Stable ohmic contacts to n-GaAs using ion-beam mixing

Stable ohmic contacts to n-GaAs using ion-beam mixing

Volume 3, number MATERIALS 7,8 STABLE OHMIC CONTACTS S.R. SMITH TO n-GaAs USING ION-BEAM MIXING 13 March 300 College Park Ave., Dayton, OH 45...

411KB Sizes 0 Downloads 9 Views

Volume

3, number

MATERIALS

7,8

STABLE OHMIC CONTACTS S.R. SMITH

TO n-GaAs USING

ION-BEAM

MIXING

13 March

300 College Park Ave., Dayton, OH 45469, USA

1985

Stable ohmic contacts to n-type (100) GaAs have been fabricated with structure was subjected to Ge + ion bombardment of sufficient energy and interface. A high-intensity incoherent lamp was used to vacuum anneal the chemical, physical, and electrical properties. The resulting structure consisted very thin interfacial layer of Ni, GeAs. This process specifically addresses eutectic alloy contact, namely, electromigration of the Au from the interface,

1. Introduction In the art of manufacturing semiconductor devices, contacts must be connected to or incorporated within the bulk of a semiconductor substrate so that electrical connections can be made to the device or devices. The contact should be non-injecting for minority carriers and the contact should not undergo electromigration under high electric fields [ 11. To be commercially successful any contacting process must be compatible with the mass production techniques employed to make the devices to which the contact is applied; it must also yield reproducibly reliable contacts. All semiconductor devices require at least one ohmic contact and often the quality of the contact is one of the most significant factors affecting the performance of III-V semiconductor devices. Segregation, interdiffusion, and reaction of the constituents of a metal-semiconductor contact, which can occur at the elevated temperatures used for contact formation or device operation, greatly affect the electrical properties of the metal-semiconductor structure. Alloying or sintering of composite metal films has been shown to produce changes in the Schottky barrier energy and contact resistance as a result of the alteration of the contact metal and by an increase in the majority carrier concentration in the semiconductor. The composition of the met294

May 1985

and J.S. SOLOMON

Uniuersity of Dayton Research Institute, Received

LETTERS

thin vapor deposited Ni films. The Ni/GaAs dose to induce a mixing effect at the Ni-GaAs contact zone which was then characterized for primarily of a Ni, GaAs layer on GaAs with a the shortcomings inherent in the Ni/Au-Ge phase separation and Au coalescence.

al film can be altered by interdiffusion and reaction within the metal fiim and by reaction with the semiconductor element(s). The net carrier concentration in the substrate can be altered by the indiffusion of a dopant from the metal or by the creation of lattice defects by the out-diffusion of a substrate element [2] . Ohmic implies, in principle, a non-injecting nature and linear current-voltage characteristics in both directions. However, in practice, a contact is acceptable if the voltage drop across the contact is much smaller than that across the device. Linearity is of less importance if the contact resistance is less than the device resistance. Piotrowska et al. [3] have compared the various contacting techniques for III-V compound semiconductors. Their results suggest that the formation of an ohmic contact is largely the result of the doping action of active species from multicomponent metallic structures. They conclude that the variety of technological approaches to ohmic contact formation reflects the difficulties in obtaining satisfactory contacts rather than the great choice of methodologies.

2. The Ni/Au-Ge

system

The Ni/Au-Ge alloyed contact to n-GaAs has been in use for more than 15 years. The system of 0 167-577x/85/$ 03.30 0 Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)

Volume 3, number 7.8

MATERIALS

Ni-capped, Au-Ge (12%) eutectic alloy was first introduced by Braslau et al. [4] for making ohmic contacts to Gunn-effect devices. Since then, it has become the most widely used contact to n-type GaAs. Even though the precise mechanism of contact formation is still debated, and even the cause of contact failure is not universally agreed upon. Wittmer et al. [5] have shown that after short annealing times the Ni/Au-Ge system is in a stable configuration in which two of the elements have formed compounds at the interface. They report that the failure of the electrical properties of ohmic contacts to GaAs may be due to insufficient doping of the GaAs by Ge, but performed no work of their own on this system. Robinson ]6] has studied the behavior of the constituent elements during the contact formation process. He reports that for heat treatments below the eutectic temperature of 360°C layers of Au/Ge are relatively unchanged, but that Ni diffuses through to the GaAs/ AuGe interface and accumulates creating a Ni/GaAs Schottky diode. Above 360°C the AuGe melts and Ge diffuses into the GaAs resulting in ohmic behavior. Ni present at the interface increases the wetting properties of the molten AuGe on the GaAs surface and enhances the uniformity of the contact. Ogawa [7] studied Ni/Au-Ge contacts using microprobe AES and X-ray analyses. His results support the important role played by the Ni at the interface, and further show the formation of the binary compounds of NiAs, and AuGa, as well as the formation of a ternary compound NiGeAs at 500°C. However, it is precisely all of these various compounds that create the non-uniformity that eventually disrupts the interface and leads to contact failure. While Grovenor [8] concluded that the interfacial phase formation played little part in determining the macroscopic properties of the contact and that Ge doping was not really the process responsible for the ohmic nature, Heiblum et al. [9] found evidence for strong phase segregation and a high-resistivity layer at the interface. All of this points up the complex nature of the alloying process, which is not well understood. In a comparative study of Au/Ni/Au-Ge, Ni/Au-Ge, and Au-Ge contacts on n-GaAs, Marlow [lo] found that the degradation of the contact nature after aging appeared to be attributable to the out-diffusion of Ga and the in-diffusion of Au and Ni.

