Accepted Manuscript Electron microscopy characterization of AlGaN/GaN heterostructures grown on Si (111) substrates A. Gkanatsiou, Ch B. Lioutas, N. Frangis, E.K. Polychroniadis, P. Prystawko, M. Leszczynski PII:
S0749-6036(16)31142-9
DOI:
10.1016/j.spmi.2016.10.024
Reference:
YSPMI 4569
To appear in:
Superlattices and Microstructures
Received Date: 5 October 2016 Accepted Date: 12 October 2016
Please cite this article as: A. Gkanatsiou, C.B. Lioutas, N. Frangis, E.K. Polychroniadis, P. Prystawko, M. Leszczynski, Electron microscopy characterization of AlGaN/GaN heterostructures grown on Si (111) substrates, Superlattices and Microstructures (2016), doi: 10.1016/j.spmi.2016.10.024. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Electron Microscopy characterization of AlGaN/GaN heterostructures grown on Si (111) substrates
A. Gkanatsioua, Ch. B. Lioutasa, N. Frangisa, E. K. Polychroniadisa, P.
RI PT
Prystawkob,c, M. Leszczynskic,b a
Department of Physics, Aristotle University of Thessaloniki 54124, Greece
b
Institute of High Pressure Physics “Unipress”, Polish Academy of Sciences
TopGaN Ltd Sokolowska 29/37, 01-142 Warsaw, Poland
M AN U
c
SC
Sokolowska 29/37, 01-142 Warsaw, Poland
Abstract.
AlGaN/GaN buffer heterostructures were grown on “on axis” and 4 deg off Si (111) substrates by MOVPE. The electron microscopy study reveals the very good
TE D
epitaxial growth of the layers. Almost c-plane orientated nucleation grains are achieved after full AlN layer growth. Step-graded AlGaN layers were introduced, in order to prevent the stress relaxation and to work as a dislocation filter. Thus, a
EP
crack-free smooth surface of the final GaN epitaxial layer is achieved in both cases,
AC C
making the buffer structure ideal for the forthcoming growth of the heterostructure (used for HEMT device applications). Finally, the growth of the AlGaN/GaN heterostructure on top presents characteristic and periodic undulations (V-pits) on the surface, due to strain relaxation reasons. The AlN interlayer grown in between the heterostructure demonstrates an almost homogeneous thickness, probably reinforcing the 2DEG electrical characteristics. Keywords: AlGaN/GaN; heterostructure; TEM; HRTEM; High Electron Mobility Transistor; Si substrate.
ACCEPTED MANUSCRIPT 1. Introduction The
aluminum
gallium
nitride
(AlGaN)/gallium
nitride
(GaN)
compound
semiconductor material system exhibits many interesting properties making it an ideal candidate for both electronic and optoelectronic devices. In addition to its large
RI PT
band gap (GaN 3.4 eV, aluminum nitride (AlN) 6.2 eV), high breakdown electric strength (1–3x1010 V/cm), high electron saturated drift velocity (2.2x1010 cm/s) and good thermal stability, the AlGaN/GaN heterojunction exhibits large conduction band
SC
discontinuity and strong spontaneous and piezoelectric polarization effect. This fact results in high concentration of 2-dimensional electron gas (2DEG) near the
M AN U
interface, allowing incorporation of a large electric field (>106 V/cm) and high sheet charge (>1013 cm−2) without any doping. So, the AlGaN/GaN high electron mobility transistor (HEMT) has a larger potential application in the high-frequency, hightemperature and large-power areas [1,2].
TE D
The heteroepitaxy of GaN on sapphire and silicon carbide (SiC) has been proven to be a great success, partly because native substrates of GaN are still very expensive. Besides sapphire and SiC, silicon (Si) is an attractive alternative substrate, offering
EP
several advantages such as high thermal conductivity (130 W/(m·K)), the availability
AC C
of low-cost large wafers (12-inch wafer available), and the opportunity of integration with other Si electronics. However, because of the high substrate temperatures (>1000°C) required during metalorganic chemical vap or deposition (MOCVD) growth, a large tensile stress can occur during cooling, causing cracks in the GaN epitaxial layers, because of the large mismatch in the thermal expansion coefficients of GaN (5.59 × 10−6 K−1) and Si (3.59 × 10−6 K−1). Moreover, this mismatch results in bowing of the Si wafers, which makes the lithography process difficult during device fabrication [3]. Furthermore, the large lattice mismatch between GaN and Si (~17%)
ACCEPTED MANUSCRIPT leads to the formation of high-density dislocations during the initial growth and cooling process. Another problem is referred as chemical reactivity between Ga and Si, leading to melt back etching at elevated temperatures during metalorganic vapor phase epitaxy (MOVPE) of GaN. Direct outcomes besides cracking of epitaxial
RI PT
layers are the presence of high dislocation density of grown material and roughening of the layers surface [4,5].
