JP4770165B2 - High-speed EEM processing method using agglomerated fine particles - Google Patents

Description

The present invention relates to a high-speed EEM processing method using agglomerated fine particles. More specifically, ultrafine particles are brought into fluid contact with a processed surface of a workpiece such as a Si wafer, a semiconductor wafer such as SOI, and a thin film substrate. The present invention relates to a high-speed EEM processing method using agglomerated fine particles in which processing proceeds without causing any thermal alteration.

  Conventionally, a suspension in which ultrafine particles are dispersed is made to flow along the processed surface of the workpiece, and the ultrafine particles are brought into contact with the processed surface under almost no load. So-called EEM (Elastic Emission Machining), in which processed surface atoms are removed by an order close to the atomic unit by an interaction (a kind of chemical bond), is already known. EEM, which is one of the surface processing techniques, is an ultra-precise processing method that uses a chemical reaction between two solid surfaces, the surface of the workpiece, the surface of the workpiece, and the surface of the fine powder particles having reactivity. Since EEM uses large fine particles as viewed from the atomic scale, the atoms in the concave portions of the microroughness do not act on the surface of the fine particles, and the atoms in the convex portions are removed preferentially. Therefore, the removal amount to be processed is very small, and efficient processing is possible. Moreover, long period roughness can be removed by widening the fine particle supply region. In order to realize EEM, it is necessary to generate a stable flow of ultrapure water for supplying and removing fine particles on the surface of the workpiece. As a method of supplying the ultrafine particles onto the processing surface, a rotating sphere type processing head method and a nozzle type processing head method have been proposed.

  In the rotating sphere processing head method, as shown in Patent Documents 1 to 3, the processing sphere is pressed against the processing surface of the work piece while being rotated by the rotation driving means, so that the suspension flow near the processing surface. And maintaining a non-contact state with respect to the processing surface by the dynamic pressure, and scanning the sphere over the entire processing surface, to continue the spot processing marks formed in a minute region on the processing surface, A wide surface is precisely processed. Further, in the nozzle type processing head system, as shown in Patent Document 4, a workpiece and a high pressure nozzle are disposed at a predetermined interval in a processing tank mainly composed of ultrapure water, A high-speed shear flow of ultrapure water sprayed from a high-pressure nozzle is generated near the surface of the workpiece, and fine particles chemically reactive with the workpiece are supplied to the workpiece surface by the flow of ultrapure water. Then, fine particles chemically bonded to the workpiece are removed by a high-speed shear flow to remove atoms on the surface of the workpiece, and the processing proceeds.

  In the rotating sphere type machining head method, a physically stable elastohydrodynamic lubrication state is generated on the surface of the workpiece, a high-speed shear flow is generated on the surface, and the medium space wavelength region (1 μm to 100 μm) is processed. Can be smoothed. However, in order to create a uniform flow in the minute gap between the rotating sphere and the workpiece, it is necessary to enlarge the processing tank so that unnecessary flow generated around the rotating sphere does not affect it. In addition, since it was necessary to use a low elastic modulus polymer material such as polyurethane as the material of the rotating sphere, not only the dimensional accuracy and durability were inferior, but also organic contamination of the processing liquid occurred, and further swelling due to water The rotating body is deformed due to slipping, and the flow stability cannot be obtained. As another problem, the rotating sphere can only obtain asymmetric point-shaped processing marks, and the micro gap is about 1 μm. Therefore, it is easy to receive the disturbance by the influence of the coarse grain in a processing liquid, and that processing efficiency is low.

Therefore, in the newly developed nozzle type machining head system, the fine particles to be used are not limited, so it is possible to select various fine particles, and about 0.1 mm (diameter for circular nozzles, and for slit nozzles) Since the fine particles can be supplied only to the local region of (width), there is an advantage that it is suitable as a head for numerical control processing. Here, the surface shape of the fine particles is a very important factor from the planarization mechanism of EEM. Therefore, as a result of emphasizing the dispersibility and uniformity of fine particles in EEM, spherical fine particles were used. However, since spherical fine particles have a small contact area with the processed surface, they have the disadvantages that the amount of removal of each fine particle is small and the processing efficiency is low.
Japanese Patent Publication No. 2-25745 Japanese Patent Publication No. 7-16870 Japanese Patent Publication No. 6-44989 JP 2000-167770 A

Therefore, in view of the above-mentioned situation, the present invention intends to solve the problem that the nozzle type machining head system takes advantage of the fact that the fine particles to be used are not limited, optimizes the kind of fine particles and the machining parameters, and creates a Si wafer. It is an object of the present invention to provide a high-speed EEM processing method using agglomerated fine particles that realizes high-efficiency processing such as semiconductor wafers containing SiC, which is difficult to process, and optical element materials for EUVL such as Zerodur and ULE.

