OSA's Digital Library

Optics Express

Optics Express

  • Editor: Andrew M. Weiner
  • Vol. 21, Iss. 18 — Sep. 9, 2013
  • pp: 21693–21701
« Show journal navigation

Adjoint shape optimization applied to electromagnetic design

Christopher M. Lalau-Keraly, Samarth Bhargava, Owen D. Miller, and Eli Yablonovitch  »View Author Affiliations


Optics Express, Vol. 21, Issue 18, pp. 21693-21701 (2013)
http://dx.doi.org/10.1364/OE.21.021693


View Full Text Article

Acrobat PDF (1083 KB)





Browse Journals / Lookup Meetings

Browse by Journal and Year


   


Lookup Conference Papers

Close Browse Journals / Lookup Meetings

Article Tools

Share
Citations

Abstract

We present an adjoint-based optimization for electromagnetic design. It embeds commercial Maxwell solvers within a steepest-descent inverse-design optimization algorithm. The adjoint approach calculates shape derivatives at all points in space, but requires only two “forward” simulations. Geometrical shape parameterization is by the level set method. Our adjoint design optimization is applied to a Silicon photonics Y-junction splitter that had previously been investigated by stochastic methods. Owing to the speed of calculating shape derivatives within the adjoint method, convergence is much faster, within a larger design space. This is an extremely efficient method for the design of complex electromagnetic components.

© 2013 OSA

1. Introduction and motivations

Silicon photonics offer the unique ability of managing light through sub-wavelength Silicon waveguides patterned on chip, enabling extremely tight integration of photonic components and conventional CMOS electronics. Consequently, functions that previously required many separate components may now all be performed on single chips, reducing their cost, energy consumption and size [1

1. P. Sandborn, N. Quack, N. Hoghooghi, J. B. Chou, J. Ferrara, S. Gambini, B. Behroozpour, L. Zhu, B. Boser, C. Chang-Hasnain, and M. C. Wu, “Linear frequency chirp generation employing opto-electronic feedback loop and integrated Silicon photonics,” in CLEO, (2013) pp. 5–6.

].

Nevertheless a number of challenges remain, one of them being the efficient management of light at these scales. Indeed, while straight Silicon waveguides can have extremely low loss and enable excellent transport of light throughout the chip, other functions (such as splitters, waveguide crossings, multimode interferometers) suffer from the presence of evanescent fields outside the waveguide and imperfect reflections at the Silicon/oxide interface. These induce scattering loss, which can be highly detrimental to the total system performance. For this reason, a significant effort in photonic device topology optimization has taken place recently. This has drastically reduced the losses in Y-splitters [2

2. A. Sakai, T. Fukazawa, and T. Baba, “Low loss ultra-small branches in a silicon photonic wire waveguide,” IEICE Trans. Electron. E85, 1033–1038 (2002).

,3

3. Y. Zhang, S. Yang, A. E. J. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef] [PubMed]

], crosstalk and insertion losses in waveguide crossings [3

3. Y. Zhang, S. Yang, A. E. J. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef] [PubMed]

,4

4. P. Sanchis, P. Villalba, F. Cuesta, A. Håkansson, A. Griol, J. V. Galán, A. Brimont, and J. Martí, “Highly efficient crossing structure for silicon-on-insulator waveguides,” Opt. Lett. 34(18), 2760–2762 (2009). [CrossRef] [PubMed]

], along with other more exotic components [4

4. P. Sanchis, P. Villalba, F. Cuesta, A. Håkansson, A. Griol, J. V. Galán, A. Brimont, and J. Martí, “Highly efficient crossing structure for silicon-on-insulator waveguides,” Opt. Lett. 34(18), 2760–2762 (2009). [CrossRef] [PubMed]

] and is effectively enabling better photonic circuits.

