Rigorous spatial point-splitting evaluation of the Casimir self-stress on a perfectly conducting spherical shell

A perfectly conducting spherical shell of radius $a$ in vacuum experiences a repulsive Casimir self-stress, with total Casimir energy $E = +0.09235\,\hbar c/(2a)$ — a result established independently by Boyer (1968), Davies (1972), Balian and Duplantier (1978), and by Milton, DeRaad, and Schwinger (MDS, 1978) via temporal point-splitting: the formally divergent vacuum expectation value of the electromagnetic stress tensor is regularized by separating the two field points in imaginary time, which after Wick rotation supplies an exponential ultraviolet cutoff in frequency. MDS computed both the total energy and the radial surface stress, found them consistent, and remarked that a spatial point-splitting — separating the two field points in space rather than time — would be a desirable independent check but is far more delicate, because spatial coincidence limits at a smooth conducting surface generate non-integrable divergences in powers and logarithms of the separation. The open question is whether such a purely spatial cutoff can be made rigorous and unambiguous. Recent calculations (2017–2024) reproduce the conducting-shell value with spatial or mixed splittings, but the two leading groups dispute whether spatial splitting suffices to regulate the divergences at all, the transverse-stress integrals diverge without temporal damping, and the finite part of the logarithmically divergent term depends on an arbitrary length-scale choice. A complete, rigorous spatial point-splitting derivation of the shell's energy and surface stress, their agreement within one spatial prescription, and a unique finite part remains to be given.

Background

The Casimir effect is the force that quantum vacuum fluctuations exert on material boundaries: the allowed modes of the electromagnetic field are altered by the presence of conductors, and the change in the zero-point energy $\frac{1}{2}\sum\hbar\omega$ of the modes produces, for example, the familiar attractive force between two parallel conducting plates. For a single closed surface, such as a perfectly conducting spherical shell of radius $a$, the analogous quantity is the self-stress: the net pressure the vacuum exerts on the shell itself, or equivalently the total Casimir self-energy $E(a)$ of the configuration, with the total self-force on the shell following from $F = -\,\partial E/\partial a$ and the average normal pressure on the surface being $F/(4\pi a^{2})$.

For the conducting sphere the self-stress was first computed by T. H. Boyer in 1968 and came as a surprise: it is repulsive, so the shell tends to expand, with total Casimir energy

$E \;=\; +\,\frac{0.09235\,\hbar c}{2a},$

where $\hbar$ is Planck's constant and $c$ the speed of light — opposite in sign to the attractive parallel-plate case. This value was reconfirmed by B. Davies (1972), by R. Balian and B. Duplantier (1978) using a multiple-scattering expansion (an iterated-reflection series for the field Green function), and by K. A. Milton, L. L. DeRaad, and J. Schwinger (1978; "MDS") using Green-function techniques; it is not in doubt.

The subtlety is regularization. The vacuum expectation value of the electromagnetic stress tensor, whose radial component gives the surface force, is formally a product of fields at the same spacetime point and is divergent. The standard cure is point-splitting (point separation): the two field factors are evaluated at distinct points $x$ and $x'$, the expectation value is expressed through the Green function — the two-point correlator of the field — and the coincidence limit $x'\to x$ is taken only after the singular part has been dealt with. In their calculation, MDS split the points in time: after a Wick rotation (analytic continuation of the frequency integrals to imaginary frequencies $\zeta$), a small imaginary-time separation $\tau$ between the two points supplies a damping factor $e^{-\zeta\tau}$ that acts as an ultraviolet cutoff on the frequency integration, rendering all intermediate expressions finite and the coincidence limit unproblematic. They computed both the total Casimir energy (inside and outside the sphere) and the radial component of the stress tensor on the surface, and found that the two methods agree.

