Scholarly article on topic 'Effect of SiC substrate properties on structural perfection and electrical parameters of AlGaN/GaN layers'

Effect of SiC substrate properties on structural perfection and electrical parameters of AlGaN/GaN layers Academic research paper on "Materials engineering"

CC BY-NC-ND
0
0
Share paper
Academic journal
Modern Electronic Materials
OECD Field of science
Keywords
{"Epitaxial layers" / "Nitride heterostructures" / "Disk-shaped inclusions" / Impurity}

Abstract of research paper on Materials engineering, author of scientific article — Kira L. Enisherlova, Tatyana F. Rusak, Vyacheslav I. Korneev, A.N. Zazulina

Abstract We have analyzed the effect of volume and surface defects in SiC substrates on the structure and electrophysical parameters of AlGaN/GaN epitaxial heterostructures grown on them. Regions with internal stresses usually induced by carbon rich disk-shaped inclusions have been detected in the initial substrates. We show experimentally that the presence of internal stresses in SiC could affect the microroughness of the epitaxial films in regions grown on the stressed areas. An abrupt deterioration of electrophysical parameters occurs in epitaxial film regions growing above internally stressed areas in the substrate. AlGaN/GaN layers contain impurities delivered to their bulk during epitaxy or preparatory operations.

Academic research paper on topic "Effect of SiC substrate properties on structural perfection and electrical parameters of AlGaN/GaN layers"

Author's Accepted Manuscript

Effect of SiC substrate properties on structural perfection and electrical parameters of AlGaN/GaN layers

Modern

Electronic

Materials

Kira L. Enisherlova, Tatyana F. Rusak, Vyacheslav I. Korneev, A.N. Zazulina

www.elsevier.con/bcate/iDem

PII: S2452-1779(17)30028-2

DOI: http ://dx. doi.org/10.1016/j.moem.2017.03.001

Reference: MOEM51

To appear in: Modern Electronic Materials

Received date: 1 December 2016 Accepted date: 24 February 2017

Cite this article as: Kira L. Enisherlova, Tatyana F. Rusak, Vyacheslav I. Korneev and A.N. Zazulina, Effect of SiC substrate properties on structura perfection and electrical parameters of AlGaN/GaN layers, Modern Electronic Materials, http://dx.doi.org/10.1016/j.moem.2017.03.001

This is a PDF file of an unedited manuscript that has been accepted fo publication. As a service to our customers we are providing this early version o the manuscript. The manuscript will undergo copyediting, typesetting, an< review of the resulting galley proof before it is published in its final citable 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

Effect of SiC substrate properties on structural perfection and electrical parameters of AlGaN/GaN layers

Kira L. Enisherlova1, Tatyana F. Rusak1, Vyacheslav I. Korneev1, A. N. Zazulina1 1JSC S&PE "Pulsar",

27 Okruzhnoiproezd, Moscow 105187, Russia

Kira L. Enisherlova (Enisherlova@pulsarnpp.ru) - author for correspondence

K. L. Enisherlova1 -Dr. Sci. (Eng.), Head of Laboratory (Enisherlova@pulsarnpp.ru); T. F. Rusak1 - Senior Researcher; V. I. Korneev1; A. N. Zazulina1

Abstract. We have analyzed the effect of volume and surface defects in SiC substrates on the structure and electrophysical parameters of AlGaN/GaN epitaxial heterostructures grown on them. Regions with internal stresses usually induced by carbon rich disk-shaped inclusions have been detected in the initial substrates. We show experimentally that the presence of internal stresses in SiC could affect the microroughness of the epitaxial films in regions grown on the stressed areas. An abrupt deterioration of electrophysical parameters occurs in epitaxial film regions growing above internally stressed areas in the substrate. AlGaN/GaN layers contain impurities delivered to their bulk during epitaxy or preparatory operations.