The management of stress and the achievement of crack-free GaN based
SC
heterostructures is a very important issue for the growth of good-quality GaN on silicon and promising results have been reported. More specifically, it was reported
M AN U
[6] that a low temperature (LT)-AlN interlayer could be used to relieve the tension stress in the GaN films grown on the Si substrate and then reduce the cracks in the grown GaN epilayer. Moreover, it could also prevent the threading dislocations from propagating into the overlying GaN. Other ways reported in order to avoid cracks in
TE D
the GaN layer make use of patterned Si substrates (offering free facets to release the large tensile stress from thermal mismatch) or compositionally graded AlGaN layers [4,5].
EP
However, the use of LT-AlN interlayers is a time-consuming complex method due to
AC C
the frequent ramping of temperature, V/III ratio and pressure during growth, giving rise to non-uniformity. Moreover, the growth on patterned silicon substrates is not convenient because it requires extra processing before growth. Finally, graded AlGaN intermediate layers have been successfully introduced in between the GaN layers, in order to prevent cracking of the GaN [7,8]. In this work, we report on the structural characterization of AlGaN/GaN heterostructures grown on Si (111) substrates by MOVPE. Two types of substrates were used: on axis and 4-deg cut off. The structural characterization was performed
ACCEPTED MANUSCRIPT by using conventional and high-resolution Transmission Electron Microscopy (TEM and HRTEM). 2. Experimental
RI PT
The epitaxial growth of the heterostructures on Si (111) substrate was carried out using EMCORE Turbodisc™ D75 MOCVD system operating at low pressures. Trimethylgallium (TMG), trimethylaluminum (TMA), and ammonia (NH3) were used as Ga, Al and N precursors, respectively. The substrates were cleaned using
SC
solvents, and then subjected to an in situ thermal cleaning in flowing H2 at 1100oC
M AN U
for 10 min.
A nucleation AlN layer was deposited with a few seconds TMA preflow before ammonia was introduced into the reactor, in order to prevent thick SiN formation and preserve in the nitride part the epitaxial orientation of trigonal Si (111) symmetry. Moreover, a high temperature AlN layer (HT-AlN) layer was deposited at low V/III
TE D
ratio, at 1230oC temperature and 30Torr pressure, in order to provide a smooth AlN surface and some selection of exact c-plane orientated nucleation grains after full
EP
layer growth. According to XRD measurements, the AlN layer growth conditions used resulted in almost normal c-plane orientation to the substrate surface
AC C
regardless of Si (111) tilt (in both substrate types: on axis and 4-deg off to (111) plane). The rest of the multi-AlGaN buffer structure was deposited in standard MOVPE growth conditions with strain/bow engineering described below. Two samples consisting of a buffer multilayer structure (Sample-A and -B) and one of a HEMT structure (Sample-C) were grown on Si (111) substrates using standard strain engineering approach of multi-AlGaN intermediate layers. The buffer multilayer structures, Sample-A and B, consist of nine different layers grown on “on-axis” and 4 deg off Si (111) substrate respectively. In the HEMT structure (Sample-C), after the
ACCEPTED MANUSCRIPT deposition of the multilayer buffer structure (grown on “on-axis” Si substrate), the AlGaN/AlN/GaN heterostructure was deposited, as well as a final GaN layer as a cap layer. The layers of the epi-structures are described in Table 1.
Nominal thickness (nm)
1
AlN-1
25
2
HT-AlN
256
3
GaN
100
5 6
8
LT-AlN interlayer
9
GaN buffer
13
43-80
102-106
88-111
191-238
155-163
216-223
195-268
227-233
170-244
215-251
468-479
203-247
55-58
460-470
88-98
30
17-21
12.7-23
17-24
1100
1005-1014
1364-1375
268 265 257 100
GaN cap
1115-1117
80
EP
GaN undoped channel AlN spacer layer AlGaN (~20%)
AC C
12
263-265
TE D
GaN
11
227-231
AlGaN-1 (~76% Al) AlGaN-2 (~35% Al) AlGaN-3 (~19% Al)
7
10
238-250
M AN U
4
Experimental Experimental Experimental thickness thickness thickness (nm) (nm) (nm) Sample-A Sample-B Sample-C On-axis 4-deg-off On-axis
SC
Layers numbering
Layers (bottom to top)
RI PT
Table 1: Nominal and measured thicknesses of the examined samples.
1
1.6-2.5
20 19-21
2-2.5
The structural characterization (using TEM and HRTEM) of the samples was carried out using a JEOL 2011 operated at 200kV. Specimen preparation for the TEM experiments, which included mechanical polishing and low-angle argon milling, is described elsewhere [9].
