Investigation of Mechanical Stress-Induced Double Stacking Faults in

dislocations and their influence on the electrical properties of the material. The defects were nucleated by one scratch performed on the (11-20) sample surface ...
2MB taille 1 téléchargements 215 vues
Materials Science Forum Vols. 527-529 (2006) pp. 379-382 online at http://www.scientific.net © (2006) Trans Tech Publications, Switzerland Online available since 2006/10/15

Investigation of Mechanical Stress-Induced Double Stacking Faults in (11-20) Highly N-doped 4H-SiC Combining Optical Microscopy, TEM, Contrast Simulation and Dislocation Core Reconstruction M. Lancin1, G. Regula1, J. Douin2, H. Idrissi1, L. Ottaviani1, B. Pichaud1 1

2

TECSEN, UMR-6122, Université Paul Cézanne, 13397 Marseille-cedex20 -France CEMES, UPR 8011, 29 rue Jeanne Marvig, BP 94347, 31055 Toulouse cedex 4 France [email protected]

Keywords: 4H-SiC, double-stacking faults, HRTEM, WBTEM, dislocation core composition

Abstract: Defects are introduced into (11-20) highly N-doped 4H-SiC by one surface scratch followed by annealing at 550°C or 700°C with or without an additional compressive stress. The defects are planar and always consist of double stacking faults dragged by a pair of partial dislocations. In a pair, the partial dislocations have the same line direction, Burgers vector and core composition. All the identified gliding dislocations have a silicon core. An analysis of their expansion during annealing proves that C(g) partial segments can be created but that C(g) partial dislocations are immobile. Introduction We introduced defects into 5×1018 cm-3 N-doped 4H SiC in order to study the dynamics of the dislocations and their influence on the electrical properties of the material. The defects were nucleated by one scratch performed on the (11-20) sample surface at 45° to the [1-100] direction (Fig. 1a,b) in order to optimize the forces applied on the dislocation sources. The defect expansion was induced by both cantilever bending so that the sources were under compression and annealing at low temperature (T ≤ 700°C). In our previous works [1-3] we were interested in the defects which expand under compressive stress. We only found double stacking faults (DSFs) dragged by 30° partial dislocations (PD) with a silicon core (Si(g)). Our result differed from those already published. Indeed, the defects created either by indentation or constant compressive strain rate consist of faults dragged by Si(g) but they are single (SSFs) or multiple faults but not DSFs [4-6]. When the defects are created by sample surface grinding, they consist of DSFs but they are dragged by 30° partials with either silicon or carbon core [7-9]. The discrepancies between the different results suggest that the deformation procedure may influence both the PD core and the stacking fault (SF) multiplicity. To check this hypothesis, we decided to characterize all the defects created by our deformation process and not only the ones which expand under the applied stress [10]. Thus, after chemical etching of the deformed samples, three kinds of defects (called A, B and C) can be sorted according to their propagation direction from the scratch (P1 or P2, see Fig.1a,b) and their expansion. The A faults expand with the applied stress and can reach 3 mm, they correspond to the DSFs already characterized [1]. The other faults, B and C, glide in the P1 or the P2 direction, respectively, but their expansion never exceeds 200 µm. They also consist of DSFs dragged by a pair of Shockley partial dislocations with a silicon core [10]. For each population of DSFs, we found only one Burgers vector direction for all the PD pairs. If we consider the influence of the compressive stress on the PDs, we find that DSFA and DSFB must expand but that DSFC should shrink. The DSFC expansion, although limited, proves that the scratch induces a stress field which plays a role during the annealing. At that 1 All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of the publisher: Trans Tech Publications Ltd, Switzerland, www.ttp.net. (ID: 193.49.32.89-31/10/07,11:30:18)

