An Exploration into Damage Repair and Manufacturing Technology of Photomask Glass Substrates

Abstract

There are strict requirements on the surface-shape accuracy, cracks, scratches, and subsurface damage of photomask glass substrates with the advancement of photo-etching technology. In this study, photomask quartz glass substrates were etched with the hydrofluoric (HF) acid etching method after chemical mechanical polishing (CMP), so that the polishing hydrolysis layer could be completely removed. In addition, the surface scratches caused by CMP were observed with a surface quality detection device and a white light interferometer. Second, the high-efficiency array-type magnetorheological polishing technology was employed to eliminate the surface-shape error on the surface of quartz photomask glass substrates. Finally, the removal function rotation angle based magnetorheological polishing technology was utilized to evenly remove the mid-frequency ripples and scratches caused in the preceding technological process. The experimental results demonstrated that these technologies can be used to realize rapid damage removal and high-quality processing of photomask glass substrates.

1. Introduction

The integrated circuit (IC) is the core and forerunner of modern high technologies. It shall be highlighted that photo-etching technology plays a crucial role in making ICs [1]. In recent years, the integration of semiconductor technology has developed rapidly, and the photo-etching technology for manufacturing semiconductor devices has also been improved, which proposes increasing higher requirements for the accuracy of photomask glass substrates [2]. The surface-shape accuracy, local tiny bubbles, cracks, scratches, subsurface damage, and other defects of photomask glass substrates would deteriorate their imaging effects and cause serious errors in the pattern layout. The fabrication requirements for the photomasks are shown in

Table 1. Photomask manufacturing requirements.

Item	Value
Surface shape errors	Less than 1 wave
Surface roughness (RMS)	Less than 2 nm
Cracks, ripples, pitting	Not allowed
Scratches	Not allowed for sizes larger than 1 µm
	0.031/cm2 allowed for sizes less than 1 µm

. Hence, it is necessary to propose higher requirements for the quality of photomask glass substrates, including high ultraviolet transmittance, high optical uniformity, high intrinsic quality, and high surface processing quality [3,4].

Photomask glass substrates can be mainly divided into quartz glass substrates and soda glass substrates according to different substrate materials [5]. With quartz glass as substrate materials, the quartz glass substrate has a high optical transmittance and low thermal expansion rate. Compared with the soda glass substrate, the quartz glass substrate is smoother and wear-resistant, with a long service life. Thus, it is mainly used for high-precision photomask substrates [6]. The processing quality of quartz glass substrate directly determines the quality of photo-etching technique [7]. Due to some limitations in equipment or technology during manufacturing, the pattern on the photomask substrate cannot be completely consistent with the design image. Further, the manufacturing defects and errors on the photomask substrate will be introduced into the chip manufacturing by photo-etching technology in the subsequent silicon wafer manufacturing process. Therefore, the surface and subsurface quality of the photomask substrate will directly affect the yield rate and stability of chip.

