APPLIED PHYSICS LETTERS 86, 021905 (2005)

Structure of the carrot defect in 4H-SiC epitaxial layers M. Benamara,a) X. Zhang, and M. Skowronskib) Department of Materials Science and Engineering, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, Pennsylvania 15213

P. Ruterana and G. Nouet SIFCOM, UMR 6176 CNRS-ENSICAEN, 6, Bd du Maréchal Juin, 14050 Caen, France

J. J. Sumakeris, M. J. Paisley, and M. J. O’Loughlin Cree Inc., 4600 Silicon Drive, Durham, North Carolina 27703

(Received 20 August 2004; accepted 3 November 2004; published online 4 January 2005) Transmission electron microscopy and KOH etching were used to determine the structure of the carrot defect in 4H-SiC epilayers. The defect consists of two intersecting planar faults on prismatic ¯ 00其 and basal {0001} planes. Both faults are connected by a stair-rod dislocation with Burgers 兵11 ¯ 0兴 with n ⬎ 3 at the crossover. A Frank-partial dislocation with b = 1 / 12关44 ¯ 03兴 vector 1 / n 关101 terminates the basal fault. © 2005 American Institute of Physics. [DOI: 10.1063/1.1849416] Silicon carbide offers an advanced semiconductor technology making it a potential candidate to replace conventional (Si, GaAs) materials for high power and high frequency electronic applications.1 One of the remaining challenges facing the development of large-area SiC devices is the elimination of morphological defects in epitaxial layers. Several types of such defects have been reported and are referred to as micropipes,2 carrots,3,4 comets,2 shallow pits,3,4 and triangular defects.3 Many of them are either known or are suspected to affect device performance.3–5 Growth parameters such as the substrate orientation, the Si/ C flux ratio, or the growth rate were found to affect the formation and/or appearance of defect morphologies. A number of authors have demonstrated using transmission electron microscopy (TEM) techniques that these defects are fingerprints of extended volume defects intersecting the surface.6,7 Detailed analysis of their structure allowed in some instances one to understand their formation mechanism. As an example, the triangular defects are due to cubic 3C-polytype inclusions in the hexagonal epilayers.7 Similarly, this letter is focused on the identification of extended defects under the carrot defect3,8,9 using cross-sectional TEM (XTEM). Currently available structural information about carrot defect includes plan-view TEM as well as x-ray topography data. Okada et al. observed several faults in the basal plane located at the carrot tip and interpreted them as Shockleytype faults.8 Their observations were limited to the nearsurface region and did not allow imaging of the crystallographic defects underneath the carrot. Our recent x-ray topography study9 reported evidence of a planar defect formation in the prismatic plane and identified a screw dislocation as the origin of the carrot. However, detailed information on the defect microstructure could not be extracted due to the limited resolution of the technique. It is thus the purpose of this letter to report the identification of stacking faults, in the prismatic and basal planes under the carrot defect, using XTEM. Both faults are connected by a stair-rod dislocation at the crossover and a Frank-partial dislocation a)

Electronic mail: [email protected] Author to whom correspondence should be addressed; electronic mail: [email protected]

b)

