Effects of ligand exchange reactions on the composition of Cd1-yZnyTe grown by metalorganic vapor-phase epitaxy

Article abstract of DOI:10.1021/jp970062k

The metalorganic vapor-phase epitaxy (MOVPE) of cadmium zinc telluride (Cd_1-yZn_yTe) from dimethylcadmium (DMCd), dimethylzinc (DMZn), diethylzinc (DEZn), and diisopropyltelluride (DIPTe) was studied using on-line infrared spectroscopy to monitor the feed and effluent gases. The film composition was measured by X-ray diffraction. No zinc was incorporated into the film when DMCd and DMZn were used due to the very low reactivity of the latter compound. When DMCd and DEZn are tried, the films were nonuniform with Cd-rich films deposited at the reactor inlet and Zn-rich films deposited near the reactor outlet. This film profile was due to alkyl ligand exchange reactions between the group II precursors in the feed, producing DMZn, methylethylzinc (MEZn), methylethylcadmium (MECd), and diethylcadmium (DECd). The decomposition rates of these precursors vary over a wide range with DECd reacting at a 250 K lower temperature than DMZn. Since the organocadmium compounds were consumed at a much faster rate, CdTe was deposited first, while ZnTe was deposited downstream. The ligand exchange reactions explain why previous workers found it difficult to grow Cd_1-yZn_yTe alloys of uniform composition by MOVPE.

Full text of DOI:10.1021/jp970062k

4882
J. Phys. Chem. B 1997, 101, 4882-4888
Effects of Ligand Exchange Reactions on the Composition of Cd1-yZnyTe Grown by
Metalorganic Vapor-Phase Epitaxy
Menno J. Kappers, Kerri J. Wilkerson, and Robert F. Hicks*
Department of Chemical Engineering, 5531 Boelter Hall, UniVersity of California at Los Angeles,
Los Angeles, California 90095-1592
ReceiVed: January 3, 1997; In Final Form: March 14, 1997X
The metalorganic vapor-phase epitaxy (MOVPE) of cadmium zinc telluride (Cd1-yZnyTe) from dimethyl-
cadmium (DMCd), dimethylzinc (DMZn), diethylzinc (DEZn), and diisopropyltelluride (DIPTe) was studied
using on-line infrared spectroscopy to monitor the feed and effluent gases. The film composition was measured
by X-ray diffraction. No zinc was incorporated into the film when DMCd and DMZn were used due to the
very low reactivity of the latter compound. When DMCd and DEZn are tried, the films were nonuniform
with Cd-rich films deposited at the reactor inlet and Zn-rich films deposited near the reactor outlet. This
film profile was due to alkyl ligand exchange reactions between the group II precursors in the feed, producing
DMZn, methylethylzinc (MEZn), methylethylcadmium (MECd), and diethylcadmium (DECd). The
decomposition rates of these precursors vary over a wide range with DECd reacting at a 250 K lower
temperature than DMZn. Since the organocadmium compounds were consumed at a much faster rate, CdTe
was deposited first, while ZnTe was deposited downstream. The ligand exchange reactions explain why
previous workers found it difficult to grow Cd1-yZnyTe alloys of uniform composition by MOVPE.
Introduction
present, none of this previous work fully explains why the
segregation of zinc in the film changes so dramatically with
temperature.
In the manufacture of large-area infrared focal plane arrays
(IRFPAs), it is highly desirable to deposit the HgCdTe photo-
diodes directly onto GaAs (001) or sapphire (0001) substrates.1-7
This is done in a two-step process, where Cd1-yZnyTe is first
deposited on the GaAs or Al2O3 by vapor-phase epitaxy, and
then Hg1-xCdxTe is deposited on the buffer layer by liquid-
phase epitaxy. In order to minimize the number of defects in
the film, the Cd1-yZnyTe buffer layer is lattice matched to the
Hg1-xCdxTe active layer by incorporating between 4.1 and 3.6%
zinc in the film. It is essential that the Cd1-yZnyTe alloy be
highly uniform in composition, thickness, and crystallinity, as
slight variations in the zinc content (>0.1%) can seriously
impact the quality of the Hg1-xCdxTe epilayer. This makes
composition control a critical issue.
Previously, Ahlgren et al.8 and Kisker9 have studied the
MOVPE of Cd1-yZnyTe using DMCd, DEZn, and diethyltel-
luride (DETe) and compared the composition in the vapor-phase
to the solid-phase composition. In Ahlgren’s study, the deposi-
tion temperature was 714 K, and it was observed that the zinc
preferentially incorporates into the film for all ratios of DMCd
to DEZn in the gas inlet. Kisker studied the growth at lower
temperatures and obtained very different results. At 653 K,
zinc does not incorporate into the film below a group II vapor
composition of 30% DEZn, and nearly pure ZnTe is deposited
when the vapor composition exceeds 50% DEZn. This makes
it very difficult to achieve a uniform zinc composition of only
4% over the entire substrate surface.
In this study, on-line infrared spectroscopy was used to
monitor the gas composition of the reactor feed and effluent
during MOVPE of CdZnTe. It was discovered that exchange
reactions occur between the alkyl ligands of DMCd and DEZn
in the feed lines on the way to the reactor. This discovery
explains the origin for the unusual alloy segregation behavior
in II-VI MOVPE. The results of this study are described
below.
