Generating multi-wavelength phase screens for atmospheric wave optics simulations using fast Fourier transforms

M. W. Hyde IV a , M. F. Spencer b,c , and M. Kalensky d · 2025

Read Generating multi-wavelength phase screens for atmospheric wave optics simulations using fast Fourier transforms by M. W. Hyde IV a , M. F. Spencer b,c , and M. Kalensky d (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-00613.txt.

Introduction

Hyde and co-authors present a fast method to build phase screens for simulating how light waves travel through turbulent air at several colors at once. Such screens help test adaptive optics systems before they are built. Their 2025 paper uses fast Fourier transforms to synthesize the screens.

The team checked the method by comparing a known theory value, the path length structure function for two colors, with what their screens produced. The two matched very well. The authors say the method will help run two color adaptive optics simulations in strong turbulence. That setting is where older screen methods tend to struggle most, they note. Such tests matter before hardware faces real sky conditions.

Catalog record

CREATORM. W. Hyde IV a , M. F. Spencer b,c , and M. Kalensky d
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE16 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-00613

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Generating multi-wavelength phase screens for atmospheric wave optics simulations using fast Fourier transforms M. W. Hyde IV a , M. F. Spencer b,c , and M. Kalensky d Abstract We present a method to synthesize phase screens for multi-wavelength, atmospheric wave optics simulations using fast Fourier transforms. We validate our work by comparing the theoretical, two-wavelength optical path length structure function to simulated results, and we find the two to be in excellent agreement. Our phase screen synthesis method will be useful in two-wavelength adaptive optics simulations in strong or deep turbulence. keywords Adaptive optics; atmospheric turbulence; wave optics simulations † † articletype: RESEARCH ARTICLE † † affiliation: a Epsilon C5I, Beavercreek, OH, 45431, USA b Department of Engineering Physics, Air Force Institute of Technology, Dayton, OH, 45433, USA c Joint Directed Energy Transition Office, Office of the Under Secretary of Defense for Research and Engineering, Washington, DC, USA d Integrated Engagement Systems Department, Naval Surface Warfare Center Dahlgren Division, Dahlgren, Virginia, USA 1 Introduction Techniques to quickly and accurately synthesize atmospheric phase screens for use in wave optics simulations has been an active area of research for over 30 years [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 ] . For much of this history, the primary goal has been to generate an accurate, discrete representation of a continuous turbulent path at a single wavelength. Recently, there has been interest in two- or multi-wavelength adaptive optics (AO) systems for beam projection applications [ 10 , 11 ] . In two-wavelength AO, the atmosphere is sensed at one wavelength and corrected at another. In astronomy, this is common practice and has been studied for decades [ 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 ] . However, in contrast to astronomical viewing, two-wavelength AO systems for beam projection generally operate in complementary conditions (i.e., small