terminates the basal fault. The overall geometry of the defect is similar to those previously studied by Blank et al.10 in ZnS, Drum11 in AlN, and Vermault et al.12 in III-nitrides ¯ 10其 prismatic stacking epilayers. These authors analyzed 兵12 faults folding to the basal plane with displacement R = I1 or ¯ 01兴 and attributed their formation to growth doR = 1 / 2关11 mains or interstitial precipitation. The 100-␮m-thick 4H-SiC epilayers were grown by chemical vapor deposition on Si-face, n-type 4H-SiC substrates. The substrates were 8° off-cut from [0001] in the ¯ 0兴 direction. The TEM experiments were conducted on a 关112 Philips Tecnai F20 and a Jeol 2000 EX transmission electron microscopes both operating at 200 kV, using conventional ¯ 0兴 zone axis. The specimens containing imaging along 关112 one carrot defect each were first etched in KOH at 500 ° C for 5 – 20 min. This step produced long continuous grooves corresponding to the intersections of prismatic planar defects with the layer surface.9 The XTEM samples were then cut from the part of the carrot where the defect intersected the surface, thinned down by grinding, mechanically dimpled, and ion milled to electron transparency. During the entire preparation procedure, we made use of the etched v-groove to ensure the presence of the defect on the TEM sample and to unambiguously locate its position during the experiments.13 Figure 1 is a dark-field TEM image taken with g ¯ ¯ 0兴 zone axis at the location directly = 1100 close to the 关112 underneath the etched groove. A vertical defect extending downward from the epilayer surface (black area in the upper portion of the figure) is clearly visible. In this projection the defect line generally follows the [0001] direction. Tilting the ¯ 0兴 axis toward either sample a few degrees off the 关112 ¯ 00兴 or 关1 ¯ 100兴 made the defect appear wider with inclina关11 tion. This characteristic demonstrates that the defect is planar ¯ 00其 priswith an average habit plane being the first-order 兵11 ¯ 0兴 matic plane. The dark-field images obtained along the 关112 zone axis with a small objective aperture centered on either ¯ reflections produced asymmetric contrast 0002 or 0002

0003-6951/2005/86(2)/021905/3/$22.50 86, 021905-1 © 2005 American Institute of Physics Downloaded 06 Jan 2005 to 128.2.105.86. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp

021905-2

Benamara et al.

Appl. Phys. Lett. 86, 021905 (2005)

FIG. 3. Bright-field images of the prismatic fault termination, (a) under multibeam conditions, the fault folds to the basal plane; and (b) with g ¯¯14, a stair-rod dislocation is visible at the crossover. = 21

¯ 00 FIG. 1. Transmission electron microscopy dark-field image with g = 11 ¯ 0兴 axis (parallel to the off-cut direction). Inset is a high magclose to 关112 nification image showing the zigzag morphology of the defect.

(white-black or black-white) of the defect as expected for a stacking fault.14,15 At higher magnification, the defect presented a more complex structure (see the inset in Fig. 1). Specifically, the fault exhibited a peculiar zigzag morphology with planar ¯ 00其 plane. Some segments inclined by about 28° from the 兵11 of these segments appeared edge-on when viewed along ¯ 0兴, allowing identification of one set of habit planes to 关112 ¯ 02其 planes. Other segments be the first-order pyramidal 兵11 appeared wide and tilting experiments did not allow one to determine their habit planes. The ability of such faults to ¯ 00其 plane was confirmed by optical bend away from the 兵11 microscopy. Figure 2 shows a typical high magnification plan view image of the as-grown epilayer surface with a horizontal segment of a different carrot ridge.9 The zigzag contrast corresponds to the groove formed at the intersection of the fault and the surface. It is apparent that such faults are ¯ 00其 plane all along the carrot not always confined to the 兵11 axis, although most grooves appeared perfectly straight as is the case for the fault analyzed in this report.13 The prismatic stacking fault ended about 90 ␮m from the epilayer surface, where it intersected a basal plane stacking fault. Figure 3(a) is a bright-field image taken under multibeam conditions. It shows the junction of prismatic

