaAs surface. (c] a sulfur-lreated Ga/~g surface and (d) a surface having an ultmtbln tin ,~) si ICL measured from the AI Schottky intcrfaces formed after various tl'oatmcnts are summarized in fig. 1. These spectra ~re separated into constituent components by the standard procedure based on the known chemical shift data. In the case of the AI/clean MBE GaAs interface shown in fig. In, the G a 2 p spectrum indicates presence of metallic Oa in addition to G a As bonding, and the AI2p spectrum indicates the presence of AI-As bonding in additiml to metallic AL respectively. Evidently an exchange reaction [15] takes place between AI and Ga and an AlAs layer is formed, The thickness of the AlAs layer is very thin being about 2 monolaycrs (ML). In the interface of the Al/chemically etched GaAs shown in fig. lb, the interface reaction seems to be quite different from that at the clean A I / M B E GaAs interface. Before metal deposition, the surface had a native oxide layer consisting of As203 and G a 2 0 ~ [15]. However, after
I£ Iimegawael aL / Fonnatiolt e?,~chani~nlof $chottky barriemon MBE-grown GuAs
metal deposition, only A I 2 0 ~ was detected at the interface, indicating that both Oa and As oxides were deoxidized by AI so as to form only AI orJde. The amounts of metallic Ga and As components in the interface region i t , ll were found to be very small, suggesting that Ga and As atoms produced by deoxidation by by AI formed GaAs. From the angle-resolved XPS measurements it was concluded that AIzO~ forms a continuous layer at the interface whose thickness was approximately l0 A. Fig. le shows the XPS spectrum of the Streated OaAs. In accordance with the literature [17-19] the S-treated GaAs surface had Ga sulfide and As sulfide components with no or very much reduced oxides before the metal deposition. After the AI deposition, the As sulfide component was not detected, and only GaAs and metallic Ga components were observed. This metallic Ga is most probably produced by the reduction of the Ga sulfide by AI. At the same time, a considerable amount of A l - S component was observed. The ;hickuess of the interfacial Ga and Al sulfide layer is estimated to be about 5 ,~. In the case of the barrier with a Si ICL of about 10 A. a well-defined Si interfaeial layer was observed [20] together with metallic Ga and A l As components shown in fig. ld. 3.2. Interface Fermi level positiou ( E r )
The positions of the interface Fermi level ( E R ) , were determined by the XPS band bending measurements for thin metal coverages (less than 50 ,~) and by l - V and C-V" measurements for thick metal coverages (about 1000 A). It has recently been pointed out that Evl measured by XPS is subjected to errors due to the surface photovoltaic effects [21,22]. Fig. 2 shows the measured core level energy shifts versos the X-ray radiation power for ion etched and MBEgrown free surfaces. The result shows existence of shifts due to the photovoltaie effect even at room temperature before metal deposition. This effect was much reduced after metal deposition. The present measurement was done keeping the X-ray radiation power in the range of 1-10 W to minimize possible error.
col ,01
Sl - ~ o o e o ~
~ r~
¢,a As ( o a ~ z )
t
~
~ ---o-
-a 1 -02 (
..... I
. . . . . . . .
)
~ ,
' :"ioo i
.
,
tO x -ray ~tce powerIW)
Fig. 2. Shnls of core luv¢l binding energy versus the X-ruy radiation power measured on some free surfaces belare metal delx)sitian.
The measured data oa Eiq are summarized in figs. 3a and 3b for various interfaces. The dashed line in fig. 3 shows the location of the hybrid orbital energy E u o which is a characteristic charge neutrality level in the DIGS model [4,1 I]. It lies at 0.47 eV above valence band maximum for GaAs [4]. For chemically etched surfaces and sulfur-treated (S-treated) surfaces data from the literature [6] are also included. As seen in figs. 3a and 3b, the EFi position of the AI Schottky barrier changes considerably depending on the surface treatments. The S-treatment reduces the SBH for the n-~ype GaAs while the ion etching and the insertion of a Si t e L increases it. When the barrier metal was changed, clear dependences of the SBH on the metal work-function Om were observed for chemically etched and S-treated surl~ces as seen in figs. 3a and 3b. For the work-function, data after Michaelsou [23] was used. The dependence can Ix', represented by the interface index S defined as S = d(SBH)/d~m.
(l)
With this definition, S = 0.2 was obtained for the barriers on the chemically etched surfaces and S = 0.3, for the barriers on the S-treated surfaces. An increase of S by the S-treatment is consistent with the results of other workers [6,7].
320
tl. Ilasegawa et at / Formation mechanism of $c~ottky barriersml MBE-grown GaAs
4. DIGS r c ~ e l including effects o f an interracial layer
[4.11] for the Fermi level pinning at interfaces. According to this model, random stress resulting from the inter[acial bonding mismatch brings back
Since all the: barriers studied here possess interracial layers (IL), their effects should be included in discussing the mechanism of the barrier formation. Hasegawa and O h n o proposed the unified d i ~ r d e r - i n d u c e d g a p state ( D I G S ) model
the hand states into the energy gap. resulting in a D I G S continuum with a continuous energy and spatial distribution. The D I G S continuum has a characteristic charge neutrality point, EHo. determined by the hybrid orbital energy in the
m
chetnc~y etched
MBE
p Ga~a o
I0~
I.( :>
£3
R
2
"0..~.
o~
~o~
a
i
i
E
~,
, o , ~,~, "3 4 5 6 u wo~kfunctio~6(eV) Schottky
ion etched 1015,-treated
~[Si :CL !
x 1~3dl o ,, ! iv,..,
~
,, ~.0
\o
d osL . . . . . . . . . . . . . . . ~\:_ Lr ~o S=(~3 o
E~ e lb ' - ' ~A~
:-C i, ,_0
~ , ~' workfunction~t~leV) e Schollky •~
~
Fig- 3. The measured data on the position of the interface Fermi level (En). (a) Balrier~ formed on MnE-grown clean surface and chemically etched (100) surfaces of GaAs. (b) Barriers formed on sulfur-treated, Ar + io:l etched, and Si ICL inserled (100) surfaces of GaAs.
