Journal of Crystal Growth 8 (1971) 165—171 © North-Holland Publishing Co.
GALLIUM DOPED EPITAXIAL SILICON P. RAE-CI-IOUDHURY Westinghouse Research Laboratories, Pittsburgh, Pennsylvania 15235, U.S.A.
Received 20 August 1970; revised manuscript received 9 September 1970
Gallium doped epitaxial silicon is grown by the pyrolysis of silane and trimethyl gallium (TMG) in an rf heated furnace. Effects of growth temperature and partial pressure of TMG on the extent of carbon contamination of the films are examined thermodynamically as well as experimentally. The temperature and the layer doping ranges covered are 1220 to 1430 °Kand 3 respectively.The overalldoping 1.3 x 10’” to 6.5 < 1018 atoms/cm
process is found to be exothermic with a heat of reaction of —89 kcal/mole of gallium incorporated in silicon. Resistivity fluctuations are found to decrease with increasing doping level. Structural and electrical inhomogeneities of these layers are cornpared with gallium doped bulk Si. Transmission electron micrographs of the heavily doped epitaxial layers indicate presence of precipitates similar to those found in gallium diffused silicon.
1. Introduction
2. Experimental
Epitaxial silicon is commonly produced either by hydrogen reduction of SiC1 4 or pyrolysis of S1H4 in an rf heated furnace. To date, boron has been the primary source of p-type doping of epitaxial as well as diffused junctions. However, boron is not a preferred dopant in that its atomic size is about 25 % smaller than that of silicon andasthe dopant stress at 1500 as much seven foldinduced in excess of the yield°Kcan stress be of silicon1’2). Thus, massive dislocation generation3) is usually observed from heavy boron doping. Also, unlike other group iii dopants, boron is known to form compounds4) with silicon such as SiB 4 and SiB6. Therefore, a need exists for developing dopant sources other than boron. Gallium is a preferred dopant in silicon because of better atomic size matching (mismatch 7.7 %) and lower diffusivity than boron. Due to the absence of stable hydrides of gallium, the dopant may be transported either as vapor from a heated elemental source, as halides, or as organometallics such as trimethyl galhum ((CH3)3Ga). Trimethyl gallium has vaporispres5) awhich sufsure of 0.29 at room temperature ficiently highatm to dope the epitaxial silicon to the limit of solid solubility. The present study consists of doping of epitaxial silicon with gallium using trimethyl gallium and examination of the layers electrically as well as metallurgically.
A conventional horizontal rf heated reactor having a SIC coated graphite susceptor was used for the epitaxial growth. Silane or silicon tetrachloride was used for the deposition of epitaxial layers. Silicon substrates used were n-type, 5 ohm-cm, oriented 20 off (111), and chemically polished on both sides. The doping profiles 6). were measured by thefilm spreading resistance technique The doped epitaxial surfaces were examined by a Sirti etching technique7) using light microscopy. The films were also evaluated by transmission electron microscopy (TEM) to reveal film quality and possible clues to the reactions taking place in the system. The levels of carbon and gallium in the films were checked by a spark source mass spectrometer. The doping of epitaxial Si by the TMG being a new process, the rectification characteristic of the grown junctions were checked by examining their current-voltage curves. 3. Results and discussion A brief thermodynamic examination of the system we are about to discuss is instructive at this point. In the temperature range of interest to the silicon epitaxial technology, trimethyl gallium (TMG) is likely to decompose to CH 4(g) and Ga(l) which may or may not further react. The following reactions may be written:
165
166
P. RAI-CHOUDHURY
(CH3)3Ga(g)+~H2(g)
=
CH4(g)
=
Ga(/)-F-3CH4(g)
C (graphite) —F2H2(g)
[I] [2]
formation at the higher temperature (PTMG > and no carbon contamination at the lower temperature equli
~PTMG < PTMG
C (graphite)+Si(s) Ga(/)
=
=
SiC(s)
Ga (dissolved in Si).
