Abstract
A new iterative technique for calculating the eikonal function of a light field providing the focusing into a set of points is introduced. This technique is a modification of the supporting quadric method widely used for design of reflecting and refracting optical surfaces for generating prescribed illuminance distributions at given discrete set of points. As an example, we design a refractive optical element which focuses an incident beam into a set of points with energy pattern forming an image of a keyboard of a calculator. It is shown that the proposed technique is well-suited for the design of diffractive optical elements producing continuous intensity distributions within the scalar theory of diffraction. It is also shown that the calculated eikonal function is a good initial guess when designing diffractive optical elements using the iterative Gerchberg-Saxton algorithm.
© 2015 Optical Society of America
1. Introduction
The problem of design of an optical element generating a required illuminance distribution belongs to the class of extremely challenging inverse problems of nonimaging optics [1]. In many cases such problems can be reduced to finding solutions to nonlinear elliptic partial differential equations of second order. In optical design, direct numerical solutions of such equations in special cases have been considered recently in [2, 3]. In general, such inverse problems of optical design are approached using various iterative techniques [4–13]. One of the universal iterative methods widely used for the design of nonimaging optical elements is the so-called supporting quadric method (SQM) [10–13]. Originally, SQM was developed for solving problems in design of mirrors and lenses transforming incoming beam of light so that the outgoing beam illuminates a given set of points with prescribed intensity distribution. Depending on the type of the problem being solved, the required optical surface is defined as an envelope of a family of quadrics, such as paraboloids, ellipsoids, or hyperboloids. In particular, when designing a mirror for illuminating a target set in the far-field, the required mirror is represented as a set of segments each of which is a piece of a paraboloid. In the near-field case the required mirror is a set of segments of ellipsoids [10, 11]. In [12], one of the present authors used SQM to give a mathematical treatment of the problem of design of a freeform segmented refracting lens transforming an incident collimated beam with given intensity into an output light beam illuminating a given target set with a prescribed intensity distribution. In either case, each segment of the appropriate quadric is defined by certain parameters determined by a special iterative procedure whose convergence is established in a mathematically rigorous manner. The same ideas were applied in [13] to design problems in which quadrics should be replaced by Cartesian ovals.
In this work, a modification of SQM is proposed enabling the method to be applied to the problem of calculating the eikonal function of a light field (specified in a certain plane) providing the focusing into a given discrete set of points. This problem is relevant in the design of diffractive optical elements (DOEs), especially of multifocal diffractive lenses, beam splitters, and diffusers [14–18]. Indeed, the height of the DOE microrelief is in direct proportion to the eikonal function [14]. In addition, the eikonal function can be used to reconstruct refractive (or reflective) optical element generating this eikonal distribution at a preset illuminating beam [19]. This makes the proposed approach well-suited for design of refractive (or reflective) optical elements under the assumption that illuminance distribution of the incident beam is changed insignificantly upon transmission through the optical element [19]. In our setting, we determine the eikonal function providing the focusing into a given set of points as an envelope of a two-parameter set of eikonals of convergent spherical beams with different foci. The proposed method is remarkable for its computational simplicity. As an example, we calculate the eikonal function providing the focusing into an array of points corresponding to a calculator keyboard image. The obtained eikonal function is then used to construct a refractive optical element and DOEs. We demonstrate that the proposed method can be used for designing DOEs producing prescribed continuous intensity distributions within the scalar theory of diffraction.
2. The eikonal function for focusing into a two-dimensional region and discrete point set
The problem of calculating the eikonal for generating a required discrete illuminance distribution can be formulated as follows. Let be the illuminance distribution in a region (aperture) in the plane , where are the Cartesian coordinates in that plane. We wish to determine the eikonal distribution of a light field providing the focusing into a given set of points in the plane with prescribed energy distribution (Fig. 1). Note that in the framework of geometrical optics the light ray directions are determined by the partial derivatives of the eikonal function: . Assuming that the surrounding medium has unity refractive index, we can write the unit vector of a ray originating from a point as
On the basis of the function , a refractive or diffractive optical element focusing into a given set of points can be designed. Several such examples are given below in sections 3 and 4.
