Why does a star change when we change wavelength bands?
A visible-light image of a galaxy can readily show older stars, while an ultraviolet image can make regions containing young, hot, massive stars particularly prominent. It is the same galaxy. We have selected another part of its light and, with it, different properties of its material.
Observing ultraviolet requires matching the whole photon path: rising above absorbing atmospheric layers for bands inaccessible from the ground, choosing suitable coatings and windows, keeping surfaces clean and using a detector with the required spectral response. Replacing the camera alone is not always enough.
We will explore this with Hubble and STIS, its spectrograph that separates light by wavelength. From the article on telescope capabilities, we need the distinction between signal amount and resolution. From the materials article, we need the idea that material and coating are chosen for a particular wavelength. Equations and numerical examples appear in visible optional branches. NASA: ultraviolet stars and galaxies.
Why did a spectrograph need a rocket?
In 1801, Johann Ritter detected radiation beyond the violet edge of the visible spectrum. Detecting ultraviolet nearby and obtaining a distant astronomical ultraviolet spectrum are different tasks, however. The atmosphere transmits some UV wavelengths while absorbing others very strongly. NASA: discovery of ultraviolet.
In 1946, Richard Tousey’s group at the US Naval Research Laboratory used a high-altitude rocket to obtain a solar spectrum down to about 210 nm. Absorption by ozone and oxygen obstructed ground instruments and the balloons available at the time. Participant William Baum described the work as access to a previously unavailable part of the spectrum. The rocket provided a short observing window; it was not a permanent orbiting observatory. Baum, memoir of Tousey, 2002, “Dawn of the Space Age”.
Orbit made it possible to revisit targets and accumulate long observing records. IUE, launched on January 26, 1978, operated until September 1996, obtaining ultraviolet spectra of stars, gas, comets and other objects. STIS was installed on Hubble during the 1997 servicing mission. Following a power failure in 2004, astronauts repaired it in 2009. Ultraviolet astronomy needed both access to altitude and instruments that retained sensitivity in that environment. NASA: IUE, NASA: STIS history.
Why does a suntan not mean all UV is visible from the ground?
Here we use a broad conventional UV range of 10–400 nm; nm, a nanometer, is one billionth of a meter. Different fields define the lower boundary and subdivisions differently. UVA/UVB/UVC are common in Earth sciences, while NUV/FUV/EUV are used in astronomy; they do not form one universal set of boundaries. NASA Space Math, exercise25, NASA: UV sub-band names.
Some longer-wavelength UV reaches the surface, mainly around 290–400 nm. This is not a perfectly transparent rectangle: the result depends on wavelength, ozone column, clouds, aerosols and path angle. Much shorter wavelengths are largely closed to ground-based astronomy. Thus both “the atmosphere blocks all UV” and “UV reaches people, so a ground telescope sees all of it” are inaccurate. NASA Earth Observatory: atmospheric conditions.
The plotted bands belong to specific STIS detectors; they do not define all natural NUV/FUV. The current handbook lists nominal intervals of FUV-MAMA 115–170 nm, NUV-MAMA 160–310 nm and CCD 164–1030 nm. They overlap, with unequal sensitivity inside. The recording range with a particular grating is another limitation. STIS Instrument Handbook v25.0, §3.2, §7.2, §7.4.
Branch · With equationsAn optional calculation; it does not replace a material's spectral characterization.
How much energy does one UV photon carry?
Photon energy in a vacuum relates to wavelength as:
E = (hc)/(λ).
Here h is Planck’s constant and c is the speed of light. 1 eV, an electronvolt, equals 1.602176634 × (10)^( − 19) J. At 300 nm, the result is about 4.13 eV; at 150 nm, 8.27 eV; and at 121.6 nm, 10.20 eV. Halving wavelength doubles the energy of one photon. The calculation uses defining SI constants, with rounding only for display.
This helps explain why the electronic properties of coatings and photocathodes matter. Energy alone, however, does not answer “will this window transmit the light?” We need the material’s actual spectral properties, thickness, surface and operating conditions. For FUV-MAMA, for example, the short-wavelength boundary depends on the MgF₂ window as well as the photocathode. STIS Handbook, §7.4.2.
