3.1. Ge Deposition
The MOVPE growth of the solar cell structures is usually carried out in mass transport regime, at temperature values around 600 °C, however, in the attempt to minimize the evaporation of the contaminants from the graphite parts of the growth chamber, the growth temperature for the Ge deposition has been reduced to 475–500 °C.
The drawback of reducing the growth temperature is the change of the growth regime: around 470 °C the Ge deposition with GeH
4 takes place in the kinetic regime [
11], therefore it is strongly influenced by temperature variation along the wafer radius, as well as by the total carrier gas flow. In
Table 1, we report the growth rate of Ge deposited on Ge substrates in a III-V contaminated MOVPE reactor, as a function of several growth parameters, along with the number of IV elements coating runs.
The Ge growth rate related to MOVPE depositions extracted from
Table 1, performed with similar germane partial pressure (from 15 Pa to 25 Pa), as a function of the deposition temperature and of the coating runs number is shown in
Figure 2.
It can be pointed out that few IV-based material coating runs are enough to obtain high growth rate, as compared with the values reported in literature (see [
11,
12]), related to samples grown at similar temperatures and precursor partial pressure. After already one IV elements coating run, Ge growth rate becomes mainly influenced by the MOVPE reactor geometry (which allows obtaining high germane utilization efficiency) and by other growth conditions, like the use of N
2 as carrier gas instead of H
2, which improves the growth rate [
9].
At this stage, the “As growth blocking role” seems irrelevant, as the growth rate can be easily controlled and optimized, for example, by changing the total flow and the precursors partial pressure, as shown by comparing the runs S3, S4 and S5. The evolution of the background carrier concentration related to the samples S1, S2, S4, S7, S8 and S10, nominally undoped, has been measured by electrochemical capacitance-voltage measurements (see
Figure 3a).
In sample S1, a carrier concentration around 1020 cm−3, (likely due to the incorporation of arsenic, see SIMS analysis reported in Figure 7) has been measured at the beginning of the run. This value is reduced to 1–2 × 1018 cm−3 after 2 µm thick Ge deposition. During Ge growth, the level of contamination is reduced, as germanium is deposited, with different efficiency, on all the growth chamber surfaces (susceptor and ceiling, etc) from which the previous deposited arsenic can evaporate. From sample S1 to sample S4, the carrier concentration drops at the end of the run from 2 × 1018 cm−3 to 3 × 1017 cm−3.
After replacing the susceptor, ceiling, quarts plate and graphite satellites, a significant arsenic contamination is still present at the beginning of run S7, probably due to the residual evaporation of arsenic from the graphite ring, however, this contamination is reduced to 4 × 1016 cm−3 after 3.5 µm thick Ge deposition. The contamination further decreases to 1 × 1016 cm−3 in run S8, after a Ge deposition 7.5–8 µm thick. This carrier background value has remained constant even after several further coating runs, as shown by looking at the carrier profile measured in the sample S10 (performed after further 21 coating runs of Ge and SiGeSn layers).
It is worthwhile to point out that the background “contamination” is also related to the growth rate, as shown in
Figure 3b. The first part of S9 has been carried out at high growth rate (117 nm/min), while the second part, with a growth rate one order of magnitude lower (10 nm/min). The background carrier concentration increases from 6 × 10
15 cm
−3 to 4.5 × 10
16 cm
−3 by switching from high to low growth rate. This result can be explained by considering that by increasing the growth rate, the contaminants which are present in the MOVPE growth chamber become much more diluted during the deposition of the material. The drop in carrier concentration registered near the surface of sample S9 could be a measurement artefact or it could be explained by the evaporation of the contaminant during the cool down phase of the MOVPE run.
We found that it is possible to increase the growth rate by 6%, without varying the growth temperature (and keeping constant the total carrier flow and the germane partial pressure), by replacing germane diluted in hydrogen with germane diluted in nitrogen (compare the growth rate values reported in
Table 1 for samples S5 and S10).
