Pith. sign in

REVIEW 4 major objections 5 minor 55 references

Algorithmic Discovery of Casimir-Polder forces: Repulsion in the Ground State

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper presents an inverse-design algorithm that automatically sculpts a structure to produce a desired Casimir-Polder force, and demonstrates it by finding a funnel-like gold shape that repels a ground-state atom.

desk verdict Clever new inverse-design pipeline for Casimir-Polder forces, but the headline repulsion rests on an unvalidated time-domain proxy—worth refereeing, not yet believed. read the letter →

arxiv 2412.01483 v1 pith:BK4NIXKJ submitted 2024-12-02 quant-ph physics.atom-phphysics.comp-ph

Romuald Kilianski , Claire M. Cisowski , Robert Bennett This is my paper · ORCID
classification quant-phphysics.atom-phphysics.comp-ph
keywords Casimir-Polderforceinversedesignadjointmethodlevel-setoptimizationrepulsivedispersionforcesground-stateatomtime-domainelectrodynamicsfinite-difference
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 establishes a general inverse-design algorithm for Casimir-Polder forces: instead of testing hand-picked geometries against a force target, the user specifies the atom, the target force direction, and an arbitrary starting shape, and the algorithm iteratively deforms the shape until the desired force is produced. The demonstration targets the most elusive case, repulsion of a ground-state atom, and reports that a gold cylinder evolves into a funnel-like structure whose merit function settles at a negative, repulsive value after 12 iterations. The final shape resembles previously known ring and plate-with-hole geometries but was reached with no user-supplied geometric input, which is the central claim of the paper. A sympathetic reader should care because the approach promises to turn force engineering from a case-by-case analytic exercise into a general computational design tool.

What carries the argument

The engine is the time-domain Casimir-Polder convolution $$U_{\rm CP}(r_A) = -\hbar \int_0^\infty dt\,{\rm Im}[g_x(-t)] $E_x^{{(1)}}$(r_A,t),$$ which converts the imaginary-frequency Green's-tensor integral into a single convolution of a source function $g_x$ with the scattered electric field, computable with a finite-difference time-domain solver. Around this, the paper builds an adjoint method: the change in the force merit function is proportional to the overlap $\int dt'\, E(r',t')\cdot E^A(r',t')$ of forward and time-reversed adjoint fields, so only two simulations are needed per iteration. This overlap defines a normal velocity $v_n$ for the level-set advection equation $\partial\Phi/\partial\tau + v_n|\nabla\Phi| = 0$, which pushes the shape boundary in the direction that makes the merit function more negative.

What would settle it

Compute the full imaginary-frequency Casimir-Polder potential of the optimized funnel with an independent frequency-domain Green's-tensor solver and evaluate the force along the axis at the atom's position; a positive (attractive) value at the location where the merit function is negative would show the proxy misled the optimization.

Watch

Extended reading notes

Core claim

On its own terms, the paper claims that repulsive Casimir-Polder forces on ground-state atoms need not be hand-engineered: they can be discovered by specifying only the atom's polarizability and a target force direction, then letting an adjoint-based level-set optimizer deform an arbitrary initial geometry. The demonstration starts with a gold cylinder in front of an x-polarized ground-state rubidium-87 atom and, after 12 iterations, the merit function reaches a negative (repulsive) value. The final geometry is funnel-like, with a central indentation that deepens into a hole, and the paper notes it is reminiscent of, but not identical to, the ring and plate-with-hole geometries previously known to give repulsion. Because the discovery is independent of user input about shape, the paper's central claim is that the method inverts the usual workflow, turning 'find a geometry that does X' into an automated optimization.

Load-bearing premise

The algorithm optimizes a time-domain proxy potential built from a truncated current pulse with a cutoff, and the paper does not independently verify that this proxy reproduces the exact frequency-domain Casimir-Polder potential for the optimized funnel; if the proxy disagrees with the exact potential, the final shape may not actually repel the atom.

