Casimir-Polder Force

• Casimir-Polder force is a quantum electrodynamic interaction between polarizable objects that transitions from the non-retarded (r⁻³ or r⁻⁶) regime to the retarded (r⁻⁴ or r⁻⁷) regime with distinct scaling laws.
• It is formulated using electromagnetic Green tensors and Lifshitz-type methods, which reveal how medium effects, finite size, and thermal fluctuations shape measurable forces at micrometer separations.
• Recent studies extend its analysis to anisotropic, chiral, and nonequilibrium systems, enabling tunable interactions and even repulsion in engineered media and structured environments.

The Casimir–Polder force is the retarded, quantum-electrodynamic generalization of van der Waals dispersion forces between polarizable objects, and the atom–surface analogue of the Casimir effect. It arises from quantum, and at finite temperature also thermal, fluctuations of the electromagnetic field coupled to fluctuating dipoles or currents in matter. Predicted by Casimir and Polder in 1948, it is conventionally described either as a position-dependent energy shift generated by modified vacuum modes or as a fluctuation-induced interaction mediated by electromagnetic Green tensors and reflection coefficients (Thiyam et al., 2014, Messina et al., 2012).

1. Definition and limiting behavior

For neutral polarizable particles, the non-retarded limit is the London–van der Waals interaction, with energy scaling as $U\sim r^{-6}$ for two atoms or molecules. When retardation becomes relevant, the interaction crosses over to the Casimir–Polder regime; in vacuum, the two-atom interaction scales as $r^{-7}$ (Thiyam et al., 2014, Messina et al., 2012). For an atom or small sphere above a planar surface, the standard asymptotic forms are

$U(z) = -\frac{C_3}{z^3}$

in the nonretarded regime and

$U(z) = -\frac{C_4}{z^4}$

in the retarded regime (Cherroret et al., 2017).

For a perfect conductor and isotropic static polarizability $\alpha(0)$ at zero temperature, the retarded atom–surface potential is

$$\overline{U}(z) = -\frac{3\hbar c \alpha(0)}{32\pi^2\epsilon_0 z^4},$$

with force

$$\overline{F}(z) = -\frac{3\hbar c \alpha(0)}{8\pi^2\epsilon_0 z^5}.$$

At finite temperature, for distances larger than the thermal wavelength

$\lambda_T = \frac{\hbar c}{k_B T},$

the mean potential above a metallic plate behaves as

$$\overline{U}(z) = -\frac{1}{16\pi}\frac{\hbar c \alpha(0)}{\epsilon_0 \lambda_T z^3},$$

so the force decays as $z^{-4}$ in the thermal regime (Cherroret et al., 2017).

The force is generally defined by the gradient of the interaction energy. In the atom–wall geometry this is written as

$r^{-7}$0

and analogous definitions apply for radial or lateral components in more general geometries (Klimchitskaya et al., 2021).

2. Field-theoretic and Lifshitz formulations

A standard formulation expresses the Casimir–Polder interaction in terms of the electromagnetic Green tensor. For an isotropic atom with scalar polarizability $r^{-7}$1 at position $r^{-7}$2 in a planar inhomogeneous medium, the energy can be written as

$r^{-7}$3

where $r^{-7}$4 and $r^{-7}$5 are reduced TE and TM Green’s functions (Milton, 2018). In planar geometries, this is the macroscopic-QED counterpart of the Lifshitz description based on reflection amplitudes.

For two polarizable particles embedded in a medium, the free energy takes a Lifshitz-type form. For two identical isotropic molecules of finite size in water,

$r^{-7}$6

with Matsubara frequencies $r^{-7}$7, excess polarizability $r^{-7}$8, and medium-modified Green-tensor components $r^{-7}$9 (Thiyam et al., 2014). Expanding the logarithm gives the familiar weak-coupling form

$U(z) = -\frac{C_3}{z^3}$0

which, for methane in water, was found to agree very well with the full non-expanded expression once finite molecular size is included (Thiyam et al., 2014).

In inhomogeneous backgrounds, renormalization is achieved by subtracting the interaction with the local inhomogeneous medium by itself. This subtraction removes the divergent self-interaction of the atom with the analytic continuation of the background and leaves the finite interaction with the nonanalytic feature, such as a plate or an interface (Milton, 2018). This framework generalizes the Dzyaloshinskii–Lifshitz–Pitaevskii interaction from homogeneous dielectric interfaces to spatially inhomogeneous dielectric media (Milton, 2018).

