How we measure

Dichroscope

The Dichroscope shows what a calcite dichroscope shows when you look through one of our seven gemstones in daylight. You see the two windows side by side as you tilt the stone and turn the dichroscope. The window colors come from measured polarized absorption spectra, by CIE colorimetry. Nothing is picked by hand: every number on the page comes from an equation below or from a documented data source. It is a teaching model of the instrument. It does not measure or identify any real stone.

The instrument

A calcite dichroscope is a short tube. It has a small rectangular window at the stone end and a lens at the eye. Between them is a cleaved rhomb of calcite (Iceland spar), with a glass prism on either side to straighten the light path. Calcite is strongly doubly refractive. So it shows two images of the window side by side, made of light that vibrates at right angles to each other.

Inside a calcite dichroscopeA calcite dichroscope seen from above: white light passes the stone and the window, then a calcite rhomb between two glass prisms splits it into two beams, and the lens shows two images of the window side by side. The light in the two images vibrates at right angles. The split is drawn larger than it is.white lightstonewindowcalcite rhomba glass prism on either sidelenseyeSeen from aboveWhat you see: two images of the windowtheir light vibrates at right angles
The calcite splits one beam into two images of the window.

Hold a stone close to the window, with white light behind it. Each image then shows the color of the stone for one direction of vibration. The type built with two polarizing filters is the London dichroscope. Its filters tint the colors, and that is why calcite is preferred.

A doubly refractive stone splits light into two rays that vibrate at right angles. In a uniaxial stone, the o-ray (ordinary ray) always vibrates across the optic axis. The e-ray (extraordinary ray) vibrates in the plane of the axis and the light. When the stone absorbs the two rays differently, the two windows differ in color: the stone is dichroic (pleochroic, with two colors). A singly refractive stone does not split the light, so its windows always match.

The model

The model is our first version: one straight path of d = 5 mm through a polished slab, in unpolarized daylight (CIE illuminant D65). The shape you choose changes only the drawing. The colors are always for the same 5 mm, so a longer path cannot make one stone darker than another.

Directions
k is the line of sight, into the screen. c is the optic axis, tilted θ from k. The o-ray vibrates along o = c × k (normalized), and the e-ray along e = k × o. When you look straight down the axis (c parallel to k), there is no split.
Absorption
αo(λ) is for light that vibrates across the axis, and αe(λ) for light that vibrates along it. Both are natural-log coefficients in cm−1. The e-ray at angle θ takes αe′ = αo cos2θ + αe sin2θ, and each ray passes T = exp(−s α d), with the saturation s at 1.
The dichroscope
When the dichroscope is turned by φ, window 1 passes vibration along p1 = cos φ u + sin φ v (u, v: the axes of the screen). Window 2 is at right angles. With cos2ψ = (p1 · o)2:
W1(λ) = ½ S(λ) [To cos2ψ + Te′ sin2ψ]
W2(λ) = ½ S(λ) [To sin2ψ + Te′ cos2ψ]
S is the illuminant. Unpolarized light splits into two unrelated (incoherent) waves, so their powers simply add. There is no interference term, whatever the retardation of the stone.
What follows
Turned 45° from the vibration directions of the stone, each window takes half of each ray, and the two windows match. Turned 90°, they swap. Down the axis, both rays are ordinary, and the windows match at any turn. An isotropic stone (diamond, cubic zirconia) has αe = αo.

For each window, the page shows its L*, C* and h° (CIELAB) and its share of each ray, cos2ψ and sin2ψ. Between the windows, it shows the CIEDE2000 color difference ΔE00 and its class.

The spectra

Optics belong to the species, and color belongs to the variety. Ruby and blue sapphire share the refractive indices of corundum, but not their absorption. The indices come from the gemstone data of the 3D Gemstone Simulator. The absorption comes from a shared spectral registry of measured, polarized spectra, each on a 5 nm grid from 380 to 780 nm.

A measured spectrum is decadic absorbance A for a sample of thickness t. We take the lowest absorbance over the whole range of the file as the baseline (reflection losses and the offset of the instrument). That range runs into the near infrared, where these stones barely absorb. We average the rest over each 5 nm bin and convert it: α = ln(10) A / t. Saturation s = 1 is the color content of the sample itself.

Gemstoneo-ray (E ⊥ c)e-ray (E ∥ c)Status
RubyGIA’s published absolute Cr3+ cross-sections in corundum, both polarizations (Dubinsky, Stone-Sundberg & Emmett 2020), at 2,164 ppma Cr: the content that gives the measured E ⊥ c absorption of a natural ruby from Chanthaburi, Thailand (GRR 1843, 0.787 mm) at 560 nmFitted, both rays
Blue sapphireMedium blue sapphire GRR 1020a, 4.289 mm, both polarizations measuredFitted, both rays
EmeraldEmerald GRR 3570, Afghanistan, 0.859 mm, both polarizations measuredFitted, both rays
AmethystAmethyst from the Anahí Mine, Bolivia: a 6.17 mm slab cut on (0001), so the light travels along the axis and vibrates across itDerived from the o-ray: the three color-center bands fitted, then each scaled by its published E ∥ c to E ⊥ c ratio: 1.06 near 545 nm, 1.92 near 410 nm, 1.40 near 357 nm (Hassan & Cohen 1974)o-ray fitted, e-ray provisional
Diamond, cubic zirconia, moissaniteColorless: no absorption in the visibleNone needed