Remarkably, we have been able to switch Ge conductivity polarity from n-type to p-type as shown in
Figure 4. The first part of sample S11 has been grown nominally undoped, then, by injecting in the growth chamber DEZn, with a molar fraction of 3.5 × 10
−4, the last part of the sample shows a p–type conductivity, with a carrier concentration around 2–3 × 10
15 cm
−3. On sample S12, we have replaced DEZn with TMGa and the dopant molar fraction has been increased to 1.8 × 10
−3. Eventually, sample S13 has been deposited with a constant TMGa molar fraction of 1.27 × 10
−2. The related ECV characterization shows a constant p-type carrier profile, with a concentration around 1 × 10
18 cm
−3. It is worth noting that the doping level of sample S13 is lower than the peak doping level of sample S12, despite a higher input TMGa molar fraction has been utilized for the former sample. This apparent contradiction is due to the higher growth temperature utilized in the deposition of sample S13 (25 °C higher than the one set for the growth of sample S12, see
Table 1). The TMGa molar fraction decreases as the temperature increases, because GeH
4 activation energy is higher than TMGa activation energy (41–42 kcal/mole versus 35.4 kcal/mole, respectively). This means that the effective TMGa molar fraction was actually lower in the sample S13 than in the sample S12.
The carrier concentration results related to runs S11, S12 and S13 show that to a large extent it is possible to modulate the p-type doping in Ge even when the IV element semiconductor is deposited in a reactor which is also utilized for III-V growth.
3.2. SiGe(Sn) Deposition
SiGe(Sn) depositions, from sample S14 to sample S17, have also been carried out in a III-V contaminated reactor and after several IV-based coating runs as reported in
Table 2.
The growth rate of run S14 (SiGe), coherently with the results presented in the previous chapter, is comparable with that obtained on Ge epitaxial samples, even if a higher GeH
4 partial pressure has been utilized. It is worthwhile to point out that an increment in GeH
4 partial pressure has been required to compensate the growth rate decrease caused by the injection of Si
2H
6 in the growth chamber (see
Figure 5).
Si
2H
6 decomposition, in fact, introduces hydrogen atoms that passivate Ge surface, as found for silicon epitaxy [
13].
A carrier concentration peak value as high as 4 × 10
19 cm
−3 has been measured by EVC on sample S14, deposited after 33 coatings runs (see
Figure 6). This value is four times higher than the value measured on sample S4, which was deposited only after five coating runs (see
Figure 3a). In order to understand the reason behind this result, we have analyzed the IV elements coating runs performed till the sample S14.
We came to the conclusion that the background carrier concentration peak value measured in sample S14 could be correlated to the three SiGeSn coating runs sequentially deposited just before the run S14. During SiGeSn growth, in fact, HCl can be formed in the growth chamber by the decomposition products of SnCl
4 and hydrides (Si
2H
6 and GeH
4). Since HCl is recognized as an etchant for Ge and SiGe [
14], during SiGeSn deposition, Ge and SiGe coatings already deposited on the reactor graphite parts can be subjected to chemical etching and, as a consequence, possible evaporation paths for the previous deposited As can be generated. The presence of a high arsenic concentration in the sample S14 and in the subsequent samples (S15 and S16) has been assessed by SIMS (see
Figure 7a).
Indeed, in all samples a high As incorporation is measured at the beginning of the MOVPE deposition. Since SiGe growth rate is one order of magnitude higher than the SiGeSn growth rate (see
Table 2), the As contaminant can be highly diluted in SiGe and its concentration strongly decreases during the SiGe growth, while it remains almost constant in the case of SiGeSn growth. The effect of the growth rate on As incorporation is also evidenced on sample S16, composed of a first Ge buffer layer and of a subsequent SiGeSn layer. For this sample, owing to the high thickness, we joined the results of SIMS and ECV measurements in order to show both As and the carrier profile along the sample depth (see
Figure 7a). Like SiGe, Ge deposition takes place at high growth rate, therefore the ECV measurement shows a rapid decrease in the carrier concentration during the deposition. On the contrary, on SiGeSn, SIMS measurement shows that As concentration rises to values as high as 10
19 cm
−3.
All these data lead to the following conclusions: (i) in a III-V contaminated MOVPE growth chamber, during SiGeSn deposition a considerable amount of arsenic can be released from the MOVPE reactor walls and be incorporated in SiGeSn and subsequent layers, (ii) the As contamination can be reduced during the deposition by growing the group IV compounds at high growth rate, around 100 nm/min, while it remains at level around 1019 cm−3 when the growth rate is one order of magnitude lower (10 nm/min).
At this point, it becomes interesting to analyze the reason of the growth rate reduction from run S14 (SiGe) to run S15 (SiGeSn), despite both MOVPE runs have been carried out with the same GeH
4 partial pressure. Neither the higher Si
2H
6 partial pressure utilized in run S15 with respect to run S14, nor the “As growth blocking role” can explain the reduction in growth rate. In fact, referring to the data shown in
Figure 5, an increase of Si
2H
6 partial pressure from 2.5 Pa to 6.25 Pa can only explain a 20% decrease in the growth rate. Furthermore, the same starting As contamination has been measured in SiGe and SiGeSn (see
Figure 7a). This means that As “carry-over” have saturated SiGe and SiGeSn surface sites in the same way. The low growth rate of SiGeSn can still be understood by considering the etching action on Ge and SiGe produced by HCl, as formed by the decomposition products of SnCl
4, Si
2H
6 and GeH
4. The contrasting actions between the deposition rate and the etching rate has also been observed in GeSn growth by GeH
4 and SnCl
4 [
15].