Experimental Section
The experimental apparatus and procedures used for this study
have been described previously in detail.11,12 An atmospheric-
pressure MOVPE reactor was used to deposit Cd1-yZnyTe on
the inner walls of a glass tube. The electronic-grade sources,
DMCd, DMZn, DEZn, and DIPTe, were contained in 200 cm3
stainless steel bubblers immersed in thermal regulating baths
at 273, 258, 288, and 306 K, respectively. The precursors were
vaporized into helium, which were then diluted further with
helium or hydrogen to a total flow rate of 400-800 cm3 min-1
and fed to the reactor. The carrier gases, He (99.995%) and
H2 (99.995%), were purified by deoxygenation traps and 13 X
molecular sieves held at 196 K.
Using on-line infrared spectroscopy, the consumption rates
of the organometallic compounds were determined.12 Infrared
spectra of the reactor feed and effluent were collected by passing
the gas stream through a small flow cell, 17.7 cm long and 1.6
cm in diameter, and sealed with KBr windows. The data were
acquired with a Digilab FTS-7 IR spectrometer, equipped with
a wide-band MCT detector. The spectra shown herein were
obtained by signal averaging 64 scans at 4 cm-1 resolution. In
Figure 1, the infrared spectra are shown of the gas feed and
effluent from the reactor during Cd1-yZnyTe growth at 673 K
from DMCd, DMZn, and DIPTe. The intense vibrational bands
of the organometallic sources detected between 1600 and 450
cm-1 are listed in Table 1 with the band assignments.12-15 The
In a recent paper, Kappers et al.10 proposed a kinetic model
for alloy growth. This model assumes that the DEZn and
DMCd compete for the same sites on the film surface and that,
after desorption of the alkyl groups, the Cd and Zn atoms may
either incorporate into the film or sublime off the surface. This
model does fit Ahlgren’s data but failed to fit Kisker’s. At
* To whom correspondence should be addressed.
X Abstract published in AdVance ACS Abstracts, June 1, 1997.
S1089-5647(97)00062-X CCC: $14.00 © 1997 American Chemical Society

Effects of Ligand Exchange Reactions on Cd1-yZnyTe
J. Phys. Chem. B, Vol. 101, No. 25, 1997 4883
inner wall. Samples from the reaction zone of the tube were
analyzed by scanning for the (111) and (331) peaks. The 2θ
ranges scanned were 22° to 26° and 61° to 68° by steps of 0.01°.
Using Vegard’s law and the measured peak position, the
Cd1-yZnyTe film composition was calculated using13
a - aCdTe
aZnTe - aCdTe
y )
(1)
where a is the lattice constant (in angstroms) found by Bragg’s
law, aZnTe is the lattice constant of ZnTe (6.1004 Å), and aCdTe
is the lattice constant of CdTe (6.4823 Å). The peak position
was taken to be the 2θ corresponding to the maximum intensity
value after the diffraction pattern was smoothed. The value of
y is calculated to within 5% error. In addition, the postdepo-
sition region of the tubes was analyzed to determine whether
cadmium or zinc metal was present. These samples were
analyzed over a 2θ range of 10°-100° by 0.1° steps.
Figure 1. Infrared spectra of the reactor feed and effluent during
Cd1-yZnyTe MOVPE with DMCd, DMZn, and DIPTe in hydrogen at
673 K, 4 × 10-4 atm of DIPTe, a II/VI ratio of 1.0, an x value of 0.3,
and a total flow rate of 200 cm3/min.
The film thickness in each 1 cm section of the tube was also
determined. The mass of the film was calculated from the
difference in weight of the glass piece before and after etching
it with a dilute bromine and methanol solution. Then the film
area was obtained by multiplying the mass of the glass piece
times the unit area per mass of the glass tube. Finally, the film
thickness was calculated by dividing the film mass by its area
and X-ray density. This method yielded about (20% error in
the thickness measurement. Film compositions averaged over
the entire reactor were calculated by summing up the composi-
tion to weight ratio of each piece and multiplying this sum by
the total film weight.
absorption bands are mostly well resolved from each other and
provide a means of independently monitoring the consumption
of each molecule during MOVPE.
Further inspection of Figure 1 reveals that the peak intensities
of the organometallic compounds in the reactor effluent are
much smaller than those of the feed and that additional sharp
peaks appear in the spectrum of the effluent at 1300 and 910
cm-1. These latter peaks correspond to the vibrational modes
of the hydrocarbon products methane and propylene. Using
the Beer’s law, the intensities of the bands for the individual
organometallic compounds can be converted into gas-phase
concentrations to within 5% accuracy. The bands used are the
metal-carbon stretching modes for DMCd, DMZn, and DEZn
at 544, 621, and 572 cm-1, respectively, and the intense skeletal
Results
mode for DIPTe at 1148 cm-1
. The corresponding molar
Ligand Exchange. Infrared spectroscopic analysis of a
mixture of the group II precursors, DMCd and DEZn, reveals
irregularities as compared to the individual compounds. Figure
2 shows the infrared spectrum of the mixture of DMCd and
DEZn as a function of the initial feed composition, x, defined
as
absorption coefficients are (57 ( 5) × 103, (76 ( 2) × 103,
(57 ( 2) × 103, and (100 ( 6) × 103 cm2 mol-1 for DMCd,
DMZn, DEZn, and DIPTe.12 The consumption rates of the
precursors are determined during film deposition by monitoring
the change in the peak intensity of the infrared bands from the
inlet to the outlet of the reactor. Using the ideal gas law, the
flux of the reactant in the feed is calculated from the total flow
rate and the calibrated partial pressure of organometallic
compound. The change in IR peak intensity between the feed
and the effluent reflects the conversion of the reactant, and hence
its consumption rate is known.