380

Silicon Carbide and Related Materials 2005

point we found necessary to check which defects could expand under the sole influence of the stress field induced by the scratch in order to verify if SSFs or DSFs dragged by partial dislocation with a carbon core could be formed. The present work is thus focussed on the nucleation mechanism of the defects. Results and discussion In the samples annealed at 550°C or 700°C without additional compressive stress (Fig. 1c), chemical etching revealed faults which expand in the P1 = [-1100] or in the P2 = [1-100] direction, their propagation length never exceeding 200 µm. We obtained high resolution images on (11-20) planar views (Fig. 2). On the images, each white dot corresponds to the projection of a dumbbell that is the projection of Si and C atomic columns which are 0.109 nm apart and which thus cannot be resolved by the technique. In figure 2, the image contrast was simulated using the multi-slice method (ems software, P. Stadelmann). From an analysis of the contrast [11], the Si and C atomic columns are localized in the experimental images. All the faults consist of six layers with cubic stack characteristic of DSFs in 4H-SiC. In both P1 and P2 directions, we observed the two possible cubic stacks labeled “a-3C” and “b-3C”. In P1, we found 7 a-3C and 11 b-3C and in P2, 5 a-3C and 19 b-3C. Note that the white dots also reveal the projections of the four possible glide planes in a 4HSiC structure. It is clear that the two partial dislocations dragging the DSF must glide in G1 and G2 to form the a-3C stack and in G3 and G4 to form the b-3C stack.

c) P1

Figure 1: Optical images of (11-20) sample surfaces annealed at 700°C for 0.5h and etched 0.16 h at 500°C with molten KOH: samples were annealed with (a and b) or without (c) additional compressive stress. The dark lines reveal the intersection with the sample surface of faults which develop in P1 or P2 directions. Figure b) is a zoom of the area drawn in a).

[000-1] (1)

(2)

(1)

(3)

(3)

b-3C

a-3C G1 G2

(2)

G3 G4

[11-20]

[1-100]

Figure 2: [11-20] representations of the faults. In simulated (1) and experimental (3) HRTEM images, the white dots correspond to the projection of the Si-C dumbbells. They also reveal the projections of the 4 possible glide planes in the 4H-SiC, G1, G2, G3 and G4. In the drawings (2), the projections of Si and C atomic columns are represented as large and small circles respectively. They are superimposed on the left and right sides of the images (1) and (2) respectively. The six cubic layers are characteristic of DSFs in 4H-SiC and they induce two possible stacks labelled a-3C and b-3C. 2

Materials Science Forum Vols. 527-529

381

Weak Beam TEM studies were performed on (0001) thin foils prepared using the focused ion beam technique to characterize the partial dislocations next to their intersection with the sample surface. Figure 3 shows typical images obtained in the P2 direction, similar results being obtained in the P1 direction. In each DSF, the partial dislocations D1 and D2 have the same line direction and Burgers vector direction (b). However, the line direction varies from defect to defect and along the same defect. Such a feature shows that the scratch induces a complex stress field. Fourteen of the fifteen DSFs characterized by WBTEM had the same characteristics as the DSFs observed in samples annealed under cantilever bending except for their PD line direction. We thus categorized them by analogy with our previous results. We identified in P1 8 DSFA (a-3C stack and b = ± a/3 [1-100] where a is the 4H-SiC cell parameter) and one DSFB (b-3C and b = ± a/3 [01-10]) and in P2 5 DSFC (b-3C and b = ± a/3 [1-100]). In P2, the DSF not yet observed and labeled D was characterized by the a-3C stack and b = ± a/3 [1010]. Note that we observed 90° segments for each population of DSFs. To fully characterize the defects, we determined the core composition of the partial dislocations. Knowing the Si and C atomic positions in the foils and the pair of glide planes {G1,G2} or {G3,G4}, we reconstructed the core of the PDs which create the four populations of DSFs. In Figure 4 we show the core reconstruction of the 90° PD dragging a DSFA and thus gliding in G1 and G2 in the P1 direction. The 90° PD is of Si(g) type and it is characterized by a line L = [-1-120] and b = a/3[1-100]. This result is in perfect agreement with the reconstruction of the 30° segment lying along the [-12-10] valley of Peierls [10].

Figure 3: WBTEM images taken near the [0001] axis.

a)

[0001]

SF

b)

SF

G1

G2 SF

[11-20]

SF

[1-100]

Figure 4: a) Core reconstruction realized to obtain the structural unit characteristic of 90° PDs: Si (large dots) and C (small dots). Doted lines indicate the G1 and G2 glide planes. The line direction [-1-120] and the Burgers vector (b=SF) of the PDs were determined using the SF/RH convention. b) Sketch of a DSFA created by the scratch when an additional cantilever bending is applied: only the Si(g) segments are mobile during annealing. 3

