Incorporation of Sb and As in MBE grown GaAs x Sb 1 − x layers

With the increasing interest in low effective mass materials for intersubband devices, mixed As-Sb compounds, like GaAsxSb1−x or AlxIn1−xAsySb1−y, gain more and more attention. The growth of these materials, however, still provides significant challenges due to the complex interaction between As and Sb. In this work, we provide an in-depth study on the incorporation of Sb into the GaAsxSb1−x layers and compare our findings to the present literature on this topic. It is found that both the composition and the crystal quality of GaAsxSb1−x layers are strongly influenced by the growth rate due to the As-for-Sb exchange reaction which takes place at the growing surface, and that high crystal quality can be achieved when the growth is performed under Sb limited conditions.

Quantum Cascade Lasers (QCLs) are a reliable source of coherent light in the mid-infrared (MIR) and terahertz (THz) regimes.While high output power and room temperature operation have been achieved in the MIR, QCLs are still facing significant challenges in the THz regime.The search for concepts which allow for higher performance of these devices shifts the focus of the research towards low effective electron mass well materials like Ga x In 1x As 1,2 and InAs, [3][4][5][6] and thus compounds containing Sb, like GaAs x Sb 1x or Al x In 1x As y Sb 1y , 7 become more and more interesting as barrier materials.The growth of high quality layers of these materials, however, is still challenging since the presence of two group V species creates a non-trivial growth environment.GaAs x Sb 1x is an ideal material to study the interaction between As and Sb, since the presence of only one group III material limits the space for interpretations of the experimental results.Molecular beam epitaxy (MBE) growth of GaAs x Sb 1x was first demonstrated by Chang  et al. in 1977. 8Nakata and co-workers 9 have grown high quality layers of GaAs x Sb 1x using As 4  and Sb 4 at substrate temperatures between 470 • C and 490 • C and a growth rate of 1.3 µm h 1 .Since then, numerous publications were devoted to understanding the incorporation of As and Sb into this alloy.
The inverse relation between the growth temperature (T g ) and the incorporation of Sb into GaAs x Sb 1x was first demonstrated by Klem et al. 10 and is generally agreed on in the literature.The role of the growth rate, however, is still subject to debate.Almuneau et al. reported on the incorporation of Sb into GaAs x Sb 1x , AlAs 1x Sb x , and Al x Ga 1x As y Sb 1y . 11For all three alloys, a linear relation between the Sb flux, normalized by the total flux of the group III elements, and the Sb mole fraction in the layer was found.From these results, it was concluded that the incorporation of Sb has to be close to unity.Similarly, Semenov et al. found an increasing As mole fraction in GaAs x Sb 1x and Al x Ga 1x As y Sb 1y alloys for increasing group III fluxes and constant group V fluxes. 12This is attributed to the fact that higher growth rates will lead to a shortage of Sb and unoccupied lattice sites will be filled by excess As.
On the other hand, the reverse has also been stated by other authors.An inverse relation between the As mole fraction in GaAs x Sb 1x layers and the arrival rate of Ga atoms was first reported by Klem et al. in 1987, 13 i.e., higher P Ga results in higher Sb incorporation.The same effect has been observed by Bosacchi and co-workers 14 for GaAs x Sb 1x layers grown on GaAs substrates.The authors concluded that the incorporation of Sb into GaAs x Sb 1x is more effective at higher growth rates since the availability of Ga sites on the growing surface promotes the dissociation of the Sb species.Sun and co-workers 15 grew layers of GaAs x Sb 1x on GaAs at growth rates between 0.2 ML s 1 and 0.9 ML s 1 .At growth rates above 0.8 ML s 1 , the authors find a saturation of the Sb mole fraction and hence conclude that the growth rate dependence of the composition can be eliminated if the growth is performed at the higher growth rates.The authors assume that high growth rates, and hence, availability of Ga sites, prevent the desorption of Sb which leads to an increase in the incorporation efficiency.Selvig et al. analyzed GaAs x Sb 1x and Al x Ga 1x As y Sb 1y layers with Sb mole fractions up to 0.2, grown on GaSb substrates. 16It was found that the incorporation of Sb is decreased at the lower growth rate.The authors conclude that a lower Sb composition which is found if the GaAs x Sb 1x layer is grown at a lower growth rate is due to an As-for-Sb exchange reaction which happens at the growing surface.