In order to solve the above-mentioned problem, the present invention arranges a workpiece and a nozzle type processing head at a predetermined interval in a liquid mainly composed of ultrapure water in a processing tank, and has a particle size of 1. A processing liquid in which aggregated fine particles, which are aggregates having an average diameter of 0.5 to 5 μm by aggregation of fine particles of ˜100 nm, are dispersed in ultrapure water is sprayed from the nozzle type processing head onto the surface of the workpiece. Generating a shearing flow of the machining fluid in the vicinity of the surface of the workpiece, and supplying aggregated fine particles chemically reactive with the workpiece to the workpiece surface by the flow of the machining fluid; The chemically bonded agglomerated fine particles were removed by a shear flow to remove atoms on the surface of the workpiece, and a high-speed EEM processing method using agglomerated fine particles for proceeding the processing was configured.
Here, the agglomerated fine particles are preferably aggregates obtained by agglomerating SiO ₂ fine particles by heating, from an industrial viewpoint.
And it is preferable that the density | concentration of the aggregation fine particle in the said processing liquid shall be 3-7 vol%. Furthermore, it is preferable that the spray angle of the processing liquid sprayed from the nozzle type processing head is in the range of 10 to 90 ° with respect to the surface of the workpiece.

The high-speed EEM processing method using the agglomerated fine particles of the present invention as described above is slightly inferior in surface smoothing performance to the surface processing on the Si (001) surface as compared with the conventional EEM using spherical fine particles. Double the processing speed can be achieved, and difficult-to-work materials such as SiC can be processed. In the EULD optical element materials such as Zerodur and ULE, the unit is about 150 to 300 times. It is possible to process at a removal rate. The present invention can realize high-efficiency and ultra-precise EEM by using it as pre-processing of EEM using conventional spherical fine particles.

  In the case of the nozzle type machining head method, fine particles are supplied to the surface of the workpiece by the flow discharged from the nozzle opening, and the adhered fine particles are removed by a shear flow. In order to realize EEM processing, various processing parameters relating to the processing head must be optimally selected. The processing parameters of EEM can be roughly divided into two. One is what kind of fine particles are selected, and the other is how to discharge a machining liquid in which fine particles are dispersed in ultrapure water to a workpiece.

  In order to bind and desorb the fine particles to / from the workpiece surface by chemical interaction, a shear flow having a certain speed gradient or more is required on the workpiece surface. In the present embodiment, a nozzle type processing head opened in a slit shape is used in order to efficiently supply fine particles to a specific position on the processing surface. FIG. 1 shows a conceptual diagram of a nozzle type machining head used in this embodiment, and Table 1 shows a range of machining fluid discharge parameters. Since the distance D between the nozzle type processing head 1 and the surface of the workpiece 2 can be set to about 1 mm, the posture control during the scanning of the nozzle type processing head is extremely easy, and mechanical rigidity, thermal deformation, and other causes The influence of the deformation of the device on machining is negligible. Further, in order to form a stable flow, the machining liquid is discharged in the liquid. Further, the spray angle of the processing liquid sprayed from the nozzle type processing head is set in a range of 10 to 90 ° with respect to the surface of the workpiece, and the high-speed shear flow is efficiently performed along the surface of the workpiece. In order to form the angle, it is preferable to set the injection angle to 30 to 45 °.

  Among the processing liquid discharge conditions, the average discharge flow velocity V is determined by the back pressure, the flow path length, and the opening gap, and the flow rate at that time is divided by the opening area. For the processing characteristics, the average discharge flow velocity V, the nozzle-workpiece distance D, and the discharge angle θ determine the kinetic energy of the fine particles when they reach the workpiece surface. It greatly affects the flow. Therefore, it is necessary to select the optimum discharge conditions in order to achieve the target processing amount and processing surface roughness.