[3] [4]
Due to the instability of TMG it may be assumed that reaction [1] goes to completion. The mechanisms and thermodynamics of SiC formation different 8). Itunder can be stated conhere ditions are discussed elsewhere that for reaction [2] to proceed, the equilibrium partial pressure of CH 4 should be 4.2>( l0~ atm at 1400 °K and 1.6>( l0~ atm at 1600 OKR) This requiles TMG partial pressures of (assuming l00°/~decomposition) 1.4>< l0~ atm and 5.2 x l0~ atm, respectively. If the TMG partial pressure is below 1.4x l0~ atm at 1400 °K,for example, then reaction [2] is thermodynamically not feasible, and carbon contamination in the form of graphite or SiC is not likely. if, on the other hand, reaction [2] proceeds then the subsequent formation of SiC, although thermodynamically feasible, is rather slow kinetically and would probably require a subsequent treatment or a higher reaction tem1 O). heat Reaction [4] involves the heat of solution perature of Ga in Si for which no measured data exists but may 1) to get an order be computed estimate. from Weiser’s of magnitude This ismodeP found to be 11 kcal/mole and is therefore an endothermic process. If liquid Ga from reaction [I] remains on the Si surface for an extended period of time (i.e., if dissolution process is relatively slow) then Ga could react with Si causing effects such as alloying or surface pitting. In order to study the interaction between TMG and the silicon substrate surface, only TMG was passed over the substrates in an H 2 ambient. At high TMG partial pressures this resulted in the formation of a skin of )~3-SiC,heavily doped with Ga. For example, at 1600 °Kand a TMG partial pressure of 7.0 x l0~atm, the SiC formed had a Ga content comparable to that of carbon from the SiC matrix (i.e., Ga to SiC ratio was unity). This SiC skin became thinner with decreasing TMG partial pressure and decreasing temperature. At a temperature of 1370 °Kand a TMG partial pressure of 7.0>< 10 ~ atm, the SiC skin was no longer detectable visually as well as by electron microscopy. The results are as expected from the thermodynamic considerations; namely, carbon contamination or SIC
Initially, S1CI4 was used to grow the doped epitaxial layers. In the presence of SiCl4, TMG partial pressures of up to 7.Ox 10 atm was used at 1370 K without any significant doping effects or carbon contamination. When the temperature was increased to about 1600 °K, the resultant Si film analysis) containedand about carbon (by mass spectrometric only1 a%trace of Ga. It is noteworthy that this Si film also indicated presence of species such as SiC, Si2C, Si2C2, and Si2C3. The reason for the unexpectedly low Ga levels in these films was attributed to the fact that TMG and SiCl4 may react at room temperature in the gas lines making the Ga unavailable for transport to the reaction zone. The carbon bearing species, on the other hand, does not seem to be trapped anywhere, and the carbon contamination is as expected. As a result of possible undesirable interaction between SiCl4 and TMG, all the subsequent data were collected using silane (SiH4) which proved to be quite satisfactory. 1 shows of dopant pressure on Figure the doping leveltheofeffect Si films grownpartial from SiH (pstH4 1.7 xconcentration l0~ atm) at of 13704 °K.The Ga in athetemperature epitaxial layer reaches a maximum value ofabout 4.0 x 1018 atoms/cm3 with increasing TMG partial pressure, and this is accompanied by a gradual deterioration of the epitaxial =
I
I A
~ ~ ~
o Spreading Resistance -
—
~
A
Mass Spectrenetry
-
~
,~i7 I
106
io~
I I
1o~
~CH3I3Ga .
.
.
Fig. 1. Effect oftrimethyl gallium partial pressure on the doping level of epitaxial Si grown at 1370 °K; PSIHI = l.7x i0~ atm.
