Enhancement and speed-up of carrier dynamics in a dielectric nanocavity with deep sub-wavelength confinement

Ali Nawaz Babar; Gaoneng Dong; Jesper M{\o}rk; Rasmus Elleb{\ae}k Christiansen; S{\o}ren Engelberth Hansen; S{\o}ren Stobbe; Yi Yu

REVIEW 3 major objections 5 minor 1 cited by

Enhancement and speed-up of carrier dynamics in a dielectric nanocavity with deep sub-wavelength confinement

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A 12 nm-gap silicon bowtie cavity confines light far below the diffraction limit, and that confinement both boosts two-photon carrier generation and cuts carrier recovery to sub-picosecond times, enabling faster all-optical switching.

desk verdict First pump-probe study of a deep-subwavelength dielectric bowtie cavity; the enhanced nonlinearity is credible, but the claimed diffusion speed-up rests on a fitted tau_diff and should be treated as provisional. read the letter →

arxiv 2412.08471 v1 pith:2AXTDZGH submitted 2024-12-11 physics.optics

Gaoneng Dong , Ali Nawaz Babar , Rasmus Elleb{ae}k Christiansen , S{o}ren Engelberth Hansen , S{o}ren Stobbe , Yi Yu , Jesper M{o}rk This is my paper · ORCID

classification physics.optics

keywords dielectricbowtienanocavitydeepsub-wavelengthconfinementtwo-photonabsorptionfree-carrierdiffusionultrafastall-opticalswitchingheterodynepump-probemodevolumetopologyoptimization

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper reports experiments on a topology-optimized silicon bowtie nanocavity whose gap is only 12 nm wide, localizing light to a mode volume well below the half-wavelength-cubed diffraction limit. The authors show that this deep sub-wavelength confinement simultaneously increases the rate at which pump light generates free carriers through two-photon absorption and shortens the time those carriers take to leave the cavity region. The measured probe transmission recovers almost fully within about 2 ps, with a fitted carrier diffusion time below 1 ps, more than an order of magnitude faster than a conventional reference microcavity. The paper concludes that using a small mode volume rather than a high quality factor to enhance light-matter interaction preserves device bandwidth, making such cavities promising for low-power and ultrafast all-optical switches and modulators.

What carries the argument

The load-bearing quantities are three mode volumes: the linear effective mode volume $V_{\mathrm{eff}}$, the free-carrier absorption mode volume $V_{\mathrm{FCA}}$, and the two-photon absorption mode volume $V_{\mathrm{TPA}}$. The carrier generation rate scales as $1/V_{\mathrm{FCA}}^2$, so shrinking $V_{\mathrm{FCA}}$ raises the number of free carriers produced by two-photon absorption at a fixed pump energy, while the small $V_{\mathrm{eff}}$ creates steep carrier gradients that drive fast ambipolar diffusion out of the mode region. These quantities are tied together by temporal coupled-mode theory with a two-region carrier-density parametrization (fast diffusion time $\tau_{\mathrm{diff}}$ and slow relaxation time $\tau_{\mathrm{slow}}$) plus coherent four-wave-mixing terms that reproduce the dynamics around zero pump-probe delay.

What would settle it

A time-resolved spatial measurement of the carrier population in the 12 nm-gap cavity that showed the hotspot emptying faster than the ambipolar-diffusion simulation predicts, or that showed the fitted recovery time not shortening as the bowtie width is reduced, would falsify the diffusion-dominated sub-picosecond recovery claim.

Watch

Extended reading notes

Core claim

The central claim is that deep sub-wavelength confinement in a dielectric nanocavity does not merely strengthen light-matter coupling; it also speeds the recovery of an optical switch, because the same small mode volume that concentrates the pump field produces steep spatial gradients in the generated free-carrier distribution. The paper demonstrates this in a topology-optimized silicon bowtie cavity with a 12 nm gap: measured and simulated probe transmission dynamics agree when the carrier generation rate is enhanced by the small free-carrier-absorption mode volume ($V_{\mathrm{FCA}} = 0.033\times10^{-18}\,\mathrm{m^3}$) and the fast diffusion time is $\tau_{\mathrm{diff}} = 0.6$ ps, versus 6 ps for the reference cavity. The fast component of the carrier relaxation is attributed to ambipolar diffusion, which dominates when the surface recombination velocity is below $10^5$ cm/s. The result is a switching window of about 1.5 ps and an extinction ratio of $-5.7$ dB at 255 fJ pump energy, with coherent four-wave mixing between pump and probe adding a further contribution near zero delay. The paper concludes that dielectric bowtie cavities improve on conventional point-defect cavities by using a small mode volume rather than a high quality factor, so bandwidth is not sacrificed.