3. Medium effects, finite size, and condensed-phase interactions

Embedding media modify Casimir–Polder interactions in three distinct ways. First, the vacuum polarizability $U(z) = -\frac{C_3}{z^3}$1 is replaced by an excess polarizability $U(z) = -\frac{C_3}{z^3}$2, which for a dielectric sphere in water is

$U(z) = -\frac{C_3}{z^3}$3

Second, the effective speed of light becomes

$U(z) = -\frac{C_3}{z^3}$4

shifting the crossover between non-retarded and retarded regimes. Third, finite molecular size regularizes the short-distance divergence of the point-dipole model; for methane, a Gaussian radius $U(z) = -\frac{C_3}{z^3}$5 was used (Thiyam et al., 2014).

For two methane molecules dissolved in water, the dispersion contribution to the binding energy at contact separation $U(z) = -\frac{C_3}{z^3}$6 is

$U(z) = -\frac{C_3}{z^3}$7

at room temperature. The interaction is attractive, finite-size regularization removes the short-distance divergence, and the series-expanded and full non-expanded theories agree at all separations once screening by water is included (Thiyam et al., 2014).

For a methane molecule in water near planar interfaces, the non-retarded molecule–surface interaction is written as

$U(z) = -\frac{C_3}{z^3}$8

where

$U(z) = -\frac{C_3}{z^3}$9

The sign is controlled by dielectric contrast across the Matsubara spectrum (Thiyam et al., 2014).

At contact distance $U(z) = -\frac{C_4}{z^4}$0, the van der Waals binding energies are as follows (Thiyam et al., 2014):

Background	Surface	$U(z) = -\frac{C_4}{z^4}$1
water	SiO$U(z) = -\frac{C_4}{z^4}$2	$U(z) = -\frac{C_4}{z^4}$3
water	hexane	$U(z) = -\frac{C_4}{z^4}$4
water	air	$U(z) = -\frac{C_4}{z^4}$5

These values imply attraction to hydrophilic SiO$U(z) = -\frac{C_4}{z^4}$6 and repulsion from hexane and air in water, showing that medium response can weaken, reshape, and even reverse dispersion interactions (Thiyam et al., 2014).

4. Nonequilibrium, dynamical, and fluctuating forces

Out of thermal equilibrium, the atom–surface force is no longer determined by a single-temperature Lifshitz formula. In the Lifshitz–Antezza–Pitaevskii–Stringari framework,

$U(z) = -\frac{C_4}{z^4}$7

where $U(z) = -\frac{C_4}{z^4}$8 is wall temperature and $U(z) = -\frac{C_4}{z^4}$9 is the environment or atomic temperature (Klimchitskaya et al., 2021). In the micrometer thermal regime, the equilibrium contribution scales as $\alpha(0)$0, whereas the nonequilibrium contribution scales as $\alpha(0)$1: for a dielectric wall,

$\alpha(0)$2

and for a metallic wall,

$\alpha(0)$3

This makes nonequilibrium terms relatively more important at large separation (Klimchitskaya et al., 2021).

For a sapphire wall coated with a $\alpha(0)$4 VO$\alpha(0)$5 film, the dielectric-to-metal phase transition near $\alpha(0)$6 K strongly enhances the out-of-equilibrium Casimir–Polder force and especially its gradient. For $\alpha(0)$7Rb at $\alpha(0)$8m with $\alpha(0)$9 K, the force magnitudes are $$\overline{U}(z) = -\frac{3\hbar c \alpha(0)}{32\pi^2\epsilon_0 z^4},$$0 at $$\overline{U}(z) = -\frac{3\hbar c \alpha(0)}{32\pi^2\epsilon_0 z^4},$$1 K, $$\overline{U}(z) = -\frac{3\hbar c \alpha(0)}{32\pi^2\epsilon_0 z^4},$$2 at $$\overline{U}(z) = -\frac{3\hbar c \alpha(0)}{32\pi^2\epsilon_0 z^4},$$3 K, and $$\overline{U}(z) = -\frac{3\hbar c \alpha(0)}{32\pi^2\epsilon_0 z^4},$$4 at $$\overline{U}(z) = -\frac{3\hbar c \alpha(0)}{32\pi^2\epsilon_0 z^4},$$5 K. At $$\overline{U}(z) = -\frac{3\hbar c \alpha(0)}{32\pi^2\epsilon_0 z^4},$$6m, the ratio of force gradients for VO$$\overline{U}(z) = -\frac{3\hbar c \alpha(0)}{32\pi^2\epsilon_0 z^4},$$7/sapphire versus SiO$$\overline{U}(z) = -\frac{3\hbar c \alpha(0)}{32\pi^2\epsilon_0 z^4},$$8 is about $$\overline{U}(z) = -\frac{3\hbar c \alpha(0)}{32\pi^2\epsilon_0 z^4},$$9 at $$\overline{F}(z) = -\frac{3\hbar c \alpha(0)}{8\pi^2\epsilon_0 z^5}.$$0 K and about $$\overline{F}(z) = -\frac{3\hbar c \alpha(0)}{8\pi^2\epsilon_0 z^5}.$$1 at $$\overline{F}(z) = -\frac{3\hbar c \alpha(0)}{8\pi^2\epsilon_0 z^5}.$$2 K (Klimchitskaya et al., 2021).