  On the other hand, since EEM is a processing method that utilizes the reactivity of the surface of the fine particles, the fine particle material greatly affects the processing characteristics. The surface shape of the fine particles is an important parameter for the processing speed and the surface roughness at the atomic level in view of the EEM flattening mechanism.

In the present embodiment, SiO ₂ fine particles are employed as the fine particle material. FIG. 2 is a TEM image of the silica fine particles used for processing. The fine particles in FIG. 2 (a) are agglomerated fine particles newly adopted in the present invention and having an average diameter of about 2 μm, and the primary particles having a particle diameter of 0.1 μm or less, actually several tens of nanometers, are aggregated by heating. It has a characteristic that its surface area is very large. The fine particles in FIG. 2 (b) are uniform spherical silica fine particles having a particle diameter of 2 μm, which have been used in EEM in the past. By using these fine particles for processing, a smooth Si (001) surface can be produced at the atomic level. It is already known that this is possible. However, in the case of spherical fine particles, the contact area with the processed surface is very small, so that the removal processing amount of each fine particle is small. Therefore, it has been found that when the agglomerated fine particles of FIG. 2 (a) are used for processing, the contact area with the SiC surface can be widened and highly efficient surface processing can be performed. In addition, it can be presumed that, similarly to the aggregated fine particles, even if single non-spherical fine particles having a large number of minute irregularities on the surface are used, they can be similarly processed with high efficiency.

  Hereinafter, the superior processing efficiency of the present invention will be demonstrated by comparing the present processing using aggregated fine particles with the conventional processing using spherical fine particles. Planar processing on the Si (001) surface with both fine particles was performed, and the difference in processing speed and surface roughness was evaluated. Table 2 shows the processing conditions. Here, it should be noted that when aggregated fine particles are used, spherical fine particles are used even though the energy given to the fine particles is small due to the discharge conditions such as the discharge flow rate and the nozzle-workpiece distance. In comparison, although the surface smoothing performance is slightly inferior, a processing speed of 300 times or more is obtained. Actually, the Si surface before processing, and each Si surface after processing with aggregated fine particles and spherical fine particles were measured 16 times at different sites with a phase shift interference microscope, and the average value was obtained. The Si surface before processing was RMS: 0.146 nm and PV: 1.421 nm, but after processing with aggregated fine particles, RMS: 0.183 nm, PV: 2.14 nm, and spherical fine particles. After processing, RMS was 0.127 nm and PV was 1.365 nm. That is, after processing with spherical fine particles, both RMS and PV are improved from the surface before processing, but after processing with aggregated fine particles, both RMS and PV are slightly worse than the surface before processing. However, the significant improvement in processing speed due to the agglomerated fine particles should be noted.

  Next, the stability of the processing removal rate is an important evaluation item in consideration of practical use. In the case of an agglomerated fine particle, since it is an aggregate of fine fine particles, the fine particle collapses over time, and the processing removal rate can be reduced. Therefore, the long-term stability of the processing removal rate when the agglomerated fine particles were used was examined. The working fluid conditions are the same as in Table 3, but the working fluid concentration is 6.8 vol%. It was confirmed that the volumetric processing rate rapidly decreased to about 1/5 until 10 to 15 days from the working fluid preparation date, but thereafter maintained a constant value until 30 days. The volume processing speed was obtained by static processing of the Si (001) surface for 10 minutes, and was obtained from the shape of the unit processing trace, and the processing was always continued except for the sample exchange time. Further, as a result of examining the surface roughness after processing with a ZYGO interference microscope, it was confirmed that a surface roughness equivalent to the aforementioned value was continuously obtained in an image of 64 μm × 48 μm.

Next, the relationship between the fine particle concentration and the processing speed was examined. As a result of investigating the relationship between the volume machining speed and the machining fluid concentration after a sufficient time has passed since the machining fluid preparation date, the machining speed is dramatically improved by increasing the concentration, but the machining fluid concentration of about 7 vol% or more is increased. In this case, since the viscosity of the machining liquid increases, it is impossible to increase the concentration any more. Therefore, the optimum range of the working fluid concentration for performing high speed machining is set to 6-7 vol%. And when the Si (001) surface was processed under the optimized processing conditions using SiO ₂ aggregated fine particles in Table 3, a region of 5 mm × 5 mm was processed at a depth of about 100 nm in just 10 minutes. all right. It was also confirmed that the surface roughness of the processed surface was slightly improved from that before processing.