Keywords: epitaxial layers, nitride heterostructures, disk-shaped inclusions, impurity. Introduction

Single crystal silicon carbide is more promising substrate material compared to sapphire and silicon for the epitaxial growth of AlGaN/GaN heterostructures especially for high-power microelectronics because of the higher heat conductivity, excellent insulating properties and a good match between the SiC and GaN lattices in the basic crystallographic planes. World's leading high power device manufacturers have already turned attention to this material. For example, Cree Inc. defined its main trend as the development and fabrication of SiC semiconductor devices. It has become the world's leader in the fabrication of silicon carbide single crystal substrates and epitaxial structures grown on this material [1]. The leaders have already started large scale fabrication of ultrafast HEMTs on SiC substrates [2]. The quality of SiC substrates depends largely on the parameters of ingots that are still imperfect, although since 2006 several international projects have been started in Europe for the improvement of 76-100 mm 4H-polytype single crystal SiC technology [3, 4]. AlGaN/GaN heterostructures are typically grown by epitaxy on single crystal SiC grown in the [0001] direction and having a 4H or 6H polytype lattice. Each polytype has a unique interchanging set of similar Si-C layer pairs. These layers are turned relative to the horizontal axis so that each of the polytypes may contain their cubic or hexagonal closely packed lattices with specific proportions [5]. Therefore the lattice parameter a of all the polytypes is almost the same. Although the lattice parameter c determining the identity period differs between the polytypes, it often turns out during the growth of a specific polytype ingot that other polytype regions form inside it. In some instances different polytype modifications sequentially displace one another throughout the whole growth surface producing the so-called sandwich like structures [6].

It was found that polytype inclusions are a cause of domain formation in growing ingots. However the most frequent cause of domain formation is the screw dislocation growth mechanism with a large number of independent precipitation sites. Domain boundaries are typically tilted low angle boundaries. The misorientation of two adjacent domains normally caused by a rotation about the [0001] axis is within 2-8 arc sec as X-ray diffraction analysis suggests. Peripheral regions may contain block structures with block misorientations of up to 40 arc sec. Silicon carbide ingots may also contain more types of structural defects, e.g. micropores, screw, edge and mixed dislocations, carbon and other inclusions. It is believed that the most important defect causing catastrophic damage to SiC high power devices are micropores -defects stretched along the [0001] direction and having core cavities. Growing silicon carbide crystals may contain local mechanical stresses due to high temperature gradients, heterogeneous distribution of priming power, polytypes and other inclusions [7]. Recent success in large diameter 4 and 6 polytype SiC technology (to 150 mm) provided for substantially higher structural perfection of the material thanks to optimized growth parameters [8]. However, there are only little if any data on the mechanism whereby the abovementioned structural features and imperfections in silicon carbide affect the quality of the epitaxial films. Silicon carbide wafer slice processing have not finally been resolved, and this technology is still proprietary in most cases.

Reduction in the number of SiC and GaN lattice mismatch defects is achieved by providing Al concentration in the AlGaN layer at which a herm cells of AlGaN structure along axic C will be close to atomic steps of a vicinal SiC surface. The parameter a of SiC (0.308 nm) is closer to that of AlN (0.311 nm) than to that of GaN (0.3189 nm). This allows to grow AlN and AlGaN layers of variable composition as a herm layers on SiC. These layers are selected as follows:

1. Maximum match between the substrate and film lattices. layers

2. Minimum lattice mismatch elastic stresses.

3 Minimum density of growth dislocations caused by stress relaxation.

SiC surfaces are only little wetted by GaN. Therefore AlN herm layers are good wetting agents. There are indications of possible impurity diffusion to GaN from substrates, e.g. oxygen from sapphire substrates [9] and silicon from silicon carbide ones [10].

Currently Russian manufacturers mainly use imported SiC substrates. Therefore the aim of this work was to assess the possibility that the abovementioned defects are present in substrates on which heterostructures are fabricated and to analyze the mechanism of their impact on the structural perfection of the material and some electrical parameters of the epitaxial layers.

Experimental

For the study we chose the most widely used heterostructure, AlGaN/GaN - a classic single-junction one [9]. This heterostructure is a typical basis for HEMT HF transistors worldwide. It contains a relatively thick (2-3 ^m) buffer GaN layer and a thin (10-25 nm) AlxGa(1-x)N layer with the aluminum concentration x = 0.26-0.3.

We tested 053 mm MOCVD AlGaN/GaN heterostructures supplied by Russian and foreign companies, with a barrier layer thickness of 200 nm. Most of these structures were not additionally doped and had upper /-AlGaN and /-GaN layers 2.0-2.5 nm in thickness. The specimens were oriented along the [0001] direction and had no additional rotation. In this case one should expect a more significant impact orientation and macrorelief errors on surface morphology and uniformity of the epitaxial film.