(vertical feature) and basal plane (horizontal feature) faults. ¯¯14其 In Fig. 3(b), taken under two-beam conditions with 兵21 reflecting planes, the contrast of the basal plane fault was extinguished while that of the prismatic fault is still visible. Such behavior demonstrates that the two faults have different displacement vectors. As a consequence, a stair-rod dislocation is expected at the intersection. This dislocation, clearly visible in Fig. 3(b), produced relatively weak contrast with ¯ 10 and g most reflections and was extinguished for g = 12 = 0004. This indicates that the Burgers vector equals 1 / n ¯ 0兴 with n ⬎ 3 and is similar to that observed by Drum.11 关101 The n ⬎ 3 assignment is based on contrast being weaker than for n = 3 dislocations and prismatic fault displacement vector analysis. The basal plane fault was found to extend for several tens of micrometers away from the stair-rod dislocation and was bound on the other side by a Frank-partial dislocation ¯ 03兴. This dislocation exhibits strong difwith bF = 1 / 12关44 fraction contrast in Fig. 4(b) in agreement with its 1 / 4关0001兴 component. The basal plane fault displacement vector 共RBPF兲 has also a c-axis component and satisfies 共RBPF + bF = T兲 with T being the separation vector between any two equivalent atomic sites and bF the Burgers vector of the partial disloca¯ 03典 contion. Accordingly, the fault exhibits the I1 = 1 / 12具44 ¯ ¯ 02 [Fig. figuration. It was in contrast with g = 1100 and g = 11 ¯ 1x and g = 21 ¯¯1x [Fig. 4(a)] and out-of-contrast with g = 12 3(b)]. Stacking faults are out-of-contrast when g · R product has an integer value. The prismatic fault produced contrast with all diffraction vectors used in this study with the exception of g = 0004. Though this does not provide sufficient information to determine its displacement vector 共RPPF兲, it shows that it does not take the I1, I2, or the E configuration, as usually observed for basal faults. RPPF is also given by the sum of the basal fault displacement RBPF and the stair-rod dislocation Burgers

FIG. 2. Optical microscopy image of the as-grown epilayer surface with the ¯ 02, carrot defect. The zigzag groove forms at the intersection of the planar FIG. 4. Dark-field images of the basal fault termination with (a) g1 = 11 defect with the sample surface. and (b) g2 = 0004. Downloaded 06 Jan 2005 to 128.2.105.86. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp

021905-3

Appl. Phys. Lett. 86, 021905 (2005)

Benamara et al.

FIG. 5. Optical image of the etched carrot defect showing groove at the intersection of prismatic fault with the epilayer surface, two basal plane dislocation etch pits bounding the basal plane fault, and a weak line of the basal plane fault itself (marked with an arrow).

vector. Based on this, it is suggested that RPPF ¯ 03典 ± 1 / n关101 ¯ 0兴. Further investigations by high= 1 / 12具44 resolution electron microscopy are required to verify this assignment. An additional confirmation of the basal plane fault frequently appearing at the bottom of the carrot was obtained by KOH etching. Figure 5 shows an optical image of a different carrot with a horizontal groove enhanced by etching. The stair-rod dislocation forms an oval-shaped etch pit at the end of the groove. Oval pits are characteristic of dislocations with the line direction in the basal plane.16 In addition, one can see another oval pit located on the lower side of the groove and separated by 8 ␮m from the groove. Both pits are connected by a straight etched line (marked with an arrow in the figure and clearly visible in the inset). These features are consistent with the basal plane stacking fault bounded by two partial dislocations emerging on the surface of the off-cut epilayer. A geometric model of the analyzed structure is presented in Fig. 6. It is a compilation of results presented in this letter

FIG. 6. Geometric model of the carrot defect. Basal plane (B) and prismatic (P) faults are connected by a stair-rod (SR) dislocation. These faults are bounded by Frank partial (FP) and threading dislocation (TD), respectively. The defect nucleates on a threading screw dislocation propagating from the substrate.

