High-power photoconductive THz emitters for fast parallel electro-optic detection

arXiv:2610.00744 · physics.optics · Submitted 2026-09-30 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "High-power photoconductive THz emitters for fast parallel electro-optic detection".

Mira: Here's the gist of the paper. High-power photoconductive THz emitters enable fast parallel electro-optic detection by demonstrating a hyperspectral THz camerain driven by a high-average-power, pulsed broadband THz source <ref:2610.00744#pg2> This system simultaneously records the full THz waveform, in amplitude and phase, across 145,000 pixels over a 14.5 mm2 field of view at an acquisition rate of 2 Hz <ref:2610.00744#pg4>.

Kai: The terahertz (THz) region is highly attractive for imaging because it penetrates many optically opaque materials and can distinguish them by composition <ref:2610.00744#pg2> THz time-domain spectroscopy (THz-TDS) is uniquely suited to retrieve broadband information, recording the full waveform of the THz electric field to recover both amplitude and phase with high sensitivity <ref:2610.00744#pg2> Translating this capability into fast, wide-field imaging has however remained a major challenge, largely because the low average power of conventional TDS sources forces a trade-off between acquisition speed, field of view, and sensitivity <ref:2610.00744#pg2>.

Mira: The conventional path to achieve spatial resolution is raster-scanning the sample point by point and using a single detector <ref:2610.00744#pg2> This scheme makes it possible to recover the full spectroscopic content at every point but is inherently serial, with typical acquisition times from minutes to hours <ref:2610.00744#pg2>.

Kai: Source Development and Power Record The system utilizes a recordhigh average power photoconductive large area emitter (LAE) that delivers a THz average power of 38.3 mW at a 400 kHz repetition rate <ref:2610.00744#pg2> This source is driven by a commercial, industrially robust Yb-based femtosecond laser and achieves a THz average power of 38.3 mW, which is to the best of our knowledge the highest reported to date from any pulsed photoconductive THz emitter <ref:2610.00744#pg2>.

Paper summary: Kai: So we're looking at this paper, "High-power photoconductive THz emitters for fast parallel electro-optic detection." Basically, the authors are showing how you can get fast imaging in the terahertz range using a new kind of light source.

Mira: It claims they've built a system that records the full amplitude and phase of the terahertz wave simultaneously across thousands of pixels in a really small area with high speed. It addresses this slow, serial problem we usually have with time domain spectroscopy.

Lev: What matters here is that this isn't just another slow setup; they use a record-high average power photoconductive large area emitter, delivering thirty-eight point three milliwatts at a four hundred kHz repetition rate. That level of power is what enables the speed and field size they claim > Page two > <ref:2610.00744#pg1>

Kai: That record power comes from driving this emitter with a commercially available Yb-based femtosecond laser amplifier that gives them up to seventeen point five watts of average power at a central wavelength of one thousand thirty nanometers with a pulse duration of two hundred ninety five femtoseconds and a maximum repetition rate of four hundred kHz > Page three > <ref:2610.00744#pg2>

Mira: The core idea they are exploring is using this high average power source in a parallel electro-optic detection geometry. They use polarization modulation to turn the THz field into an intensity difference between two spatially separated images, which are then recorded by a fast camera.

Lev: From an engineering standpoint, that parallel setup requires a quarter wave plate and a wollaston prism to handle the polarization conversion > Page two > That means they're trying to get simultaneous spatial resolution instead of just one point at a time <ref:2610.00744#pg1>.

Kai: And the resulting system achieves a reconstruction of pixel resolved terahertz waveforms with about one point five terahertz bandwidth across fifteen point five square millimeters with a single trace acquisition time of five hundred milliseconds > Page two > That bandwidth is pretty significant for what they're doing here <ref:2610.00744#pg1>.

Mira: The spatial resolution they are getting depends on the propagation limited effect, which ranges from about one point three millimeters at zero point one terahertz down to about three hundred micrometers at two terahertz > Page two > That range shows how the resolution scales with the frequency of the signal <ref:2610.00744#pg1>.

