Shine and transparency do not guarantee sharpness

A household mirror shows your face well. Why not cut a disk from one for a telescope? To see the problem, imagine a faint star: light from the whole aperture must form a coordinated pattern near one location, rather than several displaced reflections.

In a typical second-surface mirror, the main reflecting layer is behind the glass. Some light reflects at the front boundary; another part passes through the glass, reflects from the metal and passes back out. This can produce a ghost image, while glass inhomogeneity and dispersion add to the problem. Precision mirror optics often use a first-surface mirror, with the reflecting layer on the light-facing side. Edmund Optics: first and second surfaces.

A telescope needs specified material properties, the right shape and preservation of that shape in use. The names “glass” and “metal” alone guarantee none of these. Window glass is not forbidden by optical laws, but its ordinary purpose does not ensure the specification a particular telescope needs.

Seven properties hidden behind the word “quality”

First separate the properties of what looks like one smooth, transparent object.

  • Surface figure describes how closely the surface follows its designed sphere, parabola or other curve. A shiny but incorrectly curved surface directs light incorrectly.
  • Roughness is fine texture over much shorter distances along the surface. It can scatter light; a good overall radius does not prove a good polish. Waviness lies between overall figure and microroughness.
  • Homogeneity means uniform refractive index within the working volume. It matters particularly when light passes through a lens.
  • Mechanical stress means internal forces that can change shape and, in glass, cause birefringence: different polarizations encounter different optical conditions.
  • Dispersion is the wavelength dependence of refractive index. It is related to coloured fringes but is not another name for inhomogeneity.
  • Transmission describes how much light in the required band passes through the part; wavelength, thickness, absorption and surface losses matter.
  • Thermal expansion describes dimensional change with temperature. It differs from the temperature dependence of refractive index itself.

These requirements are measured separately. Choosing low-expansion glass alone does not fix roughness; a photograph of a shiny surface does not prove homogeneity. Edmund Optics: figure and metrology, SCHOTT: stress in optical glass.

Choose a material for its job, not its prestige

Optical glass, such as N-BK7, has documented refractive indices, dispersion and other properties. It can form a lens or a mirror substrate. Transparency in the visible band does not mean transparency at every wavelength: the datasheet provides conditions rather than the word “perfect.” SCHOTT: N-BK7 datasheet.

Low-expansion glass-ceramic, including ZERODUR, is used for stable precision substrates. Its strength is small dimensional change within a specified temperature interval. This does not promise zero deformation under any heating: material grade, temperature gradients and support still matter. SCHOTT: ZERODUR and temperature grades.

Metal can also maintain an optical figure. Webb’s segment substrates are beryllium: mass, stiffness and cold behaviour matter. Deposited gold reflects the required infrared light, a different job from maintaining shape. XMM’s X-ray mirrors followed another route: thin nickel shells were electroformed on precise mandrels and subsequently released. “Making a mirror” therefore does not always mean polishing a solid glass disk. NASA: Webb’s mirrors, ESA: manufacturing XMM mirrors.

A special material must answer a special requirement. Machining beryllium produces hazardous dust and belongs in controlled professional manufacturing; Webb is not a suggestion to reproduce that process at home. CDC/ATSDR: beryllium.

The substrate holds the figure; the coating handles the light

Separate a protected mirror into three functional parts. The substrate carries the figure and mechanical load. A thin metal layer provides the required reflection. A protective dielectric can improve coating durability and alter its spectral behaviour. Here, dielectric means an optical non-metallic material; real systems may contain several layers or omit this protection.

Three layers, three different jobs

Light arrives from this side ↓

123
  1. Protective dielectric: durability and spectral properties
  2. Reflecting metal: interaction with light
  3. Substrate: shape, stiffness and thermal behaviour

Functional sketch of a protected mirror; layer thicknesses are greatly exaggerated and not to scale. Coatings do not all share this structure. This is neither a Webb drawing nor a deposition recipe.

Gold is not universally best: the choice depends on wavelength, incidence angle, polarization, environment and the complete layer structure. Webb uses it for infrared observations; XMM uses gold in the very different geometry of grazing-incidence X-ray reflection. Other jobs use aluminium, silver and various protective systems. Compare verified spectral properties under the required conditions rather than mirror colour. Edmund Optics: metallic coatings.

On a lens, a coating often has the opposite job: reducing unwanted reflection and increasing useful transmission. Neither mirror metal nor an antireflection coating can repair an arbitrarily incorrect substrate geometry.

Polishing is not the end of the job

Manufacturing starts with requirements: which band, operating temperature, aperture and allowed contribution to overall error? A blank is then selected. For glass, annealing is a controlled thermal cycle that reduces excessive residual stress; the schedule depends on material and size. This is a separate task from polishing. SCHOTT TIE-27, section 1.1.

Grinding brings the shape closer to its target; polishing reduces fine irregularities and, together with correction, finishes the surface. Measurement shows where more material must be removed; shine alone cannot. Optics are checked again after coating, then mounted and checked as a system.

Measurement accompanies the whole process
  1. Requirements: band, temperature, allowed error map
  2. Blank and glass annealingCheck: certificate, homogeneity, stress
  3. Grinding towards the required shapeCheck: geometry and polishing allowance
  4. Polishing and correctionCheck: figure and roughness → repeat correction if needed
  5. Coating depositionCheck: reflection/transmission and figure change
  6. Mounting and testing under operating conditionsCheck: wavefront after mounting and at the required temperature

General route for glass optics, not a universal manufacturing procedure. Mirrors and lenses need different coatings; beryllium and XMM replication follow other routes. Measurements determine whether correction must be repeated.

