Molecular Dynamics of Sputtering Yield as a Function of Ion Incident Angle in Chlorine-Adsorbed GaN Crystal K. Harafuji1 and K. Kawamura2 1 Department of Electrical and Electronic Engineering, Ritsumeikan University, Shiga, Japan 2 Graduate School of Environmental and Life Science, Okayama University, Okayama, Japan Abstract: A molecular dynamics simulation has been made to study the dependence of sputtering yield on the ion incident angle in the wurtzite-type GaN(0001) surface with the Cl-adsorbed layer. Angular dependence of the sputtering yield is calculated for a wide range of very oblique (10 degrees) to perpendicular (90 degrees) 250 eV Ar incidence. Sputtering yields for Ga and N atoms show maximum values of 0.63 and 0.20 at 50 degrees, respectively. Keywords: dry-etching, molecular dynamics, sputtering yield, GaN 1. Introduction GaN is a suitable material for power devices and short-wavelength optoelectronic devices such as lightemitting diodes (LEDs) and laser diodes (LDs) [1,2]. Since GaN is chemically inert in acids and bases at room temperature, plasma dry etching is a necessary step in device fabrication [3]. Reactive ion etching is thus conventionally used for mesa structure to reach ntype material in LEDs and for ridge structure to attain lateral optical confinement in LDs [4]. A gas mixture of Cl-based chemicals is usually used for etching discharge [5, 6]. Control of the incident angle of energetic ions in sputtering process is an important item in achieving efficient but damage-free dry-etching especially for atomic layers in active regions. The dry etching mechanism is, however, least well-understood. Molecular dynamics (MD) simulation is a powerful tool to investigate details of physical and chemical mechanisms that would otherwise be difficult or impossible to obtain [7]. Several studies on silicon etching by MD simulation have been reported [8-10]. In this article, MD simulation is made to investigate the angular dependence of the sputtering yield for a wide range of very oblique (10 degrees) to perpendicular (90 degrees) Ar incidence. Two cases are simulated, one with a clean surface and the other with a Cl-adsorbed surface. Preliminary results have been presented elsewhere by the present authors [1113]. 2. Simulation Model 2.1. Potential and Cell Parameters The functional form of a two-body interatomic potential is modeled as follows: uij Zi Z j e2 4 r f 0 (bi b j ) exp r ) D2ij exp( 0 ij D1ij exp( 1ij ij ai aj bi rij bj r ) 2ij ij ci c j (1) rij6 where the first term is the Coulomb interaction, the second term is the Gilbert-type short-range repulsion, the third and fourth terms represent the covalent bonding and covalent repulsion of the modified Morse type, respectively, and the last term is the van der Waals potential. The variable rij is the interatomic distance between the i-th and the j-th atoms, ε0 is the dielectric constant of the vacuum, Zi is the effective charge for each atom, f0 is the constant for unit conversion [41.86 kJ/(nm・mol)], ai is the repulsion radius, bi is the softness parameter, D1, D2, β1 and β2 are covalent coefficients, and ci is the van der Waals coefficient. Potential parameters are determined based on the periodic restricted Hartree-Fock ab initio method (CRYSTAL 98) [12,14]. The obtained potential parameters are listed in Table 1. The accuracy of the interatomic potential for Ga and N atoms is checked with respect to elastic constants, phonon spectrum, melting point and phase transition [15-17]. The argon ion is neutralized before impact and therefore the potential is for neutral Ar. Table 1. Calculated potential parameters in eq.(1). atom N Ga Ar Cl Na atomatom N-Ga Z(e) a (nm) b (nm) -1.150 1.150 0 -0.48 0.48 0.1970 0.0834 0.1878 0.2061 0.1493 0.0123 0.00911 0.0117 0.0190 0.0120 D1 (kJ/mol) -5250.5 β1 (nm-1) 20.0 c (kJ/mol)1/2 (nm)3 0.0364 0.0 0.0788 0.0573 0.0184 D2 (kJ/mol) 6581.7 β2 (nm-1) 40.0 2.2. MD Model The MXDORTO code developed by Kawamura is used for the present MD simulation [18]. The molecular motion is solved by the Verlet method. The Coulomb interaction is calculated by the Ewald sum method. Details of the approach we took in the simulation are similar to those in previously reported studies [1113]. All atomic pairs among gallium, nitrogen, chlorine, sodium, and argon are taken into account using the potential of eq. (1). Figure 1 shows the model configuration, where Figs. 1(a) and 1(b) are the top view and the side