Single-Shot Electro-Optic Sampling

• Single-Shot Electro-Optic Sampling is a technique that captures full time-domain electric field waveforms in a single probe event by encoding the temporal information into spatial, spectral, or angular coordinates.
• It leverages the Pockels effect where ultrafast electric fields induce birefringence in non-centrosymmetric crystals, enabling rapid diagnostics across accelerator, plasma, and spectroscopy applications.
• Multiple encoding architectures—such as spatial decoding, spectral decoding, and photonic time-stretch—offer trade-offs in resolution, bandwidth, and sensitivity tailored for various ultrafast measurement scenarios.

Single-shot electro-optic sampling (EOS) denotes electro-optic field-measurement schemes in which the temporal information of an ultrafast electric transient is acquired without a mechanical delay scan, typically by encoding the EOS delay into a spatial, angular, or spectral coordinate of an optical probe (Szwaj et al., 2016, Roussel et al., 2020, Wu et al., 16 Sep 2025). In practice, the measured transient may be a free-propagating THz pulse, a relativistic bunch near-field, coherent transition radiation, or the electric field associated with fast electrons in laser-solid interactions; the defining feature is that one probe event yields a complete time-domain record, or at minimum a complete waveform coordinate, for that shot (Funkner et al., 2018, Steffen et al., 2019, Dastrup et al., 2023). The field has accordingly developed along several architectural lines—spectral decoding, spatial decoding, echelon-based sampling, photonic time-stretch, angular encoding, and newer vector-resolved or quantum extensions—while preserving the basic EO logic of field-induced birefringence and polarization-sensitive optical readout (2206.13572, Lafreniere-Greig et al., 12 Jun 2026).

1. Physical basis

The common physical basis of single-shot EOS is the Pockels effect. A transient electric field modifies the birefringence of a non-centrosymmetric crystal, and a synchronized optical probe converts that birefringence into a measurable polarization change. In a GaP implementation for free-propagating THz fields, the induced phase retardation is written as

$$\Delta\phi(t)=\frac{2\pi d}{\lambda}n_0^3r_{41}E_{THz}(t),$$

with $d$ the crystal thickness, $\lambda$ the probe wavelength, $n_0$ the refractive index, and $r_{41}$ the EO coefficient (Szwaj et al., 2016). Reported single-shot implementations span several crystal and wavelength choices, including 5 mm GaP at 1030 nm, 500 $\mu$m ZnTe at 800 nm, 0.2 mm ZnTe in a THz-FEL spectrally encoded geometry, 150 $\mu$m GaSe for broadband mid-IR to THz dual-comb EOS, and 1 mm CdTe for vector-resolved EOS at 1024 nm (Szwaj et al., 2016, Bisesto et al., 2018, 2206.13572, Konnov et al., 2023, Lafreniere-Greig et al., 12 Jun 2026).

The readout chain is usually balanced polarimetry. In the THz-FEL work, the analyzer relations are stated explicitly: $$I^{\text{cross}}_{\text{det}}(\Gamma)= \frac{I_0}{2}(1-\cos\Gamma)\approx \frac{I_0}{4}\Gamma^2,$$ for crossed polarizers, and

$$I^{\text{bal}}_{\text{det}}(\Gamma)=I_0\sin\Gamma\approx I_0\Gamma,$$

for balanced detection in the small-signal limit (2206.13572). The former offers high sensitivity but loses the sign of the field; the latter preserves field sign and linearity.

In ultrafast quantum-field EOS, the relevant limit is explicitly subcycle: “the subcycle structure of the THz field can be probed when $\Omega\,\Delta t < 1/2$” and both fields overlap within a small spacetime volume (Virally et al., 2021). That statement makes precise a general point that also applies classically: the single-shot architecture determines how the gate duration, phase matching, and encoding optics define the actual sampled observable.

2. Encoding architectures

Single-shot EOS has diversified into several distinct encoding schemes. Some map time to wavelength, some to transverse position, and some to detector position after angular dispersion or temporal magnification. The unifying statement, made explicitly in the reflection-grating work, is that single-shot EOS is generally achieved by encoding the EOS time delay into “a spatial, angular, or frequency coordinate” (Wu et al., 16 Sep 2025).

