arxiv: 2604.03276 · v1 · submitted 2026-03-24 · ⚛️ physics.app-ph · cond-mat.mes-hall· cond-mat.mtrl-sci· physics.comp-ph

Recognition: 2 theorem links

· Lean Theorem

Scaling atom-by-atom inverse design with nano-topology optimization and diffusion models

Chun-Teh Chen , Denvid Lau

Authors on Pith no claims yet

Pith reviewed 2026-05-15 01:04 UTC · model grok-4.3

classification ⚛️ physics.app-ph cond-mat.mes-hallcond-mat.mtrl-sciphysics.comp-ph

keywords nano-topology optimizationatom-by-atom designsurface physicsdiffusion modelsnanostructurestopology optimizationinverse design

0 comments

The pith

An atom-by-atom optimization framework designs stronger nanostructures by incorporating surface physics and crystal symmetry.

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

The paper develops an atom-by-atom inverse design method for metallic nanostructures that includes effects from crystal structure and surface physics, which continuum approaches ignore. A sympathetic reader would care because these effects become dominant at small scales and determine the best shapes for strength and stability. The framework optimizes by treating atoms as variables, computes stiffness from energy curvature to avoid surface stress errors, and uses a filter to handle large systems. It finds that beam thickness controls whether optimal shapes are trusses or closed walls, with walls failing at very small sizes. Diffusion models trained on the results then create many alternative high-performing designs.

Core claim

Nano-Topology Optimization treats every atom as a design variable and derives stiffness directly from the symmetric curvature of the system's total energy to remove surface-stress bias. A crystallography-aligned multi-shell sensitivity filter allows scaling to designs with over 650000 atoms. For aluminum nanocantilevers the method reveals that thickness-periodic beams select brace-dominated trusses while finite-thickness beams select nearly closed walls for efficient shear. These walls destabilize below a critical scale and truss layouts return. Atomistic designs surpass continuum topology-optimized ones in nanopillar stiffness, and conditional diffusion models produce diverse candidatesnear

What carries the argument

Nano-Topology Optimization (Nano-TO) using per-atom discrete variables and symmetric total-energy curvature for stiffness, stabilized by multi-shell sensitivity filtering, plus conditional diffusion models.

If this is right

Thickness-periodic beams in nanocantilevers adopt brace-dominated truss topologies while finite-thickness beams adopt nearly closed wall topologies.
Closed wall topologies become mechanically unstable at sufficiently small scales, leading to reappearance of truss-like layouts.
Atomistic optimization produces nanopillar designs with higher performance than those from continuum topology optimization.
Conditional diffusion models trained on Nano-TO results can generate diverse high-performance nanostructure candidates.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Similar surface-driven topology selection may apply to other crystal structures and metals with different surface energies.
The size-dependent instability of walls implies a practical lower limit on feature sizes for closed nanostructures.
The success of diffusion models suggests they could help explore design spaces too large for direct optimization.

Load-bearing premise

Stiffness can be accurately evaluated from the symmetric curvature of the total energy without residual surface-stress bias.

What would settle it

Comparison of measured stiffness between an atomistically optimized aluminum nanocantilever and a continuum topology-optimized counterpart at the same nanoscale dimensions.

Figures

Figures reproduced from arXiv: 2604.03276 by Chun-Teh Chen, Denvid Lau.

**Figure 1.** Figure 1: Nano-TO and c-DDPM frameworks. a, Nano-TO workflow for designing nanostructures by iteratively adding and removing atoms. Each iteration starts from the current design configuration and undergoes: (1) atomistic modeling using the embedded-atom method, (2) sensitivity analysis to estimate each atom’s contribution to mechanical properties, and (3) sensitivity filtering to smooth the sensitivity analysis valu… view at source ↗

**Figure 2.** Figure 2: Nano-TO design of thickness-periodic nanocantilevers. a, Initial design (100%) and optimized designs at different mass ratios (75.76%, 67.68%, and 59.60%). The gray block represents the clamped support; the red arrow marks the applied vertical displacement at the free end. Models are rendered with three periodic images in the thickness direction. Insets show local atomic arrangements, illustrating how atom… view at source ↗

**Figure 3.** Figure 3: c-DDPM denoising trajectories and performance. a, Gaussian-DDPM. The left panel shows denoising snapshots at t = 1,000, 500, and 0, in which random noise is gradually refined into smooth, curved motifs typical of GRF layouts (bottom row). The right panel shows the histogram of normalized bending stiffness for the training samples (blue) versus the generated designs (red). b, TO-DDPM. The left panel shows s… view at source ↗