A different nonequilibrium problem is the dynamical Casimir–Polder force for an atom initially prepared in a bare excited state near a perfectly conducting wall. In that case the force is explicitly time-dependent, oscillates between attraction and repulsion both in time and in space, and recovers the static excited-state result for times much larger than the self-dressing timescale but smaller than the atomic decay time (Armata et al., 2016).

Casimir–Polder interactions also fluctuate. For a small resonant dielectric sphere or atom scanned above a disordered metallic plate, the mean interaction is essentially independent of the electron relaxation rate $$\overline{F}(z) = -\frac{3\hbar c \alpha(0)}{8\pi^2\epsilon_0 z^5}.$$3, but the relative variance obeys

$$\overline{F}(z) = -\frac{3\hbar c \alpha(0)}{8\pi^2\epsilon_0 z^5}.$$4

so the fluctuating part is linearly proportional to dissipation and vanishes in the plasma limit $$\overline{F}(z) = -\frac{3\hbar c \alpha(0)}{8\pi^2\epsilon_0 z^5}.$$5 (Cherroret et al., 2017). At finite temperature and $$\overline{F}(z) = -\frac{3\hbar c \alpha(0)}{8\pi^2\epsilon_0 z^5}.$$6, the fluctuations are exponentially suppressed,

$$\overline{F}(z) = -\frac{3\hbar c \alpha(0)}{8\pi^2\epsilon_0 z^5}.$$7

whereas the mean potential decays only algebraically (Cherroret et al., 2017).

5. Anisotropy, chirality, and repulsive sectors

A common simplification is to regard the Casimir–Polder force as generically attractive. The literature in the data set shows that this is not generally correct. Sign changes can be induced by anisotropy, chirality, topology, and nonequilibrium driving.

For chiral matter, the Casimir–Polder potential contains a distinct chiral term $$\overline{F}(z) = -\frac{3\hbar c \alpha(0)}{8\pi^2\epsilon_0 z^5}.$$8 in addition to the electric and magnetic contributions. The chiral component exists only if the particle and the medium are both chiral, and it can be attractive or repulsive depending on the relative chirality of the molecule and the medium (Butcher et al., 2012). For an isotropic chiral particle,

$$\overline{F}(z) = -\frac{3\hbar c \alpha(0)}{8\pi^2\epsilon_0 z^5}.$$9

with the molecular chiral susceptibility $\lambda_T = \frac{\hbar c}{k_B T},$0 proportional to the optical rotatory strength $\lambda_T = \frac{\hbar c}{k_B T},$1 (Butcher et al., 2012). In a cavity of chiral metamaterials, ground-state molecules do not separate enantiomerically because the electric component dominates, but for initially excited molecules the electric component can be suppressed and the chiral component can select the enantiomer (Butcher et al., 2012).

Anisotropy alone can also generate repulsive components. For two atoms with orthogonal polarizabilities, the retarded interaction admits a $\lambda_T = \frac{\hbar c}{k_B T},$2-component of force that changes sign; for partially anisotropic atoms, repulsion requires sufficiently strong anisotropy, with the equal-anisotropy threshold $\lambda_T = \frac{\hbar c}{k_B T},$3 (Milton et al., 2011). For an anisotropic atom moving on a trajectory that does not intersect a conducting cylinder, repulsion near closest approach occurs when

$\lambda_T = \frac{\hbar c}{k_B T},$4

where $\lambda_T = \frac{\hbar c}{k_B T},$5 is cylinder radius and $\lambda_T = \frac{\hbar c}{k_B T},$6 is the closest approach distance; by contrast, no such repulsion occurs outside a conducting sphere (Milton et al., 2011).