What does UV reveal beyond a different brightness?
A hot stellar surface emits a substantial share of its light at short wavelengths. A UV survey therefore helps identify young, hot, massive stars and regions of recent star formation. This does not mean every young star is hot or that a UV map alone gives every star’s age. NASA: comparing visible and UV galaxy images.
Spectral lines provide other information. Atoms and ions absorb or emit at characteristic wavelengths. Gas between a star and a telescope can leave narrow dips in the star’s spectrum. In the IUE team’s spectrum of Zeta Cassiopeiae, narrow dark features trace interstellar absorption, while broader features come from the stellar atmosphere. We must establish where a feature arose, rather than calling every dip “a cold patch on the star.” NASA: IUE and the spectrum of Zeta Cassiopeiae.
Far ultraviolet contains the important hydrogen line Lyman-α, around 121.6 nm. Studying it requires suitable conditions along the atmospheric path, optics and a detector. The same system need not work at much shorter EUV wavelengths, which call for other coatings and designs. Line positions and shapes are analyzed alongside a gas model and instrument calibration. STIS Handbook: Lyman-α and FUV response.
Why can a glass mirror work where a glass window gets in the way?
Light does not have to pass through a mirror’s entire glass substrate. The substrate holds the required shape, while a thin surface coating reflects light. An entrance window has a different job: a photon must pass through its thickness. Saying “the telescope contains glass” therefore tells us nothing by itself about UV transmission.
Hubble’s mirrors use ULE, ultra-low-expansion glass, with a reflective aluminum layer and protective magnesium fluoride, MgF₂. The substrate maintains geometry, aluminum reflects, and the coating protects against oxidation while providing the required UV properties. Aluminum oxide can strongly absorb UV; simply leaving fresh metal unprotected is not equivalent to a finished space mirror. NASA: Hubble optics, NASA/JPL: protective fluorides.
Another coating can open shorter wavelengths and introduce a new problem. Al/LiF developments use lithium fluoride, suitable for shorter UV but sensitive to moisture. GSFC/JPL teams describe a thin additional MgF₂ layer that improves coating stability during ground preparation. This is an engineering trade-off, rather than a claim that every UV mirror uses the same stack. NASA/JPL: PVD/ALD technology, 2025.
Windows and filters must also be checked in the required band. FUV-MAMA’s MgF₂ window limits shorter-wavelength access near 115 nm, while the photocathode sets another part of the response. Ordinary window glass, quartz and a special crystal are not interchangeable throughout UV. Even a sample transparent to the eye can stop the astronomical photon we need. STIS Handbook, §7.4.2.
Where does light become a spectrum?
The telescope collects light and forms an image near the scientific instrument’s entrance. A slit selects a narrow region of that image. Collimating optics prepare the beam for a dispersive element. A diffraction grating has a regular structure: waves from different regions add with phases that depend on wavelength and direction. Different wavelengths emerge in different directions.
The spectrograph camera then forms an image on the detector. Here “camera” means optics that focus beams, rather than simply a household camera. After calibration, position along one detector direction becomes a wavelength coordinate. Along the long slit, spatial information can remain: spectra are recorded for several locations in an object at once. NASA: STIS imaging spectrograph.
- Hubble telescopeReflectors collect light and form an input image.
- Correction and slitSTIS corrects aberration; the slit selects part of the image.
- Collimating opticsPrepare the beam for spectral dispersion.
- Selected gratingSeparates wavelengths by direction.
- Camera opticsFocus dispersed light onto the detector plane.
- MAMA or CCDEvents or charge; calibration links position to wavelength.
STIS has its own corrective optics for Hubble’s primary-mirror error, a slit wheel, an optical-mode selection mechanism, camera optics and three detectors. Some modes use first-order gratings, others echelles with additional order separation. Imaging modes can select a mirror instead of spectral dispersion. The diagram’s general teaching sequence is therefore not an exact spatial ray trace for every STIS mode. STIS Handbook, §3.2.2, fig.3.1.