PDEZn
x )
(2)
PDMCd + PDEZn
where PDMCd and PDEZn are the partial pressure (in atmospheres)
of DMCd and DEZn originally fed into the reactor. Surpris-
ingly, the spectra in the figure do not show the expected
summation of DMCd and DEZn absorption bands at different
relative intensities. Instead, the DMCd and DEZn absorption
The films were analyzed using a θ/2θ scan on a Crystal Logic
powder diffractometer. The glass tube was sectioned into 1
cm pieces, which were split axially to reveal the film on the
TABLE 1: Assignment of the Infrared Bands Observed for the Organometallic Compounds between 1600 and 450 cm-1 (Refs
12-15)a
DIPTe
DMCd
DMZn
DEZn
freq
assignt
ν(M-C)
freq
527 (s)
545 (s)
assignt
freq
604 (s)
621 (s)
assignt
freq
572 (s)
612 (br)
assignt
504 (vw)
874 (vw)
923 (vw)
1000 (w)
1019 (vw)
1148 (s)
1197 (s)
ν(M-C)
ν(M-C)
ν(M-C)
ν(M-C)
ν(M-C)
F(CH2)
ν(C-C-C)
F(CH3)
}
709 (br)
F (CH3)
706 (br)
F(CH3 )
957 (vw)
986 (m)
ν(C-C)/F(CH3)
ν(C-C)/ (CH3)
}
}
}
1944 (vw)
1188 (w)
1234 vw)
1376 (sh)
1383 (m)
1421 (w)
1469 (w)
δ(CH3)
ω(CH2)
c/o
993 (vw)
1169 (vw)
1305 (w)
1327 (w)
c/o
1030(vw)
1175 (vw)
1192 (w)
1304 (w)
c/o
1303 (w)
1372 (m)
1462 (m)
δ(CH3)
δ(CH3)
δ(CH3)
δ(CH3)
}
}
}
a Abbreviations: vw ) very weak, w ) weak, m ) medium, s ) strong, br ) broad, sh ) shoulder, c/o ) combination/overtone.

4884 J. Phys. Chem. B, Vol. 101, No. 25, 1997
Kappers et al.
TABLE 2: Assignment of the Infrared Bands of the
Organometallic Compounds between 1000 and 450 cm-1
peak position (cm-1) of the organometallic compound
assignt DMZn
MEZn
DEZn
572
DMCd
MECd
DECd
504
νM-C
νM-C
604
621
704
589
601
655
523
544
702
512
525
655
δCH
612
650
3
< x < 0.85) the residual IR spectrum represents a combination
of the reaction products DECd, MECd, and MEZn. At higher
and lower values of x, the residual spectrum contains only two
of the three products, i.e., DECd and MEZn at x > 0.85 and
MECd and MEZn at x < 0.30. The spectrum of MEZn was
obtained by subtracting the DMZn spectrum from the IR
spectrum of the reactor effluent at 625 K which contained only
DMZn and MEZn (as described below). The spectrum of
MEZn was used to obtain the IR spectra of DECd and MECd
from those of the reaction mixtures at x ) 0.9 and 0.1,
respectively. The band positions in the spectral region between
800 and 450 cm-1 and the proposed assignments for all
components are summarized in Table 2.
Figure 2. Infrared spectra between 850 and 450 cm-1 of the group II
organometallic compounds in the reactor feed during Cd1-yZnyTe
MOVPE at varying ratios of DMCd to DEZn (x ) PDEZn/(PDEZn
PDMCd)).
+
Next, the concentration of each organometallic compound was
determined as a function of the initial feed composition, x. Since
we were unable to calibrate the infrared band intensities of
MEZn, MECd, and DECd, their concentrations were calculated
as follows. Consider the reaction equilibrium
aDMCd + bDEZn H
pMECd + qMEZn + rDECd + sDMZn (3)
The values of a, b, and s are determined from the intensities of
the infrared bands and the corresponding molar absorption
coefficients. In addition, the feed concentrations of DMCd and
DEZn, af and bf, are known. A mass balance on the methyl
ligands yields
2(af - a) ) 2Aaf ) p + q + 2s
(4)
where A is the conversion of DMCd. A mass balance on the
ethyl ligands yields
Figure 3. Infrared spectra between 850 and 450 cm-1 of the group II
organometallic compounds present after mixing DMCd and DEZn. All
spectra are normalized to the same scale.
2(bf - b) ) 2Bbf ) p + q + 2r
(5)
where B is the conversion of DEZn. Mass balances on the Cd
and Zn containing species give
bands have decreased in intensity, and several new bands can
be discerned in the spectra. The new bands show similar
characteristics as those of DMCd and DEZn; hence, new
organometallic compounds have been formed from the reaction
of DMCd with DEZn.
af - a ) Aaf ) p + r
bf - b ) Bbf ) q + s
(6)
(7)
Spectral subtraction reveals the absorption bands of four
newly formed dialkyl species. By subtracting judiciously chosen
IR spectra from each other, the infrared spectra of the different
reaction mixtures can be deconvoluted into their components.
