A Theoretical Treatment of the Reflexion of Light by Multilayer Structures

Interference between rays of light reflected at successive interfaces in a laminated structure has often been suggested as one of the methods by which colours of animals might be produced (e.g. Biedermann, 1914, p. 1892; Rayleigh, 1919; Onslow, 1921, p. 13; Fox & Vevers, 1960). An even more striking phenomenon is the very high reflectivity, over a substantial part of the visible spectrum, found, for example, in fish scales (Denton & Nicol, 1966), in the argentea of the eye of the scallop, Pecten (Dakin, 1910; Land, 1965) and in tapeta of the eyes of cartilaginous fishes (Denton & Nicol, 1964). In several of these cases it has recently been shown (Land (1966), for the eye of Pecten; Denton & Land (1967), for fish scales) that the reflecting structure consists of a stack of flat transparent crystals, each having an optical thickness close to a quarter of a wavelength and separated from its neighbours by layers of cytoplasm of equal optical thickness. Constructive interference therefore occurs between light reflected at successive interfaces, and a high reflectivity is obtained in the same way as in artificial ‘dielectric reflectors’ made by vacuum deposition of layers of transparent materials of alternately high and low refractive index. Other highly reflecting structures, such as the tapeta behind the retinae of some mammals, also have regularly repeating structural elements at spacings of the right order of magnitude for the same kind of interference to occur, but have not yet been worked out from the optical point of view (e.g. Pedler, 1963; Dartnall et al. 1965).

It might have been thought that the commercial development of multilayer dielectric reflectors and interference filters would have led to the publication of convenient theoretical expressions for dealing with biological systems of this kind. Discussions with Dr M. F. Land during his work on the argentea of the eye of Pecten showed however that this was apparently not the case, and the author derived a number of simple formulae, relevant to reflectors of this kind, which are not given in the accounts of multilayer theory by, for instance, Heavens (1955, 1960), Vašiček (1960), Born & Wolf (1964) or Baumeister (1965). These formulae can presumably be obtained by the powerful matrix method of Abelès (1950), but this method is of little use to those who are not familiar with matrix algebra; another advantage of the method used in the present paper is that it gives greater insight into the processes which give rise to the optical properties of the complete stack of plates than does the method of Abelès. For these reasons it seemed likely that the method, and the formulae obtained by means of it, would be useful to biologists investigating other multilayer reflecting structures.

Most of the steps in this method are the same as were used by Rayleigh (1917); he in turn was adapting for thin layers the method by which Stokes (1862) had treated the case in which the plates are thick enough for interference phenomena to be dis-regarded. These papers seem to have been to a large extent forgotten; of the authors quoted in the preceding paragraph, only Abelès (1950) refers to them.

Rayleigh did not, however, derive all the particular formulae which will be given in the present paper. The author felt it would also be useful to present the whole method afresh in a more elementary style than Rayleigh’s treatment.

In § 2, the problem is stated in its simplest form (optical thickness of spaces equal to that of plates; normal incidence; no reflexion from different materials above or below the stack itself), and symbols are defined. In § 3, the steps of the solution of this problem are outlined and the most important equations are given without proof; these equations and others are derived in § 4. In § 5 the results are extended to some more general cases.

Section 4.3 contains a discussion of the variation of phase and amplitude through the thickness of the stack.

It is assumed throughout that absorption of light is negligible. The results could no doubt be adapted to the case of finite absorption by using complex values for the phase retardations due to the plates and the spaces between them. The method is restricted to stacks consisting of a regularly repeated sequence of layers.

Figure 1 illustrates the situation that is dealt with in §§ 3 and 4. Light is incident normally from above on a stack of transparent plates, p in number, uniformly spaced in an infinite medium of different refractive index. The problem is to find the fraction of the incident intensity that is reflected.

The a’s, denoting the amplitudes (and phases) of the light at different levels within the stack are defined at the broken lines in Fig. 1, i.e. at the centres of the spaces between the plates and at corresponding levels above the top plate and below the bottom plate. This convention is chosen because it makes the structure between any two of the broken lines symmetrical, and therefore simplifies the equations.

Only the symbols which are used throughout the paper are defined here; a number of other symbols are defined where they arise in §§ 4.1, 5.2 and 5.3.

