Abstract
The effect of magnetic compound fluid (MCF) slurry on the gallium arsenide (GaAs) wafer surface after nano-precision polishing was investigated. MCF slurry was prepared by mixing CS carbonyl iron particles (CIPs), Al2O3 abrasive particles, α-cellulose, and magnetic fluid. Firstly, a polishing device was assembled by designing MCF unit for the generation of revolving magnetic field. Then, the spot polishing experiments were performed on GaAs wafer surface to clarify the effects of MCF components on the surface roughness Ra and material removal (MR) at different polishing positions. Finally, the scanning polishing experiments were conducted using water-based MCF slurry containing particles with different diameters. Results show that after spot polishing with water-based and oil-based MCFs, the initial surface roughness Ra of 954.07 nm decreases to 1.02 and 20.06 nm, respectively. Additionally, the depth of MR is increased linearly with prolonging the polishing time. It is worth noting that the MR depth of surface after polishing with water-based MCF is 2.5 times higher than that with oil-based MCF. Meanwhile, the cross-section profile of the polished zone shows the W shape, which indicates the non-uniform MR on the workpiece surface after spot polishing. After scanning polishing, the cross-section profile of the polished zone shows the U shape, which indicates that MR is uniform under specific experiment conditions, regardless of the MCF types. The smoothest work surface with Ra=0.82 nm is achieved using MCF with abrasive particles of 0.3 μm in diameter, and MR rate is 13.5 μm/h.
Gallium arsenide (GaAs) wafer is one of the most important compound semiconductor materials due to its high electron mobility and low power consumptio
Novel methods for ultra-precision surface polishing have been developed and ameliorated continuously, such as the magnetic field-assisted polishing, which uses magnetic fluid (MF) or magnetorheological fluid (MRF) as the base component of slurry. For example, the surface quality of a flat polypropylene workpiece can be significantly improved by MF polishing metho
In order to overcome these problems and to maintain the excellent performance of MF and MRF slurries, a novel magnetic compound fluid (MCF) slurry is propose
In this research, an effective polishing technique for the nano-precision surface finishing of GaAs wafer was investigated. The finishing feasibility of GaAs wafer by MCF slurry was verified. The effects of MCF slurry composition and scanning path on the material removal were discussed. The surface roughness of the workpiece at different positions in the polishing area was analyzed.

Fig.1 Schematic diagram of MCF polishing process
According to the MCF polishing principle, an experiment device was designed and assembled, as shown in

Fig.2 Appearance of experiment device
Parameter | Value |
---|---|
Dimension of Nd-Fe-B magnet | Φ18 mm×10 mm |
Magnet eccentricity, r/mm | 4.5 |
MCF carrier rotational speed, nc/r·mi | 800 |
Magnet revolution speed, nm/r·mi | 1000 |
MCF slurry volume, V/mL | 0.5 |
Working gap, ∆/mm | 0.5 |
Polishing time, t/min | 60, 120 |
Material | Approximate size/μm | MCF1 | MCF2 | MCF3 | MCF4 |
---|---|---|---|---|---|
Oil-based MF | - | 40 | - | - | - |
Water-based MF | - | - | 40 | 40 | 40 |
CS CIPs | 7 | 45 | 45 | 45 | - |
HQ CIPs | 1 | - | - | - | 45 |
α-cellulose | - | 3 | 3 | 3 | 3 |
Al2O3 abrasive particles | 1 | 12 | 12 | - | 12 |
0.3 | - | - | 12 | - |
Intra by Taylor Hobson) to determine MR state. The surface roughness was measured by the white-light interferometer (Zygo Newview 600 frp, Zygo Corp).
The schematic diagram of GaAs wafer after polishing by MCF slurry for 60 min is shown in

Fig.3 Schematic diagram of workpiece after MCF polishing for 60 min

Fig.4 Surface roughness Ra of GaAs wafer after polishing by different MCF slurries for different durations

Fig.5 Surface roughness Ra at different areas on GaAs wafer after spot polishing by MCF2 slurry for different durations
The 3D images of wafer surface before and after polishing are shown in

Fig.6 3D images of work surface before (a) and after spot polishing by MCF1 slurry (b) and MCF2 slurry (c)
The surface roughness is closely related to MR state of workpiece surface. So the cross-sectional profile of polishing area was observed to discuss the effect of MCF slurry on MR. The cross-section profile and 3D structure of typical polished area after polishing with MCF2 slurry for 30 min are shown in

Fig.7 Cross-sectional profile (a) and 3D structure (b) of typical polished area after spot polishing by MCF2 slurry
The MR depth variation of wafer surface after polishing by MCF1 and MCF2 slurries is shown in