At present, chemical mechanical polishing (CMP) is the main method for processing photomask quartz glass substrates. In CMP, free abrasive particles and grinding discs are utilized to remove materials via a chemical and mechanical composite method, through which a surface with low surface roughness can be obtained [8]. However, it is difficult to control the free polishing particles in the polishing process, which would cause failure to realize higher surface-shape accuracy of the machined component surface. In addition, the mechanical removal method would leave some subsurface damage and surface scratches. Many scholars have carried out research on the manufacture of photomask glass substrates and nondestructive processing of quartz materials. ZHAO et al. proposed a full-aperture polishing method based on the application requirements of photomask substrates. They improved the equipment according to the technological requirements, which significantly reduced the surface-shape error of the polishing pad [9]. CHOU et al. conducted an exploration into the chemical mechanical grinding method of quartz glass. The machining of large-diameter workpieces with high surface quality and high shape accuracy can be realized by actively enhancing the chemical reaction in the grinding process, weakening the bonding potential energy of quartz glass, and eliminating the surface damage caused by brittle removal [8]. Wang et al. revealed that the removal efficiency of quartz glass increases obviously and the surface roughness decreases significantly as the mass fraction of abrasive increases [10]. Li et al. investigated the influence of pH on the polishing performance of quartz glass, and they analyzed and characterized the silica sol used in CMP by dynamic light scattering and Zeta potentials. Additionally, they improved the surface quality and removal rate of quartz glass by adding guanidine carbonate as a co-solvent to the polishing solution [11]. Zhu et al. found that the polishing efficiency of the polycrystalline diamond polishing pad on quartz glass was higher than that of the single crystal diamond polishing pad, and the polycrystalline diamond polishing pad would cause less subsurface damage [12]. Liu et al. explored the relationship between dressing parameters of the polishing pad and surface defects of the workpiece after CMP. They found that excessive dressing of polishing pads would cause more polishing fragments on the surface, which would cause damage to the workpiece [13]. However, the focus of most previous studies is placed on improving the technological parameters, polishing particles, and polishing discs of CMP to reduce the subsurface damage and improve the efficiency and surface roughness. There are fewer studies on the approaches to detect and repair the damage and sub-surface damage layer caused by CMP. In this study, photomask quartz glass substrates were etched with the hydrofluoric (HF) acid etching method after chemical mechanical polishing (CMP), so that the polishing hydrolysis layer could be completely removed. In addition, the surface scratches caused by CMP were observed with a surface quality detection device and a white light interferometer. Second, the high-efficiency array-type magnetorheological polishing technology was employed to eliminate the surface-shape error on the surface of quartz photomask glass substrates. Then, the removal function rotation angle based magnetorheological polishing technology was utilized to evenly remove the mid-frequency ripples and scratches caused in the preceding technological process. Finally, much surface damage is removed and there is no mid-frequency ripple on the surface. In this method, a new technology route of photomask quartz glass substrate was proposed, which could realize rapid damage removal and high-quality processing of photomask glass substrates.

2. Equipment

In this experiment, a three-dimensional white light scanning interferometer (Zygo, as shown in Figure 1) was used to measure the surface roughness of elements. In terms of this non-contact instrument for measuring the surface roughness of optical elements, the field of view was 0.07–9.3 mm, the longitudinal resolution was 0.1 nm, the transverse resolution was 0.36–9.50 μm, and the measurement repeatability RMS was less than 0.01 nm [14].

An ultra-smooth surface laser scattering quality detection device (ZC Optoelectronic Technologies Co., Ltd., ZYGO, Middlefield, CT, the United States of America, as shown in Figure 2) was used to characterize the distribution of damage points on the surface of damaged samples. This device can be employed to count and characterize the number of damage points of various sizes on the surface of samples [15].

An array-type magnetorheological polishing device was independently developed by our laboratory, as shown in Figure 3. Under the action of the permanent magnet array arrangement principle, the original single magnetorheological polishing ribbon became double ribbons, which doubled the removal efficiency and saved processing time.

3. Mechanism of Magnetorheological Polishing Related to Subsurface Damage Removal and Mid-Frequency Ripple Error Control

3.1. CMP Subsurface Damage Model

Subsubsection

The residual subsurface damage of the surface layer of chemically polished quartz glass mainly occurred at the surface hydrolysis layer and the subsurface defect layer. The surface hydrolysis layer included the polishing impurities from the shallow surface flow layer, plastic scratches, and the polishing process, whose concentration decreased with the increase in the depth. The subsurface defect layer may include subsurface cracks, brittle scratches, and residual stresses left in the polishing process, as well as plastic scratches introduced by the polishing process [16]. The residual subsurface cracks, brittle scratches, and residual stresses in the polishing process can be eliminated gradually by optimizing the polishing technology and prolonging the polishing time. Figure 4 presents a quartz glass polishing subsurface damage model.

