How we measure

Spectroscope

The Spectroscope shows what a hand spectroscope shows when white light passes through one of our seven gemstones. You see the spectrum from violet to red, with dark lines and bands where the stone absorbs. The page calculates the spectrum from the same measured absorption spectra as the Dichroscope. Nothing is drawn by hand. It is a teaching model of the instrument. It does not measure or identify any real stone.

The instrument

A hand spectroscope is a short tube with a narrow slit at the far end and a prism or a diffraction grating inside. Strong white light passes through the stone (or reflects from inside it) into the slit. The prism or grating spreads the light into a band of colors, violet at one end and red at the other.

Where the stone absorbs, its part of the band goes dark. An ion that absorbs one narrow wavelength makes a sharp absorption line. An ion that absorbs a wide stretch makes a broad band. A spectroscope with a scale lets you read the wavelength of each line in nanometers (nm).

A grating spreads the wavelengths evenly. A prism does not: glass bends violet much more than red. So the prism squeezes the red end into a short stretch and draws out the violet end. The same light spread over a longer stretch also looks dimmer. So the violet end of a prism’s spectrum is fainter than a grating’s.

Prism and grating spectraTwo spectra from 400 to 700 nm with a mark every 50 nm. In the grating spectrum the marks are evenly spaced. In the spectrum of a 60 degree N-SF11 prism the place follows the deviation, 75.5 degrees at 400 nm to 64.7 degrees at 700 nm, so the violet end is drawn out and the red end is squeezed: 600 to 700 nm takes 13 percent of the width against 63 percent for 400 to 500 nm.400400450450500500550550600600650700700Prism: 60°, dense flint glassviolet drawn out, red squeezedGrating: even spacingwavelength, nm
Look at the marks every 50 nm: even on the grating, crowded at the red end of the prism.

The model

The light
A strong incandescent lamp, CIE illuminant A (2856 K), the classic hand set-up, or daylight, CIE D65. Illuminant A is calculated from its CIE definition: Planck’s law at 2848 K with c2 = 1.435 × 107 nm K, set to 100 at 560 nm. D65 comes from the CIE table. Either is a relative spectral power S(λ).
The stone
One straight path of d from 1 to 10 mm (5 mm to start, as in the Dichroscope), across the optic axis, as the side view draws it, in unpolarized light. A doubly refractive stone splits the light into two rays that share it equally. The o-ray (ordinary) vibrates across the axis and the e-ray (extraordinary) along it. Each ray has its own absorption spectrum:
T(λ) = ½ [exp(−αo d) + exp(−αe d)].
An isotropic stone has one spectrum. The shape you choose changes only the drawing.
Fluorescence
Under strong light, the chromium in ruby and emerald glows red in its R lines. So these lines can look bright instead of dark. With Fluorescence on, each R line is added as a Gaussian 1 nm wide. Its height is a fixed gain, times the share of the lamp’s light from 380 to 620 nm that the stone absorbs, times the lamp’s own level at the line, seen through half the path. This part is illustrative: we hold no measured fluorescence spectrum, so its heights are not measured.
The prism
A 60° prism of SCHOTT N-SF11 dense flint, set at minimum deviation for the sodium D line (589.3 nm). Its index comes from SCHOTT’s published Sellmeier fit,
n2 − 1 = Σ Bi λ2 / (λ2 − Ci), with B = 1.73759695, 0.313747346, 1.89878101 and C = 0.013188707, 0.0623068142, 155.23629 µm2. It gives nd 1.78472 and an Abbe number of 25.68, as the datasheet does. The place of a wavelength across the eyepiece follows its deviation, from 75.5° at 400 nm to 64.7° at 700 nm. In the grating view, the place is simply proportional to the wavelength. Both views show 400 to 700 nm.

The spectra

The absorption comes from the shared spectral registry, the measured polarized spectra that the Dichroscope uses. Ruby comes from GIA’s published Cr3+ absorption cross-sections. Blue sapphire, emerald and amethyst come from G.R. Rossman’s measured files (the e-ray of amethyst is derived and provisional). The Dichroscope page sets this out. The Dichroscope reads the spectra on a 5 nm grid, which is fine for color but too coarse for lines. For the Spectroscope, the registry also holds the same spectra on a 1 nm grid. These come from the same files by the same steps (the same baseline, a 1 nm bin instead of 5).

The files are sampled every 0.6 to 1.2 nm (Rossman) and every 1 nm (GIA), so 1 nm keeps what they resolve. Averaged back over 5 nm, the fine spectra give exactly the Dichroscope’s spectra (case S5).

One sample per variety means the lines of one stone. The close pair of R lines in ruby, 1.4 nm apart, is resolved in neither source. So it shows as one line near 693 nm. In this sapphire, the iron line sits at 451.4 nm in the o-ray and 453.2 nm in the e-ray. So in unpolarized light it falls near 452 nm. Its weaker iron lines near 460 and 471 nm are too faint to show.

In this emerald, the e-ray absorbs much more than the o-ray. Through more than a few millimeters, the light that reaches the eye is almost all o-ray. So the e-ray’s line at 684 nm shows only through 1 to 4 mm (in lamp light) and fades beyond. Its lines near 662 and 646 nm sit on the strong absorption of the e-ray. They never show in unpolarized light in this sample, at any path. The validation table gives their place in the e-ray alone.