Architecture	Encoded coordinate	Representative characteristics
Spatial decoding	Transverse position	$d$0 mm and $d$1 gave $d$2 ps and $d$3 fs resolution in ZnTe for laser-solid diagnostics (Bisesto et al., 2018)
Spectral encoding / spectral decoding	Optical wavelength	Used for CTR and accelerator bunch profiles; at KARA, 0.91 MHz streaming over 100,000 shots; at EuXFEL, up to 2.26 MHz (Funkner et al., 2018, Steffen et al., 2019)
Phase-diversity spectral decoding	Two complementary spectral channels	Complementary transfer-function zeros and maximal-ratio combining; about 1.5 THz over 20 ps in single-shot table-top THz recording (Roussel et al., 2020)
Photonic time-stretch EOS	Stretched electrical time trace	$d$4, 6 GHz oscilloscope bandwidth corresponding to 1.2 THz at the crystal; high-sensitivity variant demonstrated on CSR (Szwaj et al., 2016)
Echelon-based single-shot EOS	Echelon step / detector position	500 time points over $d$5 ps at 1 kHz in optical-pump/THz-probe spectroscopy under 0–9 T (Dastrup et al., 2023)
Angular encoding with reflection gratings	Grating-imaged detector coordinate	Reliable beyond 6 THz and agreement with scanning EOS to $d$6 THz using a commercial 100 lines/mm reflection grating (Wu et al., 16 Sep 2025)

These architectures are not interchangeable. Spectral decoding is compact and compatible with line-array spectrometers, but classically suffers a resolution-window tradeoff. Spatial decoding provides an intuitive time-to-space map but requires careful geometric interpretation when the signal is curved, as in oblique-incidence ZnTe measurements of fast-electron fields (Bisesto et al., 2018). Time-stretch EOS trades optical complexity for direct digitization at high repetition rate (Szwaj et al., 2016). Echelon and angular-encoding schemes shift the burden to imaging fidelity and overlap geometry rather than spectrometer resolution (Dastrup et al., 2023, Wu et al., 16 Sep 2025).

3. Resolution, bandwidth, sensitivity, and calibration

The main performance parameters of single-shot EOS are time window, temporal resolution, transfer-function flatness, and field sensitivity. In spatial decoding, the time window is set geometrically: $d$7 which in the SPARC_LAB implementation gave $d$8 ps with about 100 fs resolution (Bisesto et al., 2018). In angular encoding with a tilted pulse front, the total window is

$d$9

so the window increases with illuminated grating width and angular dispersion (Wu et al., 16 Sep 2025).

For spectrally encoded EOS, the classical resolution limit is stated as

$\lambda$0

where $\lambda$1 is the stretched probe duration and $\lambda$2 the transform-limited pulse duration (Roussel et al., 2020). The DEOS work reframed this as a transfer-function-null problem rather than an unavoidable blur, and derived complementary transfer functions

$\lambda$3

combined by maximal-ratio combining so that ideally $\lambda$4 (Roussel et al., 2020). This removed the spectral-decoding penalty after the crystal in the ideal model, shifting the bottleneck to probe bandwidth, crystal bandwidth, dispersion control, and detector sampling.

Photonic time-stretch EOS introduces a different bandwidth logic. Its temporal magnification is

$\lambda$5

and in the SOLEIL experiment $\lambda$6, so 6 GHz at the oscilloscope corresponded to 1.2 THz at the EO crystal (Szwaj et al., 2016). The transfer function is not flat, however; the same paper gives

$\lambda$7

with the familiar fading response of time-stretch systems (Szwaj et al., 2016). The practical consequence was a quasi-flat 0–300 GHz region for noise evaluation and an overall 3 dB system bandwidth of about 0.4 THz.

Sensitivity gains have come from both architecture and analyzer design. The high-detectivity time-stretch scheme inserted four ZnSe Brewster plates as a partial polarizer and predicted an enhancement

$\lambda$8

with $\lambda$9 the transmission of the attenuated polarization (Szwaj et al., 2016). The measured signal enhancement was about 4.8 and the SNR improvement about $n_0$0–$n_0$1, yielding a noise-equivalent input electric field below 1.25 V/cm RMS inside the crystal over 0–300 GHz, converted to $n_0$2 (Szwaj et al., 2016).

