CN-122020322-A - Ultra-smooth optical element surface characteristic parameter prediction method

CN122020322ACN 122020322 ACN122020322 ACN 122020322ACN-122020322-A

Abstract

The invention discloses an ultra-smooth optical element surface characteristic parameter prediction method, which relates to the technical field of optical element measurement and comprises the steps of determining the value range and sampling precision of each surface parameter, constructing a scattering rate distribution database, preprocessing, constructing a classification-parameter prediction two-step neural network, training, obtaining ultra-smooth optical element integral scattering rate distribution data, and carrying out surface PSD statistical distribution prediction and surface characteristic parameter prediction. The invention realizes the accurate classification of the PSD statistical distribution of the surface of the ultra-smooth optical element, and the high-precision prediction and evaluation of key parameters such as surface roughness, autocorrelation length and the like, thereby providing a brand new solution for the performance evaluation and quality control of the ultra-smooth optical element.

Inventors

ZHANG BIN
LIU BAO
GE YIXIN
ZHONG ZHEQIANG

Assignees

四川大学

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (9)

1. The method for predicting the surface characteristic parameters of the ultra-smooth optical element is characterized by comprising the following steps of: S1, determining the value range and sampling precision of each surface parameter based on priori knowledge of an ultra-smooth optical element; s2, constructing a scattering rate distribution database by utilizing GBK scalar scattering model theory according to the scattering measurement experimental principle of the cavity ring-down method, and preprocessing; S3, constructing a classification-parameter prediction two-step neural network based on the value range and sampling precision of each surface parameter, and training through a preprocessed scattering rate distribution database; S4, obtaining integral scattering rate distribution data of the ultra-smooth optical element through a cavity ring-down method scattering measurement experiment, and carrying out surface PSD statistical distribution prediction and surface characteristic parameter prediction by using a trained classification-parameter prediction two-step neural network.
2. The method for predicting surface characteristics of an ultra-smooth optical element according to claim 1, wherein each surface parameter in S1 and a range of values thereof are: Roughness, the value range is 0.1 to 1nm; the autocorrelation length is in the value range of 1 to 20 mu m; Slope, value range 2.0 to 2.5; Cut-off frequency, the value range is 5X 10 -4 to 0.1 μm -1 .
3. The method for predicting surface characteristics of an ultra-smooth optical element according to claim 1, wherein S2 comprises the steps of: S21, calculating integral scattering rate distribution data of different spatial solid angles of the ultra-smooth optical element under three surface PSD statistical distributions of Gaussian distribution, fractal distribution and Ke Xiluo Lorentz distribution by utilizing a GBK scalar scattering model theory according to an optical cavity ring-down method scattering measurement experimental principle, and forming a scattering rate distribution database; S22, carrying out data normalization on the scattering rate distribution database, randomly scrambling the sequence, dividing the training set and preprocessing the verification set.
4. The method for predicting surface characteristics of an ultra-smooth optical element according to claim 3, wherein in S21, the integrated scattering rate distribution expression of different spatial solid angles of the ultra-smooth optical element is: , Wherein, the In order to integrate the scatter ratio, Is a bi-directional reflection distribution function of the ultra-smooth optical element, As an incident angle in spherical coordinates, Is the scattering angle in the spherical coordinate system, cos is a cosine function, Is the integral solid angle.
5. The method for predicting surface characteristics of an ultra-smooth optical element according to claim 4, wherein in S21, the bi-directional reflection distribution function expression of the ultra-smooth optical element under gaussian distribution is: , , Wherein, the As a bi-directional reflection distribution function of an ultra-smooth optical element under gaussian distribution, As an incident angle in spherical coordinates, Is the azimuth angle of incidence in spherical coordinates, Is the scattering angle in the spherical coordinate system, For the azimuth angle of scattering in spherical coordinates, In order for the roughness to be the same, For the wavelength of the incident light, For polarization dependent surface reflectivity, As a function of the natural index of refraction, At the level