321
H. Ha¢egawa el aL / Furmutiml mecgenism o[ Schottky barriers on MBE.grown GaAs
tight-binding context, When the DIGS density in high, EFi is pinned to EHO. However, the original theory tot the Schottky barrier formation [ I l] did not take account of the presence of the ILs. Based on the XPS measurements, one can think of two models for ILs, one being an ideal insulator-llke IL without gap states, and the other, a disordered semiconductor-like IL with DIGS. The models are shown in figs. 4a and 4b where t is the thickness of the IL and 6 is that of the disordered surface region of the host semiconductor. The semiconductor-like IL can in general po~ess shallow donors with the density, NDf and shallow aeceptors with the density, NAg,us shown in fig. 4h. For the present analysis, the energy and space distribution of the DIGS density is assumed, for simplicity, to be uniform and given by N[~; and N o u ,, respectively, as shown on the top of figs. 4a and 4b. By solving the Poisson's equation, the follow expression for n-type SBH, ~bn is obtained ~be = S ( ~ b m - X ~ )
+ (l -S)(Ec-EHo
) + .",~,
(z) where X~ is the electron affinity of the semicon-
ductor. For the insulator-like IL, ~.~ = O, and S is given by S
sech(6/A) tanh(~/A) '
I +qZ(t/e,)NooA
=~
,
(4)
where e~ is the permittivity of the semiconductor. For the disordered semiconductor-like IL. A~ is approximately given by
a4, = ( qlZe,)t2( N^, - Nn,),
× [I + {e,/A)(A,/e,) tanh(~/A) - 1,
x tanh(t/A,)]
(7)
A, = ~r~t/o2NDo , .
~GS densit y
VACUUM
,o~d
~,=............
i ~
donor.N.,~
[ i ionized . Ii~.:~0ZC-,S
E~ i iot~zed
iii oIGs ior, zed
6~
E~
E~o
:.'] :
acceptor. N,, ~'~r~.i"-; . . . . . . . . . . . . . . .
i il
(6)
where Ae is the DIGS screening length in the IL given by
........ VACUUM
is,
(5)
and S is given by S = seeh(6/A ) seeh(t/A~)
OIGS,denSily
/////////r/77~
(3)
where et is the permittivity of the IL and A is the DIGS screening length given by the following equation.
t.¢ ~
i !
Ev
6.~,
; i [ (a) (b) Fig. 4. DIGS model for Schouky harrier formation including Ihe interfacia] layer (11.) (a) A thin insulator-like ]L and (h) a disordered ~cmiconduclor-like ] L
322
H. Has,sawn er at / Formation mechanism of Scln~ttky barriers~ m MBE-gnJwn GaAs
1
....
-
-
~m.. 0eL
- - 6o,OX i \~ ~:,\\V,, '",
¢,=10.9¢0 !
N~s{--~,.#) (cra'~.eV~) Fig, 5. The calculated interface index $ venus interface state density N,is = N~i& based nn the model shown in fig.4a. t is Ihu IL thickness and a is Ihat of the disordered ~,emiconduufor surface layer.
A~ is due to the net ionized charge of shallow impurities in the I L Therefore. the SBH shift due to this effect is ultimately limited by the band edges of the IL or by those of the host semiconductor. Calculated values of S are shown in fig. 5 for the insulator-like IL. It is seen that S is very much dependent on the detailed nature of the I L Obviously the same applies to the semiconductor-like IL with an increased number of parameters.
$. Ilisenssion The present XPS results show that all the Schottky barriers possess interracial layers ( I L s ) with various chemical compositions and thicknesses of 5 - 1 0 ~ . The EFi measurements show that their SBH values are more or less similar, being far from those in the ideal Schottky limit. However, they are not in the firmly pinned Bardeen limit, showing clear dependences on the properties of the IL and the metal work-function.
