Glide of threading edge dislocations after basal plane dislocation conversion during 4H–SiC epitaxial growth

Journal of Crystal Growth 418 (2015) 7–14

Contents lists available at ScienceDirect

Journal of Crystal Growth journal homepage: www.elsevier.com/locate/jcrysgro

Glide of threading edge dislocations after basal plane dislocation conversion during 4H–SiC epitaxial growth Mina Abadier a, Haizheng Song b, Tangali S. Sudarshan b, Yoosuf N. Picard a, Marek Skowronski a,n a b

Department of Materials Science and Engineering, Carnegie Mellon University, Pittsburgh, PA 15213, USA Department of Electrical Engineering, University of South Carolina, Columbia, SC 29208, USA

art ic l e i nf o

a b s t r a c t

Article history: Received 25 November 2014 Received in revised form 26 January 2015 Accepted 3 February 2015 Communicated by T. Paskova Available online 9 February 2015

Transmission electron microscopy (TEM) and KOH etching were used to analyze the motion of dislocations after the conversion of basal plane dislocations (BPDs) to threading edge dislocations (TEDs) during 4H–SiC epitaxy. The locations of TED etch pits on the epilayer surface were shifted compared to the original locations of BPD etch pits on the substrate surface. The shift of the TED etch pits was mostly along the BPD line directions towards the up-step direction. For converted screw type BPDs, the conversion points were located below the substrate/epilayer interface. The shift distances in the step-flow direction were proportional to the depths of the BPD–TED conversion points below the substrate/epilayer interface. For converted mixed type BPDs, the conversion points were exactly at the interface. Through TEM analysis, it was concluded that the dislocation shift is caused by a combined effect of H2 etching prior to growth and glide of the threading segments during high temperature epitaxy. The TED glide is only possible for converted pure screw type BPDs and could present a viable means for eliminating BPDs from the epilayer during growth by moving the conversion point below the substrate/epilayer interface. & 2015 Elsevier B.V. All rights reserved.

Keywords: A1: Transmission electron microscopy A1: Basal plane dislocation A1: Dislocation glide A3: Chemical vapor deposition epitaxy B2: Silicon carbide

1. Introduction Silicon carbide (SiC) is a wide bandgap semiconductor that has many properties advantageous for high temperature and high power electronic device applications. It has high breakdown electric field, long carrier lifetimes and high thermal conductivity [1]. Despite these properties, the commercialization of SiC power electronic devices has been limited due to the presence of structural defects in SiC epilayers. A number of extended defects have been linked to degraded performance of SiC-based devices. Defects can lead to severe reduction in blocking voltages and minority carrier lifetimes, increase in forward resistance and orders of magnitude increase of leakage currents [2–6]. The long-term goal of SiC research is to eliminate such defects. The major issue for SiC bipolar devices (e.g. pin diodes) is recombination-induced stacking faults that are formed in the active layers of devices under forward bias. The stacking faults lead to degradation of the bipolar device performance over time by increasing the forward voltage drop and the on-state energy

n

Corresponding author. Tel.: þ 1 412 268 2710. E-mail address: [email protected] (M. Skowronski).

http://dx.doi.org/10.1016/j.jcrysgro.2015.02.004 0022-0248/& 2015 Elsevier B.V. All rights reserved.

losses [7,8]. The stacking faults are formed by the motion of one of the partial dislocations comprising a perfect basal plane dislocation (BPD). The activation energy for dislocation motion is provided by electron–hole recombination [6]. In order to avoid degradation of device performance, the propagation of BPDs from substrates to epilayers must be prevented. During epitaxial growth, the dislocations present in the substrate cannot terminate at the substrate/epilayer interface. The BPDs in the substrate could either propagate into the overlying epilayer or be converted to threading edge dislocations (TEDs) by changing their dislocation line directions. The majority of BPDs in the substrate spontaneously convert to the relatively electrically benign TEDs during epitaxial growth [9–11]. The percentage of converted BPDs continuously increases during growth and is significantly higher in epilayers grown on 41 off-cut substrates compared to 81 off-cut substrates [12]. In 41 off-axis epilayers, it was experimentally found that 100% of BPDs spontaneously converted to TEDs within
16 μm of epitaxial growth [9]. Therefore, it is typical to deposit a thin heavily doped (
1018 cm  3) n þ buffer layer prior to depositing the lightly doped (
1014 cm  3) n  drift layer during SiC device fabrication [13]. This is done to insure that BPD–TED conversion occurs within the highly doped buffer layer, where the electron–hole recombination rate is