Paper summary: Lev: They also mention a figure of merit for this work, which is defined as the number of spectroscopically resolved pixels acquired per second, which is N pixels over Tacq where Tacq is the time needed to record a complete terahertz time domain waveform at every spatial pixel > Page two > That metric really tells you how much data you get in a given time <ref:2610.00744#pg1>.

Kai: The validation section confirms that this parallel detection scheme lets them reconstruct the complete terahertz field E of x y tau from the imaging dataset, meaning the spatial shape evolves through the cycle instead of just scaling up and down with delay > Page four > That's a big step for understanding how these fields behave in space and time <ref:2610.00744#pg1>.

Mira: They also found that individual pixel waveforms clearly resolve the terahertz transient, and their average single pixel waveforms reproduce the pulse shape of the whole area mean, which confirms a uniform temporal response across the entire field of view > Page four > This suggests that the system is behaving consistently across its wide field <ref:2610.00744#pg1>.

Lev: But we have to consider a limitation here; they noted that the upper frequency limit for this whole area camera signal is about two terahertz, which they attribute to spatial averaging effects and lower dynamic range compared to conventional electro-optic measurements > Page two > That means at higher frequencies, you might not see as much detail <ref:2610.00744#pg1>.

Kai: And regarding resolution, they say the achievable spatial resolution is limited by diffraction within the five millimeter thick gallium phosphide crystal, suggesting that getting a ten micrometer pixel would need thinning that crystal down to a few micrometers at the upper end of their spectrum > Page two > That tells you exactly where the physical bottleneck is right now <ref:2610.00744#pg1>.

Mira: They suggest future work could involve reflection mode probing or spectral filtering of the optical probe before and after the electro-optic crystal to recover some of that dynamic range lost by using thinner crystals > Page two > That points toward a path for improving their performance further <ref:2610.00744#pg1>.

Paper summary: Lev: Before we move on, I want to make sure we've covered what they actually built. They used a specific setup involving off-axis parabolic mirrors with reflected focal lengths of one hundred point one six millimeters and fifty point zero eight millimeters separated by the sum of their focal lengths > Page three > That’s the physical layout for collecting that THz radiation <ref:2610.00744#pg1>.

Kai: So, to summarize this paper, they've demonstrated fast, wide field terahertz imaging where they record both amplitude and phase of the full terahertz waveform in parallel at every pixel using a record high power emitter > Page one > This is achieved by driving that emitter with a Yb-based laser amplifier providing thirty-eight point three milliwatts of average power at four hundred kHz repetition rate > Page two >

Mira: The authors are essentially showing how high average power is the enabling resource for practical parallel electro-optic imaging, eliminating the need for specialized kilohertz lasers that were previously required > Supplementary Information Table S1 summarizes these different approaches and shows that high THz average power is key to this kind of detection > Supplementary Information >

Lev: For me, what this means is that running this on real hardware would mean dealing with the overhead of getting those thirty-eight point three milliwatts sustained consistently across all pixels simultaneously, which will be tough for error correction if you try to use it for complex tasks > Page two >

Kai: That's where I get curious about the practical fidelity. They mentioned that their camera-based detection reproduces the same waveform shape and spectral distribution as conventional electro-optic measurements down to a single ten micrometer pixel > Reference Method >

Mira: And they also state that the usable per pixel bandwidth, about one point five terahertz for a single trace, is set by the higher per pixel noise floor rather than by spectral distortion of the detection system > Reference Method > That's an important caveat on what we can actually measure.

