PaperlayerView exact revision on arXiv

Source-linked AI summary

Inverse Optimization Techniques for Targeted Self-Assembly

Salvatore Torquato

arXiv:0811.0040v2cond-mat.softcond-mat.stat-mech

TL;DR

The paper addresses how to design interaction potentials that stabilize targeted many-body structures while incorporating complete configurational information. It reviews inverse methods for ordered and disordered configurations, including crystal, amorphous, and quasicrystal structures, and reports deeper understanding of structure–interaction relationships and optimized potentials for novel properties.

  • Problem

    Determining classical ground states and designing interactions for targeted many-body structures remain challenging problems in condensed-matter physics and materials science.

  • Method

    The paper reviews inverse statistical-mechanical methods that optimize interactions using configurational information and targeted structural or bulk-property objectives.

  • Results

    The reviewed approaches have produced optimized potentials for ordered and disordered particle configurations, including diamond and wurtzite crystals, and have deepened understanding of relationships between collective structure and interactions.

  • Takeaways & Limitations

    The methods support tailoring self-assembly toward crystals, amorphous structures, and quasicrystals, although future work must emphasize experimentally achievable interactions.

Abstract

from arXiv · show

This article reviews recent inverse statistical-mechanical methodologies that we have devised to optimize interaction potentials in soft matter systems that correspond to stable "target" structures. We are interested in finding the interaction potential, not necessarily pairwise additive or spherically symmetric, that stabilizes a targeted many-body system by generally incorporating complete configurational information. Unlike previous work, our primary interest is in the possible many-body structures that may be generated, some of which may include interesting but known structures, while others may represent entirely new structural motifs. Soft matter systems, such as colloids and polymers, offer a versatile means of realizing the optimized interactions. It is shown that these inverse approaches hold great promise for controlling self-assembly to a degree that surpasses the less-than-optimal path that nature has provided. Indeed, we envision being able to "tailor" potentials that produce varying degrees of disorder, thus extending the traditional idea of self-assembly to incorporate both amorphous and crystalline structures as well as quasicrystals. The notion of tailoring potentials that correspond to targeted structures is motivated by the rich fundamental statistical-mechanical issues and questions offered by this fascinating inverse problem as well as our recent ability to identify structures that have optimal bulk properties or desirable performance characteristics. Recent results have already led to a deeper basic understanding of the mathematical relationship between the collective structural behavior of many-body systems and their interactions, as well as optimized potentials that enable self-assembly of ordered and disordered particle configurations with novel structural and bulk properties.

I. INTRODUCTION

Inverse statistical-mechanical methods design interactions for targeted many-body structures rather than merely predicting structures from known interactions. These approaches use broad configurational information and aim to extend self-assembly across ordered, disordered, and technologically relevant materials.

  • Forward statistical mechanics predicts structural, thermodynamic, and kinetic features from known systems and approximated interactions.
  • Inverse methods optimize interactions so systems robustly and spontaneously form targeted many-particle configurations.
  • Colloids and polymers provide versatile experimental routes for realizing optimized soft interactions, while colloidal interactions can be tuned through surface modification or electrolytes.
  • Tailored potentials are motivated by fundamental inverse problems and by target structures with desirable properties, including photonic, thermal, mechanical, catalytic, separation, sensing, and electronic applications.
  • Unlike reverse Monte Carlo, these methods can seek non-pairwise or nonspherical potentials using complete configurational information to stabilize crystals, disordered structures, or quasicrystals.
  • The reviewed approaches address both known structures and potentially new structural motifs, including metastable states and nonequilibrium configurations.

III. INVERSE METHODS FOR CRYSTAL GROUND STATES

Crystal-ground-state inverse design begins with a simplified isotropic pair-potential model and seeks interactions that make a target crystal energetically and dynamically viable. Validation combines stability criteria with self-assembly from a liquid or other initial configuration.

  • Because the space of many-body potentials is infinitely large, the analysis begins with isotropic pairwise additive interactions as a practical approximation.
  • Isotropic pair interactions cannot produce thermodynamically stable chiral structures with specified handedness, but inverse methods can probe their broader structural limitations.
  • Nondirectional interactions with strong short-range repulsions can stabilize diamond structures, which are relevant to photonic band-gap research.
  • The zero-temperature and near-melting schemes together optimize potentials whose global energy minimum is the target configuration over a specified volume or density range.
  • Crystal validation examines a positive stability pressure or density range, real normal-mode frequencies, defect energies, and self-assembly during gradual cooling.
  • A stringent criterion requires the target crystal to emerge essentially defect-free from a liquid during slow annealing, although exact ground-state achievement lacks sufficient guarantees.

