Ge contacts to n-type GaAs

Ge contacts to n-type GaAs

Solid-State Electronics Vol. 23, pp. 973-985 Pergamon Press Ltd., 1980. Printed in Great Britain Pd/Ge CONTACTS TO n-TYPE GaAs H. R. GRINOLDSt and G...

994KB Sizes 5 Downloads 31 Views

Solid-State Electronics Vol. 23, pp. 973-985 Pergamon Press Ltd., 1980. Printed in Great Britain

Pd/Ge CONTACTS TO n-TYPE GaAs H. R. GRINOLDSt and G. Y. ROBINSON Department of Electrical Engineering, University of Minnesota, Minneapolis, MN 55455, U.S.A.

(Received 20 October 1979; in revised[orm 9 January 1980) Abstract--Sintered metal-semiconductor contacts, formed by thin, evaporated layers of Pd and Ge on n-type GaAs, were studied using Auger electron spectroscopy, X-ray diffraction, X-ray photoelectron spectroscopy, secondary ion mass spectroscopy, current-voltage measurements, and capacitance-voltage measurements. Prior to sintering, the as-deposited Pd/Ge/GaAs contacts were rectifying and exhibited a reproducible Schottky barrier energy $Bn of 0.67-+0.02 eV. Auger analysis indicated the initial behavior of the contact structure, upon sintering, to be an interdiffusionand reaction of Pd and Ge on a non-reacting GaAs substrate. Two germanide phases, Pd2Ge and PdGe, were identified using X-ray diffraction and Auger analysis. The intervening Ge layer prevented the reaction of Pd with the GaAs substrate at low temperatures. Because of the Pd-Ge reaction, ~bBnincreased to approximately 0.85eV. Stntering at higher temperatures (i.e. between 300 and 400°C)produced additional reactions between Pd and the GaAs substrate. The electrical properties of the contact remained rectifying and 4,B, exhibited little change from the value of 0.85eV with the interdiffusionof Pd, Ga, and As. Sintering above 400°C resulted in the formation of ohmic contacts. The diffusion of Ge to the GaAs interface was found to correlate with the onset of ohmic behavior. Current condliction in the contact was best described by thermionic-fieldemission theory, and a specific contact resistance of 3.5 x 10-4 II-cm2 was obtained after sintering above 550°C, independent of the initial impurity concentration in the substrate. Over the entire range of sintering temperatures (i.e. at or below 600°C),the interaction between the thin-film layers appeared to be governed by diffusion-controlled, solid-phase processes with no evidence of the formation of a liquid phase. As a result, the surface of the contact structure remained smooth and uniform during sintering. I. INTRODUCTION

Sintering offers several distinct advantages over alloying for the formation of ohmic contacts to GaAs. Since the formation of alloyed contacts requires the presence of an intermediate liquid-phase, the amount of GaAs consumed and the properties of the regrown layer are highly sensitive to the process parameters (i.e. the amount of each material, the temperature and time of the alloy cycle, and the method of cooling) [I, 2]. The nonequilibrium solidification of the contact overlayers results in a rough surface with ohmic behavior that is difficult to reproduce [2, 3]. Diffusion of the contact materials in subsequent thermal processes (e.g. passivation) may adversely affect the reliability of the alloyed contact. For example, with the widely used Ni/Au-Ge alloyed contact, Ni readily diffuses through the Au-Ge layer to the surface of the GaAs at 350°C [3]. Since Ni is a relatively rapid diffuser in GaAs [4] and interaction between Ni and Ga and As has been reported [5], the alloyed Ni/Au-Ge contact may exhibit a degradation in ohmic behavior with additional processing. In addition to the above undesirable properties, the selection of a contact system with a low temperature eutectic sets an upper limit to the maximum temperature of device operation. For use with high frequency, planar devices, more sophisticated and improved methods are needed to achieve reliable and reproducible ohmic contacts to GaAs. Sintering relies on the formation of the contact by low temperature, solid-phase reactions. The introduction

tNow at Hewlett-Packard, 3500 Deer Creek Rd. Palo Alto, CA 94304, U.S.A.

of the dopant into the GaAs {vithout the necessity of a liquid phase and the possibility of the formation of stable compounds at the surface of the semiconductor potentially offers an increased stability, reliability, and reproducibility with sintered contacts that is not readily available with alloyed contacts. Sintered ohmic contacts to lightly-doped, n-type GaAs have been investigated for several specific systems. Sinha et al. [6] reported ohmic behavior for sintered Pd/Ge contacts with a specific contact resistance rc = 4 × l0 -4 fl-cm2 on substrates with a net impurity concentration IND--NAI of 1016cm-3. Ohmic contacts have been obtained with the Ni/Ge/GaAs structure where rc---3 × 10-5 fl-cm2 with No = 10~Tcm-~ [7]. Thus, the formation of ohmic contacts by sintering is a viable process that can result in relatively low specific contact resistance. However, reported models of the physical processes which govern the formation are not complete and fail to identify those parameters necessary for achieving an optimum ohmic contact [6,7]. The Pd/Ge/GaAs system is a good example. While neither Ge nor Pd alone form ohmic contacts to n-type GaAs (for INo- NAI < 10t7 cm-3), when used together they result in a contact that is ohmic [6]. From the reported work [6,7], the mechanisms responsible for the ohmic behavior are unresolved and the kinetic processes involved are unknown. It is apparent that the solid-phase reactions in the structure result in the formation of compounds, but the thin-film interactions were not studied in detail nor related to the electrical properties [6]. Similar thin-film interactions may be responsible for the ohmic behavior of the Ni/Ge/GaAs system. The object "of this research was to investigate, in detail, both the electrical and metallurgical properties of 973

H. R. {3RINOLDSand G. Y. ROBINSON

974

temperatures between 175 and 600°C in a flowing N2 atmosphere for times between 2 and 120 min.

the PdlGe contact to n-type GaAs., The relationship between the composition of the structure and the electrical properties after sintering was emphasized.