167
GALLIUM DOPED EPITAXIAL SILICON
layer quality. Optical photomicrographs of these surfaces are shown in fig. 2. The deposition temperature and dopant partial pressures are such that significant carbon contanhination is not expected from a thermodynamic point of view. Since the concentration of the electroactive gallium is also below the solid solubility limit (~3.5x 1019 atoms/cm3), it was suspected that a possible cause of such surface pitting might have been due to the presence of excess Ga at the growth interface. Excess liquid Ga at the surface could have an alloying or sintering effect during the growth. (This type of pitted surface (fig. 2c) is very similar to the Si surface caused by Al sintering and may be revealed by Sirtl etching Si surface after dissolving the Al.) Mass spectrometric analysis of these surfaces sometimes mdicatedGaconcentrationsashighas 1.9 x l020atoms/cm3 and therefore might involve considerable lattice strain. These layers were then analyzed for bulk values of Ga and C using a spark source mass spectrometer. The levels of Ga thus obtained are also shown in fig. I along with the spreading resistance values. The absolute concentration of Ga (i.e., the mass spectrometric value) is identical with the concentration of electrically active Ga, except at the high doping level. Since with increas-
ruled out as long as the dopants are present substitutionally. In other words, some Ga atoms are present as interstitials, clusters and other forms that are not . .
~ -
~ ~,
‘
-
~,
-
-
-
-
~ P Fig. 2b.
p,ciu 3i3o.~-
-
1.3
10
‘1itrn.
ing doping level, the ionization energy is expected to approach zero, a possible deionization mechanism is
-
Fig. 2a.
~
pc3,o~ — l.3~ l0” atm.
Fig. 2c. Picii3i3c.u 1.3 l0~ atm. Fig. 2. Optical photomicrographs of Ga doped epitaxial Si, Sirti etched I mm; these layers correspond to the data points shown in fig. I.
168
P. RAI-CHOUDHURY
-, -
I
8,
--
L~. 1:1g. 3a.
-
Czochralski grossn bulk Si v,ith Ga le~elat 3.0 3. atoms,cm
10’
Fig. 3h.
1,0=
Epitaxial Si with elcctroacti~eGa level at 5.5 :1017 atoms/cm3 [‘(cii 1.3 10—” atm.
I
r, -
-_4 *V. ___________________________
Fig. 3c.
~,
__
-
-__
1
0F
—H -
Fpitaxial Si with electroactivc Ga level at 4.0 :1018 Fig. 3d. Epitaxial Si s’.ith elcctroactivc Ga le~elat 4.0>: 3 Plc-H,). ~ —— 1.3>. 10—’ atm. atoms/cm3: PicH,~c.a 1.3 .. l0~ atm. 10i8 atoms/cm Fig. 3. Transmission electron micrographs of doped Si wafers showing precipitates.
acceptors. It is noteworthy that the results of doping by ion implantation i 2.13) also indicate that most group w and v dopants arc not all present substitutionally in Si. Of course, ~~henthe solid solubilit~is exceeded and precipitation occurs, the absolute concentration of Ga will he considerably higher than the electroacti~clevel. The levels of carbon ~sere typically in the region of
10’ ~ atoms, cnr~and arc considered rather high’ °), al— though no undesirable electrical effects of such high levels of carbon ha~e been reported. There is no in— dication of significant increase in the level of carbon with increased doping level. The Ga doped epitaxial layers were then examined by reflection electron diffraction (RED). The diffrac-
169
GALLIUM DOPED EPITAXIAL SILICON
tion patterns obtained were those of normal, good quality, single crystal Si and did not provide any clues to the increased defects with higher doping level. These layers were then studied by transmission electron microscopy (TEM) and compared with Ga-doped bulk Si. Figure 3 shows transmission electron micrographs of Czochralski grown bulk St with Ga level at 3 Ox 1017 atoms/cm3 together with epitaxial Si with Ga levels as shown in fig. I. The lightly doped sample compared well with bulk sample (fig. 3b and a) in that they are relatively free from precipitates. The precipitates increased in size with increased gas phase doping level and did not give any diffraction pattern. Presence of any graphite or SiC also could not be detected. The precipitates virtually remained unchanged on heating in the electron microscope up to 700 °Cand appeared to be of amorphous nature. It is inferred that the precipitates, formed on increasing the gas phase doping level, are Ga induced and probably consist of pure Ga since there is essentially no terminal solid solubihity of 19
___________________________________________
10
/ I
/ 10
/ 10
17
--
/
-
16 10
/
/
/
/
/
/
/
/
/
/o
0.7 104/0,
Fig. 4.
