Research on Three-dimensional Shape Detection of Non-parallel Light Interferometric Illumination Microscope
The rapid development of machinery manufacturing and electronics industry has put forward higher requirements for the detection technology of microscopic morphology. At present, the detection of three-dimensional shape can be divided into two categories: contact and non-contact. The contact method mainly refers to the stylus method. Its principle is to convert the small displacement in the vertical direction of the stylus into an electrical signal and amplify it, so as to obtain the three-dimensional shape distribution of the detection surface. Non-contact detection methods mainly include beam focusing method, structured light projection method and interferometry. The beam focusing method uses the focused light spot as an optical probe to scan the detection surface to obtain three-dimensional data. This method can perform three-dimensional detection of complex contours, but the measurement speed is slow. Interferometry and structured light projection methods detect surface contours by solving fringe deformations, where interferometry uses the principle of fiber coherence or parallel beam coherence, including laser interferometry and white light scanning interferometry. When using optical fiber coherence, it needs to cooperate with the objective lens with a large working distance, which limits the magnification of the objective lens. White light scanning interferometry uses wide-spectrum white light as the illumination source and uses the principle of parallel beam coherence. The error of a single measurement is within 20nm. Information such as contrast and light intensity determine the absolute depth of the surface being measured. The structured light projection method avoids the use of scanning devices and has the fastest reconstruction speed. However, when there is an angle between the projection plane and the stage plane, the fringe period needs to be corrected, so this method is not suitable for submicron-level precision morphology. Measurement In this paper, the advantages of the structured light projection method and the parallel light interferometry method are combined, the light beam is diffracted by the spatial light modulator, and two diffraction orders with close light intensity are used to interfere to generate fringes. Adjusts the fringe phase. Since the use of scanning devices and reference planes is avoided, the proposed method does not require the use of interference objectives and has no limitation on the numerical aperture of the used objectives, the reconstruction process is fast, and higher lateral resolution can be achieved. In addition, since the fringes are generated by light beam interference, the phase is distributed linearly with the pixel coordinates, and there is no phenomenon of periodic fringe changes in the projection method. Finally, this paper uses the roughness comparison module with Ra of 100nm as the tested sample to conduct experiments. The four-step phase shift method is used to obtain the three-dimensional point cloud on the surface of the tested sample. The true relative height between points.
Experimental light path
It is the light path diagram of the non-parallel light interference illumination microscope proposed in this paper. The laser beam enters the beam splitting prism of the microscope through the beam expander L3, the spatial light modulator and the focusing lens L2, forming the illumination light path of the microscope system. The spatial light modulator can modulate the amplitude of the incident light according to the uploaded image. When the uploaded image is a fringe, its function is equivalent to a reflective grating, adjusting the deflection of the spatial light modulator so that two beams of diffracted light with similar light intensities enter The dichroic prism, after being focused by the microscope objective lens, interferes with the surface of the measured sample to form interference fringes.
The spatial light modulator is the core device of the system, and the period and phase of the fringe can be precisely modulated by changing the uploaded fringe pattern during the experiment. Usually, in order to improve the lateral accuracy of the 3D reconstruction point cloud, it is necessary to adjust the fringe period to make it close to the lateral resolution of the microscope. At this time, the maximum interference angle of the two beams can be calculated from the numerical aperture NA of the objective lens.
According to the parameters of the microscope objective lens used in the system (100 , NA=0.8), the maximum interference angle of the double beams is 106°, and the system resolution calculated by Rayleigh criterion is 406nm. In the experiment, the minimum fringe period that can be adjusted is 452nm, indicating that within a fringe period, there is a corresponding relationship between the phase shift and the height of at least one pixel point, that is, the lateral accuracy of the reconstructed point cloud is 452nm, which is close to the imaging resolution of the system. Due to the small fringe period, the deformation of the fringe is more sensitive than that of the large-period fringe, so it has higher axial precision. In terms of phase adjustment, the white light scanning interferometry needs to move the interference objective lens in the axial direction with the help of a piezoelectric device, and then fit the phase by calibrating the zero optical path difference on each scanning image, so there is a certain error in the phase value. In our system, the phase adjustment is realized by controlling the pixels on the spatial light modulator without scanning device, so it has higher phase adjustment accuracy. On this basis, the phase shift method is used to calculate the phase modulation value of each point on the image. The 3D reconstruction results with high lateral resolution can be obtained under the faster reconstruction algorithm.
reconstruction algorithm
In this paper, the four-step phase shift method is used to reconstruct the three-dimensional contour of the measured sample, which is divided into three steps: image preprocessing, phase modulation image extraction, and noise point filtering. The following will take the roughness comparison module with Ra=100nm as a sample to explain the algorithm used in each step. 2.1 Image preprocessing Since the imaging system uses laser illumination, the influence of laser speckle on the interference pattern is unavoidable. In the process of preprocessing the interference fringes, this paper uses an elliptical low-pass filter, so that the filtering radius along the fringe direction in the frequency domain of the image is twice that of the vertical direction of the fringes. The fringe pattern appears as two center-symmetrical bright spots in the frequency domain, and the direction of the connecting line between the two points is perpendicular to the direction of the fringe, and the direction of the connecting line is set as the long axis of the ellipse. Since the fringe period is close to the image resolution, the major axis is set to be twice the distance of 2 bright spots, and the minor axis is equal to the distance of 2 points. Such a design can reduce the impact of speckle noise in the relative phase solution on the one hand, and on the other hand, can avoid the modulation information in the interference pattern from being filtered out as much as possible. Shown are the processing results under the isotropic and anisotropic filtering methods, the comparison can reduce the noise of the image along the fringe direction, while retaining the deformation of the fringe.