3. RESULTSANDDISCUSSION The behavior of Pd/Ge contacts to GaAs upon sintering will be discussed in three stages, each of which are depicted in Fig. 1. At each stage, the important aspects of the metallurgical and electrical behavior of the contact structure have been indicated and are shown to depend upon both the sintering temperature and the sintering time. Stage 1 is reached by sintering at low temperatures (i.e. below about 300°C). Samples sintered between 300 and 400°C exhibit the properties of Stage 2. Sintering at higher temperatures (i.e. from 400 to 600°C) or for prolonged times yields the characteristics of Stage 3.

2. EXPERIMENTALPROCEDURE

Several sets of samples were fabricated to provide optimum geometry for each of the characterization techniques. The electrical properties were evaluated with contacts defined in chemical-vapor-deposited (CVD) SiO2 with the active areas between 1 x 10-6cm2 and 1.58 × l0 -3 cm2. The largest contact was used for capacitance-voltage (C-V) measurements and also provided sufficient area for Auger analysis. Current-voltage (I-V) measurements were taken on all of the contact areas. X-ray analysis, X-ray photoelectron spectroscopy (XPS or ESCA), and secondary ion mass spectroscopy (SIMS) required the use of large area samples. The fabrication of these samples was identical to those utilized for the electrical evaluation except for the absence of pattern definition. In some cases, the thickness of the Pd and Ge thin films was varied. The substrates used in this study were polished, single crystal, (100) n-type GaAs either uniformly-doped or with an n-type epitaxial layer. An alloyed NilAu-Ge ohmic contact was formed on the backside of the substrate to provide electrical contact. For the defined contact areas, the CVD SiO2 was patterned using standard photolithographic techniques. Prior to the deposition of Pd and Ge, the surface of the GaAs was etched in 20:1 Hz0:HCI for 15 sec. to remove any native oxides and then in a GaAs polish etch (H202:NH4OH:H20-7:20:!000) for 45sec. Thin films of Ge and Pd were deposited sequentially by electron-beam evaporation at a chamber pressure of 5 x 10 -7 Torr onto the unheated GaAs substrates. The Pd and Ge thin films were defined by either standard photolithography or by a photoresist liftoff technique. After characterization of the samples in the as-deposited state, the samples were sintered at STAGE I

Pd - Ge interaction

Stage 1 The metallurgical behavior of Pd/Ge contacts upon sintering at low temperatures was investigated using AES and X-ray diffraction. The electrical properties of the contact were characterized by measurement of the Schottky barrier energy. Since we have previously reported the contact behavior for Stage 1, only a brief review will be given here [8]. In as-deposited samples and upon sintering those samples at Stage 1, Pd reacted with Ge to form two germanide phases, Pd2Ge and PdGe. The phases were identified by comparison of the observed peaks in the X-ray diffraction spectra (Cu K~ radiation) with the appropriate powder X-ray diffraction cards [9]. Both Pd2Ge and PdGe were found to be polycrystalline. In as-deposited samples, peaks at 20 = 47.2°, 54.4°, 62° were identified as Pd2Ge and peaks at 20 = 50.8°, 51.6°, 52.6°, 63.4° were identified as PdGe. One peak at 20 = 77.20 was associated with 'Pd2Ge and/or PdGe. Both the intensity and the number of Bragg peaks increased with sintering time [8]. Auger analysis was used to observe the progression of the reaction between Pd and Ge by II

STAGE

Pd

- Ga

and

Pd

STAGE III 'Ge - GaAs interaction

- As

interaction

Formation of Germanide compounds (Fd2Ge and PdGe)

Germanide redistribution

No dissociation of

Ga and As diffusion out to the s u r f a c e

GaAs

Ge

doping of G a A s

Ge and As D i f f u s i o n out to the s u r f a c e

R e c t i f y i n g Contacts (increase in CBo)

Rectifying C o n t a c t s (little change in ~Bo)

Ohmic C o n t a c t Formation

Smooth contact s u r f a c e

Smooth C o n t a c t Surface

Smooth contact s u r f a c e

SINTERING

TEMPERATURE

OR T I M E

Fig. 1. The three stages by whichthe behaviorof Pd/Oe contactsto GaAschangesas the result of sintering. For each stage, the major aspects of the metallurgicaland electrical behavior are shown.

975

Pd/Ge contacts to n-type GaAs identification of an Auger spectrum associated with Ge chemically bound to Pd. In Fig. 2, Auger spectra in the energy range of the Ge LMM transition (1100--1150eV) are shown for elemental Ge (Fig. 2a) and Ge chemically bound to Pd in the form of Pd2Ge or PdGe (Fig. 2b). Auger depth profiling of Pd/Ge GaAs structures indicated that the subpeaks at 1114 and ll30eV were absent near the Pd/Ge interface in as-deposited samples and over a greater depth range in sintered samples. Based on the identification of Pd2Ge and PdGe by X-ray diffraction, the hange in the Ge LMM spectra was assumed to result from the germanide formation[8]. However, it was not possible to distinguish between Pd2Ge and PdGe using only the characteristic spectra shown in Fig. 2(b). The inability to differentiate the two germanide phases by an alteration of the Auger spectrum may be due to the limited energy resolution of the energy analyzer (AE/E = 0.6%) [10] or the effects of the sputter-ion beam on the analyzed surface [11]. The subpeaks, at 16 and 32eV below the main Ge transition, have been previously identified as plasmon4oss peaks in covalently-bonded Ge [12]. The sensitivity of plasmonloss peaks to the chemical environment of the atom has been used in the study of oxides on GaAs [13, 14]. Thus, the changes in the Ge spectra displayed in Fig. 2 are attributed to shifts in the plasmon-loss peaks of the Ge LMM transition and allow chemical information to be obtained from the Auger analysis. In agreement withthe results of X-ray diffraction, the data from the Auger analyses indicated a reaction between Pd and Ge upon deposition of the Pd onto the Ge thin-film layer [8]. The kinetics of the formation of Pd2Ge has been reported to be approximately eight times faster than those reported for the formation of Pd2Si [15-17]. Since Pd2Si has been reported to form upon vacuum deposition of Pd on Si substrates [17], it is quite possible to have the formation of Pd2Ge and PdGe during vacuum deposition of Pd onto a Ge layer.