A. Zero-Temperature Optimization Scheme

The zero-temperature scheme parameterizes a family of pair potentials and searches its parameters for robust, defect-free assembly of a target crystal at fixed or preferably varying density.

  • The scheme selects parameters in a family ϕ(r; a0, a1, . . . , an) to produce robust, defect-free self-assembly of a target crystal.
  • Optimization is performed at absolute zero for a fixed density or, preferably, across a range of densities or equivalent pressures.
  • The resulting interaction is intended to support the target configuration under the chosen thermodynamic conditions.

B. Near-Melting Optimization Scheme

The near-melting scheme refines an initial optimized potential by simulating just below melting and minimizing a Lindemann-parameter objective to suppress liquid nucleation and promote assembly.

  • The procedure starts from the zero-temperature potential and optimizes assembly near but below the crystal melting point.
  • Simulations run at 80-95% of the melting temperature, chosen to avoid phase-transition fluctuations that would make calculations inconsistent.
  • The Lindemann parameter measures particle displacement relative to initial positions after an appropriate simulation time.
  • Simulated annealing varies the potential parameters to minimize the Lindemann-parameter objective function.
  • The ultimate test is defect-free target-crystal formation from a liquid during slow annealing to T = 0 within reasonable computer time.

IV. OPTIMIZED ISOTROPIC INTERACTIONS FOR LOW-COORDINATED CRYSTAL GROUND STATES

Inverse optimization shows that isotropic pair potentials can stabilize diverse low-coordinated crystal and cluster structures, while collective-coordinate methods extend targeting to disordered states and prescribed scattering. These results broaden the structures accessible through nondirectional interactions, although some optimized potentials remain difficult to synthesize.

  • Low-coordinated crystal ground states: Optimized isotropic pair potentials produce square and honeycomb lattices as two- and three-coordinated crystal ground states.The honeycomb crystal remains stable after temperature effects are included in the global phase diagram.
  • Robustness and realizability: The optimized square-lattice potential is simpler than the honeycomb potential, and self-assembly remains unaffected by perturbations in the potential.This robustness is identified as essential for experimental testing.
  • Unusual structures: Isotropic interactions also yield chain-like arrays, compact-cluster lattices, simple cubic, body-centered-cubic, and simple hexagonal crystals.The chain and cluster structures demonstrate that structurally anisotropic arrangements can emerge from circularly symmetric interactions.
  • Three-dimensional crystals: Strongly repulsive isotropic potentials stabilize diamond and wurtzite lattices, supported by lattice sums, phonon spectra, positive-energy defects, and molecular-dynamics self-assembly.The diamond target is challenging because it is structurally very close to wurtzite.
  • Disordered ground states: Collective-coordinate optimization generates disordered, degenerate ground states for sufficiently small χ, whereas χ of order unity produces periodic ground states.Increasing χ separates disordered, wavy-crystalline, and crystalline regimes.
  • Collective-coordinate designs: The same approach constructs disordered ground states whose radiation scattering matches prescribed patterns, including stealth, super-ideal-gas, and equi-luminous materials.Stealth configurations retain no long-range order in the infinite-volume limit, while higher χ increases mutual particle repulsion.

VI. DUALITY RELATIONS FOR CLASSICAL GROUND STATES

The paper introduces Fourier-space duality relations for analyzing classical ground states of soft pair potentials, linking real-space and reciprocal-space configurations. These relations yield ground-state results, phase-coexistence insights, and computational tools, while highlighting unresolved cases for some interactions.

  • Motivation: Ground-state determination remains a major challenge in condensed-matter physics and materials science, especially for interactions in Euclidean spaces of dimension two or higher.Although simulations can produce ground states by slowly freezing liquids, theoretical understanding remains incomplete.
  • Limitations: Ground states remain difficult to ascertain for some potentials and density regimes, including Lennard-Jones systems and square-mound interactions with nearest-neighbor distances below the cutoff.Some conclusions are supported by simulations rather than rigorous proofs, while clustered states can be rigorously analyzed for selected piecewise-constant interactions.
  • Duality framework: Fourier-transform duality relations connect ground-state energies and structures for admissible soft pair potentials with their reciprocal-space dual potentials.The framework uses bounded, absolutely integrable radial potentials whose Fourier transforms exist, and relates real-space lattices to reciprocal Bravais lattices.
  • Duality framework: When reciprocal-lattice configurations also minimize the dual potential, the duality energy inequality becomes an equality and identifies reciprocal real-space ground states.The reciprocal density is ρ^-1(2π)^-d, and equality corresponds to reciprocal lattice ground states.
  • Consequences: The duality results also reveal nonperiodic crystal-domain coexistence, infinitely many one-dimensional zero-temperature phase transitions, and a way to estimate energies or eliminate candidate ground states computationally.The framework transfers information between short-ranged and long-ranged potentials and can make computational ground-state searches more efficient.
  • Applications: Applying the relations to compactly supported and oscillatory potentials identifies ground-state lattices, degeneracies, clustered phases, and phase transitions across density ranges.Examples include fcc, bcc, simple hexagonal, and simple cubic lattices, while square-mound potentials admit rigorous clustered-ground-state analysis below the interaction discontinuity scale.

