Materials Science · Light interactions and optical materials

Optical anisotropy, scattering and disorder

Useful background: Atoms, bonding, structure and defects; Measuring and checking material claims. Each worked case explains the quantities it uses.

Back to the chapter

After reading, you can

Distinguish direction-dependent propagation from light redistribution by structure.
Explain polarisation and two transverse principal phase responses using a bounded monochromatic plate model.
Interpret matched sample-rotation and orthogonal-analyser powers without confusing polarisation with ray direction.
Use direct and total collection records to distinguish redistribution from absorption under stated conditions.
Identify controls and unresolved alternatives before assigning a structure or composition.

Structure changes optical paths

In an anisotropic crystal, refractive response depends on direction and polarisation. Scattering arises when spatial variations redirect light, such as particles, pores or interfaces. Both can occur in the same material without being the same mechanism.

Imagine two observations: a clear plate becomes bright between crossed polarisation selectors when the plate is rotated; a cloudy specimen sends little power to a narrow detector but much more to a wider collection system. The first can involve a change of relative optical phase. The second can involve light leaving in other directions. Neither observation alone identifies a composition.

The following records are separate fictional teaching cases. One is a uniform ideal phase-retarding plate; the other is a set of specimens with supplied directional and total power measurements. Their numbers are not measurements of a named crystal or ceramic, and passing one model comparison does not qualify an optical component.

Polarisation is a field orientation, not the beam direction

Take a beam travelling along z. In the exterior transverse-wave description, X and Y are perpendicular directions across the beam. Linear polarisation along X means the electric field oscillates back and forth along that fixed line. It does not mean the light ray travels along X. Polarisation can also be elliptical: at one fixed position, the field tip traces an ellipse over a cycle rather than one line.

An ideal linear analyser passes the field component along its transmission axis. For linearly polarised light at angle α to that axis, the projected field amplitude is E_a cos α and the passed power fraction is cos²α. Power follows a time-averaged squared field, not the signed field itself. Rotating an analyser can therefore change a reading even when no sample changes. A blank or known isotropic comparison and a stable incident power are essential controls.

In a birefringent material, two allowed optical polarisation modes can have different phase indices for a chosen propagation direction. These describe how the field’s phase advances, not two chemical compositions. In general the phase and energy-flow directions in an anisotropic crystal need not coincide. We avoid that extra geometry below by stipulating propagation along a principal material direction, with both model field components transverse and no spatial beam splitting.

Not every crystal has different linear optical response in every direction. Cubic symmetry can make the bulk linear response isotropic unless symmetry is broken, for example by stress or an applied condition. Conversely, oriented polymers or stressed glass can have optical anisotropy without being a single crystal. Direction and size can change observed behaviour prepares this broader direction-versus-structure distinction.

Two phase responses can change polarisation without losing power

For the first fictional case, use a monochromatic beam at vacuum wavelength λ₀ = 600 nm. A plate of thickness d = 15.0 µm has orthogonal transverse fast and slow axes with phase indices n_f = 1.500 and n_s = 1.510 for this propagation direction. Fast refers to the lower phase index. The phase-index difference is Δn = 0.0100. Temperature, wavelength, direction and state remain fixed.

Assume ideal matched boundaries, no absorption, no scattering and the same exterior medium for the input and output power comparison. These are conditions of the illustrative plate model, not consequences of calling a specimen transparent. The prepared input beam has P_b = 10.0 mW and is linearly polarised along laboratory X. The whole forward output before an analyser retains 10.0 mW under these assumptions.

A component traversing thickness d accumulates phase 2πnd/λ₀. One full cycle is 2π radians. The slow-minus-fast phase difference is δ = 2πdΔn/λ₀. With d = 15000 nm, dΔn = 150 nm, so δ = 2π(150/600) = π/2: one quarter cycle. Wavelength, refractive index and dispersion supplies the phase-index and vacuum-wavelength distinction. This phase difference is not by itself a pulse travel-time measurement.

Let θ be the angle from laboratory X to the plate’s fast axis, rotating the plate about the unchanged beam direction. Resolve the incident field into fast-axis amplitude E_a cos θ and slow-axis amplitude −E_a sin θ. The slow component acquires a different phase, then the components recombine in the same forward beam. At θ = 45.0° and δ = π/2, the output field tip traces a circle in the transverse plane. It is not simply a linear-polarisation arrow turned through a fixed angle, and it is not a ray bent into a circle.