Lev: So, to wrap up this discussion on "High-power photoconductive THz emitters for fast parallel electro-optic detection," it shows that you can achieve fast, wide field terahertz imaging by using a high average power emitter delivering thirty eight point three milliwatts at four hundred kHz repetition rate > Page two > It's a demonstration of how to reconstruct complete single cycle terahertz waveforms across about one hundred forty-five thousand pixels spanning fifteen point five square millimeters in five hundred milliseconds > Conclusion <ref:2610.00744#pg1,High-power photoconductive THz emitters for fast parallel electro-optic detection>

Conclusion: Kai: So we're wrapping up this look at this paper, "High-power photoconductive THz emitters for fast parallel electro-optic detection." Basically, they’ve shown how you can record the full amplitude and phase of a terahertz wave simultaneously across many pixels using a really powerful light source.

Mira: It’s about taking that slow, serial time domain spectroscopy and making it fast by using high average power to enable this parallel setup.

Lev: The paper is all about that record-high average power emitter delivering thirty eight point three milliwatts at four hundred kHz repetition rate > Page two > That level of power is what makes the whole thing possible for the speed and field size they claim.

Kai: So, what does this mean for us? It means we can potentially get a much faster look at how things are moving in the terahertz range than before.

Mira: It changes how we think about imaging opaque materials because you’re not just getting one snapshot; you’re getting the entire time history of the wave across the whole field at once.

Lev: But running this on real hardware means dealing with sustaining that thirty eight point three milliwatts consistently across all those pixels, which is a big engineering challenge for error correction > Page two > That’s where the overhead really lands.

Kai: Exactly. It’s not just about getting the signal; it’s about how robust the whole system is when you try to run it on actual hardware.

Mira: And while they show this reconstruction of the complete field E of x y tau, we need to remember their limitation—the upper frequency limit is around two terahertz, and that’s due to spatial averaging effects > Page two > You can’t push past that with this setup right now.

Lev: So the next step is figuring out how we can beat those diffraction limits within the crystal or find a way to recover some of the dynamic range they lost when they used thinner crystals > Page two > That points toward future work.

Photonics and Ultrafast Laser Science, Ruhr-University Bochum · Centre for Advanced Electronics, Indian Institute of Technology Indore · Helmholtz-Zentrum Dresden-Rossendorf, Institute of Ion Beam Physics and Materials Research · Research Center Chemical Science and Sustainability, University Alliance Ruhr

physics.optics

Submitted: 2026-09-30

Updated: 2026-11-14

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 83/100

The gist: High-power photoconductive THz emitters enable fast parallel electro-optic detection by demonstrating a hyperspectral THz camerain driven by a high-average-power…

Key concepts

THz Time-Domain Spectroscopy (THz-TDS)
This technique records the entire waveform of a terahertz electric field. It is crucial because it captures both the amplitude and phase information, which is necessary to fully characterize the broadband THz signal, unlike simpler methods.
Parallel Electro-Optic Detection Geometry
This geometry uses a quarter-wave plate and a Wollaston prism to split the polarization modulation into intensity differences between two spatially separated beams. These beams are then imaged simultaneously on a fast camera, allowing for rapid spatial resolution of the THz signal.
FOM (Figure of Merit)
The FOM measures imaging throughput by calculating the number of spectroscopically resolved pixels acquired per second (Npixels/Tacq). A higher FOM indicates a faster and more comprehensive hyperspectral imaging capability, showing that high average power is key to this performance.

Terminology

Summary

High-power photoconductive THz emitters enable fast parallel electro-optic detection by demonstrating a hyperspectral THz camerain driven by a high-average-power, pulsed broadband THz source <ref:2610. p. 0

Where the paper says it

Here we demonstrate a hyperspectral THz camerain which a high-average-power, pulsed broadband THz source, delivering 38. 3 mW at a 400 kHz repetition rate, drives parallel electro-optic detection over a wide field of view at high acquisition rates.

00744#pg2> This system simultaneously records the full THz waveform, in amplitude and phase, across 145,000 pixels over a 14. p. 0
Where the paper says it

3 mW at a 400 kHz repetition rate, drives parallel electro-optic detection over a wide field of view at high acquisition rates. The system simultaneously records the full THz waveform, in amplitude and phase, across 145,000 pixels over a ~14.