Editorial extensions

If this is right

  • After 12 iterations, the merit function converges to a negative (repulsive) value, and the optimized structure is a funnel-like gold shape whose central dent deepens into a hole.
  • The discovered geometry was reached without user-supplied geometric hints; only the atomic polarizability, the repulsion goal, and an arbitrary starting cylinder were specified.
  • Because the force variation is computed from only two field simulations per iteration, the algorithm can in principle scale to complex three-dimensional structures and real dispersive materials.
  • Repulsion is not universal: the atom-surface separation must fall in an intermediate window (roughly below $\lambda/2$), and a non-perforated initial structure placed too close to the atom cannot be optimized into repulsion.
  • The same machinery can be aimed at other directions or signs of the Casimir-Polder force, and the authors state it could be extended to excited atoms.

Reading between the lines

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

  • If the time-domain proxy is as faithful as the paper assumes, the same pipeline should be able to design lateral forces, specified force magnitudes, or torques simply by changing the merit function, since the paper only demonstrates one target.
  • The algorithm's independent rediscovery of ring-like and perforated shapes suggests the hole is a robust topological feature of repulsion for anisotropic ground-state atoms, rather than a quirk of the optimizer; running the optimization from several unrelated starting geometries and checking for convergence to the same shape class would test this.
  • A direct technical extension would be to insert an independent frequency-domain check of the optimized structure's Casimir-Polder force as a validation step, which would separate proxy artifacts from genuine repulsion.
  • Because the repulsion window is short, experimental realization would hinge on positioning precision: a fabricated funnel would need to be placed within that window to observe the predicted effect.
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.

Referee Report

4 major / 5 minor

Summary. The paper presents an inverse-design algorithm for producing a structure that exerts a desired Casimir-Polder force on a ground-state atom. The method combines the time-domain representation of the Casimir-Polder potential (Eq. 1), an adjoint-based shape derivative (Eqs. 2–4), and level-set advection. As a demonstration, the authors start from a gold cylinder and, after 12 iterations, obtain a funnel-like shape for which their merit function takes a negative, repulsive value (Fig. 3). The authors interpret this as algorithmic discovery of a geometry supporting ground-state Casimir-Polder repulsion, reminiscent of previously known ring-like geometries.

Significance. If the central claim is correct, the paper introduces a promising new tool: it replaces the user-driven search over candidate geometries with an automatic level-set optimization that needs only two FDTD simulations per iteration. The approach is clearly presented and builds on established adjoint and time-domain techniques. However, the reported repulsion is currently supported only by the optimizer's own time-domain proxy, without an independent validation of the physical Casimir-Polder force. The derivation of the shape-update velocity also relies on an unexamined sign/constant assumption for the material response. These issues must be resolved before the main claim can be accepted.