Calibration is architecture-specific. Spectral-encoding accelerator systems reported empirical pixel-to-time calibrations of $n_0$3 ps/pixel at KARA and 0.153 ps/pixel in the THz-FEL single-shot spectrally encoded setup (Funkner et al., 2018, 2206.13572). The reflection-grating study showed that the effective bandwidth is not reduced by angular-dispersion-induced group-delay dispersion itself; instead, the limiting factors are overlap at the EO focus, spatial chirp, and noncollinear sampling by off-axis beamlets (Wu et al., 16 Sep 2025).

4. Accelerator, plasma, and spectroscopy applications

Single-shot EOS has become a standard accelerator diagnostic because relativistic bunch fields and associated THz emission fluctuate shot to shot and often recur at MHz-class rates. At KARA, near-field spectral encoding with GaP and the KALYPSO line array produced a continuous 0.91 MHz stream of 100,000 consecutive measurements over 110.11 ms, sufficient to resolve the synchrotron oscillation at $n_0$4 kHz in low-$n_0$5 short-bunch operation (Funkner et al., 2018). At the European XFEL, a compact rack-mounted spectral-decoding system measured bunch profiles around 400 fs rms with arrival-time jitter of 35 fs rms, using a custom Yb-fiber laser synchronized to the 1300 MHz accelerator RF and KALYPSO readout up to 2.26 MHz (Steffen et al., 2019).

The SOLEIL implementation addressed a different accelerator need: pulse-by-pulse recording of coherent synchrotron radiation with high sensitivity. There the THz source was CSR from the AILES beamline, the laser repetition was phase-locked to the 104th harmonic of the storage-ring revolution frequency, and the high-detectivity time-stretch architecture improved the pulse-resolved SNR by about 6.5 while preserving true single-shot operation (Szwaj et al., 2016). A related THz-FEL application used spectrally encoded single-shot EOS in 0.2 mm ZnTe because the FEL/electron-beam arrival jitter was about 0.6 ps rms; the full waveform could then be captured in one synchronized probe shot, revealing energy-dependent pulse splitting that was blurred in multi-shot delay scans (2206.13572).

Outside storage rings and FELs, single-shot EOS has been used for laser-solid and plasma-relevant diagnostics. At SPARC_LAB, a 500 $n_0$6m ZnTe crystal placed 1 mm downstream of a metallic target measured the temporal electric-field profile of fast electrons escaping from FLAME-driven laser-solid interactions. The implementation used spatial decoding with $n_0$7 mm and $n_0$8, giving a 10 ps temporal window and about 100 fs resolution; a representative shot yielded 2.1 nC charge, 14 MeV mean energy, and about 500 fs duration (Bisesto et al., 2018).

In THz spectroscopy proper, single-shot EOS is valuable when the experiment already contains an additional scan variable. The optical-pump/THz-probe work in 0–9 T magnetic field used a reflective stair-step echelon to generate 500 delayed gate beamlets, so each shot contained 500 time points across a $n_0$9 ps window at 1 kHz. A full magnetic-field-dependent dataset was acquired in 10.3 h; the paper states that conventional detection would have required more than 11 days, and the measured noise was more than an order of magnitude lower than conventional scanning EOS for an equal number of total laser shots (Dastrup et al., 2023).

5. Quantum and vector-resolved extensions

Single-shot EOS has also expanded into quantum-optical and vector-field regimes, although these developments complicate the meaning of “single-shot.” A microscopic quantum theory of time-domain EOS showed that the measured variance contains a classical field-sampling term plus quantum contributions independent of the state of the probed THz field, arising from quantum susceptibilities and cascading processes (Kizmann et al., 2021). That theory is directly relevant to ultrashort-probe EOS because it clarifies that the sampled observable is a probe-defined mode and that some variance terms are detector-medium artifacts rather than signal.

A separate proposal for “enhanced EOS with quantum probes” replaced the classical probe by conditioned nonclassical probes derived from photon-number-entangled twin beams. For THz vacuum EOS, the relative differential noise metric $r_{41}$0 was predicted to be six times larger than with a coherent probe, and the conditioned-probe strategy also strengthened sensitivity to higher-order moments needed to distinguish, for example, vacuum from a few-photon cat state (Virally et al., 2021). The proposal remains repeated-shot and statistical rather than literal one-shot waveform capture.