This discrete problem of focusing into a set of points may be considered as a particular case of focusing into a desired one or two-dimensional (2D) region. Actually, assume that a mesh approximates a 2D region . Let us calculate the eikonal function in the plane, requiring that the rays originating in G arrive in the region . For that, let us write the eikonal function providing the focusing at point in the plane :
It can be easily shown that Eq. (2) corresponds to the eikonal of a spherical beam converging to point , written at . In this case the ray directions (Eq. (1)) are given by unit vectors directed at the focus. Since such a beam can be formed by an ideal lens, in what follows we refer to Eq. (2) as the “lens eikonal function”. In Eq. (2), the coordinates are interpreted as parameters, with defining a constant equal to the eikonal value at the focal point.
It is proposed that the eikonal function for focusing into a desired region should be defined as the envelope of a two-parameter family of functions in Eq. (2) with respect to the parameters . By definition, this envelope is a function tangent to each eikonal function from Eq. (2) at a certain point. Accordingly, the partial derivatives of the function defining the envelope are equal to the partial derivatives of the eikonal of the corresponding lens at the common point of tangency. In geometrical optics, the ray direction is defined by partial derivatives of the eikonal function, Eq. (1). Hence, the rays defined by the envelope will be directed to the points of region D where the foci of spherical beams defined by Eq. (2) are located.
Thus, the calculation of the eikonal for focusing into a desired region is reduced to constructing an envelope of a two-parameter family of lens eikonals defined by Eq. (2). In Cartesian coordinates, the envelope is given by [20]
In Eq. (3), the first equation is the lens eikonal, whereas the second and third equations define the eikonal derivatives with respect to the parameters. It is interesting that in Eq. (3), the second and third equations give a necessary condition for the extremum of function (2) with respect to the variables at fixed values of the variables . This condition corresponds to the Fermat principle enabling the eikonal function to be represented in the variables aswhereorwhere the function is given by Eq. (2).The general representation of the eikonal function in Eqs. (4)–(6) depends on the function which is to be determined from the requirement that the energy distribution in the region of focusing is equal to a prescribed distribution.
In the discrete version of the problem, light is focused into a set of focal points of the region . So, Eqs. (4)–(6) take the form
whereorand . Equations (7)–(9) define a continuous function composed of lens eikonals that have foci at points . Note that the set of values of is an discrete analog of the function . The values correspond to the eikonal values at focal points and must be defined by the given energy distribution at those points. Note that the relations (7)–(9) for the eikonal function to focus into a planar set of focal points can, in an obvious way, be extended to the case of focusing into a set of points located on an arbitrary surface. In this case, the constant in Eq. (7) should be replaced by a prespecified set of values .3. Solution of the discrete problem of focusing into a set of points
Now, we explain how to determine the values using the requirement that the desired energy distribution is to be generated at points . This problem can be addressed using an analog of the supporting quadric method described in Refs [10, 11]. for design of mirrors. In this case, the required mirror is represented as a set of segments of ellipsoids, with the number of segments equal to that of the prescribed focal points. For each ellipsoid, one focus is located at the light source while the other is at one of the focal points . The large axes of the ellipsoids are calculated by an iterative algorithm using the condition that the required energy distribution must be achieved at the focal points [10, 11]. Note that the large axes of the ellipsoids are equal to the eikonal at the focal points (i.e. the path length).
The problem of determination of the eikonal function from the condition of focusing into points with prescribed energy distribution is nearly identical to the design problem solved by the supporting quadric method of Refs [10, 11]. The functions describing ellipsoids are replaced by the lens eikonal functions:
and the eikonal values at the focal points serve as design parameters.The iterative algorithm for the calculation of the values includes multiple solution of a direct problem that involves the calculation of energy at the focal points based on a ray tracing technique. When designing mirrors composed of segments of ellipsoids, on each iteration, the rays from the source are traced and the distance from the source to each ellipsoid along the current ray is determined. Two kinds of mirrors can be constructed: the first one is built from ellipsoidal segments nearest to the source (along the ray) and the second one is built from segments which are the farthest from the source. The minimum condition is used to design elliptic mirrors (with the rays intersecting the optical axis), whereas the maximum condition is used for hyperbolic mirrors (with the rays not intersecting the optical axis).