With sintering, the conversion of Ge into the germanides continued until all the elemental Ge had reacted with Pd. In Fig. 3, Auger depth profile of a sample sintered at 200°C for 20 min is shown. Observation of the Ge Auger spectrum indicated that the subpeaks were absent throughout the depth profile, thus indicating complete conversion of Ge into Pd2Ge and PdGe. It should be mentioned that due to overlap in energy between the Ge and As LMM spectra [18], accurate determination of the germanide distribution near the GaAs interface was not possible. The oxygen level, shown in Fig. 3 at the germanide/GaAs interface, was near the detection limits of the Auger instrumentation. The presence of the oxygen was observed in only a few samples and could not be correlated reproducibly with either the electrical properties or the film composition. The thin-film interactions at Stage 1 were found to be only that between Pd and Ge. No evidence of the dissociation of the GaAs was detected in the Auger analysis and the X-ray diffraction data gave no indication of the formation of any Pd-As or Pd-Ga compounds. Since Pd has been found to readily react with both Ga and As upon sintering between 250 and 500°C [19], it was concluded that the presence of unreacted Ge in the Pd/Ge/GaAs structure prevents the reaction of Pd with GaAs [8]. Electrical measurements were performed on Pd/Ge diodes fabricated on LPE, n/n +(100) GaAs substrates with a net carrier concentration in the epitaxial layer of 1.1 x 10'6cm-3 (Sn doped). The thickness of the Pd and Ge layers was ll00 and 370,~, respectively. Prior to sintering, the contacts exhibited rectifying behavior with forward-bias characteristics displaying an exponential relationship between voltage and current for over seven decades of forward current. The Schottky barrier energy

b)

o)

SINTERED

Ge (PdGe-Pd Ge)

Ge (Ge)

100

~

\

75

I'

200*C/20rain.

dN

.IV / t,~

~ N

,o42"/I,o93 1053 / IO73

As

1042 [

25

1054

113

1176

1177

5~3 ll46 i

I000

T

i

i

II00 1200

I

I000

I

,

100

150

200

250

ION DOSE

1146 i

II00 i200

Auger Electron Energy (eV)

Fig. 2. The Oe Auger spectrum in the energy range of the LMM transition for (a) a reference sample of elemental Ge and for as-deposited samples near the Ge/GaAs interface and for (b) as-deposited and sintered samples near the Pd/Ge interface.

Fig. 3. An Augerdepth profile of a sample with layers of Pd and Ge i.I kA and 370A thick, respectively, after sintering at 200°C for 20 min. The energyof the analyzingelectron beam was 5 keV and the samples were profiled in situ with 2-keV Ar ions. The Auger peak-to-peak amplitude was normalizedusing the element sensitivites and procedure published in Ref.[18]. Ion dose (in

arbitrary units) was determined by integration of the ion beam current over the sputtering time.

H.R. GRINOLDSand G. Y. ROBINSON

976

09B, (I-V) was 0.67 -+0.02 eV with a corresponding diode factor n of 1.03+0.01. The value of Ss, (I-V) was calculated using thermionic emission theory [20]. The Schottky barrier energy of the Pd/Ge/GaAs contacts is considerably lower than that for most metals on lightly-doped, n-type GaAs (e.g. for Au/GaAs, CB, (IV) = 0.90 -+0.04 eV) [2,21,22]. Pruniaux [23] has also reported that Au/Ge contacts to lightly-doped, n-type GaAs also exhibit barriers approximately 0.16 eV lower than expected. A physical model which can be used to predict the barrier energy of a thin, evaporated Ge film on GaAs has yet to be formulated. Such a model would have to take into account the structure and conductivity of the Ge film, the role of Ge as a dopant in GaAs, and the effect of strain and electronic states at the interface. The electrical properties of the contact structure were found to vary upon sintering. The variation in ~b~. (I-V) as a function of sintering time at 175, 200 and 300°C has been plotted in Fig. 4. As shown, there is a significant increase in &B,(I-V) until a maximum is reached where 4~B,(I-V) = 0.85 eV. After sintering at 300°C, Ca,(I-V) is shown to remain relatively constant. Auger analysis and the change in the Ge LMM spectra were used to estimate the amount of elemental Ge left in each diode structure after sintering. The analysis indicated that the Ge thin film had completely reacted to form Pd2Ge and PdGe for diodes which exhibited CB,(I-V)=0.85eV. Thus, the formation of a germanide layer is responsible for the increase in SB,(I-V) shown in Fig. 14 and the value of &~,(I-V) = 0.85 eV is characteristic of the palladium germanide/GaAs interface. Stage 2 The befmvior at Stage 2 (see Fig. 1) is characterized by the reaction of the substrate constituents, Ga and As, with Pd. The metallurgical behavior of the contact structure is shown in Fig. 5(a), and 5(b), where Auger analysis was performed on samples after sintering at 300°C for 10 and 103 rain. It is apparent, from a comparison with the profile in Fig. 3, that considerable interaction has taken place between the thin film overlayers and the GaAs substrate. The displacement of the Ge distribution to the surface, the interaction between Pd and the GaAs substrate, and the out-diffusion of both Ga and As are shown to progress with sintering time. The Auger spectra for Ge LMM transition exhibited 0.9],/,. tl

it

O.8

'

,I

'

'

L /

'1

. . . .

i

"Jl

200"C

before

I

'

sintering

I

. . . .

175~C

/

1 0 .//i

i

,I

,

,

I.O

,

,I

,

,

,

to Sintering

-rime

,I

1oo

,

,

,

L

iooo

(minutes)

Fig. 4. ~bs, (I-V) as a function of sintering time at 175,200 and 300°C.

SIN'rERED

1OO

300"C/10 rain.

Pd

75 v

a~ 5 0 co

o 25

.---'/,\ O

50

I00

150

200

250

iON DOSE

(a)

1°0t

SINTERED 3OO°C

P~-Oe I Pd-A'I 75J-

PO-Oo

rain.