fields of view and long, nearly horizontal paths [ 22 ] ). The latter causes scintillation which gives rise to phase discontinuities known as branch points [ 23 , 24 ] . The effect of these phenomena on two-wavelength AO performance has yet to be fully quantified [ 10 ] . This, of course, motivates research into two-wavelength AO systems in strong turbulence conditions and, as a consequence, techniques to generate turbulence realizations with the proper statistics at two or more wavelengths. Indeed, the two-wavelength phase covariance function B_{S} , first derived by Ishimaru [ 25 ] , is \begin{split}B_{S}\left(\boldsymbol{\rho},k_{1},k_{2}\right)&=\left\langle\phi\left(\boldsymbol{\rho}_{1},k_{1}\right)\phi\left(\boldsymbol{\rho}_{2},k_{2}\right)\right\rangle=\iint_{-\infty}^{\infty}\Phi_{S}\left(\boldsymbol{\kappa},k_{1},k_{2}\right)\exp\left(\text{j}\boldsymbol{\kappa}\cdot\boldsymbol{\rho}\right)\text{d}^{2}\kappa\hfill\\ \Phi_{S}\left({\kappa},k_{1},k_{2}\right)&=\pi k_{1}k_{2}z\Phi_{n}\left(\kappa\right)\left\{\operatorname{sinc}\left[\frac{z}{2}\left(\frac{1}{k_{1}}-\frac{1}{k_{2}}\right)\kappa^{2}\right]+\operatorname{sinc}\left[\frac{z}{2}\left(\frac{1}{k_{1}}+\frac{1}{k_{2}}\right)\kappa^{2}\right]\right\},\hfill\end{split} (1) where \boldsymbol{\rho}=\boldsymbol{\rho}_{1}-\boldsymbol{\rho}_{2} , \Phi_{n} is the index of refraction power spectrum, z is the propagation distance, and \operatorname{sinc}\left(x\right)=\sin\left(x\right)/x . Letting k_{1}=k_{2}=k and making the approximation that the second \operatorname{sinc}\left(x\right) function is equal to one (better known as the geometrical optics approximation) simplifies Eq. ( 1 ) to the expression used for decades to generate atmospheric phase screens. However, multi-wavelength simulations involving turbulence should use the full expression in Eq. ( 1 ) to synthesize screens, as it is necessary to capture the physical correlation of the turbulent phase across the wavelengths of interest. Charnotskii, in Ref. [ 26 ] , was the first to present such a technique. Therein, he describes how to synthesize two-wavelength phase screens using the sparse spectrum (SS) method [ 7 , 8 ] . Our purpose in this short paper is to show how to generate such screens using the more popular Fourier or spectral method [ 2 , 5 , 9 ] . It is important to state that although the Fourier technique may be more popular than SS, Charnotskii makes compelling arguments for the superiority of the SS method in Refs. [ 9 , 26 ] . We do not dispute his conclusions in those papers. Rather, our purpose here is merely to generalize the existing single-wavelength Fourier method to any number of wavelengths. In what follows, we present the theory, procedure, and validation of our multi-wavelength Fourier/spectral phase screen method. 2 Theory We begin with the traditional Fourier method for generating an atmospheric phase screen at a single wavenumber k_{1}=2\pi/\lambda_{1} [ 2 , 5 , 9 ] : \begin{gathered}\phi\left[i,j,k_{1}\right]=\operatorname{Re}\left\{\sum_{m,n}\left(r^{\text{r}}\left[m,n,k_{1}\right]+\text{j}r^{\text{i}}\left[m,n,k_{1}\right]\right)\sqrt{\Phi_{S}\left[m,n,k_{1},k_{1}\right]\frac{2\pi}{L_{x}}\frac{2\pi}{L_{y}}}\right.\hfill\\ \left.