The infrared spectrum of each organometallic compound is
shown in Figure 3. One of the spectra is identical to that of
DMZn. The infrared bands of another product agree well with
those recorded previously for DECd.13 However, no literature
data are available for the other two species. These bands are
assigned to the mixed alkyl compounds MECd and MEZn, since
they are obvious products from the ligand exchange of DMCd
and DEZn. Nevertheless, the infrared assignment for these
species will be further justified below.
The concentration of MEZn immediately follows after rewriting
eq 7:
q ) Bbf - s
(8)
The concentration of DECd follows from the substitution of
eqs 6 and 7 into eq 4:
r ) Bbf - Aaf + s
(9)
After substitution of eq 9 into eq 6, the MECd quantity follows
from
The spectra of DECd, MECd, and MEZn were obtained from
those of the mixture as follows: The IR spectra of the
compounds DMCd, DEZn, and DMZn were subtracted from
the spectra shown in Figure 2, obtained after mixing DMCd
and DEZn at the different ratios. At medium values of x (0.30
p ) 2Aaf - Bbf - s
(10)
The exchange reactions may be assumed to be at equilibrium,
since varying the flow rate has no effect on the gas-phase
concentrations.

Effects of Ligand Exchange Reactions on Cd1-yZnyTe
J. Phys. Chem. B, Vol. 101, No. 25, 1997 4885
Figure 4. Fractional partial pressures of the group II organometallic
compounds in the reactor feed during Cd1-yZnyTe MOVPE at varying
x values.
Figure 6. Dependence of the consumption rates of the organometallic
compounds on the vapor composition during Cd1-yZnyTe MOVPE at
673 K and a II/VI ratio of 1.0: (a) DMCd, DMZn, and DIPTe; (b)
DMCd, DEZn, and DIPTe.
completely decomposed, while MEZn and DMZn remain
unaffected. The pyrolysis of MEZn and DMZn begins at 670
and 715 K, respectively. The shift toward more DMZn
production at higher temperatures and the high thermal stability
of the MEZn and the DMZn results in a low conversion of the
organozinc compounds. At 700 K, only 11% of the zinc species
in the feed is deposited, while all the organocadmium reactants
are converted.
In summary, the experiments verify that alkyl ligand exchange
occurs between DMCd and DEZn in the gas phase. From the
experiments conducted, there is no indication that the reaction
could have occurred on the walls of the stainless steel tubing;
the composition of the feed did not change with either the total
flow rate or the temperature (<370 K) of the tubing leading to
the IR cell. However, we cannot absolutely rule out that ligand
exchange is catalyzed by the tube wall. Of the newly formed
species in the feed, the DECd and MECd are much more
reactive than DMZn and MEZn. The reactivity of the precursors
increases in order: DMZn < MEZn < DEZn < DMCd <
MECd < DECd.
MOVPE of Cd1-yZnyTe Using DMZn. Films of Cd1-yZnyTe
were deposited using DMCd, DMZn, and DIPTe at temperatures
between 673 and 748 K and at II/VI ratios of 1.0 and 2.0. In
Figure 6a, the consumption rates of the organometallic com-
pounds are plotted as a function of the feed composition during
MOVPE at 673 K, a II/VI ratio of 1.0, and a DIPTe partial
pressure of 4 × 10-4 atm. The only precursors that are
consumed are DMCd and DIPTe. The cadmium species
consumed in excess of the DIPTe are deposited as a metal film
or particulates in the reactor outlet.11 In line with this, no zinc
metal is deposited downstream. The zinc from DMZn is not
incorporated into the film at any x value other than 1.0, at which
point pure ZnTe is grown.
During the growth of CdTe from DMCd and DIPTe (x ) 0),
the consumption rate of DMCd is greater than that of DIPTe
for the reaction conditions selected. With decreasing partial
pressure of DMCd, the consumption rate of DMCd remains
constant, while that of DIPTe increases. The consumption rates
cross each other between x ) 0.0 and 0.3. A similar change of
consumption rates with decreasing II/VI ratio has been observed
before during CdTe OMVPE.11 At x values higher than 0.5,
the consumption rates of DMCd and DIPTe both decrease as
the zinc mole fraction in the feed increases. This latter trend is
obviously due to the decline in the DMCd partial pressure.
MOVPE of Cd1-yZnyTe Using DEZn. Shown in Figure
6b are the consumption rates of DIPTe, all the cadmium sources
(depicted as DRCd), and all the zinc sources (depicted as DRZn)
during growth at 673 K, a DIPTe partial pressure of 1.1 × 10-3
atm, and a II/VI ratio of 1.0. The combined consumption rate
of the organocadmium compounds is higher in the mixture than
Figure 5. Fractional partial pressures of the group II organometallic
compounds in the reactor effluent at varying temperatures for an x value
of 0.5.
Equations 3-10 have been used to calculate the partial
pressures of MECd, DECd, and MEZn in the mixtures fed to
the MOVPE reactor. The calculated values agree well with the
trends in the infrared band intensities of these three compounds.
Shown in Figure 4 is the dependence of the fractional partial
pressure of each organometallic compound on the feed com-
position, x. The fractional partial pressure is taken to be the
partial pressure of the species divided by the total partial pressure
of all the group II species. On the left side are the curves
obtained for DMCd and DEZn. As expected, the fractional
partial pressure of DEZn increases as the value of x increases,
and the fractional partial pressure of DMCd decreases with
increasing x. Both curves, however, are concave-up because
their partial pressures in the mixture have decreased compared
to those initially in the feed. The partial pressures of DMZn
and MECd formed by the exchange reaction increase with x
and reach a maximum at x ) 0.3. The amount of MECd formed
is about twice that of DMZn at the maximum. The partial
pressures of MEZn and DECd slowly increase with x and attain
their highest value at x ) 0.6. Note that, at x values higher
than 0.6, the DMCd is almost completely converted into MECd
and DECd.