8

M. Abadier et al. / Journal of Crystal Growth 418 (2015) 7–14

minimal compared to the drift layer. On the other hand, the lowdoped active region of the device is free of BPDs. The typical thickness of the n þ buffer layer is 5–20 μm. The BPDs in the buffer layer are not expected to form stacking faults due to the low carrier recombination rate. However, it was recently reported that even BPDs converted in the buffer layer (within 5–10 μm of the buffer/drift layer interface) could still form stacking faults at current injection below 1000 A/cm2 [14]. The deeper the conversion point in the buffer layer, the higher the current injection levels required to cause BPD faulting to a stacking fault. One possible solution is to grow thicker buffer layers in order to insure that conversion occurs at greater depths. However, this is not desirable, as it will increase the total duration of epitaxial growth and consequently the cost of devices. In this study, we report the glide of the TED segments towards the up-step direction after conversion of screw BPDs during epitaxial growth. The TED glide is energetically favorable for all converted screw BPDs, as it decreases the total dislocation length. TED glide moves the conversion point deeper in the epilayer even below the substrate/epilayer interface. Therefore, the TED glide could present a viable mechanism to avoid the formation of stacking faults at high current densities with a thin buffer layer or no buffer layer at all. The TED glide was observed while analyzing the BPD–TED conversion in epilayers grown on substrates etched with molten KOH related chemistry prior to growth. Defect-selective etching of SiC substrates with molten KOH related chemistry is an approach used to induce the conversion of all the substrate BPDs to occur at the substrate/epilayer interface [15,16]. Substrate etching leads to the formation of oval etch pits at locations where BPDs intersect the substrate surface. The BPD etch pits on the substrate surface have a different step morphology compared to areas away from etch pits. This leads to the epitaxial growth proceeding laterally at the bottom of etch pits, blocking the path for BPD propagation and forcing it to convert to a TED at the substrate/epilayer interface [17]. Typically molten KOH was used as the etchant; however most recently a milder eutectic mixture of KOH–NaOH–MgO is used instead in order to avoid deterioration of the epilayer surface morphology [18]. Song et al. [19] have shown that 99.9% of BPDs convert to TEDs in an epilayer grown on a 41 off-axis substrate etched in the eutectic mixture. Along with the BPD conversion, a very interesting phenomenon was also observed [18,19]. The locations of the TED etch pits on the epilayer surface were shifted relative to the original locations of the BPD etch pits on the substrate surface. The shift was mostly along the BPD line directions toward the up-step direction. On the other hand, the propagation of TEDs and threading screw dislocations (TSDs) from the substrate to the epilayer was not associated with any shift. The reason behind the shift in the locations of BPDs associated with their conversion to TEDs was not understood. In this study, KOH etching and transmission electron microscopy (TEM) are used to analyze the BPD–TED conversion and explain the nature of the dislocation shift. It was found that the shift was caused by a combined effect of the in situ H2 etching of the substrate prior to growth, as well as glide of the threading segment after conversion. The glide mechanism was confirmed by (gb) analysis and precise measurements of the conversion point depth.

etch pits (5–6 μm). The large etch pits are useful in mapping the locations of dislocations before and after epitaxial growth. The etched surfaces were then mapped by Nomarski optical microscopy (NOM) to record the locations of BPD etch pits. Chemical vapor deposition (CVD) was used to grow a
4 μm-thick 4H–SiC homoepitaxial layer at a temperature of 1600 1C and a pressure of 300 Torr. The precursors used were tetrafluorosilane (SiF4) and propane (C3H8), at C/Si ratio¼ 1.5. Hydrogen was used as the carrier gas at a flow rate of 9 standard liters per minute (slm). The thickness of the epilayer was confirmed by Fourier transform infrared spectroscopy (FTIR) and cross-sectional scanning electron microscopy (SEM). KOH etching was then performed in order to create etch pits at emergence points of dislocations at the epilayer surface. In order to analyze the dislocation conversion and shift behavior, a FEI NOVA-600 Nanolab focused ion beam (FIB) system was used to prepare cross-sectional TEM specimens at several locations of converted BPDs. TEM is then used to observe the BPD– TED conversion and perform (gb) dislocation analysis and explain the shift of dislocations towards the up-step direction. The TEM analysis was performed using a JEOL 2000EX TEM operating at 200 keV.

3. Results 3.1. NOM analysis Fig. 1(a) shows an optical micrograph of the substrate after KOH–NaOH–MgO etch but prior to epitaxial growth. Several dislocation etch pits appear on the surface including three oval BPD etch pits outlined by black ellipses. The BPD lines directions are along the long axes of the corresponding etch pits. Besides the oval BPD etch pits, four hexagonal etch pits are also visible; three

2. Experimental The substrates used were 8 mm  8 mm square pieces diced from a 41 off-axis n þ doped (0001) 4H–SiC wafer. The step-flow direction is defined as [112̄ 0]. The substrates were etched by KOH–NaOH–MgO for an extended period of time in order to create relatively larger

Fig. 1. Nomarski optical micrographs of (a) substrate after KOH–NaOH–MgO etching and (b) after epilayer growth and KOH etching. (c) Superposition of micrographs in (a) and (b).