/IO3

I

Pd

5c

f-As

o 25

5O

1OO

150

2O0

250

ION DOSE

(h) Fig. 5. Augerdepth profilesof the samples with layers of Pd and Ge 1.1 k/~ and 370,~ thick, respectively, after sintering at 300°C for (a) 10min and (b) 103 min. the same shape as that previously determined for Pd2Ge and PdGe (Fig. 2b). X-ray diffraction also indicated the presence of both germanide compounds after sintering similar samples of larger area for 100 min at 300°C. From the work of Olowolafe et aL [19], Pd reacts with both Ga and As to form several compounds. Sinha et al. [6], in their study of Pd/Ge/GaAs contacts, detected the presence of PdAs2 and PdGa after heat treatment. However, the only results reported were those obtained after sintering at the higher temperature of 500°C. The reaction between Pd and both Ga and As is also suggested by the Auger profiles displayed in Fig. 5. In Fig. 5(b), two regions between the germanide layer and.the GaAs substrate are apparent; one region between an ion dose of 60 and 100 and labeled Pd-As, and another region between an ion dose of 100 and 180 and labeled Pd-Ga. Several methods were utilized in this study in an attempt to identify the chemical composition of the compounds of Pd, Ga, and As which are likely to form in the Pd/Ge/GaAs contact structure. X-ray diffraction was found to lack sufficient sensitivity to detect either Pd-As or Pd-Ga compounds, but did indicate the absence of elemental Pd after sintering for 100 min at 300°C. The

Pd/Ge contacts to n-type GaAs samples used for the X-ray analysis had relatively thin layers of both Pd and Ge (1250 and 410/~ respectively); thus the sample volume may have been too small for the measurement technique. ESCA was also employed without success. Reference samples of reacted Pd on GaAs (with the identity of the compouds verified by X-ray diffraction) were analyzed. No discernible shift was detected in the binding energy for either Pd, As, or Ga when in different chemical states. SIMS was also used, but the ion yields were too small for detection by the instrumentation. Even though the compounds of Pd-As and Pd-Ga in the Pd/Ge/GaAs structure after sintering could not be identified in this study, it is likely from the reported work of others [6, 19] and from the Auger data in Fig. 15 that one or more compounds do exist. After sintering at 300°C for 100 rain, Ga, As and O were detected at the start of the Auger analysis. As shown in Fig. 5 the amount detected of first As and then of Ga, increases with sintering time. Since Ga and As are not present in germanide layer, the diffusion out to the surface is most likely along "high-diffusivity paths" or grain boundaries in the Pd2Ge and PdGe [24, 25]. The Auger spectrum indicated that As and Ga were in the form of oxides in a thin layer on the surface [13, 14]. The surface morphology was found to undergo little change throughout the interactions described above. The arrival of the germanides to the free surface as a result of sintering coincided with a slight change in surface color as observed by optical microscopy. However, the surface texture remained as smooth as the as-deposited surface. The electrical properties of the contact were found to be relatively insensitive to the compositional behavior for sintering at Stage 2. The contacts were rectifying, with the diode factor n less than 1.07 for all samples. The Schottky barrier energy @,.(I-V) is plotted in Fig. 4 as a function of sintering time at 300°C. As shown, the value of 4',, (I-V) remains between 0,86 and 0.82 eV and is quite close to the maximum values obtained at 175 and 200°C. Sintering at the higher temperatures of 350°C (for less than 40 min) and 400°C (for less than 10 min) resulted in the same high value of ~,, (I-V)-~ 0.84 eV.

Stage 3 Sintering at Stage 3 was characterized by the penetration of Ge to the GaAs substrate. In Fig. 6, the Auger depth profiles of two samples sintered at 400 and 500°C are shown. Comparison of Fig. 6(a) with the Auger depth profile of Fig. 5(b) indicates the compositional distributions to be similar, but with some significant differences near the GaAs substrate. Upon sintering, Ge is observed not only near the region closest to the surface but also throughout the contact overlayers, notably near the GaAs substrate. In general, the redistribution of the Ge profile toward the GaAs substrate was found to be more extensive for those samples sintered at higher temperatures or for longer times. The form of the Ge near the surface region in Fig. 6(a) was that of Pd2Ge or PdGe by observation of the

9"17

SINTERED 4OO°C// 112 rain,

I0C

75

Pd" A

o--t

GB

50

o

Ge

25

r.O 0

0

50

1OO

150

__

200

250

10N DOSE

(a)

SINTERED 5 0 0 ° C//120 rftin.

I00

75

Go

Pd

~z 5o

oF

"

5'0 ~

,;o

,.~o

~oo

~o

ION OOSE

(b) Fig. 6. Augerde2thprofilesof sampleswith layersof Pd and Oe l.Ik~, and 370A thick, respectively,after sinteringat (a) 400°C for ! 12min,(b) 550°Cfor 120min. characteristic Auger spectra. However, interferencewith the As Auger peaks again prevented the form of the Oe to be identified at all depths into the structure. X-ray diffraction was performed on a sample sintered at 500°C for 30 rain. Within the limited sensitivity of the X-ray analysis, only one at a peak 20 = 49.5°, besides that of the GaAs substrate, was obtained. This peak could be due to either PdGe (in a 030) orientation) or Pd2Ge (in a (121) orientation), or both [9]. Elemental Ge could also be present in the structure, initially in the region near the GaAs interface (Fig. 6a) and, perhaps, throughout the overlayers after prolonged sintering (Fig. 6b). The source of any elemental Ge would have to be the result of dissociation of the palladium germanide(s). Limited information is available in the literature on the stability of the germanide phases. Hutchins and Shepela [26] have reported that PdGe is stable to at least 560°C, but their study was with Pd layers on Ge substrates and thus without the competing reaction between Pd and Ga or As.

