Laser-Sustained Plasma (LSP) Light Sources Explained

What a laser-sustained plasma is

A laser-sustained plasma (LSP), also called a continuous optical discharge, is a steady-state plasma kept hot entirely by absorbing a focused CW laser beam. The phenomenon was first demonstrated around 1970 by Yu. P. Raizer and coworkers using CO₂ lasers in high-pressure gases. It has since become the basis of commercial broadband lamps: a sealed fused-silica bulb filled with xenon, a near-infrared pump laser, and optics that collect the plasma's emission.

The architecture separates two jobs that a conventional arc lamp combines. The laser delivers the power, and the plasma converts it into broadband light. There are no electrodes in the hot zone, so there is no electrode erosion to set the lifetime or the position of the arc. The emitting volume is defined by the laser focus rather than by an electrode gap. Both features matter for the applications that use LSP sources.

Basic layout of an LSP lamp. Real designs differ in bulb shape, collection geometry (ellipsoidal reflectors are common), and how they filter out the pump wavelength. Every design needs some way to ignite the plasma, dispose of unabsorbed pump light, and collect emission over a large solid angle.

Why brightness, not power, is the figure of merit

Optical inspection and metrology tools collect light through small fields at modest numerical aperture. The geometric quantity that limits how much light they can accept is the étendue,

\[ G = n^2 A\,\Omega \approx A\,\pi\,\mathrm{NA}^2 , \]

where \(A\) is the illuminated field and \(\Omega\) the projected solid angle of the pupil. For a 0.1 mm field at NA 0.1, \(G \approx 2.5 \times 10^{-4}\) mm²·sr, a very small number. A passive optical system conserves radiance (power per unit area per unit solid angle). The most power a source can deliver into such a system is therefore its radiance multiplied by the system étendue, \(\Phi_{max} \approx L\,G\), however much total power the lamp emits.

So a source that is twice as powerful but has twice the emitting area delivers no more light to the wafer. What matters is radiance, and in particular spectral radiance at the wavelengths the tool uses. An LSP concentrates its emission into a region a few hundred micrometers across at a temperature of roughly 10,000–20,000 K. Its spectral radiance in the ultraviolet can far exceed that of electrode arc lamps with similar input power. That is the core reason LSP sources became the standard illuminator for broadband inspection and metrology. The blackbody calculator shows how strongly UV radiance depends on temperature.

How the laser sustains the plasma

Ignition

A CW laser cannot ionize cold gas on its own. A hundred watts focused to a 50 µm spot is about 10⁷ W/cm² at peak. Optical breakdown of a neutral gas needs intensities orders of magnitude higher, normally reached only with focused pulsed lasers. LSP lamps therefore ignite the plasma by an auxiliary method, usually a high-voltage discharge between electrodes in the bulb, or sometimes a pulsed laser. Once enough free electrons and hot gas exist at the focus, the CW beam takes over.

Absorption: inverse bremsstrahlung and photoionization

In a hot, ionized gas, free electrons absorb laser photons during collisions with ions (and, to a lesser extent, neutral atoms). This is inverse bremsstrahlung. In the classical approximation the absorption coefficient scales as

\[ \kappa_{IB} \;\propto\; \frac{n_e n_i}{\sqrt{T}}\,\lambda^3\left[1 - e^{-hc/\lambda k_B T}\right] \;\xrightarrow{\;hc/\lambda \ll k_B T\;}\; \frac{n_e n_i\,\lambda^2}{T^{3/2}} . \]

Two things follow. First, absorption grows with the square of the charged-particle density, and therefore with gas pressure. Second, it grows rapidly with wavelength. Free–free absorption alone would favor CO₂-laser light at 10.6 µm by more than a hundred times over 1.07 µm light. At 1 µm, however, photoionization of thermally excited xenon (bound–free absorption) contributes about as much as inverse bremsstrahlung at around 10⁴ K, so in practice CO₂ light is absorbed roughly 50–100 times more strongly in the same plasma. That is why the early optical discharges used CO₂ lasers. Modern lamps use near-infrared diode or fiber lasers instead: they are compact and reliable, and their wavelengths pass through fused silica, which is opaque at 10.6 µm. High fill pressure makes up for the weaker absorption at 1 µm.