5 mm2 field of view at an acquisition rate of 2 Hz <ref:2610. p. 0
Where the paper says it

Here we demonstrate a hyperspectral THz camerain which a high-average-power, pulsed broadband THz source, delivering 38. 3 mW at a 400 kHz repetition rate, drives parallel electro-optic detection over a wide field of view at high acquisition rates.

00744#pg4>.

Introduction to the Problem

The terahertz (THz) region is highly attractive for imaging because it penetrates many optically opaque materials and can distinguish them by composition <ref:2610.00744#pg2> THz time-domain spectroscopy (THz-TDS) is uniquely suited to retrieve broadband information, recording the full waveform of the THz electric field to recover both amplitude and phase with high sensitivity <ref:2610. p. 0

Where the paper says it

de The terahertz (THz) region is highly attractive for imaging because it penetrates many optically opaque materials and can distinguish them by composition. Among the various methods that allow us to access this spectral region, THz time-domain spectroscopy (THz-TDS) is uniquely suited to retrieve broadband information, recording the full waveform of the THz electric field to recover both amplitude and phase with high sensitivity.

00744#pg2> Translating this capability into fast, wide-field imaging has however remained a major challenge, largely because the low average power of conventional TDS sources forces a trade-off between acquisition speed, field of view, and sensitivity <ref:2610. p. 0
Where the paper says it

Among the various methods that allow us to access this spectral region, THz time-domain spectroscopy (THz-TDS) is uniquely suited to retrieve broadband information, recording the full waveform of the THz electric field to recover both amplitude and phase with high sensitivity. Translating this capability into fast, wide-field imaging has however remained a major challenge, largely because the low average power of conventional TDS sources forces a trade-off between acquisition speed, field of view, and sensitivity.

00744#pg2>. The conventional path to achieve spatial resolution is raster-scanning the sample point by point and using a single detector <ref:2610. p. 0
Where the paper says it

Extending these capabilities from spatially averaged spectroscopy to imaging is an active topic of research and many efforts are currently dedicated to spatially resolving the THz field at sufficient speed. The conventional path to achieve spatial resolution is raster-scanning the sample point by point and using a single detector, including near-field schemes to reach sub-diffraction resolution6,7.

00744#pg2> This scheme makes it possible to recover the full spectroscopic content at every point but is inherently serial, with typical acquisition times from minutes to hours <ref:2610. p. 0
Where the paper says it

The conventional path to achieve spatial resolution is raster-scanning the sample point by point and using a single detector, including near-field schemes to reach sub-diffraction resolution6,7. This scheme makes it possible to recover the full spectroscopic content at every point but is inherently serial, with typical acquisition times from minutes to hours.

00744#pg2>.

Source Development and Power Record

The system utilizes a recordhigh average power photoconductive large area emitter (LAE) that delivers a THz average power of 38.3 mW at a 400 kHz repetition rate <ref:2610. p. 0

Where the paper says it

Here we demonstrate a hyperspectral THz camerain which a high-average-power, pulsed broadband THz source, delivering 38. 3 mW at a 400 kHz repetition rate, drives parallel electro-optic detection over a wide field of view at high acquisition rates.

00744#pg2> This source is driven by a commercial, industrially robust Yb-based femtosecond laser and achieves a THz average power of 38. p. 1
Where the paper says it

Here we demonstrate wide-area, spatially and spectrally-resolved EO detection at high speed, thanks to a recordhigh average power photoconductive large-area THz emitter (LAE). The source is driven by a commercial, industrially robust Yb-based femtosecond laser and achieves a THz average power of 38.

3 mW, which is to the best of our knowledge the highest reported to date from any pulsed photoconductive THz emitter <ref:2610. p. 1
Where the paper says it

The source is driven by a commercial, industrially robust Yb-based femtosecond laser and achieves a THz average power of 38. 3 mW, which is to the best of our knowledge the highest reported to date from any pulsed photoconductive THz emitter.