Fig.8 MR depth of wafer surface after spot polishing by MCF1 and MCF2 slurries for different durations
MR is related to the friction coefficient, polishing pressure, and relative velocity between abrasive particles in MCF slurry and workpiece surface under the external magnetic field. The friction coefficient between the abrasive particles and workpiece surface is constant, whereas the polishing pressure of MCF slurry acting on the polishing area varies according to the magnetic field intensity, which results in different polishing pressures of particles. The relative velocity depends on the rotational speed of magnet. In order to obtain the uniform MR, the forces acting on single CIP and abrasive particle under the magnetic field are analyzed, as shown in
(1) |

Fig.9 Schematic diagram of forces acting on CIP and abrasive particle under external magnetic field
where HB is the Brinell hardness of the GaAs workpiece (7350 N/m
(2) |
According to
(3) |
Then, the embedding depth of the abrasive particle acting on the workpiece surface can be expressed as follows:
(4) |
where di is the embedding depth of single abrasive particle acting on the workpiece surface.
In the process of MCF polishing, CIPs in MCF fluid is affected by the magnetic force, gravity, viscous resistance, Van der Waals force, and buoyancy. As shown in
(5) |
where µ0 is the permeability of vacuum; m is the magnetic dipole moment of CIP; H is the magnetic field; ∇ indicates the Hamiltonian operator. CIP can be regarded as the magnetic dipole owing to its extremely small size.
The value of m can be further determined by CIP volume (VCIP), CIP magnetic susceptibility (χm), CIP magnetic permeability (μ), vacuum permeability (μ0), and magnetic field H. The relationship equation is as follows:
(6) |
Then, the coordinate system (x, y, z) is defined, as shown in
(7) |
where i, j, and k are the unit vectors in the coordinate system. Thus, the magnetic field vector can be obtained by
(8) |
where Hx, Hy, and Hz are the components of the magnetic field in the x, y, and z directions, respectively.
Then, the forces on CIP is expressed as follows:
(9) |
with
(10) |
(11) |
(12) |
where FCIP-x, FCIP-y, and FCIP-z are the components of magnetic force in x, y, and z directions, respectively; ∂Hx/∂x, ∂Hx/∂y, and ∂Hx /∂z are the gradient of Hx in x, y, and z directions, respec-tively; ∂Hy/∂x, ∂Hy/∂y, and ∂Hy/∂z are the gradient of Hy in x, y, and z directions, respectively; ∂Hz/∂x, ∂Hz/∂y, and ∂Hz/∂z are the gradient of Hz in x, y, and z directions, respectively.
In fact, the single abrasive particle is also affected by many other forces during the polishing process. Among them, the magnetic levitation force is the primary factor affecting the working particles, and the force on single abrasive particles is as follows:
(13) |
where VAP is the volume of abrasive particle and M is the magnetization intensity of the magnetic fluid. M can be defined as follows:
(14) |
where χ′m is MCF magnetic susceptibility. Thus, the force on single abrasive particle is as follows:
(15) |
with
(16) |
(17) |
(18) |
where FAP-x, FAP-y, and FAP-z are the components of the magnetic force in x, y, and z directions, respectively.
Substituting
(19) |
In addition, the number of abrasive particles per circle can be defined, as follows:
(20) |
where rx is the radius of the active abrasive particle during the polishing process.
MCF process can remove a very small amount of material by penetration and rotation of abrasive particles on the wafer surface. The MR depth by abrasive particles is equal to the product of indentation depth of abrasive particles and finishing length. Thus, the MR depth (dMR) is calculated, as follows:
(21) |
where nc is the rotational speed of the permanent magnet.
MR depth can be obtained from
(22) |
In this research, CIPs and abrasive particles are considered as spherical particles with diameter of 7 and 1 µm and density of 7.8×1

Fig.10 Experimental and simulated MR depth of GaAs wafer surface after polishing
It can be seen that the theoretical curve shows a W shape after MCF polishing, which is consistent with the results of polishing experiment. In addition, the values of the fitting curve are slightly smaller than the experimental values, because the influence of magnetic force on the single abrasive particle is considered in the calculation in order to simplify the model. In the simulation study, the flow behavior of MCF is considered as laminar flow for simplification. In the tests, the flow behavior of MCF is too complex to be directly simulated, resulting in errors. As a result, MR on the workpiece surface using MCF is related to not only the polishing force, but also to the velocity of relative movement between the abrasive particles in MCF slurry and the workpiece surface. Therefore, the uniform MR can be obtained by adjusting the polishing track, which changes the polishing force and velocity of relative movement.
The water-based MCF slurry was used to polish the GaAs wafer along the scanning path due to its better MR effect and smoother surface, compared with the oil-based MCF slurry. In order to obtain the uniform MR and fine surface roughness, the selection of appropriate step length is crucial. Therefore, the step length should be less than 4.6 mm, which is the maximum MR depth, according to the cross-section profile in
The polishing unit moves along the polishing path, as shown in