Judging from the chemical compositions of the ILs and the electrical behavior of the correspond. ins bulk materials, ILs studied here are most likely either scmieonducting (AlAs, Si ICL) or insulating (AIzOs). The exact nature of the IL in the S-treated interface is not known yet, but dominant presence of the AI-S bonding suggests its semicondueting character. In this connection, it should be noted that most of the metallic Ga component in fig. 1 comes from the top AI layer, and not from the interface region itself, as indicated from the metal coverage dependences of various core-level signals. In any case, it is highly unlikely that all these thin lLs possess metallic character without energy gaps. Furthemlore, even for a thin metal layer, there is some indication based on the STM study that an energy gap opens up if the layer is ordered with respect to the semiconductor underneath [24]. Thus. to understand the present resuits in terms of the metal-induced gap state (MIGS) model [2,25,26], one has to assume that the exponential tail of the metal wavefunction penetrates through the IL and reaches the semiconductor underneath sufficiently. However, t,.te Imnetrafior depth of the MIGS defined in terms of the exponential decay length of the charge density, is estimated to be only about 3 A for GaAs and St, and much smaller for materials with wider energy gaps [25]. Thus, it is extremely difficult to understand the present results consistently in terms of the MiGS model. It is also difficult to understand the present results in terms of the effective work-function (EWF) model [3] or in terms of the advanced defect model [27], because of absence of significant amount of elemental As in the ILs. According to these models, As should determine an effective work-functien or control the density of the As antisite donors. On the other hand, the observed variation of SBH can be explained consistently by the present phenomenological theory taking account of the presence of the ILs. An upward shift of EFt of the AI barrier on the clean MBE surface can be explained by the formation of an ordered epitaxial A i / A I A s / G a A s structure with an AlAs IL having a small DIGS density. The behavior of
H. Hasegawa el aL / Fo~ation mechmd~n of Schottky barriers on MBE-gr~n GOAS
SBH on chemically etched surfaces with S = 0.2 can be explained well by the presence of an insulator-like IL with a thickness of about 10/~ and the semiconductor surface state density (Nss = N u n ) of about 1Oj4 em -z eV - t according to fig. 5. The strong tendency towards a firm pinning at Erie, seen in fig. 3b, which is caused by Ar + ion etching, is obviously consistent with the DIGS model. SBH on the S-treated and Si ICL inserted barriers can be explained by the presence of semiconductor-like ILs, although a more detailed study is necessary to determine which parameter of the IL is dominant to cause the observed behavior. Clearly, the usual argument of reduced Nss or "unpinning" of Fermi level after the Streatment cannot be justified at all from the present theory. The reduced Nss which will cause unpinning, should be at least of the order of 1012-1013 cm -2 eV -I. As seen in fig. 5, this should then to lead to a much larger value of S than the observed value of S = 0.3. The treatment may cause some reduction of Nss, but it still lies in the range of high 10 ~3 to 1014 cm 2 eV i. It should also be noted that the observed value may well be explained by very probable values of other parameters in the theory. From the viewpoint of the SBH control, doping of shallow impurities to the Si ICLs reported by Waldrop and Grant [8] offers an attractive possibi!ity. The increased SBH in fig. 3 may indeed be due to this effect caused by ample Ga doping which is consistent with the XPS result in fig. 1. According to eq. (3), the doping level should at least be of the order of 10 20 cm -~ to cause shifts of SBH of a few hundred meV.
6. Conclusion To clarify the mechanism of Schottky barrier formation, barriers were formed on MBE-grown (100) GaAs surfaces subjected various treatments, and characterized by XPS, I - V and C - V measurements. The main conclusions are listed below: (1) All the barriers including the A l / e l e a n MBE GaAs were found to posses interfacial lay-
323
ers (IL) of different chemical compositions and thicknesses of 5 - 1 0 ~. (2) All the barriers showed pinned behavior being far from the ideal Sehottk3, limit. But, they are not in the firmly pinned Bardeen limit either, and t h e S c h o t t k y b a r r i e r h e i g h t s showed c l e a r dependences on the properties of the IL and the metal work-function. S = 0.2 was obtained for chemically etched surfaces, and S = 0.3, for sulfur-treated surfaces, (3) A phcnomenological theory for SBH based on the DIGS model was developed which includes the effects of the ILs. (4) The observed behavior of SBH can be explained semi-quantitatively by the theory.
Acknowledgement The present work is supported partially by a Grant-in-Aid on the Priority Area of M e t a l Semiconductor Interfaces (No. 02232103) from the Ministry of Education, Science and Culture.
References I l l W.E. Spieer, P.W. Chye, P.R, Sk~ath, C.Y. Su and I. Lindau, J. Vac. Sci, TechnoL 16 (1979) 1422. [2J V, Heine, Phys, Rev. 138 (1965l 1689, [3] J.M, Woodall and J,F. Freeouf, J, Vac. Sci. Technol, 19 (1981)794.
[4] H. Hasegawaand H. Ohno. J. Vac. Sci. Technol. B 4 (1986) 1130. [5] CJ. Sandroff, R.N. Nonenbprg, LC. Bisehoff and R. Bhal, Appl. Phys. Lcn, 51 (1987)33. [6] J. Fan, Y. Kurata and Y. Nannichi,Jpn. J. Appl. Phys. 28 0989) L2255. [7] M.S, Carpenter, M.R. Melloch,B.A. Cowans, Z. Dardas and W.N. Dclgass,J. Vac. ScL TechnoL B 7 (1989)845. [8] J.R. Waldrop and R.W. Grant, J. Vac. Sci. Technol. n 6 (1988) 1432. [9] J.C. Costa. F. Willlamson,TJ. Miller, K. l~yzavi, M.L Nathan, D.S.L Mui, S. Strite and H. Morkoc, Appl. Phys. LeII, S8 (1991)382. [10] H. Hasegawa,M. Akazawa,K. Malsu;,~ak,H. Ishii and H. Ohno, Jpn. J. Appl. Phys. 27 (1988)1.2205. [I I] H. Hasegawa,in: Proc- ISlh Int. Conf. on the Physi~¢of Semicondt~ctors, Vol. 1 (1986)p. 291. [12] E.A. Kraut, Phys. Roy. n 28 (1983) 1965. [13] R.W. Grant, J.R. Wa[drop, S.P. Kowalczyk and E.A. Kraut, J. Vac. Sci. Technol. 19 (1981)477.