VII. CONSTRUCTION OF CONFIGURATIONS WITH TARGET PAIR CORRELATIONS

The section examines whether prescribed pair correlations can be realized by many-particle configurations and describes optimization-based constructions that test this problem. Numerical studies identify terminal densities and configurations realizing selected target correlations, while general realizability remains unresolved.

  • Realizability problem: The realizability problem asks whether a candidate g2(r) corresponds to a many-particle configuration at positive number density.Necessary conditions include nonnegative g2(r), nonnegative structure factor S(k), and constraints on local density fluctuations.
  • Realizability problem: Checking realizability is difficult because suitable higher-order correlation functions g3, g4, and beyond may not exist for a specified density and g2.The known necessary conditions are difficult or potentially impossible to verify in arbitrary dimension.
  • Model evidence: A one-dimensional nearest-neighbor-exclusion model exhibited pair-correlation realizability over a nonzero density range and violated the Kirkwood superposition approximation for g3.The study also found reverse Monte Carlo inappropriate for generating typical many-body configurations from a candidate pair correlation.
  • Optimization framework: A g2-invariant process preserves a prescribed nonnegative pair correlation as density varies, with terminal density defined as the maximum density satisfying known necessary conditions.Terminal densities for selected target correlations have been determined using numerical and analytical optimization.
  • Configuration construction: Stochastic optimization evolves an initial particle configuration until it matches targeted correlation functions up to specified cutoff distances.This construction strategy was used to test whether target correlations at terminal density are realizable by sphere packings.
  • Configuration construction: For a two-dimensional target shown in Fig. 6, 500 particles form only dimers at terminal packing fraction φ*=0.5 with average contact value Z=1.0.The targeted g2 form is realized up to dimensionless distance r/D=2.5.
  • Open problems: Although numerical evidence supports realizability of the selected pair correlation for sphere packings in dimensions d≥2, the general realizability problem remains an active research question.A conjecture relating realizability to S(k)≥0 in asymptotically large dimensions remains unproved.

VIII. DESIGNING ISOTROPIC PAIR POTENTIALS FOR TARGETED BULK PROPERTIES

Inverse methods can optimize interactions to produce targeted bulk properties, illustrated through thermal expansion coefficients and Poisson’s ratio.

  • Targeted bulk properties: Inverse methods optimize many-particle interactions to achieve targeted novel bulk properties, including thermal expansion coefficients and Poisson’s ratio.The section uses these two properties as specific target examples.

A. Thermal Expansion Coefficients

The section describes inverse design of isotropic interactions for negative thermal expansion in equilibrium solid systems. An optimized softened interior core potential produces negative, zero, and positive expansion across temperature before melting.

  • Motivation: Zero, very large, and negative thermal expansion coefficients have distinct technological motivations, including dimensional stability and actuation.Zero expansion can improve longevity of space structures, bridges, and piping systems.
  • Prior context: Before this work, negative thermal expansion had been observed only in multicomponent materials with open unit cells and highly directional bonding.Ice and zirconium tungstate are cited as examples of materials exhibiting negative thermal expansion.
  • Inverse design result: An optimized isotropic potential produces negative thermal expansion in equilibrium single-component solids in two and three dimensions across a wide temperature and pressure range, including zero pressure.The result established this behavior without directional interactions.
  • Mechanism: A softened interior core within a basin of attraction is a sufficient condition for negative thermal expansion in the described systems.The mechanism is illustrated schematically in Fig. 7(a).
  • Mechanism: As temperature increases, the optimized softened interior core potential yields negative, zero, and then positive thermal expansion before melting in both two and three dimensions.The result comes from optimization and constant-pressure Monte Carlo simulations.

B. Poisson’s Ratio

The section considers inverse design of negative Poisson’s ratio, or auxetic behavior, using isotropic interactions. Pair potentials require negative pressure and specific constraints, while a three-body potential can impose the behavior at zero temperature and positive pressure.

