WorksPhysics and Astronomy
The electromagnetic stress tensor in cubic crystals and amorphous solids
Abstract
The electromagnetic vacuum stress tensor in a system consisting of classical point-like atoms on a cubic lattice is derived directly from the electric field and, alternatively, using the known more phenomenological formula involving the (usually unknown) derivative of the dielectric tensor. The results agree. The stress tensor contains a nontrivial constant and therefore cannot be obtained from macroscopic electrodynamics alone. A system consisting of a mosaic of randomly oriented grains of a cubic crystal is used as a model of an amorphous solid. The electromagnetic stress tensor in this system is deduced by averaging the corresponding quantity in a cubic crystal over all orientations. Two components remain: the component in the direction of the electric field and the component perpendicular to it. The transverse component agrees with the analogous quantity for liquids derived by Peierls in an entirely different way.
Introduction
Macroscopic electrodynamics is an astonishingly successful re-formulation of the microscopic Maxwell equations, applicable when the characteristic length scales of the system in question are much larger than the atomic scale. It correctly describes quasi-static electromagnetic phenomena in condensed matter, as well as the reflection and refraction of light by dielectric media for different polarizations and the intensities of reflected and transmitted light in the isotropic and anisotropic case [1, 2]. Material properties enter solely through constitutive coefficients such as susceptibilities and electrical conductivity.
The fact that the theory, which is essentially linear in the fields, also correctly reproduces the bilinear energy density, is an additional bonus. The underlying physical principle is rarely stated. Atomic polarization currents are additive and are equivalent to a time-dependent macroscopic surface charge at the boundary of a medium, for instance on the plates of a capacitor. Similarly, the microscopic circular currents in the atoms of a magnetizable cylinder cancel in the interior but are equivalent to a macroscopic current around the cylinder. The macroscopic quantities then determine the energy.
Difficulties arise, however, for another bilinear quantity, the electromagnetic stress tensor, or momentum flux. This limitation is the origin of longstanding debates. Two more recent reviews are [3, 4]. Rather than addressing the problem in a purely formal manner, we here calculate the stress tensor exactly for microscopic models. Such models have been used for various purposes [5, 6, 7, 8]. The microscopic stress tensor, however, has not been previously evaluated.
Electromagnetic stress tensor in a lattice of point dipoles
The central tool used in this work is the Fourier series of the electric field in an orthorhombic lattice of point dipoles. We derive this sum from a model of small, but finite, classical polarizable atoms.
The result reproduces the electric field of the discrete-dipole approximation [7, 5] if the atoms are small, and also coincides with the Ewald-Kornfeld sum [9, 8] if the convergence parameter of this formalism is chosen in such a way that the reciprocal lattice sum dominates.
A direct result is a formula expressing the components of the dielectric tensor in terms of the polarizability of an atom, which reduces to the Clausius-Mossotti relation in the case of cubic crystals, but also yields the dependence of on deformations of the crystal. This dependence is needed to compute the stress tensor from the known phenomenological formula.
The electric field
The atoms are assumed to consist of two interpenetrating spherically symmetric charge distributions with a charge . We assume that the radial scale of satisfies , where are the lattice constants of the orthorhombic crystal. Three simple charge distributions and their Fourier transform are
We assume that is large. The displacements of the charges then are small, and the charge density of the atom at the origin in a local field is
where
is the polarization of an atom, and the unit cell volume. The symbol ‘#’ always denotes exact, microscopic, quantities. The polarizability of an atom is . The charge density of the crystal is periodic and can be written as a Fourier series,
We now assume that and the average (macroscopic) field point in direction. The Fourier series of the electric field then is
It suffices to verify that the Maxwell equations and are satisfied. The usual definition of the dielectric constant in terms of polarization density and average field leads to
This equation can be used to replace the microscopic quantity in the expression (4) for the electric field with a macroscopic quantity. The computation of the stress tensor then is a purely formal task.