Shown in Figure 5 is the dependence of the fractional partial
pressures of the group II precursors in the reactor outlet on the
reactor temperature for a zinc mole fraction in the feed gas of
0.5. On the left side of the figure, the fractional pressures of
DMCd and DEZn are shown. On the right side, the products
of the ligand exchange reaction, DMZn, MEZn, MECd, and
DECd, are shown. The overall equilibrium does not change
significantly between 300 and 470 K, leaving the partial
pressures of each compound constant. However, the partial
pressures of all the precursors, except DMZn, decrease above
a temperature of 470 K. The sudden decrease in partial pressure
of the reaction components is attributed to the decomposition
of DECd, which starts at about 470 K. The original precursors,
DMCd and DEZn, only start to decompose at 570 and 630 K,
respectively. At 650 K, the Cd precursors and DEZn are

4886 J. Phys. Chem. B, Vol. 101, No. 25, 1997
Kappers et al.
a and b, respectively. A 10-fold increase in II/VI ratio and a
40 K increase in temperature have no significant effect upon
the trends observed. It should be noted that the film composi-
tions calculated from the infrared data produce segregation
curves in reasonably good agreement with the results presented
in the figure.
Discussion
Ligand Exchange. The analysis of the feed gases by infrared
spectroscopy reveals that the group II organometallic compounds
undergo rapid exchange of their methyl and ethyl ligands during
transport to the MOVPE reactor. Equilibrium between the
reaction components is reached within the time scale of mixing
in the feed lines and the IR gas cell (20 s). From the infrared
spectra in Figures 2 and 3, four of the six reaction components
can be readily identified, i.e., the two starting reactants, DMCd
and DEZn, and their fully exchanged counterparts, DECd and
DMZn. There is no precedent in the literature for the vibrational
spectra of the two remaining zinc and cadmium species.
However, the vibrational spectra of mixed alkyl complexes of
lead,16 germanium,17 and arsenic18,19 are well described. It was
observed that the values of the metal-carbon stretching
frequencies for the mixed alkyl species lie between the values
of the fully exchanged alkyl species. For example, the average
value of the degenerate Pb-C stretching frequency shifts
gradually to lower values with each consecutive exchange of a
methyl group for an ethyl group: from 469 cm-1 for tetrameth-
yllead, to 465 cm-1 for trimethylethyllead, to 459 cm-1 for
dimethyldiethyllead, to 456 cm-1 for methyltriethyllead, and
to 452 cm-1 for tetraethyllead.16 The average values for the
degenerate Cd-C and Zn-C stretching frequencies shift in an
analogous fashion: from 534 cm-1 for DMCd to 519 cm-1 for
MECd to 504 cm-1 for DECd and from 613 cm-1 for DMZn
to 595 cm-1 for MEZn to 572 cm-1 for DEZn (from Table 2).
In total, 14 different alkyl ligand exchange reactions occur
simultaneously in the mixture of DMZn, MEZn, DEZn, DMCd,
MECd, and DECd. The reactions occurring between zinc and
cadmium compounds with different ligands that lead to ligand
exchange are listed in Table 3. The self-exchange reactions,
of which there are six, and the exchange reactions of solely
methyl, or ethyl ligands, of which there are two, have been
omitted from this Table.
Figure 7. Dependence of the composition (left) and the thickness
(right) of the film on the relative axial position in the reactor during
Cd1-yZnyTe MOVPE with DMCd, DEZn, and DIPTe at 673 K, a II/VI
ratio of 1.0, and x values of 0.4 and 0.9.
Figure 8. Composition of the solid versus the vapor for Cd1-yZnyTe
MOVPE with DMCd, DEZn, and DIPTe for (a) II/VI ratios of 0.7 and
7.0 at 693 K and (b) growth temperatures of 653, 673, and 693 K at a
II/VI ratio of 1.0. The solid curves are the result of calculations
described in the text.
when DMCd reacts with DIPTe to form CdTe (i.e., compare x
) 0.3 to x ) 0.0). Moreover, the data points at x > 0.5
correspond to 100% conversion of the cadmium species. On
the other hand, the combined consumption rate of the zinc
species only slightly exceeds that of the DEZn during ZnTe
MOVPE (x ) 1.0). Once again, these results are indicative of
ligand exchange reactions which produce less stable organocad-
mium compounds, MECd and DECd, relative to DMCd, and
more stable organozinc compounds, MEZn and DMZn, relative
to DEZn. The excess cadmium plates out as a metal film in
the reactor outlet; no zinc metal is deposited downstream.
There is a distinct difference in the uniformity of the
Cd1-yZnyTe film grown at low and high x values. Figure 7
shows how the film composition (left) and the film thickness
(right) changes down the length of the reactor tube for x values
of 0.4 and 0.9 and a growth temperature of 673 K. For x )
0.4, a fairly uniform film containing about 4% zinc is deposited
throughout the heated zone. At the higher x value, an alloy
film containing 90% CdTe is deposited over the first 40% of
the reactor, while a film of 90% ZnTe is deposited over the
last 20% of the reactor. The film rich in CdTe at the inlet of
the reactor is almost 4 times as thick as the film rich in ZnTe
near the outlet. The transition from cadmium- to zinc-rich film
growth occurs at a point where all the cadmium-containing
species have been depleted from the gas phase. This point tends
to shift toward the inlet with increasing growth temperature.