See Fig. 1. Defined as amplitudes of electric vector, as a complex quantity so as to indicate phase as well as absolute amplitude.

⁠: downward wave, at centre of space below jth plate.

⁠: upward wave, at centre of space below jth plate.

a_i, a_r, a_t : defined in § 4.1 and Fig. 2.

n_b, d_b : each plate in the stack.

n_a, d_a: each space between adjacent plates.

n₀, d₀; n_z, d_p: defined in § 5.3 and Fig. 4.

Expressions are for normal incidence; oblique incidence is dealt with in § 5.2.

ϕ_b : in each plate of the stack, = 2π d_bn_b/λ.

ϕ_a : in each space between plates, = 2π d_an_a/λ.

ϕ: equal to ϕ_a or ϕ_b when ϕ_a= ϕ_b.

δ: defined in § 4·1.

ϕ₀, ϕ_z,χ :defined in § 5.3.

r₁, t₁, r_2,t₂, r₃, t₃, R, T are defined for light incident from above; the same symbols with primes are for light incident from below on the same interfaces or structures. In general, complex, but r, t, r_0,r_z are real.

r, t: at n_b/n_a interface.

r₀: at n₀/n_a interface (§ 5·3).

r_z: at n_a/n_z interface (§ 5·3).

ρ, τ: one plate of the stack, with one half-space above and below.

R, T: whole stack, with one half-space above the top plate and one half-space below the bottom plate.

r₁, r₂, r₃, t₂, r₃, t₃ :defined in § 4·1 (see also Fig. 2).

R_z, R_0z : defined in § 5·3.

μ, μ_1,μ₂ : ratio of amplitudes in successive layers, defined by equations (6)–(10), (15), (20), (21), (23), (24), (44). (|μ₁| < |μ₂| when μ’s are real.)

h, h_1,h₂: ratio of upgoing to downgoing wave, defined by equations (7) – (9), (11), (12), (16), (17), (22), (26), (45), (46). (|h₁| < | h₂| when h’s are real.)

m: = ⁠, and equation (32).

α: defined by equations (13) and (14).

θ: defined by equations (25) or (49).

2k : coefficient of h in equation (12), (17) or (46).

2k′ : coefficient of μ in equations (10), (15) or (44).

p : number of plates in the stack.

j: ordinal number of a plate in the stack, from j = 1 for top plate to j = p for bottom plate.

v_a, v_b : defined in § 5·2.

The formulae developed in this paper were used by Land (1966) in connexion with the reflecting layer in the eye of Pecten. The symbols used by him differ from those used in the present paper in the respects shown in Table 1.

The main steps in the procedure are listed here, numbered to agree with the sub-sections of § 4 in which the formulae are derived.

The expressions given in this section are appropriate to the case where the optical thicknesses of plates and spaces are equal, incidence is normal to the interfaces, and there is no reflexion from structures above the topmost plate or below the bottom one. Extensions of these results to more general cases are given in § 5.

The equations are numbered according to their sequence in § 4.

Derive a pair of recurrence relations connecting the amplitudes of the upward-and downward-propagating waves at successive layers in the stack.

This is a complex quantity, indicating the phase as well as the absolute amplitude of the reflected light.

Take the square of the modulus of R to obtain the reflectance of the stack (fraction of incident intensity that is reflected).

This approaches unity (complete reflexion) as the number of plates is increased.

Suppose that two partially reflecting surfaces are separated by a layer of material which introduces a phase delay of δ (Fig. 2).

r₁, t₁, r₂ and t₂ are the reflexion and transmission coefficients for light incident from above, and and are the corresponding quantities for light incident from below.

Since the system now under consideration (one plate flanked by two half-spaces) is symmetrical about the centre of the plate, these expressions for ρ and τ are appropriate for light incident either from above or from below (i.e. ρ′ = ρ and τ′ = τ).

When this same plate, the jth, is in place in the stack, light is incident on it from below as well as from above because of reflexion from lower layers in the stack (Fig. 3).

These equations are the required recurrence relations.