Next, Zerodur and ULE were processed with the above-mentioned SiO ₂ aggregated fine particles. Table 4 shows each processing speed ratio when the processing speed of the Si (001) surface is 1. Further, as a result of observing the processed surfaces of Zerodur and ULE with a phase shift interference microscope, it was found that the surface roughness was slightly improved, and a surface equivalent to processing with spherical fine particles was obtained. As a result, it was confirmed that if the aggregated fine particles are used, the EUVL optical material can be processed at a unit removal rate of about 150 to 300 times that of EEM using conventional spherical fine particles. In the high spatial frequency region, the improvement of roughness progresses, but in the case of the nozzle type machining head method, the improvement in the roughness of the intermediate frequency region is low in sensitivity, and in this region, flattening in the pre-processing process is necessary It is.

  In recent years, silicon carbide (SiC) single crystal, which is a wide band gap semiconductor, has attracted attention as a material for next-generation power semiconductor devices, and research and development of SiC devices exceeding the performance of Si has been conducted. In order to further improve the performance of SiC elements, technological developments such as reduction of crystal defects in SiC single crystals and enhancement of the quality of oxide film interfaces are desired. As a basic technology for developing these technologies, it is important to establish a surface treatment technology and a surface processing technology for obtaining a clean and flat SiC surface. When processing of the SiC surface by EEM is considered, since the SiC surface is highly hard and chemically stable, it is expected that processing is difficult with EEM, which is a processing method using a chemical reaction. For this reason, the processing is performed by employing aggregated fine particles having a large surface area so that the reaction area between the fine particles and the SiC surface becomes wide during processing.

  In this embodiment, a 4H—SiC (0001) Si surface is used as a workpiece, and as a pre-processing treatment, the treatment is immersed for 10 minutes in a 50% HF solution for the purpose of removing the natural oxide film on the surface. Went. Table 5 shows the processing conditions in this processing.

  The result of processing the SiC surface using the agglomerated fine particles of FIG. Fig. 3 shows the differential image and cross-sectional profile obtained by measuring the SiC surface before processing with a phase shift interference microscope, and Fig. 4 shows the differential image and cross-sectional profile obtained by measuring the SiC surface after processing with a phase shift interference microscope. The microscope used for the measurement is NewView200CHR manufactured by ZYGO, and the measurement area is 64 μm × 48 μm.

  The processing amount at this time was 50 nm, and the processing speed was 15.6 nm / h. By adjusting the concentration of the processing liquid and the discharge speed of the fine particles, the current processing speed can be increased several times, so that the processing can be performed with higher efficiency. 3 (a) and 3 (b), it can be seen that there are innumerable polishing marks (scratches) on the pre-processed surface, and the surface roughness is composed of irregularities with a spatial wavelength of 1 to 10 μm. Then, from FIGS. 4A and 4B, the scratch existing on the pre-processed surface is completely removed by processing, the roughness of the spatial wavelength of 5 to 10 μm is removed, and a smooth surface is formed. I understand that. FIG. 5 shows the result of performing spatial frequency analysis PSD (Power Spectral Density) based on the measurement data. Smoothing of the surface can be confirmed in a spatial wavelength region of 1 μm to 10 μm.

  Atomic force microscope: The results of measuring the SiC surface before and after processing using AFM (Atomic Force Microscopy) will be described. The AFM used for the measurement was SPA400 + SPI3800N manufactured by SII Nano Technology, and was measured in the DFM mode under atmospheric pressure. 6A, 6B, and 6C are AFM images obtained by measuring the surface before and after processing in a 2 μm × 2 μm region. The micro-roughness of the pre-processed surface is Ra: 0.605 nm [PV: 6.169 nm, RMS: 0.767 nm], whereas the micro-roughness of the processed surface when using the agglomerated fine particles of FIG. The microroughness was Ra: 0.074 nm [PV: 1.108 nm, RMS: 0.093 nm]. From the AFM image, scratches seen on the pre-processed surface are completely removed after processing, and the surface is smoothed, so that a surface with high efficiency and sufficient surface accuracy is formed. Recognize. When the surface of FIG. 6B is processed using the spherical fine particles of FIG. 2B, the microroughness of the surface becomes Ra: 0.042 nm [PV: 0.475 nm, RMS: 0.053 nm]. It has also been confirmed that a smooth surface can be obtained at the atomic level.