324
H. HasegaavJel al. / kimnatum tmg'hani~m of Schotdo, barriers oil MBE-gmn~I GaA~
[14] S.M. Sze, Physics of Semiconductor Dt:vicgs, 2nd ed. [Wil~3', New York, 198l I pp. 245. [15] R.Z, Bachrach, in: MelaI-Semiconductor Scholtky Barrier Juncfons and Their Applications. Ed. B,L+ Sharma (Plenum. New York, 19821oh. 2. p, IO3. [16] G,P, Schwahz, GJ. Gualtlcrl. G.W. KamnlelOt! and B. Schwahz, J. Elecffachem. Soci 126 (19791 1737. 117] ILM. G¢ih, J, Shin and C.W. Wilmsen+ J+ Vac. Sci. Technol+ B 8 11990) 838. [18] C J+ Sandroff. M.S. Hcgde, L.A. Farrow. C,C, Chang and J.p. Halrbison, Appl, Phys. Letl, 54 (19891 362. [19] C.J. Splndt, D. LUi, K, Miyano, P. Mcis,~ncf. T,T, Chiang, 1"+ Kendelewi~, I. Lindau and W+E. Spicer, Appl, Phys. Lett. 55 (198q1 8fJ3.
[20] H. Haseguw~, M. Akazawa. H, lshii and K+ Malsuzaki, J. Vac. Scl. Teehnol+ B 7 (19891 870. [21] M.H. ilecht, J. Vac. Sci. Teehnol. B 8 (19~0) 1018. [22] M. Alonso, R, Cimino, Ch. Maierhofer, Th. Chass~, W. Braun and K. Horn, J. Vac. Sci, Technol. B 8 (19901 955. 123] H.B. Michaelson, IBM J. Res. Dev. 22 (19781 72. [24] R.M Feenstra, Appl, Sur£ Sci. 56-58 (It~2) 104, [25] M.L Cohen. Electrons at Inlerface. in: Adwances in Electronics and Eleclron Physics, VoL 51 (Academic Press, New york, 19801 p. 21. [26] J. Tersoff. Phys. Rev. B 32 (198516968, [27] W.E. Spicer, T. Kendelewicz, N. Newman, R. Cap. C. McCanls, K. Miyano, L Lindau. Z. LilienlaI-Weber and E,R. Weber, AppL SurL Sci. 33/34 (19881 1009.
appeal
surface ~3ience
Formation mechanism of Schottky barriers on MBE-grown GaAs surfaces subjected to various treatments Hideki Hasegawa, Hirotatsu Ishii and Ken-ichi K0yanagi Deparrme/ar of Elcclrical En~kJeering and Reseereh Cenler for Interface Qumtgml Electronics, Hokkaido Umrersiry,. Sapporo 060, Japan Received fi May 1991:accepted for puhllcaUon 12 July 1991
AI. MS and Au Scholtk'j barriers were forme,~ on MBE-grown GaAs(100) surfaces subjected lo various treatments including cbamical etching, ion etching, sulfur treatmcnl and insertion of an uhralhin $i interface control layer (Si-ICL), They were characterized by the X-ray plmmelectron spectroscopy (XPS), current-voltage (l-V) and capacilance-vuhagc (C-V) techniques, The XPS lesults indicated the presence el an interfacial layer (ILl in each case. Ahhougb Ihc bchavinr is far from Ihe ideal Schouky limil, the barrier height showed dependences on the properties of ILs and the metal work-ttmction. Based on the disorder-lnduced saP state (DIGS) model, a theory including Ihe effect or an insulator-like or sumiconductor-lik¢ IL was developed. Tba theory explains the observed behavior rca~mably well. showing that the detailed nature of the IL is the key feature for the understanding and control of Scholtky barriers.
I. Introduction In spite of a long history of research which has led to numerous models [1-4], the formation mechanism of the Sghottky barriers has not been amply clarified yet. Recently, artificial changes of Schottky barrier heights (SBH) over a wide range have been achieved by sulfur treatments [5-7], insertion of Si interlayers [8,9], ctc., whose mechanism, however, still is not clear. T h e present study was carried out to identify the dominant mechanism determining the SBH. AI, M g and An Schottky barrier were formed on MBE grown (100) G a A s surfaces subjected to various treatments including chemical etching, ion etching, sulfur treatment and insertion of an ultrathin Si inter layer or interface control layer (ICL) [10]. T h e X-ray photoelectron spectroscopy (XPS), current-voltage ( l - V ) and capacitancevoltage ( C - V ) techniques were used for characterization. The X P S measurements indicated the presence of interracial layers (ILl with various compositions in all cases including the ease of the bar-
tier on the clean M B E GaAs. $BH were found to depend on the properties of II..s and the metal work-function. A phenomenological theory of Schottky barrier formation based on the unified disorder-induced gap state ( D I G S ) model [4,11] was then L.eveloped where the presence o f insulator-like or semiconductor-like IL is taken into aCCount, it is concluded that the observed behavior of 8 B H can he explained reasonably well by the theory and that the detailed nature of the IL is the key feature for understanding and control of Schottky barriers.
2. Experimental The experiment was done in a system having the MBE, deposition, XPS, and electrical characterization chambers, all connected by a U H V transfer chamber. Clean G a A s surfaces were prepared by the conventional M B E technique. The surface exhibited a clean As-stabilized (2 × 4) reconstruction pattern at the growth temperature on the substrate (Tsuh) of 580°C.