Load-bearing premise

The sub-picosecond recovery time is set by a fitted diffusion constant in the coupled-mode model rather than a direct measurement; if the fast recovery actually comes from surface recombination, thermal effects, or a measurement artifact, the central diffusion-speed-up narrative would collapse.

Editorial extensions

If this is right

All-optical switches made from deep sub-wavelength dielectric cavities can recover in about a picosecond at femtojoule pump energies without relying on a high quality factor, so the operating bandwidth is not compromised.
At the same 255 fJ pump energy, the bowtie cavity reaches a $-5.7$ dB extinction ratio against $-2.3$ dB for a higher-$Q$ reference cavity, showing that nonlinear mode volume, not linear $Q$, sets the switching contrast.
Narrowing the bowtie from 12 nm toward 2 nm is calculated to reduce both the free-carrier-absorption mode volume and the fast diffusion time, reaching tens of femtoseconds, though faster diffusion also limits carrier accumulation.
The coherent four-wave-mixing contribution near zero delay is itself a usable fast switching mechanism, confirmed by detecting the idler signal only in that time window.
Replacing silicon with a material with a larger two-photon absorption coefficient, such as InP or GaAs, would further lower the pump power needed for a given resonance shift.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A testable extension of the reported mechanism is that any nonlinearity localized to the hotspot, not just two-photon carrier generation, should inherit the same fast recovery because the steep carrier gradients are a geometric property of the small mode volume.
The diffusion speed-up implies a design trade-off the paper does not optimize: reducing the gap improves contrast and speed but lowers the number of carriers that can accumulate, so a specific bowtie width should minimize switching energy per bit at a given bit rate.
A decisive check of the mechanism would be to vary the surface recombination velocity or use time-resolved spatial imaging; if recovery time ceases to scale with the diffusion gradients as the gap narrows, surface recombination or thermal effects would be the true cause.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The first thing to know: this is a credible experimental demonstration that a dielectric bowtie cavity with a 12 nm gap enhances two-photon absorption and shortens the carrier recovery transient compared to a reference nanobeam cavity. It is also the first pump-probe study of ultrafast dynamics in such deep-subwavelength dielectric cavities, and the observation that the FCA mode volume drops steeply with bowtie width while the TPA mode volume barely changes is a genuinely useful design insight. The idler signal measurement independently confirms the FWM contribution near zero delay, which is careful and should be credited.

The main soft spot is exactly what the reader flagged: the sub-picosecond diffusion time is not measured, it is a fitted parameter in a coupled-mode model. tau_diff = 0.6 ps comes from Table S3, and the simulation gives 0.3 ps—a factor of two off. The paper also assumes surface recombination below 1e5 cm/s, but for etched silicon nanostructures, 1e6 cm/s is plausible, and Figure S3 shows that at that rate surface recombination dominates. So the mechanism for the fast recovery—diffusion versus surface recombination—is not uniquely established. That matters because the title and abstract make the diffusion speed-up the central narrative. I don't think this invalidates the work, but the claim should be softened or supported with a direct measurement (e.g., time-resolved carrier imaging) or at least a sensitivity analysis showing the fit is robust to S.

Minor issues: no error bars, data not public, and the abstract says "more than an order of magnitude" for the diffusion time when the quoted numbers (0.6 vs 6 ps) are exactly a factor of ten. The switching-window comparison (1.5 vs 18 ps) does support the stronger phrasing, but the diffusion-time comparison does not. Should be fixed. One more thing: the stress-test worry that the sub-ps transient is just coherent FWM overlap is addressed by the idler measurement—the coherent process is real and separately confirmed. The unresolved issue is not the FWM; it's that the recovery tail is fit by a diffusion time instead of being pinned down independently.