978

H. R. GRINOLDSand G. Y. ROBINSON

The surface texture in the active area of the contact remained relatively smooth at Stage 3. In Fig. 7, SEM micrographs of the as-deposited surface and the surface after sintering for 5 rain at 550°C are displayed. Some roughness develops in the contact area upon sintering, but significantly less than that obtained with the Au-Ge based alloyed contacts [3]. To be noted is the extensive roughening ot the Pd and Ge on the SiO2 around the contact. This was observed on all samples after sintering above 400°C. The roughness and the lack of good adhesion (the overlayer was easily scratched during probe measurements) would preclude the use of Pd/Ge as an interconnect metallization between contacts. Upon sintering at 550°C for more than 30 min, or at 600°C for shorter times, the texture of the contact became increasingly rough but did not exhibit a degree of roughness that would indicate the presence of a liquid phase during heating. In Stage 3, the Pd/Ge/GaAs structure was found to change from a Schottky diode to an ohmic contact. This was evident from both contact resistance measurements at room temperature and detailed current-voltage measurements over a range of temperatures. The electrical data will be discussed from measurements made on two sets of samples. Substrates of LPE n/n +, (100) GaAs with a net carrier concentration in the epitaxial layer of 1.1 x 1016cm -3 (Sn doped) and bulk, n-type, (100) GaAs with a net carrier concentration of 2.0 × 1017 cm -3 (Te doped) were used. The samples were fabricated concurrently using the procedures previously discussed with layers of Pd and Ge that were 1250 ~ and 410/~ thick, respectively. The progression to ohmic behavior with sintering is shown in Fig. 8. Both the forward and reverse characteristics of the diodes become more linear as the temperature of sintering is increased. The characteristics shown in Fig. 8(d), after sintering at 550°C for 20 rain, indicates that nearly identical behavior is obtained for both polarities of bias up to a diode current density of approximately 750Amp/cm2. However, both a nonlinearity and an asymmetry was observed in the I-V characteristics of the same diode at higher current densities. The specific contact resistance rc was evaluated after sintering and the results are presented in Fig. 9. The method of Cox and Strack [28], with modification for use with square contact geometries [3] was used to estimate re. As indicated in Fig. 9, the value of rc decreases by over eight orders of magnitude (to about 4.0× 10-4l~cm:) as the sintering temperature is increased from 400 to 600°C. The contacts fabricated on the higher-doped substrates are shown to yield lower values of rc than those fabricated on the lower-doped material for sintering below 55&C. Upon sintering at 600°C, the lowest value of rc was obtained and was found to be independent of the doping level in the substrate. This is in contrast to the results of several studies on both alloyed [29-31] and sintered systems [6], where the minimum value of rc was found to decrease with increasing substrate doping. This difference will be discussed more fully below. The uncertainty in the value of re, as depic-

ted by the error bar in Fig. 9, reflects the difficulty in accurate separation of the series and the bulk spreading resistance from that of the contact resistance. The Schottky barrier energy was evaluated by several measurement techniques. The value of &~, (I-V) determined from I-V measurements and that of tb~o (C-V) determined from C-V measurements [20] is plotted as a function of sintering temperature in Fig. 10. For completeness, the behavior is shown for all three stages. Also indicated is the variation of the diode factor n which was determined from the I-V measurements [20]. In Fig. 10, the difference in the barrier energy determined by the two techniques is due to an image-force lowering of approximately 0.025 eV [32] for these samples sintered throughout Stages 1 and 2. However, in Stage 3 the difference between &~n (I-V) and &so (C-V) is much greater than 0.025 eV and there is a rapid increase in n with increasing sintering temperature. In the calculation of Son (I-V), thermionic emission was assumed to govern current conduction in the contact. From the large value of n and the difference in the barrier energy determined by I-V and C-V measurements, the assumption of conduction by only thermionic emission is not justified at Stage 3 [32]. In order to determine if tunneling of electrons through the metal-semiconductor interface was significant, the I-V characteristics of contacts sintered at Stage 3 were measured as a function of temperature T [33]. The relative importance of thermionic and field emission to the total current can be estimated by the characteristic energy Eoo given by [32]

Eoo (qh/4~')(No[m*e,) 1:2 =

(1)

where h is Planck's constant, m* is the effective mass and es is the permittivity of the semiconductor. In Fig. 11, Vo, which is the inverse of the slope of In I vs V, has been plotted as a function of KTlq for an as-deposited sample and two samples at Stage 3 (one sintered at 400°C and one sintered at 450°C, each for 20 min). The value of Vo extrapolated to absolute zero is defined as Voo,where Eoo = qVoo. Eoo is found to increase after sintering from 10.8 to 34 meV, indicating an increase in the doping density. For the as-deposited sample, a discrepancy exists between the value of No calculated from Eoo = 10.9 meV (2.6×10'7cm-3) and that obtained from the C-V measurements (1.1 x 1016cm-3). Although an alteration of the doping density near the surface region of the GaAs substrate may occur upon deposition of the Ge, a more likely reason for the discrepancy is diode edge effects. Saxena [34] has indicated that "guard rings" are necessary for I-V measurements at low temperatures on contacts that have thermionic emission as the dominant conduction mechanism. Without restriction of the current flow to a path perpendicular to the contact interface, the higher field about the diode periphery introduces a tunneling component in the saturation current which becomes significant as the temperature is reduced. This effect was investigated in this study by measurements made on diodes with different peripheries. The value of

Pd/Ge contacts to n-type GaAs

Before

Sinterin 9

979

Af'~rer

Sinteein 9

5 5 0 ~C / 5 minufes, Pd - (~ ~r~ S i'

- (,5c, -(

¢

Fig. 7. Scanning electron micrographs of the contact surface before sintering and after sintering at 550°C for 5 min.

Pd/Ge contacts to n-type GaAs

a)

b)

V R (volt-) I0

20

I

16

~=

v. (volts)

30

0

i

I

2

Stoge I

O.91

5

981

I

I

I

StogeII

]

StogeIll

4

16 ~

,2

\

12

/

F~

O.E

> 1.5

u.~° 8

; ol ~

"~mO.7

O'

0.2

0.4

0.2

VF (volte)

c) 16 =

t2

~:8 u- 4

i

I

>

n

-

1.2 I

b.I

I

Sintered 4 5 0 * C

o I O.!

d)

~

1.4

0.4

vF (valve)

oe-depo$ited

J"

i

i.o

I IO0

*

I

*

200

1

J

300

Sinlering Temperolure

I

400 (°C)

I

1

500

Fig. lO. The behavior of @B, ( I - V ) , n, and @Bo (C-V) as a f u n c t i o n o f sintering t e m p e r a t u r e . T h e sintering t i m e w a s 20 m i n for e a c h s a m p l e . T h e stages, as defined in Fig. 1, are indicated. R

~a i

i

O

0.2

I

i

J

0.08

0.4

VF,VR (volts) Sintered 5 0 0 ° C

4a

0.16

VF, MR(VOlts) Sintered 550"C

• ~

kyo_o-3#_~v_. . . . .

ered

~ ~

4~ooc/zomio:

Sintering Time =2Drain. for b ) , c ) , d )

Fig. 8. Current-voltage characteristics of

Pd/Ge/GaAs diodes

with an active area of 2.6 x 10-5 cm2. The forward bias characteristic is denoted by F where IF and VF are the forward current and forward voltage (positive with respect to the metal), respectively. The reverse-bias characteristic is denoted by R where IR and Ve are the reverse current and reverse voltage, respectively. The I-V characteristics are shown for (a) before sintering, (b) after sintering at 450°C for 20 min, (c) after sintering at 500°C for 20 rain and (d) after sintering at 550°C for 20 min.