00744#pg2>. The LAE is excited by a commercially available Yb-based laser amplifier delivering up to 17. p. 2
Where the paper says it

8 mm, separated by the sum of their focal lengths. The LAE is excited by a commercially available Yb-based laser amplifier delivering up to 17.

5 W of average power at a central wavelength of 1030 nm with a pulse duration of 295 fs and a maximum repetition rate of 400 kHz <ref:2610. p. 2
Where the paper says it

The LAE is excited by a commercially available Yb-based laser amplifier delivering up to 17. 5 W of average power at a central wavelength of 1030 nm with a pulse duration of 295 fs and a maximum repetition rate of 400 kHz (TruMicro2000, TRUMPF).

00744#pg2>.

Parallel Electro-Optic Detection Geometry

The parallel EO imaging geometry combines the high average power and high repetition rate of the source with parallel EO imaging for rapid, spatially resolved THz-TDS <ref:2610.00744#pg2> A quarter-wave plate (QWP) and a Wollaston prism (WP) convert the polarization modulation into an intensity difference between two spatially separated probe images with orthogonal polarizations that are imaged simultaneously on a fast camera <ref:2610. p. 4

Where the paper says it

1a, with a modified EOS detection arm as shown in Fig. In this implementation, a quarter-wave plate (QWP) and a Wollaston prism (WP) convert the polarization modulation into an intensity difference between two spatially separated probe images with orthogonal polarizations that are imaged simultaneously on a fast camera.

00744#pg2>. A 4f-imaging system with unit magnification, consisting of two identical lenses of focal length f = 150 mm, relayed the crystal plane onto the camera detector <ref:2610. p. 4
Where the paper says it

In this implementation, a quarter-wave plate (QWP) and a Wollaston prism (WP) convert the polarization modulation into an intensity difference between two spatially separated probe images with orthogonal polarizations that are imaged simultaneously on a fast camera. To preserve the spatial correspondence between the EO crystal plane and the sensor of the camera, a 4f-imaging system with unit magnification, consisting of two identical lenses of focal length f = 150 mm, relayed the crystal plane onto the camera detector.

00744#pg2>.

Performance Metrics and Spatial Resolution

The resulting system achieves a reconstruction of pixel-resolved THz waveforms with a 1.5 THz bandwidth across a 14. p. 1

Where the paper says it

Employing this source in a parallel EO imaging geometry, we reconstruct pixel-resolved THz waveforms with a ~1. 5 THz bandwidth across a 14.

5 mm2 circular field of view (4. p. 1
Where the paper says it

5 THz bandwidth across a 14. 5 mm² circular field of view (4.

3 mm equivalent diameter, 145,000 pixels at 10 µm pixel size) with a single-trace acquisition time of 500 ms <ref:2610. p. 1
Where the paper says it

5 mm² circular field of view (4. 3 mm equivalent diameter, ~145,000 pixels at 10 µm pixel size) with a single-trace acquisition time of 500 ms.

00744#pg2>. The estimated propagation-limited THz spatial resolution ranges from 1. p. 4
Where the paper says it

2b, and the resulting bandwidth is discussed below. The estimated propagation-limited THz spatial resolution ranges from ~1.

3 mm at 0.1 THz down to 300 µm at 2 THz <ref:2610. p. 4
Where the paper says it

The estimated propagation-limited THz spatial resolution ranges from ~1. 1 THz down to ~300 µm at 2 THz.

00744#pg2>. The FOM for this work is defined as the number of spectroscopically resolved pixels acquired per second, FOM = Npixels/Tacq, where Tacq is the time required to record a complete THz time-domain waveform at every spatial pixel <ref:2610. p. 9
Where the paper says it

Second, the inherent sparsity of the THz signal can be exploited: sparse or adaptive delay sampling reduces the number of temporal samples when only a restricted spectral range or a known waveform structure is needed, as recently demonstrated by compressed-sensing reconstruction of field-resolved traces beyond the Nyquist frequency40, while spatial compressive schemes reduce the number of points sampled across the field of view12 , together offering substantial reductions in both acquired data and processing load. Finally, synchronous detection schemes such as dynamic-subtraction or lock-in-type camera readout, synchronized to a modulation of the THz beam, are known to suppress technical noise in camera-based EO imaging25,41 and would narrow the remaining gap between the whole-area camera signal and conventional lock-in EOS, while also reducing the averaging, and hence the data, required to reach a given dynamic range.