view of the crystal part, respectively. The small blue and intermediate green circles denote Ga and N atoms, respectively. The small yellow and orange circles are the Cl and Na atoms, respectively. The large pink circle is the Ar atom. An MD basic cell consists of 1920 Ga atoms and 1920 N atoms. The top surface has half-monolayer Cl coverage (60 Cl atoms with -0.48e charge). Counter 60 Na atoms with +0.48e charge are set on the bottom surface to satisfy the charge neutrality condition for the Ewald sum method. In Fig. 1(a), Na atoms on the bottom surface are omitted to clearly show the distribution of Cl atoms. Potential parameters of Cl and Na are determined on the basis of the rock-salt crystal of NaCl. The charge of Cl atoms is estimated by the ab initio method (Gaussian 03) with MP2/321G for GaCl2 and GaCl3 molecules [19]. The bottom one-pair layer of Ga and N atoms of the MD basic cell, 240 atoms, and 60 Na atoms are set to the thermostat of 300 K. The surface area of the MD basic cell is 3.21×3.33 nm2. The height is 4.31 nm. A vacuum region with a height of 12.28 nm is set along the C direction of the cell. Periodic boundary conditions are employed three dimensionally on the boundary of the cell. MD simulations are performed under the NVE ensemble except for the bottom thermostat layers, where N (number of atoms), V (volume), and E (energy) are kept constant. First, the relaxation calculation of 15 ps without Cl and Na atoms and that of 14 ps with Cl and Na atoms are carried out to attain an equilibrium state. Second, an energetic Ar atom with 250 eV is brought into the surface with the angle θ from 10 to 90 degrees with the increment of 10 degrees. The statistical data of sputtering yield is obtained averaged over 30 Ar ions for each incident angle. The incident position is given by random number. A time step of 0.6-0.7 fs is used depending on the Ar energy. At each Ar incidence, the motion of all atoms is followed at least for 6 ps. It takes about 3 h CPU time for each Ar impact to complete the simulation by the use of a 2.4 GHz personal computer. C 0001 Ar NG C N a l a θ 01 1 0 21 10 (a) Vacuum Region 300K thermostat (b) Fig. 1. Simulation model. (a)Top view. (b)Side view. 3. Calculation Result Figure 2 shows the result of Ga and N sputtering yields for a wide range of very oblique (10 degrees) to perpendicular (90 degrees) Ar incidence. Circle and triangle denotes Ga and N sputtering yields, respectively. Blue line denotes the case of clean surface without Cl-adsorbed layer. Sputtering yields for Ga and N atoms show maximum values of 0.33 and 0.60 at 50 degrees, respectively. On the other hand, red line denotes the case with the Cl-adsorbed layer. Sputtering yields for Ga and N atoms show maximum values of 0.63 and 0.20 at 50 degrees, respectively. In the case with the clean surface, for very oblique Ar incidence with the angle less than 20 degrees, the Ar atom grazes over the crystal surface, and does not generate definite collision-cascade and hot-spot. At 30 degrees incidence, the Ar atom brings about both the collision-cascade and hot-spot up to the top three layers, but seldom generates the sputtering. At 40 degrees incidence, the sputtering rate increases largely. The crystal randomization takes place up to the top five layers. For incident angles with 40-70 degrees, a cooperation of the horizontal and vertical momenta efficiently kicks atoms in the crystal, and generates collision-cascade to a great deal. This brings about the high sputtering rate. Nitrogen atoms are preferably sputtered. The reason is as follows. When an incident Ar atom collides with a heavy Ga atom, Ar is reflected upward, that is, backscattered. The heavy Ga atom serves as a reflecting wall for the incident Ar atom. The backscattered Ar atom gives the upward momentum and energy of crystal atoms, especially those of light N atoms. In the case with the Cl-adsorbed layer, chemical sputtering of Ga is promoted, whereas N sputtering is much suppressed due to the Coulomb repulsive force between N and Cl atoms. Ga atom is chemically sputtered mostly in the form of Ga-Cl2 or Ga-Cl. These products escape from the surface in the time range of mainly 200-500 fs after the impact of the incident Ar ion. There are small amounts of products escaping in the time range of 500-2000 fs. N atom is mostly sputtered in the form of Ga-N in the time range less than 200 fs. It is to be noted that there is a finite sputtering