of infinity and at the level of infinity, The number of convergence stages, | is the factorial operator, As a function of the PSD in a gaussian distribution, In order to achieve a peripheral rate of the material, In order to be of an autocorrelation length, Is that The spatial frequency in the direction is such that, Is that The spatial frequency in the direction is such that, Intermediate parameters for the GBK scalar scattering model; the bidirectional reflection distribution function expression of the ultra-smooth optical element under fractal distribution is as follows: , , Wherein, the As a bi-directional reflection distribution function of the ultra-smooth optical element under fractal distribution, As a function of the PSD in a fractal distribution, For the fractal distribution of the first parameter, For the fractal distribution of the second parameter, A third parameter is fractal distribution; The bi-directional reflection distribution function expression of the ultra-smooth optical element under Ke Xiluo Lorentz distribution is: , , Wherein, the As a bi-directional reflection distribution function of an ultra-smooth optical element at Ke Xiluo lorentz distribution, As a PSD function in Ke Xiluo lorentz distribution, For a Ke Xiluo lorentz distribution of the first parameter, Is a cut-off frequency of Ke Xiluo lorentz distribution.
6. The method for predicting surface characteristics parameters of an ultra-smooth optical element according to claim 5, wherein said GBK scalar scattering model intermediate parameters The expression of (2) is: , Wherein, the For the refractive index of the incident medium, For the refractive index of the exit medium, Is the relevant surface roughness; The expression of (2) is: , Wherein, the To integrate the cut-off frequency for the spatial frequency, To take the following measures And PSD function as argument; Spatial frequency in direction The expression of (2) is: , wherein sin is a sine function and cos is a cosine function; Spatial frequency in direction The expression of (2) is: 。
7. The method for predicting surface characteristics parameters of an ultra-smooth optical element according to claim 1, wherein the classification-parameter prediction two-step neural network of S3 comprises: The classification neural network is used for analyzing which class of surface PSD statistical distribution is Gaussian distribution, fractal distribution and Ke Xiluo Lorentz distribution according to the integral scattering rate distribution data of the ultra-smooth optical element; the Gaussian distribution parameter prediction network is used for predicting roughness and autocorrelation length aiming at the ultra-smooth optical element integral scattering rate distribution data of Gaussian distribution type; The fractal distribution parameter prediction network is used for predicting roughness, autocorrelation length and slope aiming at the integral scattering rate distribution data of the ultra-smooth optical element of the fractal distribution type; Ke Xiluo Lorentz distribution parameter prediction network is used for predicting roughness, autocorrelation length and cut-off frequency for the ultra-smooth optical element integral scattering rate distribution data of Ke Xiluo Lorentz distribution type.
8. The method for predicting surface characteristic parameters of an ultra-smooth optical element according to claim 7, wherein the classification neural network, the gaussian distribution parameter prediction network, the fractal distribution parameter prediction network and the Ke Xiluo lorentz distribution parameter prediction network have the same structure and each comprise a first convolution block, a second convolution block, a third convolution block, a multi-scale pooling layer, a full-connection layer and an output layer which are sequentially connected; the first convolution block and the second convolution block have the same structure and comprise two convolution layers and a maximum pooling layer which are sequentially connected; The third convolution block is a convolution layer; the multi-scale pooling layer is two parallel pooling branches and comprises a global average pooling layer and a full-connection layer which are sequentially connected, a global maximum pooling layer and a full-connection sub-layer which are sequentially connected, wherein the global average pooling layer and the full-connection sub-layer are respectively connected with the multi-scale pooling layer; The full-connection layer comprises three full-connection sublayers which are connected in sequence; the output layer is a full connection sub-layer.
9. The method of claim 7, wherein S3 trains the classification neural network using a cross entropy loss function, trains the Gaussian distribution parameter prediction network, the fractal distribution parameter prediction network, and the Ke Xiluo Lorentz distribution parameter prediction network using an MSE loss function.