major comments (4)
  1. [Eq. (1) and Fig. 3] The central numerical evidence for the headline claim is the time-domain merit function F_x obtained from Eq. (1) with one particular source profile J(t) and cutoff gamma=2.5c/L0. No independent evaluation of the physical Casimir-Polder force on the final structure is reported, so the negative value in Fig. 3 is only the optimizer's own objective. Please validate the method against a known result (for example, an infinite plane or the analytic ring/plate-with-hole geometries of Refs. [25,28,29]), and recompute the final funnel's Casimir-Polder potential with an independent frequency-domain imaginary-frequency integral, or at least with a substantially different source cutoff and simulation window. Without such a check, the sign of the proxy cannot be equated with a physical repulsive force.
  2. [Eqs. (3)–(4)] The derivation of the shape-update velocity replaces P(r,t) with E(r,t) and drops the proportionality constant, stating that it is positive. For a Drude metal in vacuum, the susceptibility epsilon(omega)-1 is negative over the low-frequency range that dominates the Casimir-Polder interaction, so the sign of the shape derivative is not fixed by the argument given. The chosen level-set velocity v_n = partial_{x'} int dt' E(r',t') . E_A(r',t') is therefore not guaranteed to be a descent direction for the actual merit function. Please derive the variation of F_x with the explicit frequency-dependent material contrast for the gold Drude model, or otherwise justify the sign of the velocity.
  3. [Setup, p. 4 and Eq. (1)] The calculation keeps only the x-x component of the Green's tensor, written as alpha(omega)=alpha(omega)delta_{ix}. However, the setup is described as a rubidium-87 atom, which has an isotropic scalar polarizability. For an isotropic atom the Casimir-Polder potential requires the trace alpha(G_xx+G_yy+G_zz), and the omitted y and z polarization components could contribute an attractive force that overwhelms the x-component signal. Please state clearly that the demonstration applies to a model atom with only x-direction polarizability, or perform the full trace calculation to support the general claim of ground-state repulsion.
  4. [Fig. 3 and numerical convergence] No convergence or error analysis is reported for the FDTD optimizations: no cell size, timestep, spectral resolution, or dependence on the finite time window and damping rate gamma is given. Since the method's usefulness rests on the reliability of the optimized geometry, at least one robustness test (for example, varying gamma or the grid resolution and observing whether the final shape and sign of F_x remain unchanged) should be included.
minor comments (5)
  1. [Fig. 3 and sign convention] The text says the merit value is negative (repulsive), but from F_CP = -grad U and Eq. (1), a positive F_x would usually denote a force pointing away from the structure. Please define the sign convention used in Fig. 3 explicitly.
  2. [Abstract and p. 2] The statement that the geometry is discovered 'completely independently of any input from the user' is overstated, because the user provides the initial cylinder, the simulation domain, the goal of repulsion, and the atom parameters.
  3. [Eq. (2)] The integration domains T and T' and the perturbed volume V' are not defined in the text. Please specify the time intervals and how they relate to the finite FDTD simulation window.
  4. [Polarizability notation] The notation alpha(omega)=alpha(omega)delta_{ix} is ambiguous for a tensor; a clearer form would be alpha_{ij}(omega)=alpha(omega)delta_{ix}delta_{jx}.
  5. [General] There are several typographical issues, including 'an non-isotropic' on page 4 and 'is an stumbling block' in the introduction. A careful proofread would improve the presentation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the optimized geometry is an output of the stated objective, and the time-domain formula is an exact convolution identity from an external reference.

full rationale

None of the paper's load-bearing steps reduces to its own inputs. Equation (1) is presented as a convolution identity relating the standard frequency-domain Casimir-Polder potential to a time-domain scattered field, with g_x(omega) defined so that the convolution reproduces U_CP; this is a mathematical rewriting, not a fit. The optimization uses this U_CP (through the merit function F_x) only as the objective, and the final funnel geometry is the output of the level-set/advection dynamics, not a parameter fitted to any subset of data. The time-domain formulation and the specific source J(t) are taken from Ref. [49], which is an external (non-author) reference. The authors' own prior works [39,44] appear only as contextual citations and are not load-bearing for the algorithm or for the sign of the result. The absence of an independent frequency-domain validation of the truncated-source proxy is a correctness/robustness concern, not circularity: a wrong proxy would make the result incorrect, but it would not make the derivation equivalent to its assumptions. No uniqueness theorem and no author-specific ansatz is invoked to force the outcome, and the discovered geometry is not a renamed version of a known result. Therefore no circular step is exhibited.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The ledger captures the user-chosen simulation parameters and modeling choices that the central claim depends on. No new physical entities, particles, forces, or dimensions are introduced; the work is a numerical design method.