M. Abadier et al. / Journal of Crystal Growth 418 (2015) 7–14

of which are due to TSDs and one (at the top most part of the figure) due to a TED. Fig. 1(b) shows an optical micrograph of the same location after epitaxial growth and repeated KOH etching. Oval shaped surface depressions appear on the epilayer surface at the locations of all the etch pits produced by etching of the substrate. All of the surface depressions are elongated along the step flow directions presumably due to interaction between the step flow and the surface corrugation. In addition to depressions, well-defined new etch pits mark the locations where the dislocations intersect the epilayer surface. It is clear from the inspection of Fig. 1(b) that the epilayer etch pits coincide with the bottom part of the surface depressions for TEDs and TSDs. Since TSDs typically propagate perpendicular to the epilayer surface, the bottom of the surface depression is expected to coincide with the substrate etch pit that produced it. The black ellipses in Fig. 1(b) contain three new etch pits corresponding to TEDs instead of BPD etch pits, which means that the BPDs in the substrate converted to TEDs in the epilayer. However, the etch pits of the three converted BPDs are displaced relative to the surface depressions marking the intersection of the BPD with the substrate surface. The shift direction is along the BPD line directions towards the up-step direction. The shift distances of converted BPDs could be estimated by measuring the distance between the etch pits and the bottom part of the corresponding surface depressions on the epilayer surface. Fig. 1(c) shows a superposition of the two micrographs in (a) and (b). The etch pits of two TSDs (outlined by white circles) are used to align the two micrographs on top of each other. For the converted BPDs, the depressions on the epilayer surface coincide with the locations of the BPD etch pits on the substrate. This implies that the observed shift could be due to dislocation motion towards the up-step direction after conversion. 3.2. TEM analysis In order to analyze the BPD–TED conversion, a total of three cross-sectional TEM specimens were prepared at the locations of BPDs that converted to TEDs in the epilayer. Fig. 2(a) and (b) shows plan view optical micrographs of the etched surfaces of the substrate and epilayer, respectively. The location of the first TEM specimen A is marked by the black rectangle in Fig. 2(b). The BPD line direction is along the [112̄ 0] step-flow direction and the shift distance is about 20 μm towards the up-step direction. The plane normal of the TEM specimen is along the [11̄00] direction. A

Fig. 2. Nomarski optical micrograph of the (a) substrate after KOH–NaOH–MgO etching and (b) the epilayer after KOH etching. Locations of TEM specimens A and C are marked in (b).

9

Fig. 3. Bright-field TEM micrographs of a BPD converting to a TED in Specimen A at ̄ (a) [11̄00] zone-axis, (b) g112̄ 0 two-beam, (c) g0004̄ two-beam and (d) g1108 twobeam orientations.

second TEM specimen B (not shown in figure) was prepared at a converted BPD with a similar configuration to that of the A specimen. The shift distance in this case is
13 μm. The white rectangle in Fig. 2(b) shows the location and orientation of the third TEM specimen C. The BPD line direction is inclined at 641 relative to the [112̄ 0] step-flow direction. The plane normal of the TEM specimen is also inclined at 641 relative to the [11̄00]direction (i.e., inclined at 41 relative to the [101̄0] direction). The TED etch pit is shifted by
53 μm along the BPD line direction or by 24 μm along the up-step direction. The figure also shows two other BPDs in the substrate that converted to TEDs in the epilayer. Fig. 3(a) is a bright-field TEM micrograph of the specimen A at the [11̄00] zone-axis orientation. The black contrast at the top of the image is the protective Pt layer deposited on the epilayer surface during the preparation of the TEM specimen. The horizontal white line represents the epilayer surface. The BPD appears to have converted to a TED at 4.28 μm below the epilayer surface and the converted TED propagates up to the etch pit on the epilayer surface. The epilayer thickness was precisely measured by cross-sectional SEM to be 4.04 7 0.07 μm. This implies that the BPD–TED conversion point is actually below the substrate/epilayer interface. The depression on the epilayer surface (outside the field of view of the image) is further to the right of the etch pit. It is interesting to note that the BPD in the substrate is actually located under the TED etch pit and is not related to the location of the surface depression (which marks the intersection of the BPD with the substrate surface prior to epitaxial growth). The TED is also inclined relative to the [0001] direction towards the down-step direction. The down-step inclination of the TED is contrary to what would be expected based on the up-step shift of the etch pit locations after epitaxial growth. The inclination angle of the threading segment changes along the dislocation length. The specific inclination angle is denoted next to different segments of the TED line in Fig. 3(a).

10

M. Abadier et al. / Journal of Crystal Growth 418 (2015) 7–14

TED is also inclined towards the down-step direction. The Burgers ̄ 0] ̄ direction. vector was estimated by (gb) analysis to be along [211 The Burgers vector is at 601 relative to the BPD line direction. This indicates that this BPD is a mixed type dislocation. Fig. 4(b) is a bright-field TEM micrograph of the specimen C at the ̄ zone-axis orientation. The step-flow direction is towards the [1010] right side of the figure. Two dislocation line segments were captured and observed in the specimen. The dislocation extending downward from the etch pit is the TED, whereas the dislocation in the bottom left corner is the BPD. The exact location where the BPD converts to TED was milled away during FIB thinning of the specimen. By extrapolating the dislocation lines and based on the dislocation line profiles in the 2 previous specimens, the BPD–TED conversion is expected to have happened 4.32–4.41 μm below the epilayer surface. Similarly to the specimen A, the conversion point is located below the substrate/epilayer interface. By performing (gb) analysis, the ̄ direction (i.e., the Burgers vector is estimated to be along the [1̄210] BPD is a screw dislocation). Based on the analysis of the three TEM specimens, Fig. 4(c) shows a plot of BPD–TED conversion depth versus shift distance along the [112̄ 0] step-flow direction. The plot shows almost a linear relationship between conversion depth and shift distance along step-flow direction. The error bars for the conversion depth were estimated based on the magnification calibration of the TEM. The error is significantly higher for specimen C, since the conversion point was milled away during thinning of the TEM specimen. For the shift distances, the error bars were estimated based on the resolution of the optical micrographs of the substrate and the epilayer, as well as the error in measuring the distance using an image processing software.