This paper deserves a serious referee. The experimental work is solid, the mode-volume analysis is valuable, and the central qualitative result—stronger and faster switching in a bowtie cavity—is probably right. But the quantitative diffusion claim and the missing uncertainty analysis need to be addressed before I'd take the speed-up at face value. For a reading group it would provoke a good discussion about model fitting in nanophotonics; I'd bring it.

Recommendation: send to peer review, and ask the authors for error bars, data availability, and a sensitivity analysis on the surface recombination rate before acceptance.

Referee Report

3 major / 5 minor

Summary. The paper reports an experimental study of a topology-optimized silicon bowtie nanocavity with a 12 nm gap, showing deep sub-wavelength optical confinement. Using heterodyne pump-probe measurements, the authors observe a larger and faster probe transmission change compared to a conventional nanobeam reference cavity. They interpret the dynamics with a temporal coupled-mode model that includes two-photon-absorption-induced free carriers, carrier diffusion, and coherent four-wave mixing. The central claim is that the small mode volume simultaneously enhances the carrier generation rate and shortens the carrier recovery time to below 1 ps, attributed to fast carrier diffusion, and that this represents an advantage for ultrafast all-optical switching.

Significance. If the central claim holds, the paper demonstrates a practical advantage of dielectric bowtie cavities over conventional point-defect cavities: deep sub-wavelength confinement can improve both the strength and the speed of all-optical switching. The work combines careful fabrication, heterodyne pump-probe measurements, and an established coupled-mode theory framework, and it includes an independent idler-signal measurement that corroborates the coherent four-wave-mixing contribution. These are notable strengths. However, the key mechanistic claim—that the sub-picosecond recovery is due to carrier diffusion—rests on a fitted parameter and an unverified assumption about surface recombination, so the significance is conditional on additional evidence.

major comments (3)

[Wavelength-dependent dynamics / Table S3] The central claim of a sub-picosecond carrier diffusion time relies on tau_diff = 0.6 ps being a fit parameter in the coupled-mode model (Table S3) with no reported uncertainty. The model also assumes a surface recombination rate S < 1e5 cm/s (Supplement S2) without experimental support. Figure S3 shows that for S = 1e6 cm/s the relaxation is dominated by surface recombination rather than diffusion, and such values are plausible for etched silicon nanostructures. The paper therefore does not uniquely establish that the fast recovery is caused by diffusion. Please provide direct evidence (e.g., passivation experiments, temperature dependence, or spatially resolved carrier measurements) or rephrase the claim to 'fast carrier recovery' and clearly state that the diffusion mechanism is inferred rather than measured.
[Wavelength-dependent dynamics, first paragraph] The simulated fast diffusion time for the bowtie cavity is approximately 0.3 ps, while the fitted value is 0.6 ps, a factor-of-two discrepancy. The slow tail has only one-tenth of the total amplitude, so the data provide weak constraint on tau_diff. The manuscript should report the fit quality, parameter uncertainties, and a sensitivity analysis showing that the sub-picosecond recovery time is required by the data rather than enforced by the model structure.
[Mode volumes and carrier diffusion / Supplement S2] The comparison between the bowtie and reference cavities is based on fitted tau_diff values (0.6 ps vs 6 ps, Table S3) that are not measured directly, and the two-region carrier model uses a fitted volume ratio R12 (1/10 vs 1/1.3) whose physical basis is not discussed. Since the pump pulse widths (0.93 ps vs 1.12 ps) and loaded Q factors (700 vs 1200) differ between the two cavities, the relative contributions of mode volume, Q factor, and pulse width to the observed speed-up and extinction ratio should be disentangled to support the claim that the mode volume is the dominant cause.

minor comments (5)

[Figures 1c and 1d] The experimental traces in Figures 1c and 1d do not show error bars or measurement uncertainty; adding them or stating the noise floor would help assess the fit quality.
[Supplementary equations S8 and S9] The typesetting of equations S8 and S9 is garbled in the supplement, making it difficult to read the terms involving the carrier densities and the perturbation variables; please provide a clean version.
[Abstract] The first sentence contains a subject-verb agreement error: 'The emergence of dielectric bowtie cavities enable' should be 'enables'.
[Data Availability Statement] The statement 'available from the corresponding author upon reasonable request' is not a public data-availability statement; consider depositing the raw pump-probe traces and processed data in a repository.
[Supplement S4, Eq. (s12)] The derivation of the lock-in signal should clarify the role of the finite reference pulse width and the condition that the reference is much shorter than the probe; currently it is stated as an aside and would benefit from a short derivation.