Fig.11 Schematic diagram of MCF polishing along scanning path
The experiment results of workpiece surface after scanning polishing for 60 min are shown in

Fig.12 Cross-section profile (a) and 3D structure (b) of typical polished area after scanning polishing

Fig.13 Surface roughness Ra at different areas on GaAs wafer after scanning polishing by MCF3 slurry for different durations

Fig.14 Surface morphologies at the center of polishing area after scanning polishing with MCF2 slurry (a), MCF3 slurry (b), and MCF4 slurry (c)
After scanning polishing with different MCF slurries for 60 min, the surface roughness and MRR are shown in

Fig.15 Surface roughness (Ra) and MR rate (MRR) after scanning polishing for 60 min with different MCF slurries
1) GaAs wafer surface can be smoothened by the magnetic compound fluid (MCF) slurry. Larger material removal and smoother surface can be achieved through the water-based MCF slurry, compared with the oil-based MCF slurry. The cross-section profile of workpiece after spot polishing shows the W shape, indicating that the material removal is uneven. The material removal is increased linearly with prolonging the polishing time.
2) The wafer surface after scanning polishing with water-based MCF slurry shows the U shape. The optimal surface has the roughness of 0.82 nm after spot polishing with the MCF slurry containing smaller abrasive particles. Besides, the maximum surface roughness is 5.49 nm after polishing with the MCF slurry containing the carbonyl iron particles (CIPs) and abrasive particles of the same diameter.
3) Bigger CIPs and abrasive particles are beneficial to rapidly remove the material, whereas the same diameter of CIPs and abrasive particles in MCF slurry is harmful to simultaneously meet the requirements of high material removal and low surface roughness. Compared with the spot polishing results, the surface roughness of GaAs wafer improves by 85% after scanning polishing using specific slurry.
References
Sun Weizhong, Niu Xinhuan, Wang Haiyun et al. Rare Matel Materials and Engineering[J], 2006, 35(10): 1544 (in Chinese) [Baidu Scholar]
Chen H C, Lin C C, Han H V et al. Solar Energy Materials and Solar Cells[J], 2012, 104: 92 [Baidu Scholar]
Englhard M, Klemp C, Behringer M et al. Joural of Applied Physics[J], 2016, 120(4): 45301 [Baidu Scholar]
Yi D, Li J, Cao J. Bulgarian Chemical Communications[J], 2017, 49(K1): 113 [Baidu Scholar]
Matovu J B, Ong P, Leunissen L H A et al. ECS Journal of Solid State Science and Technology[J], 2013, 2(11): 432 [Baidu Scholar]
Yi D, Li J, Zhu P. Crystals[J], 2018, 8(8): 321 [Baidu Scholar]
Yu B J, Gao J, Chen L. Wear[J], 2015, 330–331: 59 [Baidu Scholar]
Signgh A K, Jha S, Pandey P M. Materials and Manufacturing Processes[J], 2012, 27(10): 1139 [Baidu Scholar]
Kawakubo H, Tsuchiya K, Sato U. Journal of the Japan Society for Abrasive Technology[J], 2010, 54(1): 31 [Baidu Scholar]
Yin S, Shinmura T, Wang D. Key Engineering Materials[J], 2006, 304–305: 320 [Baidu Scholar]
Shimmada K, Wu Y, Wong Y C. Journal of Magnetism and Magnetic Materials[J], 2033, 262(2): 42 [Baidu Scholar]
Shimada K, Toyohisa F, Oka H. Transactions of the Japan Society of Mechanical Engineers[J], 2001, 67(664): 3034 [Baidu Scholar]
Wu Y, Shimada K, Wong Y C. International Journal of Machining and Machinability of Materials[J], 2009, 5(2–3): 232 [Baidu Scholar]
Wu Y, Wang Y, Fujimoto M et al. Journal of the Korean Society of Manufacturing Technology Engineers[J], 2014, 23(6): 547 [Baidu Scholar]
Shimada K, Akagami Y, Kamiyama S et al. Journal of Intelligent Material Systems and Structures[J], 2002, 13(7): 405 [Baidu Scholar]
Guo H R, Wu Y B, Lu D et al. Journal of Materials Processing Technology[J], 2014, 214(11): 2759 [Baidu Scholar]
Wang Y L, Wu Y B, Nomura M. Precision Engineering[J], 2016, 45: 67 [Baidu Scholar]
Wang Y L, Wu Y B, Nomura M. Precision Engineering[J], 2017, 48: 32 [Baidu Scholar]
Zhao Zifen, Shao Huiping, Lin Tao et al. Rare Metal Materials and Engineering[J], 2016, 45(6): 1609 (in Chinese) [Baidu Scholar]