ACCEPTED MANUSCRIPT 3. Results and discussion For all the three types of samples the electron diffraction study reveals that the layers were grown epitaxially on the Si substrate, besides the lattice mismatch on the interface plane. An example is shown in Fig. 1, obtained from a selected area
RI PT
containing part of the Si substrate and the first layers of Sample-B. It is clear for the epitaxial relationships that the [111] Si direction is parallel to the [0001] AlN direction, the [112] Si direction is parallel to [10-10] AlN direction and the [110] Si direction is
SC
parallel to [11-20] AlN direction. The main spots of the two phases are indicated with
M AN U
solid and dashed line circles for Si and AlN respectively.
Fig. 1: A typical electron diffraction
pattern obtained from a selected area containing Si substrate and the first AlN layers along [110].
TE D
The very good epitaxial growth of AlN
layers
on
the
Si
(111)
substrate is clear.
EP
For the structural characterization of epilayers and interfaces cross-sectional TEM
AC C
analysis was carried out. The conventional bright field images in Fig. 2 (a) and (b) show the multilayer structures of Sample-A and B, respectively. In the upper part of the structure, the glue used for the preparation of the samples, clearly indicates the end of the sample. Table 1 gives the nominal and experimental thicknesses of the layers (from bottom to top) measured on the cross-sectional images of Fig. 2.
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
Fig. 2: Low magnification TEM images showing the buffer multilayer structure grown on Si (111) substrate for (a) Sample-A and (b) Sample-B. Contrast changes reveal the nine layers that were grown epitaxially on the substrate.
In the following paragraphs we discuss the obtained results starting from bottom to top.
ACCEPTED MANUSCRIPT 3.1. Multilayer buffer Samples (A and B) 3.1.1. The Si substrate/AlN-1 interface In both cases (Sample-A and -B), a thin amorphous interlayer, probably SiNx, was revealed by HRTEM analysis at the AlN-1 (layer1)/Si substrate interface (Fig. 3), as
RI PT
also mentioned in the work of Kaiser et al. [10]. The layers’ thickness is not homogeneous in all areas varying from around 0.3 up to 2nm. The presence of this layer indicates the efficient relaxation of the large lattice mismatch f300 K=+23.4% at
SC
the AlN-1/Si (111) interface.
Generally, it is well known that GaN is not suitable for direct epitaxial growth on Si
M AN U
substrate suffering from a strong tendency towards the formation of amorphous Si3N4 when the silicon substrate is exposed to ammonia [11]. Moreover, Ishikawa et al. [12] reported that LT-GaN is not suitable as buffer layer because of the melt-back etching of Si by Ga. Thus, in this study similarly to others
TE D
(with Amano et al. [13] being the first who used AlN as an intermediate buffer layer between GaN and Si), the insertion of AlN epitaxial layers is used for GaN growth on silicon and serves not only as wetting layers for GaN but also as barrier layers to
AC C
EP
prevent Ga melt-back etching.
Fig. 3: A HRTEM image showing the amorphous layer between the Si substrate and the AlN-1 layer for both specimens.
SC
RI PT
ACCEPTED MANUSCRIPT
Fig. 4: A HRTEM image showing the existence of an amorphous layer in Sample-A.
M AN U
Moreover, FFT (shown as inset) and filtered IFFT from areas indicated by white curves, disclose the Volmer-Weber island growth in AlN layer.
According to Kai Cheng et al. [8], the growth of the AlN epitaxial layers starts with a Volmer-Weber island growth due to the large lattice mismatch between AlN and the
TE D
Si substrate (17%). In both samples, the first AlN-1 grown layer has a grain-like structure showing a characteristic contrast with bright and dark areas due to a slight misorientation between them. An example of this structure is presented in the
EP
HRTEM image of Fig. 4 for Sample-A, making clear the origin of the contrast and the growth model. The Fast Fourier Transform (FFT) of the region is presented as inset.
AC C
It is clearly observed in both HRTEM image and FFT that AlN is grown in small areas slightly tilted of c-axis. The reconstruction of the lattice fringes via an IFFT using only the indicated by white circles spots, allows the estimation of the tilting angle compared to the substrate perfect orientation. Thus, a typical 8.7 deg tilted AlN grain is illustrated at Fig. 4. On the other hand, there are also regions that are almost caxis (0001) orientated, indicating that the Si (111) crystal orientation is transferred through the amorphous interlayer to the AlN-1 layer [13]. However, at the 4-deg off
ACCEPTED MANUSCRIPT case (Sample-B), we did not find tilted grains exhibiting more than 2 degrees angle from the perfect c-axis orientation. It is also observed that in the slightly tilted regions, the amorphous interlayer thickness varies as compared to the dimensions and the shape for the perfect
RI PT
oriented ones, probably resulting to this change of crystal orientation. The thickness variation is illustrated in the HRTEM image of Fig. 4 with arrows in the amorphous layer, indicating the difference in thickness in the almost c-axis oriented and in the
SC
slightly tilted one.
