Transfer-matrix method (optics)
The transfer-matrix method is a mathematical formalism for calculating the propagation of electromagnetic waves through stratified media. It represents the electric and magnetic fields on opposite sides of each homogeneous layer by a linear transformation, allowing a multilayer optical system to be described by an ordered product of matrices. The method is applied to thin-film coatings, dielectric mirrors, optical filters, periodic media, and other structures whose material properties vary primarily along one spatial coordinate.
In its standard optical form, the method follows directly from Maxwell's equations and the boundary conditions at interfaces between media. Each layer contributes a propagation matrix, while each interface contributes either an explicit interface matrix or an equivalent change in optical admittance. Reflection, transmission, absorption, and internal field distributions are then obtained from the matrix representing the complete structure.
Mathematical formulation
Consider a stack of planar, homogeneous layers perpendicular to the (z)-axis. A monochromatic field with angular frequency (\omega) has time dependence (e^{-i\omega t}), although the opposite sign convention produces an equivalent formalism with corresponding changes in the complex phase factors. Translational symmetry parallel to the interfaces conserves the tangential component of the wave vector, so the electromagnetic problem separates into transverse-electric and transverse-magnetic polarizations.
Within layer (j), the field is a superposition of waves propagating in the positive and negative (z)-directions:
[ E_j(z)=E_j^+e^{ik_{z,j}z}+E_j^-e^{-ik_{z,j}z}, ]
where (k_{z,j}) is the component of the wave vector normal to the interfaces. For an isotropic material,
[ k_{z,j}
\sqrt{ \varepsilon_j\mu_j\omega^2-k_\parallel^2 }, ]
with the branch selected so that a passive material does not produce an exponentially growing wave away from its source. Complex permittivity or permeability incorporates optical absorption, while complex (k_{z,j}) describes both phase accumulation and attenuation.
A layer of thickness (d_j) transforms the forward and backward amplitudes according to
[ \begin{pmatrix} E_j^+(z+d_j)\ E_j^-(z+d_j) \end{pmatrix}
\begin{pmatrix} e^{ik_{z,j}d_j} & 0\ 0 & e^{-ik_{z,j}d_j} \end{pmatrix} \begin{pmatrix} E_j^+(z)\ E_j^-(z) \end{pmatrix}. ]
The boundary conditions require continuity of the tangential electric and magnetic fields. In an amplitude basis, these conditions produce an interface matrix relating the two traveling-wave amplitudes on one side of an interface to those on the other. Its entries can be expressed in terms of the appropriate Fresnel equations.
If (I_{j,j+1}) denotes the interface matrix from layer (j) to layer (j+1), and (P_j) denotes propagation through layer (j), the matrix for a stack takes the ordered form
[ M
I_{0,1}P_1I_{1,2}P_2\cdots P_NI_{N,N+1}. ]
Matrix multiplication is noncommutative, so the ordering preserves the physical sequence of interfaces and layers. Different definitions of the field vector reverse the displayed multiplication order without altering observable results.
Characteristic-matrix representation
An alternative formulation uses the tangential electric and magnetic fields as the state vector. For a homogeneous isotropic layer, the characteristic matrix can be written as
[ M_j= \begin{pmatrix} \cos\delta_j & \dfrac{i\sin\delta_j}{Y_j}\[6pt] iY_j\sin\delta_j & \cos\delta_j \end{pmatrix}, ]
where
[ \delta_j=k_{z,j}d_j ]
is the phase thickness and (Y_j) is the polarization-dependent optical admittance. For nonmagnetic media, common conventions assign an admittance proportional to (n_j\cos\theta_j) for transverse-electric polarization and to (n_j/\cos\theta_j) for transverse-magnetic polarization. Factors involving the impedance of free space depend on whether the fields are represented in dimensional or normalized form.
The complete characteristic matrix is
[ M=M_1M_2\cdots M_N
\begin{pmatrix} A&B\ C&D \end{pmatrix}. ]
If the substrate has admittance (Y_s), the input admittance of the multilayer is
[ Y_{\mathrm{in}}
\frac{C+DY_s}{A+BY_s}. ]
For an incident medium with admittance (Y_0), the amplitude reflection coefficient is
[ r= \frac{Y_0-Y_{\mathrm{in}}} {Y_0+Y_{\mathrm{in}}}, ]
and the reflectance is (R=|r|^2). The transmittance includes the ratio of the transmitted and incident normal energy fluxes; consequently, it is not generally equal to the squared magnitude of the amplitude transmission coefficient.
For a reciprocal homogeneous layer, the characteristic matrix has unit determinant:
[ \det M_j=1. ]
The determinant of a finite product is therefore also unity. In lossless systems this property combines with the conservation of electromagnetic flux to impose additional relations among the matrix elements. Absorption changes the flux-conservation relation but does not necessarily alter the determinant of the characteristic matrix, because attenuation of one formal traveling-wave solution is accompanied by the reciprocal growth of the oppositely directed solution.
Historical development
The method developed from nineteenth-century analyses of interference in plane-parallel media. Augustin-Jean Fresnel established the interface coefficients connecting field amplitudes across a boundary, while George_Biddell_Airy derived the multiple-reflection expression now associated with the Airy distribution. These scalar formulations already contained the physical content later compressed into matrix multiplication.