The macroscopic picture of elastic stress fields distribution in the substrate bulk was controlled with a pilot fast polarized light control instrument designed at our institute [10]. The screen image shows the elastic stress distribution across the entire wafer. More detailed examination of elastic stress areas was carried out under an optical microscope in Nomarsky contrast. The composition of the inclusions was analyzed under a CAMSCAM-S4 scanning electron microscope with an INCAENERGY X-ray attachment at a 20 keV accelerating voltage. To

avoid specimen surface charging by the electron beam we sputtered some specimen surfaces with a gold layer.

Structural defects in the GaN layer were studied by selective etching in phosphoric acid at 200°C. Silicon carbide substrates were etched in a KOH melt at 500°C. To measure dislocation density profiles in the epitaxial layers we used a special method including sequential plasma polishing etching with step-by-step removed layer depth control followed by selective chemical etching.

To study the distribution of dislocation pile-ups along the diameter of the epitaxial layers we used the new local point etching method.

To assess the effect of structural defects on the electrical parameters we recorded HF C-V and VI curves of the mercury probe / heterostructure surface Schottky barrier at room temperature (Troom) on an MDC CSM/Win Semiconductor Measurement System with a 0.005 cm-2 mercury probe with a planar position of the probe and the second electrode on the tested surface. The use of a mercury probe allows measurements to be carried out directly on the as-grown structures, analyzing and comparing different heterostructure areas.

Results and Discussion

Figure 1 shows (a-c) the most typical pictures of distribution and locations of stressed regions on the heterostructure surfaces and (d-f) the most stressed regions as seen under a microscope in Nomarsky contrast. One can see contrasting or smeared bands and local stress rosettes with a strong black and white contrast. Table shows results of electrical measurements which showed the following:

- even very large internal stress regions have only little if any effect on the Hall data;

- internal stress regions exhibit abrupt leakage current growth (the reverse I-V curve branch);

- internal stress regions cause a stronger negative shift of the cutoff voltage Uut and a growth of Cmin at lower frequencies (100-1000 Hz). Moreover, heterostructures with internal stress regions have a larger scatter of the electrical parameters.

Table. Electrophysical parameters of test AlGaN/GaN heterostructures.

Structure Hall measurements C- V measurements V-I measurements

Mobility cm2/V • s Carrier concentration , cm-2 Layer resistivity, Ohm/sq. Cmax, pF Cutoff voltage, V Reverse currents, A

1 2070 1.35 • 1013 224 1450— 1550 -6.5 ... -7 3 • 10-4 - 1.5 • 10-3

2 2070 1.3 • 1013 228 1300— 1400 -7 ... -8 5 • 10-4 (reg. 1) - 1 • 10-8 (reg. 2)

3 2080 1.3 • 1013 227 1300— 1330 -4.3 ... -4.5 2 • 10-10 - 1 • 10-8

The insensitivity of Hall measurements to stressed regions is most probably caused by the averaging of the carrier concentration data and mobility in the channel over the whole heterostructure.

The C-V and V-I measurements are carried out in the local regions of the heterostructure. This approach provides a statistically meaningful correlation of the tested parameters with stressed regions. Local leakage current rise and Cmin growth at low frequencies may indicate the presence

of parasitic channels in the GaN buffer layer and at the AlGaN/GaN interface. The negative shift

of the cutoff voltage suggests the presence of positive charge in the barrier layer above the elastically stressed regions.

These divergence of electrical parameters can be caused by the inheritance of elastic stresses in the growing GaN layer or an evolution of the growth structural defect system in the epitaxial film.

The direct effect of elastic stresses may occur through the interaction between the elastic stress fields and the macroscopic polarization pattern produced at the nitride compound heterostructure interface. The electric field produced by macroscopic polarization hinders electron movement in one direction thus favoring the formation of two-dimensional electron gas [11]. Without external fields the overall macroscopic polarization is determined by the sum of the spontaneous polarization in the equilibrium structure and the piezoelectric polarization caused by deformations at the heterostructure interface. Our calculations show that the sign of the piezoelectric polarization is the same as that of the spontaneous polarization. Therefore the overall macroscopic polarization increases.