L1:TTERS

May 1985

Dell et al. [ 1 l] attempted to determine the minimum amount of Ge necessary to form a contact by varying the thickness of the Au-Ge film. They did not attempt to separate the effects of the Au and Ge, however. In an attempt to enhance the doping efficiency, Vidimari [ 121 used an SiO cap over the Ni/Au-Ge layers to some success. The difficulties inherent in the alloy system remain, however, and a new approach to produce ohmic contacts on n-GaAs is clearly called for.

3. The Au/Ni-Ge/GaAs

system

In 1983, Kuan et al. [ 131 showed conclusively that the culprit in the alloyed contact was one of the constituent elements: namely Au. They studied the alloy system on GaAs(lOO) surfaces using SEM and STEM/ EDX. The disruptive effect of the phase segregation at the interface was clearly demonstrated. The growth of Au(Ga, As) and Ni,GeAs grains larger than 200 nm appears to raise the barrier height and the contact resistance and form areas of restricted current flow as mentioned by Braslau [4]. From this and other work [2,4,7,10] we are able to point to Au as the cause of poor aging performance and surface inhomogeneity, and to develop a solution to the problems inherent in the alloyed Ni/Au-Ge system. If Au is the elemental constituent that causes the failure of the system, it seems fairly obvious that a contact made without Au should not be subject to the shortcomings induced by the presence of Au, such as: phase segregation at the interface, and the subsequent formation of a Au/GaAs Schottky barrier; “balling up” of the metallic gold, which disrupts the contact morphology; and the in-diffusion of Au beyond the contact region which seems to allow, or cause, the dissolution of the contact and produce the poor aging characteristics inherent with this method. In 1971, Bower was granted a patent [ 141 for an ohmic contact process with n-GaAs using ion-beam mixing [ 151. Nothing is found in the literature concerning the contacting method of Bower, perhaps because he also relied on Au as the metallic component of his system. However, since that time, ion-beam mixing has become increasingly more useful as a research and fabrication tool. In the present work ion-beam mixing was incorporated into the contact formation process. The contact 295

Volume

3, number

7,8

MATERIALS

May 1985

LETTERS

is made with Ni and implanted Ge; Au is deposited after the contact is formed to provide a wire interconnect. Thus this system is referred to as a Au/NiGe/GaAs contact.

4. Experimental Nickel was vapor deposited to a thickness of 100 nm by electron-beam evaporation onto (100) GaAs in an ultra-high vacuum chamber at pressures below 1 X lop7 Torr at a rate of x0.3 rim/s.. The GaAs was subjected to the following treatments prior to Ni deposition: (1) Rinsed in methanol for 15 s; (2) rinsed in distilled HZ0 for 15 s; (3) etched in a solution of 5 H,SO, : 1 HZ02: 1 HZ0 for 10 min at 6O’C; (4) rinsed in distilled HZ0 for 30 s; (5) etched in 1 HCl: 5 HZ0 for 30 s; (6) blown dry with filtered dry nitrogen; (7) UV ozone cleaned * 1 ; (8) incoherent lamp heated to 580°C for 5 min. The UV ozone cleaning step removed nearly all traces of carbon while the 58O’C heat treatment removed the remaining surface oxide just prior to Ni deposition

[lb].

The Ni/GaAs specimen was removed from the deposition chamber and placed into an ion implanter where it received a dose of 1 X 1015 atoms/cm* at an energy of 300 keV, which produces a peak in the Ge distribution just beyond the Ni/GaAs interface in the GaAs. The specimen was then re-inserted into the deposition chamber where it was vacuum annealed with a 1000 W high-intensity incoherent lamp assembly for ~10 s. The characteristic time versus temperature curve obtained with a chromel-alumel thermocouple attached to the deposited Ni surface is shown in fig. 1. Following the annealing step Au was deposited over the Ni to a thickness of 500 nm. Auger electron measurements were made with a Physical Electronics Industries model 10-I 50 analyzer using 4 keV electrons with a current density of 0.05 A/cm*. Inert ion sputtering for depth profiling was accomplished with argon ions generated with a Physicai *I Ultraviolet Ozone Cleaner, model 0306, Montgomeryville, PA 19836.

296

UVOCS,

Inc.,

TIME

(set

1

Fig. 1. Temperature versus time characteristics for in vacua heating of Ni films on GaAs with a 1000 W high-intensity incoherent lamp.