Shown in Figure 8 are zinc segregation curves determined
from the weight-averaged composition of the films deposited
in the reactor. The solid curves in the figures are the result of
calculations to be described later in the text. The scatter of the
data is particularly large at the higher values of x, due to the
difficulties in obtaining accurate measurements of film composi-
tion and thickness down the length of the tube. Nevertheless,
it is evident that the composition of the Cd1-yZnyTe film does
not vary consistently with the II/VI ratio, or temperature, graphs
Ligand exchange between different group II organometallic
compounds was investigated during the late 1960s. For instance,
McCoy and Allred studied the reaction between DMCd and
DMZn in solution by nuclear magnetic resonance.20 The effect
of the solvent on the reaction kinetics for methyl exchange
between DMCd and DMZn and for interactions of DMCd with
group III organometallic compounds were investigated by Oliver
and co-workers.21,22 It was found that these reactions proceed
rapidly under mild conditions. The alkyl ligands transfer from
one metal center to the other via a four-centered transition
state:21,22
R
M
R
R
M
R
R
M
R
(11)
+
+ R′
R′
M′ R′
M′ R′
R′
M′ R′
where R, R′ are CH3 and/or C2H5 and M, M′ are Zn and/or Cd.
To our knowledge, the reactions listed in Table 3 have never
been studied before in the gas. We assume that the same
mechanism and thermodynamics hold for the reactions in the
1
gas as for those in solution. Recent H NMR measurements
on a 1:1 mixture of DMCd and DEZn in the liquid reveal that,
to within the experimental error, the same ligand exchange
products are formed at the same relative concentrations as that

Effects of Ligand Exchange Reactions on Cd1-yZnyTe
J. Phys. Chem. B, Vol. 101, No. 25, 1997 4887
TABLE 3: Alkyl Ligand Exchange Reactions Observable by
Infrared Spectroscopy
only a small amount of the cadmium but more of the zinc in
the gas feed was available for film growth, hence the preferential
incorporation of zinc in their films.
reaction
no.
Kisker9 obtained very different results. At 653 K, zinc was
not incorporated in the film below x ) 0.3, and nearly pure
ZnTe was deposited when the vapor composition exceeds x )
0.5. This S-shaped zinc segregation behavior tends to stretch
out and become more gradual as the growth temperature was
raised to 673 K. Kisker also found a composition gradient over
the length of the reactor. In spite of this, he chose to measure
the composition at only one point in the system. Our observa-
tions of a strong gradient in the film composition with changing
value of x in Figure 7 helps to explain the S-shaped segregation
behavior seen in Kisker’s data. If the location in the reactor
for film analysis had been closer to the inlet, then he would
have observed a cadmium-rich film over the complete composi-
tion range of the gas phase. However, the location in the reactor
chosen by Kisker for film analysis must have been at a point in
the reactor, where the film composition can change suddenly
with gas-phase composition from a cadmium-rich to a zinc-
rich film. The exact location of this point in a horizontal reactor
depends on the experimental conditions, such as temperature
and flow rate.
Modeling of the Cd1-yZnyTe growth kinetics can be quite
complicated with five reactive species in the feed (DMCd,
MECd, DECd, DEZn, and DIPTe). Before the exchange
reactions between the group II sources were discovered, we
proposed a model for the growth of Cd1-yZnyTe that is based
upon the following surface reaction mechanism: dissociative
adsorption of the organometallic compounds, desorption of the
alkyl ligands, film growth, and desorption of the excess zinc or
cadmium metal.10 This model predicted that the zinc segrega-
tion behavior would be relatively insensitive to the substrate
temperature and the II/VI ratio. However, it would be strongly
influenced by the intrinsic kinetic parameters, i.e., the difference
in the cadmium and zinc sublimation energies and the relative
sticking probabilities of the organometallic precursors.
The experimental results presented in Figure 8 are consistent
with the trends predicted by the model. In spite of the large
scatter of the data, it is evident that the distribution of zinc and
cadmium between the phases is insensitive to the process
conditions. According to our model, if the heats of sublimation
of Cd and Zn are equal, then the alloy composition is determined
by the relative adsorption rates of the group II compounds. This
assumption is allowed in the present case because only cadmium
sublimation is observed during MOVPE. The mole fraction of
zinc in the solid phase, y, is given by (eq 36 in ref 10)
DEZn + DMCd H MEZn + MECd
MEZn + MECd H DMZn + DECd
MEZn + DMCd H DMZn + MECd
DEZn + MECd H MEZn + DECd
DEZn + DMZn H 2MEZn
R1
R2
R3
R4
R5
R6
DMCd + DECd H 2MECd
measured in the gas.23 The standard heat of formation for DEZn
and DMZn differs by 2 kcal/mol, while those for DECd and
DMCd differ only by 0.1 kcal/mol.24 Although the values are
unknown for MECd and MEZn, the heat of formation for these
species is expected to have a value close to the average of their
dimethyl and diethyl counterparts. This suggests that the change
in the free energy for the exchange reactions R1-R6 should
be less than about 2 kcal/mol.