M AN U
3.1.2. The high temperature (HT)-AlN/GaN layers
On the top of the AlN-1 layer, a HT-AlN layer (layer 2) is deposited. The already described grain structure extends through the 25nm thick (nominal thickness) AlN-1 into the HT-AlN layer and is responsible for the AlN-1/HT-AlN interface not being
TE D
very clearly visible, especially in the case of Sample-A [10]. In Sample-B, due to the almost c-axis oriented grain structure of the AlN-1 layer, the AlN-1/HT-AlN is visible enough as compared to Sample-A. Moreover, in order to minimize the misorientation
EP
of GaN layer islands grown on top of HT-AlN layer, a smooth surface of HT-AlN is required. According to Kai Cheng et al [8], as the thickness of the AlN epitaxial layer
AC C
increases, the island density is reduced with the islands size becoming larger. Furthermore, with AlN layer thickness reaching 160 nm, the surface is almost smooth and coalesced. In our case, the AlN layer shows an even greater thickness, ranging from 200 nm to 240 nm. As a result, a uniform and almost sharp AlN surface is achieved, being a good template for further AlGaN or GaN growth and thus leading to well orientated GaN islands on top.
3.1.3. The AlGaN intermediate layers
ACCEPTED MANUSCRIPT In Fig. 5, the rough interface between GaN (layer 3) and AlGaN-1 (layer 4) is indicated by white colored arrows, showing the partially coalesced islands formed
SC
RI PT
during growth of the GaN layer in Sample-A.
EP
TE D
indicate the rough interface.
M AN U
Fig. 5: A TEM image showing the HT-AlN/GaN/AlGaN-1 layers for Sample-A. The arrows
Fig. 6: A TEM image showing the HT-AlN/GaN/AlGaN-1 layers for Sample-B. The arrows
AC C
indicate the sharp interface.
In Sample-B, the GaN layer thickness is greater leading to a uniform and fully coalesced continuous layer, as shown in Fig. 6. In order to completely avoid or slow down the compressive stress relaxation rate three layers of AlGaN with different compositions were introduced, leading to only modest changes in the lattice constant, i.e., the Al content was varied in steps. This strain buffer approach results in build-up compressive strain and therefore convex
ACCEPTED MANUSCRIPT shape of the structure at growth temperatures. After growth process, the wafer is cooled down to the room temperature and its shape becomes concave, as a result of huge difference in thermal expansion coefficient (CTE) between Si (111) substrate
between them, preserving good in-plane structural quality.
RI PT
and nitride layers. The intermediate layers tend to relax mostly on interfaces
The TEM images shown in Fig. 2 (a), (b) and Fig. 5 and 6, present an enhanced contrast making evident that the HT-AlN/GaN interface is highly defective in the case
SC
of Sample-A. Moreover, many dislocations are observed in the first grown layers in
M AN U
both samples (with the dislocations density being greater in 4 deg cut off Sample-B), due to the lattice mismatch of the materials. However, as moving towards the structure surface, the situation changes.
The AlGaN-1/AlGaN-2 (layer 4/layer 5) interface is indicated by the white colored
TE D
arrows and the AlGaN-2/AlGaN-3 (layer 5/layer 6) interface by the grey colored arrows in Fig. 7 (a) and (b) for Sample-A and Sample-B respectively. Rough interfaces are observed in Sample-A, with undulations following the rough GaN
EP
surface grown on. Some of the interfacial roughness was intentionally introduced in order to assist relief during cool down from MOVPE growth temperature of above
AC C
1050oC.
On the other hand, almost sharp interfaces were observed in between the AlGaN intermediate layers in Sample-B (as already mentioned GaN/AlGaN-1 (layer 3/layer 4) shows an almost sharp interface, resulting to sharp interfaces of epilayers grown on). Moreover, it is reasonable to conclude that the 4 deg cut off Si substrate probably results in less interfacial roughness in between the structure than the onaxis one.
RI PT
ACCEPTED MANUSCRIPT
Fig. 7: TEM images showing (a) rough interfaces between intermediate AlGaN layers of the
SC
multilayer structure of Sample-A, (b) almost sharp interfaces between intermediate AlGaN
M AN U
layers of the multilayer structure of Sample-B. The arrows indicate the interfaces.
The AlGaN-3/GaN (layer 6/layer 7) interface is clearly of better structural quality as compared to previous grown layers for both samples. It demonstrates sharp, crack free plane (that is more evident in Sample-A, where rough interfaces were previously observed). Characteristic TEM and HRTEM images obtained from Sample-A, shown
AC C
EP
Sample-B.