A narrow input line does not guarantee that we can resolve it. Narrow radiation at the entrance becomes a finite-width spot at the output. Optics, slit, source size and detector all affect that width. Smaller pixels alone cannot repair a broad optical profile.
- R=1000 · FWHM=0.15 nm
- R=10000 · FWHM=0.015 nm
Branch · With equationsA synthetic comparison; this is not a specification for a particular STIS or COS mode.
How R differs from nanometers per pixel
Spectral resolving power is often written as:
R = (λ_(0))/(Δλ).
We must define Δλ. Here it is FWHM, the full width at half maximum of the response to one unresolved narrow line. At λ₀=150 nm, teaching values R=1000 and 10000 give Δλ=0.15 nm and 0.015 nm. This is the instrumental profile width, rather than the chart’s sampling interval.
The chart specifies two infinitely narrow, equal lines at 149.975 and 150.025 nm. Each has integrated weight 1; convolution with a normalized Gaussian profile preserves its area. Both curves contain the same amount of arbitrary signal, although the narrower one has taller peaks. At lower R the lines blend into one maximum; at higher R two are visible. The axis shows spectral density of arbitrary weight per nanometer, rather than measured counts or stellar flux.
The actual STIS profile depends on configuration; its handbook lists FWHM in pixels for different slits and sources. COS tables also define R through FWHM but explicitly state that the real LSF is not Gaussian. Our symmetric curves cannot stand in for any instrument’s accuracy while ignoring wings, noise and calibration. STIS: first-order LSFs, COS Handbook, table5.1.
How does MAMA count individual photons?
A photon that passes through the chosen optics reaches a photocathode, a sensitive layer that can release an electron after absorbing light. Not every incoming photon is registered: the probability depends on wavelength and design. The photoelectron enters a microchannel plate, MCP, containing many tiny channels. An electric field accelerates electrons, and wall collisions produce secondary electrons. A small input becomes an electron pulse.
Behind the plate, a multi-anode structure and electronics determine the event’s position. Hence MAMA, Multi-Anode Microchannel Array. Accumulation mode adds events into an image; time-tag mode can retain position and time for each registered event. The distribution of positions after a spectrograph allows the spectrum to be reconstructed. MCP gain itself does not measure wavelength from pulse height. STIS Handbook, §7.4.1.
hν · Photon → e⁻ · Photoelectron → Electron multiplication → Position + count.
STIS’s two MAMAs have an important construction difference. In FUV-MAMA, the opaque CsI, cesium-iodide photocathode is deposited directly on the plate’s entrance face. A photon passes through a separate MgF₂ window and is absorbed near the channel entrances. In NUV-MAMA, semi-transparent Cs₂Te, cesium telluride, sits on the window’s inner surface; electrons then cross a gap to the MCP. These designs have curved channels, rather than a pair of identical flat plates compulsory in every instrument. STIS Handbook, figs.7.14–7.15.
Such a detector needs controlled brightness. Excessive flux makes events count incorrectly, and still stronger illumination can damage the sensitive system. A faint target and an arbitrarily bright target are therefore not equivalent observing cases: the configuration’s official brightness screening is required. STIS Handbook, §3.2.1.
Must a UV detector be a MAMA?
No. A CCD, a charge-coupled device, can also work in part of the ultraviolet. Light creates charge carriers in silicon; pixels accumulate charge during an exposure, after which it is transferred and read out. Properties near the surface matter: the desired light must reach the sensitive region, and the resulting charge must be collected instead of lost at the surface.
STIS has a thinned, backside-illuminated CCD. Special backside treatment and an antireflection coating provide near-UV sensitivity. Its nominal range in the current table is 164–1030 nm, so calling every CCD a visible-only camera is wrong. Yet this range does not make that particular CCD a universal receiver across UV10–400 nm. STIS Handbook, §7.2.1, table7.1.
For another solid-state example, the JPL team describes silicon CCDs with specially processed back surfaces and deposited multilayer UV filters. The filter transmits the desired band while suppressing visible light that would otherwise add background. Choosing silicon alone is again insufficient: surface, layers and the entire optical path matter. NASA/JPL: UV sensor coatings.