30

sintered

i

20

::mv

~

//vT'~ as-depo sited

I0 -~Voo= lO.gmVIo3F~!

,

,

102~ ~

i'

'

'

/~

I ,ND.NI= i] I X I016 qm3

I01~

"<-( Voo = 2.25mV)

O(

I

IIo -kT//q

I (mV)

I

20 •

I

30

Fig. 11. The behavior of Vo as a function kTIq for an asdeposited contact and contacts sintered at 400 and 450°C for

20 min. The active area was 1.58 x 10-3 cm2 for each contact. The extrapolated value of Voo is shown for each sample. If diode edge effects were negligible in the as-deposited contact, then a value of V0o= 2.25 mV would be predicted [32].

'a~L" ~ '3"0 ',~o ' 3 ; o ' ,;o z,~ Sintering

Temperature (°C)

Fig. 9. The specific contact resistance rc as a function of tintering temperature for samples with an initial substrate impurity concentration of 1.1 x 10~6cm -3 and 2.0 x lO v7cm -3. The sintering time was 20 rain at each temperature.

Eoo was found to decrease with decreasing periphery. Thus, the discrepancy in the value of No for the asdeposited sample in Fig 11 is attributed to edge effects associated with the diode geometry used in this study. Therefore, the correct value of Eoo was assumed to be 2.25 meV for the as deposited sample in Fig. 11, corresponding to N o = l . l × 1 0 ' 6 c m -3 from the C-V measurements. With the value of Eoo and using the thermionic-field emission theory of Padovani and Stratton [33, 35], the Schottky barrier energy at absolute zero @eo (I-V-T) can be estimated for each of the three samples in Fig. 11. In

982

H.R. GRINOLDSand G. Y. ROBINSON

Fig. 12, [J~ cosh (EoolkT)]lT is plotted vs l/Eo where the slope is proportional to -4,Bo (I-V-T)+qV, and qV, = kT[ln (Nc/No] is the energy difference between the Fermi level and the conduction band level in the semiconductor and Nc is the conduction band density of states [35]. The value of qV, was estimated using No obtained from C-V measurements for the as-deposited sample and from Eoo for the sintered samples. The results shown in Figs. 11 and 12 are summarized in Table 1. For the as-deposited sample, the values of the Schottky barrier energy determined by I-V, C-V, and I-V-T measurements are in reasonable agreement. After sintering, the Schottky barrier energy remains at a high value. This is shown by the values of both ~b~o (C-V) and ~ o (I-V-T). This is in contrast to the barrier energy 4~B, (I-V) obtained using only thermionic emission theory. From evaluation by these three methods, the contacts at Stage 3 appear to be described by thermionic-field emission. Hence, the conduction mechanism is not solely that of thermionic emission over the barrier; tunneling becomes more important and must be responsible for the ohmic behavior observed at Stage 3. A direct correlation between the specific contact resistance and the net impurity concentration tNo - NAI can be found from the values of Eoo determined from Fig. 11. In Fig. 13, INo-N~I is plotted along with the measured values of r,, as a function of sintering tern-

-"41

,

,

,

/ 166 z,

,

,

,

O

OS- deposited

A

sintered 4 0 0 * C

,

\~

20rain.

o

IO 2

-2 IO

sintered 4 5 0 " C

41--

T Z°

~-7

o

"8 16~ ILl o° ,-,0, -IF

)Bo: 1.0

[ ,68

I 2[0

i

4~

i

i

70

1

9%~

iO26

,/E0 I.v;' Fig. 12. A plot of [Js cosh (EoolkT)]lTvs llEowhere the slope is proportional to - ~ , + qV, for the samples of Fig. 11. The value of ~bBowas calculated from the data using a least-squares method of logarithmic-regression analysis.

Table 1. Summary of the electrical data obtained froml-V,C-V and 1-V-T measurements on three samples.Thediode area of each sample was 1.58 × 10-3 cm2. The values in parentheses are corrections for diode edge effects As-deposited

~Bn(I-V)

0.69 + 0.01

Sintered Sintered 400°C/20 minutes 450°C/20 minutes

0.79 + 0.01

0.57 + 0.01

i.i0

1.49

in eV n at 230C ¢Bo (C-V)

1.01 0.72+0.02

0.85+0.02

0.87+0.02

in eV I ND - NA I

l.lxl016

i.i x 1016

1.2xi016

in cm-3 (from C-V)

E(~o in meV IN D - NA J

in cm -3

10.8

17.3

34

(2.25) 2.6

x

1017

6.5xi017

2.5 x 1018

(i.i x 1016

(from Eoo )

~Bo (I-V-T) in eV

0.68+0.03

0.94+0.03

1.0+0.15

Pd/Ge contacts to n-type GaAs perature. With the assumption that the energy of the contact barrier ~Bo remains near 1.0 eV after sintering at 600°C, 'INo-NA I was estimated from thermionic-field emission theory and is indicated by the dotted line in Fig. 13 [36]. At Stage 3, the formation of a heavily doped n-type layer near the surface of the GaAs substrate is suggested by both the electrical results presented in Fig. 13 and by the distribution of the Ge shown in Fig. 6. Since Ge is known to be an amphoteric dopant in GaAs with a shallow donor level when occupying a Ga vacancy [3739] and since Pd does not form ohmic contacts on lightly-doped, n-type GaAs [6, 19], it is concluded that the incorporation of Ge into the GaAs as an n-type impurity is responsible for the ohmic behavior of the contact at Stage 3. The Ge doping results in the formation of a contact barrier that is high in energy but is thin enough to have a significant portion of the current conduction by tunneling of the majority carrier electrons. The spatial extent of the Ge-doped layer can be inferred only from the electrical measurements. Quantitative Auger analysis is limited by both the sensitivity of the technique and the interference of the Ge and As spectra [40, 18]. In addition, it was found that the relatively poor ion yield of Ge prevented SIMS from being used effectively. As noted in Table 1, the values of INo - Nal determined from Eoo are higher than the values calculated from the C-V data. Hence, the presence of the Ge-doped layer appears transparent to the C-V analysis. This implies that the edge of the depletion region, where the change in charge is measured for the C-V technique, is at a depth equal to or greater than the Ge-doped layer. This is also in agreement with the dependence of r~ on