4. Discussion

Fig. 4. Bright-field TEM micrographs of (a) specimen B and (b) specimen C at the [11̄00] and [101̄0] zone-axis orientations, respectively. (c) Plot of conversion depth versus shift distance along the step-flow direction for the three converted BPDs analyzed by TEM.

The Burgers vector of the dislocation was determined through (gb) analysis. Fig. 3(b)–(d) shows bright-field TEM micrographs ̄ ̄ two-beam conditions, respectively. The at g112̄ 0, g0004̄ and g1108 ̄ ̄ two-beam conditions. dislocation is invisible at g0004̄ and g1108 Consequently, the Burgers vector has to be along the [112̄ 0] direction making the BPD a screw dislocation. Fig. 4(a) is a bright-field TEM micrograph of the TEM specimen B ̄ at [1100] zone-axis orientation. The BPD appears to have converted to a TED 4.08 μm below the epilayer surface (i.e., at the substrate/ epilayer interface). BPD–TED conversion at the interface is expected due to the KOH–NaOH–MgO etching of the substrate prior to epitaxial growth. Similarly to specimen A, the BPD in the substrate is located under the etch pit rather than the surface depression. The

A total of three cases of BPD–TED conversion with different shift distances and BPD line directions were analyzed by the TEM. In all cases, the BPDs in the substrate were located under the TED etch pits and not under the epilayer surface depressions. This is surprising especially after Fig. 1(c) confirmed that the epilayer surface depressions coincide with the locations of the BPD etch pits on the substrate. Another striking observation was the location of the conversion point. In Fig. 4(a), the BPD–TED conversion occurred at the substrate/epilayer interface, which agrees with the KOH–NaOH– MgO substrate etching mechanism. However for the dislocation shown in Fig. 3(a), the conversion point is below the substrate/ epilayer interface. The conversion of the BPD appears to have happened in the substrate, something that has not been reported before and is inconsistent with the optical micrograph (Fig. 1(a)) showing BPD etch pits visible after substrate etching. In all three cases, the TEDs in the epilayer were inclined towards the downstep direction. A direct proportionality was also noticed between the depth of the BPD–TED conversion point below the substrate/ epilayer interface and the distance by which the etch pit locations were shifted for a converting BPD, as shown in Fig. 4(c). 4.1. Effect of in situ H2 etching One possible mechanism that can explain the above observations is that a certain thickness of the substrate is etched away during the in situ H2 etching prior to epitaxial growth. H2 etching of 4H–SiC substrates is typically performed prior to growth in order to create atomically flat surfaces and remove surface scratches after chemical mechanical polishing [20]. The H2 etching rate depends on the temperature and the gas flow rate. The sample used in this study was etched at 1600 1C for

M. Abadier et al. / Journal of Crystal Growth 418 (2015) 7–14

Fig. 5. Illustration showing the progression of a BPD in an off-axis 4H–SiC substrate (a) before etching, (b) after defect selective etching by KOH–NaOH–MgO, (c) after in situ H2 etching and (d) after epilayer growth and KOH etching. The shift distance between the surface depression and the TED etch pit in (d) depends on the amount of material removed by H2 etching in (c).

5 min and the H2 flow rate was 9 slm. This corresponds to a thickness of 0.9 μm etched away from the 4H–SiC substrate, as calculated from a separate experiment where only in situ H2 etching was performed. Fig. 5 is an illustration that shows how H2 etching can lead to a shift in the location of dislocation etch pits before and after growth. A BPD in the substrate intersects the substrate surface at an angle of 41 (off-cut angle), as shown in Fig. 5(a). Upon etching the substrate with KOH–NaOH–MgO, an oval etch pit is formed at the location where the BPD intersects the substrate surface (Fig. 5 (b)). During the substrate surface preparation by in situ H2 etching, a certain thickness of the substrate is etched away. The dashed horizontal line in Fig. 5(c) represents the substrate surface prior to H2 etching, whereas the solid line represents the surface after H2 etching. The etching of an off-cut substrate by H2 obviously leads to a change in the location where the BPD intersects the substrate surface (Fig. 5(c)) prior to epitaxial growth. After H2 etching, a depression is formed on the surface of the substrate at the original location of the BPD etch pit. At the initial stage of epitaxial growth, the BPD converts to a TED at the substrate/epilayer interface (solid line), as shown in Fig. 5(d). The TED produces an etch pit on the epilayer surface after KOH etching. On the other hand, the depression at the epilayer surface is formed at the location of the BPD etch pit on the substrate surface. This was confirmed by the superposition of Nomarski optical micrographs of the sample surface before and after growth, as shown in Fig. 1(c). The distance between the TED etch pit and the