The elastic stress field distribution patterns shown in Table 1 are quite inhomogeneous and are formed in the bulk of the substrate. However, the elastic stress of SiC surface atomic bonds is inherited by the growing film during the epitaxial growth of GaN on SiC. Local inhomogeneous stress distribution patterns inherited by the film distort the homogeneity of the macroscopic polarization. This is a probable cause of the degradation and scatter of the electrical parameters in heterostructures with clearly visible stressed regions.

Stressed regions in the substrate and in the film may also lead to predominant redistribution and accumulation of epitaxial film impurities in the stressed regions. This is confirmed by SEM with X-ray additional attachment for the stressed regions on the surface layer (Fig. 2). A region with a high stress concentration is found to contain a number of occasional elements, and the SiC surface after GaN removal only contained oxygen because of SiC oxidation. To establish a correlation between the experimental stress pattern and the formation of defects in the growing film we sequentially etched GaN layers in phosphoric acid at 200°C. As we used droplet etching method, we selected the etchant volume and etching time individually for each experiment.

Figure 3 shows results for successive etching of a stressed region. After the initial short etching step, one can clearly see, in Polaroid light, circular etching patterns corresponding to site points that showed stress rosettes before etching (Fig. 3 a).

After the second etching step (photo with Polaroids off) the dislocation distribution is generally uniform (Fig. 3 b). As at least a 0.1 ^m layer was removed after the first two etching steps, the pattern seen in Fig. 3 b corresponds to the dislocation distribution in the top part of the GaN layer in the vicinity of the AlGaN working layer.

In regions with the highest stress concentration the GaN layer is etched down to the SiC substrate forming a slightly faceted etching pattern. Around these etching patterns and in some other areas (e.g. I, II and III) one can see arc-shaped dislocation pile-ups (Fig. 3 b). Further etching led to fast growth of these etching patterns at very low GaN vertical etching rates (Fig. 3 c). The overall stress distribution pattern remains generally the same but is less clear due to multiple etch pits. The SiC surface inside major etching patterns exhibits small black disk-shaped inclusions (Fig. 3 d). Element analysis in these areas showed the presence of silicon and carbon. The carbon content was far greater than expected according to stoichiometry. Along with oxygen which is unavoidable after SiC etching, no other impurities were found and therefore these inclusions were pure carbon.

One can see by changing Polaroid rotation degree that these inclusions are high stress concentration centers (Fig. 3 c).

It can be seen from Fig. 3 that after long-term etching the etching patterns growing on dislocations hinder further observation. To analyze the dislocation pattern in deeper GaN layers we removed a 1.5-2 ^m GaN layer from the heterostructure surface with plasma etching and then

sequentially etched the surface in phosphoric acid. Tracing the evolution of discrete etching patterns one can judge about the layer dissolution sequence and hence the epitaxial film growth sequence (Fig. 4).

After short-term selective etching one can see etching patterns in the form of hexagonal pyramids with sharp nodes (Fig. 4 a). The dislocation density is almost the same as in the upper layer (Fig. 4 a). Further etching steps show flat-bottomed hexagonal etch pits with depths of less than 0.1 ^m. Later on their size increased and their depth decreased until complete dissolution (Fig. 4 b). The dislocation density in the lowermost layer is far lower (Fig. 4 c). The size of the largest etching patterns the bottom of which reached the substrate increased dramatically. Their central parts often contained carbon inclusions as shown in Fig. 3.

These observations suggest that the material is moved by layers. Each newly formed etch pit reaches the next layer and starts growing in size while dissolving until the whole layer is removed. The last to dissolve are areas around etching patterns the bottom of which reached the substrate within the first minutes of etching. Figure 4 c shows that the thickness of the remaining layer in these areas is greater compared to the rest of the surface. One can assume that the layer grows in a reverse sequence. Large blocks with relatively perfect structure grow following a screw mechanism. When the first blocks start agglomerating their boundary lattice mismatch defects act as new layer precipitation centers. The structure of these new layers is less perfect than the layer of the primary blocks as can be seen from Fig. 4 a-c.