In order to reduce the weight of the etching action, a higher germane partial pressure could be used. For example, S. Wirths at al., have shown that it is possible to reach SiGeSn growth rates as high as 125 nm/min with a Ge
2H
6 partial pressure of 120 Pa (a value four time higher than the GeH
4 partial pressure used to grow our samples) [
16]. Therefore, in principle, by increasing SiGeSn growth rate it should be possible to reduce the As background concentration near the value obtained in Ge and SiGe.
However, we have also assessed a further difficulty in controlling SiGeSn conductivity, due to the presence of impurities in SnCl
4 precursor. This is evidenced by SIMS analysis related to P incorporation in the samples S14, S15 and S16 as shown in
Figure 7b. It can be pointed out that in run S14 (SiGe) and in the first part of run S16 (Ge growth), where SnCl
4 is not used, a P concentration, respectively, around 4 × 10
16 cm
−3 and 3 × 10
17 cm
−3 has been measured, instead, during the growth of SiGeSn (S15), P concentration increases to 3 × 10
19 cm
−3.
Thus, we can conclude that by using SnCl
4, a considerable amount of P is introduced in the growth chamber. Like As, P introduces electrons in IV-based compounds and therefore it influences the material’s conductivity. The P contamination of the growth chamber produced by SnCl
4 allows explaining the higher P concentration measured in Ge, in run S16, than in SiGe, in run S14. Ge deposition, in fact, has been carried out after SiGeSn growth (S15). During SiGeSn deposition, P is introduced in the growth chamber, can be deposited on the MOVPE reactor walls and then subsequently be released and be incorporated in Ge during the growth. Therefore, for controlling SiGeSn conductivity, regardless the contamination introduced from the previous III-V deposition, it is important to reduce the amount of P contained in SnCl
4, or to consider alternative precursors, like, for example, TESn (Triethyltin) as, reported for the growth of GeSn [
17].
3.3. III-V-Based Semiconductor Deposition
We have not observed growth rate variation in III-Vs owing to the previous IV elements growth. Therefore, we have focused our investigation on how to control the III-Vs conductivity, trying to reduce to acceptable levels the IV elements contamination in III-Vs. The III-Vs structures analyzed are reported in
Table 3.
MOVPE runs S18 and S19 are related to AlAs/GaAs distributed Bragg reflector (DBR) structures, whose period has been repeated 15 times. These samples have been deposited in mass transport regime (Tgrowth = 888 K) after one and eight III-V coating runs, respectively. In
Figure 8, SIMS concentrations of Ge, Si and Sn, measured along the thickness of the samples S18 and S19 are depicted. The incorporation of Ge, Si and Sn is different in AlAs and GaAs, therefore, along the DBR structure thickness, the concentration of these atoms oscillates. After one coating run, the maximum group IV elements concentration in the III-V Bragg test structure is around 5 × 10
17 cm
−3. After eight coating runs (with an equivalent deposition of 8 µm thick III-Vs), the maximum background contamination is reduced to values lower than 2 × 10
17 cm
−3. This value is more than half of the previously reported value measured in GaAs (4.6 × 10
17 cm
−3), in a Si-Ge MOVPE contaminated growth chamber, after a 10 µm thick coating run [
18].
In order to further reduce IV elements incorporation in III-Vs, we assessed the influence of the growth temperature on the background carrier concentration. The evaporation of the contaminants from the MOVPE reactor walls is tightly related to the growth temperature and, in particular, by decreasing the growth temperature, we expect reducing the background IV elements concentration in III-V semiconductors.
We have then investigated a InGaP/InGaAs/InGaP/Ge(GaAs) structure (S20), in which the bottom InGaP and InGaAs layers have been deposited at 903 K, while InGaP top layer has been deposited at lower temperature (773 K). Moreover, in run S21 we have replicated the same growth process of run S20 concerning the InGaP and InGaAs bottom layers, while the InGaP top layer has been replaced by the InGaAs one. For both samples, the IV elements concentration has been indirectly measured by ECV, as shown in
Figure 9. It is worthwhile to point out that the MOVPE deposition of sample S20 and S21 have been carried out after 20 and 32 III-V coating runs, respectively.