How a UV chain is made, assembled and checked
Start with requirements: band, resolution, flux and target type. A mirror needs a stable substrate, grinding and polishing, and checks of surface figure and roughness. Controlled thin reflective and protective layers follow. Documented GSFC/JPL developments combine physical vapor deposition with atomic layer deposition; thickness and layer order affect the spectral result. Reflectance is assessed at operating wavelengths, beyond visible shininess. NASA/JPL: making UV coatings.
The detector is assembled as a separate unit: window, photocathode in the correct position, MCP, anodes and electronics, or a prepared solid-state chip with its window and readout. Integration checks the relative positions of optics, slit, grating and sensitive plane. The figure’s section explains component functions; it is not a dimensional specification for making a flight MAMA.
Cleanliness is part of optical performance. Particles scatter light, while thin molecular films can absorb UV even when a surface looks acceptable. A 1997 NASA/GSFC report on returned WF/PC-1 hardware described organic contamination of a pickoff mirror. Cleaning restored its UV reflectance: contamination, rather than destroyed coating, caused the problem. The proposed photopolymerization mechanism is explicitly presented as the authors’ hypothesis. This is a historical WF/PC-1 result, rather than our experiment or a measurement of present-day STIS. NASA-CP-3361, printedp.83 / PDF91.
Assembly therefore controls cleanliness and volatile materials, and sensitivity is checked again after integration. STIS even uses a warm CCD dewar window to reduce contaminant buildup. This does not put every detector component at the same temperature. STIS Handbook, §7.2.1.
Acceptance also needs calibration. Line lamps with known spectral features link position to wavelength; uniform illumination helps measure spatial response variations. Dark signal, background, linearity and sensitivity changes over time are checked. STIS has separate lamp subsystems for these tasks, and corresponding calibration corrections are applied to its data. STIS Handbook, §3.2.2, §13.2: sensitivity and line profiles.
Branch · Numerical exampleA hypothetical photon budget at one wavelength, not a measurement of STIS end-to-end sensitivity.
Why small losses at each stage accumulate
Imagine four reflections with equal reflectivity r, a grating efficiency of 0.5, a window transmission of 0.9 and a detector with registration probability 0.2 for each photon reaching it. In this example an event is one registered photon; additional CCD electron yield is not modeled.
p_(det) = (r)^(4) × 0.5 × 0.9 × 0.2.
At r=0.8, around 3.69% of incoming photons become counts; at r=0.7, 2.16% do. For 10000 photons at the entrance, these are expected values of 368.64 and 216.09 events, rather than fractional photons in one actual measurement. Every coefficient is a teaching assumption; the reflection count is not attributed to the real STIS.
This shows why cleanliness and suitable coatings matter along the entire path. An electrical pulse can be amplified after a photon is lost, but that cannot return a photon absorbed by a window. An actual end-to-end budget also includes wavelength dependence, slit losses, background, noise and temporal changes. STIS Handbook: throughput definition and sensitivity-table limits.
Documentation and teaching calculations were checked here. No UV telescope, vacuum detector or coating was physically made or tested for this article.
What must change for UV to become data?
An ultraviolet telescope begins with the desired scientific signature, rather than one identical set of parts for every wavelength. A band inaccessible from the ground needs a suitable location. Each subsequent stage must transmit or reflect the desired light without adding unwanted background. The detector converts light into events or accumulated charge; the spectrograph ahead of it links position to wavelength.
Try explaining: will a more sensitive detector help if the entrance window has absorbed the required band? It cannot return the lost photons. Does a glass mirror substrate mean light must pass through that glass? No: the reflective layer performs a different function. Does a smaller pixel always produce a narrower spectral line? No: an appropriate optical profile and calibration are needed.
For more on mirror shape and image field, read the article on Cassegrain, Gregorian and Ritchey–Chrétien telescopes. Ultraviolet adds requirements for atmosphere, materials, cleanliness and electronic detection to that geometry.