From spectrum to what you see

No screen can show the color of a single wavelength: every spectral color lies outside its gamut. Each wavelength is drawn in the most saturated screen color of that light. The page converts its CIE 1931 color to linear sRGB (the IEC 61966-2-1 matrix), sets the negative primaries to zero, and scales it until its strongest primary is at full strength.

Its brightness is that of the empty spectroscope, compressed the way the eye adapts to a bright spectrum, times the light that the stone lets through. With u(λ) = 0.08 × S(λ) × ȳ(λ) × p(λ) / Yc(λ), where ȳ is the CIE luminosity function, p is how tightly the view packs that light (1 for the grating) and Yc is the luminance of the drawn color, the brightness is [1 − exp(−u)] × T(λ). The compression keeps both ends of the spectrum visible, as they are to an eye that looks into a bright spectroscope. The factor T keeps the contrast of every line exactly as the stone makes it. The stone’s own color in the side view comes from the 5 nm spectra through the same path in daylight, with the shared colorimetry.

Reading the lines

The page names each line and band of the stone’s reference list and measures it in the model. A line is the darkest point in its window. A parabola through the absorbance places it between grid points. Its contrast is c = 1 − T/Tc against the light 2.5 to 7 nm on either side. The contrast of a band is against the brighter of the two 12 nm stretches just outside it. Its span is where it is more than half as deep as at its darkest.

A line under 0.04 is too faint to see, under 0.15 is weak, under 0.4 is distinct, and above that is strong. A broad band needs a deeper dip to be seen, so its steps are 0.1, 0.35 and 0.7. These steps are our own choice, not a published scale. The reading names the coloring element when lines show: chromium for ruby and emerald, iron for sapphire. Wavelengths are given to the whole nanometer, as you read a hand spectroscope.

What the model leaves out

Validation

These cases run on the page’s own data each time we build the site. The build stops if one fails. The reference positions are the standard hand-spectroscope values. Where the model differs by more than 1 nm, or a value has no source that we hold, we mark it for checking and do not change it.

CaseWhat it checksResultNeedsOutcome
S1Every reference line within 1 nm of the model (a band: its darkest point inside its range)17 of 20 agree; sap-450, sap-460, sap-471 off, all flaggedevery one agrees or is flagged for checkingPass
S2Prism and grating name the same wavelength under the cursor (position, then back)7.5e-12 nmunder 0.05 nmPass
S3Diamond, cubic zirconia and moissanite: every color passes, no lines or bandslargest loss 0, 0 features0 and 0Pass
S4A longer path never lets more light through (1, 5 and 10 mm)largest rise 0.0e+00Pass
S5One source: the 1 nm spectra average back to the Dichroscope's 5 nm spectralargest difference 0.0%under 3%Pass
S6Illuminant A against the CIE table (380, 560, 780 nm)largest difference 0.0004under 0.001Pass
S7N-SF11 from its Sellmeier fit: nd and the Abbe number against the SCHOTT datasheet (1.78472, 25.68)nd 1.78472, vd 25.68within 0.00005 and 0.05Pass
S8The prism squeezes the red and stretches the violet; the grating spreads them evenly400 to 500 nm: 0.63, 600 to 700 nm: 0.13 of the prism's width; grating 0.33 and 0.33prism red under violet; grating equalPass
S9Fluorescence: bright R lines only with it on, only in ruby and emerald, each within 1 nm of its lineRuby 693.8 nm; Emerald 683.0, 680.4 nmruby and emerald onlyPass
S10The empty spectrum is smooth: no step between neighboring nm that could pass for a linelargest step 0.12under 0.15 (of 3)Pass

Every reference feature against the model, in nm. A line agrees when it is within 1 nm of each reference value; a band when its darkest point is inside its range. “To check” marks a value we could not back with a source we hold, listed for review.

GemstoneReferenceModelWhereAgrees
Ruby692.8 and 694.2693.3seen from 1 mm; weak at 1 mmYes
Ruby668667.8seen from 3 mm; weak at 3 mmYes
Ruby659659.0seen from 2.5 mm; weak at 2.5 mmYes
Ruby500 to 610547.0darkest at 5 mm, strongYes
Ruby475 and 476.5475.9seen from 1 mm; weak at 1 mmYes
Ruby468.5467.6seen from 1.5 mm; weak at 1.5 mmYes
Ruby400 to 430410.0darkest at 5 mm, strongYes
Sapphire450452.0seen from 1 mm; weak at 1 mmNo, to check
Sapphire460Not foundnot resolved in the measured spectrumNo, to check
Sapphire471Not foundnot resolved in the measured spectrumNo, to check
Sapphire540 to 700700.0darkest at 5 mm, distinctYes, to check
Emerald683684.0seen at 1 to 4 mm; distinct at 1 mmYes, to check
Emerald680.5681.4seen from 2.5 mm; distinct at 2.5 mmYes
Emerald662662.0too faint to see in unpolarized light; a peak of the e-ray aloneYes
Emerald646647.0too faint to see in unpolarized light; a peak of the e-ray aloneYes
Emerald637637.0seen from 1.5 mm; weak at 1.5 mmYes, to check
Emerald580 to 630614.0darkest at 5 mm, strongYes
Emerald477.4477.0too faint to see in unpolarized light; a shoulder of the o-ray aloneYes, to check
Emerald400 to 440424.0darkest at 5 mm, strongYes
Amethyst500 to 600545.0darkest at 5 mm, weakYes

Sources

We are not affiliated with any of these publishers.