11

surface depression depends on the amount of material removed by H2 etching. For the sample used in this study, H2 etching removed 0.9 μm from the substrate. The corresponding shift distance along the step-flow direction is 0.9 μm/tan(41)¼ 12.9 μm. The H2 etching mechanism can explain why the shift in the locations of etch pits is only observed for converted BPDs and is not observed for TEDs and TSDs, which generally propagate along the [0001] direction. H2 etching could also explain the change in the locations where BPDs intersect the substrate surface and convert to TEDs. Moreover, H2 etching could explain how the etch pits were shifted towards the up-step direction after growth, although the TEDs in all three TEM specimens were inclined towards the down-step direction. However, several experimental observations suggest that H2 etching is not the only mechanism responsible for the etch pit shifts. The shift distances were measured for 32 converted BPDs. Fig. 6 shows a histogram of the count of converted BPDs versus their shift distances along step-flow direction. A wide range of shift distances (8–19 μm) was observed, even for dislocations that are located very near to each other on the wafer, and had presumably the same H2 etching rate. Based on the H2 etching mechanism, the shift distances for all BPDs should be 12.9 μm or less. The red dashed line and arrow in Fig. 6 show the range of shift distances that could be explained by the H2 etching mechanism. Clearly, the H2 etching mechanism alone cannot explain the wide range of shift distances as well as shift distances in excess of 12.9 μm. This implies the existence of an additional mechanism. For TEM specimens A and C, the conversion point was below the substrate/ epilayer interface. However, the conversion point was exactly at the interface for specimen B. It is important to note that for the dislocation in specimen B, the shift distance was
13 μm (i.e., consistent with H2 etching mechanism) and the BPD was a 601 dislocation (i.e., not a screw dislocation). On the contrary, the shift distances for dislocations in specimens A and C were
20 and
24 μm, respectively. Both values are significantly larger than the 12.9-μm value expected from the H2 etching mechanism. Moreover, the BPDs in both specimens were screw dislocations. These results suggest a possible relationship among the Burgers vector of the BPD, the location of the conversion point with respect to the interface and the observed shift distance of the TED etch pit relative to the depression at the epilayer surface. 4.2. Effect of TED glide Fig. 4(c) shows a linear relation between the conversion depth and the shift distance along step-flow direction. This linear relation is similar to what was observed previously by Zhang

Fig. 6. Histogram of number of dislocations versus shift distance. The red dashed line and arrow mark shift distances that could be explained by H2 etching. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

12

M. Abadier et al. / Journal of Crystal Growth 418 (2015) 7–14

et al. [21] during high temperature annealing of 4H–SiC epilayers. These authors used annealing at 1800–2000 1C in order to convert the BPDs that propagated to the epilayer surface to TEDs. During annealing, the BPDs convert to TEDs under the effect of the image force between the BPD and the epilayer surface [10]. The conversion starts at the epilayer surface, and then the TED glides along the up-step direction. The glide of the threading segment leads to a shift in the location where it intersects the epilayer surface and the conversion point is moved deeper in the epilayer. The larger the glide distance, the deeper the conversion point becomes. TED glide is energetically favorable since it leads to decrease the total length of the dislocation. The TED glide mechanism observed by Zhang et al. [21] during annealing could also occur during the epitaxial growth. The upstep glide of TEDs after BPD–TED conversion could explain how the conversion point could move below the substrate/epilayer interface as well as shift distances in excess of the 12.9 μm allowed by H2 etching. In order to confirm if this is the case, a second epilayer sample was grown where the effect of H2 etching was avoided. It is wellknown that adding C3H8 to H2 lowers the etching rate [22]. For that epilayer, substrate surface preparation was performed by a reduced H2 flow rate combined with a certain flow rate of C3H8 prior to growth. This surface preparation condition was first tested in an only H2 etching experiment (i.e., no epilayer growth). The sample was weighed before and after surface preparation by an analytical semi-/micro-balance. There was no detectable change in the sample weight, which implied that there is no H2 etching effect at this surface preparation condition. Similar to the first epilayer sample, the substrate of the second sample was etched by KOH–NaOH–MgO before epitaxial growth. The epilayer was also etched by KOH after growth and etch pits were mapped on the substrate and epilayer surface. Optical micrographs obtained before and after epitaxial growth were compared. The epilayer thickness was confirmed by cross-sectional SEM to be 4.48 70.07 μm. If H2 etching is the only mechanism responsible for the shift of etch pits, no shift should be observed in this sample. Fig. 7(a)–(c) shows superposition of optical micrographs of the substrate and epilayer surfaces at three different locations. The substrate and epilayer images were aligned by TSDs (not shown in the figure). In these micrographs, all of the BPD etch pits are located on the substrate surface and all of the TED etch pits are on the epilayer surface. In Fig. 7(a), the BPD etch pit on the substrate surface coincides with the TED etch pit on the epilayer surface. The BPD converted to a TED without any shift. In Fig. 7(b), a BPD ̄ ̄20] upconverted to a TED with a shift of 9.5 μm towards the [11 step direction. It is important to note that the shift of the etch pits is strictly along the up-step direction, even though the BPD line direction was inclined at
201 relative to the step-flow direction. This is different than what was observed in the first epilayer sample (e.g. Fig. 1(c)) where the shift was along the BPD line directions towards the up-step direction.