TE D
in Fig. 8 (a) and (b) respectively, clearly present this fact. A similar situation exists in
Fig. 8: (a) A TEM image showing the interface between the AlGaN-3 and GaN layers for Sample-A. The arrows indicate the interface. (b) A HRTEM image of the same sample revealing the good quality of the interface.
Moreover, it is clearly evident that the AlGaN intermediate multilayers perform very well to prevent the propagation of the threading dislocations from the GaN (layer 3)
ACCEPTED MANUSCRIPT film into the upper GaN layer (layer 7). As clearly observed in Fig. 2 (a) and (b) where the whole multilayer structure is shown, dislocations are strongly reduced from the substrate towards the structure surface, for both samples. Thus, the compositionally graded AlGaN buffers work as a dislocation filter, producing an
RI PT
effective way to terminate or annihilate dislocations. Similarly to our study, other studies [8,10,12] have also shown that AlGaN layers of linearly graded composition can reduce dislocation density. Moreover, resulted structure is free of cracks on the
SC
surface, despite the fact that CTE differs almost 2 times between Si and nitride
M AN U
epilayers.
3.1.4. The GaN layer/LT-AlN interlayer/GaN buffer In Sample-A the surface morphology of the GaN layer (layer 7) grown on the stepgraded AlGaN intermediate layers is smooth and crack-free, as shown in Fig. 9. After
TE D
the LT-AlN (layer 8) (15-20nm measured thickness) and buffer GaN (layer 9) (~1µm measured thickness) layers growth, the surface of the multilayer buffer structure grown on “on-axis” Si (111) substrate exhibits even better characteristics (Fig. 12): it
EP
is extremely sharp and crack free, giving promising results for the following HEMT
AC C
structure implementation.
Fig. 9: A TEM image illustrating the GaN layer surface morphology and the almost sharp LTAlN layer in Sample-A.
ACCEPTED MANUSCRIPT On the other hand, in Sample-B, in the [11-20] cut direction, a high density of Vshaped pits (having their end connected with threading dislocations) was observed in GaN layer, which in this case exceeds 400nm thickness, as shown in Fig. 10 (probably because of strain relaxation reasons). The sides of the pits appear to be
RI PT
aligned with specific crystallographic directions. Considering the growth direction and the angle of the V-pits of about 60°, the sidewalls lie in the {1-101} planes. The size of the V-pits varies from 120 up to 230 nm concerning width and from 100 up to 200
SC
nm in depth. The presence of V-pits, initiated by threading dislocations, is probably caused by the increased strain energy and the reduced Ga incorporation on the [10-
M AN U
11] pyramid planes as compared to the [0001] plane [14].
However, the cross section TEM in the [10-10] direction elucidated the nature of these defects. The preferential orientation of the defects observed in our samples suggests the presence of a non-uniform strain distribution during the growth of GaN
TE D
onto off-axis Si substrates, i.e., the strain mainly being relaxed along the [11-20] axis. As a result, the overgrown LT-AlN interlayer, with nominal thickness 30nm,
AC C
EP
“follows” the growth and mounds the underlying V-shaped GaN.
Fig. 10: A TEM image showing the V shaped formations in the [11-20] cut direction in Sample-B.
Fig. 11 (a) shows in higher resolution the right part of the V structure presented in Fig. 10. A Fast Fourier Transform (FFT) was performed in the area indicated by the
ACCEPTED MANUSCRIPT white rectangular (noted in Fig. 11 (a)), as presented in Fig 11 (b). The enlarged part, shown as inset, makes clear the existence of double spots and reveals the presence of materials with slightly different lattice constants. According to the theoretical values for the lattice constants for AlN and GaN layers (4.997 and 5.189
RI PT
respectively), the lattice constant ratio towards c-axis, is 1.038. Taking into account the FFT results, the distance of each one of the double spots gives for the c-axis a ratio equal to 1.029, close enough to the theoretically obtained ratio. The
SC
reconstruction of the lattice fringes via a filtered inverse FFT (IFFT) using only the indicated by no1 and no2 spots represents the areas that give rise to the distinct
M AN U
spots (Fig. 11 (c) and (d) respectively). Moreover, the bright area (right part of a V structure) of HRTEM image of Fig. 11 (a) is proved to be the AlN layer. On the other hand it is clear that the dark areas surrounding the bright one, represent GaN layer. However, it must be noted that the non-sharp boundaries of the GaN reconstructed
TE D
area (Fig.11 (d)) indicate an inter-diffusion of the two materials. This is probably the reason of the presence of modulated brightness fringes in nano-areas very close to
AC C
EP
the AlN layer (Fig.11 (a)).