Most of these optimizations are based on heuristic optimization methods such as genetic optimization [4

4. P. Sanchis, P. Villalba, F. Cuesta, A. Håkansson, A. Griol, J. V. Galán, A. Brimont, and J. Martí, “Highly efficient crossing structure for silicon-on-insulator waveguides,” Opt. Lett. 34(18), 2760–2762 (2009). [CrossRef] [PubMed]

], particle swarm optimization [3

3. Y. Zhang, S. Yang, A. E. J. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef] [PubMed]

,5

5. Y. Zhang, S. Yang, E. J. Lim, G. Lo, T. Baehr-Jones, and M. Hochberg, “A CMOS-compatible, low-loss and low-crosstalk silicon waveguide crossing,” IEEE Photon. Technol. Lett. 25(5), 422–425 (2013). [CrossRef]

], or other hybrid methods tailored for specific problems [6

6. T. Tanemura, K. C. Balram, D. S. Ly-Gagnon, P. Wahl, J. S. White, M. L. Brongersma, and D. A. B. Miller, “Multiple-wavelength focusing of surface plasmons with a nonperiodic nanoslit coupler,” Nano Lett. 11(7), 2693–2698 (2011). [CrossRef] [PubMed]

]. Heuristic optimization relies on a somewhat limited parameterization of the solution space and subsequent random testing of a large number of different parameter sets. Because of the high computational cost of solving Maxwell’s equations, these optimization methods may only be applied to relatively simple geometries, as they require the testing a very large number of different solutions in order to find a satisfactory one.

While this is perfectly suitable for the simple problems mentioned above, these methods will fail to perform in a reasonable amount of time for more complex geometries and functions. It is therefore necessary to have a more efficient way of performing topology optimization for general purposes. In our shape optimization approach, shape derivatives play an important role. In this paper we present an adjoint method to calculate shape derivatives by wrapping an inverse algorithm around commercial Maxwell solvers. Such efficient gradient descent methods unlock the possibility to optimize particularly complex structures, which has not previously been possible.

2. Presentation of the adjoint method for electromagnetic problems

The adjoint method enables the computation of shape derivatives at all points in space, with only two electromagnetic simulations per iteration. It has been extensively used for shape optimization in mechanical engineering [7

7. M. P. Bendsøe and N. Kikuchi, “Generating optimal topologies in structural design using a homogenization method,” Comput. Meth. In Appl. M. 71, 197–224 (1988).

9

9. T. Borrvall and J. Petersson, “Topology optimization of fluids in Stokes flow,” Int. J. Numer. Methods Fluids 41(1), 77–107 (2003). [CrossRef]

] but has seen more limited use for photonic components [10

10. J. S. Jensen and O. Sigmund, “Topology optimization for nano-photonics,” Las. Photon. Rev. 5(2), 308–321 (2011). [CrossRef]

14

14. G. Veronis, R. W. Dutton, and S. Fan, “Method for sensitivity analysis of photonic crystal devices,” Opt. Lett. 29(19), 2288–2290 (2004). [CrossRef] [PubMed]

], and more recently quantum electronics [15

15. A. F. J. Levi and I. G. Rosen, “A novel formulation of the adjoint method in the optimal design of quantum electronic devices,” SIAM J. Contr. Optim. 48(5), 3191–3223 (2010). [CrossRef]

]. Mathematical derivations of the adjoint method are available in optimization textbooks [8

8. M. P. Bendsoe and O. Sigmund, Topology Optimization Theory, Methods and Applications. (Springer, 2003)

,16

16. G. Strang, Computational Science and Engineering , (Wellesley-Cambridge, 2007).

], but we will limit ourselves to a very simple example that intuitively illustrates the mathematical procedure when it is used in the context of electromagnetism.

In our example, we want to maximize the absolute value of the electric field at a given point x0, given a geometrical region Ω in which we can change the electric permittivityε at every point. That Figure-of-Merit is
FoM=|E(x0)|2
(1)
(vectors are written in bold). The change in figure of merit for a small change of dielectric permittivity Δεr of volume ΔV at x in Ω is
ΔFoM=Re[Eold(x0)¯ΔE(x0)]
(2)
where Eold(x0)is the value of the electric field at a given point before any change and ΔE(x0) represents the change in electric field when the small dielectric modification is performed.