Fig. 7(c) shows the conversion of three BPDs to TEDs associated with a shift at 601 relative to the step-flow direction. Similar to the dislocations in Fig. 7(b), the shift is strictly along a crystallographic direction ([2̄110]). In total,
80% of substrate BPDs in the sample converted with no shift, whereas
20% of BPDs converted with a shift toward a 〈112̄ 0〉 direction. The shift associated with BPD conversion in Fig. 7(b) and (c) confirms that H2 etching is not the only mechanism responsible for the shift of etch pits. The shifts were along specific 〈112̄ 0〉 directions towards the up-step direction. This suggests that TED glide could be the other mechanism responsible for the observed shift of etch pits. Glide of dislocations is only possible in the planes containing the dislocations Burgers vectors and line directions. The Burgers vectors of BPDs in 4H–SiC substrates are along 〈112̄ 0〉 directions [23]. In order to confirm the TED glide mechanism, TEM specimens were prepared at the locations of the two converted BPDs shown in Fig. 7(a) and (b). The plane normals of the specimens were along the [11̄00] direction. Fig. 8(a) and (b) shows bright-field TEM micrographs at the [11̄00] zone-axis orientation of the converted BPDs shown in Fig. 7(a) and (b), respectively. In Fig. 8(a), the BPD– TED conversion point is 4.45 μm below the epilayer surface. This length scale corresponds to the actual epilayer thickness (4.48 7 0.07 μm). Therefore, the conversion point is at the substrate/epilayer interface. The conversion point in Fig. 8(b), on the other hand, is located 5.18 μm below the epilayer surface (or 0.7 μm below the substrate/epilayer interface). By performing (gb) analysis, the dislocation Burgers vector was determined to be along the [112̄ 0] direction. Therefore, the TED could glide in the (11̄00) plane towards the up-step direction driven by the decrease in the total dislocation length. Based on the TED glide mechanism, the shift distance and the conversion depth should be linearly related to each other by the tangent of the off-cut angle (i.e., tan(41)). For a shift distance of 9.5 μm, the conversion point should be (9.5 μm  tan(41) ¼ 0.66 μm) below the substrate/epilayer interface. By assuming an

̄ Fig. 8. Bright-field TEM micrographs at the [1100] zone-axis orientation for a BPD that converted to a TED (a) with no shift (in Fig. 7(a)) and (b) with a shift towards the up-step direction (in Fig. 7(b)). The conversion point in (b) is below the substrate/epilayer interface.

Fig. 7. Superposition of optical micrographs of the substrate and epilayer at three different locations in sample 2 showing BPD–TED conversion with (a) no shift of etch pits, ̄ ̄20] direction and (c) shift towards the [2̄110] direction. All of the BPD etch pits are located on the substrate surface and all of the TED etch pits are on (b) shift towards the [11 the epilayer surface.

M. Abadier et al. / Journal of Crystal Growth 418 (2015) 7–14

Fig. 9. Schematic illustrations of a BPD in the (0001) basal plane progressing in steps (a–d) by the TED glide process. Step-flow direction is upwards. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)

epilayer thickness of 4.45 μm (based on Fig. 8(a)), the conversion point in Fig. 8(b) should be (4.5 μm þ0.66 μm ¼5.11 μm). This value is very close to the experimentally observed value of 5.18 μm. This confirms that the TED glide mechanism is responsible for the shift in dislocation etch pits observed in this sample, where the effect of H2 etching was avoided. The shift of etch pits in the first epilayer sample is caused by a combined effect of H2 etching as well as glide of TEDs after BPD– TED conversion. The H2 etching mechanism explains the shift of etch pits along the BPD line directions towards the up-step direction. The TED glide mechanism explains the large shift distances (13–19 μm), as well as BPD–TED conversion points being located below the substrate/epilayer interface. 4.3. Model of TED glide TED glide is only possible in specific planes and along specific crystallographic directions. However, as shown in Fig. 8(b), the BPD converted with a shift even though its line directions were not along a specific crystallographic direction. Fig. 9 shows a series of schematic diagrams showing the steps for TED glide for a converted BPD whose line direction was at a random orientation in the basal plane of the substrate. The plane of Fig. 9 is the (0001) basal plane. The horizontal line represents the substrate/epilayer interface and the step-flow direction is upwards. The vertical line represents projection of the (11̄00) prismatic plane perpendicular to the (0001) basal plane. Fig. 9(a) shows a BPD inclined at an angle (θ) relative to the stepflow direction. Let us assume that the Burgers vector of the BPD is along the [112̄ 0] step-flow direction. BPDs in 4H–SiC are known to dissociate into two partial dislocations with a narrow (30–70 nm) stacking fault in between [24]. The fault is marked by a red shaded area between the lines of the two partial dislocations. At the substrate surface, the partial dislocations constricted and the BPD converted to a TED (Fig. 9(a)). First, the BPD glides in the basal plane and its line direction aligns with the [112̄ 0] direction, as shown in Fig. 9(b). The partial dislocations can then constrict and the constricted BPD becomes