For our discrete problem, it follows from Eqs. (7)–(9) that in order to trace a ray from the point it suffices to determine the minimal or maximal element of the array .
In the general case, the energy distribution between the focal points for specified values of is calculated as follows. Let denote the nodes of a rectangular mesh with cell size that approximates the domain in which the eikonal function is defined. Each ray is assumed to carry the energy . For every starting point of the ray, the destination point is derived by finding the minimal (maximal) element of the array . At the destination point (denoted by index ), the energy is incremented by the magnitude .
In the case of mirrors, a detailed description of the monotonically converging iterative algorithm for calculating the ellipsoid parameters can be found in Refs [10, 11]. The said algorithm can be extended without any alterations to calculate the values of in Eq. (10). At each step of the algorithm, the energy distribution between the focal points is calculated, followed by the correction of the values of . If the energy at a point is lower than the specified value then, based on the minimum condition in Eq. (8), the value of needs to be decreased by a magnitude . With ellipsoids, a decrease in the large axis makes the ellipsoid closer to the source, in this way increasing the number of rays directed to the destination point (second focus of the ellipsoid).
Thus, the eikonal function to focus the beam into a set of focal points can be derived using the following iterative algorithm:
- 1) An array of focal points with energy distribution is specified.
- 2) Initial guesses for the eikonal values at the focal points are defined.
- 3) A mesh is introduced in the eikonal definition domain and a matrix of values is calculated.
- 4) Energy distribution between the focal points is calculated.
- 5) If , where is the required accuracy of the target energy distribution, the solution is saved and the iterative process is stopped. Otherwise, continue to the next step.
- 6) The values of are corrected using a method described in Refs [10, 11]. The matrix is calculated for new values of . Return to step 4.
Note that the software implementation of the algorithm involves a single computation of the matrix at step 3. Afterwards, the solution process consists in the iteration of steps 4–6. In that case, the energy distribution between the focal points in step 4 is deduced by calculating the minimal (maximal) elements in the columns of the matrix , whereas the matrix is recalculated in step 6 via altering the row elements by values .
For a detailed description of the method for correcting the values in step 5, which enables the error to decrease monotonically during the iteration process, see Refs [10, 11]. At the same time, the method is characterized by a relatively slow convergence rate, especially for a large number of focal points. The analysis conducted by the authors has shown the following simple heuristic approach to the correction of the values of to be efficient:
where is a positive constant. The signs in Eq. (11) are related to the eikonal type chosen. The plus sign corresponds to an elliptic-type eikonal (with the rays intersecting the optical axis), whereas the minus sign corresponds to a hyperbolic-type eikonal (with the rays not intersecting the optical axis). In Eq. (11), the eikonal increment in the focal points is chosen to be proportional to the difference between the calculated and required energy values in the focal points.4. Design example
To demonstrate the capability of the proposed method to generate complex images, we calculated the eikonal function for the generation of a calculator image defined on a 2D equidistant mesh of pixels (Fig. 2). Note that Fig. 2 contains white pixels that form the required image. The eikonal was calculated from the condition of focusing into a discrete set of points corresponding to white pixels in Fig. 2. The points are located in the plane mm on a mesh with step , the energy distribution between the points is assumed to be constant . The eikonal function was specified in the plane in a square-shaped domain with the side on an equidistant mesh of pixels with step . The illuminance in the domain was assumed to be constant.
The calculated eikonal function is depicted in Fig. 3 and provides the generation of an almost uniform illuminance distribution at the desired points, with the root-mean-square deviation from the specified uniform distribution being less than 6% (Fig. 4). Let us note that the resolution of Fig. 3 does not allow to see the details of the constructed eikonal function. In the next section, a central fragment of this function is shown which demonstrates that it consists of multiple segments connected in a continuous manner.