The energy balance

The plasma settles where the laser power it absorbs balances what it loses through radiation (the useful output), thermal conduction, and convection. Because the plasma absorbs on the side facing the incoming beam, it tends to sit slightly upstream of the geometric focus, where the converging beam still provides enough intensity to sustain it. Below a minimum maintenance power the balance cannot be met and the plasma goes out. That threshold falls steeply with gas pressure and is much lower in heavy, easily ionized gases than in air. Not all of the pump power is absorbed. The transmitted remainder must be dumped or recycled, and it has to be kept out of the collected light.

Why high-pressure xenon

It is inert. Xenon does not react with the fused-silica envelope or the electrodes, which allows long sealed-bulb lifetimes.
It ionizes easily for a noble gas. Its first ionization energy, 12.13 eV, is the lowest of the stable noble gases. That gives high electron densities, and therefore strong absorption, at a given temperature.
It is heavy. Low thermal conductivity reduces conduction losses from the hot core and lowers the maintenance power.
It emits a dense continuum. Recombination and bremsstrahlung produce a broad continuum through the UV and visible. Xenon also adds strong neutral-atom lines between roughly 800 and 1000 nm.
Pressure multiplies everything. Fill pressures from several to tens of atmospheres, which rise further in operation, increase absorption, emissivity, and radiance. They also make the bulb a pressure vessel, and it must be engineered and enclosed as one.

The spectrum compared with a blackbody

For a thermal plasma in local thermodynamic equilibrium, Kirchhoff's law sets an upper bound. Spectral radiance cannot exceed the Planck function at the plasma temperature, \(L_\lambda = \varepsilon_\lambda B_\lambda(T)\) with \(\varepsilon_\lambda = 1 - e^{-\tau_\lambda} \le 1\). Where the plasma is optically thick (\(\tau_\lambda \gg 1\)) it radiates close to a blackbody at its temperature. Where it is optically thin it radiates less. The temperature dependence of the Planck function shows why a 10⁴ K source is so valuable in the UV.

Planck spectral radiance normalized to the 15,000 K peak, which falls at 193 nm. At 250 nm a 15,000 K blackbody is about 320 times brighter than one at 6,000 K (roughly the solar surface). At 500 nm the ratio is about 21. These curves are the upper bounds that a real plasma approaches where it is optically thick.

The emitted spectrum is shaped by several factors that a blackbody model leaves out:

Optical depth varies with wavelength. The hot core may be thick at some wavelengths and thin at others, so the spectrum is not a single Planck curve. Its brightness temperature changes across the band.
Cooler gas surrounds the core. The cooler xenon around the plasma reabsorbs at strong line positions, which can produce self-reversed line profiles.
The envelope and path limit the UV. Below about 200 nm, transmission depends strongly on the fused-silica grade and wall thickness. Molecular oxygen absorbs strongly there too (the Schumann–Runge bands and continuum), so the beam path needs a nitrogen purge or vacuum. Below about 242 nm oxygen photodissociates and forms ozone, which affects both safety and optics contamination.
Pump light leaks into the output. Scattered and transmitted laser light at the pump wavelength is far brighter than the plasma continuum at that wavelength. Dichroic optics or notch filters have to remove it.

Stability: where the engineering effort goes

For inspection, a source that is bright but fluctuates is a source of false defects. Noise and drift in an LSP source come from a few identifiable places:

Convection. The hot plasma drives a buoyant plume and recirculating flow inside the bulb. Unsteady flow moves and reshapes the plasma, causing low-frequency intensity noise and wander of its position. Bulb geometry and orientation are design variables for exactly this reason.
The pump laser. Power fluctuations and pointing jitter couple directly into plasma temperature and position. Back-reflections into the laser have to be managed.
Position sensitivity of the collection optics. Collection optics image a sub-millimeter source onto a field stop or fiber. A plasma shift of a few tens of micrometers can change coupled power far more than it changes total emission, so position control is often more important than power control.
Aging. Deposits and devitrification of the bulb wall slowly reduce UV transmission. Where ignition electrodes are used, their condition affects ignition reliability even though they carry no current in steady operation.