MOVPE of Cd1-yZnyTe. Based on our previous studies of
ZnTe MOVPE from DMZn and DIPTe, substantial zinc
incorporation is expected for alloy growth from DMCd, DMZn,
and DIPTe at 673 K.12 However, from Figure 6a, it is apparent
that the growth rate of ZnTe is zero for essentially all x values
except for the case where no DMCd is fed. Evidently, DMCd
and DMZn compete for the same adsorption sites on the growth
surface. The decomposition of DMCd seems to be energetically
more favorable, thus inhibiting ZnTe growth. Similar observa-
tions were reported for Cd1-yZnySe growth during photoassisted
MOVPE.25 Using DMCd, DMZn, and DESe, no zinc was
incorporated into the film for any gas-phase zinc content less
than 85%.
The metalorganic vapor-phase epitaxy of Cd1-yZnyTe alloys
has been studied by Ahlgren et al.8 and Kisker.9 The organo-
metallic compounds examined were DMCd, DEZn, and DETe.
The ligand exchange reactions between the group II precursors
account for the erratic zinc segregation observed by these
researchers. We have found that ligand exchange in the feed
influence the alloy growth in two distinct ways. First, the partial
pressure of DEZn in the feed is lowered at the expense of the
more stable exchange products, DMZn and MEZn. This trend
is enhanced at temperatures exceeding 550 K (see Figure 5).
This makes the incorporation of zinc below 720 K difficult,
because little of the DEZn is left, and DMZn and MEZn do not
decompose readily at those temperatures. In the case of DMZn,
this is confirmed by the inability to grow the alloy using this
precursor (see Figure 6a). Second, the DMCd is converted into
the less stable products, DECd and MECd. For x values greater
than 0.6, there is virtually no DMCd left in the feed gas, and
only MECd and DECd remain (see Figure 4). Since methyl-
ethylcadmium and diethylcadmium are thermally unstable, the
incorporation of cadmium into the film is strongly favored.
Consequently, near the reactor inlet, only pure CdTe is deposited
in a thick layer, as seen in Figure 7.
Ahlgren and co-workers8 observed that at 714 K the zinc
preferentially incorporates into the film for all ratios of DMCd
to DEZn in the gas inlet. At this substrate temperature, zinc
incorporation may have been higher than in the present study,
because more of the MEZn will have decomposed along with
the DEZn. Nevertheless, the DMZn should still have been
unreactive. Moreover, the gas flow velocities used in their study
must have been very high to prevent the organocadmium
compounds from reacting to form CdTe on the leading edge of
the substrate. The use of high gas velocities is supported by
their observation of an increasing zinc content over the length
of the substrate, even though it was only 1 in. long. From the
few details given of their growth conditions, we speculate that
y ) 1/(A + 1)
(12)
in which
1/2
MDEZn
S
DRCd 1 - x
A )
(13)
(
)
MDRCd SDEZn
x
where M and S are the molecular weight (g/mol) and the initial
sticking probability of the organometallic compound, and the
subscript DRCd refers to the average value for the three reactive
Cd sources, DMCd, MECd, and DECd. There is no need to
consider MEZn and DMZn since they do not decompose
appreciably under the reaction conditions studied. The solid
curves shown in Figure 8 are the best fit of eq 12 to the data.
This fit yields a ratio of the sticking probability of the cadmium
compounds to diethylzinc, SDRCd/SDEZn, of 3.0. This ratio is
consistent with the higher reactivity of the cadmium sources.
The wide variation of data about the solid curve is due in part

4888 J. Phys. Chem. B, Vol. 101, No. 25, 1997
Kappers et al.
to the uncertainty in the composition data and in part to the
fact that the model does not account for variations in the
concentrations of the reactants along the reactor. This latter
problem could be solved by incorporating the model with
differential mass balances for each reactant and then integrating
these equations over the reactor length.
Chemistry program (Grant No. DMR-9422602), and Chemical
Thermal Systems program (Grant No. CTS-9531785). In
addition, material in this study is based upon work supported
by a National Science Foundation Graduate Fellowship to
K.J.W. All of this support is greatly appreciated.
References and Notes
Alkyl ligand exchange of the group III organometallic
precursors in III-V MOVPE has been reported by Grady et
al.26 They observed ligand exchange between trimethylamine
alane and trimethylgallium to yield different gallane species and
trimethylaluminum. The large difference in thermal stability
between these compounds lead to nonuniformities in the
composition of the AlxGa1-xAs films deposited in the MOVPE
reactor. These results are consistent with our current investiga-
tion of InxGa1-xAs MOVPE. We have discovered that. even
under low-pressure conditions, the group III precursors trim-
ethylindium and triethylgallium exchange ligands and that this
affects the segregation behavior of indium in the film.27
From the results presented in this study, we may conclude
that growing Cd0.96Zn0.04Te with ∆y ) 0.001 cannot be
accomplished in a horizontal flow reactor because of the
thorough upstream mixing of DMCd with DEZn. On the other
hand, Cd0.96Zn0.04Te has been successfully grown on GaAs/Si
substrates at 523-553 K using a high-speed, rotating-disc
reactor.28-30 We suspect that in this reactor the DMCd and
DEZn are fed through separate injection manifolds in the reactor
headplate. This design, together with rapid mixing of the gases
above the substrate, makes it possible to deposit alloy films of
uniform composition over the entire substrate. Complex reactor
designs of this type are essential when ligand exchange reactions
of the sources leads to large differences in their thermal stability.
(1) Dornhauf, R.; Nimtz, G. Narrow Gap Semiconductors. In Hohler,
G. Tracts in Modern Physics; Springer: Berlin, 1983.
(2) Tung, J. J. Cryst. Growth 1988, 86, 161.