Description

Ultra-smooth optical element surface characteristic parameter prediction method Technical Field The invention relates to the technical field of optical element measurement, in particular to a method for predicting surface characteristic parameters of an ultra-smooth optical element. Background With the advancement of gravitational wave detection technology, the surface quality of ultra-smooth optical elements has become one of the key factors affecting system performance. The gravitational wave satellite-borne telescope needs to have a receiving and transmitting integrated function, and the gravitational wave signal to be measured is extremely weak, so that extremely severe requirements are put on the stray light level of the satellite-borne telescope. Stray light from spaceborne telescopes is primarily limited by the surface characteristics of ultra-smooth optical elements, which typically require sub-nanometer surface roughness and extremely low optical loss. Any minor surface defect or scattering can cause signal attenuation and stray light increase, thereby seriously affecting the measurement accuracy and reliability of the gravitational wave telescope. However, the detection of the surface characteristics of the ultra-smooth optical element still faces significant challenges, on one hand, a unified measurement standard and evaluation system is not established, and on the other hand, the traditional characterization method has limitations in detection efficiency, precision, nondestructive comprehensive evaluation and the like, and is difficult to meet the requirements of high-precision and rapid measurement of the surface morphology, scattering distribution and loss parameters. Among the various optical characterization methods, GBK (Generalized Beckmann Kirchhoff, generalized beckman-kirchhoff) scalar scattering theory can accurately describe the scattering characteristics of ultra-smooth optical elements, including bidirectional scattering distribution, angle-resolved scattering, and integrated scattering rate, but its input needs to depend on the PSD (Power SPECTRAL DENSITY ) of the element surface and its characteristic parameters. It should be noted that, although the cavity ring-down method is used as a high-sensitivity optical loss measurement technique, it can reflect the overall loss characteristics of the device, but it is difficult to directly obtain the surface morphology and the related parameters thereof. Therefore, there is an urgent need to develop a new method that has high precision and high efficiency, and can realize multi-parameter non-destructive detection and prediction. Disclosure of Invention Aiming at the defects in the prior art, the method for predicting the surface characteristic parameters of the ultra-smooth optical element solves the problem of insufficient performance in the aspects of detection efficiency, precision, nondestructive comprehensive evaluation and the like in the prior art. In order to achieve the aim of the invention, the invention adopts the following technical scheme: a method for predicting surface characteristic parameters of an ultra-smooth optical element comprises the following steps: S1, determining the value range and sampling precision of each surface parameter based on priori knowledge of an ultra-smooth optical element; s2, constructing a scattering rate distribution database by utilizing GBK scalar scattering model theory according to the scattering measurement experimental principle of the cavity ring-down method, and preprocessing; S3, constructing a classification-parameter prediction two-step neural network based on the value range and sampling precision of each surface parameter, and training through a preprocessed scattering rate distribution database; S4, obtaining integral scattering rate distribution data of the ultra-smooth optical element through a cavity ring-down method scattering measurement experiment, and carrying out surface PSD statistical distribution prediction and surface characteristic parameter prediction by using a trained classification-parameter prediction two-step neural network. Further, each surface parameter in S1 has a value range: Roughness, the value range is 0.1 to 1nm; the autocorrelation length is in the value range of 1 to 20 mu m; Slope, value range 2.0 to 2.5; Cut-off frequency, the value range is 5X 10 -4 to 0.1 μm -1. Further, the step S2 includes the steps of: S21, calculating integral scattering rate distribution data of different spatial solid angles of the ultra-smooth optical element under three surface PSD statistical distributions of Gaussian distribution, fractal distribution and Ke Xiluo Lorentz distribution by utilizing a GBK scalar scattering model theory according to an optical cavity ring-down method scattering measurement experimental principle, and forming a scattering rate distribution database; S22, carrying out data normalization on the scattering rate distribution database, randomly s