We can see that the reports on the growth of GaAs x Sb 1x by MBE strongly disagree on the incorporation behavior of Sb.While some authors find an increase in the Sb mole fraction with the growth rate, others find the opposite behavior.Moreover, the authors reporting to find an increase in the Sb mole fraction with the growth rate present very different explanations for this effect.Hence, an in-depth study on the effect of the growth rate on the composition of GaAs x Sb 1x is necessary in order to understand the mechanisms that define the final layer composition.
In order to investigate the incorporation of Sb into GaAs x Sb 1x , a set of samples was grown in a Riber Compact 21 MBE system on free-standing n + InP (001) wafers.Within this set, the growth rate (R GaAsSb ) was varied between 0.195 ML s 1 and 1.56 ML s 1 .The range of P Sb 2 was between 5 × 10 −7 Torr and 1.0 × 10 −6 Torr, a P As 2 between 2.8 × 10 −6 Torr and 1.1 × 10 −5 Torr was used, and the T g was varied from 400 • C to 460 • C. All pressures in this work are given as absolute values without correction for the specific molecular species.For the group V materials, valved cracking cells were used with the cracking zone temperatures set to 850 • C and 1000 • C to produce As 2 and Sb 2 , respectively.For all samples, first, the oxide was thermally desorbed under P As 2 of 1.1 × 10 −5 Torr at 520 • C. The temperature was then reduced to the growth temperature where a 50 nm thick Ga 0.47 In 0.53 As buffer layer and a 400 nm thick GaAs x Sb 1x layer were grown.All samples were measured with high resolution X-ray diffraction (HRXRD) to determine their crystal quality and composition.
To understand the influence of the growth rate on the composition of GaAs x Sb 1x , four sets of samples have been grown at a T g of 400 • C and 460 • C. For each set, T g , P As 2 , and P Sb 2 were kept constant, while the R GaAsSb was varied.The results are given in Figure 1.All samples show a decrease in the As mole fraction with increasing R GaAsSb on the low growth rate side (R GaAsSb between 0 ML s 1 and 0.5 × 10 −6 ML s 1 ), which is in agreement with Refs.13-16.On the high growth rate side, however, the As mole fraction is increasing with rising R GaAsSb as reported in Refs.11 and 12. Depending on the actual set of parameters chosen in each of these publications, the authors only saw either the low or the high growth rate side which resulted in the discrepancy in their reports.
In order to gain further insight, samples were grown at different values for T g , P As 2 , and P Sb 2 .We find that a decrease in P Sb 2 , under otherwise identical conditions, leads to a shift of the measured compositions in the GaAs x Sb 1x layers towards higher As mole fractions (Figure 1(b)).While this seems to be an obvious result, it should be noted that the effect is much more dramatic on the high growth rate side of the experiment than on the low growth rate side.If we assume a temperature dependent but otherwise constant incorporation coefficient for Sb, as described in Refs.11 and 12, the arsenic mole fraction should be given by x = 1 − C P Sb 2 P Ga , with C being constant for a given T g .On the high growth rate side, the compositions of the GaAs x Sb 1x layers approach this trend, as indicated by the dashed lines in Figure 1(a).We can conclude that the composition at high growth rates is strongly dependent on the arrival rate of Sb 2 molecules, while a different effect limits the incorporation of Sb into the GaAs x Sb 1x layer at the low growth rates.In order to understand the via C. On the low growth rate side, the As mole fraction is rising with the falling growth rate.Supplying less Sb 2 (b) leads to a higher As mole fraction.Supplying less As 2 (c) leads to a lower As mole fraction in the GaAs x Sb 1x layer.On the high growth rate side, samples grown at a T g of 400 • C and 460 • C show an almost identical behavior (d).On the low growth rate side, the incorporation of Sb into GaAs x Sb 1x is strongly decreased at a higher temperature.
influence of the T g , samples were grown at a T g of 460 • C under otherwise identical conditions (Figure 1(d)).We can see that there is an almost perfect overlap with the samples grown at T g of 400 • C on the high growth rate side.This indicates that the sticking of Sb does not depend on T g in this temperature range, and that this slope is solely dependent on the P Sb 2 P Ga fraction.On the low growth rate side, however, we see a strong increase in the As mole fraction at a higher T g .