0169.4332/92/$05.00 © 1992 - Elsevier Science Publishers B.V. All rights reserved
318
I~ Hulegawael aL / FomuJli~mnu~'hm#smof Schoaky tm~iers m~ MBE,gnmw GtMs
The chemical etching was done in air by an etehant of NH~OH : H.,O z : H zO = 8 : 1.4 : 296. Ar ÷ ion etching was done at room temperature with ion acceleration voltage of 500 V and the ion dose of 1014 cm -2, The sulfur-treatment (S-treatmeat) [5-7] was dope by immersing the GaAs sample into an (NH4)2S IS: 6%) solution for several hours at room temperature after chemical etching. Extra sulfur on surface was removed by heating the surface at 150°g2 in the vacuum chamber. The altrathin Si ICL [10] was formed at T,ur, = 25D°C in the MBE chamber just ;,filer the GaAs growth using the Si Knudsen-ceE ,source and the background As~ pressure of I x 10 -x Tom For Schottk'y barrier formation, AI was evaporated from the AI K-¢¢1| in the MBE chamber whose background pressurewas below 1O-~ Torr, and AI, An and Mg were evaporated using W-coils in the deposition chamber whose backgruund pressure was below 10 -7 T o m The evaporation rate of AI and Au was about I ,g,/s. and Ihat of Mg was about 10 ,g./s. The XPg measurements were carried out using a Perkin-Elmer PHI-SI0(} spectrometer. The A I K a radiation (1486~6 eV) was used for the excitation. The surface and interface Fermi level positions (Ev,] were measured by XPS, I - V and C - V measurements. To obtain EFi from XPS data, the previously reported relationships [12,13] of Evi = E(;~3d(in G a A s ) - 1 8 . 8 0 eV and Eri=E/~,~d(in OaAs) - 40.79 eV were used. For the analysis of I - V data, the thermionie emission theory [14] was used taking account of the image force lowering [14]. The values of the Richardson constants used in this study were 8.2 A c m - - " K z for n-OaAs and 48 A c m -" K -z for p-GaAs, respectively. C - V data was analyzed using the ( l / C - ' ) - V plots.
3. Experimental results 3.1. X P S analysis o f blterfaee stnwture
The structure of the Schnttks, barrier interface was investigated by the in situ XPS analysis. The spectra of the Ga2p~/,_ and the AI2p core levels
Ga2pa/a
Al2p
]
(a) AI(TA)/MBE g r o w n GaAs
|
0~pa
~ps
( b ) A l ( l S . ~ ) / e h e m i e a l l y eLehed GaAsl ~,
C~aAS
~,c , m G~
~2°3
rneta~c AJ
GaAs meta~c ~ ~ _ S r~et~c
,to) A l ( a ~ ) / S - t r e a t e d
I5~
(d) Alta,~)/Si-ICL/GiAs met~,c t4-~s
......
1125
/
~etatr+z
. . . . . .
1120 1115 80 75 70 BINDING ENERGY (eV)
65
Fig. I. XPSspcclraof Ga2p~/-andAI2pcorelevehfromAI S¢lsatky interfaces formud on [a) an MBE clcan GaAs surface, (b) a chemicallyetched GaAs surface. (c] a sulfur-lreated Ga/~g surface and (d) a surface having an ultmtbln tin ,~) si ICL measured from the AI Schottky intcrfaces formed after various tl'oatmcnts are summarized in fig. 1. These spectra ~re separated into constituent components by the standard procedure based on the known chemical shift data. In the case of the AI/clean MBE GaAs interface shown in fig. In, the G a 2 p spectrum indicates presence of metallic Oa in addition to G a As bonding, and the AI2p spectrum indicates the presence of AI-As bonding in additiml to metallic AL respectively. Evidently an exchange reaction [15] takes place between AI and Ga and an AlAs layer is formed, The thickness of the AlAs layer is very thin being about 2 monolaycrs (ML). In the interface of the Al/chemically etched GaAs shown in fig. lb, the interface reaction seems to be quite different from that at the clean A I / M B E GaAs interface. Before metal deposition, the surface had a native oxide layer consisting of As203 and G a 2 0 ~ [15]. However, after
I£ Iimegawael aL / Fonnatiolt e?,~chani~nlof $chottky barriemon MBE-grown GuAs
metal deposition, only A I 2 0 ~ was detected at the interface, indicating that both Oa and As oxides were deoxidized by AI so as to form only AI orJde. The amounts of metallic Ga and As components in the interface region i t , ll were found to be very small, suggesting that Ga and As atoms produced by deoxidation by by AI formed GaAs. From the angle-resolved XPS measurements it was concluded that AIzO~ forms a continuous layer at the interface whose thickness was approximately l0 A. Fig. le shows the XPS spectrum of the Streated OaAs. In accordance with the literature [17-19] the S-treated GaAs surface had Ga sulfide and As sulfide components with no or very much reduced oxides before the metal deposition. After the AI deposition, the As sulfide component was not detected, and only GaAs and metallic Ga components were observed. This metallic Ga is most probably produced by the reduction of the Ga sulfide by AI. At the same time, a considerable amount of A l - S component was observed. The ;hickuess of the interfacial Ga and Al sulfide layer is estimated to be about 5 ,~. In the case of the barrier with a Si ICL of about 10 A. a well-defined Si interfaeial layer was observed [20] together with metallic Ga and A l As components shown in fig. ld. 3.2. Interface Fermi level positiou ( E r )
The positions of the interface Fermi level ( E R ) , were determined by the XPS band bending measurements for thin metal coverages (less than 50 ,~) and by l - V and C-V" measurements for thick metal coverages (about 1000 A). It has recently been pointed out that Evl measured by XPS is subjected to errors due to the surface photovoltaic effects [21,22]. Fig. 2 shows the measured core level energy shifts versos the X-ray radiation power for ion etched and MBEgrown free surfaces. The result shows existence of shifts due to the photovoltaie effect even at room temperature before metal deposition. This effect was much reduced after metal deposition. The present measurement was done keeping the X-ray radiation power in the range of 1-10 W to minimize possible error.
col ,01
Sl - ~ o o e o ~
~ r~
¢,a As ( o a ~ z )
t
~
~ ---o-
-a 1 -02 (
..... I
. . . . . . . .
