Controlling Tin Droplets and Dual-Pulse Coupling

Abstract

An EUV source succeeds only when a moving tin target, pre-pulse, and main pulse occupy the same microscopic volume at the correct times. I treat target formation and coupling as a high-rate precision-control problem. The droplet generator must maintain diameter, velocity, spacing, trajectory, temperature, purity, and shape while operating near a harsh plasma environment. High-speed sensing predicts each droplet's intercept. The pre-pulse transforms the sphere into a larger target whose expansion depends on delivered energy and delay. The main carbon-dioxide laser pulse must overlap that target with appropriate intensity and temporal profile. Coupling error appears as lost EUV power, altered spectrum, residual tin, fast ions, or collector burden. I propose an event-level state vector and diagnostics chain that links every droplet to its laser telemetry, plasma emission, and debris signature. Average source power is not enough to optimize this interaction because different failure modes can share the same mean.

The target is manufactured in flight

The source does not aim at a solid object held in a fixture. It creates a stream of molten metal targets moving through vacuum. Each target must be present for microsecond-scale laser timing while previous plasma and debris have not fully vanished from the environment.

For a spherical droplet of diameter $d$, density $\rho$, and speed $v$, target mass and kinetic energy are

$$m=\rho\frac{\pi d^3}{6}, \qquad E_k=\frac{1}{2}mv^2.$$

Using $d=25\ \mu\mathrm{m}$, liquid-tin density near $7{,}000\ \mathrm{kg/m^3}$, and $v=70\ \mathrm{m/s}$ gives a mass near $5.7\times10^{-11}\ \mathrm{kg}$ and kinetic energy near $1.4\times10^{-7}\ \mathrm{J}$. The laser energy dominates the interaction, but droplet mass and position determine how that energy couples.

Diameter variation scales mass with $d^3$. A one-percent diameter change causes roughly a three-percent mass change before other effects. That sensitivity makes droplet control part of plasma control.

Jet breakup must become a clock

A droplet generator can use controlled jet breakup to form periodic droplets. Temperature sets viscosity and surface tension. Pressure and nozzle geometry set flow. Excitation controls breakup frequency and phase. Tin purity and wetting affect nozzle behavior. Vibration moves the trajectory.

At 50 kilohertz, droplets are emitted every 20 microseconds. A 70-meter-per-second stream travels 1.4 millimeters in that interval. The target train is orderly enough for timing and unforgiving enough that a small drift can miss the laser focus.

I define a droplet state

$$\mathbf{x}_d=[x,y,z,v_x,v_y,v_z,d,T,\epsilon_s,t_a],$$

where $\epsilon_s$ describes shape and $t_a$ is predicted arrival. Sensors estimate the state, and controls adjust generator excitation, pressure, temperature, timing, or beam steering.

Droplet variable	Source consequence	Diagnostic need
Position and velocity	Laser overlap and timing	Multi-point, high-speed trajectory sensing
Diameter and shape	Mass and coupling distribution	Calibrated shadowgraphy or scattering
Temperature	Viscosity, surface tension, breakup	Generator thermal model and measurement
Spacing	Event timing and collision risk	Phase-resolved stream observation
Purity and nozzle state	Stability, clogging, satellite droplets	Process control and maintenance evidence

The pre-pulse creates the main target

A compact sphere presents limited area to the main laser. The pre-pulse reshapes it into a disk, sheet, or cloud whose dimensions and density distribution improve coupling. The best shape is not necessarily the largest. It must remain dense enough for efficient absorption and produce favorable plasma and debris.

Expansion evolves during delay $\Delta t$ between pulses. A simple scale is

$$R(\Delta t)\approx R_0+u_e\Delta t,$$

where $u_e$ is an effective expansion speed. Real expansion is anisotropic and depends on pre-pulse wavelength, energy, spot, duration, and droplet state.

The main-pulse spot must overlap the expanded mass distribution. If the target is too small, laser light bypasses it. If it is too diffuse, absorption and plasma conditions shift. If the pulse arrives too early or late, usable emission falls and residue changes.

FormBreak the liquid jet PredictEstimate target intercept ShapeApply pre-pulse CoupleDeliver main pulse DiagnoseClassify photon and debris outcome

Timing error becomes spatial error

For a target moving at speed $v$, timing error $\delta t$ creates along-track error

$$\delta x=v\delta t.$$

At 70 meters per second, 100 nanoseconds corresponds to 7 micrometers, a significant fraction of the original droplet diameter. Beam pointing, focus position, optical delay, sensor latency, and clock synchronization must be understood on comparable scales.

Controls can predict the target and trigger the laser, steer the beam, or reject an event. Rejecting a bad event may protect the collector but creates a missing pulse. The scanner dose system then needs to know what happened. Event classification should propagate downstream instead of being hidden inside an average.

The coupling metric should include more than EUV energy:

$$J=w_EE_{13.5}-w_mM_{residual}-w_iE_{ion}-w_vV_{variation}.$$

The weights express source value and lifetime. A setting that maximizes photons while sharply increasing damaging debris may be inferior over a maintenance cycle.

Diagnostics must survive the source

Useful diagnostics include high-speed imaging, laser energy and wavefront measurement, timing telemetry, EUV photodiodes, spectroscopy, plasma imaging, ion and debris sensors, collector reflectivity indicators, and intermediate-focus power.

No instrument observes the whole interaction. Imaging may estimate target geometry and saturate during plasma. Spectroscopy observes emitted states but integrates along a path. Power sensors see useful output after collection but cannot uniquely identify cause. State estimation must fuse them with calibrated uncertainty.

BeforeDroplet and laser state

Records position, shape, timing, pulse energy, focus, and pointing.

DuringPlasma and spectral state

Observes emission timing, band content, geometry, and ion behavior.

AfterPhoton and debris outcome

Measures intermediate-focus energy, residual tin, and collector consequence.

Across timeGenerator and chamber health

Links event distributions to drift, contamination, and maintenance need.

Sampling only occasional events can miss rare modes that dominate contamination. Full-rate raw imaging may be impractical, so the system needs edge processing, feature extraction, triggered capture, and statistically designed archival. It should retain representative ordinary events and all anomalies.

Control drift before it becomes downtime

Nozzle erosion, deposits, tin chemistry, thermal state, pressure regulation, alignment, laser optics, and sensor calibration change slowly. Their signatures can appear as increased trajectory variance, satellite droplets, timing correction, missed-event rate, spectral shift, or debris accumulation.

Predictive maintenance must distinguish drift from harmless variability. A model trained only on normal production can flag novelty and still fail to estimate remaining life. End-of-life campaigns and controlled fault insertion provide more valuable labels.

The source should support a planned transition to a spare generator or service state before unstable droplets damage the collector. Exchange procedures must control tin temperature, vacuum contamination, alignment, calibration, and restart qualification.

Optimize the distribution, not one event

The industrial target is a narrow distribution of event outcomes sustained for billions of cycles. I want the source to report coupling efficiency, missed-event rate, variance by time scale, debris proxy, diagnostic confidence, and remaining maintenance margin.

A record-setting pulse demonstrates physics. A stable event population demonstrates manufacturing. The next frontier lies in making every droplet observable enough that source power, collector lifetime, and downtime can be optimized from the same evidence rather than traded through trial and surprise.