Original monochromatic ideal-retarder field diagram. The upper panel is a transverse view, with laboratory X rightwards and Y upwards. A dot-in-circle symbol separately denotes propagation z out of this page; the beam does not travel along the X input electric-field direction. The fast plate axis is at 45 degrees from X, and the orthogonal slow axis is dashed. Supplied phase indices are 1.500 and 1.510 at vacuum wavelength 600 nanometres, and thickness 15.0 micrometres gives relative phase pi/2. Along this stipulated principal propagation direction, no spatial beam splitting is included. In the lower panel the same ideal output is viewed at one fixed position during a cycle. Its normalised field components are E_X/E_a = [cos u + sin u]/2 and E_Y/E_a = [cos u − sin u]/2. Equal horizontal and vertical scales are 160 pixels per unit input field amplitude. The green locus is a circle of radius one divided by square root of two. A blue circle and solid radial line identify u=0, at (0.5,0.5); an orange square and dashed radial line identify u=pi/2, at (0.5,−0.5). The field tip changes orientation in time; this is not a circular ray, a material trajectory or an optical-power plot. All values are original fictional model inputs.

Read a rotation contrast with both analyser channels

Compare ideal orthogonal analyser channels: X is parallel to the prepared input polarisation and Y is crossed. They can be separate calibrated port measurements or matched observations of the same unchanged output beam. Keep the incident beam, wavelength, specimen position, normal incidence, aperture and detector response fixed while rotating the plate about z. Do not mix a plate rotation with a changing analyser axis.

For readers following the field calculation, write u = 2πft for the common phase in radians at a fixed output position, with f in cycles per second and t elapsed time; an overall phase origin is arbitrary. Projection back onto crossed Y gives E_Y/E_a = sin θ cos θ[cos u − cos(u − δ)]. The minus sign follows from the original slow-axis projection. Averaging its square over a full cycle and dividing by the input average E_a²/2 gives P_Y/P_b = sin²(2θ) sin²(δ/2). The parallel channel carries P_X/P_b = 1 − P_Y/P_b in this loss-free model.

For δ = π/2, sin²(δ/2) = 1/2. At θ = 0° the crossed channel is dark; at 22.5° it carries 2.50 mW; at 45.0° it carries 5.00 mW. The parallel channel correspondingly carries 10.0, 7.50 and 5.00 mW. Their sum remains 10.0 mW. A lower one-channel reading is therefore not evidence of absorption in this model.

The curve is a conditional phase-retarder prediction. Real rotation dependence can also involve unequal attenuation of polarisations, called dichroism, changing reflection, stress, coatings or polarisation-dependent scattering. Constant total power and a matched blank constrain some alternatives but do not uniquely prove the microscopic model. Additional phase/polarisation analysis and wavelength or thickness comparisons must test the remaining alternatives.

Original ideal plate prediction at λ₀ = 600 nm and δ = π/2; input P_b = 10.0 mW. These are orthogonal analyser powers, not direct-versus-diffuse transmission.
Plate θ (degrees)	Crossed Y (mW)	Parallel X (mW)	Sum (mW)
0	0	10.0	10.0
22.5	2.50	7.50	10.0
45.0	5.00	5.00	10.0
67.5	2.50	7.50	10.0
90.0	0	10.0	10.0

Original loss-free model prediction for the separate ideal retarder: vacuum wavelength 600 nanometres, phase-index difference 0.0100, thickness 15.0 micrometres, relative phase pi/2 and prepared input beam power 10.0 milliwatts. Only the plate rotates about the fixed beam direction; input X polarisation, both analyser axes and collection stay fixed. The horizontal axis is plate angle theta from 0 to 180 degrees, at eight-thirds pixels per degree. Vertical power is 0 to 10 milliwatts, at 40 pixels per milliwatt. Dashed orange crossed-Y power is 5 sin-squared(2 theta) milliwatts; solid blue parallel-X power is 10 minus that value; dotted grey is their constant 10-milliwatt sum. Circles on Y and squares on X identify the tabulated plate angles 0,22.5,45,67.5,90 degrees. At 22.5 degrees, Y/X are 2.50/7.50 milliwatts; at 45 degrees they are both 5.00. Coincident markers there represent the same power in different channels, not missing data. These are coherent polarisation analyser channels, not direct-versus-diffuse collection, absorbed power, experimental data or a unique microscopic mechanism.