However, to relate the stress tensor to the conventional expression we also need the dielectric constant, which can be derived by writing the field at the origin in a different way. The alternative expression is the sum of and the field generated by the charge distribution (2) of the atom at the origin at ,
Equating now and the expression (4) for and eliminating with (5) leads to
Equations (7) and (8) determine the dielectric constant of orthorhombic crystals. Atoms with a Gaussian charge distribution yield a rapidly convergent sum. In fact, with this choice Eq. (8) agrees with the reciprocal lattice component of the corresponding Ewald-Kornfeld sum for point dipoles, the radius plays the role of the convergence parameter. The lattice component is of the order and negligible for . This confirms that the limit is equivalent to point dipoles. Consequently, full Ewald summation is often unnecessary with modern computing power; the reciprocal-lattice sum converges quickly, and the real-lattice contribution remains negligible for sufficiently small .
For cubic crystals and for the sum in (8) can be approximated by an integral if the missing term is added. This gives and . This value together with (7) reproduces the usual Clausius-Mossotti equation for in a cubic crystal.
Stress tensor perpendicular to the electric field
We now use the microscopic electric field (4) to compute the electromagnetic vacuum stress tensor
Our sign convention for is that of [10]: a positive diagonal component compresses the medium in the given direction. We begin with the interstitial vacuum plane defined by , parallel to the field . The computation in the plane is analogous.
The granular part (4) of can be written as
The derivative in allows to restrict the sum in to instead of merely . The sum over in can now be performed using formula (35) from Appendix A.1. The result is
with . For this leads to
The field component vanishes. The stress tensor component perpendicular the field is .
We are only interested in the average over the area in . The average of a product of two Fourier series like Eq. (13) is the sum of the product of the Fourier components, see Appendix A.2. The result is
where is a constant. We have omitted the area in the final expression, since it follows from the indices, and because is the same in any vacuum plane parallel to .
Stress tensor in direction of the electric field
To get the electric field in the interstitial planes perpendicular to the field we now perform the sum over in Eq. (11). We split the sum into contributions with and ,
The condition in the part can be ignored, as we are only interested in contributions that depend on . The sums over in Eq. (16) were performed using the formulas (36) and (35) in Appendix § 7.§ 1. From at it follows that , but there remains
The constant in Eq. (17) implies that the actual non-granular field in is instead of . Averaging over an area yields (see Appendix § 7.§ 2)
Expressions (15) and (18) for cubic crystals are the main result of this section. The constant indicates that the electromagnetic stress tensor, in general, cannot be derived from macroscopic electrodynamics alone.
Electromagnetic stress tensor from phenomenological formula
In textbooks one finds the expression [2, 10]
for the electromagnetic stress tensor in non-magnetizable liquids. The symbols and denote the macroscopic (average) fields, is the relative dielectric constant of the liquid, and its mass density. The derivative is to be taken at constant temperature. This expression follows by equating the change of the electromagnetic energy in a volume to the work done when the volume is changed [10]. The practical value of formula (19) is limited, however, because the derivative is usually unknown. In general it also seems difficult to identify (19) with some average of the exact microscopic . An average over a plane in the liquid would contain contributions from electromagnetic forces within the atoms, which are compensated by some other forces, but differ from zero also for .
The phenomenological derivation of the stress tensor (19) can straightforwardly be generalized to isotropic solids and to orthorhombic crystals if the electric field points in the direction of a crystal axis, see Appendix § 7.§ 3. We here again use the axis. We only consider amorphous media or cubic crystals, originally of volume and with . An uniaxially compressed liquid remains isotropic and thus . This is not true for solids, where a uniaxial deformation changes the aspect ratio of the unit cells. In general thus . The energy balance leads to
We now compare the phenomenological expressions (20) and (21) with the exact results derived above.