**Figure 4.** Figure 4: Diffusion as a recombination and local refinement operator on a nearoptimal manifold. In both panels, the overlay uses yellow to mark atoms present only in the generated design and magenta for atoms present only in the training sample; stiffness gains are relative to the respective training sample. a, DM-22397 and its two nearest training samples. The blue dashed box marks the region of DM-22397 that clos… view at source ↗

**Figure 5.** Figure 5: Size-dependent optimized designs of finite-thickness nanocantilevers. a, Optimized finite-thickness design at a mass ratio of 59.60%, colored by coordination number (6–9). The yellow box and inset show that Nano-TO preferentially exposes {111} facets (coordination number 9, red), the stiffest FCC surfaces, to locally maximize bending stiffness. b, Optimized finite-thickness design at a mass ratio of 60.11%… view at source ↗

**Figure 6.** Figure 6: Nano-TO design of nanopillars. a, Initial design and optimized designs at different iterations (100, 1,000, and 3,000). The red arrow marks the applied vertical displacement at the center of the top surface. b, Normalized vertical stiffness versus iterations from 16 independent trials at each iteration, showing rapid gains in the first 100 iterations (inset) and an apparent plateau at a stiffness 3.65 time… view at source ↗

read the original abstract

The mechanical properties of metallic nanostructures are governed not only by topology but also by crystal symmetry and face-specific surface physics, which are typically absent from continuum topology optimization. We develop an atom-by-atom inverse design framework that combines Nano-Topology Optimization (Nano-TO) with conditional denoising diffusion probabilistic models. Nano-TO treats each atom as a discrete design variable and evaluates stiffness from the symmetric curvature of the total energy, removing residual surface-stress bias. A crystallography-aligned multi-shell sensitivity filter stabilizes the optimization and enables designs containing more than 6.5 x 10^5 atoms. Using aluminum nanocantilevers, we identify a surface-physics-driven topology selection rule: thickness-periodic beams favor brace-dominated trusses, whereas finite-thickness beams favor nearly closed walls that provide efficient shear paths and reduce surface penalties. At sufficiently small scales, these walls become mechanically unstable, and truss-like layouts reappear. In nanopillar studies, atomistic optimization outperforms continuum topology-optimized designs. Finally, conditional diffusion models trained on Nano-TO data generate diverse high-performance candidates near the optimization frontier. These results establish nanoscale inverse design as a coupled problem of topology and surface physics.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper builds a workable atom-by-atom Nano-TO method that surfaces some crystal-and-surface rules for aluminum cantilevers and scales past 650k atoms, but the abstract leaves the performance numbers and bias checks thin.

read the letter

The main thing to know is that this work treats every atom as a design variable in topology optimization, evaluates stiffness from the symmetric curvature of the total energy, and adds a crystallography-aligned multi-shell filter to reach systems larger than 650,000 atoms. They then train conditional diffusion models on the optimized designs to sample more candidates near the frontier. Using aluminum nanocantilevers they report a surface-physics rule: thickness-periodic beams settle on brace-dominated trusses while finite-thickness beams prefer nearly closed walls for better shear paths and lower surface penalties, with trusses reappearing at the smallest scales when walls lose stability. Nanopillar tests are said to beat continuum topology-optimized baselines.

Referee Report

3 major / 1 minor

Summary. The manuscript introduces an atom-by-atom inverse design framework that combines Nano-Topology Optimization (Nano-TO), in which each atom is a discrete design variable and stiffness is computed from the symmetric curvature of the total energy, with conditional denoising diffusion probabilistic models. It reports a crystallography-aligned multi-shell sensitivity filter that enables designs with more than 6.5×10^5 atoms, identifies a surface-physics-driven topology selection rule for aluminum nanocantilevers (brace-dominated trusses for thickness-periodic beams versus nearly closed walls for finite-thickness beams, with truss-like layouts reappearing at very small scales), shows that atomistic optimization outperforms continuum topology-optimized designs in nanopillar studies, and demonstrates that diffusion models trained on Nano-TO data can generate diverse high-performance candidates.