This growth sequence hypothesis suggests that the dislocation pattern of the upper working GaN layer is not inherited from the substrate; rather, it develops depending on the growth conditions of the previous layer. However, the height and homogeneity of the film surface microroughness depend largely on the sizes of the initial blocks growing on the substrate. To clarify the role of structural defects in silicon carbide in the formation of epitaxial films we etched SiC surfaces (Fig. 5) after GaN layer moving in the same area as shown in Fig. 3 in a KOH melt at 500°C. Typically SiC is etched at 600°C or higher, but the rate of high-temperature etching in the vicinity of inclusions is far higher than on the rest of the surface, and there is the risk of stripping the inclusions. The etching patterns observed after oriented etching exactly correspond to the micropore etching patterns reported earlier [7]. These etching patterns do not provide exact information on the initial cause of their formation, i.e. a micropore or an inclusion. Slow etching at 500°C retains carbon inclusions. Comparison between Figs. 3 a and 5 a showed that GaN layer moving did not change the stress pattern. The etched silicon carbide surface exhibited etching patterns, some of which are discrete and random while others form low angle boundaries between domains. There is no direct correlation between etching patterns on silicon carbide surface and in the GaN layer even after long-term etching because the dislocation density in silicon carbide (102-103 cm-2) is 2-3 orders of magnitude lower than that in GaN. Stress free regions (A Region) exhibit two types of etching patterns forming low angle boundaries: larger dark hexagonal pits and poorly seen smaller triangular pits.

In accordance with an earlier classification [7], these pits correspond to micropores and edge component dislocations, respectively. In stressed regions (B Region) hexagonal etching patterns are replaced by carbon inclusions acting as stress concentrators. Micropores are believed to be the most harmful defects in device epitaxial structures because some of micropores contain carbon inclusions forming large stressed regions.

The presented epitaxial film growth model suggests that the larger the block size in the device fabrication layer the larger the block height. Thus the growth of large blocks around inclusions and stress concentration regions leads to an increase in the size and height of surface microroughness in the stressed regions. Prophilographic measurements showed that the height of surface microroughness in stress free regions is within the AlGaN working layer thickness (20 nm) while features up to 50-60 ^m in height may occur in stressed regions (Figs. 6 and 7). This is far greater than the thickness of the working layer, and these features may cause layer inhomogeneity and structural imperfection.

Summary

We have shown experimentally that elastically stressed regions in SiC substrate cause abrupt deterioration of the properties of AlGaN/GaN epitaxial film grown on these substrates, e.g. higher leakage currents, the cutoff voltages shift to negative magnitudes and parameter scatter. We established that the stress concentrators determining the elastic stress pattern are disk-shaped carbon-rich inclusions up to 5 ^m in size. These inclusions may concentrate along low angle boundaries or occur randomly.

The SiC dislocation structure does not determine the density and distribution of dislocations in

c /T O ^ O

the upper GaN layer (5 • 10-10 cm- ). The dislocation density in SiC is 10 -10 cm- outside low angle boundaries.

The epitaxial films contain impurities not complying with the stoichiometric composition of the multilayered structure; these impurities are incorporated into the film during its growth. Experimental proof is provided that inclusions acting as highest stress concentrators in SiC provide favorable conditions for the predominant growth of large precipitate blocks during epitaxy, and therefore the surface morphology changes. The height of microroughnesses in stressed regions may be far greater than the thickness of the working AlGaN/GaN layer. This may cause layer inhomogeneity and structural imperfection.

References

1. Kurysheva E. Power equipment company Cree silicon carbide. Komponenty i tekhnologii = Components and Technologies, 2011, no. 6, pp. 106-110. (In Russ.)

2. Maiskaya V. Wireless components - millimeter wave. Elektronika: NTB, 2011, no. 6 (00112), pp. 42-49. (In Russ.)

3. Loboda M. J., Chung G., Carlson E., Drachev R., Hansen D., Sanchez E., Wan J., Zhang J. Advances in SiC substrates for power and energy applications. CS Mantech Conf. Palm Springs (CA, USA), 2011. http://csmantech.org/0ldSite/Digests/2011/papers/8b.3.pdf

4. Waltereit P., Muller S., Kirste L., Prescher M., Storm S., Weber A., Schauwecker B., Hosch M., Splettstober J. Development of an epitaxial growth process on European SiC substrates for a low leakage GaN HEMT technology with power added efficiencies around and 65 %. CS Mantech Conf New Orleans (LA, USA), 2013. Pp. 121-124. http://csmantech.org/0ldSite/Digests/2013/papers/036.pdf

5. Danilin V. N., Dokuchaev Yu. P., Zhukova T. A., Komarov M. A. Powerful high-temperature and a new generation of radiation-resistant microwave devices on wideband heterojunction structures of AlGaN/GaN. Electronnaya Tekhnika. Series 1: Elektronika SVCH = Electronic Engineering. Ser. 1: Microwave Electronics, 2001, no. 1, p. 137. (In Russ.)