Cubic crystal
To actually get a value from the expressions (20) and (21) we also need the variation of of a cubic crystal under deformations with . This variation can be accounted for in the sum (8) by replacing with . The numerical evaluation of (8) for small deformations in the direction or leads to
where and is the constant from Eq. (15).
We first consider a deformation in the direction perpendicular to the electric field. The derivative follows as
The value is negative for . Inserting the derivative (23) into the formula (21) reproduces the exact average vacuum stress tensor component (15).
The derivative for a deformation in the direction of the electric field is
and with this value the formula (20) also reproduces the exact vacuum stress tensor component (18) in the direction of the electric field.
A two-dimensional model
Essential aspects of the problem can be illustrated with an effectively two-dimensional model: a medium consisting of thin polarizable rods arranged at the lattice points of a square lattice in the --plane (Figure 1). The electric field and all rods are oriented along the -direction. No polarization charge arises in the medium. The electric field thus is constant, and macroscopic and microscopic electric fields are identical. Only the component of the dielectric tensor plays a role. If the system has length in -direction, then , where is a constant. Eq. (21) then again reproduces the (here trivial) vacuum stress tensor

Figure 1. A dielectric medium consisting regularly arranged polarizable rods, all oriented in -direction. The electric field in the medium also points in -direction.
in the planes parallel to the rods. In the long wavelength limit, the model also permits exact solutions for signal propagation in directions perpendicular to the -axis, for instance along . Such signals only sense the dielectric constant, not the individual rods. The dispersion relation is . Adding the contribution from the magnetic field according to (9) yields (all time-average factors are suppressed)
The boundaries of a signal with a finite width in -direction are pulled inward because of . Strictly speaking, a finite width signal no longer represents purely two-dimensional electrodynamics; electric and magnetic field cannot terminate abruptly, and there is a transition region with oblique components of the fields. However, the difference of the stress tensor inside and outside the signal remains unchanged. If the rods form a two-dimensional liquid, this inward force is compensated by a pressure. The effective longitudinal stress tensor of a stationary signal thus is
The corresponding momentum density is , which gives
This is what is found in experiments in real liquids [11, 12]. Although the two-dimensional model is not directly applicable to real liquids, it serves as an example where the underlying physics can be understood starting from a microscopic framework.
Arbitrarily oriented electric field
We now generalize the computation of the stress tensor in a cubic crystal in Section 2 to a situation with an arbitrarily oriented field . The electric field in the general case is the linear superposition of the field in the situation with along considered above, with the field from the situation with along or . The field as above comprises the non-constant Fourier components of .
It suffices to consider the plane given by , the field in the planes and follows by a cyclic permutation of the coordinates. To simplify the notation we write . The field component perpendicular to considered above in (17) yields the first line of
The field components and are parallel to , exactly as is parallel to in Eq. (13), and analogously contribute diagonally as and , and in the form . This gives the second and third line in Eq. (28).
The requisite averages according to Appendix A.2, are again given by the sum of the products of the Fourier components. One must only keep in mind that the constant in combines with to . The result is
The constant is defined in Eq. (14), the constant is new. Using Eq. (9) one can now compute . Adding the contribution from the constant fields yields
The constant has dropped out. The expressions (30) and its cyclic permutations of the coordinates can be combined to
This is our final result for a cubic crystal. Omitting the areas is justified since is independent of the specific areas in a translationally invariant system. Areas containing dipoles can be slightly deformed to avoid possible complications inside the dipoles. The vacuum stress tensor (31) evaluated over interstitial surfaces enclosing some atoms determines the electromagnetic force acting on them.
The stress tensor (31) contains the dielectric tensor and the constant from (14). For it reproduces Eq. (15) and Eq. (18). It strongly depends on the direction of the field. One can now, for instance, compute for an electric field along a space diagonal , and then project onto to get the electromagnetic stress in the direction of the field.