With this convention for the signs, |μ₁| < 1 irrespective of the signs of r and sin ϕ, so that the (μ₁, h₁) solution is always the one with amplitude decreasing in the down-ward direction through the stack.

The upper sign makes |h| < 1 so that the (μ₁h₁) solution is the one in which the net energy flux is downwards.

This quantity (reflectance, the fraction of the incident light intensity that is reflected) is the square of the modulus of R. It is convenient to use different procedures for obtaining the modulus when the μ’s are real and when they are complex.

This quantity is real, and lies between 0 and 1.

With an infinite number of plates m approaches zero and equation (35) shows that the reflectance then approaches unity, i.e. all the incident light is reflected.

As the wavelength is changed, ϕ, and therefore also θ, changes, and this function fluctuates, the fluctuations becoming closer as the number of plates, or their optical thickness, is increased. This behaviour of multilayer dielectric reflectors is well known (e.g. Vašiček, 1960). The reflectance varies between zero when sin pθ = 0 and r²/cos²ϕ when sin pθ = ±1.

This may be approached from either equation (35) or equation (39); the result is the same in both cases but the approach from equation (35) is the simpler.

The appropriate expressions for ρ and τ have already been derived (equations (1b) and (2b)).

Equations (27)–(31), (33), (34), (36)–(38) are still valid provided that 2k is taken as the coefficient of h in equation (46) instead of in equation (17), and θ is defined by equation (49) instead of equation (25). The signs to be chosen in evaluating the μ’s and h’s should be selected by the following criteria:

(a) When the μ’s are real, μ₁ is the solution of equation (44) whose absolute value is less than unity. The correct sign for h₁ is obtained by substituting μ₁ into equation (45).

(b) When the μ’s are complex, μ₁ is the solution of equation (46) whose absolute value is less than unity. The correct sign for μ₁ is obtained by substituting h₁ into equation (45) and rearranging to obtain μ.

Calculation of the reflectance of the whole stack is a little more laborious than when ϕ_a = ϕ_b, but it can be carried out by means of equations (34) or (38). When μ is real (|k′| > 1), equation (34) is used. k′, the coefficient of 2μ in equation (44), must be evaluated, and from it, μ₁ and μ₂. m² is then obtained as (μ₁/μ₂)^p. k, the coefficient of 2h in equation (46), is evaluated, and k and m² are substituted into equation (34), giving the required result. When μ is complex, θ is evaluated from equation (49), and k and θ are substituted into equation (38).

When the incident light is not directed along the normal to the reflecting surfaces in the stack, different expressions must be used for r, ϕ_a and ϕ_b. Equations (1b) and (2b) will then give the new values for ρ and τ, and the rest of the treatment is unchanged.

Note that if ϕ_a = ϕ_b for normal incidence, this will no longer be true for obliquely incident light.

The situation to be considered is shown in Fig. 4, where the stack is deposited on a material of refractive index n₅ and may also be covered with a material of refractive index n₀. r_z and r₀ are the amplitude reflexion coefficients for n_a/n_a and n₀/n_a interfaces respectively.

The whole stack, from the upper to the lower broken line in Fig. 4, can be regarded as a single partially reflecting surface with reflexion and transmission coefficients R and T defined by equations (30) and (31).

If these results are used when the incident light is oblique, it is important that the same sign convention should be used throughout for r₁ (see equation (50)).

R_s, and R_0s, like R, are complex quantities indicating the phase as well as the absolute amplitude of the reflected light. To obtain the corresponding reflectances, it is necessary to evaluate R_z or R_0z as a complex quantity and take the sum of the squares of its real and imaginary parts.

It has so far been assumed that each repeat in the stack consists of one plate of refractive index n_b and one space of refractive index n_a. If the repeating unit is more complex, the reflexion and transmission coefficients ρ and τ for a single unit can be obtained by repeated application of equations (1) and (2); this is the procedure described for example by Vašiček (1960), and used in § 5·3. μ, h, k and k′ are then found by substituting this ρ and τ into equations (10), (11) and (12), and from them, m and θ. The reflectance of the stack is then found from equation (34) or (38) according as the μ’s are real or complex.

I wish to acknowledge the helpful suggestions of Professor W. A. H. Rushton, who read the typescript, and of Professor O. S. Heavens.