I Ig

/ J

,o4

T

,

J

,

t

,

200 $intering

n

,

i

400

,

J

,

ot

6 0

,b4

Temperature (*C)

Fig. 13. The net impurity concentration near the semiconductor interface as calculated from the value of Eoo and as estimated from Ref. [36] ( - - - ) as a function of sintering temperature. The sintering time was 20 min for each sample. The value of JNo- NAI at 250C was obtained from C-V data. Also

shown is the behavior of rc for samples with an initial substrate impurity concentrationof 1.1 x l016cm-3.

983

the background doping level of the substrate for sintering below 550°C (Fig. 9). For rc to be dependent on the initial impurity concentration in the substrate, current conduction must be governed in part by the potential distribution beyond the Ge-doped layer. Thus, the layer does not extend far enough, nor is doped heavily enough, to contain all of the depletion region. Under these conditions, the barrier energy from the C-V analysis assuming a uniform impurity concentration, would predict a value lower than the actual barrier. This would account for the lower value of &Bo (C-V) compared to 4'~o (I-V-T) in Table 1.

4. SUMMARY

In Fig. I, the important aspects of the metallurgical and electrical behavior of the contact structure are summarized for all three stages. At Stage 1, the metallurgical behavior of the contact was characterized by the reaction of Pd and Ge and the formation of the germanide phases, Pd2Ge and PdGe. The reaction was found to proceed until all of the elemental Ge had completely reacted with the overlying Pd. Reaction between Pd and the GaAs substrate was prevented while elemental Ge was still present in the contact. With the change from a Pd/Ge/GaAs structure to that of a Pd/Pd2Ge-PdGe/GaAs structure, the Schottky barrier energy increased significantly. It is evident from the electrical properties of the contact that the germanide compounds are metallic and thus are analogous to silicide compounds (e.g. Pd2Si) [15, 26]. Stage 2 was characterized by the reaction of Pd with both Ga and As. The Pd-Ga and Pd-As reactions continued with sintering until all of the elemental Pd was consumed. This resulted in a structure depicted by the Auger profile in Fig 5(b). In the interaction between Pd and the GaAs substrate, the diffusing species may be either Pd or the germanides (or both) as well as Ga and As. If Pd diffuses through the germanide layer, then the Ge distribution in Fig. 5 may be viewed as a diffusion marker to measure the progress of the interaction between Pd and the GaAs substrate. The identification of the diffusing species was not pursued in this study. Throughout the interactions at Stage 2, the electrical properties of the contact remained rectifying. Since the Schottky barrier energy remained unchanged, the metalsemiconductor interface was not altered by the Pd-Ga and Pd-As interaction or if a new metal-semiconductor interface was formed, it has approximately the same barrier energy. Sintering at Stage 3 resulted in the formation of ohmic contacts by incorporation of the Ge into the GaAs as a donor impurity. Germanium, perhaps in elemental form, was observed by Auger analysis to redistribute upon sintering toward the surface of the GaAs. The presence of Ge at the GaAs surface was found to correlate with the degradation of the rectifying properties of the contact and the onset of ohmic behavior. Current conduction in the contact was described by thermionic-field emission theory and the existence of a Ge-doped layer was postulated. The Ge-doped layer evidently alters the potential

984

H. R. GRINOLDSand G. Y. ROBINSON

distribution near the MS interface such that electrons can easily tunnel through the contact barrier. The surface of sintered Pd/Ge contacts remained uniform at all stages and was significantly smoother than the surface of alloyed contacts to GaAs [3]. The lowest value of rc (3.5 x 10-41I-cm2) was obtained on the lighterdoped substrates only after extensive sintering and the accompaniment of some surface roughening. However, it was possible to form contacts with rc less than 10-3 ftcm2 to lightly-doped substrates and obtain very smooth surfaces similar to that shown in Fig. 7. Upon sintering at all stages, the interactions between the constituents of the thin-film structure appeared to be governed by diffusion-controlled, solid-phase processes. Over the entire range of sintering temperatures used in this study, there was no evidence from the Auger data, the electrical results, or the smooth appearance of the contact surface that would indicate the formation of a liquid phase during the heating process. The metallurgical interactions in the structure are instrumental in the formation of the ohmic contact. Since neither Ge nor Pd alone are able to form good ohmic contacts to lightly-doped n-type GaAs [6, 7, 19], the sintered Pd/Ge ohmic contacts must involve a process which is dependent upon the presence of both Pd and Ge. Thin films of only Ge on GaAs have been reported [7], and were observed in this study by Auger analysis, to form a relatively inert structure with no evidence of dissociation of the GaAs after sintering at 500°C. Although some diffusion of Ge into the GaAs substrate is expected to occur [41, 42], none was detectable by the Auger analysis. With both Pd and Ge in the structure, the relative number of Ga vacancies over As vacancies could be altered by the interaction between Pd and GaAs and the diffusion of both Ga and As out to the surface. Since the results presented for Stage 3 indicate that Ge is incorporated into the GaAs lattice as a donor impurity, the Pd-Ga and Pd-As interactions evidently provide a situation favorable for greater occupation by Ge on Ga sites rather than As sites. The usefulness of sintered PdlGe ohmic contacts to n-type GaAs is limited unless the specific contact resistance can be reduced by at least another order of magnitude (from 3 x 10-4 ~-cm 2 to around 10-~ l)-cm~). Alloyed contacts (e.g. Ni/Au-Ge) result in values of rc near 10-6 l)-cm2 on substrates doped at 1017cm -3 [43]. In those applications where specific contact resistance is not of primary importance, sintered Pd/Ge contacts are useful and offer several distinct advantages over alloyed contacts. A superior degree of surface planarity, a more uniform coverage of the contact area, a greater reproducibility of the solid-phase reactions, and an expected higher reliability of the contact are attributes possessed by this system which are not readily available with an alloyed contact system. It is these qualities that make the sintered Pd/Ge/GaAs system a viable ohmic contact, suitable for use in the fabrication of GaAs devices. The Pd/Ge/GaAs contact has application as a

Schottky barrier diode. In the as-deposited state, the contact has a highly reproducible barrier height and conduction that is characterized by thermionic emission over a wide range of forward current.