/
/
/
/
/
/
1
OK’
Temperature dependence of gallium concentration in epitaxial Si; PTMO = 1.0x10” atm.
9
~~l .~
-
~ 0°.
10
-
~
0°
o
.~
0 0
0
00
00 00
00 OOo
uP
~ ~ ~ I
1016 ~
10
20
30
40
50
Depth. microns . Fig. 5. Electroactive Ga concentration of a Czochralski grown bulk Si as a function of depth (i.e., a longitudinal profile) showing
an apparent concentration fluctuation.
Si in Ga. It should be mentioned that the Ga induced precipitates in the present study are morphologically very similar to those observed by Knopp and Stickler’4) on Ga diffused Si. Figure 4 shows the effect of deposition temperature on the gallium concentration in the epitaxial layer at a TMG partial pressure of 1.Ox 10-6 atm. The dependence is exponential giving a value of AH equal to —89 kcal which is the sum of the heats of reactions [11 and [4]. The overall doping reaction is thus an exothermic process. The enthalpy of reaction [4], i.e. AH 4 or AHS, is the heat of solution of gallium in Si, and is obtained by subtracting the heat of reaction [1] from the measured value of —89 kcal. Limited thermochemical data on TMG do not, however, permit calculation of the heat of solution of gallium. Resistivity microinhomogeneities of the epitaxial layers were examined using a high resolution spreading resistance probe and compared with bulk material. Figure 5 shows the electroactive Ga concentration as a function of depth (longitudinal profile) of a Czochralski grown bulk Si (see also fig. 3a). Figure 6 shows a similar profile of an epitaxial Si of comparable doping level (see fig. 3b for the corresponding TEM). The bulk Si shows a slightly greater resistivity fluctuation that the epitaxial Si which in general is the case with most common dopants. In the present study these resistivity fluctuations were found to decrease with increase in the doping level. Figure 7 shows the electroactive Ga concentration as a function of depth of an epitaxial Si
170
P. RAI-CHOUDHURY
doped heavily with Ga until precipitation occurred (see figs. I and 3d). This seemed to be an apparent contradiction; namely the sample with the heaviest precipita10 18
-
tion showed thetoleast resistivity inhomogeneities. Before attempting explain these resistivity fluctuations
~
it is necessary to elaborate on the general observations made. First of all, the damping of the resistivity fluc-
E 0
lOU’ 16
0
Njype
-
ype
.
10 -
io15 -
0
0
-
io14 0
10
20
‘0
Depth, microns . . Fig. 6. Impurity concentration profile of an epitaxial Si. otcomparable Ga doping level as the bulk Si of fig. ~,
10 19
solubihity matrixinreaches a saturation limit level,the andhardening therefore, of thethe variation the contact area with the spreading resistance probe will also decrease or disappear. In general (for cases of B, Ga
0
io18
and As) the size of the observed precipitates are of the
-
-
~ io17
order of a fraction of a micron which is rather large to contribute significantly to further hardening of the matrix. Since the tip diameter of the spreading resistance probe is 8 microns, it is unlikely that these precipitates have any significant effect in our observed resistivity fluctuations. An attempt was made to measure microhardness of the epitaxial layers as a function of doping level using a Knoop hardness test and a 100 g load. It is difficult to attach any numerical values for
0
-
o
N-Type PType
--
,~
io16
--
10 15
-
tuations with increasing doping levels has been ohserved consistently not only for Ga doped epitaxial Si, but B and As doped epitaxial Si as well. Onset of precipitation is not particularly connected with this. In the case of B doping, resistivity fluctuations were found to be completely damped near the solid solubihity limit which was then followed by a drastic increase in fluctuations due to the formation of SiB4 precipitates4) on a rather massive scale. Such phase transformations, of course, were not possible for the cases of Ga and As doping (e.g., SiAs is relatively unstable4)) and consequently, resistivity fluctuations once damped out could not be increased again. A possible explanation for the variations of resistivity fluctuations is that with increase in the doping level solid solution hardeiiing is taking place which causes variation in the extent of contact with the spreading resistance probe. Near the solid
heavily doped layers due to extensive fracture occurring during measurement. However, a positive indication of
-
I
I
Depth, micr~s
I 50
Fig. 7. Impurity concentration profile of an epitaxial Si doped heavily with Ga (PTMG = 1.3 x 10— atm) showing damping out of the apparent concentration fluctuations.
