1

Summer Research Intern Report

On

Modeling Binary Optical Surfaces through

Comprehensive Phase Analysis and Curve Fitting
By
Pranov Raj
MSc. Tech (Photonics), NIT Warangal
Carried out at

CSIR-CSIO(Chandigarh)
For the duration of (20th may – 20th July)
Under the supervision of
Ms. Ishani Salaria
(Scientist, Micro and Nano Optics Centre)
 2

ACKNOWLEDGEMENT

I extend my heartfelt gratitude to CSIR-CSIO, Chandigarh for the enriching opportunity to

partake in the summer Research internship program. This experience has been profoundly
influential in my professional and personal growth.

I would like to express my deepest appreciation to Ms. Ishani Salaria (Scientist, Optical devices
and systems Research Area) for their unwavering support, mentorship, and guidance throughout
the duration of the internship.

In conclusion, I also thank my Parents and Friends for their support and encouragement, which
played a crucial role in this Internship.

This internship has been a transformative experience, and I am thankful for the knowledge gained,
the skill developed, and the relationships formed. The lesson learned will undoubtedly shape my
future endeavors in the field of optical engineering.
 3

Date: 19/07/2024

DECLARATION
I affirm that the project titled ‘Modeling Binary Optical Surfaces through Comprehensive Phase
Analysis and Curve Fitting’ faithfully represents my independent work conducted at CSIR- Central
Scientific Instrumental Organization (CSIO) as a Summer Research Intern. This research spanned
two months from May to July 2024, under the guidance of Ms. Ishani Salaria.

Pranov Raj

(Id- T2405212)

Certified that the above statement made by the student is correct to the best of our knowledge and
belief.

Ms. Ishani Salaria

(Scientist, Micro and Nano Optics Centre)

 4

TABLE OF CONTENT

ABSTRACT 5

Introduction 5

Methods 6

Curve fitting 10

Modeled Plano convex lens 14

Binary equivalent surface 17

Fourier Transforming property of thick lens 20

Optical imaging system 22

Computational approach for lens combination 23

Zemax model for lens combination 29

Binary equivalent surface 30

Conclusion 31
 5

ABSTRACT
This report outlines the process of calculating the target phase profile required for designing
diffractive metalenses. Metalenses are a type of optical lens that use nanostructured materials to
manipulate light at a subwavelength scale. By adjusting the shape and arrangement of these
nanostructures, we can control how light behaves as it passes through the lens. First, we defined
the desired phase profile for the metalens. Using a Binary 2 surface, we performed curve fitting to
obtain the coefficients for the phase term, specifically focusing on the ρ^2 term. These coefficients
were then entered into Zemax, an optical design software, to verify their accuracy and ensure the
design would function as intended.

In addition to single-lens modeling, we extended our approach to a multi-lens system. This system
consisted of two lenses separated by free space. By calculating the phase introduced by each lens
and the propagation phase through the free space, we determined the total phase introduced by the
optical system immediately after the second lens. We then used the derived binary surface
coefficients to create a binary equivalent of the lens system and validated the results using Zemax.

INTRODUCTION
Lens elements are fundamental components used for imaging purposes. Starting with the human
eye, lens systems play a crucial role in commonly used imaging devices such as cameras and
microscopes. Their primary function is to focus light, moving an image at infinity to an image at
some finite distance. However, traditional lenses can be bulky and heavy, which is a significant
drawback in modern optical systems where compactness and weight are critical considerations.

To address these limitations, metalenses have emerged as a promising solution. Metalenses are a
type of optical lens that use nanostructured materials to manipulate light at a subwavelength scale.
By adjusting the shape and arrangement of these nanostructures, metalenses can control light with
high precision, offering the potential to replace bulky traditional lenses with much thinner and
lighter alternatives.

Designing metalenses requires precise knowledge of the target phase profile that the lens must
impart on incident light. This phase profile dictates how the light waves are manipulated as they
 6

pass through the lens, influencing properties such as focal length and aberrations. Achieving the
desired phase profile involves computational methods, including the use of Binary 2 surfaces for
curve fitting. These methods enable the extraction of coefficients that define the phase terms
necessary for accurate simulation and validation using optical design software like ZEMAX.

METHODS
The main focus of our study is to accurately determine and validate the target phase profile of a
thick lens.

Unlike thin lens, thick lens is having some thickness providing more accurate modeling for real-
world applications.

In Fourier optics, A thick lens is an optical element characterized by its finite axial thickness and
refractive properties, which introduce a complex phase shift to the incident wavefront.