)
~ ,
' :"ioo i
.
,
tO x -ray ~tce powerIW)
Fig. 2. Shnls of core luv¢l binding energy versus the X-ruy radiation power measured on some free surfaces belare metal delx)sitian.
The measured data oa Eiq are summarized in figs. 3a and 3b for various interfaces. The dashed line in fig. 3 shows the location of the hybrid orbital energy E u o which is a characteristic charge neutrality level in the DIGS model [4,1 I]. It lies at 0.47 eV above valence band maximum for GaAs [4]. For chemically etched surfaces and sulfur-treated (S-treated) surfaces data from the literature [6] are also included. As seen in figs. 3a and 3b, the EFi position of the AI Schottky barrier changes considerably depending on the surface treatments. The S-treatment reduces the SBH for the n-~ype GaAs while the ion etching and the insertion of a Si t e L increases it. When the barrier metal was changed, clear dependences of the SBH on the metal work-function Om were observed for chemically etched and S-treated surl~ces as seen in figs. 3a and 3b. For the work-function, data after Michaelsou [23] was used. The dependence can Ix', represented by the interface index S defined as S = d(SBH)/d~m.
(l)
With this definition, S = 0.2 was obtained for the barriers on the chemically etched surfaces and S = 0.3, for the barriers on the S-treated surfaces. An increase of S by the S-treatment is consistent with the results of other workers [6,7].
320
tl. Ilasegawa et at / Formation mechanism of $c~ottky barriersml MBE-grown GaAs
4. DIGS r c ~ e l including effects o f an interracial layer
[4.11] for the Fermi level pinning at interfaces. According to this model, random stress resulting from the inter[acial bonding mismatch brings back
Since all the: barriers studied here possess interracial layers (IL), their effects should be included in discussing the mechanism of the barrier formation. Hasegawa and O h n o proposed the unified d i ~ r d e r - i n d u c e d g a p state ( D I G S ) model
the hand states into the energy gap. resulting in a D I G S continuum with a continuous energy and spatial distribution. The D I G S continuum has a characteristic charge neutrality point, EHo. determined by the hybrid orbital energy in the
m
chetnc~y etched
MBE
p Ga~a o
I0~
I.( :>
£3
R
2
"0..~.
o~
~o~
a
i
i
E
~,
, o , ~,~, "3 4 5 6 u wo~kfunctio~6(eV) Schottky
ion etched 1015,-treated
~[Si :CL !
x 1~3dl o ,, ! iv,..,
~
,, ~.0
\o
d osL . . . . . . . . . . . . . . . ~\:_ Lr ~o S=(~3 o
E~ e lb ' - ' ~A~
:-C i, ,_0
~ , ~' workfunction~t~leV) e Schollky •~
~
Fig- 3. The measured data on the position of the interface Fermi level (En). (a) Balrier~ formed on MnE-grown clean surface and chemically etched (100) surfaces of GaAs. (b) Barriers formed on sulfur-treated, Ar + io:l etched, and Si ICL inserled (100) surfaces of GaAs.
321
H. Ha¢egawa el aL / Furmutiml mecgenism o[ Schottky barriers on MBE.grown GaAs
tight-binding context, When the DIGS density in high, EFi is pinned to EHO. However, the original theory tot the Schottky barrier formation [ I l] did not take account of the presence of the ILs. Based on the XPS measurements, one can think of two models for ILs, one being an ideal insulator-llke IL without gap states, and the other, a disordered semiconductor-like IL with DIGS. The models are shown in figs. 4a and 4b where t is the thickness of the IL and 6 is that of the disordered surface region of the host semiconductor. The semiconductor-like IL can in general po~ess shallow donors with the density, NDf and shallow aeceptors with the density, NAg,us shown in fig. 4h. For the present analysis, the energy and space distribution of the DIGS density is assumed, for simplicity, to be uniform and given by N[~; and N o u ,, respectively, as shown on the top of figs. 4a and 4b. By solving the Poisson's equation, the follow expression for n-type SBH, ~bn is obtained ~be = S ( ~ b m - X ~ )
+ (l -S)(Ec-EHo
) + .",~,
(z) where X~ is the electron affinity of the semicon-
ductor. For the insulator-like IL, ~.~ = O, and S is given by S
sech(6/A) tanh(~/A) '
I +qZ(t/e,)NooA
=~
,
(4)
where e~ is the permittivity of the semiconductor. For the disordered semiconductor-like IL. A~ is approximately given by
a4, = ( qlZe,)t2( N^, - Nn,),
× [I + {e,/A)(A,/e,) tanh(~/A) - 1,
x tanh(t/A,)]
(7)
A, = ~r~t/o2NDo , .