Disorder can redistribute light without the same phase signature

A pore, particle, second phase, rough surface or change in optical orientation can create a spatial variation in refractive response. Light interacting with that variation can leave the original beam direction. The strength and angular distribution depend on optical contrast, size and shape relative to wavelength, orientation, concentration and correlations. More interfaces do not imply one universal scattering law or a guaranteed opacity.

Randomly oriented birefringent grains can present different local optical responses even with the same chemical formula. Pores and chemically different boundary layers supply other contrasts. In a cubic material with isotropic bulk linear response, grain rotation alone need not create that bulk-index contrast; residual pores, second phases, boundary regions or stress can still matter. Phases, grains and microstructure distinguishes those structural accounts.

Transparent polycrystalline alumina specimens have been reported under particular dense, fine-grained conditions. That observation rules out the blanket claim that every polycrystal of an optically anisotropic material must be opaque; it does not supply a universal grain-size cutoff. A single crystal can also contain inclusions, damage or surface roughness. Chemistry and the labels single crystal or polycrystal leave the actual optical comparison unresolved.

Scattered light need not be distributed equally in all directions or be completely unpolarised. A scattering pattern can depend on incident polarisation even when the question is not a bulk crystal’s directional phase index. Thus a polarisation-sensitive narrow reading cannot automatically be assigned to the retarder model.

Choose measurements that separate effects

Use controlled polarisation, angles and collection geometry. A transparent single crystal and a polycrystal of the same chemistry can differ because boundaries and orientation variations affect propagation. Surface roughness can add another contribution.

For the second fictional case, compare the table at λ₀ = 600 nm with P₀ = 10.0 mW incident on each specimen. All have matched thickness, illuminated area, surface-state protocol, incidence and temperature. The X/Y labels now specify two orthogonal incident linear polarisations with the specimen fixed. There is no output polarisation analyser: every collection accepts both output polarisations. This protocol and its powers are separate from the rotating ideal plate.

P_dir is power accepted by a stated narrow angular aperture around a fixed reference forward direction. P_T is the total power crossing the far-side hemisphere, including the narrow part and other transmitted directions; P_R is total power leaving the incident-side hemisphere. Assume valid complete collection, calibrated common incident normalisation, no optical gain, no wavelength conversion and no changing stored energy. Then P_A = P₀ − P_T − P_R, and the transmitted power outside the narrow aperture is P_other = P_T − P_dir.

A/X gives P_other = 8.0 − 6.0 = 2.0 mW and P_A = 10.0 − 8.0 − 1.0 = 1.0 mW. B/X gives P_other = 5.0 mW but the same P_A = 1.0 mW. Its lower direct power is accompanied by more transmitted power outside the aperture, not more absorbed power in this exact record. This establishes redistribution across the stated collection boundary, without identifying a unique pore or grain mechanism.

C/X has the same 3.0 mW direct reading as B/X, yet total transmission is only 5.0 mW. Its supplied total reflection is 1.0 mW, so P_A = 4.0 mW. The identical narrow readings conceal different full power accounts. Reflection, absorption and transmission develops those definitions; an integrating sphere or wider collection must still be validated for its ports, losses and calibration rather than assumed complete from its name.

Separate invented complete power records, all in mW; P₀ = 10.0 mW for each. X/Y refers to incident polarisation, with specimen and collection geometry fixed.
Record	P_dir	P_T total	P_R total	P_other = P_T − P_dir	P_A
A / X	6.0	8.0	1.0	2.0	1.0
B / X	3.0	8.0	1.0	5.0	1.0
B / Y	5.0	8.0	1.0	3.0	1.0
C / X	3.0	5.0	1.0	2.0	4.0

Original separate fictional specimen records at vacuum wavelength 600 nanometres and common incident power 10.0 milliwatts. Unlike the retarder figure, X/Y here labels incident polarisation and there is no output analyser; collection accepts both output polarisations. Four horizontal stacked bars have a common scale of 44 pixels per milliwatt and a full 10-milliwatt account. Plain blue is narrow-aperture direct transmission; diagonal-hatched orange is other transmitted power outside that fixed angular aperture; vertical-striped violet is total reflected power; cross-hatched pink is absorbed power under the complete passive nonconverting balance. A/X is 6,2,1,1 milliwatts respectively. B/X is 3,5,1,1. B/Y is 5,3,1,1. C/X is 3,2,1,4. All four sum to 10. A/X and B/X have the same total transmitted power 8 and absorbed power 1 but different angular collection. B incident-polarisation change moves 2 milliwatts into the aperture while total transmission, reflection and absorption stay fixed. C/X shares the direct reading 3 with B/X yet has absorption 4 rather than 1. Power outside the aperture may include a displaced or split beam, not automatically diffuse scattering. All values are supplied exact teaching records, not actual ceramic data or a unique structural identification.