(3) Tennant, W. E.; Kozlowski, L. J.; Bubulac, L. O.; Gertner, E. R.;
Vural, K. IEEE Conf. Series 1990, 321, 15.
(4) Filippozzi, J. L.; Therez, F.; Este`ve, D.; Fallahi, M.; Kendil, D.;
da Silva, M.; Barbe, M.; Cohen-Solal, G. J. Cryst. Growth 1990, 101, 1013.
(5) Willardson, R. K., Beer, A. C., Eds. Semiconductors and Semi-
metals; Academic Press: New York, 1981; Vol. 18.
(6) Johnson, S. M.; Kalisher, M. H.; Ahlgren, W. L.; James, J. B.;
Cockrum, C. A. Appl. Phys. Lett. 1990, 56, 946.
(7) de Lyon, T. J.; Johnson, S. M.; Cockrum, C. A.; Wu, O. K.;
Hamilton, W. J.; Kamath, G. S. J. Electrochem. Soc. 1994, 141, 2888.
(8) Ahlgren, W. L.; Johnson, S. M.; Smith, E. J.; Ruth, R. P.; Johnston,
B. C.; Kalisher, M. H.; Cockrum, C. A.; James, T. W.; Arney, D. L.; Ziegler,
C. K.; Lick, W. J. Vac. Sci. Technol. A 1989, 7, 331.
(9) Kisker, D. W. J. Cryst. Growth 1989, 98, 127.
(10) Kappers, M. J.; McDaniel, A. H.; Hicks, R. F. J. Cryst. Growth
1996, 160, 310.
(11) McDaniel, A. H.; Wilkerson, K. J.; Hicks, R. F. J. Phys. Chem.
1995, 99, 3574.
(12) Wilkerson, K. J.; Kappers, M. J.; Hicks, R. F. J. Phys. Chem., in
press.
(13) Butler, I. S.; Newbury, M. L. Spectrochim. Acta 1977, 33A, 669.
(14) Kaesz, H. D.; Stone, F. G. A. Spectrochim. Acta 1959, 360.
(15) Bellamy, L. J. The Infra-red Spectra of Complex Molecules;
Wiley: New York, 1956.
(16) Jackson, J. A.; Nielson, R. J. J. Mol. Spectrosc. 1964, 14, 320.
(17) Cross, R. J.; Glockling, F. J. Organomet. Chem. 1965, 3, 146.
(18) Cullen, W. R.; Deacon, G. B.; Green, J. H. S. Can. J. Chem. 1966,
44, 717.
(19) Green, J. H. S.; Kynaston, W.; Rodley, G. A. Spectrochim. Acta
1968, 24A, 853.
(20) McCoy, C. R.; Allred, A. L. J. Am. Chem. Soc. 1962, 84, 912.
(21) Henold, K.; Soulati, J.; Oliver, J. P. J. Am. Chem. Soc. 1969, 91,
3171.
(22) Soulati, J.; Henold, K.; Oliver, J. P. J. Am. Chem. Soc. 1971, 93,
5694.
(23) Higa, K. T. Unpublished results.
(24) Pilcher, G.; Skinner, H. A. The Chemistry of the Metal-Carbon
Bond; Wiley: New York, 1982.
(25) Bao, K. X.; Mo, R.; Kalisetty, S.; Gokhale, M.; Robinson, J.; Ayers,
J. E.; Jain, F. C. Presented at the 8th International Conference on MOVPE,
1996.
(26) Grady, A. S.; Markwell, R. D.; Russell, D. K.; Jones, A. C. J. Cryst.
Growth 1990, 106, 239.
(27) Kappers, M. J.; Warddrip, M. L.; Hicks, R. F. Manuscript in
preparation.
Conclusions
We have discovered that ligand exchange reactions occur
between the group II precursors during II-VI MOVPE. This
appears to be a general phenomenon in the growth of II-VI
and III-V alloys by this technique.26,27 Although ligand
exchange is well-documented in organometallic chemistry, its
impact on the MOVPE process has not been widely recognized.
With regard to Cd1-yZnyTe film growth, ligand exchange
produces unstable cadmium compounds, DECd and MECd, and
very stable zinc compounds, DMZn and MEZn. The large
difference in reactivities of these sources makes it very difficult
to uniformly incorporate controlled amounts of zinc into the
alloy film over the entire area of the heated substrate.
(28) Tompa, G. S.; Salagaj, T; Cook, L.; Stall, R. A.; Nelson, C. R.;
Anderson, P. L.; Wright, W. H.; Ahlgren, W. L.; Johnson, S. M. J. Cryst.
Growth 1991, 107, 198.
(29) Johnson, S. M.; Vigil, J. A.; James, J. B.; Cockrum, C. A.; Cockrum,
C. A.; Konkel, W. H.; Kalisher, M. H.; Risser, R. F.; Tung, T.; Hamilton,
W. J.; Ahlgren, W. L.; Myrosznyk, J. M. J. Electron Mater. 1993, 22, 835.
(30) Anderson, P. L.; Erbil, A.; Nelson, C. R.; Tompa, G. S.; Moy, K.
J. Cryst. Growth 1994, 135, 383.
Acknowledgment. The authors thank Dr. Kelvin Higa for
useful discussion of the results. Funding for this research was
provided by the Office of Naval Research (Award No. N00014-
95-1-0904), the National Science Foundation, Solid-State