  • Motivation: Negative Poisson’s ratio is an unusual target property in which stretching in one direction causes expansion in an orthogonal direction.Such auxetic behavior has been observed in relatively few elastically isotropic materials, often with intricate structures.
  • Pair-potential design: Isotropic two-body potentials can produce negative Poisson’s ratio in two- and three-dimensional crystal phases under tension when specified equalities and inequalities are satisfied.This result contrasts with the complex anisotropic interactions common in previously known auxetic materials.
  • Scope and limitations: Under pair interactions with elastic isotropy, the analysis suggests auxetic behavior occurs only in crystals under the nonequilibrium condition of negative pressure.This scope creates a practical boundary for pair-potential realizations.
  • Many-body design: A three-body potential yields negative Poisson’s ratio by construction in close-packed two- and three-dimensional lattices at zero temperature and positive pressure.Its design includes an energy cost for deforming equilateral triangles in the triangular lattice and analogous close-packed structures.
  • Pair-potential design: For Lennard-Jones interactions, the negative-Poisson’s-ratio region occurs at negative pressure, while the lattice becomes unstable beyond the permitted lattice-constant range.The figure marks positive pressure to the left of the dotted line and negative pressure to the right.

IX. FUTURE WORK AND CONCLUSIONS

The section signals a discussion of future directions followed by concluding remarks.

  • The section discusses future directions.
  • The section closes with concluding remarks.
  • The section combines forward-looking discussion with a formal conclusion.

A. Interaction Potentials for Targeted Configurations at Positive Temperature

Inverse interaction-design methods can be extended beyond zero-temperature targets to positive-temperature configurations and multicomponent systems. Patchy and anisotropic interactions broaden the structures that can be targeted, including crystals, amorphous structures, and quasicrystals.

  • Interaction Potentials for Targeted Configurations at Positive Temperature: Inverse techniques developed for ground-state structures can be extended to positive-temperature configurations and defect control.The passage identifies point, line, and planar crystal defects under varied growth conditions as desirable targets.
  • Interaction Potentials for Targeted Configurations at Positive Temperature: Multicomponent optimization enlarges the parameter space through species composition and effective particle-size ratios, requiring careful potential and target selection.The search can be made manageable by restricting potentials to forms consistent with colloidal interactions.
  • Interaction Potentials for Targeted Configurations at Positive Temperature: Anisotropic pair interactions offer greater flexibility for targeting structures, although some nonconventional lattices can be stabilized with radial pair potentials.The reviewed work reports that angle-dependent or non-additive interactions are not always necessary.
  • Interaction Potentials for Targeted Configurations at Positive Temperature: Patchy particles have produced chains, sheets, rings, icosahedra, square pyramids, tetrahedra, and twisted and staircase structures through surface-pattern design.These particles are proposed as building blocks for unique colloidal structures and as candidates for targeting low-coordinated crystals, amorphous structures, and quasicrystals.

D. Inverse Optimization Methods for Novel Targeted Bulk Properties

The paper identifies unresolved opportunities in inverse optimization for bulk properties, unusual phase behavior, experimentally realizable interactions, and nonequilibrium self-assembly. It argues that broader optimization schemes and dynamical approaches could expand control over material structures and properties.

  • Inverse Optimization Methods for Novel Targeted Bulk Properties: A general optimization scheme for interactions targeting bulk properties across broad potential families and conditions has not yet been devised.The paper gives auxetic behavior over wide temperature and pressure ranges as a challenging example.
  • Inverse Optimization Methods for Novel Targeted Bulk Properties: Inverse melting reverses conventional behavior because heating the liquid at constant pressure causes crystallization, but robust optimized interactions for this behavior remain undeveloped.The phenomenon is described as a first-order crystal–liquid transition in which the crystal has higher entropy than the coexisting isotropic liquid.
  • Inverse Optimization Methods for Novel Targeted Bulk Properties: Future work should develop robust, experimentally synthesizable potentials for targeted structures and bulk properties.The paper notes that some optimized interactions are already standard or potentially laboratory-feasible, while others cannot currently be synthesized.
  • Inverse Optimization Methods for Novel Targeted Bulk Properties: Effective pair potentials must be related to intrinsic three-body and higher-order interactions present in nondilute materials.The paper identifies this correspondence as important because real many-particle interactions are necessarily nonadditive.
  • Inverse Optimization Methods for Novel Targeted Bulk Properties: Inverse optimization of many-particle dynamics remains undeveloped, despite nonequilibrium systems that can self-organize into quiescent states separated from fluctuating states by dynamical phase transitions.The reviewed equilibrium-focused approaches therefore leave a conjugate kinetic problem for future development.
Loading 0811.0040v2…