free parameters (3)
  • current cutoff gamma = 2.5 c / L0 with L0 = 100 nm
    Chosen by hand to set the source pulse cutoff; controls the spectral window of the time-domain convolution, so the optimization objective is a proxy for the physical Casimir-Polder force.
  • operational length scale L0 = 100 nm
    User-chosen length scale; all dimensions, distances, and gamma are expressed in units of L0, and it determines which distance regime of the force is probed.
  • initial cylinder geometry = radius 1.5 L0, height 0.4 L0, center at 1.1 L0 from the dipole
    User-specified starting geometry for the level-set optimization; the algorithm's convergence to repulsion may depend on this initialization.
assumptions (4)
  • domain assumption The time-domain convolution in Eq. (1) with the chosen current source J(t) exactly reproduces the imaginary-frequency Casimir-Polder potential U_CP.
    Invoked from Ref. [49]; the optimization objective inherits any error from source spectral coverage or finite-time truncation.
  • domain assumption The electromagnetic medium is reciprocal, allowing the Green's tensor source-observer swap used to compute the adjoint field backwards in time.
    Stated after Eq. (2); gold is reciprocal, so this holds for the demonstration, but the claimed generality would fail for non-reciprocal media without extra simulations.
  • ad hoc to paper P(r,t) is proportional to E(r,t) in the perturbed volume, allowing positive constants to be dropped in the shape derivative.
    Explicitly assumed after Eq. (3); for dispersive gold this is an approximation and the shape update may not be an exact gradient of the true force.
  • ad hoc to paper The level-set velocity v_n = partial_x' integral dt' E dot E_A gives a descent direction for the force objective.
    By construction delta_Fx = -integral dA v_n^2 delta_tau is non-positive, but this only guarantees monotonic decrease of the proxy, not of the physical force unless the preceding gradient derivation is exact.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Algorithmic Discovery of Casimir-Polder forces: Repulsion in the Ground State." pith.science (2026). https://pith.science/paper/BK4NIXKJ

@misc{pith2026241201483,
  author       = {Pith},
  title        = {Pith review of: Algorithmic Discovery of Casimir-Polder forces: Repulsion in the Ground State},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BK4NIXKJ}},
  note         = {Machine review of arXiv:2412.01483}
}
read the original abstract

We present a general-purpose algorithm for automatic production of a structure that induces a desired Casimir-Polder force. As a demonstration of the capability and wide applicability of the method, we use it to develop a geometry that leads to a repulsive Casimir-Polder force on a ground-state atom. The results turn out to be reminiscent of the ring-like geometries previously used to induce repulsion, but with some new features and -- importantly -- discovered completely independently of any input from the user. This represents a powerful new paradigm in the study of atom-surface forces -- instead of the user testing various geometries against a desired figure of merit, the goal can be specified and then an appropriate geometry created automatically.

Figures

Figures reproduced from arXiv: 2412.01483 by the authors.

Figure 1
Figure 1. FIG. 1. Overview of the concept behind this work. All simple [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Overview of the algorithm. (a) An atomic dipole (in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. The change in the normalized merit function [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 50 canonical work pages

  1. [49]

    Rodriguez, M

    A. Rodriguez, M. Ibanescu, D. Iannuzzi, J. D. Joannopoulos, and S. G. Johnson, Phys. Rev. A 76, 032106 (2007)

    work page 2007

  2. [1]

    This in turn changes the forward and adjoint fields due to their having new boundary conditions, so the process repeats with the new geometry until the goal is reached

    incrementally migrate toward a new position that will be advantageous with respect to our merit function Fx, shown in (d). This in turn changes the forward and adjoint fields due to their having new boundary conditions, so the process repeats with the new geometry until the goal is reached. [52], operating on the principle of modifying and stretch- ing an...

    work page

  3. [2]

    H. B. G. Casimir, Proc. K. Ned. Akad. 360, 793 (1948)

    work page 1948

  4. [3]

    P. W. Milonni, The quantum vacuum: an introduction to quantum electrodynamics(Academic press, 2013)

    work page 2013

  5. [4]

    S. K. Lamoreaux, Physical Review Letters 78, 5 (1997)

    1997

  6. [5]

    M. J. Sparnaay, Nature 180, 334 (1957)

    work page 1957

  7. [6]

    R. S. Decca, D. L´ opez, E. Fischbach, G. L. Klimchit- skaya, D. E. Krause, and V. M. Mostepanenko, Physical Review D 75, 077101 (2007)

    work page 2007

  8. [7]

    Mohideen and A

    U. Mohideen and A. Roy, Physical Review Letters 81, 4549 (1998)

    1998

Show all 55 references
  1. [8]

    J. L. Garrett, D. A. Somers, and J. N. Munday, Physical Review Letters 120, 040401 (2018)

    2018

  2. [10]