Amorphous solids
A simple model of an amorphous solid is a mosaic of randomly oriented grains of a cubic crystal. The dielectric tensor of the mosaic is diagonal and agrees with that of the cubic crystal. The electromagnetic stress tensor within the grains, however, is obtained by averaging the stress tensor (31) over the orientations of the grains. The electromagnetic stress tensor in the grain boundaries is complicated, but the field is finite, and the contributions to the average stress tensor are negligibly when the grains are large.
One must distinguish between the stress tensor component in the direction of the electric field and the component perpendicular to it. Projecting Eq. (31) onto the direction of and averaging over all directions of yields an expression for the longitudinal electromagnetic stress tensor in an amorphous solid,
The angular average is denoted by . To obtain we have inserted Eq. (31) and written . The result then follows from
see Appendix A.4. The expression (32) differs from the longitudinal stress tensor (18) along an axis of a cubic crystal at order .
To obtain the transverse electromagnetic stress tensor, one must project Eq. (31) onto a direction perpendicular to a given , average over , and then over the directions of ,
The average over in can be written as an average over an angle by expressing as a sum of two unit vectors orthogonal to ,
This gives . The last equal sign simply states that the squares of the components in each row of the orthogonal matrix sum to one. The last line in (34) then follows from
see also Eq. (33). The stress tensor components (32) and (34) in an amorphous solid no longer depend on the constant . This, however, does not mean that they can be derived from macroscopic electrodynamics alone. The transverse component (34), notably, agrees with Peierls’s result for amorphous solids and liquids [13].
Conclusion
Idealized models with point dipoles on a lattice are usually quantitatively useful for real crystals as well. We have computed the electric field in an orthorhombic lattice of point dipoles, first with the macroscopic electric field oriented along a crystal axis. This has allowed us to derive the electromagnetic stress tensor in a cubic crystal in two ways: directly from the microscopic field, and alternatively, using the conventional phenomenological formula, involving the macroscopic electric field and a derivative of the dielectric tensor. The results agree. The fact that the stress tensor contains a constant depending on the microscopic details, rules out its derivation from macroscopic electrodynamics alone [14, 15], at least for cubic crystals.
The generalization to an arbitrarily oriented macroscopic electric field in a cubic crystal is essentially a linear superposition of the more symmetric case. The generic electromagnetic stress tensor is still defined by its value in the interstitial vacuum planes.
Modeling an amorphous solid as a mosaic of arbitrarily oriented grains of a cubic crystal has allowed to compute the longitudinal and transverse components of in an amorphous solid as an average of over crystal orientations. The constant drops out.
The propagation of light in a liquid represents a physical system where the electromagnetic stress tensor can be directly measured, making it the subject of theoretical investigations. Although dynamic phenomena are generally beyond the scope of this work—apart from the two-dimensional model in Section 3.2—several concluding remarks are in order.
The experiments [11, 12] utilized a signal of finite width, and the measured signal momentum is given by the ratio (27).
In his analysis, Peierls [13] started from a partially microscopic model and also arrived at the stress tensor component (34) in propagation. The corresponding usual magnetic contribution directly follows from the macroscopic Maxwell equations.
What then would be needed to get the ratio (27) is a mechanism eliminating the unusual factor of from the electric contribution (34). The two-dimensional model discussed in Section 3.2 indicates at least that the finite width of the signal—and the stress tensor components in the directions of and at the lateral borders of the signal—play a role.
Appendix
Fourier sums
To compute the electric field in interstitial planes, the following equation was used:
This equation is valid in the interval for any , and must be extended periodically such that . To verify the Fourier sum, it suffices to check that both expressions satisfy the differential equation
The r.h.s. satisfies the equation for since it is a linear combination of . The inhomogeneous differential equation is satisfied because . This discontinuous derivative at generates a -function in the second derivative. One also needs the Fourier series
valid in the interval . Instead of deriving as a limiting case of one could alternatively verify that satisfies the differential equation .