MDS remarked that a spatial point-splitting — separating the two points in space rather than time, so that the cutoff is a short distance rather than a short time — would be a more direct independent check, but is considerably more delicate. The reason for the delicacy was made precise by D. Deutsch and P. Candelas (1979): near a smooth conducting boundary, the local vacuum stress contains non-integrable surface divergences — terms growing as inverse powers of the distance $\Delta$ from the surface, plus a term logarithmic in $\Delta$ — which can only be absorbed into local surface properties such as a surface tension. A temporal cutoff suppresses these before the limit is taken; a spatial cutoff meets them head-on, as explicit divergences in powers and logarithms of $\Delta$ that must be separated from the finite, physical part by a stated principle.

The question has retained live weight. Since 2017, the group of K. A. Milton (with P. Parashar, K. V. Shajesh, I. Brevik, and G. Kennedy) and the group of U. Leonhardt (with Y. Avni) have debated exactly this point for spherical geometry. The Milton school computes with temporal point-splitting, supplemented at most by a transverse (angular) split, and argues that spatial splitting alone cannot regulate all divergences; the Leonhardt school computes the surface stress by splitting the field points spatially across the boundary and argues that the divergent terms are cutoff-dependent artifacts attributable to surface physics. Both recover the established value in the conducting limit, but the rigor and the uniqueness of the spatial procedure remain contested: the finite part of the logarithmically divergent term depends on the choice of length scale inside the logarithm, a choice neither school claims is water-tight. What no one has provided in the four decades since MDS is a complete and rigorous purely spatial point-splitting evaluation of the conducting spherical shell — one that derives the finite energy and surface stress, demonstrates the energy/stress-tensor agreement within the same spatial regularization, and states unambiguously how every regulator-dependent term is handled so that the finite part is unique.

Problem Statement

The problem is to give a complete and rigorous evaluation of the Casimir self-energy and self-stress of a perfectly conducting spherical shell of radius $a$ in which the ultraviolet regularization is implemented by spatial point-splitting: the two field points of the (formally coincident) stress-tensor expectation value are separated in space at finite separation, the regularized quantities are computed, and the coincidence limit is taken only at the end. Such an evaluation must satisfy all four of the following requirements, which are jointly necessary to resolve the problem:

1. Define the spatial cutoff precisely enough that all intermediate quantities are finite and the limiting procedure is unambiguous, with the ultraviolet control supplied by the spatial separation itself and not by any hidden temporal splitting or frequency damping. The distinction concerns which separation supplies the ultraviolet regulator taken to zero last: a temporal separation (equivalently, an exponential frequency-damping factor such as $e^{-\zeta\tau}$ in the frequency integral) versus a spatial separation of the field points (radial and/or angular); the use of Wick-rotated imaginary-frequency Green functions is not by itself temporal splitting.

2. Reproduce from the spatially regularized expressions the established finite results — the total Casimir energy $E = +0.09235\,\hbar c/(2a)$ and the corresponding repulsive total self-force $F = -\,\partial E/\partial a = +0.09235\,\hbar c/(2a^{2})$, i.e., the force constant $0.04617$–$0.04618$ in units of $\hbar c/a^{2}$ quoted in the stress-tensor comparisons of the literature.

3. Demonstrate, within the same spatial regularization, the agreement between the energy method (computing $E$ from the Green function and differentiating with respect to $a$) and the stress-tensor method (computing the radial stress on the surface directly) that the temporal-splitting calculation established.

4. State unambiguously how every regulator-dependent term generated by the spatial cutoff — the non-integrable surface divergences in powers of the separation $\Delta$ and its logarithm, including curvature-squared contributions — is separated, absorbed into surface quantities, or proven not to contribute, so that the finite part is unique (either independent of the choice within the class of spatial cutoffs, or fixed by a stated principle that is justified independently of reproducing the established value $E = +0.09235\,\hbar c/(2a)$).