Some algebraic manipulations are needed to arrive at the derivative. The change in field at x0 can be written for a small enough volume perturbation ΔV:
ΔE(x0)=GEP¯¯(x0,x)pind=ε0ΔεrΔVGEP¯¯(x0,x)Enew(x)
(3)
where GEP¯¯(x0,x)is the Maxwell Green’s function relating the electric field at x0 to the induced polarization density pindat x in the infinitesimal volume ΔV. Enew is the electric field given the new dielectric distribution. If the change Δεr is small enough, we may approximateEnew(x)Eold(x). Note that for binary structures Δεr is not small, but ΔVcan be the small parameter for the derivative. A similar line of reasoning results in almost the same final equation, albeit taking care to distinguish which components of E and D are continuous across the boundary [14

14. G. Veronis, R. W. Dutton, and S. Fan, “Method for sensitivity analysis of photonic crystal devices,” Opt. Lett. 29(19), 2288–2290 (2004). [CrossRef] [PubMed]

,17

17. O. D. Miller, PhD thesis (2012), University of California at Berkeley, http://optoelectronics.eecs.berkeley.edu/ThesisOwenMiller.pdf

,18

18. S. G. Johnson, M. Ibanescu, M. A. Skorobogatiy, O. Weisberg, J. D. Joannopoulos, and Y. Fink, “Perturbation theory for Maxwell’s equations with shifting material boundaries,” Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 65(6), 066611 (2002). [CrossRef] [PubMed]

].

We can then rewrite (2) as
ΔFoMΔεr=ε0ΔVRe[Eold(x0)¯(GEP¯¯(x0,x)Eold(x))]
(4)
Using the reciprocity of the Green’s function GEP¯¯(x0,x)=GEP¯¯(x,x0)T we have
ΔFoMΔεr=Re[(ε0ΔVGEP¯¯(x,x0)Eold(x0)¯)Eold(x)]Re[Eadj(x)Eold(x)]
(5)
The mathematical method can be understood from the new adjoint electric field:
Eadj(x)=ε0ΔVGEP¯¯(x,x0)Eold(x0)¯
(6)
which is the electrical field induced at x from an electric dipole at x0driven with amplitude ΔVε0Eold(x0)¯, as illustrated in Fig. 1
Fig. 1 Adjoint method schematic: two simulations are needed for every iteration; the direct and the adjoint simulation. Sources for each simulation are drawn in red.
. Thus, the gradient of the Figure-of-Merit can be obtained with only a single simulation, even though it provides the derivative with respect to permittivity at every point in the computational region Ω. The term Eold(x0) is readily available from the original forward simulation.

Therefore with just one forward simulation (which is needed to calculate the FoM in all optimization schemes) plus one adjoint simulation, the shape derivative can be obtained over the entire design region, for arbitrarily many degrees of freedom. With the gradient of the Figure-of-Merit calculated, changes in the geometry can be introduced proportional to the gradient, known as the gradient descent method. Applied iteratively, this can then lead to an optimum. For a more detailed and general study of the adjoint method and more complex Figures-of-Merit we refer to [17

17. O. D. Miller, PhD thesis (2012), University of California at Berkeley, http://optoelectronics.eecs.berkeley.edu/ThesisOwenMiller.pdf

].

The adjoint method is also extremely attractive since the overall iterative scheme can be wrapped around a commercial forward solver, such as the one used in [19

19. Lumerical FDTD Solutions, www.lumerical.com

].

3. Y-Splitter optimization example using the level set method for shape representation

A Y-splitter for λ = 1550nm vacuum wavelength light was optimized by the adjoint method to compare with state of the art Silicon photonic components optimized up to date [3

3. Y. Zhang, S. Yang, A. E. J. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef] [PubMed]

]. The material system (Silicon waveguide, Silicon dioxide cladding) and the constraints of small overall dimensions and minimum feature size were kept the same as in [3

3. Y. Zhang, S. Yang, A. E. J. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef] [PubMed]

]. For the minimum feature size a minimum radius of curvature of 200nm was imposed. The waveguide is 220nm thick, the most common choice for Silicon photonics. The two waveguide branches and their junction at the end of the splitter were left to be the same as in [3

3. Y. Zhang, S. Yang, A. E. J. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef] [PubMed]

], although they also could have easily been optimized. The design region was the central 2µm × 2µm domain.

Fig. 2 Top view of the optimized silicon splitter geometry obtained after 51 iterations of the Steepest Descent algorithm. Only the designable region geometry was allowed to change. The Silicon waveguide is 220nm thick, and the cladding is Silicon dioxide.

By contrast, the adjoint method provides shape derivatives over the entire design region. The level set method, developed by Sethian and Osher [20

20. S. Osher and J. A. Sethian, “Fronts propagating with curvature-dependent speed: algorithms based on Hamilton-Jacobi formulations,” J. Comput. Phys. 79(1), 12–49 (1988). [CrossRef]

], was chosen to represent the geometry. This enables a more flexible representation of a larger design space than, for example, spline interpolations used in [3

3. Y. Zhang, S. Yang, A. E. J. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef] [PubMed]

5

5. Y. Zhang, S. Yang, E. J. Lim, G. Lo, T. Baehr-Jones, and M. Hochberg, “A CMOS-compatible, low-loss and low-crosstalk silicon waveguide crossing,” IEEE Photon. Technol. Lett. 25(5), 422–425 (2013). [CrossRef]

]. Level sets are particularly usable within an adjoint approach, since a very large number of shape derivatives are inside the Level Set, compared to the feasible number of variables in stochastic optimization. Note also that level set methods impose two-phase, binary materials throughout the optimization, compatible with practical engineering, but in contrast with [10

10. J. S. Jensen and O. Sigmund, “Topology optimization for nano-photonics,” Las. Photon. Rev. 5(2), 308–321 (2011). [CrossRef]

], which optimizes a continuously variable permittivity.

The figure-of-merit that we employed was transmission into the fundamental mode of the bent output waveguides, which can be obtained from Poynting vectors:
FoM=18|E×Hm¯dS+Em¯×HdS|2Re(Em×Hm¯)dS
(7)
where Emand Hm are the field profiles of the fundamental mode at the surface S, while E and H are the actual fields from the direct simulation at that surface. Thus Eq. (7) is the power transmission, corrected for the mode overlap.

Adapting the adjoint Eq. (6) to the new figure of merit (7) (and employing an additional magnetic Green’s function, GEM, and magnetic symmetries [17

17. O. D. Miller, PhD thesis (2012), University of California at Berkeley, http://optoelectronics.eecs.berkeley.edu/ThesisOwenMiller.pdf

]), we arrive at the adjoint field:
Eadj(x)=A(GEP¯¯(x,x')Hm(x')¯×nGEM¯¯(x,x')n×Em(x')¯μ0)dS
(8)
with
A=14ε0ΔV(Eold×Hm¯dS+Em¯×HolddS)¯Re(Em×Hm¯)dS
(9)
where GEM¯¯(x,x') is the electromagnetic Green’s function expressing the electric field at x due to a magnetic dipole at x'.

The adjoint simulation described Eq. (8) consists of sending the desired mode backwards into the splitter. This is analogous to Eq. (6), where the adjoint source was located at the measurement point of the Figure-of-Merit. This source problem can be solved with a standard Maxwell solver. FDTD is perfectly suited for this propagating wave problem. Also analogous to Eq (6) the phase of the adjoint source is set using Eoldand Hold, from the forward simulation, as described in Eq (9). Once the adjoint simulation is performed, the derivative of the Figure-of-Merit with respect to dielectric permittivity at every point in the design region is calculated by combining the forward and adjoint simulations results into Eq. (5). FDTD is perfectly suited to solve the direct and adjoint problem, which consists of propagating waves in a dielectric.

This derivative is then used to modify the geometry of the splitter. Since we employed a level set description of geometry, the derivative is used as a velocity field to modify the level set shape. This has the effect of pushing out the geometry boundary when the derivative is positive and pushing it in when it is negative. Since the refractive index if Silicon is higher than that of Silicon dioxide this implements the imperative of the derivative at every point: The Figure-of-Merit benefits from an increase in the dielectric permittivity where the derivative is positive and vice-versa. The step-size criterion for each iteration is a fixed area of changing type in 2d, and a fixed volume in 3d.

The device was first optimized using 2d finite difference time domain (FDTD) simulations of a structure extruded infinitely in the 3rd dimension. In 2d, the effective index method is used and the Silicon is assigned the fictitious refractive index = 2.8, which mimics the proper in-plane wavevector of the correct 3d mode. Once iterative progress stopped in 2d (41 iterations), the problem was transferred to 3d for more iterations. Naturally the first 3d iteration is not as good as the optimized 2d device, since the effective index method is only an approximation. The optimal structure was computed within 51 iterations (102 simulations), achieving a record low insertion loss −0.07dB. By comparison, ref [3

3. Y. Zhang, S. Yang, A. E. J. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef] [PubMed]

]. achieved a minimal insertion loss −0.13dB, after 1500 simulations using particle swarm optimization. (Note that for such small attenuation, the simulation results are very sensitive to the simulation parameters, which may not have been perfectly identical to ref [3

3. Y. Zhang, S. Yang, A. E. J. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef] [PubMed]

].).

Thus adjoint steepest descent, with much lower computational cost, can yield as good or better results than particle swarm optimizations, which take no advantage of the underlying Maxwell equation physics.

4. Conclusion

As photonic and wireless components become an increasingly important part of electronics, it is evident that many problems will require electromagnetic optimization. The computational cost of solving Maxwell’s equations is significant, and inefficient design optimization algorithms will become unacceptable. We have shown that the adjoint gradient decent method for shape optimization of sub-wavelength photonic devices can be readily implemented by embedding commercial Maxwell solvers within an inverse optimization algorithm.

For exploration of larger solution spaces where local optima may exist, this method may be augmented with a clever choice of Figure-of-Merit, as well as global optimization routines such as simulated annealing to provide efficient and powerful automated design of photonic components.

Acknowledgments

This work is supported by the Defense Advanced Research Projects Agency (DARPA) E-PHI program under Grant No. HR0011-11-2-0021, the NSF E3S Center under NSF award 0939514, and the U.S. Department of Energy “Light–Material Interactions in Energy Conversion” Energy Frontier Research Center under Grant DE-SC0001293.

References and links

1.

P. Sandborn, N. Quack, N. Hoghooghi, J. B. Chou, J. Ferrara, S. Gambini, B. Behroozpour, L. Zhu, B. Boser, C. Chang-Hasnain, and M. C. Wu, “Linear frequency chirp generation employing opto-electronic feedback loop and integrated Silicon photonics,” in CLEO, (2013) pp. 5–6.

2.

A. Sakai, T. Fukazawa, and T. Baba, “Low loss ultra-small branches in a silicon photonic wire waveguide,” IEICE Trans. Electron. E85, 1033–1038 (2002).

3.

Y. Zhang, S. Yang, A. E. J. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express 21(1), 1310–1316 (2013). [CrossRef] [PubMed]

4.

P. Sanchis, P. Villalba, F. Cuesta, A. Håkansson, A. Griol, J. V. Galán, A. Brimont, and J. Martí, “Highly efficient crossing structure for silicon-on-insulator waveguides,” Opt. Lett. 34(18), 2760–2762 (2009). [CrossRef] [PubMed]

5.

Y. Zhang, S. Yang, E. J. Lim, G. Lo, T. Baehr-Jones, and M. Hochberg, “A CMOS-compatible, low-loss and low-crosstalk silicon waveguide crossing,” IEEE Photon. Technol. Lett. 25(5), 422–425 (2013). [CrossRef]

6.

T. Tanemura, K. C. Balram, D. S. Ly-Gagnon, P. Wahl, J. S. White, M. L. Brongersma, and D. A. B. Miller, “Multiple-wavelength focusing of surface plasmons with a nonperiodic nanoslit coupler,” Nano Lett. 11(7), 2693–2698 (2011). [CrossRef] [PubMed]

7.

M. P. Bendsøe and N. Kikuchi, “Generating optimal topologies in structural design using a homogenization method,” Comput. Meth. In Appl. M. 71, 197–224 (1988).

8.

M. P. Bendsoe and O. Sigmund, Topology Optimization Theory, Methods and Applications. (Springer, 2003)

9.

T. Borrvall and J. Petersson, “Topology optimization of fluids in Stokes flow,” Int. J. Numer. Methods Fluids 41(1), 77–107 (2003). [CrossRef]

10.

J. S. Jensen and O. Sigmund, “Topology optimization for nano-photonics,” Las. Photon. Rev. 5(2), 308–321 (2011). [CrossRef]

11.

P. Seliger, M. Mahvash, C. Wang, and A. F. J. Levi, “Optimization of aperiodic dielectric structures,” J. Appl. Phys. 100(3), 034310 (2006). [CrossRef]

12.

W. R. Frei, D. A. Tortorelli, and H. T. Johnson, “Geometry projection method for optimizing photonic nanostructures,” Opt. Lett. 32(1), 77–79 (2007). [CrossRef] [PubMed]

13.

V. Liu and S. Fan, “Compact bends for multi-mode photonic crystal waveguides with high transmission and suppressed modal crosstalk,” Opt. Express 21(7), 8069–8075 (2013). [CrossRef] [PubMed]

14.

G. Veronis, R. W. Dutton, and S. Fan, “Method for sensitivity analysis of photonic crystal devices,” Opt. Lett. 29(19), 2288–2290 (2004). [CrossRef] [PubMed]

15.

A. F. J. Levi and I. G. Rosen, “A novel formulation of the adjoint method in the optimal design of quantum electronic devices,” SIAM J. Contr. Optim. 48(5), 3191–3223 (2010). [CrossRef]

16.

G. Strang, Computational Science and Engineering , (Wellesley-Cambridge, 2007).

17.

O. D. Miller, PhD thesis (2012), University of California at Berkeley, http://optoelectronics.eecs.berkeley.edu/ThesisOwenMiller.pdf

18.

S. G. Johnson, M. Ibanescu, M. A. Skorobogatiy, O. Weisberg, J. D. Joannopoulos, and Y. Fink, “Perturbation theory for Maxwell’s equations with shifting material boundaries,” Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 65(6), 066611 (2002). [CrossRef] [PubMed]

19.

Lumerical FDTD Solutions, www.lumerical.com

20.

S. Osher and J. A. Sethian, “Fronts propagating with curvature-dependent speed: algorithms based on Hamilton-Jacobi formulations,” J. Comput. Phys. 79(1), 12–49 (1988). [CrossRef]

21.

http://optoelectronics.eecs.berkeley.edu/PhotonicInverseDesign/

OCIS Codes
(130.3120) Integrated optics : Integrated optics devices
(230.1360) Optical devices : Beam splitters
(230.7370) Optical devices : Waveguides

ToC Category:
Integrated Optics

History
Original Manuscript: July 26, 2013
Revised Manuscript: August 19, 2013
Manuscript Accepted: August 20, 2013
Published: September 6, 2013

Citation
Christopher M. Lalau-Keraly, Samarth Bhargava, Owen D. Miller, and Eli Yablonovitch, "Adjoint shape optimization applied to electromagnetic design," Opt. Express 21, 21693-21701 (2013)
http://www.opticsinfobase.org/oe/abstract.cfm?URI=oe-21-18-21693