Topological backgrounds can produce still different behavior. In the conical spacetime of an ideal straight cosmic string, the Casimir–Polder force on a microparticle with isotropic polarizability is always repulsive; at large separations it varies inversely as the fifth power of the distance, and in the non-retarded regime as the inverse fourth power (Saharian et al., 2011). For anisotropic polarizability, the sign depends on the eigenvalues and orientation of the polarizability tensor, and the interaction can change sign with separation; the orientation dependence also produces a torque on the particle (Saharian et al., 2011).

Not all anisotropic systems support repulsion. An anisotropic atom interacting with an anisotropic dielectric semispace does not exhibit Casimir–Polder repulsion (Milton et al., 2011).

6. Structured media, thin films, graphene, and analogues

Engineered media substantially broaden the phenomenology of Casimir–Polder forces. In the classical limit for a thin dielectric film of thickness $\lambda_T = \frac{\hbar c}{k_B T},$7, the free energy and force decrease more quickly with separation than for a thick plate: $\lambda_T = \frac{\hbar c}{k_B T},$8 whereas for a semispace $\lambda_T = \frac{\hbar c}{k_B T},$9 and $$\overline{U}(z) = -\frac{1}{16\pi}\frac{\hbar c \alpha(0)}{\epsilon_0 \lambda_T z^3},$$0 (Klimchitskaya et al., 2014). For metallic films, the leading electric contribution for purely electrically polarizable atoms is independent of whether the metal is described by the Drude or plasma model, but magnetic contributions depend on both film thickness and the chosen low-frequency model (Klimchitskaya et al., 2014).

Graphene provides a particularly distinctive large-distance regime. For an atom or nanoparticle above a graphene-coated fused-silica substrate at $$\overline{U}(z) = -\frac{1}{16\pi}\frac{\hbar c \alpha(0)}{\epsilon_0 \lambda_T z^3},$$1, the large-separation regime in which the zero Matsubara term gives at least $$\overline{U}(z) = -\frac{1}{16\pi}\frac{\hbar c \alpha(0)}{\epsilon_0 \lambda_T z^3},$$2 of the total Casimir–Polder force is reached at

$$\overline{U}(z) = -\frac{1}{16\pi}\frac{\hbar c \alpha(0)}{\epsilon_0 \lambda_T z^3},$$3

independently of the graphene energy gap $$\overline{U}(z) = -\frac{1}{16\pi}\frac{\hbar c \alpha(0)}{\epsilon_0 \lambda_T z^3},$$4 and chemical potential $$\overline{U}(z) = -\frac{1}{16\pi}\frac{\hbar c \alpha(0)}{\epsilon_0 \lambda_T z^3},$$5 (Klimchitskaya et al., 2023). In this regime the force is

$$\overline{U}(z) = -\frac{1}{16\pi}\frac{\hbar c \alpha(0)}{\epsilon_0 \lambda_T z^3},$$6

and asymptotically

$$\overline{U}(z) = -\frac{1}{16\pi}\frac{\hbar c \alpha(0)}{\epsilon_0 \lambda_T z^3},$$7

The classical limit itself can occur at much larger separations depending on $$\overline{U}(z) = -\frac{1}{16\pi}\frac{\hbar c \alpha(0)}{\epsilon_0 \lambda_T z^3},$$8 and $$\overline{U}(z) = -\frac{1}{16\pi}\frac{\hbar c \alpha(0)}{\epsilon_0 \lambda_T z^3},$$9; the asymptotic results agree better for larger chemical potential and smaller energy gap (Klimchitskaya et al., 2023).

For freestanding graphene, the classical regime at room temperature is reached at about $z^{-4}$0m separation, the magnetic contribution is suppressed relative to the electric one, and at separations above $z^{-4}$1m the Casimir–Polder interaction is of the same strength as with an ideal-metal plane (Klimchitskaya et al., 2014).

Finally, there are non-photonic analogues. In a one-dimensional tight-binding nanowire with two impurity atoms, virtual single-electron hopping induces an “electronic Casimir-Polder” interaction whose energy and force decay exponentially with impurity separation,

$z^{-4}$2

rather than with the power laws characteristic of QED Casimir–Polder forces (Yang et al., 2014). This is not the electromagnetic Casimir–Polder effect, but it isolates the general mechanism of fluctuation-mediated forces in a different mediating field.