The microscopic removal mechanism of polishing materials can be explained according to the established polishing subsurface damage model. Firstly, the hydrolysis of quartz glass broke its inherent network structure, and the silicic acid gel was formed on the surface, which eventually induced a decrease in the surface density, especially the decrease in the surface hardness. Subsequently, the large-scale polishing particles embedded in the polishing film would remove the material on the soft surface hydrolysis layer by the two-body wear manner, and produce plastic scratches. The polishing particles bearing a large normal load can penetrate the hydrolysis layer to wear the matrix material. However, small-scale polishing particles rolled and moved between the polishing film and the quartz glass surface, and they can remove the material or be embedded in the hydrolysis layer by adhesion. With the progress of polishing, the temperature rise caused by the friction between the polished film and the quartz glass surface, the instantaneous high temperature of microprotrusion on the friction surface, the residual structural defects during polishing, and the stress exerted during polishing would promote the hydrolysis of quartz glass and the softening of the hydrolyzed layer. Finally, the friction force between the polishing film and the quartz glass surface promoted the plastic flow of the shallow surface flow layer. This plastic flow can quickly cover the grinding damage exposed in the polishing process and the plastic scratch introduced by the polishing, thus ensuring the smoothness of the surface.

3.2. Mechanism of Magnetorheological Polishing Related to Subsurface Damage Removal

In magnetorheological polishing technology, the rheological properties of the magnetorheological polishing fluid in a magnetic field can be utilized to polish the workpiece, with the relevant principle shown in Figure 5 [17]. Before the magnetorheological fluid entered the polishing zone, the magnetic moments of magnetically sensitive particles were randomly arranged since there was no external magnetic field. The magnetorheological fluid would not show magnetism to the outside. At this moment, the magnetic sensitive particles and polishing abrasive particles were uniformly dispersed in the base carrier liquid. After the magnetorheological fluid was brought into the polishing area by the polishing wheel, the magnetically sensitive particles were magnetized to generate the dipole moment under the action of the high-intensity gradient magnetic field. In order to minimize the energy, the magnetically sensitive particles were connected into chains to form a “flexible polishing film”. Under the magnetic buoyancy of the magnetic field, magnetorheological fluid was precipitated from non-magnetic polishing particles, which were “embedded” on the surface of the “flexible polishing film”. For the reason that the magnetorheological fluid had the properties of Bingham medium in the magnetic field, the rigid “core” formed by Bingham medium would generate a wedge-shaped zone in the polishing area after the optical element was pressed into a magnetorheological forging belt. The polishing abrasive particles would adhere to the surface of the optical element under the action of hydrodynamic pressure and pass through the wedge-shaped zone at a high speed, thus achieving higher polishing efficiency.

During the CMP of quartz glass, the force exerted on the polishing disc was transmitted to the surface of optical elements through polishing particles. The positive pressure of a single polishing particle on the surface of optical elements was about 10⁻³ N [18]. Meanwhile, during the magnetorheological polishing of quartz glass, the positive pressure exerted by polishing particles on the surface of optical elements was composed of three parts, including gravity (G), magnetic buoyancy (Fz), and normal component of hydrodynamic pressure (Fwn). Among them, gravity (G) can be ignored, as shown in Figure 6. According to the parameters of magnetorheological polishing and the measured hydrodynamic pressure in the polishing area, it can be found that the positive pressure exerted by a single polishing particle on the surface of optical elements in magnetorheological polishing was about 10⁻⁷ N, which was far less than that exerted by polishing particles in traditional polishing. Therefore, magnetorheological polishing can effectively remove subsurface plastic scratches introduced by traditional polishing.

3.3. Generation and Control Mechanism of Mid-Frequency Ripple Errors in Magnetorheological Polishing

In actual machining, the magnetorheological removal function moved along a certain trajectory with a certain feed step. However, the removal function had a certain size, which can remove materials outside the trajectory. Hence, it was inevitable to generate periodic residual errors under convolution. The speed achievement based on a regular scanning path was not a continuous process. For the raster scanning path, the speed in the scanning direction can be changed continuously, but the speed during line changes was discontinuous. Due to the discrete data and discontinuous speed, residual errors in the non-resident area of the workpiece were distributed on the surface of the machined workpiece with a certain frequency. This caused mid-frequency ripple errors on the machined surface, as shown in Figure 7.