We assume the radii of curvatures for the two lens surfaces of thick lens to be R1 and R2,
respectively, as shown in Fig.

Fig: Thickness function (a) Front View (b) Side View [1]

Where 𝜟𝟎 is the maximum thickness of the lens and ∆ is the thickness at coordinate (x, y).
Therefore, the total phase delayed suffered by the lens is:
 7

𝝓(𝒙, 𝒚) = 𝑲𝒏∆(𝒙, 𝒚) + 𝒌[𝜟𝟎 − ∆(𝒙, 𝒚)]

Because we want to represent the lens in a wave-optics sense, the lens may also be a simple
multiplicative phase change of the field:

𝒕𝒍 = 𝐞𝐱𝐩(𝒋𝝓(𝒙, 𝒚))……(i)

The complex field U’ immediately behind the lens is related to the complex field U incident on
the plane immediately in front of the lens:

𝑼′ = 𝒕𝒍 (𝒙, 𝒚)𝑼(𝒙, 𝒚)

In order to compute the Phase introduced by the lens, we need to first determine the thickness
function ∆(𝒙, 𝒚).

To find the thickness ∆(x, y), we split the lens into three parts, as shown in Fig and write the total
thickness function as the sum of three individual thickness functions
 8

Fig: calculation of Thickness function (i) Geometry of ∆𝟏 (ii) Geometry of ∆𝟐 (iii) Geometry of
∆𝟑

By calculating thickness function and substituting its value in phase delay equation, we can
determine the total phase delay introduced by the lens.

Computational approach:
Initial Parameters

format long
% Parameters
lambda = 660e-9; % wavelength in meters (660 nm)
 9

k = 2 * pi / lambda; % wavenumber

n = 1.5168; % refractive index of the lens material (BK7)

A0 = 0.005e-3; % maximum thickness of the lens

R1 = 0.113696e-3; % radius of curvature of the first lens surface

R2 = 10000; % radius of curvature of the second lens surface

The meshgrid function creates a rectilinear grid for representing and manipulating spatial
phase data. This grid is essential for modeling complex spatial variations in phase across a
2D surface, making it useful for analyzing optical elements like lenses or meta surfaces with
intricate phase profile.

% Create meshgrid for x and y

[X, Y] = meshgrid(-15e-6:4.5e-7:15e-6, -15e-6:4.5e-7:15e-6);
% Compute rho
rho = sqrt(X.^2 + Y.^2);

% Compute phi using the given equation

phi = k * n * A0 - k * (n - 1) * R1 * (1 - sqrt(1 - rho.^2 / R1^2)) +
k * (n - 1) * R2 * (1 - sqrt(1 - rho.^2 / R2^2));

% Circular aperture mask

cir = sqrt(X.^2 + Y.^2) <= 0.022e-3; % lens radius

Because we want to represent the lens in a wave-optics sense, the lens may also be a simple
multiplicative phase change of the field:

% Compute the multiplicative phase transformation

t_l = exp(1j * phi);

The field after the lens U1 is simply

 10

U1= ta *U

where U= field in front of the lens = 1

ta = t_l .* cir;

U1 = ta;

angle(U1) returns the phase angle in the interval [- for each element of the complex
array U1.

Colorbar display the current colormap and indicate the mapping of data values into the
colormap.

colormap hsv return the colormap with m color.

% Plot the phase transformation

figure;

imagesc(angle(U1)); % plot the phase

colorbar;
colormap hsv;
title('Phase Transformation Introduced by the Lens');
xlabel('x (meters)');
ylabel('y (meters)');
 11

In OpticStudio, the Metasurface is represented using the diffractive surface type, which introduces
an additional phase modulation over the base refractive or reflective surface. This phase profile
induces precise ray bending effects. In this report, I employ the Binary 2 surface type for modeling,
specifically utilizing a least squares curve fitting approach. By employing this technique, the
designed metalens can effectively manipulate light, ensuring optimal performance across various
optical applications.

Curve Fitting:
For curve fitting I use least square fitting approach, the least squares curve fitting method is chosen
for its ability to accurately model complex phase profiles using polynomial expansions, ensuring
that the designed metalens functions effectively in manipulating light as intended for various
optical applications.

Why Choose Least Squares Curve Fitting?

1. Minimization of Residuals: Least squares fitting minimizes the sum of the squares of the
differences between the observed values (phase data) and the values predicted by the
polynomial curve (based on your fitted coefficients). This approach ensures that the fitted
curve is as close as possible to the actual data points.

2. Flexibility with Polynomial Orders: It allows you to fit polynomials of various degrees,
giving flexibility to approximate complex phase profiles with higher-order polynomial
terms like 𝝆𝟐 .

Steps Involved in Least Squares Curve Fitting:

1. Normalization: Initially, the 𝝆𝟐 values are normalized within a specified radius (e.g., 0.02
mm in y case). This normalization step scales the values relative to the radius, ensuring
consistent treatment across different datasets or regions of interest.
 12

2. Masking: A logical mask is created to select data points within the defined normalization
radius. This ensures that the curve fitting focuses on the relevant region of interest where
the phase profile needs to be accurately represented.

3. Vectorization: The normalized 𝝆𝟐 values are flattened into a vector, alongside the
corresponding phase values ϕ. This prepares the data for polynomial fitting by organizing
it into suitable input-output pairs.

4. Polynomial Fitting: Using the polyfit function, a polynomial of a specified degree (e.g.,
5th order) is fitted to the normalized data. The function calculates the coefficients that best
fit the data in a least-squares sense, ensuring that the polynomial closely matches the phase
profile over the selected range of 𝝆𝟐 values.

5. Coefficient Adjustment: After obtaining the coefficients, they are typically reversed
(using flip) to align with the standard polynomial representation (e.g., highest degree first).
This adjustment is necessary because the binary optic 2 surface applies phase contributions
based on descending powers of 𝝆𝟐 .

Computational approach for curve fitting:

The reason for computing we need to fit the given equation into a form This
means we want to express as a polynomial in terms of even powers of .

% Compute rho^2
rho2 = rho.^2;

1. norm_radius = 0.02e-3; This sets the normalization radius to 0.2 mm

2. mask = rho <= norm_radius: This creates a logical mask array where the condition is true
(1) for elements of rho that are less than or equal to norm_radius, and false (0) otherwise.
3. rho2_normalized = rho2(mask) / norm_radius^2; This step performs the normalization
of the elements of rho2 within the specified radius. Specifically:

• rho2(mask) selects the elements of rho2 that correspond to rho values within the
norm_radius.
• These selected values are then divided by the square of norm_radius to normalize them.
 13

Purpose: This normalization is typically used to scale rho2 values within a specific radius to a
common scale. This approach is useful when focusing on a particular region and normalizing its
values based on a defined radius.

% Normalize rho^2 within a radius of 5 mm

norm_radius = 0.02e-3; % 0.02 mm
mask = rho <= norm_radius;
rho2_normalized = rho2(mask) / norm_radius^2;

% Flatten the phi matrix to a vector and apply the same mask
phi_vec = phi(mask);

polyfit(x, y, n) returns the coefficients for a polynomial p(x) of degree n that is a best fit (in a least-
squares sense) for the data in y. The coefficients in p are in descending powers, and the length of
p is n+1 where

n values goes from 0 to n.

% Perform polynomial fitting on normalized data to 4th order

coeffs_normalized = polyfit(rho2_normalized, phi_vec, 5);

flipping is required because we want to get the result in the form of increasing powers like in

P = flip(coeffs_normalized);

As we all know that the binary optic 2 surface adds phase to the ray according to the following
polynomial expansion: .so the first term of our result has no contribution to the binary
surface.

% Display the coefficients

disp('Polynomial coefficients:');
disp(P);
for i=2:length(P)
disp(P(i))
end
 14

Polynomial coefficients:

-10.471283030631579

-0.100237788504564

Hence the coefficient of 𝝆𝟐 = −𝟏𝟎. 𝟒𝟕𝟏𝟐𝟖𝟑𝟎𝟑𝟎𝟔𝟑𝟏𝟓𝟕𝟗

And 𝝆𝟒 = −𝟎. 𝟏𝟎𝟎𝟐𝟑𝟕𝟕𝟖𝟖𝟓𝟎𝟒𝟓𝟔𝟒

Now verify this result in the OpticStudio

Modeled Plano convex lens:

Fig: lens data of a plano convex lens having entrance pupil diameter 0.022mm and Effl 0.2mm
 15

fig: lens layout

 16

fig: FFT MTF

 17

Binary equivalent surface:

After putting the coefficient for the binary 2 surface and keeping normalization radius same as
that of taken in MATLAB code.

Polynomial
coefficients:

-10.471283030631

-0.100237788504564

Fig: lens data

Fig: layout
 18

Fig: FFT MTF

As we can see the analysis of the plano-convex lens and its binary equivalent yielded identical
results, confirming the accuracy of the MATLAB code used to determine the binary surface
coefficients. The code effectively computes the coefficients necessary for the design of a binary 2
surface that mimics the phase profile of a plano-convex lens. This validation ensures that the
computational methods and algorithms implemented in the code are reliable and can be utilized
for further optical system design and analysis involving binary optical elements.
 19

Fourier Transforming Property of Thick lens

Now, I will discuss the very important 2D Fourier transforming property of a lens which is key to
several imaging and optical processing applications. I will consider a configuration involving a
lens and free-space propagation as shown in Fig: An object transparency with transmittance t(x’,
y’ ) is placed at a distance d in front of the lens. The output field is observed at the just after the
lens.

Assuming illumination of the object by a plane wave propagating along the optic axis, the
calculation of the output g(x, y) in terms of the input t(x’ , y’) involves the following steps:

1. Fresnel propagation from input plane to the lens.

2. Multiplication by lens transmission function.

g(x, y)

Fig: Fourier transforming of a thick lens

A common propagation routine is based on equation

And uses the transfer function given by

 20

Computation approach for Fresnel propagation involves creating .M file in MATLAB, name
“PropTF”
 21

Optical Imaging system

Having derived the Fourier transforming property of a convex lens, we will now consider an
important imaging system—Combination of two lenses. This is a general imaging system model
which may be used to describe a large number of imaging systems with minor modifications. The
system of interest is as shown in Fig.

Fig: combination of two lens layout

In this layout, the input incidence field Ulens is initially set to 1. The field just after the first lens,
referred to as U1, has been calculated earlier on page 9 of this report. Here, Z represents the
propagation distance, and the wave propagation over this distance Z is determined using the
Fresnel propagation method discussed previously.

To obtain the field information just after the second lens, or more importantly, to determine the
combined phase profile of the system, we can replicate this lens system using a binary equivalent
surface. This replication is essential for creating a Metalens with the desired combined phase
profile, allowing us to achieve the intended optical performance with a single, compact optical
element.
 22

COMPUTATIONAL APPROACH:

format long
% Define parameters
lambda = 550e-9; % wavelength in meters (550 nm)
k = 2 * pi / lambda; % wavenumber
n = 1.5168; % refractive index of the lens material (BK7)
A0 = 10e-3; % maximum thickness of the lens
R1 = 59.320e-3; % radius of curvature of the first lens surface
R2 = 350.120e-3; % radius of curvature of the second lens surface

% Define the lens thickness profile A(x, y)

[X, Y] = meshgrid(-10e-3:5e-4:10e-3, -10e-3:5e-4:10e-3); % 20mm x 20mm area
with 0.5 mm step

% Circular aperture mask

cir = sqrt(X.^2 + Y.^2) <= 10e-3;

% Lens thickness profile

A1 = R1 * (1 - sqrt(1 - (X.^2 + Y.^2) / R1^2));
A2 = R2 * (1 - sqrt(1 - (X.^2 + Y.^2) / R2^2));
A = A0 - A1 + A2;

% Calculate the total phase delay suffered by the wave at coordinate (X,Y)
% in passing through the lens

phi_total = k * (A0 + (n - 1) * A);

Because we want to represent the lens in a wave-optics sense, the lens may also be a simple
multiplicative phase change of the field:

% Compute the multiplicative phase transformation

tLens = exp(1j * phi_total);

The field after the lens U′ Lens is simply

U'Lens= tLens*Ulens

where Ulens= field in front of the lens

ta = tLens .* cir;
U1 = ta; %field behind the lens
 23

angle(U1) returns the phase angle in the interval [- for each element of the complex array
U1.

Colorbar display the current colormap and indicate the mapping of data values into the
colormap.

colormap hsv return the colormap with m color.

% Plot the phase transformation

figure;
imagesc(angle(U1)) % plot the phase
colormap hsv;
colorbar;
title('Phase Transformation Introduced by the Lens');
xlabel('x (meters)');
ylabel('y (meters)');

Propagation phase delay

% Define propagation distance

z = 1e-3; % propagation distance in meters
u2=propFF(U1,20e-3,lambda,z); %called propTF function and stored in
variable u2.
%u2 is the propagated field
%phase of the propagated field
phi_del=angle(u2);
imagesc(phi_del)
title('Propagation phase delay');
 24

xlabel('x (meters)');
ylabel('y (meters)')

colormap hsv;
colorbar;

Phase just after the lens two

% 2nd lens surface

R3 = 59.32e-3; % radius of curvature of the first surface of 2nd lens
R4 = 350.120e-3; % radius of curvature of the second surface of 2nd
lens

A3 = R3 * (1 - sqrt(1 - (X.^2 + Y.^2) / R3^2));

A4 = R4 * (1 - sqrt(1 - (X.^2 + Y.^2) / R4^2));
A11 = A0 - A3 + A4;

% Calculate the total phase delay

phi_total1 = k * (A0 + (n - 1) * A11);

ta=exp(1i*phi_total1).*u2.*cir;
phi3=angle(ta);
 25

imagesc(phi3)
title('Phase Transformation Introduced by second the Lens');
xlabel('x (meters)');
ylabel('y (meters)')

% Compute final phase just after the 2nd lens

phi_final = phi_total + phi_del + phi_total1;

curve fitting

norm_radius = 5e-3; define a variable norm_radius and assigned its value 5 mm

mask = rho <= norm_radius: This creates a logical mask array where the condition is true (1) for
elements of rho that are less than or equal to norm_radius, and false (0) otherwise.

rho2_normalized = rho2(mask) / norm_radius^2; This step performs the normalization of the

elements of rho2 within the specified radius. Specifically:

• rho2(mask) selects the elements of rho2 that correspond to rho values within the
norm_radius.
• These selected values are then divided by the square of norm_radius to normalize them.
 26

Purpose: This normalization is typically used to scale rho2 values within a specific radius to a
common scale. This approach is useful when focusing on a particular region and normalizing its
values based on a defined radius.

% Compute rho
rho = sqrt(X.^2 + Y.^2);
% Normalize rho^2 within a radius of 5 mm
norm_radius = 5e-3;
mask = rho <= norm_radius;
rho2 = rho.^2;
rho2_normalized = rho2(mask) / norm_radius^2;

% Flatten the phi matrix to a vector and apply the same mask
phi_vec = phi_final(mask);

% Perform polynomial fitting on normalized data to 2nd order

coeffs_normalized = polyfit(rho2_normalized, phi_vec, 2);
P = flip(coeffs_normalized);

% Display the coefficients

disp('Polynomial coefficients:');
disp(P);
for i=2:length(P)
disp(P(i))
end

Polynomial coefficients:

-2.067678649212336e+03

-2.752913670422028

Hence the coefficient of 𝝆𝟐 = −𝟐. 𝟎𝟔𝟕𝟔𝟕𝟖𝟔𝟒𝟗𝟐𝟏𝟐𝟑𝟑𝟔𝐞 + 𝟎𝟑

And 𝝆𝟒 = −2.752913670422028
 27

Now verify the result in ZEMAX OpticStudio:

Two lens combination having EFFL=70:

Fig: lens data

fig: Spot diagram

 28

fig: FFT MTF

fig: Lens Layout

 29

Binary equivalent surface to replicate this lens system by substituting the

coefficient of rho’s

Fig: lens data

Fig: substituting binary phase coefficient which is extracted from MATLAB

 30

fig: SPOT Diagram

fig: FFT MTF

 31

Fig: Binary equivalent layout

Conclusion
This project successfully explored the phase characteristics of multi-lens optical systems through
detailed phase analysis and curve fitting for binary surface modeling. By utilizing specific lens
parameters, the phase delay introduced by individual lenses was computed and subsequently
curve-fitted to obtain binary surface coefficients. These coefficients were used to create a binary
equivalent of the lens, which was validated using Zemax software.

The comparative analysis between the original lens data and the binary surface approximation
showed similar phase profiles, demonstrating the effectiveness of the binary surface modeling
approach. However, discrepancies in the spot diagrams and MTF results indicated areas for further
refinement. These differences, particularly the defocus observed in the MTF of the binary
equivalent surface, suggest the need for more precise modeling techniques or additional
considerations in the binary surface approximation.

Overall, this work provides a robust framework for phase analysis and binary surface modeling in
optical systems, contributing valuable insights into the design and simulation of complex lens
systems. Future research could focus on refining the curve fitting process and exploring advanced
modeling techniques to enhance the accuracy of binary surface approximation
 32

References

1. Introduction to Fourier Optics: Goodman

2. Computational Fourier Optics: SPIE

3. Fourier Optics and Computational Imaging: Springer

4. Introduction to aberration and optical Imaging system

5. Zemax user Manual