\quad\times\exp\left(\text{j}2\pi\frac{m}{M}i\right)\exp\left(\text{j}2\pi\frac{n}{N}j\right)\right\},\hfill\end{gathered} (2) where \operatorname{Re}\left(c\right) is the real part of c ; M,N are the numbers of points along the horizontal and vertical grid directions; i,j are indices corresponding to x,y ; and m,n are indices corresponding to spatial frequencies \kappa_{x},\kappa_{y} . Also in Eq. ( 2 ), L_{x}=M\Delta x , L_{y}=N\Delta y , \Delta x,\Delta y are the horizontal and vertical grid spacings, r^{\text{r}},r^{\text{i}} are N\times M matrices of independent, standard normal random numbers, and lastly, \Phi_{S} is the phase power spectrum in Eq. ( 1 ) [ 27 , 28 ] . Equation ( 2 ) is equivalent to a discrete inverse Fourier transform; therefore, the fast Fourier transform (FFT) algorithm can be used to synthesize the screen. Note that the imaginary part of the sum in Eq. ( 2 ) is statistically independent of the real part and also has the proper spatial statistics. Either or both can be used to form \phi . Our goal is to generate other phase screens at k_{2},\,k_{3},\cdots,\,k_{Q} using Eq. ( 2 ) that have the proper covariance. Proceeding, we compute the moment \left\langle\phi\left[i_{1},j_{1},k_{q}\right]\phi\left[i_{2},j_{2},k_{p}\right]\right\rangle using Eq. ( 2 ) and, after expansion, obtain \begin{gathered}\left\langle\phi\left[i_{1},j_{1},k_{p}\right]\phi\left[i_{2},j_{2},k_{q}\right]\right\rangle=\sum_{m_{1},n_{1}}\sum_{m_{2},n_{2}}\frac{2\pi}{L_{x}}\frac{2\pi}{L_{y}}\sqrt{\Phi_{S}\left[m_{1},n_{1},k_{p},k_{p}\right]}\sqrt{\Phi_{S}\left[m_{2},n_{2},k_{q},k_{q}\right]}\hfill\\ \quad\times\left\{\left\langle r^{\text{r}}\left[m_{1},n_{1},k_{p}\right]r^{\text{r}}\left[m_{2},n_{2},k_{q}\right]\right\rangle\cos\left[2\pi\left(\frac{m_{1}}{M}i_{1}+\frac{n_{1}}{N}j_{1}\right)\right]\cos\left[2\pi\left(\frac{m_{2}}{M}i_{2}+\frac{n_{2}}{N}j_{2}\right)\right]\right.\hfill\\ \quad\left.+\,\left\langle r^{\text{i}}\left[m_{1},n_{1},k_{p}\right]r^{\text{i}}\left[m_{2},n_{2},k_{q}\right]\right\rangle\sin\left[2\pi\left(\frac{m_{1}}{M}i_{1}+\frac{n_{1}}{N}j_{1}\right)\right]\sin\left[2\pi\left(\frac{m_{2}}{M}i_{2}+\frac{n_{2}}{N}j_{2}\right)\right]\right.\hfill\\ \quad\left.-\,\left\langle r^{\text{r}}\left[m_{1},n_{1},k_{p}\right]r^{\text{i}}\left[m_{2},n_{2},k_{q}\right]\right\rangle\cos\left[2\pi\left(\frac{m_{1}}{M}i_{1}+\frac{n_{1}}{N}j_{1}\right)\right]\sin\left[2\pi\left(\frac{m_{2}}{M}i_{2}+\frac{n_{2}}{N}j_{2}\right)\right]\right.\hfill\\ \quad\left.-\,\left\langle r^{\text{i}}\left[m_{1},n_{1},k_{p}\right]r^{\text{r}}\left[m_{2},n_{2},k_{q}\right]\right\rangle\sin\left[2\pi\left(\frac{m_{1}}{M}i_{1}+\frac{n_{1}}{N}j_{1}\right)\right]\cos\left[2\pi\left(\frac{m_{2}}{M}i_{2}+\frac{n_{2}}{N}j_{2}\right)\right]\right\}.\hfill\end{gathered} (3) Letting \left\langle r_{1}^{\text{r}}r_{2}^{\text{r}}\right\rangle=\left\langle r_{1}^{\text{i}}r_{2}^{\text{i}}\right\rangle and \left\langle r_{1}^{\text{r}}r_{2}^{\text{i}}\right\rangle=\left\langle r_{1}^{\text{i}}r_{2}^{\text{r}}\right\rangle simplifies Eq. ( 3 ) to \begin{gathered}\left\langle\phi\left[i_{1},j_{1},k_{p}\right]\phi\left[i_{2},j_{2},k_{q}\right]\right\rangle=\sum_{m_{1},n_{1}}\sum_{m_{2},n_{2}}\frac{2\pi}{L_{x}}\frac{2\pi}{L_{y}}\sqrt{\Phi_{S}\left[m_{1},n_{1},k_{p},k_{p}\right]}\sqrt{\Phi_{S}\left[m_{2},n_{2},k_{q},k_{q}\right]}\hfill\\ \quad\times\left\{\left\langle r^{\text{r}}\left[m_{1},n_{1},k_{p}\right]r^{\text{r}}\left[m_{2},n_{2},k_{q}\right]\right\rangle\cos\left[2\pi\left(\frac{m_{1}}{M}i_{1}+\frac{n_{1}}{N}j_{1}-\frac{m_{2}}{M}i_{2}-\frac{n_{2}}{N}j_{2}\right)\right]\right.