References

1. Kimball A. Milton, Lester L. DeRaad, Jr., and Julian Schwinger, "Casimir self-stress on a perfectly conducting spherical shell", Annals of Physics 115, 388-403 (1978). DOI: 10.1016/0003-4916(78)90161-6, https://doi.org/10.1016/0003-4916(78)90161-6
2. Timothy H. Boyer, "Quantum Electromagnetic Zero-Point Energy of a Conducting Spherical Shell and the Casimir Model for a Charged Particle", Physical Review 174, 1764-1776 (1968). DOI: 10.1103/PhysRev.174.1764, https://doi.org/10.1103/PhysRev.174.1764
3. B. Davies, "Quantum Electromagnetic Zero-Point Energy of a Conducting Spherical Shell", Journal of Mathematical Physics 13, 1324-1329 (1972). DOI: 10.1063/1.1666141, https://doi.org/10.1063/1.1666141
4. Roger Balian and Bertrand Duplantier, "Electromagnetic waves near perfect conductors. II. Casimir effect", Annals of Physics 112, 165-208 (1978). DOI: 10.1016/0003-4916(78)90083-0, https://doi.org/10.1016/0003-4916(78)90083-0
5. D. Deutsch and P. Candelas, "Boundary effects in quantum field theory", Physical Review D 20, 3063-3080 (1979). DOI: 10.1103/PhysRevD.20.3063, https://doi.org/10.1103/PhysRevD.20.3063
6. Kimball A. Milton, "The Casimir effect: recent controversies and progress", Journal of Physics A: Mathematical and General 37, R209-R277 (2004). DOI: 10.1088/0305-4470/37/38/R01, arXiv:hep-th/0406024, https://doi.org/10.1088/0305-4470/37/38/R01
7. Prachi Parashar, Kimball A. Milton, K. V. Shajesh, and Iver Brevik, "Electromagnetic δ-function sphere", Physical Review D 96, 085010 (2017). DOI: 10.1103/PhysRevD.96.085010, arXiv:1708.01222, https://arxiv.org/abs/1708.01222
8. Yael Avni and Ulf Leonhardt, "Casimir self-stress in a dielectric sphere", Annals of Physics 395, 326-340 (2018). DOI: 10.1016/j.aop.2018.06.002, arXiv:1710.09174, https://arxiv.org/abs/1710.09174
9. Kimball A. Milton, Prachi Parashar, Iver Brevik, and Gerard Kennedy, "Self-stress on a dielectric ball and Casimir–Polder forces", Annals of Physics 412, 168008 (2020). DOI: 10.1016/j.aop.2019.168008, arXiv:1909.05721, https://arxiv.org/abs/1909.05721
10. Ulf Leonhardt, "Comment on 'Self-Stress on a Dielectric Ball and Casimir-Polder Forces'", arXiv:1910.02604 (2019). https://arxiv.org/abs/1910.02604
11. Iver Brevik, "Possible connection between time-splitting parameter and surface tension in Casimir-type problems", International Journal of Modern Physics A 37, 2241010 (2022). DOI: 10.1142/S0217751X2241010X, https://doi.org/10.1142/S0217751X2241010X; preprint arXiv:2205.04247, https://arxiv.org/abs/2205.04247, posted under the differing preprint title "Connection between time-splitting parameter and surface tension in Casimir problems. Related phenomena in relativistic collapse of a singular shell"
12. Yang Li, "Casimir stresses of the dielectric ball: Inhomogeneity and divergences", Physical Review D 110, 116031 (2024). DOI: 10.1103/PhysRevD.110.116031, arXiv:2411.07911, https://arxiv.org/abs/2411.07911
13. Theodorus Maria Nieuwenhuizen, "How the vacuum rescues the Lorentz electron and imposes its Newton and geodesic motion, and the equivalence principle", The European Physical Journal Special Topics 235, 3101-3110 (published online 7 July 2025; volume dated 2026). DOI: 10.1140/epjs/s11734-025-01769-8, https://doi.org/10.1140/epjs/s11734-025-01769-8