Choose the next observation for the remaining ambiguity

For the fixed specimen B, changing incident X to Y raises P_dir from 3.0 to 5.0 mW while P_other falls from 5.0 to 3.0 mW. Total transmitted, total reflected and absorbed powers stay 8.0, 1.0 and 1.0 mW. This is polarisation-dependent redistribution relative to the same aperture in the supplied record. It does not establish the ideal plate’s phase indices, an absorption change or a new composition.

Match the measurement to the proposed explanation. To test a pure phase-retarder account, examine polarisation and phase with stable beam collection, a blank and known axes. To test redistribution, compare angular or aperture-dependent data while also collecting total outgoing power. To test absorption, require the complete common-denominator balance or an appropriate independent absorption measurement. Changing thickness, finish or illumination during a comparison introduces new possible causes.

Power outside the aperture can include a displaced or split directional beam as well as a broad distribution. P_other is not automatically diffuse-scattering power. Angularly resolved measurements and a check of beam alignment can distinguish these possibilities before a random-scattering model is assigned.

A measured angular pattern may support a scattering description without uniquely determining pore size, grain orientation, refractive contrast or absorption coefficient. Different structures can produce similar collected powers. A microscope, density/porosity evidence or another wavelength can help discriminate specified alternatives, but no one measurement name supplies a unique inverse solution.

Keep uncertainty with each comparison. Small differences in total power cannot be treated as exact mechanism changes when calibration, collection or sample variation is comparable. Calibration, uncertainty and repeatability prepares this assessment. The exact records here teach which extra observation is informative; they do not imply a universal material ranking, safe-use claim or composition test.

Worked examples

Work through the case

A dense ceramic and a single crystal share chemistry but transmit different amounts of visible light. What should be checked?

Compare pores, boundaries and optical anisotropy.
Inspect surface finish and thickness.
Separate absorption from scattering.

Chemistry alone leaves several structural explanations unresolved.

A phase change becomes a channel contrast

Use the ideal 15.0 µm plate, Δn = 0.0100, λ₀ = 600 nm and P_b = 10.0 mW. Find δ and the crossed/parallel powers at θ = 22.5° and 45.0°. Does the crossed reading measure absorbed power?

Convert d to 15000 nm. Then dΔn = 150 nm and δ = 2π × 150/600 = π/2.
At θ = 22.5°, sin²(2θ) = sin²45° = 1/2; sin²(δ/2) = 1/2. The crossed fraction is 1/4, giving 2.50 mW; the parallel fraction is 3/4, giving 7.50 mW.
At θ = 45.0°, the crossed fraction is 1/2, giving 5.00 mW; the parallel channel also gives 5.00 mW.
Both sums are 10.0 mW, and the model stipulates no absorption. The analyser changes which field component is selected; the sample’s absorbed power is not the rejected single-channel power.

δ = π/2. Powers are (Y, X) = (2.50, 7.50) mW and (5.00, 5.00) mW. The channel contrast is not an absorptance measurement.

One narrow reading hides two mechanisms

B/X and C/X both give 3.0 mW in the narrow aperture. Use their complete table entries to find other transmitted and absorbed powers. Can the common direct reading establish the same optical state?

For B/X, P_other = 8.0 − 3.0 = 5.0 mW and P_A = 10.0 − 8.0 − 1.0 = 1.0 mW.
For C/X, P_other = 5.0 − 3.0 = 2.0 mW and P_A = 10.0 − 5.0 − 1.0 = 4.0 mW.
Each total closes: 3.0 + 5.0 + 1.0 + 1.0 = 10.0 for B/X; 3.0 + 2.0 + 1.0 + 4.0 = 10.0 for C/X.
The same narrow power is compatible with more angular redistribution in B/X and more absorption in C/X. These records do not identify the structural cause.

B/X has 5.0 mW other transmission and 1.0 mW absorption; C/X has 2.0 and 4.0 mW. Equal direct transmission does not select the same full response.

Rotate the input without confusing protocols

For fixed specimen B, change incident polarisation X to Y while retaining the same incident power and both-polarisation collection. What changes, and does it demonstrate the ideal phase-retarder model?

