基于PSD评价及伪随机轨迹的单晶碳化硅表面粗糙度研究

Study on surface roughness of monocrystalline silicon carbide based on PSD evaluation and pseudo-random tool path

随着新能源、特高压需求爆发,以单晶碳化硅为代表的第三代半导体技术近几年得到了飞速发展,大口径单晶碳化硅材料制备已经成为现实,相比于目前已成熟应用的RB-SiC材料,单晶碳化硅不需要通过CVD或PVD改性就可以获得1 nm甚至更优的表面粗糙度,在光学元件领域的应用具有广阔前景,但同时加工难度高是亟待解决的问题。为了解决单晶碳化硅材料在光学加工过程中的粗糙度问题,提出了一种基于PSD评价及熵增理论的伪随机轨迹加工改善粗糙度的方法。相较于传统单一的Ra值评价方法,通过引入PSD曲线丰富了粗糙度评价的维度;利用对熵增理论的分析,从理论上讨论了确定性抛光轨迹和伪随机轨迹对粗糙度尺度下累计误差影响的区别。通过对6 in(1 in=2.54 cm)单晶碳化硅进行多轮抛光实验,结果表明:在相同初始粗糙度情况下,确定性轨迹与伪随机轨迹虽均得到了Ra约1 nm的粗糙度值,但PSD曲线可以明显看出确定性轨迹出现了尖峰,而伪随机轨迹则更为平滑。验证了特定采样区间下的PSD曲线作为粗糙度评价手段的有效性,同时论证了伪随机轨迹相较于确定性轨迹在单晶碳化硅材料抛光上的优势。

Objective As a novel material leading the third-generation semiconductor technology revolution,monocrystalline silicon carbide has a very excellent prospect in the application of semiconductor field.And because of its high thermal conductivity,high elastic modulus,and high temperature stability,it is a highly competitive material in traditional imaging optics,high-power laser optics and other fields.Under some circumstances,such as application scenarios as high-power laser optics or EUV optics,the surface roughness Ra needs to be less than 1 nm or even much lower.Obviously,such high-performance specifications necessitate much more precise optical manufacturing for these types of optical applications.In traditional optical manufacturing,the technology of CCOS(Computer Controlled Optical Surfacing)is a commonly utilized manufacturing procedure in the whole process of optical manufacturing.Although some researchers at home and abroad have conducted detailed investigations on the influence of CCOS processing on MSF(Middle Spatial Frequency)errors for optical surfaces,there is a lack of research on the influence of CCOS processing on HSF(High Spatial Frequency)errors for optical surfaces.The intensity of the HSF errors of optical surfaces directly determines the surface roughness.Therefore,it is necessary to find a proper solution to how to evaluate the HSF errors for optical surfaces and how to reduce the HSF errors,which determines the surface roughness,when the overall HSF errors and surface roughness don’t meet expectations.Methods PSD(Power Spectral Density)is the most commonly utilized indicator to evaluate the distribution of intensities of different frequencies for a certain signal.The sudden-peak form on a PSD curve indicates a sudden increase of the intensity of certain frequency band for the said signal,and at the same time,the peak form on a PSD curve will directly lead to the increase of the surface roughness.Inspired by the principle of the increase of entropy,experiments were conducted on two monocrystalline SiC flat surfaces with similar initial surface roughness distributions.One surface was processed with a pseudo-random tool path which was based on the Gilbert space-filling curve(Fig.2),while the other was processed with a conventional deterministic rasterized trajectory(Fig.1(a)).Finally,the surface roughness distributions and PSD curves of the two surfaces after the experiment were analyzed.Results and Discussions Through the comparison of the experiment of the two surfaces,it can be seen that both two monocrystalline SiC surfaces have an approximate initial roughness Ra=7 nm(Fig.4)and then get experimented with 10 sets of 40 minutes'polishing.And after the polishing process is completed for both surfaces,the PSD curve of the surface processed with a deterministic rasterized tool path contains a sudden peak nearby frequency domain 0.01μm−1(Fig.6),whereas the PSD curve of the surface processed with pseudo-random tool path appears to be much smoother(Fig.8).In the meantime,the test results show that the surface processed with the pseudo-random tool path has lower roughness,which in turn indicates that the surface quality is higher after processing with the pseudo-random tool path.Conclusions PSD is one of the most versatile indicators when it comes to signal analysis.And enlightened by law of the increase of entropy in thermodynamics,and all other things being equal,the deterministic rasterized tool path is simply replaced with a pseudo-random one.And the final testing results are significantly different.That is,when making use of CCOS technology to process monocrystalline SiC,the pseudo-random tool path can be utilized to reduce the relative intensity of HSF errors of a certain surface.And it proves that the pseudo-random tool path in the CCOS processing stage has a great inhibiting effect on the HSF errors of optical surfaces and therefore facilitates lower surface roughness and better surface quality.