    E. L. I.E. Dzyaloshinskii and L. Pitaevskii, Advances in Physics 10, 165 (1961)

    1961

  3. [11]

    E. M. Lifshitz, Zh. Eksp. Teor. Fiz. 29, 94 (1956), [Sov. Phys. JETP. 2, 73 (1956)]

    1956

  4. [12]

    Some authors refer to interactions at electrostatic dis- tances as van der Waals forces

    It is crucial for us to point out that we refer to any in- teraction between a polarizable atom and a macroscopic surface, regardless of the distance regime, as CP forces. Some authors refer to interactions at electrostatic dis- tances as van der Waals forces

  5. [13]

    S. Y. Buhmann, Dispersion Forces I - Macroscopic Quantum Electrodynamics and Ground-State Casimir, Casimir-Polder and van Der Waals Forces , Vol. 247 (Springer, Berlin, 2012)

    2012

  6. [14]

    A. M. Marvin and F. Toigo, Physical Review A 25, 782 (1982)

    1982

  7. [15]

    M. J. Mehl and W. L. Schaich, Surface Science 99, 553 (1980)

    1980

  8. [16]

    Brevik, M

    I. Brevik, M. Lygren, and V. N. Marachevsky, Annals of Physics 267, 134 (1998)

    1998

  9. [17]

    J. M. Wylie and J. E. Sipe, Physical Review A 32, 2030 (1985)

    1985

  10. [18]

    D. A. R. Dalvit, P. A. Maia Neto, A. Lambrecht, and S. Reynaud, Physical Review Letters100, 040405 (2008)

    2008

  11. [19]

    L. H. Ford and N. F. Svaiter, Physical Review A 62, 062105 (2000)

    2000

  12. [20]

    Buks and M

    E. Buks and M. L. Roukes, Phys. Rev. B 63, 033402 (2001)

    2001

  13. [21]

    S. Y. Buhmann, D.-G. Welsch, and T. Kampf, Physical Review A 72, 032112 (2005)

    2005

  14. [22]

    S. J. Rahi, M. Kardar, and T. Emig, Phys. Rev. Lett. 105, 070404 (2010)

    2010

  15. [23]

    Maluf and K

    N. Maluf and K. Williams, An Introduction to Microelec- tromechanical Systems Engineering, Artech House Mi- croelectromechanical Systems (Artech House, 2004)

    2004

  16. [24]

    J. N. Munday, F. Capasso, and V. A. Parsegian, Nature 457, 170 (2009)

    2009

  17. [25]

    T. H. Boyer, Phys. Rev. A 9, 2078 (1974)

    1974

  18. [26]

    K. A. Milton, P. Parashar, N. Pourtolami, and I. Brevik, Physical Review D 85, 025008 (2012)

    2012

  19. [27]

    K. A. Milton, E. K. Abalo, P. Parashar, N. Pourtolami, I. Brevik, and S. ˚A. Ellingsen, Physical Review A 83, 062507 (2011)

    2011

  20. [28]

    P. P. Abrantes, Y. Fran¸ ca, F. S. S. da Rosa, C. Farina, and R. de Melo e Souza, Phys. Rev. A 98, 012511 (2018)

    2018

  21. [29]

    K. A. Milton, E. K. Abalo, P. Parashar, N. Pourtolami, I. Brevik, and S. ˚A. Ellingsen, Journal of Physics A: Mathematical and Theoretical 45, 374006 (2012)

    2012

  22. [30]

    A. W. Rodriguez, J. D. Joannopoulos, and S. G. Johnson, Phys. Rev. A 77, 062107 (2008)

    2008

  23. [31]

    Eberlein and R

    C. Eberlein and R. Zietal, Phys. Rev. A 83, 052514 (2011)

    2011

  24. [32]

    Jameson, Journal of Scientific Computing 3, 233 (1988)

    A. Jameson, Journal of Scientific Computing 3, 233 (1988)

    1988

  25. [33]

    M. P. Bendsøe and N. Kikuchi, Computer Methods in Applied Mechanics and Engineering 71, 197 (1988)

    1988

  26. [34]