increased hardness with increased doping level was observed. In the presence of electric current (current density i~0.0lmA/mm2) considerable softening (as much as 60%) of Si has been observed by Westbrook and Gilrnan15). This softening effect probably also reaches saturation under the conditions of measure‘
ment
(current
4
density ~2 x 10 mA/mm
2
) in heavily
GALLIUM DOPED EPITAXIAL SILICON
doped material. Consequently, if both the solid solution hardening and the softening due to electromechanical effects reach saturation at heavy doping, a damping out of resistivity fluctuation may be expected. A better understanding of the process may be gained through a direct measurement of microhardness under a given electric current as a function of doping level. Diodes were fabricated from the Ga doped epitaxial Si and normal diode characteristics were obtained even with the heaviest doped material (see fig. 3d). The epitaxial layers being more heavily doped than the substrate (donor level in the substrates is I x 10’ atoms/ cm3); this is rather an insensitive test for layer quality. 4. Summary The possibility of using organometallics at relatively high temperatures for doping epitaxial silicon is analyzed thermodynamically and demonstrated using tnmethyl gallium (TMG). An acceptable quality of epitaxial layers is obtained by proper choice of TMG partial pressure and deposition temperatures. At high doping levels the absolute concentration of Ga is found to be greater than the concentration of electroactive Ga, and this difference increases with the doping level. The overall process of dopant incorporation is found to be an exothermic process with gallium concentration increasing with decreasing temperature. Resistivity microinhomogeneities are found to decrease with increasing doping level and do not seem to be influenced by the presence of precipitates.
171
Acknowledgment The author wishes to thank Dr. A. J. Noreika for assistance with the electron microscopic work, W. Cifone and R. J. Pfeil for assistance with the experimental work, and Dr. J. F. Johnson for helpful discussions. References I) G. L. Pearson, W. T. Read, Jr. and W. L. Feldmann, Acta Met. 5 (1957) 181.
2) J. U. Lawrence, J. AppI. Phys. 37 (1966) 4106.
3) E. Levine, J. Washburn and G. Thomas, J. AppI. Phys. 38 81. 4) (1967) P. Rai-Choudhury and F. 1. Salkovitz, J. Crystal Growth 7 (1970) 353, 361. 5) G. E. Coated, M. L. H. Green and K. Wade, Organometallic Compounds (Methuen, London, 1967) p. 343. 6) R. G. Mazur and D. H. Dickey, J. Electrochem. Soc. 113 (1966) 255. 7) E. Sirtl and A. Adler, Z. Metallk. 52 (1961) 529. 8) P. Rai-Choudhury and N. P. Formigoni, J. Electrochem. Soc. 116 (1969) 1440. 9) JANAF Thermochemical Tables (Dow Chemical Co., Midland, Michigan, Aug. 1965). 10) P. Rai-Choudhury, A. J. Noreika and M. L. Theodore, J. Electrochem. Soc. 116 (1969) 97. 11) K. Weiser, J. Phys. Chem. Solids 7 (1958) 118. 12) J. A. Davis, J. Denhartog, L. Eriksson and J. W. Mayer, Phys. 45 4053.S. T. Picraun and J. A. Davis, 13) Can. J. W. J.Mayer, L. (1967) Eriksson, Can. J. Phys. 46 (1968) 663. 14) A. N. Knopp and R. Stickler, J. Electrochem. Soc. 11 (1964) 15) J.1372. H. Westbrook and J. J. Gilman, J. AppI. Phys. 33 (1962) 2360.