Note: This is a draft of a paper submitted for publication. Contents of this paper should not be quoted or referred to without permission of the author(s).

I1 11 To be presented at 6th International Conference on Silicon Carbide and

Related Materials-1995, Kyoto, Japan, September 18-21, 1995

To be published in meeting Proceedings

MOLECULAR- JET CHEMICAL VAPOR DEPOSITION OF S i c D. Lubben, G. E. Jellison, and F. A. Modine

Solid State Division, Oak Ridge National Laboratory P.O. Box 2008, Oak Ridge, Tennessee 37831-6030

“The submitted manuscript has been authored by a contractor of the U.S. Government under contract No. DE-AC05-840R21400. Accordingly, the U.S. Government retains a nonexclusive, royalty-free license to publish or reproduce the published form of this contribution, or allow ethers to do so. for U.S. Government purposes.

Prepared by SOLID STATE DIVISION

OAK RIDGE NATIONAL LABORATORY Managed by

LOCKHEED MARTIN ENERGY SYSTEMS under

Contract No. DE-AC05-840R21400 with the

U.S. DEPARTMENT OF ENERGY Oak Ridge, Tennessee

September 1995

Molecular-jet chemical vapor deposition of S i c

D. Lubben, G. E. Jellison and F. A. M o d i n e

Solid State Division, Oak Ridge National Laboratory, Oak Ridge, TN 37531- 6030

Abstract. Sic films have been deposited by molecular-jet chemical vapor deposition (MJCVD) on Si(OO1) substrates. Methylsilane (MS) diluted in He was used as a precursor for deposition under conditions which produced a MS molecular beam with 0.365 eV translational energy. Films grown at temperatures between 1000 and 1130 "C and above 1200 "C were single crystal as judged by electron channeling, while those grown at intermediate temperatures were polycrystalline. Films grown at lower temperatures gen- erally had a smoother surface morphology for moderate thicknesses, although all films showed at least some degree of faceting. The best thick films, up to 4 pm, were obtained for substrate temperatures of x 1210 "C under flow conditions which produced a deposition rate of 1200 -\ per minute.

1. In t roduct ion

Sic is a material with whose excellent thermal properties and wide bandgap make it a potentially important candidate for high-voltage, high-power and high-temperature device applications. Although many polymorphs exist, the two most extensively studied are the 6H (hexagonal) and 3C (cubic) material.

The major impediments to realizing the potential of S i c for power applications are structural imperfections present in the material. Bulk 6H Sic contains large features known as micropipes which are holes through the material that are capable of short- ing vertical device structures. These micropipes are propagated through epitaxial thin films grown on 6H substrates. 3C-Sic is more difficult to produce in bulk and is usually formed by epitaxial growth on single-crystal substrates, e.g. Si. However, high-quality single-crystal substrates sufficiently lattice-matched to S i c to produce defect-free mate- rial are not available. This problem is exacerbated by the high temperatures required for epitaxial growth, leading to thermal expansion mismatches as well.

Since the first successful heteroepitaxy of S ic on Si over a decade ago 111, there has been a large interest in film growth. Recently, several authors used different precursor gases for chemical vapor deposition (CVD) in an attempt to lower the deposition tem- perature and/or eliminate the need for buffer layers. These gaies include methylsilane.

1

SiCH3H3 (MS) [2], silacyclobutane (31, hexamethyldisilane [4. 51. and triethylsilane [SI. Reactive sputtering with a low-energy particle bombardment of the growing film [GI, has also been used to improve the crystallinity.

In this paper we discuss a preliminary investigation into the growth of Sic by molecular jet chemical vapor deposition (MJCVD). MJCVD is similar to conventional CVD except that a molecular beam from a "free jet" expansion is utilized as the source for deposition. There are two main benefits to this technique: 1) very high fluses can be obtained in a collisionless beam which eliminates unwanted homogeneous reactions and may allow higher deposition rates, and 2) seeded beams can be used to achieve superthermal energies which can lead to improved epitayy at reduced temperatures.

Under the conditions used in our experiments, the e n e r a of the MS precursors may be calculated from [7]

E; CpTo . Wi WN,, -=-

where E; and W; are the energy and molecular weight of species i, respectively, cp is the molar average heat capacity, To is the temperature of the gas behind the orifice, and is the molar average molecular weight. Inserting values for 10% >IS in He at TO = 300K yields an energy of 0.363 eV. In the 100% dilution limit, this value would be 0.510 eV; if Hz were used rather than He, the energy would be 1.02 e\'.

2. Experimental Procedures

The films were grown in a load-locked, turbomolecular-pumped vacuum system with a base pressure of 2 x Torr. MS? diluted to 10 % in He was introduced through a 100 pm diameter orifice at a distance of 2 to 3 cm above the substrate. The flow through the orifice was regulated by a servo-driven leak valve that controlled by feedback from the measured chamber background pressure Pb. The substrates, both on-axis and 4" miscut Si(OO1) wafers, were radiatively heated using a BS-encapsulated graphite heater. .A cutout in the substrate platten allowed direct radiative coupling to the sample.

The substrates were degreased, etched in dilute HF, reoxidized in a UI- ozone reactor and etched again in dilute HF immediatelx prior to inserting them into the deposition chamber. Before deposition, the wafers were annealed for 1 minute at 900 "C. In some cases, propane gas was introduced into the chamber (not t.hrough the nozzle) at pressures between 10 mTorr and 1 Torr. The flow through the nozzle was initiated and the temperature held at 900 "C for 2 minutes. The temperature was raised to 1000 and then lOS0 degrees for 10 minutes each before finally being raised to the growth temperature.