Sort:  Author  |  Year  |  Journal  |  Reset  

References

  1. P. Sandborn, N. Quack, N. Hoghooghi, J. B. Chou, J. Ferrara, S. Gambini, B. Behroozpour, L. Zhu, B. Boser, C. Chang-Hasnain, and M. C. Wu, “Linear frequency chirp generation employing opto-electronic feedback loop and integrated Silicon photonics,” in CLEO, (2013) pp. 5–6.
  2. A. Sakai, T. Fukazawa, and T. Baba, “Low loss ultra-small branches in a silicon photonic wire waveguide,” IEICE Trans. Electron.E85, 1033–1038 (2002).
  3. Y. Zhang, S. Yang, A. E. J. Lim, G. Q. Lo, C. Galland, T. Baehr-Jones, and M. Hochberg, “A compact and low loss Y-junction for submicron silicon waveguide,” Opt. Express21(1), 1310–1316 (2013). [CrossRef] [PubMed]
  4. P. Sanchis, P. Villalba, F. Cuesta, A. Håkansson, A. Griol, J. V. Galán, A. Brimont, and J. Martí, “Highly efficient crossing structure for silicon-on-insulator waveguides,” Opt. Lett.34(18), 2760–2762 (2009). [CrossRef] [PubMed]
  5. Y. Zhang, S. Yang, E. J. Lim, G. Lo, T. Baehr-Jones, and M. Hochberg, “A CMOS-compatible, low-loss and low-crosstalk silicon waveguide crossing,” IEEE Photon. Technol. Lett.25(5), 422–425 (2013). [CrossRef]
  6. T. Tanemura, K. C. Balram, D. S. Ly-Gagnon, P. Wahl, J. S. White, M. L. Brongersma, and D. A. B. Miller, “Multiple-wavelength focusing of surface plasmons with a nonperiodic nanoslit coupler,” Nano Lett.11(7), 2693–2698 (2011). [CrossRef] [PubMed]
  7. M. P. Bendsøe and N. Kikuchi, “Generating optimal topologies in structural design using a homogenization method,” Comput. Meth. In Appl. M. 71, 197–224 (1988).
  8. M. P. Bendsoe and O. Sigmund, Topology Optimization Theory, Methods and Applications. (Springer, 2003)
  9. T. Borrvall and J. Petersson, “Topology optimization of fluids in Stokes flow,” Int. J. Numer. Methods Fluids41(1), 77–107 (2003). [CrossRef]
  10. J. S. Jensen and O. Sigmund, “Topology optimization for nano-photonics,” Las. Photon. Rev.5(2), 308–321 (2011). [CrossRef]
  11. P. Seliger, M. Mahvash, C. Wang, and A. F. J. Levi, “Optimization of aperiodic dielectric structures,” J. Appl. Phys.100(3), 034310 (2006). [CrossRef]
  12. W. R. Frei, D. A. Tortorelli, and H. T. Johnson, “Geometry projection method for optimizing photonic nanostructures,” Opt. Lett.32(1), 77–79 (2007). [CrossRef] [PubMed]
  13. V. Liu and S. Fan, “Compact bends for multi-mode photonic crystal waveguides with high transmission and suppressed modal crosstalk,” Opt. Express21(7), 8069–8075 (2013). [CrossRef] [PubMed]
  14. G. Veronis, R. W. Dutton, and S. Fan, “Method for sensitivity analysis of photonic crystal devices,” Opt. Lett.29(19), 2288–2290 (2004). [CrossRef] [PubMed]
  15. A. F. J. Levi and I. G. Rosen, “A novel formulation of the adjoint method in the optimal design of quantum electronic devices,” SIAM J. Contr. Optim.48(5), 3191–3223 (2010). [CrossRef]
  16. G. Strang, Computational Science and Engineering, (Wellesley-Cambridge, 2007).
  17. O. D. Miller, PhD thesis (2012), University of California at Berkeley, http://optoelectronics.eecs.berkeley.edu/ThesisOwenMiller.pdf
  18. S. G. Johnson, M. Ibanescu, M. A. Skorobogatiy, O. Weisberg, J. D. Joannopoulos, and Y. Fink, “Perturbation theory for Maxwell’s equations with shifting material boundaries,” Phys. Rev. E Stat. Nonlin. Soft Matter Phys.65(6), 066611 (2002). [CrossRef] [PubMed]
  19. Lumerical FDTD Solutions, www.lumerical.com
  20. S. Osher and J. A. Sethian, “Fronts propagating with curvature-dependent speed: algorithms based on Hamilton-Jacobi formulations,” J. Comput. Phys.79(1), 12–49 (1988). [CrossRef]
  21. http://optoelectronics.eecs.berkeley.edu/PhotonicInverseDesign/

Cited By

Alert me when this paper is cited

OSA is able to provide readers links to articles that cite this paper by participating in CrossRef's Cited-By Linking service. CrossRef includes content from more than 3000 publishers and societies. In addition to listing OSA journal articles that cite this paper, citing articles from other participating publishers will also be listed.


« Previous Article  |  Next Article »

OSA is a member of CrossRef.

CrossCheck Deposited