On the basis of the derived eikonal function, a refractive surface was reconstructed. Exact analytical formulas to calculate a refractive surface that would generate a required eikonal distribution were presented in Ref [19]. At the same time, for the chosen “paraxial” parameters, the height of the refractive surface can be obtained using the thin element approximation [14]:
where is the refractive index of the optical element material.To verify the validity of the design technique given above, the optical element defined by Eq. (12) and the eikonal in Fig. 3 was simulated using the commercial software TracePro implementing a ray tracing technique [21]. For the simulation, the surface defined by Eq. (12) was approximated by a system of non-uniform rational Bezier splines (NURBS) using the computer-aided design software Rhinoceros [22]. The input surface of the optical element was assumed to be the plane perpendicular to the propagation direction of a uniform incident beam. Figure 5 shows the refractive optical element and the simulated illuminance pattern generated by the element. The illuminance distribution was calculated using 1,000,000 rays and confirms that the incident beam is focused into a desired pattern. A certain non-uniformity in the energy distribution between the focal points is due to the spline-aided approximation of the element surface and the inherent random noise of the ray tracing technique.
5. Analysis of the diffraction effects
The problem of designing phase diffractive optical elements (DOEs) is often formulated as that of calculating the eikonal function (or a related phase function , where is the wavenumber and is the free-space wavelength) from the condition of generating a desired illuminance pattern [14]. The DOE microrelief height is given by
The eikonal function in Fig. 3 to focus into a set of points in Fig. 2 was calculated in the framework of geometrical optics. In this context, it is worth studying the performance of a DOE defined by Eq. (13) within the scalar theory of diffraction. Below, the illuminance distribution generated by the eikonal function of Fig. 3 is analyzed using the Fresnel-Kirchhoff approximation. The complex amplitude of the light field in the DOE plane is defined as . In the Fresnel-Kirchhoff approximation, the illuminance distribution in the plane takes the form [14]
For the convenience of the calculations, the eikonal function is represented aswith the first term defining the eikonal of a paraxial lens [14]. In this case, the calculation of is reduced to the calculation of the following Fourier integral:Fig. 6 (a) depicts the central fragment of the function modulo . For the calculation of the integral in Eq. (16), the said eikonal function was calculated using Eqs. (7) and (9) on an equidistant mesh of pixels with the step . The resulting illuminance pattern obtained in the Fresnel approximation and corresponding to the calculator image is shown in Fig. 6(d). The illuminance in Fig. 6(d) was calculated on an equidistant mesh composed of pixels, with the step . Note that in the Fresnel approximation, the magnitude corresponds to a focal spot size generated by a lens of focus f and square aperture G with the side length [14]. For the given parameters (, , ), . In contrast with Fig. 5, individual points forming the keyboard image are significantly less evident in Fig. 6(d). The reason is that due to diffraction the points turned into finite-size spots, which have merged to form an almost continuous image of the calculator. Let us derive a qualitative estimate of an average size of the diffraction spots for the aforesaid example. In accordance with the employed calculation method, the eikonal function in Figs. 3 and 6(a) corresponds to a set of lens eikonal segments (subapertures) of equal-area that are connected in a continuous manner. As an estimate of the focal-spot average size, we will use the quantity
which corresponds to the diffraction spot size of a lens with the aperture size , where is the number of focal points. Note that the distance between the image points is assumed to be larger than , meaning there is no subaperture interplay. For the above given parameters, we have . It should be reminded that the calculator image was defined on a rectangular mesh with the step . Thus, for the studied example, , which accounts for a continuous illuminance pattern in Fig. 6(d).These arguments show that the proposed method is suitable for calculating the eikonal function when designing DOEs intended to form given continuous illuminance patterns. Note that the continuity of the distribution is achieved by properly choosing the distance between points while calculating the eikonal function. Let us study the magnitude of diffraction effects with decreasing the mesh step relatively to the value of . It can be easily shown [14] that in the paraxial approximation, the eikonal function
enables focusing into a scaled set of points with coordinates, where is the scale factor. Figures 6(b), and 6(c) depict the eikonal functions at and , respectively. The illuminance patterns calculated in the framework of geometrical optics (not shown here) confirmed that for the functions in Eq. (18) at and , scaled calculator images similar to that in Fig. 4 are generated. The illuminance distributions obtained in the Fresnel approximation for the above-given functions are depicted in Figs. 6(e) and 6(f). The images suggest that with the decrease in the distance between points, the quality of the calculator image deteriorates. In particular, at the digits become unreadable. The deterioration of the image quality is caused by the overlapping of diffraction spots from different (not just adjacent) segments, leading to their interference. Thus, the method for calculating the eikonal function when designing a DOE to form continuous illuminance patterns has its application limits. In the considered example, an image of acceptable quality can only be obtained if the distance between points is in the interval .Within the scalar theory of diffraction, iterative and gradient methods have been widely used to design DOEs [14, 23]. The most popular methods are the Gerchberg-Saxton (GS) algorithm and its modifications [14, 23]. Both the convergence of the GS algorithm and the quality of the obtained solution essentially depend on the initial guess for the DOE phase function. Let us analyze the performance of the GS algorithm using the geometrical optics solution as an initial guess. As the required distribution, the continuous image of the calculator in Fig. 2 is used with the points being replaced by square pixels of size .