Circularity Check

1 steps flagged · score 4.0 of 10

Sub-ps diffusion time is a fitted parameter in the coupled-mode model, so the headline speed-up is partly circular despite independent diffusion simulations.

fitted input called prediction [Abstract; Supplement S3 Table S3; main text, section 'Wavelength-dependent dynamics' (p. 9)]
"The fast diffusion time used for the fitting is 0.6 ps, which is comparable to the numerically simulated result of 0.3 ps obtained using the ambipolar-diffusion model."

The abstract reports 'A diffusion time below 1 ps is achieved for the bowtie cavity,' but in the coupled-mode model (S3 eqs. s8-b, s8-c, s9-c, s9-d) tau_diff is a free decay constant of the carrier density in the cavity-mode region, and Table S3 sets tau_diff = 0.6 ps for the bowtie and 6 ps for the reference. The paper explicitly says this value was 'used for the fitting.' The near-zero-delay transient is independently shown to be dominated by FWM via the idler signal, and the slow tail is only one-tenth of the total amplitude, so the measured trace constrains tau_diff loosely; a larger surface recombination (S >= 1e6 cm/s, Fig. S3) would mimic fast relaxation.

full rationale

The paper's main derivation chain is largely self-contained: the mode volumes Veff, VFCA, and VTPA are computed from simulated field profiles via eqs. s5-s7 rather than fitted, and the TPA/FCA scaling arguments follow from those independent quantities. The coherent FWM contribution is corroborated by the measured idler signal in Supplement S6, so the reliance on the authors' earlier coupled-mode theory (Refs. 43, 44, 13) is supported by independent experimental evidence rather than being circular. The one significant circular element is the diffusion-time claim: tau_diff = 0.6 ps is a free fitting parameter in Table S3, and the abstract's 'diffusion time below 1 ps is achieved' is essentially that fitted number. The ambipolar-diffusion simulation of 0.3 ps is an independent estimate and prevents full circularity, but it is a factor of two away from the fitted value and rests on the same assumed surface-recombination bound (S < 1e5 cm/s). Because the fast transient around zero delay is FWM-dominated and the slow tail is weak, the fitted tau_diff is not uniquely identified with carrier diffusion. This is a partial circularity of the 'fitted input called prediction' type, not a collapse of the whole derivation; the central confinement-speed-up picture retains substantial independent grounding, so the score is 4 rather than higher.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claims rest on a small number of fitted parameters, chiefly the fast diffusion time and the carrier volume ratio, and on standard semiconductor and optical modeling assumptions. No new physical entities are introduced. The fitted parameters are the main source of uncertainty in the quantitative claims.

free parameters (3)

tau_diff (fast carrier diffusion time) = 0.6 ps (bowtie), 6 ps (reference)
Fitted to the measured probe recovery dynamics in the coupled-mode model; the independently simulated value is 0.3 ps, so this parameter carries much of the claimed speed-up.
R12 (ratio of effective carrier volumes) = 1/10 (bowtie), 1/1.3 (reference)
Fitted to match the relative amplitudes of fast and slow recovery components; no independent measurement is provided.
tau_slow (slow carrier decay time) = 100 ps (both cavities)
Set to a common value for both cavities, presumably from prior carrier lifetime literature; it shapes the slow tail but is not the focus of the claims.

assumptions (5)