and the overall carrot structure observed in x-ray topography images.9 Similarly to single prismatic faults observed under micropipe-related trenches,17 the carrot defect nucleates on a dislocation with a Burgers vector close to c axis propagating from the substrate. It consists of two stacking faults intersecting along the stair-rod dislocation. The prismatic fault is bounded by a threading dislocation with its Burgers vector primarily along [0001] on the up-step side and by a stair-rod dislocation along the basal plane. The threading dislocation is usually tilted in the down-step direction.9 The basal plane stacking fault has a Frank character and is bounded by a Frank partial on one side and the stair-rod dislocation on the other. The whole structure nucleates at the epilayer–substrate interface and might result from the cross-slip of the starting threading dislocation or from the dissociation of a Frankpartial dislocation18 at the interface. Furthermore, similarities with the Lomer–Cottrell lock frequently observed in fcc metals19 are obvious and let us think that the observed structure might as well result from the interaction of the starting threading dislocation with a basal plane dislocation. At this time, the driving force for formation of this defect is not known and is a subject of an ongoing study. The authors acknowledge many fruitful discussions with Dr. S. Ha and N. T. Nuhfer. This work was supported in part by ONR Grant Nos. N00014-02-1-0427 and N00014-02-C0302, monitored by Dr. Harry B. Dietrich. 1

R. Rupp and I. Zverev, Mater. Sci. Forum 433-6, 805 (2003). S. Ferrero, S. Porro, F. Giorgis, C. F. Pirri, P. Mandracci, C. Ricciardi, L. Scaltrito, C. Sgorlon, G. Richieri, and L. Merlin, J. Phys.: Condens. Matter 14, 13397 (2002). 3 T. Kimoto, N. Miyamoto, and H. Matsunami, IEEE Trans. Electron Devices 46, 471 (1999). 4 T. Kimoto, Z. Chen, S. Tamura, and S. Nakamura, Jpn. J. Appl. Phys., Part 1 40, 3315 (2001). 5 P. G. Neudeck and J. A. Powell, IEEE Electron Device Lett. 15, 63 (1994). 6 T. Okada, T. Kimoto, K. Yamai, H. Matsunami, and F. Inoko, Mater. Sci. Eng., A 361, 67 (2003). 7 W. L. Zhou, P. Pirouz, and J. A. Powell, Mater. Sci. Forum 264-2, 417 (1998). 8 T. Okada, T. Kimoto, H. Noda, T. Ebisui, H. Matsunami, and F. Inoko, Jpn. J. Appl. Phys., Part 1 41, 6320 (2002). 9 X. Zhang, S. Ha, M. Benamara, M. Skowronski, M. J. O’Loughlin, and J. J. Sumakeris, Appl. Phys. Lett. 85, 5209 (2004). 10 H. Blank, P. Delavignette, and S. Amelinckx, Phys. Status Solidi 2, 1660 (1962). 11 C. M. Drum, Philos. Mag. 11, 313 (1965). 12 P. Vermaut, P. Ruterana, G. Nouet, and H. Morkoc, Philos. Mag. A 75, 239 (1997). 13 M. Benamara, X. Zhang, J. Grim, P. Ruterana, G. Nouet, J. J. Sumakeris, M. J. Paisley, and M. Skowronski, Materials Research Society, Spring Meeting 2004 (unpublished). 14 D. B. Williams and C. B. Carter, in Transmission Electron Microscopy (Plenum, New York, 1996), p. 386. 15 Z. Liliental-Weber, M. Benamara, J. Washburn, I. Grzegory, and S. Porowski, Phys. Rev. Lett. 83, 2370 (1999). 16 S. Ha, W. M. Vetter, M. Dudley, and M. Skowronski, Mater. Sci. Forum 389–393, 443 (2002). 17 N. Vouroutzis, R. Yakimova, M. Syväjärvi, H. Jacobson, J. Stoemenos, and E. Janzén, Mater. Sci. Forum 389–393, 395 (2002). 18 J. P. Hirth and J. Lothe, Theory of Dislocations, 2nd ed. (Wiley, New York, 1982), pp. 320–324. 19 D. Hull and D. J. Bacon, Introduction to Dislocations, 3rd ed. (Pergamon, New York, 1984), p. 106; J. P. Hirth, and J. Lothe, in Ref. 18, p. 799. 2

Downloaded 06 Jan 2005 to 128.2.105.86. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp