Multi-objective, robust constraints enforced global topology optimizer for optical devices

Patent metadata

Field	Value
Publication number	US-11796794-B2
Application number	US-202117317624-A
Country	US
Kind code	B2
Filing date	May 11, 2021
Priority date	May 12, 2020
Publication date	Oct 24, 2023
Grant date	Oct 24, 2023

Abstract

A method for optimization of photonic devices is disclosed. The method includes receiving a set of unconstrained latent variables; mapping the set of unconstrained latent variables to a constrained space to generate a constrained device; calculating the permittivity across each element of the constrained device; determining a permittivity-constrained width gradient based at least partially on the permittivity across each element; and optimizing the set of unconstrained latent variables by at least partially using the permittivity-constrained width gradient.

First claim

What is claimed is:

1. A method of manufacturing photonic devices, the method comprising:
using a generator to generate a set of unconstrained latent variables;
mapping the set of unconstrained latent variables to a constrained space to generate a constrained photonic device;
calculating the permittivity across each element of the constrained photonic device;
determining a permittivity-constrained variable gradient based at least partially on the permittivity across each element of the constrained photonic device;
iteratively updating the generator at least partially based on the permittivity-constrained variable gradient;
generating using the generator an optimized set of unconstrained latent variables; and
manufacturing the generated constrained photonic device.

2. The method of claim 1 , wherein mapping the unconstrained space to a constrained space comprises reparameterization of the unconstrained latent variables.

3. The method of claim 2 , wherein the reparameterization of the unconstrained latent variables comprises solving for a set of constrained width space {w i } through the following equations:

v
i

=

Sigmoid
(

u
i

)

,

v
M

=
1

;

⁢

k
i

=

v
i

⁢

1
-

∑

j
=
1

i
-
1

k
j
2

,

k
1

=

v
1

;
and

⁢

w
i

=

k
i

∑

j
=
1

M

k
j

⁢

(

L
-

Mw
min

)

+

w
min

,

wherein the set of unconstrained latent variables are {u i } and wherein w min is a minimum width of each feature of the constrained width space, M is the total number of features, and Σ i=1 M w i =L.

4. The method of claim 3 wherein calculating the permittivity across each element of the constrained device comprises performing a gray-space relaxation of the set of constrained width space to determine a function of refractive index across each element in the set of constrained width space.

5. The method of claim 4 , wherein the gray-space relaxation comprises solving for

n
i

(
x
)

=

[

exp
(

±

x
2

-

(

w
i

2

)

2

T

)

+
1

]

-
1

,

wherein x is the position across the width of a certain element, where x=0 is at the middle of the element, and T controls the binarization of the pattern.

6. The method of claim 4 , further comprising determining an efficiency-constrained variable gradient of the constrained device based on the permittivity-constrained variable gradient.

7. The method of claim 6 , wherein the efficiency-constrained variable gradient is further based on an efficiency-permittivity gradient.

8. The method of claim 7 , wherein the efficiency-constrained variable gradient is calculated using the following equation:

∂
Eff

Assignees

• Univ Leland Stanford Junior

Inventors

• Fan Jonathan
• Jiang Jiaqi
• Chen Mingkun

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0475

  Generative networks · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G02B27/0012 Primary

  Optical design, e.g. procedures, algorithms, optimisation routines · CPC title

• G01M11/0228

  by measuring refractive power · CPC title