~GS densit y
VACUUM
,o~d
~,=............
i ~
donor.N.,~
[ i ionized . Ii~.:~0ZC-,S
E~ i iot~zed
iii oIGs ior, zed
6~
E~
E~o
:.'] :
acceptor. N,, ~'~r~.i"-; . . . . . . . . . . . . . . .
i il
(6)
where Ae is the DIGS screening length in the IL given by
........ VACUUM
is,
(5)
and S is given by S = seeh(6/A ) seeh(t/A~)
OIGS,denSily
/////////r/77~
(3)
where et is the permittivity of the IL and A is the DIGS screening length given by the following equation.
t.¢ ~
i !
Ev
6.~,
; i [ (a) (b) Fig. 4. DIGS model for Schouky harrier formation including Ihe interfacia] layer (11.) (a) A thin insulator-like ]L and (h) a disordered ~cmiconduclor-like ] L
322
H. Has,sawn er at / Formation mechanism of Scln~ttky barriers~ m MBE-gnJwn GaAs
1
....
-
-
~m.. 0eL
- - 6o,OX i \~ ~:,\\V,, '",
¢,=10.9¢0 !
N~s{--~,.#) (cra'~.eV~) Fig, 5. The calculated interface index $ venus interface state density N,is = N~i& based nn the model shown in fig.4a. t is Ihu IL thickness and a is Ihat of the disordered ~,emiconduufor surface layer.
A~ is due to the net ionized charge of shallow impurities in the I L Therefore. the SBH shift due to this effect is ultimately limited by the band edges of the IL or by those of the host semiconductor. Calculated values of S are shown in fig. 5 for the insulator-like IL. It is seen that S is very much dependent on the detailed nature of the I L Obviously the same applies to the semiconductor-like IL with an increased number of parameters.
$. Ilisenssion The present XPS results show that all the Schottky barriers possess interracial layers ( I L s ) with various chemical compositions and thicknesses of 5 - 1 0 ~ . The EFi measurements show that their SBH values are more or less similar, being far from those in the ideal Schottky limit. However, they are not in the firmly pinned Bardeen limit, showing clear dependences on the properties of the IL and the metal work-function.
Judging from the chemical compositions of the ILs and the electrical behavior of the correspond. ins bulk materials, ILs studied here are most likely either scmieonducting (AlAs, Si ICL) or insulating (AIzOs). The exact nature of the IL in the S-treated interface is not known yet, but dominant presence of the AI-S bonding suggests its semicondueting character. In this connection, it should be noted that most of the metallic Ga component in fig. 1 comes from the top AI layer, and not from the interface region itself, as indicated from the metal coverage dependences of various core-level signals. In any case, it is highly unlikely that all these thin lLs possess metallic character without energy gaps. Furthemlore, even for a thin metal layer, there is some indication based on the STM study that an energy gap opens up if the layer is ordered with respect to the semiconductor underneath [24]. Thus. to understand the present resuits in terms of the metal-induced gap state (MIGS) model [2,25,26], one has to assume that the exponential tail of the metal wavefunction penetrates through the IL and reaches the semiconductor underneath sufficiently. However, t,.te Imnetrafior depth of the MIGS defined in terms of the exponential decay length of the charge density, is estimated to be only about 3 A for GaAs and St, and much smaller for materials with wider energy gaps [25]. Thus, it is extremely difficult to understand the present results consistently in terms of the MiGS model. It is also difficult to understand the present results in terms of the effective work-function (EWF) model [3] or in terms of the advanced defect model [27], because of absence of significant amount of elemental As in the ILs. According to these models, As should determine an effective work-functien or control the density of the As antisite donors. On the other hand, the observed variation of SBH can be explained consistently by the present phenomenological theory taking account of the presence of the ILs. An upward shift of EFt of the AI barrier on the clean MBE surface can be explained by the formation of an ordered epitaxial A i / A I A s / G a A s structure with an AlAs IL having a small DIGS density. The behavior of
H. Hasegawa el aL / Fo~ation mechmd~n of Schottky barriers on MBE-gr~n GOAS
SBH on chemically etched surfaces with S = 0.2 can be explained well by the presence of an insulator-like IL with a thickness of about 10/~ and the semiconductor surface state density (Nss = N u n ) of about 1Oj4 em -z eV - t according to fig. 5. The strong tendency towards a firm pinning at Erie, seen in fig. 3b, which is caused by Ar + ion etching, is obviously consistent with the DIGS model. SBH on the S-treated and Si ICL inserted barriers can be explained by the presence of semiconductor-like ILs, although a more detailed study is necessary to determine which parameter of the IL is dominant to cause the observed behavior. Clearly, the usual argument of reduced Nss or "unpinning" of Fermi level after the Streatment cannot be justified at all from the present theory. The reduced Nss which will cause unpinning, should be at least of the order of 1012-1013 cm -2 eV -I. As seen in fig. 5, this should then to lead to a much larger value of S than the observed value of S = 0.3. The treatment may cause some reduction of Nss, but it still lies in the range of high 10 ~3 to 1014 cm 2 eV i. It should also be noted that the observed value may well be explained by very probable values of other parameters in the theory. From the viewpoint of the SBH control, doping of shallow impurities to the Si ICLs reported by Waldrop and Grant [8] offers an attractive possibi!ity. The increased SBH in fig. 3 may indeed be due to this effect caused by ample Ga doping which is consistent with the XPS result in fig. 1. According to eq. (3), the doping level should at least be of the order of 10 20 cm -~ to cause shifts of SBH of a few hundred meV.