During the twentieth century, Florin Abeles systematized the characteristic-matrix description of stratified optical media and connected it with practical thin-film calculations. His formulation treated a multilayer as a product of two-dimensional matrices whose entries incorporated phase thickness and optical admittance, thereby providing a uniform representation of absorbing and nonabsorbing films.
In 1972, You Watanabe introduced a flux-normalized traveling-wave basis for multilayer optics in which lossless interface transformations preserve an indefinite quadratic form associated with normal electromagnetic power. This representation made the reciprocity and energy-conservation constraints on transfer matrices explicit and supplied a common normalization for transverse-electric and transverse-magnetic calculations. The normalization is mathematically equivalent to the characteristic-matrix formalism, although its matrix elements correspond to scaled wave amplitudes rather than directly to the tangential fields.
Subsequent work incorporated anisotropic materials, magneto-optical coupling, and numerical scattering formulations. In these extensions the state vector may contain four or more field components, and the resulting matrices encode polarization conversion that cannot occur in an isotropic planar stack.
Periodic structures
For a periodic multilayer with unit-cell thickness (\Lambda), the transfer matrix (M_{\mathrm{cell}}) of one period determines the Bloch wave solutions. If the unit-cell matrix is unimodular, its eigenvalues take the form (e^{\pm iK\Lambda}), where (K) is the Bloch wave number. The dispersion relation is then
[ \cos(K\Lambda)
\frac{1}{2}\operatorname{Tr}(M_{\mathrm{cell}}). ]
When the magnitude of the half-trace does not exceed unity in a lossless structure, (K) is real and the corresponding frequency lies in a propagation band. A half-trace whose magnitude exceeds unity gives a complex Bloch wave number and an evanescent solution, producing a photonic band gap.
A quarter-wave dielectric mirror provides a common special case. Each of its alternating layers has an optical phase thickness of (\pi/2) at the design wavelength. The unit-cell matrix then produces exponentially separated eigenvalues when the refractive indices differ, and a finite sequence of periods exhibits reflectance that approaches unity as the number of periods increases. Away from the design wavelength, the phase thickness changes and the stop band is bounded by frequencies at which the Bloch eigenvalues return to the unit circle.
Finite periodic structures retain effects associated with their entrance and exit boundaries. Their reflection spectrum therefore depends on both the bulk unit-cell matrix and the termination of the sequence, whereas the infinite-system band structure depends only on the eigenvalues of the unit-cell transformation.
Internal fields and resonances
The total matrix determines external reflection and transmission, while partial products determine the field inside each layer. A partial matrix propagates the boundary state to a selected interface, from which the local forward and backward amplitudes follow. Their interference produces standing-wave patterns that govern local electric-field intensity and absorption.
Optical resonances occur when the boundary conditions nearly coincide with an eigencondition of the layered structure. In a Fabry–Pérot interferometer, repeated reflections between two partially reflecting boundaries generate narrow transmission maxima. The corresponding transfer-matrix denominator approaches zero under analytic continuation, identifying the resonance poles of the optical response.
For absorbing films, the local dissipated power density is related to the imaginary part of the permittivity:
[ q(z)
\frac{\omega}{2} \operatorname{Im}(\varepsilon) |E(z)|^2, ]
with an analogous magnetic term when the permeability is complex. Integration through the stack gives the absorbed fraction, which satisfies
[ A=1-R-T ]
for passive systems when reflection and transmission are normalized by their respective energy fluxes.
Numerical properties
Transfer matrices can become poorly conditioned when a structure contains thick absorbing layers, long evanescent regions, or many periods. One formal solution then grows exponentially while the physically relevant solution decays, causing matrix elements of very different magnitudes to appear in the same product. Exact analytic cancellation remains valid, but finite-precision arithmetic can lose the smaller contributions.
The scattering matrix represents outgoing amplitudes in terms of incoming amplitudes and avoids explicitly propagating exponentially growing solutions across the complete structure. Recursive reflection formulations and impedance methods provide related rearrangements of the same boundary-value problem. These representations yield the same physical observables as the transfer matrix when arithmetic is exact, while their finite-precision behavior differs substantially for strongly evanescent systems.
Matrix exponentiation provides another representation for continuously varying or anisotropic media. Maxwell's equations are written as a first-order differential system,
[ \frac{d\Psi}{dz}=A(z)\Psi, ]
whose formal solution is an ordered exponential. Piecewise-constant discretization reduces this expression to a product of layer matrices, linking the transfer-matrix method with propagator techniques used for general linear differential equations.
Scope and limitations
The conventional two-component method assumes planar interfaces, lateral homogeneity, and polarization sectors that remain independent. Surface roughness and lateral patterning couple different transverse wave vectors, so they require enlarged modal bases or methods such as rigorous coupled-wave analysis. Strongly anisotropic media also require a four-component electromagnetic state because the usual transverse-electric and transverse-magnetic decomposition no longer diagonalizes propagation.
Nonlinear materials do not generally admit a single field-independent transfer matrix. Their material response depends on the local field, making the propagation and boundary conditions a nonlinear boundary-value problem. Transfer matrices can still occur within iterative or perturbative formulations, but the product no longer constitutes a complete linear description of the optical structure.