In addition to the composition of the GaAs x Sb 1x layers, their crystal quality might give a hint on the incorporation of Sb.In order to understand the influence of the growth rate on the crystal quality, samples grown at different growth rates but similar compositions were examined by HRXRD.A comparison of the crystal quality of GaAs x Sb 1x layers grown at 0.227 ML s 1 and at 0.870 ML s 1 is shown in Figure 2. Subfigure (a) shows the rocking curve scans of the two layers.The sample grown at a lower growth rate shows a full width at half maximum (FWHM) of 258 arc sec while the peak corresponding to the layer grown at a higher growth rate shows a FWHM of only 36 arc sec.HRXRD reciprocal space maps (RSMs) around the InP (224) diffraction peaks of the two layers are shown in subfigures (b) and (c).The in-plane lattice constant measured for the layer grown at the low growth rate deviates from that of the InP substrate, indicating partial relaxation.Using the measured in-plane and out-of-plane lattice constants and using the stiffness components given in Ref. 17, a relaxation of 25% was calculated.The layer grown at a higher growth rate (subfigure (c)) shows no deviation in the in-plane lattice constant and hence can be assumed to be fully strained.The effect of early relaxation can also be seen by comparing the crystal quality of highly lattice mismatched layers.The two samples shown in Figure 1(b) which have been grown at a lower P As 2 and at a R GaAsSb of ∼0.45 ML s 1 and 0.88 ML s 1 have almost identical compositions.However, the layer grown at a higher growth rate shows a rocking curve peak FWHM of 1140 arc sec while the layer grown at a lower growth rate shows a peak width of 1710 arc sec which is an increase of about 50%.This observation indicates that the mechanism, controlling the Sb incorporation at low growth rates, facilitates the strain relaxation in the GaAs x Sb 1x layers.From these results we can conclude that the crystal quality of GaAs x Sb 1x layers is not solely defined by the relaxation due to lattice mismatch.Layers grown at low growth rates already show partial relaxation at compositions which yield a fully strained layer when a high growth rate is used.
Different mechanisms for the incorporation of Sb into GaAs x Sb 1x layers were proposed in the literature.Semenov and co-workers found an increase in the As mole fraction with increasing growth rates and attributed this to a shortage of Sb.These observations can be confirmed for the high growth rate side of the experiment.Three independent studies found a decrease in the As mole fraction with an increasing growth rate and provided different theories to explain their observations.The sample grown at a lower growth rate of 0.227 ML s 1 shows FWHM of 258 arc sec, while the layer grown at 0.0870 ML s 1 shows a FWHM of only 36 arc sec.The RSMs around the InP (224) diffraction peak show a relaxation of 25% for the GaAs x Sb 1x layer grown at the low growth rate and a fully strained layer for the sample grown at the high growth rate.Sun et al. suggested that this increase in the Sb mole fraction is related to the availability of Ga sites which prevent the desorption of Sb species and hence increase the incorporation of Sb into the GaAs x Sb 1x layers.Under this assumption, the sticking coefficient of Sb would have to increase super linearly with the availability of Ga sites since the P Sb P Ga fraction is decreasing when the R GaAsSb is raised.Furthermore the Sb sticking coefficient should be independent of the P Sb 2 .Hence, increasing P Sb 2 would have a similar effect on the high and on the low growth rate side of the experiment.In Figure 1, however, we see that increasing the P Sb 2 has a much stronger effect on the high growth rate side.Moreover, the incorporation of As would have to be independent of the R GaAsSb , since it would counteract the increase in the Sb mole fraction.Bosacchi et al. suggested that the dissociation of Sb 4 into Sb 2 is enhanced when a higher density of Ga sites is available.The arguments which speak against this mechanism proposed, by Sun et al., can also be applied to this theory.Moreover, an enhanced incorporation of Sb at higher growth rates was also found when Sb 4 is cracked into Sb 2 by using a cracker cell.This indicates that the dissociation of Sb 4 is not the predominant mechanism in this experiment.