6. Conclusion To clarify the mechanism of Schottky barrier formation, barriers were formed on MBE-grown (100) GaAs surfaces subjected various treatments, and characterized by XPS, I - V and C - V measurements. The main conclusions are listed below: (1) All the barriers including the A l / e l e a n MBE GaAs were found to posses interfacial lay-
323
ers (IL) of different chemical compositions and thicknesses of 5 - 1 0 ~. (2) All the barriers showed pinned behavior being far from the ideal Sehottk3, limit. But, they are not in the firmly pinned Bardeen limit either, and t h e S c h o t t k y b a r r i e r h e i g h t s showed c l e a r dependences on the properties of the IL and the metal work-function. S = 0.2 was obtained for chemically etched surfaces, and S = 0.3, for sulfur-treated surfaces, (3) A phcnomenological theory for SBH based on the DIGS model was developed which includes the effects of the ILs. (4) The observed behavior of SBH can be explained semi-quantitatively by the theory.
Acknowledgement The present work is supported partially by a Grant-in-Aid on the Priority Area of M e t a l Semiconductor Interfaces (No. 02232103) from the Ministry of Education, Science and Culture.
References I l l W.E. Spieer, P.W. Chye, P.R, Sk~ath, C.Y. Su and I. Lindau, J. Vac. Sci, TechnoL 16 (1979) 1422. [2J V, Heine, Phys, Rev. 138 (1965l 1689, [3] J.M, Woodall and J,F. Freeouf, J, Vac. Sci. Technol, 19 (1981)794.
[4] H. Hasegawaand H. Ohno. J. Vac. Sci. Technol. B 4 (1986) 1130. [5] CJ. Sandroff, R.N. Nonenbprg, LC. Bisehoff and R. Bhal, Appl. Phys. Lcn, 51 (1987)33. [6] J. Fan, Y. Kurata and Y. Nannichi,Jpn. J. Appl. Phys. 28 0989) L2255. [7] M.S, Carpenter, M.R. Melloch,B.A. Cowans, Z. Dardas and W.N. Dclgass,J. Vac. ScL TechnoL B 7 (1989)845. [8] J.R. Waldrop and R.W. Grant, J. Vac. Sci. Technol. n 6 (1988) 1432. [9] J.C. Costa. F. Willlamson,TJ. Miller, K. l~yzavi, M.L Nathan, D.S.L Mui, S. Strite and H. Morkoc, Appl. Phys. LeII, S8 (1991)382. [10] H. Hasegawa,M. Akazawa,K. Malsu;,~ak,H. Ishii and H. Ohno, Jpn. J. Appl. Phys. 27 (1988)1.2205. [I I] H. Hasegawa,in: Proc- ISlh Int. Conf. on the Physi~¢of Semicondt~ctors, Vol. 1 (1986)p. 291. [12] E.A. Kraut, Phys. Roy. n 28 (1983) 1965. [13] R.W. Grant, J.R. Wa[drop, S.P. Kowalczyk and E.A. Kraut, J. Vac. Sci. Technol. 19 (1981)477.
324
H. HasegaavJel al. / kimnatum tmg'hani~m of Schotdo, barriers oil MBE-gmn~I GaA~
[14] S.M. Sze, Physics of Semiconductor Dt:vicgs, 2nd ed. [Wil~3', New York, 198l I pp. 245. [15] R.Z, Bachrach, in: MelaI-Semiconductor Scholtky Barrier Juncfons and Their Applications. Ed. B,L+ Sharma (Plenum. New York, 19821oh. 2. p, IO3. [16] G,P, Schwahz, GJ. Gualtlcrl. G.W. KamnlelOt! and B. Schwahz, J. Elecffachem. Soci 126 (19791 1737. 117] ILM. G¢ih, J, Shin and C.W. Wilmsen+ J+ Vac. Sci. Technol+ B 8 11990) 838. [18] C J+ Sandroff. M.S. Hcgde, L.A. Farrow. C,C, Chang and J.p. Halrbison, Appl, Phys. Letl, 54 (19891 362. [19] C.J. Splndt, D. LUi, K, Miyano, P. Mcis,~ncf. T,T, Chiang, 1"+ Kendelewi~, I. Lindau and W+E. Spicer, Appl, Phys. Lett. 55 (198q1 8fJ3.
[20] H. Haseguw~, M. Akazawa. H, lshii and K+ Malsuzaki, J. Vac. Scl. Teehnol+ B 7 (19891 870. [21] M.H. ilecht, J. Vac. Sci. Teehnol. B 8 (19~0) 1018. [22] M. Alonso, R, Cimino, Ch. Maierhofer, Th. Chass~, W. Braun and K. Horn, J. Vac. Sci, Technol. B 8 (19901 955. 123] H.B. Michaelson, IBM J. Res. Dev. 22 (19781 72. [24] R.M Feenstra, Appl, Sur£ Sci. 56-58 (It~2) 104, [25] M.L Cohen. Electrons at Inlerface. in: Adwances in Electronics and Eleclron Physics, VoL 51 (Academic Press, New york, 19801 p. 21. [26] J. Tersoff. Phys. Rev. B 32 (198516968, [27] W.E. Spicer, T. Kendelewicz, N. Newman, R. Cap. C. McCanls, K. Miyano, L Lindau. Z. LilienlaI-Weber and E,R. Weber, AppL SurL Sci. 33/34 (19881 1009.