The two-port EOS protocol for Gaussian quantum light made that statistical character explicit. It uses two spatially separated EOS channels, each with its own probe delay, and reconstructs a temporal coherence matrix $r_{41}$1 from the joint distribution $r_{41}$2. The paper’s main result is that the two-port cross-correlation cancels the NIR shot-noise and cascading terms that contaminate one-port variance measurements, enabling full multimode Gaussian-state reconstruction of ultrabroadband MIR pulses in theory (Yang et al., 2 Jun 2025). This is single-shot only at the level of ultrafast gating; the tomography itself remains ensemble based.

Vector-resolved EOS introduced a different extension. With a circularly polarized probe, a $r_{41}$3-cut zinc-blende crystal, and a polarization-sensitive camera, the full in-plane THz vector field can be reconstructed directly in the time domain from one camera exposure per delay, without sequential polarization analysis (Lafreniere-Greig et al., 12 Jun 2026). The method recovered $r_{41}$4 and $r_{41}$5 components simultaneously and reported a dynamic range of 58.7 dB per transverse component under a $r_{41}$6 wire-grid-polarizer reference, $r_{41}$7 for polarization-angle validation, and angular sensitivity of $r_{41}$8 (Lafreniere-Greig et al., 12 Jun 2026). Crucially, the same paper states that “single-shot” here means single-acquisition vector retrieval at each delay; the full time-domain waveform still required a conventional delay scan in 66.7 fs steps.

6. Limitations, terminology, and outlook

Single-shot EOS remains constrained by alignment, calibration, and transfer-function fidelity. In high-sensitivity time-stretch EOS, polarization alignment, EO-crystal orientation, PM-fiber coupling, balanced-detector delay matching, and model-based voltage-to-field calibration all directly affect performance, and the quoted field is the field inside the crystal rather than necessarily the incident free-space field (Szwaj et al., 2016). In spatial-decoding laser-solid measurements, nonuniformities of the ZnTe surface and of the probe transverse profile were explicitly identified as image-quality limitations (Bisesto et al., 2018). In CTR spectral encoding at A0, the measured waveform was the CTR field after transport and crystal response, not a direct bunch-current profile, and the authors emphasized thick-crystal distortions, phase mismatch, and spectral-encoding artifacts (Maxwell et al., 2012).

A recurrent misconception is that all scan-free EOS should be classed as single-shot full-waveform capture. The literature in the supplied corpus draws sharper boundaries. Cryogenic in-situ EOS of superconducting circuits used pump-probe delay scans and asynchronous/equivalent-time sampling rather than any one-shot temporal encoding, so it is repeated-sampling EOS, not single-shot EOS (Priyadarshi et al., 2024). High-resolution dual-comb EOS likewise used scan-free asynchronous sampling of repeated events, with one interferogram every 0.0145 s at up to 69 Hz, but not one-pulse waveform capture (Konnov et al., 2023). “Photonic time stretch fieldoscopy” is even more distant: it is presented as an alternative field-resolved method inspired by time-stretch ideas but explicitly “not standard EOS” (Gommel et al., 3 Dec 2025).

The present outlook in the cited work follows three directions. First, sensitivity enhancement remains active: time-stretch EOS showed SNR gains of $r_{41}$9–$\mu$0 with available laser power and argued that higher amplifier output should move performance toward the theoretical enhancement factor $\mu$1 (Szwaj et al., 2016). Second, bandwidth extension continues through both decoding theory and optics: DEOS removed transfer-function nulls from spectral decoding in the ideal model (Roussel et al., 2020), while reflection-grating angular encoding demonstrated reliable operation beyond 6 THz and argued that angular-dispersion-induced GDD is not the bandwidth-limiting mechanism (Wu et al., 16 Sep 2025). Third, the scope of what EOS measures is widening, from scalar THz waveforms to in-plane vector fields and multimode quantum correlations (Lafreniere-Greig et al., 12 Jun 2026, Yang et al., 2 Jun 2025). This suggests that “single-shot EOS” is no longer a single technique but a family of EO sampling strategies unified by one requirement: preserving field-resolved information in an acquisition fast enough that the event need not be repeated for temporal reconstruction.