rate even at 20 degrees for both Ga and N atoms. For both cases with and without Cl-adsorbed layer, sputtering rate is peaked at 50 degrees. The obliquely incident Ar atom efficiently penetrates into the inner atomic layers through the gap between atoms, reflected obliquely by the Ga atom, and kicks upward both light N atom and heavy Ga atom especially on the top layer. This is the reason why obliquely incident Ar atom brings about the large sputtering rate not only of N atom but also of Ga atom. Figure 3 shows the snapshot of the side view of the top five layers after the Ar impact to the Cl-adsorbed surface. The incident angle is 20 degrees. Ar atom reaches the surface at 93 fs. Figure 3(a) shows the snapshot at 643 fs. The incident Ar atom collides with a Cl atom, and is reflected from the surface. The Cl atom itself subsequently collides with Ga-N atoms on the first layer shown by dashed line. A very thin hot spot on the adsorbed Cl and the first Ga-N layers is generated around the collided Cl incident point. The original hexagonal crystal structure is disordered there. The Cl atom is also reflected from the surface. Reflected Ar 643 fs Sputtered Cl Hot-Spot 0.7 S puttering Yield 0.6 0.5 N Ga N (C l) G a(C l) 0.4 (a) 0.3 1755 fs 0.2 GaCl2 product is leaving. 0.1 0 0 10 20 30 40 50 60 70 80 90 Incident A ngle (deg) Fig.2 Angular dependence of the sputtering yield for Ar ion incidence with 250 eV, where 90-degrees denotes the normal incidence. (b) Fig.3 Snapshot side view of near surface layers after the impact of Ar atom with 20 degrees incidence. (a)643 fs, (b)1755 fs. . Figure 3(b) shows the snapshot at 1755 fs. The hot spot shrinks, and then disappears. The original hexagonal crystal structure is almost recovered. That is, recrystallization takes place. A volatile product of a GaCl2 molecule is generated, and this molecule begins to leave the surface. 4. Conclusion A molecular dynamics simulation has been made to study the dependence of sputtering yield on the ion incident angle in the wurtzite-type GaN(0001) surface with and without a Cl-adsorbed layer. Angular dependence of the sputtering yield is calculated for a wide range of very oblique (10 degrees) to perpendicular (90 degrees) 250eV Ar incidence. In the case with the Cl-adsorbed layer, chemical sputtering of Ga is promoted, whereas N sputtering is much suppressed due to the Coulomb repulsive force between N and Cl atoms. Ga atom is chemically sputtered mostly in the form of Ga-Cl2 or Ga-Cl. Sputtering yields for Ga and N atoms show maximum values of 0.63 and 0.20 at 50 degrees, respectively. References [1] S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, H. Kiyoku, Y. Sugimoto, T. Kozaki, H. Umemoto, M. Sano, and K. Chocho, Jpn. J. Appl. Phys. 37, L309 (1998). [2] S. Nakamura, Phys. Status Solidi A 176, 15 (1999). [3] L. Zhang, J. Ramer, J. Brown, K. Zheng, L. F. Lester and S. D. Hersee, Appl. Phys. Lett. 68, 367 (1996). [4] Y. Lacroix, T. Nakanishi and S. Sakai, Proc. Int. Workshop on Nitride Semiconductors, IPAP Conf. Series 1, Nagoya, 2000, p.782. [5] I. Adesida, A. Mahajan, E. Andideh, M. A. Khan, D. T. Olsen, and J. N. Kuznia: Appl. Phys. Lett. 63, 2777 (1993) . [6] S. J. Pearton, C. R. Abernathy, and F. Ren: Appl. Phys. Lett. 64, 2294 (1994). [7] M. E. Barone, T. O. Robinson, and D. B. Graves, IEEE Trans. Plasma Sci. 24, 77 (1996). [8] D. E. Hanson, A. F. Voter and J. D. Kress, J. Appl. Phys. 82, 3552 (1997). [9]M. E. Barone and D. B. Graves, J. Appl. Phys. 78, 6604 (1995). [10] H. Feil, J. Dieleman, and B. J. Garrison, J. Appl. Phys. 74, 1303 (1993). [11] K. Harafuji and K. Kawamura, Jpn. J. Appl. Phys. 47, 1536 (2008). [12] K. Harafuji and K. Kawamura, Jpn. J. Appl. Phys. 49, 08JE03 (2010). [13] K. Harafuji and K. Kawamura, Proc. 20th Int. Symp. on Plasma Chemistry, Philadelphia, 2011, FDM08. [14] V. R. Sanders, R. Dovesi, C. Roetti, M. Causa, N. M. Harrison, R. Orlando, and C. M. Zicovich Wilson, (1999) CRYSTAL 98. User’s manual. University of Torino (Italy) and CLRC Daresburg Laboratory (UK). [15] K. Harafuji, T. Tsuchiya, and K. Kawamura, Jpn. J. Appl. Phys. 43, 522 (2004). [16] K. Harafuji, T. Tsuchiya, and K. Kawamura, J. Appl. Phys. 96, 2501 (2004). [17] K. Harafuji, T. Tsuchiya, and K. Kawamura, J. Appl. Phys. 96, 2513 (2004). [18] K. Kawamura, Japan Chemistry Program Exchange (1996) P029. [19] Gaussian 03, Revision E.01, User’s Reference (Gaussian, Inc., Wallingford CT, 2004).
© Copyright 2026 Paperzz