domain assumption Ambipolar diffusion model with a single ambipolar diffusion coefficient describes carrier spreading in the silicon cavity
Used in Supplement S2 and Fig. 3b to simulate carrier relaxation; assumes bulk diffusion dominates and surface recombination is below 1e5 cm/s for the presented results.
domain assumption Coupled-mode theory with the probe treated as a small perturbation to the pump field is valid
Supplement S3, Eqs. (s9); underlies the entire simulation of probe transmission and idler generation.
domain assumption A two-region carrier density model (cavity mode region and background) with a fixed volume ratio R12 represents the spatial carrier dynamics
Supplement S3, Eqs. (s8-b,c) and (s9-c,d); R12 is a fitted parameter in Table S3.
domain assumption Literature values for silicon TPA, Kerr, FCA, and free-carrier dispersion coefficients apply at the relevant carrier densities
Table S3 lists beta_TPA = 9.95e-12 m/W, n2 = 4e-18 m2/W, K_Car = 0.89e-12 m3/s, sigma = 1.45e-21 m2; these are not measured in this work.
standard math The topology optimization design problem uses time-harmonic Maxwell equations and fabrication constraints
Supplement S1, Eqs. (s1)-(s4); standard electromagnetic modeling, not a source of concern.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Enhancement and speed-up of carrier dynamics in a dielectric nanocavity with deep sub-wavelength confinement." pith.science (2026). https://pith.science/paper/2AXTDZGH

@misc{pith2026241208471,
  author       = {Pith},
  title        = {Pith review of: Enhancement and speed-up of carrier dynamics in a dielectric nanocavity with deep sub-wavelength confinement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2AXTDZGH}},
  note         = {Machine review of arXiv:2412.08471}
}

read the original abstract

The emergence of dielectric bowtie cavities enable optical confinement with ultrahigh quality factor and ultra-small optical mode volumes with perspectives for enhanced light-matter interaction. Experimental work has so far emphasized the realization of these nanocavities. Here, we experimentally investigate the ultrafast dynamics of a topology-optimized dielectric (silicon) bowtie nanocavity, with device dimensions down to 12 nm, that localizes light to a mode volume deep below the so-called diffraction limit given by the half-wavelength cubed. This strong spatial light concentration is shown to significantly enhance the carrier generation rate through two-photon absorption, as well as reducing the time it takes for the carriers to recover. A diffusion time below 1 ps is achieved for the bowtie cavity, which is more than an order of magnitude smaller than for a conventional microcavity. Additionally, parametric effects due to coherent interactions between pump and probe signals are also enhanced in the bowtie cavity, leading to an improved extinction ratio. These results demonstrate important fundamental advantages of dielectric bowtie cavities compared to conventional point-defect cavities, laying a foundation for novel low-power and ultrafast optical devices, including switches and modulators.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Strong coupling between a dielectric nanocavity and a monolayer transition metal dichalcogenide
physics.optics 2025-02 conditional novelty 7.0 of 10

An InP extreme-confinement nanocavity and a monolayer of MoTe2 show clear avoided crossing with a light-matter coupling of about 5 meV, exceeding the system losses by a factor of two.

Reference graph

Works this paper leans on

14 extracted references · 14 canonical work pages · cited by 1 Pith paper

[1]

Inverse design of optical mode converters by topology optimization: tutorial,

Christiansen, R. E., "Inverse design of optical mode converters by topology optimization: tutorial," J. Opt. 25, 083501 (2023)

work page 2023
[2]

Jin, J.-M., The finite element method in electromagnetics (John Wiley & Sons, 2015)

work page 2015
[3]

Compact 200 line MATLAB code for inverse design in photonics by topology optimization: tutorial,

Christiansen, R. E. & Sigmund, O., "Compact 200 line MATLAB code for inverse design in photonics by topology optimization: tutorial," J. Opt. Soc. Am. B 38, 510 (2021)

work page 2021
[4]

On projection methods, convergence and robust formulations in topology optimization,

Wang, F., Lazarov, B. S. & Sigmund, O., "On projection methods, convergence and robust formulations in topology optimization," Struct. Multidisc. Optim. 43, 767-784 (2010). 32

work page 2010
[5]

Minimum length scale in topo logy optimization by geometric constraints,

Zhou, M., Lazarov, B. S., Wang, F. & Sigmund, O., "Minimum length scale in topo logy optimization by geometric constraints," Comput. Methods Appl. Mech. Eng. 293, 266-282 (2015)

work page 2015
[6]

A class of globally convergent optimization methods based on conservative convex separable approximations,

Svanberg, K., "A class of globally convergent optimization methods based on conservative convex separable approximations," SIAM J. Optim. 12, 555-573 (2002)

work page 2002
[7]

High quality factor photonic crystal nanobeam cavities,

Deotare, P. B., et al., "High quality factor photonic crystal nanobeam cavities," Appl. Phys. Lett. 94, 121106 (2009)

work page 2009
[8]