  Next, FIG. 7 shows the results of spatial frequency analysis from data obtained by measuring 500 nm, 1 μm, 2 μm, and 4 μm squares of the pre-processed surface and the processed surface. From the spatial frequency analysis, an improvement of about 1/10 to 1/100 is observed in all regions with a spatial wavelength of 1 μm to 0.01 μm. This indicates that the average amplitude of the roughness is reduced from 1/3 to 1/10 or less in each spatial wavelength region. In addition, uniform smoothing in the entire spatial wavelength region can be confirmed with processing without any difference in the spectral structure of the microroughness.

  The crystallinity of the surface before and after processing was evaluated using LEED (Low Energy Electron Diffraction). Even after immersing SiC in a 50% HF solution for 10 minutes before processing and observing the surface with LEED, a diffraction pattern was not obtained. From this, it can be seen that there are at least about 1 to 2 layers of damaged layers due to polishing during wafer fabrication on the SiC surface subjected to high concentration HF cleaning. Next, after processing the SiC surface cleaned with high concentration HF using the agglomerated fine particles of FIG. 2A, the surface was observed with LEED. As a result, the processed surface had a bulk crystal structure of 1 × 1. The diffraction pattern was confirmed. From this, it was demonstrated by direct observation of diffraction spots by LEED that the damage layer resulting from the polishing process during wafer fabrication was removed by processing, and a surface free of crystallographic atomic arrangement disturbance was obtained. .

  As described above, it was confirmed that by processing SiC with the aggregated fine particles, the scratch existing on the pre-processed surface was completely removed by the process, and a surface having no disorder in the crystal structure was formed. Further, the microroughness of the processed surface was confirmed by RMS to be 0.1 nm or less (2 μm × 2 μm region) by AFM measurement, and it was shown that sufficient surface accuracy can be obtained by EEM.

It is a conceptual diagram of the process by a nozzle type processing head. The TEM image of microparticles | fine-particles is shown, (a) shows an aggregation microparticle, (b) shows a spherical microparticle. The result of having measured the SiC surface before a process with a phase shift interference microscope is shown, (a) is a differential image, (b) is a cross-sectional profile. The result of having measured the SiC surface after a process with a phase shift interference microscope is shown, (a) is a differential image, (b) is a cross-sectional profile. It is a graph which shows the result of the spatial frequency analysis (PSD) of the SiC surface after a process. It is an AFM image of the surface before and after processing, (a) is the surface before processing, (b) is the surface after processing with aggregated particles, (c) is the surface after processing with spherical particles on the surface of (b), respectively. Show. It is a graph which shows the result of having performed the spatial frequency analysis from the data which measured 500nm, 1 micrometer, 2 micrometers, and 4 micrometers square of the pre-processed surface and the processed surface (processed surface).

Explanation of symbols

1 Nozzle type processing head 2 Work piece

Claims (4)

An object to be processed and a nozzle type processing head are arranged at a predetermined interval in a liquid mainly composed of ultrapure water in a processing tank, and fine particles having a particle diameter of 1 to 100 nm are aggregated to have an average diameter of 0. A machining liquid in which aggregated fine particles that are aggregates of 5 to 5 μm are dispersed in ultrapure water is sprayed from the nozzle type machining head onto the surface of the workpiece, and the machining liquid is placed near the surface of the workpiece. In addition to generating a shear flow, agglomerated fine particles that are chemically reactive with the workpiece are supplied to the workpiece surface by the flow of the machining fluid, and the aggregated fine particles chemically bonded to the workpiece are removed by the shear flow. A high-speed EEM processing method using agglomerated fine particles, characterized in that atoms on the surface of a workpiece are removed and processing proceeds. The aggregating fine particles, fast EEM working method using aggregation particles of claim 1, wherein an aggregate obtained by aggregating by heating the SiO ₂ particles. The high-speed EEM processing method using the aggregated fine particles according to claim 1 or 2, wherein the concentration of the aggregated fine particles in the processing liquid is 3 to 7 vol%. The high-speed EEM processing method using agglomerated fine particles according to any one of claims 1 to 3 , wherein an injection angle of the processing liquid injected from the nozzle type processing head is in a range of 10 to 90 ° with respect to a surface of the workpiece.

Images