NASA describes Webb’s loop: after initial polishing, cold-induced figure changes were measured and incorporated into correction before coating. ESA separately checked XMM shells after attachment. These examples explain why inspection occurs between operations as well as at the end. Vacuum deposition and electroforming are described here as professional processes, without a home recipe.

A finished figure still has to survive its surroundings

A part may be correct on a metrologist’s bench and deform in a different mount. Excessive clamping, different support points or uneven temperature change its condition. Testing therefore examines the mounted system under representative operating conditions as well as the free surface. Supporting thin shells was a distinct engineering problem in XMM manufacturing. ESA: support, integration and deformation.

A freely expanding, uniformly heated material mainly changes scale. Uneven heating or constrained expansion can change figure and stress. Computing how much a length grows therefore does not yet compute stellar blur or focus shift.

Branch · Thermal expansionOptional · safe to skip

100 mm of N-BK7 warmed by 20 degrees

For a small relative length change and approximately constant coefficient:

ΔL ≈ αLΔT.

L is initial length, α the linear expansion coefficient in K⁻¹, and ΔT temperature change in K. A difference of 20°C equals 20 K. The N-BK7 datasheet lists mean α = 7.1 × (10)^( − 6) (K)^( − 1) over −30…+70°C. Take a freely and uniformly heated length L = 100 mm from 20 to 40°C:

ΔL ≈ 7.1 × (10)^( − 6) × 100 × 20 mm = 0.0142 mm = 14.2 μm.

K⁻¹×mm×K gives mm; cooling reverses the sign, and no temperature change gives zero. Applying the mean coefficient over a narrower interval is a teaching approximation. It does not describe Webb’s cryogenic operation, a tightly clamped part, temperature gradients or refractive-index changes. The 14.2 µm result is not a measured surface error.

Surface nanometres and optical-path nanometres are different quantities

Light from different aperture regions must arrive with the required relative phases. A wavefront, a surface of equal phase, helps describe this coordination. Wavefront error is often expressed as an equivalent optical-path difference, in nanometres or fractions of a wavelength.

If one mirror region rises towards the light, the outward and return paths shorten. A surface-height value therefore cannot simply be called the same wavefront error. For a lens the relationship differs: refraction, thickness and refractive index enter the problem.

An uneven surface changes relative optical path

Surface height h, nm

025−101

Path difference δℓ, nm

0−50−101

Arbitrary transverse coordinate x

h = 25 nm → |δℓ| = 50 nm; λ = 550 nm → 1/11

Teaching map h(x)=25 cos²(πx/2), −1≤x≤1; δℓ≈−2h for near-normal incidence and small slopes. Vertical scales are arbitrary and exaggerated; the two panels use different scale increments. This is not a ray trace or a blur prediction. A common displacement of the whole surface would produce only piston.

Branch · Path geometryOptional · safe to skip

Where the factor of two comes from near normal incidence

Let a common reference plane lie a distance z from the undisplaced mirror, with a local rise h towards the light. For a horizontal patch at normal incidence, the round-trip path is:

ℓ(h) = 2(z − h), δℓ = ℓ(h) − ℓ(0) = − 2h.

For h = 25 nm, the difference has magnitude 50 nm. At λ = 550 nm this is 50/550 = 1/11 wave. The smooth map in the figure assumes near-normal incidence and small slopes; it is not an exact ray trace over the artificially exaggerated slope.

A crucial check: if the whole surface shifts by the same 25 nm, every path changes equally. This common “piston” phase shift does not itself distort wavefront shape. In the figure, centre and edges have different heights. Even then, maximum height alone does not determine spot size: the full map, aperture, wavelength and system are needed.

Questions about “λ/10” matter more than the number alone

Metrology provides traceable measurement results with known conditions and uncertainties. Profilometry yields a surface profile; interferometry compares wavefronts. Fine roughness may need different instruments and a different range of spatial scales. One test cannot establish every property at once. Edmund Optics: measuring roughness.

When you see “λ/10”, ask: surface or wavefront, which wavelength, which clear aperture, free or mounted, and at what temperature? Which reference figure was subtracted, were tilt and defocus removed, and which spatial scales remain in the data? Without this context, the numbers may describe different things.

Branch · Reading a mapOptional · safe to skip

PV and RMS cannot replace one another

PV is the difference between the map’s highest and lowest values. RMS is root-mean-square deviation from a chosen reference; with the mean removed:

\mathrm{RMS}=\sqrt{\frac1N\sum_{i=1}^N(e_i-\bar e)^2}.

e_(i) are errors at equally weighted samples, \bar e their mean and N their number. The result has the map’s units. A narrow high region and a broad ripple can have equal PV but different RMS. Real aperture calculations account for sample area and the mask. Comparisons must preserve the reference, processing and working region. Neither definition sets a universal tolerance for all telescopes.

Good optics requires a coordinated chain

We can now answer the opening question. A household part may be transparent or shiny without guaranteeing the figure, spectral behaviour and stability a telescope needs. Material sets possibilities; processing achieves the figure; coating controls interaction with light; mounting and temperature influence preservation of the result. Measurement closes the chain.

Try explaining three cases: why a perfectly smooth but incorrectly curved surface gives a poor image; why glass-ceramic cannot be declared unchanging at every temperature; and why shifting the whole mirror equally differs from a local bump. If your answers invoke different mechanisms rather than the word “quality”, you have the central idea.

Later parts of the course use these requirements to explain lens colour correction and mirror manufacture. Revisit the light path in “How does a telescope work?”. The numbers here are documentary and computational examples, not physical test results from a manufactured telescope.