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
EP
Fig. 11: (a) A HRTEM image showing the right part of V-structure, (b) the resulting FFT. An enlarged part shown as inset makes clear the splitting of the diffracted spots. (c) IFFT using
AC C
only the spots noted with no1 that correspond to AlN and (d) IFFT using only the spots noted with no2 that refer to GaN.
Subsequently, the GaN buffer layer deposited on the top of LT-AlN interlayer starts growing inside these V-pits, resulting in a very good quality, crack-free smooth layer of 1.3µm thickness that can be used as a buffer for the forthcoming growth of the AlGaN/GaN heterostructure for HEMT device application (Fig. 12).
ACCEPTED MANUSCRIPT
RI PT
Fig. 12: A HRTEM image showing a magnified part of the upper GaN buffer layer surface for both samples.
SC
3.2. HEMT Sample (C) The AlGaN/GaN heterostructure
M AN U
HEMT Sample-C consists of a multilayer buffer structure (grown on “on-axis” Si substrate) similar to Sample-A with the AlGaN/GaN heterostructure overgrown. The first nine layers grown representing the buffer structure show quite similar characteristics to Sample-A structure, with differences in thicknesses as presented in
TE D
Table 1. An AlN spacer layer of 1.5-2.5nm thickness was inserted in between the heterostructure, in order to improve the 2DEGs performance [15]. The AlN thickness is almost uniform across the sample, giving rise to the electrical characteristics of the
EP
electron channel. In Fig. 13 (a) and (b), the free surface of Sample-C is shown,
AC C
where characteristic and periodic undulations (V-pits) are revealed, with a height of around 15nm. The GaN-cap layer (indicated by arrows in Fig. 13 (b)), which is not always clearly visible, follows these V-pits of the AlGaN layer. As already reported [16,17], in order to obtain a maximal electron density in the 2DEG electron channel any relaxation of the top Al0.2Ga0.8N should be avoided. The onset of relaxation is determined by the strain via the Al content of the AlxGa1-xN layer and by the thickness of this layer. In our case, in Sample-C, although the low Al concentration (20%) should lead to uniform surface, TEM study showed V-shaped
ACCEPTED MANUSCRIPT pits on the top of the structure. This phenomenon could be attributed to the thickness of the upper layer (19-21 nm), probably exceeding the critical relaxation thickness. Thus, there is a certain critical layer thickness of the AlxGa1-xN at which relaxation
M AN U
SC
RI PT
and stress reduction sets in, regardless of the presence of low Al content.
Fig. 13: (a) A HRTEM image showing a magnified part of the upper layers of Sample-C, (b) A HRTEM image showing the surface V-pits. The arrows indicate the GaN cap layer,
4. Conclusions
TE D
although not clear enough.
In this paper, we studied the structural characteristics of buffer AlGaN/GaN
EP
multilayer structures epitaxially grown on “on-axis” and 4 deg. off Si (111) substrates and HEMT structures grown on “on-axis” Si (111) substrate. The very good epitaxial
AC C
growth of the layers, one on top of the other, was revealed by electron diffraction study. During growth, the step graded AlGaN buffer inserted in between the GaN layers, behaves as a defect filter, preventing the threading dislocations existing in the first grown layers from propagating into the GaN layer (grown after the AlGaN layers). Moreover, the upper GaN layer in 4 deg. cut off Sample-B shows V-pits, with the AlN layer grown upon it beginning to grow into these pits. Therefore, a good quality crack-free and smooth GaN buffer surface is achieved for both buffer Samples-A and B (despite the fact that CTE differs almost 2 times between Si and
ACCEPTED MANUSCRIPT nitride epilayers), suitable for the forthcoming AlGaN/GaN heterostructure growth, despite the lattice mismatch in the Si/AlN interface and the cut off angle of Sample B (4 deg.). Finally, after the growth of the AlGaN/GaN heterostructure on the buffer structure
RI PT
similar to Sample-A (on “on-axis” Si substrate), using an AlN strain relieving interlayer, characteristic and periodic undulations (V-pits) are revealed on the surface, due to strain relaxation reasons. The AlN interlayer demonstrates an almost
SC
uniform thickness, probably reinforcing the 2DEGs electrical properties.
M AN U
Acknowledgements
Funding: This work was supported by the IKY Fellowships of Excellence for Postgraduate Studies in Greece-SIEMENS Program; the Greek General Secretariat for Research and Technology, contract SAE 013/8-2009SE 01380012; and the JU
TE D
ENIAC Project LAST POWER Large Area silicon carbide Substrates and heteroepitaxial GaN for POWER device applications [grant number 120218].