382

Silicon Carbide and Related Materials 2005

All the PDs dragging the DSFs, including the DSFD, are of the Si(g) type. The results of this study confirm those obtained in samples annealed under additional compressive stress. Moreover they suggest that 90° PDs are first created by the scratch. We can thus describe the DSF expansion using the results obtained in samples annealed with and without bending stress. As for the DSFA, a 90° Si(g) is created by the scratch (Fig.4 b). When the loop expands to release the stress, a 30° Si(g) is created. Because a 90° PD is more mobile than a 30° PD, the 90° PD glides faster and disappears. Such an interpretation is in good agreement with the results of Pirouz et al.[4-6]. When the 30° Si(g) reaches the neutral fiber where the stress is equal to zero it cannot glide deeper. As the DSFA glides further in P1, this implies that C(g) segments are created. And, indeed, we do observe in other experiments not shown here a zigzag shape of the dislocation with Si and C segments. The expansion of the DSFA over long distance in P1 proves that C(g) can be created and its lack of expansion in P2 proves that C(g) cannot glide. This fact confirms those obtained during plastic deformation [3-6]. Let us consider now the three other types of DSFs detected which are dragged by Si(g) but the expansion of which is limited. As we mentioned above, the DSFC loops should shrink under the applied compressive stress and indeed their length decreases with this stress. The DSFD loop should also shrink and indeed its expansion is very limited. On the contrary, the DSFB loops should expand but they do not. For a given stress, the compressive forces are two times smaller on DSFB loops than on DSFA loops. It is known that in a material where there are several glide systems, the primary system is the first excited. If it is efficient to release the stress, the other glide systems are not excited. This may explain the different loop expansion as a function of b. But a question still remains: why did we observe only DSFs and not SSFs? Two driving forces are proposed to account for the DSF formation: - the gain in energy due to the 3C4H phase transformation, - the gain in energy due to the quantum well action in highly n-doped 4H-SiC [7,12,13]. But these two driving forces can also be invoked during plastic deformation whereas only SSFs or multiple SFs are observed. We suggest that we should take into account a third contribution: the dislocation interaction has an energy cost which may counterbalance the gain in energy due to the local phase transformation or the QWA. This should be the case during plastic deformation, where a high density of dislocations is nucleated and where the DSF formation is not observed. On the contrary in our deformation procedure which induces a low density of long dislocations, the energy cost due to the dislocation interaction is too low to prevent the DSF formation. In summary, the performed study shows that only DSFs dragged by Si(g) partial dislocations are created by our deformation procedure. The different DSF expansions are a consequence of their Burgers vectors direction and thus of the deformation procedure. References [1] H. Idrissi, M. Lancin, J. Douin, G. Regula, B. Pichaud, R. El Bouayadi, and J.M. Roussel : Mat. Res. Soc. Symp. Proc. 815 (2004), p. J72.1 [2] H. Idrissi, G. Regula, M. Lancin, J. Douin and B. Pichaud: Phys. Stat. Sol. (c) 6 (2005), p.1998-2003 [3] L. Ottaviani, H. Idrissi, P. Hidalgo, M. Lancin, B. Pichaud: Phys. Stat. Sol. (c) 6 (2005), p.1792-96 [4] A.V. Samant, M.H. Hong and P. Pirouz: Phys. Stat. Sol. (b) 222, (2000), p.75 [5] P. Pirouz and J.W. Yang: Ultramicroscopy 51, (1993), p. 189 [6] P. Pirouz, J.L. Demenet and M.H. Hong: Phil. Mag A, 81, (2001), p. 1207 [7] J.Q. Liu, H.J. Chung, T.A. Kuhr, Q. Li and M. Skowronski: Appl. Phys. Lett. 80, 12, (2002), p. 2111 [8] T.A. Kuhr, J.Q. Liu, H.J. Chung, M. Skowronski and F. Szmulowicz: J. Appl. Phys. 92, (2002), p. 5863 [9] H.J. Chung, J.Q. Liu and M. Skowronski: Appl. Phys. Lett. 81 (2002), p.3759 [10] G. Regula, M. Lancin, H. Idrissi, B. Pichaud and J. Douin: Phil Mag Lett. 85 (2005), p. 259-267 [11] M. Lancin, C. Ragaru and C. Godon: Phil. Mag. B 81, (2001), p. 1633 [12] M.S. Miao, S. Limpijumnong and W.R.L. Lambrecht: Appl. Phys. Lett. 79, (2001), p.4360 [13] H.P. Iwata, U. Lindefelt, S. Oberg and P.R. Briddon: J. Appl. Phys. 93, (2003), p.1577; ibid. J. Appl. Phys. 94, (2003) p.4972

4