    S. Wang, S. Wen, Z.-L. Deng, X. Li, and Y. Yang, Phys- ical Review Letters 130, 123801 (2023)

    2023

  27. [35]

    Molesky, Z

    S. Molesky, Z. Lin, A. Y. Piggott, W. Jin, J. Vuckovi´ c, and A. W. Rodriguez, Nature Photonics 12, 659 (2018)

    2018

  28. [36]

    Cordaro, B

    A. Cordaro, B. Edwards, V. Nikkhah, A. Al` u, N. En- gheta, and A. Polman, Nature Nanotechnology 18, 365 (2023)

    2023

  29. [37]

    Sen and A

    S. Sen and A. K. Mitchell, Physical Review Letters 133, 076501 (2024)

    2024

  30. [38]

    S. Goel, S. Leedumrongwatthanakun, N. H. Valencia, W. McCutcheon, A. Tavakoli, C. Conti, P. W. H. Pinkse, and M. Malik, Nature Physics 20, 232 (2024)

    2024

  31. [39]

    Bordiga, E

    G. Bordiga, E. Medina, S. Jafarzadeh, C. B¨ osch, R. P. Adams, V. Tournat, and K. Bertoldi, Nature Materials 23, 1486 (2024)

    2024

  32. [40]

    Bennett, Physical Review A 103, 013706 (2021)

    R. Bennett, Physical Review A 103, 013706 (2021)

    2021

  33. [41]

    Bennett and S

    R. Bennett and S. Y. Buhmann, New Journal of Physics 22, 093014 (2020)

    2020

  34. [42]

    J. R. Capers, S. J. Boyes, A. P. Hibbins, and S. A. R. Horsley, Communications Physics 4, 209 (2021)

    2021

  35. [43]

    Matuszak, S

    J. Matuszak, S. Y. Buhmann, and R. Bennett, Physical Review A 106, 013515 (2022)

    2022

  36. [44]

    C. M. Cisowski, M. C. Waller, and R. Bennett, Phys. Rev. A 109, 043533 (2024)

    2024

  37. [45]

    Miguel-Torcal, J

    A. Miguel-Torcal, J. Abad-Arredondo, F. J. Garcia- Vidal, and A. I. Fernandez-Dominguez, Nanophotonics 11, 4387 (2022)

    2022

  38. [46]

    S. Y. Buhmann and D.-G. Welsch, Applied Physics B 82, 189 (2006)

    2006

  39. [47]

    O. D. Miller, Ph. D. Thesis (2012)

    2012

  40. [48]

    A. W. Rodriguez, A. P. McCauley, J. D. Joannopoulos, and S. G. Johnson, Phys. Rev. A 80, 012115 (2009)

    2009

  41. [50]

    A. F. Oskooi, D. Roundy, M. Ibanescu, P. Bermel, J. Joannopoulos, and S. G. Johnson, Computer Physics Communications 181, 687 (2010)

    2010

  42. [51]

    P. T. Kristensen, B. Beverungen, F. Intravaia, and K. Busch, Phys. Rev. B 108, 205424 (2023)

    2023

  43. [52]

    Osher and J

    S. Osher and J. A. Sethian, Journal of Computational Physics 79, 12 (1988). 6

    1988

  44. [53]

    and scikit-fmm [54] to extend the velocity over the entire domain), in turn given by our simulated forward and adjoint fields, we drive the evolution of our shape’s boundaries as prescribed by the particular velocity field. We now have all the tools ready to proceed with the a...

    2020

  45. [54]

    The term forward is conventionally defined as opposite- to-adjoint, and not as forward-time; however, this inter- pretation applies in this work

  46. [55]

    J. E. Guyer, D. Wheeler, and J. A. Warren, Computing in Science & Engineering 11, 6 (2009)

    2009

  47. [56]

    Stein and contributors, scikit-fmm: The fast marching method for python (2024), accessed: 2024-07-01

    D. Stein and contributors, scikit-fmm: The fast marching method for python (2024), accessed: 2024-07-01

    2024

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.