Y.-F. Wu, D. Kapolnek, J.P. Ibbetson, P. Parikh, B.P. Keller, U.K. Mishra,
AC C
[1]
EP
References
Very-high power density AlGaN/GaN HEMTs, IEEE Trans. Electron Devices. 48 (2001) 586–590. doi:10.1109/16.906455.
[2]
W. Luo, X. Wang, H. Xiao, C. Wang, J. Ran, L. Guo, J. Li, H. Liu, Y. Chen, F. Yang, J. Li, Growth and fabrication of AlGaN/GaN HEMT based on Si (111) substrates by MOCVD, Microelectronics J. 39 (2008) 1108–1111. doi:10.1016/j.mejo.2008.01.083.
[3]
T. Watanabe, J. Ohta, T. Kondo, M. Ohashi, K. Ueno, A. Kobayashi, H.
ACCEPTED MANUSCRIPT Fujioka, AlGaN/GaN heterostructure prepared on a Si (110) substrate via pulsed sputtering, Appl. Phys. Lett. 104 (2014) 13–17. doi:10.1063/1.4876449. [4]
J. Liu, J. Zhang, Q. Mao, X. Wu, F. Jiang, Effects of AlN interlayer on growth of GaN-based LED on patterned silicon substrate, CrystEngComm. 15 (2013)
[5]
RI PT
3372. doi:10.1039/c3ce27059a.
J. Cheng, X. Yang, L. Sang, L. Guo, J. Zhang, J. Wang, C. He, L. Zhang, M. Wang, F. Xu, N. Tang, Z. Qin, X. Wang, B. Shen, Growth of high quality and
SC
uniformity AlGaN/GaN heterostructures on Si substrates using a single AlGaN
doi:10.1038/srep23020. [6]
M AN U
layer with low Al composition., Sci. Rep. 6 (2016) 23020.
A. Dadgar, J. Bläsing, A. Diez, A. Alam, M. Heuken, A. Krost, Metalorganic Chemical Vapor Phase Epitaxy of Crack-Free GaN on Si (111) Exceeding 1 µm in Thickness, Jpn. J. Appl. Phys. 39 (2000) L1183–L1185.
[7]
TE D
doi:10.1143/JJAP.39.L1183.
K. Cheng, H. Liang, M. Van Hove, K. Geens, B. De Jaeger, P. Srivastava, X. Kang, P. Favia, H. Bender, S. Decoutere, J. Dekoster, J.I. Del Agua Borniquel,
EP
S.W. Jun, H. Chung, AlGaN/GaN/AlGaN double heterostructures grown on
AC C
200mm silicon (111) substrates with high electron mobility, Appl. Phys. Express. 5 (2012). doi:10.1143/APEX.5.011002.
[8]
K. Cheng, M. Leys, S. Degroote, B. Daele, S. Boeykens, J. Derluyn, M. Germain, G. Tendeloo, J. Engelen, G. Borghs, Flat GaN epitaxial layers grown on Si(111) by metalorganic vapor phase epitaxy using step-graded AlGaN intermediate layers, J. Electron. Mater. 35 (2006) 592–598. doi:10.1007/s11664-006-0105-1.
[9]
L.M. Sorokin, A.V. Myasoedov, A.E. Kalmykov, D.A. Kirilenko, V.N. Bessolov,
ACCEPTED MANUSCRIPT S.A. Kukushkin, TEM investigation of semipolar GaN layers grown on Si (001) off cut substrates, Semicond. Sci. Technol. 30 (2015) 114002. doi:10.1088/0268-1242/30/11/114002. [10] S. Kaiser, M. Jakob, J. Zweck, W. Gebhardt, O. Ambacher, R. Dimitrov, a. T.
RI PT
Schremer, J. a. Smart, J.R. Shealy, Structural properties of AlGaN/GaN
heterostructures on Si(111) substrates suitable for high-electron mobility
transistors, J. Vac. Sci. Technol. B Microelectron. Nanom. Struct. 18 (2000)
SC
733. doi:10.1116/1.591268.
[11] W. Wang, H. Wang, W. Yang, Y. Zhu, G. Li, A new approach to epitaxially
M AN U
grow high-quality GaN films on Si substrates: the combination of MBE and PLD, Sci. Rep. 6 (2016) 24448. doi:10.1038/srep24448. [12] H. Ishikawa, K. Yamamoto, T. Egawa, T. Soga, T. Jimbo, M. Umeno, Thermal stability of GaN on (111) Si substrate, J. Cryst. Growth. 189–190 (1998) 178–
TE D
182. doi:10.1016/S0022-0248(98)00223-1.
[13] H. Amano, N. Sawaki, I. Akasaki, Y. Yoyoda, Metalorganic vapor phase epitaxial growth of a high quality GaN film using an AlN buffer layer, Appl.
EP
Phys. Lett. 48 (1986) 353–355. doi:10.1063/1.96549.