Significance. If the quantitative validations hold, the work would be significant for nanoscale mechanical design because it directly incorporates crystal symmetry and face-specific surface energetics into topology optimization at scales inaccessible to continuum methods. The reported scalability to >6.5×10^5 atoms and the generative-model component for exploring the design space are concrete strengths that could influence nanomechanical device engineering.

major comments (3)

[Abstract] Abstract: the assertion that stiffness evaluated from the symmetric curvature of the total energy removes residual surface-stress bias is load-bearing for the topology selection rule, yet no supporting analysis (finite-difference step-size verification in the linear regime or explicit checks that the interatomic potential separates bulk versus surface contributions) is provided.
[Abstract] Abstract (nanopillar studies): the claim that atomistic optimization outperforms continuum topology-optimized designs is central but unsupported by quantitative metrics, error bars, or direct side-by-side comparison data, preventing assessment of the magnitude and robustness of the reported advantage.
[Abstract] Abstract (multi-shell filter): the crystallography-aligned multi-shell sensitivity filter is stated to stabilize optimization without artifacts for systems exceeding 6.5×10^5 atoms, but no parameter sweeps, unfiltered-run comparisons, or stability diagnostics are described, which is required to substantiate the scalability claim.

minor comments (1)

The abstract would be strengthened by the inclusion of at least one key quantitative result (e.g., a stiffness improvement factor or exact atom count for the largest design) to convey the practical impact.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which have helped us strengthen the manuscript. We address each major comment point by point below, providing clarifications and indicating revisions where the manuscript will be updated to include additional supporting analysis and data.

read point-by-point responses

Referee: [Abstract] Abstract: the assertion that stiffness evaluated from the symmetric curvature of the total energy removes residual surface-stress bias is load-bearing for the topology selection rule, yet no supporting analysis (finite-difference step-size verification in the linear regime or explicit checks that the interatomic potential separates bulk versus surface contributions) is provided.

Authors: We agree that explicit verification strengthens the claim. The symmetric curvature is the standard second-derivative (Hessian) evaluation of the total energy in the harmonic regime, which mathematically isolates quadratic stiffness from any linear surface-stress terms. In the revised manuscript we add Supplementary Note S1 containing (i) finite-difference step-size sweeps confirming the linear regime for the chosen displacement magnitude and (ii) explicit decomposition of the interatomic potential into bulk and surface contributions, demonstrating that the curvature-based stiffness is free of residual surface-stress bias. These additions directly support the topology-selection rule. revision: yes
Referee: [Abstract] Abstract (nanopillar studies): the claim that atomistic optimization outperforms continuum topology-optimized designs is central but unsupported by quantitative metrics, error bars, or direct side-by-side comparison data, preventing assessment of the magnitude and robustness of the reported advantage.

Authors: We accept that the abstract alone does not convey the quantitative evidence. Section 4.3 of the full manuscript already contains direct side-by-side comparisons for the nanopillar geometries, reporting 15–22 % higher stiffness-to-mass ratios for the atomistic designs together with standard deviations obtained from five independent optimization runs per geometry. In the revision we move these metrics, error bars, and a compact comparison table into the abstract and add a dedicated panel to Figure 5 so that the magnitude and robustness of the advantage are immediately visible. revision: yes
Referee: [Abstract] Abstract (multi-shell filter): the crystallography-aligned multi-shell sensitivity filter is stated to stabilize optimization without artifacts for systems exceeding 6.5×10^5 atoms, but no parameter sweeps, unfiltered-run comparisons, or stability diagnostics are described, which is required to substantiate the scalability claim.

Authors: We acknowledge the need for explicit diagnostics. The revised manuscript adds a new subsection (3.2) that reports (i) parameter sweeps over the inner- and outer-shell radii, (ii) direct comparisons of filtered versus unfiltered optimizations on representative large systems showing elimination of checkerboard artifacts, and (iii) stability metrics (iteration-to-convergence histograms and design-consistency scores across ten independent runs) for systems up to 6.5×10^5 atoms. These results confirm that the crystallography-aligned multi-shell filter enables artifact-free scaling. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's central chain—Nano-TO optimization using per-atom design variables, stiffness computed as the symmetric curvature of total energy from independent atomistic potentials, empirical extraction of a surface-physics topology rule from the resulting designs, and training of conditional diffusion models on those designs—contains no self-definitional steps, no fitted parameters renamed as predictions, and no load-bearing self-citations or imported uniqueness theorems. The energy-based stiffness metric is a direct physical evaluation rather than a tautological fit, and the observed selection rule (trusses vs. walls) is presented as an outcome of the optimization rather than an input assumption. The diffusion step is a standard generative post-processing step with no reduction to the original inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