Figs. 7 (а)-(c) depict the eikonal functions derived after 30 iterations of the GS algorithm. The eikonal functions in Figs. 6 (а)-(c) were used as initial guesses. The illuminance distributions generated by the eikonal functions in Figs. 7 (а)-(c) are presented in Figs. 7 (d)–(f), showing the required calculator images at different scales. In all three cases, the root-mean-square deviation values of the numerically calculated illuminance patterns from the desired ones do not exceed 5%. Thus, the use of the geometrical optics solution as the initial guess enables the fast convergence of the iterative algorithm. Note, for comparison, that with a random initial guess of the phase, the root-mean-square deviation of the numerically obtained illuminance pattern from the required one stays in the range of 10–15% after 50–100 iterations. The authors believe that the problem of choosing the initial guess is of special significance when generating required illuminance patterns on non-planar or inclined planar surfaces. In that case, the relation between the field in the aperture region and the target region cannot be expressed by the Fourier integral (16), which can be calculated using known fast algorithms. At the same time, as we indicated at the end of Section 1, the proposed algorithm for calculating the eikonal function can be extended almost without change to the case when the target points are located on an arbitrary surface.
6. Conclusion
An iterative method for calculating the eikonal function of a light field from the condition of focusing into a set of points was proposed. The developed method is a modification of the supporting quadric method developed for designing mirrors and refractive optical elements generating discrete intensity or illuminance patterns on a given set of points. The method has been developed as a particular solution of the problem of calculating the eikonal to focus light into a desired region, with the eikonal function being expressed as the envelope of a two-parameter family of eikonals of lenses. High efficiency of the proposed approach is demonstrated by calculating the eikonal function to form a complex illuminance pattern corresponding to a calculator keyboard image. The method is suited for the design of DOEs to generate desired continuous illuminance patterns using the scalar theory of diffraction. The continuity of the resulting patterns has been achieved by choosing a proper distance between the focal points while calculating the eikonal function. The eikonal function offers a good initial guess when designing DOEs using the iterative Gerchberg-Saxton algorithm.
Acknowledgments
The work of L. L. Doskolovich, M.A. Moiseev and E.A. Bezus (sections 2, 4–6) was funded by the Russian Science Foundation, project # 14-19-00969; the work of V. Oliker (sections 1 and 3) was partially supported by grants from the US Air Force Office of Scientific Research (AFOSR) and US-Israel Binational Science Foundation (BSF).
