Dummy growth is being performed frequently when people are using MOVPE system, for both of research and manufacture, although the content and process of dummy growth varies depending on the desired sample structure. Dummy growth is a part of the daily routine maintenance of MOVPE systems to maintain the stable growth conditions of the reactor. As a process of a dummy growth, baking at high temperature (> 1100 ℃) under atmosphere of ammonia and hydrogen can clean the reactor and gas lines. Dummy growth can keep the inner environment of the reactor to be stable, this is critical for some growth which is sensitive to it, such as GaN growth on Si. The content of a dummy growth depends on the
Fig. 2.15 Surface of GaN grown with TMAl pre-flow time of (a) 0 s; (b) 2 s; (c) 4 s; (d) 8 s.
(a) (b)
(c) (d)
31
film which would be grown after it. Generally, it is GaN dummy growth for the growth of GaN on sapphire. But for the growth of GaN on Si, the content of dummy growth should be AlN. There are two examples of growth series without an AlN dummy. In section 2.2.2, AlN dummy was applied to eliminate the problem of Si melt-back. However, it has been proved that AlN dummy could not stop Si melt-back.
Example I: Continuously repeating growth in Fig. 2.16. Three samples were grown continuously without AlN coating dummy growth in between; the growth mode of GaN for all three samples was 3D growth. But there was an AlN dummy growth prior to the first sample (a). The conditions for sample (a) and (b) are exactly the same; the only difference is that there was no AlN coating dummy prior to sample (b). From curvature observation in Fig.
2.16, no difference could be distinguished before the growth stage of GaN, i.e. thermal desorption, AlN growth and temperature ramping up. In curvature transition curves, they differ from the stage of island formation, as marked in Fig. 2.16. The 3D growth of sample (a) lasted much longer than that of sample (b), as shown by the straight line duration with no variation. In sample (b) the 3D growth was very short and coalesced rapidly. After coalescence, growth mode of GaN turned from 3D growth to 2D growth, so that the stress in GaN evolved from slightly compressive to tensile. From Stoney equation we know that the slope of curvature curve indicates the stress magnitudes and higher slope means larger stress.
Tensile stress in GaN of sample (b) is much higher than that of sample (a). Consequently, as curvature curves shown, sample (b) cracked during cooling while crack in sample (a) was much less. The same as sample (b), before the growth of sample (c) there was no AlN coating dummy growth also, but the difference was that during thermal desorption, there was 5 min
Fig. 2.17 FWHM of XRD rocking curve of continuously grown samples.
Fig. 2.16 Effect of AlN dummy growth, example I.
32
extra thermal cleaning in nitrogen atmosphere. 3D growth period in sample (c) was shorter than that in sample (a) but longer than in sample (b). The stress in GaN of sample (c) lies also between (a) and (b). The final residual stress during and after cooling down of sample (c) was also in between sample (a) and (b), as well as the amount of cracks. As indicated in Fig. 2.
17, the crystal quality of GaN also shows the same tendency as that of the slope of curvature evolution, namely the stress strength.
Example II: V/III ratio tests of AlN buffer layer. This was designed to investigate the effect of V/III ratio of AlN buffer layer on the properties of GaN without AlN dummy growth in between them. From sample 1 to sample 4, V/III ratio of AlN buffer layer varied from 470 to 4019 while all other conditions were kept the same. Growth mode of GaN was 2D layer-by-layer growth to avoid the influence from coalescence during 3D growth. Sample 5 was designed to repeat the growth of sample 1. AlN coating dummy growth was performed only prior to sample 1 and no AlN dummy before the following samples. As shown by the curvature curves in Fig. 2.18, except sample 1, the stress behavior of all other four samples are exactly the same. From sample 2 to sample 5, during the growth of GaN, it showed no dependence of GaN performance on the V/III ratio of AlN buffer layer. The quality of GaN shown in Fig. 2.19 also acts the same as it of curvature transition. Actually, as would be shown later, GaN performance should be dependent on it.
The detailed and exact mechanism behind the phenomena of uncontrollable growth is not well understood and yet to be investigated further. However, the possible process can be discussed. In the two examples above, the difference between the first sample and the following sample is that prior to the first sample there was an AlN coating dummy growth.