On sample S20, a starting background carrier concentration near 1 × 10
18 cm
−3 has been measured when the material is grown at 903 K on Ge substrate. After depositing 1 µm of InGaP at 773 K, the background carrier concentration is reduced to 6 × 10
15 cm
−3, a value around three order of magnitude lower than the value measured by E. Welser, in InGaP layers grown in a Ge-contaminated MOVPE growth chamber [
8]. This comparison confirms the key role of the growth temperature for decreasing the incorporation of IV elements contaminants in III-Vs. By growing InGaP on a GaAs substrate, the carrier concentration can be further reduced by an order of magnitude, down to 3 × 10
14 cm
−3. By using a GaAs substrate, in fact, the Ge auto-doping and Ge solid state diffusion that take place during the InGaP growth on a Ge substrate can be both suppressed [
19].
A strong decrease in the background carrier concentration owing to the reduction of the growth temperature has also been measured on sample S21 (see
Figure 9b). During InGaP nucleation, which takes place at 903 K, the starting carrier concentration level is as high as 2 × 10
18 cm
−3. During the InGaAs deposition carried out with a growth rate twenty-four times higher than the InGaP one, the level of contamination is reduced to 2 × 10
17 cm
−3. As soon as the growth is stopped, before the temperature ramping, there is an accumulation of contaminants on the sample surface. Eventually, after the growth of 0.4 µm InGaAs at 773 K, the carrier concentration sharply decreases to 6 × 10
16 cm
−3. This value is still higher than the value measured in the top InGaP layer of sample S20, because the top InGaAs layer of sample S21 has been deposited with a thickness which is more than half of the InGaP one.
It is interesting to compare the background carrier concentration measured at the end of InGaAs layer grown at higher temperature on the sample S21 with the IV elements atoms concentration measured by SIMS on sample S19 (see
Figure 9b and
Figure 8). These values are almost equal, showing that after 8 µm thick III-V coating deposition, the subsequent III-V coatings are not effective in further reducing the back ground carrier concentration.
3.4. Morphological and Structural Assessment
The effect of the cross influence between group IV and III-V elements on the morphology and crystal structure of III-V and IV-based semiconductors has been assessed. As shown in
Figure 10, Ge surface morphology does not degrade when Ge is grown in a III-V contaminated MOVPE reactor. The surface morphologies of Ge layers are excellent and the best ones have been obtained when Ge is doped with Gallium (sample S12). Similar morphological results have been obtained for SiGe (run S14).
SiGeSn morphology, on the other hands, has been found strongly dependent on the level of As contamination in the MOVPE growth chamber. If SiGeSn is grown by MOVPE, around 756 K, in an arsenic free growth chamber (i.e., in a growth chamber with clean graphite parts that did not see any previous III-V deposition), tin precipitation takes place and the layer’s morphology deteriorates (see
Figure 11). In order to explain the key role of As to avoid tin precipitation, a theory on the role of the bond length of adatoms in inhibiting tin segregation has been proposed [
20].
Concerning III-Vs semiconductors, we found that when they are deposited in a IV elements contaminated MOVPE growth chamber their morphology changes as a function of the substrate type and orientation. On sample S20, for example, the top InGaP layer, deposited at low temperature, presents a featureless morphology on GaAs substrate and different oriented defects on Ge substrate (see
Figure 12).
In particular, regardless of the growth temperature, we assessed that III-V’s morphology on Ge substrate strongly degrades in Sn-contaminated MOVPE growth chamber, eventually leading to a polycrystalline material.
In
Figure 13, for example, we show the morphology and the structural characterization of a AlGaAs/GaAs/AlGaAs heterostructure, deposited on Ge and GaAs substrates in a Sn-contaminated MOVPE growth chamber.
The results reported in
Figure 13 could be explained by assuming that during the heating phase of the MOVPE run, Sn evaporates from the reactor walls and can be adsorbed on the substrate surface at different concentrations according to the substrate orientation. In particular, the (100) 6° off towards <111> orientation has more steps and kinks, where Sn can be adsorbed, than (100), 2° off towards <110> orientation. If Sn adsorbed somehow disturbs the atoms incorporation in the crystal, the epitaxial growth could be prejudiced on substrates whose orientation favors Sn adsorption. In order to confirm this hypothesis and exclude the influence of the chemical nature of the substrate, further experiments will be reported in a next publication by considering the deposition of III-Vs on Ge substrates with the same orientation of the GaAs ones.