Average over a surface of a unit cell
The average stress tensor (9) in the plane of a cubic crystal contains averages of the type
The components of the electric field in the area are given as a Fourier series
as in (13). Since is real it follows . Inserting the sum (37) yields
The field in only contributes as , where is the component of .
Energy balance in a solid
The phenomenological expression (19) for the electromagnetic stress tensor of a liquid directly follows from energy conservation [10]. Its generalization (20, 21) to an orthorhombic crystal or an amorphous solid is straightforward. Consider a planar capacitor with plates of area perpendicular to the -axis, separated by a distance , and an electric field . At constant charge , the total energy is
Fringe effects at the borders are negligible if the area is large. Under a strictly uniaxial deformation, the variation of the energy with distance yields the stress tensor component
It is clear from the derivation that the stress tensor does not include the electromagnetic stress within the atoms, which is compensated by some other force. It could be computed, however, from the microscopic field in an interstitial vacuum plane, if such a plane exists. The expression (21) for the stress tensor in the direction perpendicular to the field is obtained by varying of .
Average over orientations
In § 5 the averages (33) of unit vector components over all orientations were used to calculate the stress tensor of an amorphous solid from that of a cubic crystal. These averages can be derived using spherical coordinates and the unit vector component ,
References
- [1]A. Lipson, S. G. Lipson, and H. Lipson. Optical physics. Cambridge University Press, Cambridge, 2010.
- [2]J. D. Jackson. Classical electrodynamics. Wiley, New York, 1999.
- [3]P. W. Milonni and R. W. Boyd. Momentum of light in a dielectric medium. Adv. Opt. Photonics, 2:519 – 553, 2010.DOI
- [4]B. A. Kemp. Resolution of the Abraham-Minkowski debate: Implications for the electromagnetic wave theory of light in matter. J. Appl. Phys., 109:111101, 2011.DOI
- [5]B. T. Draine and J. Goodman. Beyond Clausius-Mossotti: wave propagation on a polarizable point lattice and the discrete dipole approximation. Astrophys. J., 405(2):685–697, 1993.DOI
- [6]P. de Vries, D. V. van Coevorden, and A. Lagendijk. Point scatterers for classical waves. Rev. Mod. Phys., 70(2):447–466, 1998.DOI
- [7]B. T. Draine. The discrete-dipole approximation and its application to interstellar graphite grains. Astrophys. J., 333:848–872, 1988.DOI
- [8]C. K. Lo, J. T. K. Wan, and K. W. Yu. Effects of geometric anisotropy on local field distribution: Ewald-Kornfeld formulation. J. Phys.: Condens. Matter, 13:1315–1321, 2001.DOI
- [9]H. Kornfeld. Die Berechnung elektrostatischer Potentiale und der Energie von Dipol- und Quadrupolgittern. Z. Phys., 22:27–43, 1924.DOI
- [10]L. D. Landau and E. M. Lifshitz. The classical theory of fields. Pergamon press, New York, 1975.
- [11]R. V. Jones and J. C. S. Richards. The pressure of radiation in a refracting medium. Proc. R. Soc. Lond. A, 221(1147):480 – 498, 1954.DOI
- [12]R. V. Jones and B. Leslie. The measurement of optical radiation pressure in dispersive media. Proc. R. Soc. Lond. A, 360(1702):347 – 363, 1978.DOI
- [13]R. Peierls. The momentum of light in a refracting medium. Proc. R . Soc. Lond. A, 347(1651):475–491, 1976.
- [14]A. Shevchenko and B. J. Hoenders. Microscopic derivation of electromagnetic force density in magnetic dielectric media. New J. Phys., 12:053020, 2010.DOI
- [15]W. Frias and A. I. Smolyakov. Electromagnetic forces and internal stresses in dielectric media. Phys. Rev. E, 85(046606), 2012.pubmed.ncbi.nlm.nih.gov/22680597