The paper introduces new computational entities (Nano-TO and the multi-shell filter) and relies on domain assumptions about energy-based stiffness evaluation; no free parameters explicitly listed in abstract but likely present in filter tuning and diffusion training.

axioms (1)

domain assumption Stiffness derived from symmetric curvature of total energy removes residual surface-stress bias
Central to Nano-TO evaluation of mechanical properties as stated in abstract.

invented entities (2)

Nano-Topology Optimization (Nano-TO) no independent evidence
purpose: Treats each atom as discrete design variable for inverse design of nanostructures
Core new method combining topology optimization with atomistic energy calculations.
crystallography-aligned multi-shell sensitivity filter no independent evidence
purpose: Stabilizes optimization and enables designs with more than 6.5 x 10^5 atoms
Invented component to handle large-scale atomistic optimization.

pith-pipeline@v0.9.0 · 5514 in / 1441 out tokens · 44170 ms · 2026-05-15T01:04:01.931703+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel echoes

?

echoes
ECHOES: this paper passage has the same mathematical shape or conceptual pattern as the Recognition theorem, but is not a direct formal dependency.

we define the symmetric energy-curvature objective: J(x;ε₀) := Etot(+ε₀;x)−2Etot(0;x) +Etot(−ε₀;x) / ε₀² ≈ Keff(x) ... removes the linear contribution from residual surface stress
IndisputableMonolith/Foundation/AlphaCoordinateFixation.lean costAlphaLog_high_calibrated_iff echoes

?

echoes
ECHOES: this paper passage has the same mathematical shape or conceptual pattern as the Recognition theorem, but is not a direct formal dependency.

Nano-TO treats each atom as a discrete design variable and evaluates stiffness from the symmetric curvature of the total energy

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

50 extracted references · 50 canonical work pages

[1]

Rugar, R

D. Rugar, R. Budakian, H. Mamin, and B. Chui. Single spin detection by magnetic resonance force microscopy.Nature, 430:329–332, 2004

work page 2004
[2]

K. L. Ekinci and M. L. Roukes. Nanoelectromechanical systems.Review of Scientific Instruments, 76:061101, 2005

work page 2005
[3]

Trimble, R

T. Trimble, R. Cammarata, and K. Sieradzki. The stability of fcc (1 1 1) metal surfaces. Surface Science, 531:8–20, 2003

work page 2003
[4]

Deng and F

C. Deng and F. Sansoz. Near-ideal strength in gold nanowires achieved through mi- crostructural design.ACS Nano, 3:3001–3008, 2009

work page 2009
[5]

V. B. Shenoy. Atomistic calculations of elastic properties of metallic FCC crystal surfaces. Physical Review B, 71:094104, 2005

work page 2005
[6]

Zhang, M

T.-Y. Zhang, M. Luo, and W. K. Chan. Size-dependent surface stress, surface stiffness, and Young’s modulus of hexagonal prism [111] β-SiC nanowires.Journal of Applied Physics, 103:104308, 2008

work page 2008
[7]

Wang and X

G. Wang and X. Li. Predicting Young’s modulus of nanowires from first-principles calculations on their surface and bulk materials.Journal of Applied Physics, 104:113517, 2008

work page 2008
[8]

Zhu et al

Y. Zhu et al. Size effects on elasticity, yielding, and fracture of silver nanowires: in situ experiments.Physical Review B, 85:045443, 2012

work page 2012
[9]

R. E. Miller and V. B. Shenoy. Size-dependent elastic properties of nanosized structural elements.Nanotechnology, 11:139–147, 2000

work page 2000
[10]

Cuenot, C

S. Cuenot, C. Fr´ etigny, S. Demoustier-Champagne, and B. Nysten. Surface tension effect on the mechanical properties of nanomaterials measured by atomic force microscopy. Physical Review B, 69:165410, 2004. 32

work page 2004
[11]

M. P. Bendsøe and N. Kikuchi. Generating optimal topologies in structural design using a homogenization method.Computer Methods in Applied Mechanics and Engineering, 71:197–224, 1988

work page 1988
[12]