As shown in the above figure, the mid-frequency ripple height is mainly determined by the removal function characteristics and residence time. Therefore, it is necessary to control the removal function characteristics and residence time to control the mid-frequency ripples caused by the convolution effect. Reducing the raster scanning step and the specific removal function rotation angle can reduce the mid-frequency ripple height by two orders of magnitude. At this ripple scale, the residual ripple would be submerged in the noise, thus achieving the control of the mid-frequency ripple errors indicated in the component [19].

4. Damage Detection and Repair of Photomask Quartz Glass Substrates

4.1. Damage Detection of Photomask Quartz Glass Substrates

In this experiment, the photomask quartz glass substrates were three-inch glass substrates purchased from Shenzhen Precise Trading Co., Ltd., (Shenzhen, China) as shown in Figure 8. These photomask quartz glass substrates were treated by CMP at first. An ultra-smooth surface laser scattering quality detection device was used to measure the surface damage and defects, as shown in Figure 9a. It can be seen in the dark field test picture that the surface was basically free of defects, due to the fact that the surface hydrolysis layer completely covered the damage. Subsequently, the surface of photomask quartz glass substrates was fully etched by the hydrofluoric (HF) acid etching method, so that the surface hydrolysis layer can be completely dissolved. Then, the dark field test was performed on the surface, as shown in Figure 9b. It can be seen that there were many random scratches on the surface, which would seriously affect the imaging quality of photomask glass substrates. Therefore, it was required to remove scratches and damage on the surface after CMP.

In addition, a white light interferometer was adopted to test the surface roughness before and after pickling and magnetorheological polishing, as shown in Figure 10. Figure 10a presents the surface roughness before pickling, Figure 10b presents the surface roughness of the non-damaged area after pickling, Figure 10c presents the surface roughness after magnetorheological uniform scanning, and Figure 10d presents the surface roughness of the scratched area after pickling. The scratch morphology after removing the hydrolysis layer can be seen from the results of Figure 10d. The scratch depth was about 0.05 μm, which would seriously affect the imaging quality of photomask glass substrates. Hence, it is necessary to introduce magnetorheological processing technology to remove surface damage.

4.2. Repair and Manufacturing Process of Photomask Quartz Glass Substrates Based on Array-Type Magnetorheological Polishing Technology

Firstly, a vertical interferometer was employed to measure the surface shape of the photomask glass substrate, as shown in Figure 11a. Then, an array-type magnetorheological device was utilized to modify the surface shape. The removal function for surface shape modification is shown in Figure 12. The removal efficiency was approximately doubled compared with the original magnetorheological polishing efficiency based on a single ribbon. The repaired surface shape is shown in Figure 11b. It can be seen that the array-type magnetorheological repair can quickly and effectively reduce the surface-shape error of photomask glass substrates. However, there would be mid-frequency ripple errors caused by magnetorheological polishing, as shown in Figure 13 (the sampling frequency band was 0.8–1.6 mm⁻¹). Thus, it was required to use the removal function rotation angle to uniformly remove the sub-surface damage and the mid-frequency ripple error caused by surface shape modification.

In addition, a magnetorheological polishing method based on the removal function rotation angle was adopted in the study. The simulation analysis results showed that the PV value of surface mid-frequency ripples decreased by two orders of magnitude [18,19] when the removal function rotated 24°. Then, the array-type magnetorheological rotation angle was adopted to perform uniform scanning removal on the modified surface of photomask substrates with mid-frequency ripples. The obtained ripple errors in the mid-frequency band are shown in Figure 14 (the sampling frequency band is 0.8–1.6 mm⁻¹). Finally, the photomask substrates with scratches on their surface after pickling were uniformly scanned for removal of 20 nm, 50 nm, and 100 nm based on the array-type magnetorheological rotation angle method. The dark field test pictures on the surface are shown in Figure 15a–c, respectively. The evolution of scratches and surface roughness measured by a white light interferometer are shown in Figure 16a–c.