In Figure 1 we also see that the growth temperature has a stronger influence on the composition of the GaAs x Sb 1x layer at low growth rates.The fact that the high growth rate side for both curves overlaps shows that the desorption of Sb is only slightly dependent on the T g in this temperature range.This agrees with the line-of-sight QMS measurements by Kaspi and co-workers. 18,19Since the desorption of Ga can be neglected at these temperatures 20 and, hence, the availability of Ga sites is temperature independent, the difference between the GaAs x Sb 1x layers grown at different temperatures has to be explained by a different effect.Both mechanisms cannot provide an explanation for (1) the lower Sb incorporation at elevated T g and ( 2) an enhanced tendency for the relaxation of GaAs x Sb 1x layers grown at low growth rates.
Selvig and co-workers conclude that the increase in the Sb mole fraction with higher growth rates is related to an As-for-Sb exchange reaction which takes place at the growth surface. 16According to this mechanism, the Sb mole fraction increases with the growth rate since the time an Sb site is exposed to As flux decreases, and hence, the rate at which this event takes place is reduced.The As-for-Sb exchange reaction was previously studied by Losurdo and co-workers 21 by exposing steady GaAs and GaSb surfaces to Sb and As fluxes, respectively.The authors found a transformation of the GaSb into GaAs while the reverse reaction is not found and conclude that the As for Sb exchange is favored due to the difference in the enthalpy of formation of the two reactions, TEM analysis of GaAs layers exposed to Sb 2 flux shows relatively smooth interfaces while GaSb layers exposed to As 4 flux show As-Sb clusters.This mechanism provides a good explanation for all the phenomena which have been found regarding the incorporation of Sb into GaAs x Sb 1x layers.Since the exchange reaction takes place at the growing surface, the chance for an Sb atom to be replaced by an As atom is larger at a lower growth rate.At higher T g the reaction happens more frequently since the energy barrier for this reaction is more easily overcome.Since the imperfections such as particulates and impurities strongly influence the relaxation of strain in epitaxial layers, 22,23 the early relaxation of GaAs x Sb 1x layers grown at low growth rates can be linked to the presence of AsSb clusters originating from the As-for-Sb exchange reaction.Of the three mechanisms proposed in the literature to describe the incorporation of Sb into GaAs x Sb 1x layers, the As-for-Sb exchange reaction is the only one which can describe all aspects found in the experiments.
In conclusion, we have studied the incorporation of Sb into GaAs x Sb 1x layers grown on InP by MBE and compared our results to the studies on this topic found in the literature.It was found that the incorporation of Sb is strongly dependent on the R GaAsSb .At high growth rates, the Sb mole fraction is limited by the P Sb 2 P Ga ratio.At low growth rates, the As-for-Sb exchange reaction inhibits the incorporation of Sb and leads early relaxation of lattice mismatched layers.It is evident that the growth rate is an important parameter for both composition and crystal quality of mixed As-Sb compounds.

FIG. 1 .
FIG. 1.(a) Composition of GaAsx Sb 1x over the growth rate R GaAsSb at different P As2 , P Sb2 , and T g .On the high growth rate side, the As mole fraction rises with the growth rate since for each curve the P Sb2 is constant and hence the ratio P Sb 2 P Ga is

FIG. 2 .
FIG. 2. Triple axis rocking curve (a) and RSM ((b) and (c)) scans of GaAs x Sb 1x layers grown at different growth rates.The sample grown at a lower growth rate of 0.227 ML s 1 shows FWHM of 258 arc sec, while the layer grown at 0.0870 ML s 1 shows a FWHM of only 36 arc sec.The RSMs around the InP (224) diffraction peak show a relaxation of 25% for the GaAs x Sb 1x layer grown at the low growth rate and a fully strained layer for the sample grown at the high growth rate.