13

locally a screw dislocation, as shown in Fig. 9(c). As a consequence, the TED is pulled by its line tension and glides towards the up-step direction, as shown in Fig. 9(d). This process of BPD glide, constriction and TED glide can occur several times for short segments of the BPD during growth. This will eventually lead to a significant shift in the location where the dislocation intersects the epilayer surface. Based on Fig. 9(d), TED glide decreases the dislocation length from (L) to ((L  sin(θ))þ (L  cos(θ)  sin(41))). Assuming L ¼9.5 μm and θ ¼201, the dislocation length decreases to E3.9 μm. This confirms that TED glide significantly decreases dislocation length and energy, even for BPDs whose line directions are not exactly along their Burgers vectors. It is also important to note that there is a consistent difference in ̄ the TED line direction in the (1100) prismatic plane for the case of BPD conversion with TED glide as compared to no glide. In the case of TED glide (Figs. 3(a), 4 and 8(b)), the TED is typically a curved line and is inclined at a large angle (24–351) relative to the [0001] direction. For the case of no TED glide (Figs. 4 and 8(a)), the TED is typically a straight line and is inclined at a small angle (7–151) relative to the [0001] direction. Moreover, if the BPD is not a screw dislocation then the BPD–TED conversion point is exactly at the substrate/epilayer interface. This is consistent with Chung et al. [11] who observed BPD–TED conversion of a 601 BPD at the substrate/ epilayer interface. In their paper, the TED was also a straight line and was inclined at 191 relative to the [0001] direction. Therefore, the curvature and large inclination angles (in Figs. 3(a), 4 and 8(b)) could be an indication that the TED is being pulled by its line tension to decrease the total dislocation length. Based on our observations, it is clear that TED glide occurs at typical epitaxial growth temperatures (
1600 1C). TED glide occurs for all screw BPDs that converted to TEDs and moves the BPD–TED conversion point deeper below the substrate/epilayer interface. It is well-known that all BPDs propagating into 4H–SiC epilayers are screw dislocations [25]. In 41 off-axis epilayers, all these propagating BPDs spontaneously convert to TEDs within
16 μm of epitaxial growth [9]. The TED glide discussed here could be used as a mechanism to move the BPD–TED conversion point deeper in the epilayer or even below the substrate/epilayer interface. This will decrease the density of BPDs present within the epilayers and help to avoid degradation of device performance. Generally, the glide of dislocations requires high temperatures. For instance, the partial dislocations in the basal plane are only mobile above 1100 1C [26]. The epilayers used in this study were only 4–5 μm thick and were grown at a high growth rate (20– 25 μm/h). Therefore, the duration at which the epilayer was subjected to a high temperature is much shorter than a typical 50–100 μm thick epilayers used in high power devices. For instance, the samples analyzed in this study were subjected to a high temperature (i.e., above 1500 1C) for only
17 min including heating and cooling stages in the epitaxial growth process. The extent of TED glide and the motion of the BPD–TED conversion point deeper in the epilayer could be higher in a thicker epilayer compared to what is observed in this study. It is worthwhile to note that this is not the first time that dislocation glide was generally reported to occur during 4H–SiC homoepitaxial growth. BPDs were previously reported to glide in the basal planes, leading to the formation of dislocation half-loop arrays and interfacial dislocations [27]. However, this study is the first to report the glide of TEDs in prismatic planes at typical epitaxial growth temperatures, and the associated motion of the BPD–TED conversion point during epitaxial growth. Future work could focus on a better understanding of the effect of several growth parameters on the TED glide process. Maximizing the TED glide could be a possible way to remove any BPDs in the epilayer, even if they are not induced to convert to TEDs at the substrate/ epilayer interface during growth.

14

M. Abadier et al. / Journal of Crystal Growth 418 (2015) 7–14

5. Conclusions In conclusion, we have analyzed the motion of the threading dislocation segment of a converted basal plane dislocation. The shift of the associated etch pits is caused by a combined effect of H2 etching prior to growth and glide of TEDs after conversion during epitaxy. H2 etching causes the shift of all BPDs along their line directions towards the up-step direction. TED glide only occurs for converted screw BPDs. The shift due to TED glide is always along the dislocation Burgers vector directions (i.e., 〈112̄ 0〉 directions) towards the up-step direction. TED glide causes the conversion point to move deeper below the substrate/epilayer interface and decreases the total dislocation energy by decreasing its length. TED glide occurs at typical epitaxial growth temperatures (
1600 1C) and is relevant to all BPDs that propagate in 4H– SiC epilayers and convert to TEDs within the epilayer, since all these BPDs are screw dislocations. Acknowledgments This work was supported by ONR Grant nos. N00014.10.10532 (CMU) and N00014.10.10530 (USC) (Scott Coombe, Program Manager). M.A. acknowledges support of John and Claire Bertucci through the Bertucci Graduate Fellowship Award. References [1] Y. Goldberg, M. Levinshtein, S. Rumyantsev, Silicon carbide, in: M.E. Levinshtein, S. Rumyantsev, M.S. Shur (Eds.), Properties of Advanced Semiconductor Materials: GaN, AIN, InN, BN, SiC, SiGe., John Wiley & Sons, 2001, pp. 93–147 , http://www.wiley.com/WileyCDA/WileyTitle/productCd-0471358274.html. [2] R.A. Berechman, M. Skowronski, Q. Zhang, Electrical and structural investigation of triangular defects in 4H–SiC junction barrier Schottky devices, J. Appl. Phys. 105 (2009) 074513. [3] H. Fujiwara, T. Kimoto, T. Tojo, H. Matsunami, Characterization of in-grown stacking faults in 4H–SiC (0001) epitaxial layers and its impacts on highvoltage Schottky barrier diodes, Appl. Phys. Lett. 87 (2005) 051912. [4] P. Neudeck, J. Powell, Performance limiting micropipe defects in silicon carbide wafers, Electron Device Lett. IEEE 15 (1994) 63–65. [5] P.G. Neudeck, Electrical impact of SiC structural crystal defects on high electric field devices, Mater. Sci. Forum 338–342 (2000) 1161–1166. [6] M. Skowronski, S. Ha, Degradation of hexagonal silicon-carbide-based bipolar devices, J. Appl. Phys. 99 (2006) 011101. [7] J.P. Bergman, H. Lendenmann, P.Å. Nilsson, U. Lindefelt, P. Skytt, Crystal defects as source of anomalous forward voltage increase of 4H–SiC diodes, Mater. Sci. Forum 353–356 (2001) 299–302.