5. Discussion

Chemical mechanical polishing (CMP) is usually used in the manufacture of photomask quartz glass substrates. However, this technology is prone to induce random scratches on the surface, which would seriously affect the imaging quality of photomask substrates. However, after the surface after CMP was measured by an ultra-smooth surface laser scattering quality detection device, it can be found that there were basically no scratches on the surface. This may be explained that a hydrolyzed layer formed on the surface to cover these scratches and defects. Therefore, the surface hydrolysis layer was fully pickled by HF acid to obtain a surface with many random scratches, as shown in Figure 9. Moreover, the pickling technology could dissolve the hydrolysis layer and expose the subsurface, which significantly increased the surface roughness, as shown in Figure 10. Therefore, it was required to remove scratches and damage on the surface after CMP. Based on that, magnetorheological polishing technology was introduced to perform non-destructive machining of the photomask substrate surface.

In order to explore the feasibility of magnetorheological polishing technology, the surface roughness change of photomask glass substrates was measured after magnetorheological polishing. As shown in Figure 10, the surface roughness of photomask glass substrates decreased, instead of getting deteriorated. Hence, magnetorheological polishing can meet the surface roughness requirements of photomask substrates. In addition, the depth of surface scratches was also measured after pickling, with the depth ranging from 0 nm to 80 nm. Hence, the surface scratches can be completely removed by magnetorheological uniform scanning repair. Meanwhile, due to the large size of photomask glass substrates and higher demands, it is necessary to improve the removal efficiency of magnetorheological polishing. Thus, an array-type magnetorheological polishing device newly developed by our laboratory was used for processing. Due to the fact that the magnetorheological uniform scanning would produce surface mid-frequency ripple errors, as shown in Figure 13, the newly developed magnetorheological rotation angle polishing method can be used to suppress these errors, thus obtaining the photomask glass substrates without ripples on the surface.

Firstly, the array-type magnetorheological polishing technology was adopted to repair the initial surface. As shown in Figure 12, the PV value of the surface shape increased from 0.657 wavelengths to 0.336 wavelengths. This demonstrated that the array-type magnetorheological polishing technology was effective in repairing the surface shape of photomask substrates, and this technology could improve the quality of photomask substrates. Secondly, due to the existence of mid-frequency ripple errors on the repaired surface, the uniform removal of 100 nm on the surface was performed based on the removal function rotation angle, and the mid-frequency ripple errors on the surface basically disappeared, as shown in Figure 14. Meanwhile, after the uniform removal of 100 nm on the surface was performed, the scratches on the surface were basically eliminated when observing the evolution of the dark field and scratches on the surface treated by uniform removal based on the magnetorheological rotation angle, thus meeting the manufacturing requirements of photomask glass substrates.

6. Conclusions

In this study, the feasibility of the high-efficiency array-type magnetorheological polishing technology combined with the removal function rotation angle technology in manufacturing and repairing photomask glass substrates was explored. Moreover, a new technological route related to photomask quartz glass substrates was proposed. In addition, a surface quality detection device was used to detect the surface defects, and the HF acid etching method was used to etch the photomask quartz glass substrates after CMP. Based on that, the polished hydrolysis layer can be eliminated, and much damage and many defects were found on the subsurface. Moreover, the surface was modified by the array-type magnetorheological device developed by our laboratory, and the experimental results of surface-shape errors demonstrated the effectiveness of surface shape modification. Furthermore, the removal function rotation angle was used to uniformly remove the subsurface damage and the surface mid-frequency ripple errors caused by surface shape modification. A surface quality detection device and white light interferometer were employed to observe the evolution of the surface scratches caused by CMP. This confirmed the effectiveness of the removal function rotation angle in removing the damage and the mid-frequency ripple errors. Finally, much surface damage can be eliminated and there was no mid-frequency ripple error on the surface. These techniques can meet the performance requirements of photomask glass substrates and realize the rapid damage removal and high-quality manufacture of photomask glass substrates.