Superior Ohmic Contacts to III-V Semiconductor by Virtue of Double Donor Impurity

This invention relates to the fabrication of semiconductor devices and especially to the improvement of ohmic contacts in group III-V semiconductor materials.

It has been a problem to provide satisfactory ohmic contacts in III-V type semiconductor materials, such as semiinsulating (S.I.) GaAs. When heavily implanted with Si, for example, self-compensating Si-Si complexes are frequently formed which are electrically neutral, thereby preventing formation of good ohmic contacts.

An object of this invention is to break up electrically neutral, or inactive, donor or acceptor ion complexes in III-V semiconductor substrates and substitute electrically active complexes in their places so that good ohmic contacts can be formed therewith.

Another object is to improve ohmic contacts in group III-V semiconductor compounds, such as GaAs.

The present invention is a method to dope group III-V compounds semiconductors to degeneracy such that tunnelling (nonalloyed) ohmic contacts of extremely reliable refractory metals can be made. The method employs use of a chalcogen material (in the case of donors) from column VI of the periodic table of elements to (1) enhance the diffusion of column IV materials into III-V materials by preventing the formation of covalently bonded IV--IV complexes on the surface which are too large to in-diffuse, and (2) to form IV-VI complexes within the III-V material which act as donors on their respective cation (Ga) and anion (As) lattice sites within the III-V crystal (i.e., double-sided doping effect) and which, once formed, become too large to further diffuse, thus providing extremely reliable and stable donor pairs. Diffusion into the selected region of the extremely high concentration (e.g., >8.times.10.sup.18 /cm.sup.3) of donors is further aided by the selective implantation of silicon. The diffusion effect can be further enhanced if this implant is done while the III-V substrate is held at a temperature of 77 degrees K. If the cryogenic implant is performed, anneal/activation of the implant should be at about 580 degrees C. for about 24 hours versus the conventional anneal of 850 degrees C. for 30 minutes.

FIGS. 1A-D show the various steps in the process of fabricating improved ohmic contacts for III-V type semiconductor materials.

As may be seen in FIG. 1A, a III-V type semiconductor wafer 10 is selected as a substrate for fabricating a semiconductor device. The following description will relate specifically to a semi-insulating (S.I.) GaAs substrate 10, since the description is easier in terms of a specific III-V compound, and to a FET semiconductor device.

In step 1, a thin layer 12 of Si.sub.3 N.sub.4 is laid down on the surface of the substrate. Using conventional photoresist techniques, areas 11 are opened in the Si.sub.3 N.sub.4 layer; these areas are the regions upon which ohmic-contact electrodes are later to be placed. New photo resist is applied over the top of the device and then removed from the regions 11 which will comprise the ohmic contact region and the active channel region (under region 13 of the Si.sub.3 N.sub.4 layer).

In step 2, Si.sup.+ ions are beamed onto the top of the device at the energy level necessary to create the optimum depth of the electrically active channel (e.g., 1200 A for a power FET) and at a fluence level necessary to activate the appropriate impurity donor density (e.g., 2.times.10.sup.17 /cm.sup.3 for a power FET). Note that the Si ions are not energetic enough to penetrate both the photoresist layer 14 and the Si.sub.3 N.sub.4 layer 12, while their penetration depth through the Si.sub.3 N.sub.4 only region 13 is less than their unencumbered penetration into the regions 17 and 19 which are to become the ohmic-contact regions.

FIG. 1C shows a further deposit of a layer 18 of Ge to a depth of about 300 A. (The above steps are part of the prior art). A layer 20 of Se is now deposited on the Ge layer 18 to a depth of about 300 A. The device is now coated with a layer 22 of SiO.sub.2, or silox, or Si.sub.3 N.sub.4, or a mixture of these, and put through an annealing process at a temperature of about 750.degree. to 850.degree. C. for about 30 minutes. This is to rapidly diffuse the Ge and Se into the regions 24 and 26 of of the Si-ion implant which will become the ohmic contacts. Diffusion does not take place in the active channel region 16 which is protected by a Si.sub.3 N.sub.4 layer 13. Lateral diffusion is limited as Si implant damage is annealed out.

The diffusion effect can be further enhanced if the Si implant is done while the III-V substrate is held at a temperature of 77.degree. K. If this cryogenic implant is done, anneal/activation of the implant should be at a temperature of about 580.degree. C. for about 24 hours.

The SiO.sub.2, Si.sub.3 N.sub.4, and excess surface Ge and Se are now removed and a layer 28 of refractory metal, preferably, is deposited over each of the regions 26, 16 and 24, which may constitute the source, gate and drain regions of the FET. This can be done in a single conventional step. Metallization over the Ge/Se diffused regions 17 and 19 results in ohmic tunneling contacts of low specific contact resistivity (e.g., <1.times.10.sup.-6 ohm-cm.sup.2) whereas metal deposited over the shallow Si.sup.+ implant region 16 results in a rectifying, nonohmic, Schottky barrier contact.

Since metallization of source, gate and drain electrodes can be accomplished in one step with this approach, lithographic alignment errors are virtually eliminated. Metallization can be of the well-known Ti-W-Au type which is the most reliable known. Its reliability is not reduced by the requirement of placing a Au-Ge-Ni alloy between the Ti-W and the III-V semiconductor as is conventionally required.

The factor of importance herein is that the group IV layer (e.g., Ge) should be augmented by a group VI chalcogen layer (e.g., Se or S) such that covalent Ge-Ge bonds are broken and atomic Ge permitted to diffuse into the Si implant-damaged region. The SiO.sub.2 -Si.sub.3 N.sub.4 encapsulant should prevent unwanted diffusion of the chalcogen or Ge into the active channel region. The Ge-chalcogen complexes formed within the silicon implant-damaged region will be physically too large to diffuse further.

The process may also be employed in III-V group materials in which acceptor atoms are implanted. Acceptor atoms of group II may be used to fill some of the Ga sites and atoms of group IV to fill some of the As sites. Thus, the first layer after the Si implant might be a layer of Zn (group II) instead of Ge and the second a layer of Ge (group IV) instead of Se. This would give a P-type of semiconductor material after anneal, which would be effected at roughly the same temperature and time conditions.

In more general terms, any impurity ions which act as donors or acceptors (i.e., elements which can be ionized to provide charge carriers) for a III-V semiconductor compound can be implanted in the substrate. These ions can come from elements in groups II, IV or VI. Although indiffusion elements have been selected and although the indiffusion elements which have been illustrated herein have been Ge and Se, they may be any pair of elements which, when diffused into the implanted region of the semiconductor, become substitutional on opposite sub-lattice sites (i.e., on the III and V element lattice sites) and act together as either a pair of donors or a pair of acceptors.

The advantages of the above-described method are:

1. It provides a breakup of the Ge-Ge complexes on the GaAs surface and thus permits a more rapid diffusion of Ge into the Si implant-damaged region such that this entire region is filled with Ge prior to damage anneal.

2. It provides for a double-sided donor activity (e.g., Ge on Ga site, chalcogen on adjacent As site).

3. It provides for a greater concentration of donors within the Si implant-damaged region.

4. Rapid impurity diffusion is limited to the Si implant-damaged region.

5. The probability of degenerate donor doping to the level required for a tunnelling contact to any metal is greatly enhanced.

6. The use of the potentially unreliable gold alloy contact is eliminated.

7. The probability of obtaining tunnelling contact action by ultra-reliable refractory metal (e.g., Pt or W) contacts is enhanced.

Obviously, many modifications and variations of the present invention are possible in light of the above teachings. It is therefore to be understood that, within the scope of the appended claims, the invention may be practiced otherwise than as specifically described.