AC C
[14] C.J. Sun, M. Zubair Anwar, Q. Chen, J.W. Yang, M. Asif Khan, M.S. Shur, a. D. Bykhovski, Z. Liliental-Weber, C. Kisielowski, M. Smith, J.Y. Lin, H.X. Xiang, Quantum shift of band-edge stimulated emission in InGaN–GaN multiple quantum well light-emitting diodes, Appl. Phys. Lett. 70 (1997) 2978. doi:10.1063/1.118762. [15] J. Su, B. Krishnan, A. Paranjpe, G.D. Papasouliotis, Influence of MOCVD Growth Conditions on the Two-dimensional Electron Gas in AlGaN/GaN Heterostructures, in: CS Mantech Conf., 2014: pp. 251–254.
ACCEPTED MANUSCRIPT [16] C. He, Z. Qin, F. Xu, L. Zhang, J. Wang, M. Hou, Mechanism of stress-driven composition evolution during hetero-epitaxy in a ternary AlGaN system, Nat. Publ. Gr. 6 (2016) 1–5. doi:10.1038/srep25124. [17] Z.-Y. Liu, J.-C. Zhang, H.-T. Duan, J.-S. Xue, Z.-Y. Lin, J.-C. Ma, X.-Y. Xue, Y.
RI PT
Hao, Effects of the strain relaxation of an AlGaN barrier layer induced by
various cap layers on the transport properties in AlGaN/GaN heterostructures,
AC C
EP
TE D
M AN U
SC
Chinese Phys. B. 20 (2011) 97701. doi:10.1088/1674-1056/20/9/097701.
ACCEPTED MANUSCRIPT
FIGURE CAPTIONS
Fig. 1: A typical electron diffraction pattern obtained from a selected area containing Si
RI PT
substrate and the first AlN layers along [110]. The very good epitaxial growth of AlN layers on the Si (111) substrate is clear.
Fig. 2: Low magnification TEM images showing the buffer multilayer structure grown on Si
M AN U
that were grown epitaxially on the substrate.
SC
(111) substrate for (a) Sample-A and (b) Sample-B. Contrast changes reveal the nine layers
Fig. 3: A HRTEM image showing the amorphous layer between the Si substrate and the AlN-1 layer for both specimens.
Fig. 4: A HRTEM image showing the existence of an amorphous layer in Sample-A.
TE D
Moreover, FFT (shown as inset) and filtered IFFT from areas indicated by white curves, disclose the Volmer-Weber island growth in AlN layer.
Fig. 5: A TEM image showing the HT-AlN/GaN/AlGaN-1 layers for Sample-A. The arrows
EP
indicate the rough interface.
AC C
Fig. 6: A TEM image showing the HT-AlN/GaN/AlGaN-1 layers for Sample-B. The arrows indicate the interface.
Fig. 7: TEM images showing (a) rough interfaces between intermediate AlGaN layers of the multilayer structure of Sample-A, (b) almost sharp interfaces between intermediate AlGaN layers of the multilayer structure of Sample-B. The arrows indicate the interfaces.
ACCEPTED MANUSCRIPT Fig. 8: (a) A TEM image showing the interface between the AlGaN-3 and GaN layers for Sample-A. The arrows indicate the interface. (b) A HRTEM image of the same sample revealing the good quality of the interface.
Fig. 9: A TEM image illustrating the GaN layer surface morphology and the almost sharp LT-
RI PT
AlN layer in Sample-A.
Fig. 10: A TEM image showing the V shaped formations in the [11-20] cut direction in
SC
Sample-B.
Fig. 11: (a) A HRTEM image showing the right part of V-structure, (b) the resulting FFT. An
M AN U
enlarged part shown as inset makes clear the splitting of the diffracted spots. (c) IFFT using only the spots noted with no1 that correspond to AlN and (d) IFFT using only the spots noted with no2 that refer to GaN.
Fig. 12: A HRTEM image showing a magnified part of the upper GaN buffer layer surface for
TE D
both samples.
Fig. 13: (a) A HRTEM image showing a magnified part of the upper layers of Sample-C, (b) A HRTEM image showing the surface V-pits. The arrows indicate the GaN cap layer,
AC C
EP
although not clear enough.
ACCEPTED MANUSCRIPT
HIGHLIGHTS
RI PT
SC M AN U TE D
•
EP
• •
Structural properties of AlGaN/GaN buffer heterostructures grown on “on axis” and 4 deg off Si (111) substrates are investigated. The introduced step-graded AlGaN layers work as a dislocation filter. Almost c-plane orientated nucleation grains are achieved after full AlN layer growth. A crack free smooth surface of the multilayer buffer structure is achieved in both cases, ideal for the forthcoming AlGaN/GaN heterostructure growth.
AC C
•