Direct transmitted power rises by 5.0 − 3.0 = 2.0 mW.
Other transmitted power falls by 3.0 − 5.0 = −2.0 mW. Their total remains 8.0 mW.
Total reflected power remains 1.0 mW, so the absorbed remainder remains 1.0 mW.
The record concerns angular collection without an output analyser. It supports polarisation-dependent redistribution across that aperture boundary, not unique phase retardance or a change in absorption.

Two mW moves into the narrow collection while total transmission and absorption stay fixed. Polarisation dependence alone does not identify the phase mechanism.

Try the reasoning

Is every cloudy appearance chemical absorption?
Show the answer and reasoning
No.
Scattering can reduce direct transmission.
Why rotate polarisation?
Show the answer and reasoning
It can reveal anisotropic response.
Different optical directions may be sampled.
With the sample removed, a linearly X-polarised beam gives 10.0 mW through an X analyser, 5.00 mW through an analyser at 45.0° and zero through Y. Has a sample’s optical anisotropy been established?
Show the answer and reasoning
No sample response has been measured; the readings follow ideal analyser projection.
cos²0° = 1, cos²45° = 1/2 and cos²90° = 0. An instrument-axis rotation can create contrast without changing any material. Compare matched sample and blank records.
Keep the ideal phase indices, λ₀ = 600 nm and P_b = 10.0 mW, but double thickness to 30.0 µm. Find δ and the crossed power at plate angles 22.5° and 45.0°.
Show the answer and reasoning
δ = π. Crossed powers are 5.00 mW at 22.5° and 10.0 mW at 45.0°.
Doubling d doubles δ, but does not in general double intensity. Here sin²(δ/2) becomes 1, so the crossed fractions are sin²45° = 1/2 and sin²90° = 1.
At thickness 60.0 µm the same supplied Δn gives δ = 2π. Predict crossed power versus θ. Would darkness for every plate rotation prove that Δn = 0?
Show the answer and reasoning
Crossed power is zero for all θ in this model, despite nonzero Δn = 0.0100.
sin²(δ/2) = sin²π = 0. A whole-cycle relative phase is indistinguishable from zero phase in these intensity channels at this wavelength. Other thickness or wavelength evidence is needed; darkness alone does not prove isotropy.
For a separate supplied 750 nm record, explicitly keep Δn = 0.0100 and d = 15.0 µm, with the same ideal assumptions and P_b = 10.0 mW. Find crossed power at θ = 45.0°. Could that unchanged Δn have been assumed from the 600 nm record alone?
Show the answer and reasoning
δ = 0.4π and P_Y ≈ 3.45492 mW. The new Δn is an added stipulation, not a consequence of the earlier wavelength.
dΔn/λ₀ = 150/750 = 0.2, so δ = 0.4π and sin²(δ/2) = sin²(0.2π) ≈ 0.345492. Dispersion can change the indices; one wavelength does not establish their difference at another.
A new passive nonconverting record has P₀ = 10.0 mW, complete P_T = 8.0 mW and P_dir = 3.0 mW, but total reflected power is unknown. Can absorption be fixed at 1.0 mW?
Show the answer and reasoning
No. P_other is 5.0 mW, but the exact balance only gives 0 ≤ P_A ≤ 2.0 mW.
P_A = 2.0 − P_R with nonnegative reflected and absorbed powers. P_R = 1.0 would give 1.0 mW absorbed, but P_R = 0.5 would give 1.5 mW. Direct versus total transmission alone does not supply the missing reflected account.
Two matched specimens have the same total R and T, and one has more transmitted power outside a fixed aperture. Must it contain more pores or have a different chemical composition?
Show the answer and reasoning
Neither conclusion is established by those powers alone.
The extra collected angular redistribution is an observation. Pores, orientation variation, inclusions, roughness or other spatial optical contrasts can produce related signatures. Test specified alternatives with appropriate structural or angular evidence and preserve the matched conditions.

What to carry forward

Separate absorption, scattering and anisotropy.
Match structure and optical geometry.
A field polarisation direction differs from a propagation direction; phase contrast need not remove total beam power.
Keep incident-polarisation rotation, sample rotation, analyser rotation and collection changes distinct.
A direct-versus-total contrast can reveal redistribution under its stated power account without uniquely identifying a microstructure.
Neither cloudy appearance, crossed-channel brightness nor a crystal label proves composition or a universal transparency rule.

References

Teaching resources and research supporting this lesson’s explanations and examples.