Electronics Industries model 04-303 ion gun operated at 2 kV and a current density of 8.3 X lop3 A/cm*. Sputter rates were determined using a Sloan Dektak II surface profiler. The profiles obtained are shown in fig. 2. The simple electrical characteristics of the contacts were determined by current-voltage (I- v) measurements. The output from a Keithley 610C electrometer in the current measuring mode was fed to they axis of an Hewlett-Packard 7064A X-Y, Y recorder; the x axis was connected directly across the voltage leads.

+

v)

z

+

2.

l

4

+++++

or

0

I

100

200 SPUTTER

400

300 DEPTH

500

(nm)

Fig. 2. Auger sputter depth profiles of As, Ga, and Ni from Ge implanted contact following 800°C anneal. Numbers in parentheses are normalization factors applied to the Auger signals of the designated elements.

Volume

3, number

7 .8

MATERIALS

LETTERS

May 1985

5. Results Electrical measurements were taken on an array of six contact dots and an array of four dots. All exhibited a linear dependence of the current on the applied voltage. No quantitative determination of the specific contact resistance or of the interfacial resistance was possible with the samples in this configuration. An example of the behavior seen is shown in fig. 3. Further tests using transmission line measurements (TLM) or four terminal Kelvin resistors are planned. An important point to note is that gold was used only as a bonding pad. There is no evidence, even after six months, that the gold has in any significant way diffused through the Ni GaAs layer. Certainly not enough has passed the interface to disrupt the morphology of the contact or raise the barrier height. Based on Auger data the composition of the NiGa-As layer was Ni,GaAs. This is in agreement with work by Ogawa [ 171 and Lahav et al. [18]. Germanium was not detected by AES because the detectability limit for that element in GaAs with our instrument is ~3 at%. The SEM micrograph in fig. 4 shows the cross section of a fractured contact zone. The Ni,GaAs layer appears to be uniform and is suspected to have an epitaxial relationship of [llOl]/[lOO] to the GaAs substrate as reported by Lahav et al. [ 181.

I(A)

Fig. 4. SEM micrograph (100) GaAs.

of fractured

Au/Ni-Ge

contact

on

6. Conclusions A practical method of applying ohmic contacts to n-type GaAs using ion-beam mixing technology has been demonstrated. This approach eliminates the problems usually associated with alloyed contacts and provides a stable interface through which electrical connection to GaAs devices can be made.

Acknowledgement

T 5x10+

We would like to recognize the support of Mr. Jim Ehret , AFWAL/AADR in performing the implantations of Ge, Mr. Don Thomas for preparing the Ni evaporations, Mr. Paul VonRichter for his contributions to the electrical measurements, and Mrs. Judith Mescher-Smith for the SEM work. This work was performed under Air Force Contract No. F336 15-81 -C5095.

References I-4

l-5 Fig. 3. Current-voltage (Z-v) characteristics of two contact GaAs sample. The total resistance includes both the substrate resistance and the interface resistance. The sample was a planar array of 1 mm dots on the (100) GaAs surface.

111 V.L. Rideout, Solid State Electron. 18 (1975) 541. I21 G.Y. Robinson, Thin Solid Films 72 (1980) 129. A. Guivarc’h and G. Pelous. Solid State (31 A. Piotrowska, Electron. 26 (1983) 179. [41 N. Braslau. J.B. Gunn and J.L. Stables, Solid State Electron. 10 (1967) 381.

297

Volume 3, number 7,8

MATERIALS LETTERS

[5] M. Wittmer, R. Pretorius, J.W. Mayer and M.-A. Nicolet, Solid State Electron. 20 (1977) 433. [6] G.Y. Robinson, Solid State Electron. 18 (1975) 331. [7] M. Ogawa, J. Appl. Phys. 51 (1980) 406. [8] C.R.M. Grovenor, Solid State Electron. 24 (1981) 792. [9] M. Heiblum, M.I. Nathan and C.A. Chang, Solid State Electron. 25 (1982) 185. [lo] G.S. Marlow, M.B. Das and L. Tongson, Solid State Electron. 26 (1983) 259. [ 111 J. Dell, H.L. Hartnagel and A.G. Nassibian, J. Phys. D16 (1983) L243.

298

May 1985

[12] F. Vidimari, Electron. Letters 15 (1979) 675. [13] T.S. Kuan,P.E. Batson,T.N. Jackson, H. Ruprecht and E.L. Wilkie, J. Appl. Phys. 54 (1983) 6952. [14] R.W. Bower, U.S. Patent No. 3,600,797 (24 August 1971). [15] S. Matteson and M.-A. Nicolet, Ann. Rev. Mat. Sci. 13 (1983) 339. [ 161 A.Y. Cho, Thin Solid Films 100 (1983) 291. [17] M. Ogawa, Thin Solid Films 70 (1980) 181. [ 181 A. Lahav, M. Eizenberg and Y. Komen, in: Proceedings of the Materials Research Society, Symposium D, 1984 Fall Meeting, Boston, MA, to be published.