The temperature was monitored using a two-color optical pyrometer from IRCON operating near 1000 nm. Because the Sic is nearly transparent at this wavelength, an interference effect gave rise to apparent oscillations in the measured temperature whose period is proportional to the film thickness.

3. Results

Figure 1 shows the pyrometer output as a function of time for a typical deposition. Once deposition begins and until the oscillations die out the temperature is only approximately

r)

\

0 6 1 3 0 0 - . , . I . . I .

d) IR emission oscillations 8 1250

E

W

. .

B g 1200

5 3 1150

a

E 0 1000 2000 3000 4000 5(

Growth Time (seconds)

Figure 1. Temperature measured with an infrared pyrometer as a function of growth time. The actual temperature during deposition was 1214 "C.

IO

"4 Sic ~ r o w t h ~ a t e a p-.*-...

E, = 4.3 eV

"?. A Pb=20mTotr e.. o P,=l.OmTon

8 0.65 0.70

lOOO/T (K)

1000

100

Figure 2. Film deposition rate measured as a function of growth temperature. Square points: Pb = 1.0 mTorr; triangles, Pa = 2.0 mTorr.

known. A good estimate of the final temperature could be obtained by averaging the temperature in the first two valleys, and then averaging this result with the temper- ature of the first peak of the oscillation. This averaging technique suggests that the temperature does not change much during the course of a deposition.

Figure 2 shows the deposition rate as a function ofgrowth temperature for two dif- ferent pressures. The growth rate was obtained by measuring the time required to move between the first two valleys and the temperature using the method described above. While it should be possible to calculate the actual thickness based on the wavelength of the pyrometer and the index of refraction, the situation is complicated by the non- uniform film thickness. Circular thickness fringes show that the maximum film thickness occurs where the central portion of the beam impinges on the sample and decreases radi- ally from that spot. To calibrate this method of thickness determination, one deposition was performed with half of the substrate masked and the final thickness was measured with a profilometer yielding a value of 3000 A per cycle at the beam center. An activation energy of 4.3 eV was extracted from the data in Fig. 2. A levelling off of the deposition rate a t high temperatures suggest that the growth is flux limited. Therefore, the data with 1000/T 5 0.70 for Pl, = 1.0 mTorr together with the data for Pb = 2.0bmTorr was used in the determination.

The morphology and crystallinity of the films were examined using scanning elec- tron microscopy (SEM) and electron channeling. No difference was observed in either the morphology or crystallinity between films which were deposited directly on the Si surface using the above-described thermal ramp and those with an intermediate "buffer layer" formed with C3Hg and Hz. For films grown at T 2 1250 "C, the ramping process was critical. Without this initial step, the Si substrate was etched by in the presence of MS. In fact, profilometer traces of surfaces etched in this manner were a mirror image of the deposition profiles for Sic films.

Films deposited at T 5 1100 "C and T 3 1180 "C were single crystal, while those grown near 1150 "C were polycrystalline. The best quality films, both in terms of mor-

pliolog (smoothness) and sharpness of the electron channeling pattern, were obtained at intermediate growth temperatures under high (saturated) flux corresponding to the triangular points in Figure 2. Films grown at 1080 "C showed good morphology but poorer crystallinity while those at T 2 1200 O C and unsaturated flux had reasonable crystallinity but were more strongly faceted. This suggests that higher temperatures and hence higher growth rates may be achievable if the flux can be increased accordingly. Finally. under our deposition conditions no difference was noted in either crystallinity or morphology between films grown on aligned and miscut (4' along <110>) Si(OO1) substrates.

4. Futu re Work and Conclusions

In order to improve the morphology for thick films. it may be necessary to find substrates with a better lattice and thermal match to Sic. We have begun to investigate the growth of Sic on Tic, a cubic material with an excellent lattice match to Si (< 1% mismatch). Previous attempts at deposition on Tic [8] have had some success, although the results were dependent on the precursor gas and surface morphology was still a problem.

In summary, MJCVD appears to be a promising technique for obtaining single- crystal Sic films at high deposition rates, but much work remains to be done to charac- terize the process.

Acknowledgments

The authors gratefully acknowledge the financial support of the Office of Energy Man- agement, U.S. Department of Energy, under contract number DE-.4C05-840R21400 with Lockheed Martin Energy Systems.

References

[l] Sishino J, Powell J A and Will H A 1983 Appl. Phys. Lett. 42, 460-62 [2] Golecki I, Reidinger F and Marti J 1992 Appl. Phys. Lett. 60, 1703-5 [3] Steckl A J, Yuan C and Li J P1993 Appl. Phys. Lett. 63, 3347-9 [4] Takahashi K, Nishino S and Saraie J 1992 J. Electrochem. Soc. 139,3565-71 [5] Takahashi K, Xishino S, Saraie J and Harada K 1992 Amorphous and Crystalline Silicon

[SI IYahab Q, Sardela M R, Hultman L, Henry A, LVillander M, JantCn E and Sundgren J-E

[7] Miller D. R., in Atomic and Molecular Beam Methods vol 1, ed. by Scoles, G., Oxford

Carbide IV, Springer Proc. in Physics. 71, 78-83

1994 Appl. Phys. Lett. 65, 725-27

University Press, Oxford 1988, p. 24 [8] Kruaval G B and Parsons J D 1994 J. Electrochem. SOC. 141, 765-71; Parsons J D and

Kruaval G B 1994 J . Electrochem. SOC. 141, 771-7

A