[8] H. Lendenmann, F. Dahlquist, N. Johansson, R. Söderholm, P.Å. Nilsson, J.P. Bergman, et al., Long term operation of 4.5 kV PiN and 2.5 kV JBS diodes, Mater. Sci. Forum 353–356 (2001) 727–730. [9] R.L. Myers-Ward, B.L. VanMil, R.E. Stahlbush, S.L. Katz, J.M. McCrate, S.A. Kitt, et al., Turning of basal plane dislocations during epitaxial growth on 41 off-axis 4H–SiC, Mater. Sci. Forum 615–617 (2009) 105–108. [10] S. Ha, P. Mieszkowski, M. Skowronski, L.B. Rowland, Dislocation conversion in 4H silicon carbide epitaxy, J. Cryst. Growth 244 (2002) 257–266. [11] S. Chung, V. Wheeler, R. Myers-Ward, C.R. Eddy, D.K. Gaskill, P. Wu, et al., Direct observation of basal-plane to threading-edge dislocation conversion in 4H–SiC epitaxy, J. Appl. Phys. 109 (2011) 094906. [12] R. Myers-Ward, D. Gaskill, Managing basal plane dislocations in SiC: perspective and prospects, ECS Trans. 50 (2012) 103–108. [13] R.L. Myers-Ward, D.K. Gaskill, B.L. VanMil, R.E. Stahlbush, C.R. Eddy Jr., Reduction of basal plane dislocations in epitaxial SiC (2011).Pub. No.: US 2011/0045281 A1, Pub. Date: Feb. 24, 2011. [14] N. a. Mahadik, R.E. Stahlbush, M.G. Ancona, E.A. Imhoff, K.D. Hobart, R.L. Myers-Ward, et al., Observation of stacking faults from basal plane dislocations in highly doped 4H–SiC epilayers, Appl. Phys. Lett. 100 (2012) 042102. [15] J.J. Sumakeris, Method to reduce stacking fault nucleation sites and reduce forward voltage drift in bipolar devices; 2006. [16] Z. Zhang, T.S. Sudarshan, Basal plane dislocation-free epitaxy of silicon carbide, Appl. Phys. Lett. 87 (2005) 151913. [17] Z. Zhang, E. Moulton, T.S. Sudarshan, Mechanism of eliminating basal plane dislocations in SiC thin films by epitaxy on an etched substrate, Appl. Phys. Lett. 89 (2006) 081910. [18] H. Song, T.S. Sudarshan, Basal plane dislocation mitigation in SiC epitaxial growth by nondestructive substrate treatment, Cryst. Growth Des. 12 (2012) 1703–1707. [19] H. Song, T.S. Sudarshan, Basal plane dislocation conversion near the epilayer/ substrate interface in epitaxial growth of 41 off-axis 4H–SiC, J. Cryst. Growth 371 (2013) 94–101. [20] V. Ramachandran, M.F. Brady, A.R. Smith, R.M. Feenstra, D.W. Greve, Preparation of atomically flat surfaces on silicon carbide using hydrogen etching, J. Electron. Mater. 27 (1998) 308–312. [21] X. Zhang, H. Tsuchida, Conversion of basal plane dislocations to threading edge dislocations in 4H–SiC epilayers by high temperature annealing, J. Appl. Phys. 111 (2012) 123512. [22] B.L. VanMil, K.-K. Lew, R.L. Myers-Ward, R.T. Holm, D.K. Gaskill, C.R. Eddy, et al., Etch rates near hot-wall CVD growth temperature for Si-face 4H–SiC using H2 and C3H8, J. Cryst. Growth 311 (2009) 238–243. [23] W.M. Vetter, M. Dudley, Partial dislocations in the X-ray topography of asgrown hexagonal silicon carbide crystals, Mater. Sci. Eng. B. 87 (2001) 173–177. [24] M.H. Hong, A.V. Samant, P. Pirouz, Stacking fault energy of 6H–SiC and 4H–SiC single crystals, Philos. Mag. A 80 (2000) 919–935. [25] H. Tsuchida, M. Ito, I. Kamata, M. Nagano, Fast epitaxial growth of 4H–SiC and analysis of defect transfer, Mater. Sci. Forum 615–617 (2009) 67–72. [26] A.V. Samant, M.H. Hong, P. Pirouz, The relationship between activation parameters and dislocation glide in 4H–SiC single crystals, Phys. Status Solidi 222 (2000) 75–93. [27] X. Zhang, M. Skowronski, K.X. Liu, R.E. Stahlbush, J.J. Sumakeris, M.J. Paisley, et al., Glide and multiplication of basal plane dislocations during 4H–SiC homoepitaxy, J. Appl. Phys. 102 (2007) 093520.