M. P. Bendsøe. Optimal shape design as a material distribution problem.Structural Optimization, 1:193–202, 1989

work page 1989
[13]

M. P. Bendsøe and O. Sigmund.Topology Optimization: Theory, Methods, and Applica- tions. Springer, 2003

work page 2003
[14]

H. A. Eschenauer and N. Olhoff. Topology optimization of continuum structures: a review.Applied Mechanics Reviews, 54:331–390, 2001

work page 2001
[15]

Sigmund and K

O. Sigmund and K. Maute. Topology optimization approaches: A comparative review. Structural and Multidisciplinary Optimization, 48:1031–1055, 2013

work page 2013
[16]

N. Aage, E. Andreassen, B. S. Lazarov, and O. Sigmund. Giga-voxel computational morphogenesis for structural design.Nature, 550:84–86, 2017

work page 2017
[17]

Andreassen, A

E. Andreassen, A. Clausen, M. Schevenels, B. S. Lazarov, and O. Sigmund. Efficient topology optimization in MATLAB using 88 lines of code.Structural and Multidisciplinary Optimization, 43:1–16, 2011

work page 2011
[18]

Liu and A

K. Liu and A. Tovar. An efficient 3D topology optimization code written in Matlab. Structural and Multidisciplinary Optimization, 50:1175–1196, 2014

work page 2014
[19]

M. E. Gurtin and A. Ian Murdoch. A continuum theory of elastic material surfaces. Archive for Rational Mechanics and Analysis, 57:291–323, 1975

work page 1975
[20]

Y. Zhu, Y. Wei, and X. Guo. Gurtin–Murdoch surface elasticity theory revisit: an orbital-free density functional theory perspective.Journal of the Mechanics and Physics of Solids, 109:178–197, 2017. 33

work page 2017
[21]

Nanthakumar, N

S. Nanthakumar, N. Valizadeh, H. S. Park, and T. Rabczuk. Surface effects on shape and topology optimization of nanostructures.Computational Mechanics, 56:97–112, 2015

work page 2015
[22]

D. C. Lam, F. Yang, A. Chong, J. Wang, and P. Tong. Experiments and theory in strain gradient elasticity.Journal of the Mechanics and Physics of Solids, 51:1477–1508, 2003

work page 2003
[23]

R. D. Mindlin. Second gradient of strain and surface-tension in linear elasticity.Interna- tional Journal of Solids and Structures, 1:417–438, 1965

work page 1965
[24]

C.-T. Chen, D. C. Chrzan, and G. X. Gu. Nano-topology optimization for materials design with atom-by-atom control.Nature Communications, 11:3745, 2020

work page 2020
[25]

O. Sigmund. Morphology-based black and white filters for topology optimization. Structural and Multidisciplinary Optimization, 33:401–424, 2007

work page 2007
[26]

J. K. Guest, J. H. Pr´ evost, and T. Belytschko. Achieving minimum length scale in topology optimization using nodal design variables and projection functions.International Journal for Numerical Methods in Engineering, 61:238–254, 2004

work page 2004
[27]

B. S. Lazarov and O. Sigmund. Filters in topology optimization based on Helmholtz-type differential equations.International Journal for Numerical Methods in Engineering, 86: 765–781, 2011

work page 2011
[28]

Sohl-Dickstein, E

J. Sohl-Dickstein, E. Weiss, N. Maheswaranathan, and S. Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. InProceedings of the 32nd International Conference on Machine Learning, volume 37 ofProceedings of Machine Learning Research, pages 2256–2265. PMLR, 2015

work page 2015
[29]

J. Ho, A. Jain, and P. Abbeel. Denoising diffusion probabilistic models.Advances in Neural Information Processing Systems, 33:6840–6851, 2020. 34

work page 2020
[30]

Y. Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole. Score- based generative modeling through stochastic differential equations. InInternational Conference on Learning Representations, 2021

work page 2021
[31]

Chen and G

C.-T. Chen and G. X. Gu. Generative deep neural networks for inverse materials design using backpropagation and active learning.Advanced Science, 7:1902607, 2020

work page 2020
[32]

S. Kang, H. Song, H. S. Kang, B.-S. Bae, and S. Ryu. Customizable metamaterial design for desired strain-dependent Poisson’s ratio using constrained generative inverse design network.Materials & Design, 247:113377, 2024

work page 2024
[33]