The measured files are G.R. Rossman’s, on the Mineral Spectroscopy Server at Caltech. We checked two assignments before use. The o-ray of emerald is the yellowish green and its e-ray is the bluish green, from its polarized spectra (Wood & Nassau 1968; Webster 1955). A common cutting rule of thumb says the reverse. The e-ray of amethyst absorbs more and is the more reddish (Haidinger, in Holden 1925), and the derived spectrum follows this. Real amethyst is zoned and its color centers are slightly biaxial, so its two colors vary across a stone.

From spectrum to color

We weight the spectrum of each window by the CIE 1931 2° color-matching functions and add it up over the 81 wavelengths, with D65 as the light. We normalize the result so that an empty window (no stone) has Y = 1. That empty window is the white point for CIELAB. L*, C* and h° and the color difference come from these values. ΔE00 is CIEDE2000, checked against all 34 test pairs of Sharma, Wu & Dalal (2005).

To draw the windows, the color goes from XYZ to Display P3 on screens that show it, and to sRGB (the IEC 61966-2-1 matrix) on other screens. Deep reds and greens like those of ruby and emerald lie outside sRGB. The page brings them in by lowering chroma at constant lightness and hue until they fit. It says when it has done so and by how much. The figures always come from the true color, before that step.

How strong

The class names the color difference across the optic axis, with the dichroscope lined up, through 5 mm in daylight. Under 1 is none, 1 to 5 is weak, 5 to 16.5 is distinct, and over 16.5 is strong. We set these limits so that the seven gemstones land where the trade puts them, and then we froze them. Sapphire 41.6 and ruby 17.9 are strong, emerald 15.3 is distinct, amethyst 3.3 is weak, and diamond, cubic zirconia and moissanite are 0. Ruby and emerald sit close, so the line between distinct and strong falls halfway between them. The page also names the class for whatever tilt and turn you set, so it reads none at 45°.

What the model leaves out

Validation

We run our test cases on the page’s own data each time we build the site. The thin-limit case (V7) asks for 1 µm. But this ruby absorbs 75 cm−1 at 400 nm, so 1 µm really does tint white light slightly (ΔE00 0.19). So we run it at 0.1 µm, and check that the difference falls in proportion to the path. We also checked by hand. When you turn the eyepiece 90°, the windows visibly swap. The ring stops at 0°, 45° and 90°. The six lesson links open their exact view.

CaseWhat it checksResultNeedsOutcome
V1Isotropic null: diamond and cubic zirconia, 1,000 random orientations and turns eachlargest ΔE00 0.0e+0ΔE00 < 0.01Pass
V2Optic-axis null: every doubly refractive stone within 0.5° of its axis, any turnlargest ΔE00 0.0014ΔE00 < 0.05Pass
V3The 45° null: across the axis, turned 45°, window 1 equals window 2largest difference 1.4e-14within 1e-9Pass
V4Swap: window 1 at ψ equals window 2 at ψ + 90°largest difference 1.4e-14within 1e-9Pass
V5Energy: window 1 + window 2 is the same at every turnlargest change 3.0e-16within 1e-9Pass
V6Monotonic: lined up, the spectral difference never falls as the stone tilts from 0° to 90°holds for all four dichroic stonesall dichroic stonesPass
V7Thin limit: 0.1 µm of each stone, both windows against the empty window, and ten times the difference at 1 µmlargest ΔE00 0.0191 at 0.1 µm, 0.19 at 1 µmΔE00 < 0.1 at 0.1 µm, in proportion to the pathPass
V8Colorless and doubly refractive: moissanite, 1,000 random orientationslargest ΔE00 0.0e+0ΔE00 < 0.01Pass
V9Hue order across the axis, lined up: ruby e-ray toward orange; sapphire o-ray toward violet, e-ray toward green; emerald e-ray bluer; amethyst e-ray more reddishruby 353° to 14°, sapphire 266° and 139°, emerald 161° to 173°, amethyst 327° to 343° (o to e)all holdPass
V10Strength classes: each stone across its axis, lined up, in its class in the tradediamond 0.0 none; moissanite 0.0 none; cubic zirconia 0.0 none; ruby 17.9 strong; sapphire 41.6 strong; emerald 15.3 distinct; amethyst 3.3 weaknone < 1 ≤ weak < 5 ≤ distinct < 16.5 ≤ strongPass
V11Pipeline: the empty window in sRGB; CIEDE2000 against all 34 Sharma, Wu & Dalal (2005) pairswhite (1.0000, 1.0001, 0.9999); largest ΔE00 error 4.9e-5±0.005; within 1e-4Pass
V12Units: every measured dataset round-trips absorbance to α and back, its thickness recorded5 datasets, largest error 2.2e-16within 1e-9Pass

Sources

We are not affiliated with any of these publishers.