Mitigations include active laser-power stabilization, imaging-based feedback on plasma position, bulb and flow design, and monitoring the delivered optical output rather than the electrical or laser input.

Characterizing an LSP source

Development and qualification of these sources depends on careful photometry. The core measurements are:

Spectral radiance

Measure absolute spectral radiance by comparison with a calibrated transfer standard: typically a deuterium lamp in the UV, and a tungsten-halogen standard in the visible and near-IR. Use identical collection geometry, with the same field stop, aperture, and spectrometer. Report the results through a defined étendue, because that is how the downstream tool will see the source. The grating & spectrometer calculator helps match spectrometer bandpass and resolution to the measurement.

Brightness temperature

Inverting the Planck function for a measured spectral radiance gives a brightness temperature,

\[ T_b(\lambda) = \frac{hc}{\lambda k_B}\left[\ln\!\left(1 + \frac{2hc^2}{\lambda^5 L_\lambda}\right)\right]^{-1}, \]

which is a lower bound on the plasma temperature. It equals the true temperature where the plasma is optically thick. For optically thin lines, Boltzmann plots of line intensities give an excitation temperature, assuming local thermodynamic equilibrium.

Electron density

Stark broadening of isolated lines scales with electron density and is the standard spectroscopic estimate of \(n_e\). Separating it from Doppler and instrumental broadening requires a spectrometer whose resolution is well characterized. The line broadening calculator estimates the Doppler and natural contributions.

Imaging

Image the plasma through neutral-density and bandpass filters onto a camera with known magnification. This gives size, shape, and centroid position over time. Imaging at several wavelengths shows that UV emission comes from a smaller, hotter core than visible emission. Time series of centroid position, together with a laser power monitor, separate convective wander from pump-driven fluctuations.

Noise and drift

Record the output with fast photodiodes behind the same field stop the application uses. Analyze relative intensity noise spectra for short-term behavior and Allan deviation for drift. Correlating these with the position and laser-power channels identifies root causes, much as it would in any production SPC problem. The photon flux calculator converts the measured power into photon rates for shot-noise and detector budgets.

Where LSP sources are used

The combination of high UV radiance, broad spectral coverage, and good stability suits work where a single broadband illuminator must serve many wavelengths through small étendue:

Optical inspection. Broadband brightfield inspection of patterned wafers, and some mask-inspection applications. Short wavelengths improve resolution and contrast for small defects, and spectral flexibility helps optimize contrast for each layer.
Optical metrology. Spectroscopic ellipsometry, reflectometry, and scatterometry for film thickness and critical dimensions. These need broad, calibrated spectra with high radiance in the UV.
Laboratory instruments. UV–visible spectroscopy, microscopy, and other uses that need a stable, bright continuum source in place of deuterium plus tungsten lamp pairs or electrode arc lamps.

Development continues on the same physics problems: extending usable output to shorter wavelengths, raising radiance without giving up stability, and improving lifetime.

Key takeaways

An LSP is a plasma kept hot by absorbing a focused CW laser, through inverse bremsstrahlung and, at 1 µm, photoionization of excited atoms. It needs separate ignition, and it sits where absorbed power balances its losses.
High-pressure xenon and near-IR pumping trade weaker 1 µm absorption for compact lasers and fused-silica compatibility.
Radiance, not total power, determines how much light reaches an étendue-limited inspection or metrology tool. A small, roughly 10⁴ K plasma excels in the UV.
Real spectra fall below the Planck limit where the plasma is optically thin, and the envelope, the beam path, and pump-light filtering shape them further.
Characterization relies on calibrated spectral radiance, plasma imaging, noise and drift analysis, and spectroscopic temperature and density diagnostics.

References

Yu. P. Raizer, Laser-Induced Discharge Phenomena, Consultants Bureau, New York (1977).
Yu. P. Raizer, Gas Discharge Physics, Springer (1991).
H. R. Griem, Principles of Plasma Spectroscopy, Cambridge University Press (1997).