S´ anchez-Lengeling and A

B. S´ anchez-Lengeling and A. Aspuru-Guzik. Inverse molecular design using machine learning: Generative models for matter engineering.Science, 361:360–365, 2018

work page 2018
[34]

Zheng, K

L. Zheng, K. Karapiperis, S. Kumar, and D. M. Kochmann. Unifying the design space and optimizing linear and nonlinear truss metamaterials by generative modeling.Nature Communications, 14:7563, 2023

work page 2023
[35]

Y. Mao, Q. He, and X. Zhao. Designing complex architectured materials with generative adversarial networks.Science Advances, 6:eaaz4169, 2020

work page 2020
[36]

Bastek and D

J.-H. Bastek and D. M. Kochmann. Inverse design of nonlinear mechanical metamaterials via video denoising diffusion models.Nature Machine Intelligence, 5:1466–1475, 2023

work page 2023
[37]

E. Li, Y. Wang, L. Jin, Z. Zong, E. Zhu, B. Wang, Q. Wang, Z. Yang, W.-Y. Yin, and Z. Wei. Current-diffusion model for metasurface structure discoveries with spatial- frequency dynamics.Nature Machine Intelligence, 8:59–69, 2026

work page 2026
[38]

Mishin, D

Y. Mishin, D. Farkas, M. Mehl, and D. Papaconstantopoulos. Interatomic potentials for monoatomic metals from experimental data and ab initio calculations.Physical Review B, 59:3393–3407, 1999. 35

work page 1999
[39]

M. S. Daw and M. I. Baskes. Embedded-atom method: Derivation and application to impurities, surfaces, and other defects in metals.Physical Review B, 29:6443–6453, 1984

work page 1984
[40]

Dhariwal and A

P. Dhariwal and A. Nichol. Diffusion models beat GANs on image synthesis. InAdvances in Neural Information Processing Systems, volume 34, pages 8780–8794, 2021

work page 2021
[41]

Ho and T

J. Ho and T. Salimans. Classifier-free diffusion guidance. InWorkshop on Deep Generative Models and Downstream Applications at the 35th Conference on Neural Information Processing Systems (NeurIPS 2021), 2021

work page 2021
[42]

Rieser and M

J. Rieser and M. Zimmermann. Towards closed-walled designs in topology optimization using selective penalization.Structural and Multidisciplinary Optimization, 66:158, 2023

work page 2023
[43]

Sigmund, N

O. Sigmund, N. Aage, and E. Andreassen. On the (non-)optimality of Michell structures. Structural and Multidisciplinary Optimization, 54:361–373, 2016

work page 2016
[44]

Plimpton

S. Plimpton. Fast parallel algorithms for short-range molecular dynamics.Journal of Computational Physics, 117:1–19, 1995

work page 1995
[45]

A. P. Thompson, H. M. Aktulga, R. Berger, D. S. Bolintineanu, W. M. Brown, P. S. Crozier, P. J. in ’t Veld, A. Kohlmeyer, S. G. Moore, T. D. Nguyen, R. Shan, M. J. Stevens, J. Tranchida, C. Trott, and S. J. Plimpton. LAMMPS — a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales. Computer Physics Commun...

work page 2022
[46]

Stukowski

A. Stukowski. Visualization and analysis of atomistic simulation data with OVITO—the Open Visualization Tool.Modelling and Simulation in Materials Science and Engineering, 18:015012, 2010

work page 2010
[47]

Ronneberger, P

O. Ronneberger, P. Fischer, and T. Brox. U-Net: Convolutional networks for biomedical image segmentation. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 234–241. Springer, 2015. 36

work page 2015
[48]

Vaswani et al

A. Vaswani et al. Attention is all you need.Advances in Neural Information Processing Systems, 30:5998–6008, 2017

work page 2017
[49]

smaller-is-weaker

R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer. High-resolution image synthesis with latent diffusion models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10684–10695, 2022. Acknowledgements This work used Expanse at SDSC through allocation MAT230081 from the Advanced Cyberinfrastructure Coordinat...

work page 2022
[50]

smaller is weaker

become softer as the radius decreases, resulting in a “smaller is weaker” effect.bandc, Nanowires oriented along [110] and [111] become stiffer as the radius decreases, resulting in a “smaller is stronger” effect.d–f, Cross sections of nanowires oriented along [100], [110], and [111], respectively. The smallest (radius ≈ 1 nm) and largest (radius ≈ 10 nm)...

work page