6. Avrov D. D., Dorozhkin S. I., Tairov Yu. M., Lebedev A. O. Domain formation during syntaxy of polytypes of silicon carbide. Semiconductors, 2007, vol. 41, no. 12, pp. 1389-1393. DOI: 10.1134/S1063782607120019

7. Avrov D. D., Bulatov A. V., Dorozhkin S. I., Tairov Y. M., Fadeev A. Y., Lebedev A. O. Mechanisms of defect formation in ingots of 4H Silicon Carbide Polytype. Semiconductors, 2011, vol. 45, no. 3, pp. 277-283. DOI: 10.1134/S1063782611030055

8. Ruland G., Wu P., Xu X., Rengarajan V., Zwieback I., Gupta A., Ramm M. Status of large diameter SiC single crystals. ECS Transaction, 2014, vol. 64(7), pp. 27-33. DOI: 10.1149/06407.0027ecst

9. Alekseev A., Krasovitskii D., Petrov S., Chalyi V. Multilayer heterostructures AlN/AlGaN/GaN/AlGaN - the foundation of a new component base of solid-state microwave electronics. Komponenty i tekhnologii = Components and Technologies, 2008, no. 2, p. 138. (In Russ.)

10. Guskov B. L., Kontsevoi Yu. A. Rapid method of control defects SiC wafers and heterostructures AlGaN/GaN/SiC. Elektronnaya tekhnika. Series 2. Poluprovodnikovye pribory = Electronic engineering. Ser. 2. Semiconductor devices, 2010, no. 2(225), pp. 69-70. (In Russ.)

11. Supryadkina I. A., Bazhanov D. I., Abgaryan K. K., Mutigullin I. V. Study of the polarizations of (Al,Ga,AlGa)N nitride compounds and the charge density of various interfaces based on them. Semiconductors, 2013, vol. 47, no. 12, pp. 1621-1625. DOI: 10.1134/S106378261312018X

Heterostructure 1 Region A

Heterostructure 2

Region B

Heterostructure 3 Region C

500 urn ;

/aflbs«

Fig. 1. Typical distributions of elastic stress fields in heterostructures:

(a-c) macroimage of elastic stress field distribution in heterostructure bulk; (d-f) local

heterostructure region images in Nomarsky contrast.

Figure 2. Element analysis of inclusions in A region of GaN layer above elastically stressed region and SiC surface:

(a) stressed region in GaN film; (b) element weight percentages in point A; (c) SiC surface after GaN stripping and KHO etching; (d) element weight percentages in point B.

Element Element content, %

weight atomic

CK 59.59 76.55

OK 3.01 2.91

SiK 37.40 20.55

Total 100

Carbon Inclusion

Figure 3. Successive selective etching of GaN epitaxial layer in elastically stressed region: (a) 5 min etching; (b) 10 min, Polaroids off; (c) B region, 40 min etching, Polaroids on; (d) C region, 60 min etching, element analysis in inclusions region.

Figure 4. GaN epitaxial film dissolution and growth sequence:

(a-c) etching pattern transformation during dissolution; (d) epitaxial film growth and dissolution sequence schematic.

Figure 5. Silicon carbide surface etching after GaN layer stripping:

(a) stress distribution pattern in the region shown in Fig. 3 a; (b) same region with Polaroids off.

Fig. 6. Surface features in elastically stressed regions:

(a) comparison of large surface features and device fragments; (b and c) carbon inclusions in the center of largest surface features.

Surface range of microroughness measuring, \im

Fig. 7. Comparison of (a-c) surface features for different surface regions and (d-f) respective profilograms for (a, b, c and d) AlGaN structure on SiC and (c and f) AlGaN structure on AI2O3: (a and d) stressed regions and (b, c, e and f) stress free regions.