Nonlinear response of silicon photonic crystal microresonators excited via an integrated waveguide and fiber taper,

Barclay, P. E., Srinivasan, K. & Painter, O., "Nonlinear response of silicon photonic crystal microresonators excited via an integrated waveguide and fiber taper," Opt. Express 13, 801- 820 (2005)

work page 2005

Show all 14 references

[9]

T wo regimes of confinement in photonic nanocavities: bulk confinement versus lightning rods,

Albrechtsen, M., Vosoughi Lahijani, B. & Stobbe, S., "T wo regimes of confinement in photonic nanocavities: bulk confinement versus lightning rods," Opt. Express 30, 15458 - 15469 (2022)

2022
[10]

Generalized effective mode volume for leaky optical cavities,

Kristensen, P. T., Van Vlack, C. & Hughes, S., "Generalized effective mode volume for leaky optical cavities," Opt. Lett. 37, 1649-1651 (2012)

2012
[11]

Ultrafast coherent dynamics of a photonic crystal all-optical switch,

Colman, P., Lunnemann, P., Yu, Y. & Mork, J., "Ultrafast coherent dynamics of a photonic crystal all-optical switch," Phys. Rev. Lett. 117, 233901 (2016)

2016
[12]

Ultrafast parame tric process in a photonic-crystal nanocavity switch,

Lunnemann, P., Yu, Y., Joanesarson, K. & Mørk, J., "Ultrafast parame tric process in a photonic-crystal nanocavity switch," Phys. Rev. A 99, 053835 (2019)

2019
[13]

Switching characteristics of an InP photonic crystal nanocavity: Experiment and theory,

Yu, Y., et al., "Switching characteristics of an InP photonic crystal nanocavity: Experiment and theory," Opt. Express 21, 31047-31061 (2013)

2013
[14]

Theory of heterodyne pump –probe experiments with femtosecond pulses,

Mecozzi, A. & Mørk, J., "Theory of heterodyne pump –probe experiments with femtosecond pulses," J. Opt. Soc. Am. B 13, 2437-2452 (1996)

1996

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

Inverse design of optical mode converters by topology optimization: tutorial,

Christiansen, R. E., "Inverse design of optical mode converters by topology optimization: tutorial," J. Opt. 25, 083501 (2023)

work page 2023

[2] [2]

Jin, J.-M., The finite element method in electromagnetics (John Wiley & Sons, 2015)

work page 2015

[3] [3]

Compact 200 line MATLAB code for inverse design in photonics by topology optimization: tutorial,

Christiansen, R. E. & Sigmund, O., "Compact 200 line MATLAB code for inverse design in photonics by topology optimization: tutorial," J. Opt. Soc. Am. B 38, 510 (2021)

work page 2021

[4] [4]

On projection methods, convergence and robust formulations in topology optimization,

Wang, F., Lazarov, B. S. & Sigmund, O., "On projection methods, convergence and robust formulations in topology optimization," Struct. Multidisc. Optim. 43, 767-784 (2010). 32

work page 2010

[5] [5]

Minimum length scale in topo logy optimization by geometric constraints,

Zhou, M., Lazarov, B. S., Wang, F. & Sigmund, O., "Minimum length scale in topo logy optimization by geometric constraints," Comput. Methods Appl. Mech. Eng. 293, 266-282 (2015)

work page 2015

[6] [6]

A class of globally convergent optimization methods based on conservative convex separable approximations,

Svanberg, K., "A class of globally convergent optimization methods based on conservative convex separable approximations," SIAM J. Optim. 12, 555-573 (2002)

work page 2002

[7] [7]

High quality factor photonic crystal nanobeam cavities,

Deotare, P. B., et al., "High quality factor photonic crystal nanobeam cavities," Appl. Phys. Lett. 94, 121106 (2009)

work page 2009

[8] [8]

Nonlinear response of silicon photonic crystal microresonators excited via an integrated waveguide and fiber taper,

Barclay, P. E., Srinivasan, K. & Painter, O., "Nonlinear response of silicon photonic crystal microresonators excited via an integrated waveguide and fiber taper," Opt. Express 13, 801- 820 (2005)

work page 2005