The Laser-Produced Tin-Plasma Source

Abstract

The EUV source is an industrial plasma experiment repeated tens of thousands of times per second. A generator launches microscopic molten-tin droplets through vacuum. A first carbon-dioxide laser pulse reshapes each droplet, and a main pulse creates a hot plasma whose emission includes a narrow band around 13.5 nanometers. A collector sends part of that radiation toward the scanner while debris and out-of-band energy must be controlled. I treat the source as an energy-conversion and availability system rather than a peak-power demonstration. Its useful output is stable in-band power at the intermediate focus over manufacturing time, with acceptable dose variation, collector life, tin use, gas load, heat, maintenance, and recovery. The source's conversion efficiency is necessarily low because electrical energy passes through laser generation, beam transport, target coupling, plasma radiation, angular collection, spectral acceptance, and optical transmission. Every improvement must be evaluated against the new burden it places on plasma conditions and nearby optics.

Manufacture a star on schedule

EUV radiation at 13.5 nanometers is strongly absorbed by air and ordinary materials, so the source operates in vacuum and uses a plasma rather than a conventional transmissive lamp. ASML describes molten tin droplets around 25 micrometers in diameter traveling near 70 meters per second. Two laser pulses interact with each target, and the sequence repeats up to roughly 50,000 times per second.

The astonishing part is not that one droplet emits EUV. Laboratories can create hot plasma in many ways. The industrial achievement is placing and coupling each target accurately enough that the average source supports patterned wafers for long periods.

The source chain is

$$P_{electric} ightarrow P_{laser} ightarrow P_{absorbed} ightarrow P_{plasma} ightarrow P_{13.5}\rightarrow P_{IF}.$$

I define wall-plug efficiency to the intermediate focus as

$$\eta_{wall\rightarrow IF}=\frac{P_{IF}}{P_{electric}}.$$

Each arrow contains a different loss and reliability mechanism. Improving plasma conversion while degrading collector lifetime can reduce annual photon delivery.

Tin is selected for its spectrum

Highly charged tin ions emit strongly near 13.5 nanometers through a dense set of transitions. The source seeks a plasma temperature and density that place useful emission inside the scanner’s spectral band while limiting kinetic debris and unwanted radiation.

The plasma is neither uniform nor stationary. Laser energy heats and expands the target. Ionization state changes in space and time. Radiation can be reabsorbed within the plasma. Too little coupling gives weak emission. Excess energy can produce fast ions, broad expansion, and contamination without proportional in-band output.

Conversion efficiency is commonly screened as

$$CE=\frac{E_{EUV,band}}{E_{laser}}.$$

The exact definition must state bandwidth, solid angle, and measurement plane. A source can improve total radiated EUV while failing to improve the collected in-band energy that reaches the scanner.

Source state	Desired outcome	Failure signature
Droplet flight	Repeatable position, size, and timing	Missed or off-center target
Pre-pulse	Controlled target expansion	Poor shape or unstable density
Main pulse	Efficient plasma formation	Under-coupling, overshoot, or misalignment
Plasma emission	Stable 13.5 nm band power	Spectral shift or temporal variation
Collection	High accepted radiance	Debris, contamination, heat, or misfocus

Pulse energy is only one source control

The laser system must deliver pulse energy, temporal shape, focus, pointing, polarization where relevant, and timing matched to a moving target. Beam path contamination or thermal distortion changes coupling. Amplifier chains, optics, cooling, and controls all contribute to availability.

The pre-pulse changes the target from a compact sphere into a geometry that the main pulse can couple to more effectively. Delay between pulses matters because the target continues to expand. Main-pulse focus and arrival must overlap a shape moving through three dimensions.

If repetition rate is $f$ and useful in-band energy per event at the intermediate focus is $E_{IF}$, average power is

$$P_{IF}=fE_{IF}.$$

At 50 kilohertz, one additional millijoule per event would correspond to 50 watts of average power. The simple scaling is attractive, but higher event energy or repetition can raise heat, debris, gas load, laser stress, and control demand.

LaunchGenerate molten-tin target ShapeApply controlled pre-pulse IonizeCouple the main laser pulse CollectAccept in-band radiation MeasureStabilize intermediate-focus power

Dose stability begins at the source

The scanner can compensate some source variation through dose control and exposure timing. It cannot turn arbitrary missing or unstable pulses into perfect imaging without throughput and uniformity consequences.

I separate source variation by time scale. Pulse-to-pulse variation affects local dose statistics and control. Field-scale variation can produce slit or scan nonuniformity. Wafer-scale drift can bias critical dimension. Longer degradation changes throughput, thermal state, and maintenance need.

Diagnostics should record droplet observation, laser timing and energy, plasma emission, spectral indicators, collector response, intermediate-focus power, gas conditions, and fault events with synchronized time. A single average power number erases the signatures needed to find cause.

Control authority includes droplet timing, laser timing, pulse energy, focus, beam steering, gas flow, target parameters, and scanner exposure response. Feedback must distinguish a true source change from sensor drift or collector degradation.

Debris is coupled to photon production

Not all tin becomes useful radiation. Neutral atoms, clusters, droplets, ions, and vapor can travel toward the collector or source vessel. Fast particles damage surfaces. Deposited tin changes reflectivity. Gas flows and magnetic fields can redirect some species, while source geometry and plasma conditions influence what is produced.

The best debris mitigation starts before the collector. Better laser-target coupling can reduce unvaporized remnants. Plasma shaping can change ion distribution. Hydrogen and other source-vessel controls can support cleaning or transport, but add vacuum and chemistry constraints.

TargetDroplet generator and tracking

Creates repeatable mass, trajectory, position, and timing.

DriverHigh-power CO2 laser chain

Shapes and ionizes the target with controlled spatial and temporal energy.

PlasmaIn-band photon conversion

Balances tin ionization, opacity, expansion, and debris production.

BoundaryCollector, gas, vacuum, and diagnostics

Extracts useful radiance while preserving optical and chamber lifetime.

Annual photons matter more than peak photons

A source demonstration often reports achieved power. A fab needs the integral of usable power during scheduled production:

$$E_{annual}=\int_{production}P_{IF}(t)\,dt.$$

The integral falls during planned maintenance, unplanned recovery, power derating, instability, collector degradation, and source faults. It should be weighted by whether the scanner can use the light at required dose stability.

Availability improvement can therefore equal a power upgrade. Faster droplet-generator exchange, predictable collector cleaning, modular laser service, better fault diagnosis, and controlled restart all raise annual patterned-wafer capacity.

Maintenance itself should be treated as a state transition with entry conditions, contamination controls, configuration records, verification, and post-work qualification. A source that returns with high power but altered stability or debris behavior is not fully restored.

The research frontier

I would pursue source improvement through synchronized physics and lifecycle evidence. Plasma experiments should report debris and collector consequence. Laser upgrades should report electrical efficiency and component life. New target shapes should report generator stability. Predictive models should be tested on withheld operating periods rather than explained after failure.

The source is already industrialized, which makes its remaining frontier more demanding. The question is no longer whether tin plasma can emit EUV. It is how to deliver more controlled photons per unit electrical energy, collector damage, service hour, and manufacturing interruption.

That ratio links a plasma lasting nanoseconds to a factory expected to operate for decades. Closing it is the source’s real mission.