\hfill\\ \quad\left.-\,\left\langle r^{\text{r}}\left[m_{1},n_{1},k_{p}\right]r^{\text{i}}\left[m_{2},n_{2},k_{q}\right]\right\rangle\sin\left[2\pi\left(\frac{m_{1}}{M}i_{1}+\frac{n_{1}}{N}j_{1}+\frac{m_{2}}{M}i_{2}+\frac{n_{2}}{N}j_{2}\right)\right]\right\}.\hfill\end{gathered} (4) Equation ( 4 ) must equal B_{S} in Eq. ( 1 ) for the screens to be physical. This requires \left\langle r_{1}^{\text{r}}r_{2}^{\text{i}}\right\rangle=0 and \begin{gathered}\left\langle r^{\text{r}}\left[m_{1},n_{1},k_{p}\right]r^{\text{r}}\left[m_{2},n_{2},k_{q}\right]\right\rangle=\frac{\Phi_{S}\left[m_{1},n_{1},k_{p},k_{q}\right]\delta\left[m_{1}-m_{2}\right]\delta\left[n_{1}-n_{2}\right]}{\sqrt{\Phi_{S}\left[m_{1},n_{1},k_{p},k_{p}\right]}\sqrt{\Phi_{S}\left[m_{2},n_{2},k_{q},k_{q}\right]}},\end{gathered} (5) where \delta\left[x\right] is the Kronecker delta function. Substituting Eq. ( 5 ) into Eq. ( 4 ) and evaluating the trivial sums over m_{2},n_{2} , we obtain the Riemann sum form of Eq. ( 1 ). The above analysis reveals that we can synthesize phase screens at k_{p} and k_{q} with the proper, physical covariance by using correlated Gaussian random numbers in Eq. ( 2 ). Such numbers are easy to generate using Cholesky factors. Referring to Eq. ( 5 ), the covariance matrix for r^{\text{r}}_{p},r^{\text{r}}_{q} and r^{\text{i}}_{p},r^{\text{i}}_{q} (the real and imaginary parts are statistically independent) is equal to \mathbf{\Sigma}_{pq}=\begin{bmatrix}1&R_{pq}\\ R_{pq}&1\end{bmatrix}, (6) where R_{pq} is R_{pq}=\frac{\Phi_{S}\left[m,n,k_{p},k_{q}\right]}{\sqrt{\Phi_{S}\left[m,n,k_{p},k_{p}\right]}\sqrt{\Phi_{S}\left[m,n,k_{q},k_{q}\right]}}. (7) It is easy to generalize Eq. ( 6 ) to any number of wavenumbers, such that \mathbf{\Sigma}=\begin{bmatrix}1&R_{12}&R_{13}&\dots&R_{1Q}\\ R_{12}&1&R_{23}&\dots&R_{2Q}\\ R_{13}&R_{23}&1&\dots&R_{3Q}\\ \vdots&\vdots&\vdots&\ddots&\vdots\\ R_{1Q}&R_{2Q}&R_{3Q}&\dots&1\end{bmatrix}. (8) The Cholesky decomposition of Eq. ( 8 )—i.e., \mathbf{\Sigma}=\mathbf{L}\mathbf{L}^{\text{T}} , where \mathbf{L} is a lower triangular matrix—is [ 29 , 30 ] \begin{split}L_{qq}&=\left(\Sigma_{qq}-\sum\limits_{k=1}^{q-1}L_{qk}^{2}\right)^{1/2}\\ L_{pq}&=\frac{1}{L_{qq}}\left(\Sigma_{pq}-\sum\limits_{k=1}^{q-1}L_{pk}L_{qk}\right)\quad p>q.\end{split} (9) The Gaussian random numbers r^{\text{r}}_{1,\cdots,Q} and r^{\text{i}}_{1,\cdots,Q} , which are substituted into Eq. ( 2 ) to synthesize phase screens at k_{1},\,k_{2},\,\cdots,\,k_{Q} , are generated from 2Q N\times M matrices of independent, standard normal random numbers according to \begin{bmatrix}r^{\text{r}}_{1},r^{\text{i}}_{1}\\ r^{\text{r}}_{2},r^{\text{i}}_{2}\\ r^{\text{r}}_{3},r^{\text{i}}_{3}\\ \vdots\\ r^{\text{r}}_{Q},r^{\text{i}}_{Q}\end{bmatrix}=\begin{bmatrix}L_{11}&0&0&\dots&0\\ L_{21}&L_{22}&0&\dots&0\\ L_{31}&L_{32}&L_{33}&\dots&0\\ \vdots&\vdots&\vdots&\ddots&\vdots\\ L_{Q1}&L_{Q2}&L_{Q3}&\cdots&L_{QQ}\end{bmatrix}\begin{bmatrix}r_{1},r_{Q+1}\\ r_{2},r_{Q+2}\\ r_{3},r_{Q+3}\\ \vdots\\ r_{Q},r_{2Q}\end{bmatrix} (10) with L_{pq} computed recursively using Eq. ( 9 ). Figure 1: Two-wavelength OPL structure function D_{l}\left(\rho,\lambda_{p},\lambda_{1}\right) results plotted on linear and log-log scales. 3 Procedure In summary, the procedure for generating multi-wavelength phase screens using FFTs is as follows: 1. Generate two sets of Q standard normal random numbers. 2.

Read the complete text