References and links
1. R. Winston, J. C. Miñano, P. Benítez, N. Shatz, and J. C. Bortz, Nonimaging Optics (Elsevier Academic Press, 2005).
2. R. Wu, L. Xu, P. Liu, Y. Zhang, Z. Zheng, H. Li, and X. Liu, “Freeform illumination design: a nonlinear boundary problem for the elliptic Monge-Ampére equation,” Opt. Lett. 38(2), 229–231 (2013). [CrossRef] [PubMed]
3. R. Wu, P. Benítez, Y. Zhang, and J. C. Miñano, “Influence of the characteristics of a light source and target on the Monge-Ampére equation method in freeform optics design,” Opt. Lett. 39(3), 634–637 (2014). [CrossRef] [PubMed]
4. Z. Feng, Y. Luo, and Y. Han, “Design of LED freeform optical system for road lighting with high luminance/illuminance ratio,” Opt. Express 18(21), 22020–22031 (2010). [CrossRef] [PubMed]
5. Y. Luo, Z. Feng, Y. Han, and H. Li, “Design of compact and smooth free-form optical system with uniform illuminance for LED source,” Opt. Express 18(9), 9055–9063 (2010). [CrossRef] [PubMed]
6. M. A. Moiseev and L. L. Doskolovich, “Design of refractive spline surface for generating required irradiance distribution with large angular dimension,” J. Mod. Opt. 57(7), 536–544 (2010). [CrossRef]
7. M. A. Moiseev, S. V. Kravchenko, and L. L. Doskolovich, “Design of efficient LED optics with two free-form surfaces,” Opt. Express 22(S7), A1926–A1935 (2014). [CrossRef] [PubMed]
8. K. Wang, F. Chen, Z. Liu, X. Luo, and S. Liu, “Design of compact freeform lens for application specific Light-Emitting Diode packaging,” Opt. Express 18(2), 413–425 (2010). [CrossRef] [PubMed]
9. X. Mao, H. Li, Y. Han, and Y. Luo, “Polar-grids based source-target mapping construction method for designing freeform illumination system for a lighting target with arbitrary shape,” Opt. Express 23(4), 4313–4328 (2015). [CrossRef] [PubMed]
10. V. I. Oliker, “Mathematical aspects of design of beam shaping surfaces in geometrical optics” in Trends in Nonlinear Analysis, M. Kirkilionis, S. Krömker, R. Rannacher, and F. Tomi, eds. (Springer, 2003).
11. S. A. Kochengin and V. I. Oliker, “Determination of reflector surfaces from near-field scattering data II. Numerical solution,” Numer. Math. 79(4), 553–568 (1998). [CrossRef]
12. V. Oliker, J. Rubinstein, and G. Wolansky, “Supporting quadric method in optical design of freeform lenses for illumination control of a collimated light,” Adv. Appl. Math. 62, 160–183 (2015). [CrossRef]
13. D. Michaelis, P. Schreiber, and A. Bräuer, “Cartesian oval representation of freeform optics in illumination systems,” Opt. Lett. 36(6), 918–920 (2011). [CrossRef] [PubMed]
14. V. A. Soifer, V. V. Kotlyar, and L. L. Doskolovich, Iterative Methods for Diffractive Optical Elements Computation (Taylor & Francis Ltd., 1997).
15. M. A. Golub, L. L. Doskolovich, N. L. Kazanskiy, S. I. Kharitonov, and V. A. Soifer, “Computer generated diffractive multi-focal lens,” J. Mod. Opt. 39(6), 1245–1251 (1992). [CrossRef]
16. A. Zlotnik, S. B. Yaish, O. Yehezkel, K. Lahav-Yacouel, M. Belkin, and Z. Zalevsky, “Extended depth of focus contact lenses for presbyopia,” Opt. Lett. 34(14), 2219–2221 (2009). [CrossRef] [PubMed]
17. N. Kazanskiy and R. Skidanov, “Binary beam splitter,” Appl. Opt. 51(14), 2672–2677 (2012). [CrossRef] [PubMed]
18. R. Bitterli, T. Scharf, H.-P. Herzig, W. Noell, N. de Rooij, A. Bich, S. Roth, K. J. Weible, R. Voelkel, M. Zimmermann, and M. Schmidt, “Fabrication and characterization of linear diffusers based on concave micro lens arrays,” Opt. Express 18(13), 14251–14261 (2010). [CrossRef] [PubMed]
19. L. L. Doskolovich, A. Yu. Dmitriev, and S. I. Kharitonov, “Analytic design of optical elements generating a line focus,” Opt. Eng. 52(9), 091707 (2013). [CrossRef]
20. L. P. Eisenhart, A Treatise on the Differential Geometry of Curves and Surfaces (Schwarz Press, 2008).
23. J. R. Fienup, “Phase retrieval algorithms: a comparison,” Appl. Opt. 21(15), 2758–2769 (1982). [CrossRef] [PubMed]