As discussed in the part of 2.2.2, the effect of AlN coating is to cover the GaN and metal gallium deposition inside the reactor. Influence of gallium origins from the reaction between it and silicon and may lead to Si melt-back. From the problem of uncontrollability here, although after using new gas purifiers for H2 and N2, the problem of Si melt-back had been solved due to the quality improvement of AlN buffer layer even without AlN coating dummy growth, GaN and metal gallium deposition still affects the growth. We suspect that the influence started from the stage of thermal cleaning. If there is no AlN coating dummy prior
Fig. 2.18 Irreproducibility of GaN on AlN buffer layer grown under various V/III ratio. I-470 II-1010 III-4015 IV-2019 V-470 800
1200 1600 2000 2400 2800 3200 3600 4000
FWHM (arc sec)
V/III ratio
(0002) (10-12)
Fig. 2.19 FWHM of XRD rocking curve of GaN on AlN buffer layer grown under various V/III ratio.
FWHM of (10-12) plane exceeded the upper limit, so it was shown the same for sample 2 to 4.
33
to a growth, Si wafer will be exposed to the surroundings of the deposition of GaN or metal gallium. Then during thermal cleaning or desorption at high temperature above 1000 ℃, some gallium atoms may be adsorbed on the surface of Si but the amount is not enough to initiate Si melt-back. Since the quality of AlN is good enough to isolate the TMGa source and Si surface during GaN growth and stop the reaction between them, no Si melt etching occurs.
But the adsorbed gallium atoms might change the surface state of Si substrate and affects the growth quality of AlN buffer layer and then the growth of following GaN layer. In next chapters we can know that gallium adsorption should have worsened the quality of AlN buffer layer but not worse enough to cause Si melt-back. Although how gallium atoms adsorption influence the growth of AlN buffer layer is not clear. For the samples with AlN coating prior to it, deposition of GaN or gallium was covered by AlN coating and the influence from it had been eliminated. On the contrary, such problem doesn’t exist for the growth of GaN on sapphire. Ga or GaN deposition inside the reactor doesn’t affect the quality of GaN growth on sapphire. The comparison between GaN growth on Si and sapphire shows that the difference between them lies in the substrate surface state which may be contaminated by a trace of metallic GaN.
So the difference between growth with and without AlN coating prior to them might be the gallium adsorption state on Si wafer surface. AlN dummy coating is an effective way to realize controllable growth. Similar to GaN and gallium deposition, there might be also some AlN and aluminum deposition inside the reactor, but the Al adsorption on Si surface is not harmful to AlN growth. On the contrary, proper amount of it is good for the growth of AlN, like the effect of TMAl pre-flowing to stop silicon nitride formation. Besides AlN coating, there is also some other method to eliminate the influence of GaN and gallium, such as thermal cleaning in HCl atmosphere which is more convenient than AlN coating but not available in our system.
Fig. 2.20 Thermal cleaning was performed in the end of a growth in H2 and NH3 atmosphere to improve controllability.
Fig. 2.21 Curvature of GaN growth on Si, V/III ratio for AlN buffer layer varied from 115 to 9156.
34
In example I, from sample (c) it can be known that thermal cleaning in N2 ambience is also working to improve controllability somehow but not good enough. Thermal desorption in nitrogen might be helpful to remove adsorbed H atoms from the surface of Si and then the adsorption of gallium atoms. However, on the other hand, annealing in nitrogen may cause Si nitridation and deteriorate the growth on III-nitrides on silicon. This argument is reasonable because the edge area of the surface of sample (c) in example I was rough. It is hopeful that the combination of AlN coating and thermal cleaning in nitrogen might be more effective to improve growth controllability by reducing more adsorbed H atoms, yet to be tested.
Some other efforts also have been tried to improve reproducibility and avoid AlN coating.
Although AlN is working to enhance reproducibility, it is time consuming and costly. Adding thermal cleaning in mixture atmosphere of hydrogen and ammonia at high temperature in the end of a growth prior to cooling was tested, as shown in Fig. 2.20. Effect of this measure is not stable and more or less works. Like thermal desorption of Si wafer in nitrogen, it also might be harmful to the growth. Annealing in hydrogen at high temperature may lead to GaN decomposition and destroy the surface of a sample, as discussed in part 2.2.2. So this method was not applied anymore.
Applying AlN coating, experiments of V/III ratio test for AlN buffer layer like that in example II of uncontrollability test were repeated. The result of curvature was shown in Fig.
2.21. It is clear that the performance of GaN is strongly dependent on the growth conditions of AlN buffer layer, such as V/III ratio which shown here. Based on all the experiments above, AlN dummy coating is a practical and effective method to achieve high controllability for the growth of GaN on Si, although it consumes much time and is not recommendable to industrial production.