Acknowledgements--The authors would like to acknowledge L. E. Toth for his comments in the early stages of this study, Mike Ross for his help and time with the ESCA and SIMS equipment, and R. E. Lee for the epitaxial GaAs material. The study was supported by the National Science Foundation under grant DMR 76-14997.

REFERENCES

I. V. L. Rideout, Solid-St. Electron. 18, 541 (1975). 2. A. Y. C. Yu, H. J. Gopen and R. K. Waits, Solid-St. Electron. 14, 515 (1971) 3. G. Y. Robinson, Solid-St. Electron. 18, 331 (1975). 4. D. L. Partin, A. (3. Milnes and L. F. Vassamillet, J. Electron. Math. 7,279 (1978). 5. J. M. Poate, K. N. Tu and J. W. Mayer (Eds.) Thin FilmsInterdiffusion and Reactions. Wiley, New York (1978). 6. A. K. Sinha, T. E. Smith and H. J. Levinstein, IEEE Trans. Electron. Dev. ED-22, 218 (1975). 7. W.T. Anderson,Jr., A. Christou and J. E. Davey, IEEE Trans. Electron. Dev. Meeting; Dig. Tech. Papers, p. 192, (1977). 8. H. R. Grinolds and G. Y. Robinson, Appl. Phys. Lett. 34, 575 0979). 9. Powder diffraction file, Americal Society for Testing and Materials, Card Numbers: Pd--5-0681, GaAs--14-450, PdGe--7-286, Pd2Ge---6-558. 10. Operation manual and specifications for Phi 115-110 CMA, Physical Electronics Industries (1974). 11. J. A. Roth and C, R. Crowell J. Vac. Sci. Tech. 15, 1317(1978). 12. R. Ludeke and A. Korea, Phys. Rev. B13, 739 (1976). 13. C. C. Chang, R. P. H. Changand S. P. Murarka, J. Electrochem. Soc. 124, 922 (1977). 14. L. A. Chesler and G. Y. Robinson, Appl. Phys. Lett. 32, 60 (]978). 15. D, J. Fertig and G. Y. Robinson, Solid-St, Electron. 19, (407) 75 (1976). 16. G. Ottaviani, C. Canali, G. Ferrari, G. Majni, M. Prudenziati and S. S. Lau, Thin Solid Films, 47, 187 (1977). 17. W. D. Buckley,S. C. Moss and R. Scott, Solid-St. Electron. 15, 1331 (1972). 18. L. E. Davis, P. W. Palmberg, G. E. Riach, R. E. Weber and N. C. MacDonald, Handbook of Auger Electron Spectroscopy, Physical Electronics Industries (1976). 19. J. O. Olowolafe, P. S. Ho, H. J. Hovel, J. E. Lewis and J. M. Woodall, 3", AppL Phys. 50, 955 "(1979). 20. S. M. Sze, Physics of Semiconductor Devices. Wiley, New York (1%9). 21. F. A. Padovani and G. G. Sumner, J. Appl. Phys. 36, 3744 (]%5). 22. S. Guha, B. M. Arora and V. P. Salvi, Solid-St. Electron. 20, 431 (1977). 23. B. R. Pruniaux, J. Appl. Phys. 42, 3575 (1971). 24. J. M. Morabito, Proc. 7th Intern. Vac. Con[. and 3rd Intern. Conf. Solid Surfaces, p. 2267, Vienna (1977). 25. P. G. Shewmon, Diffusion in Solids. McGraw-Hill,New York (]%3). 26. G. A. Hutchins and A. Shepela, Thin SolidFilms 18, 343 (1973). 27. C.C. Chang,P. H. Citrin and B. Schwartz,J. Vac. Sci. Tech. 14, 943 (1977), 28. R. H. Cox and H. Strack, Solid-St. Electron. 10, 1213 (1%7). 29. H. Matino and M. Tokunaga, jr. Electrochem. Soc. 116, 709 (1%9). 30. K. Heime, U. Konig,E. Kohn and A. Wortman, Solid-St.Electron. 17, 835 (1974).

H. R. GR1NOLDSand G. Y. ROaZNSON 31. Y. A. Goldberg and B. V. Tsarenkov, Soviet Phys.--Semiconductors, 3, 1447 (1970). 32. E. H. Rhoderick, Metal-Semiconductor Contacts. Oxford Press (1978). 33. F. A. Padovani in Semiconductors and Semimetals (Edited by Willardson and Beer), Vol. 7A. Academic Press New York, 097D. 34. A. N. Saxena, Surface Sci. 13, 1.51 (1969). 35. F. A. Padovani and R. Stratton, Solid-St. Electron. 9, 695 (1966). 36. C. Y. Chang, Y. K. Fang and S. M. Sze, Solid-St. Electron. 14, 541 (1971). 37. A. G. Milnes, Deep Impurties in Semiconductors. Wiley, New York (1973). 38. J. O. McCaldin and R, Harada, J. Appl. Phys. 31, 2065 (1960).

985

39. A. M. Andrews and N. Holonyak, Jr., Solid-St. Electron. 15, 601/1972). 40. A. W. Czanderna led), Methods o/Surface Analysis. Elsevier Scientific, New York (1975). 41. D. L. Kendall in Semiconductors and Semimetals, (Edited by Willardson and Beer), Vol. 4. Academic Press, New York 0968). 42. H. Morkoc, S. G. Bandy, R. Sankaran, G. A. Antypas and R. L. Bell, IEEE Trans. Electron. Dev. ED-,~, 619 (1978). 43. M. Otsubo, H. Kumabe and 14. Miki, Solid-St. Electron. 20, 617 (1977). 44. A. Christou, Solid-St. Electron. 22, 141 (1979). 45. L. P. Erickson, M. S. Thesis, University of Minnesota Graduate School, (1979).