00744#pg2>.

Validation and Spatiotemporal Reconstruction

The parallel detection scheme allows for the reconstruction of the complete THz field E(x, y,τ) from the imaging dataset <ref:2610.00744#pg2> The transverse field distribution therefore does not simply scale up and down with delay; its spatial shape evolves through the cycle, which is captured simultaneously across the field of view during the delay sweep <ref:2610. p. 6

Where the paper says it

4a shows snapshots of the transverse field at selected delays spanning the singlecycle THz transient (a corresponding video of the field evolution is provided in the supplementary material). The transverse field distribution therefore does not simply scale up and down with delay; its spatial shape evolves through the cycle, which is captured simultaneously across the field of view during the delay sweep.

p. 6
Where the paper says it

4a shows snapshots of the transverse field at selected delays spanning the singlecycle THz transient (a corresponding video of the field evolution is provided in the supplementary material). The transverse field distribution therefore does not simply scale up and down with delay; its spatial shape evolves through the cycle, which is captured simultaneously across the field of view during the delay sweep.

00744#pg2>. Individual pixel waveforms clearly resolve the THz transient, and the average single-pixel waveforms closely reproduce the pulse shape of the whole-area mean, confirming uniform temporal response across the field of view <ref:2610. p. 6
Where the paper says it

dNormalized THz waveforms: a single 10 µm camera pixel (single trace), its 100-trace average, and the spatial average over the whole aperture, compared with balanced-photodetector EOS measured using the same expanded probe and 5-mm GaP crystal. eCorresponding power spectra for the single pixel, single-pixel average, whole-area average, and EOS detection.

00744#pg2>.

Discussion on Limitations and Future Directions

The observed upper-frequency limit of 2 THz for the whole-area camera signal is consistent with spatial averaging effects and lower dynamic range compared to conventional EOS measurements <ref:2610.00744#pg2>. The achievable spatial resolution is limited by diffraction within the 5-mm-thick GaP crystal, which suggests that reaching a pixel-limited resolution of 10 µm would require thinning the crystal to a few micrometres at the upper end of our spectrum <ref:2610. p. 8

Where the paper says it

Reaching a pixel-limited resolution of 10 µm would require thinning the crystal to a few micrometres at the upper end of our spectrum (≈3 µm at 1 THz, ≈6 µm at 2 THz), and to sub-micrometre thickness at 0. Near-field EO imaging with micrometre-thick LiNbO₃ crystals has demonstrated 5–9 µm (λ/600) resolution at 100 GHz 16 , confirming that this regime is experimentally accessible.

00744#pg2>. Future improvements could include reflection-mode probing or spectral filtering of the optical probe before and after the EO crystal to recover part of the DR sacrificed by thinner crystals <ref:2610. p. 8
Where the paper says it

In future systems where higher spatial resolution may be desired, several additional measures can recover part of the DR sacrificed by thinner crystals. Reflection-mode probing with a high-reflection-coated crystal surface doubles the effective interaction length at fixed thickness37.

00744#pg2>.

Methods for Power Characterization

The THz average powers reported in the Results were measured by replacing the EO detection crystal with a calibrated pyroelectric THz power meter <ref:2610.00744#pg2> The calibrated THz power meter requires a modulated THz signal at 18 Hz with 50% duty cycle <ref:2610. p. 9

Where the paper says it

The routes to higher resolution, faster acquisition and broader bandwidth discussed above could extend high-average-power THz-TDS as a practical platform for high-speed, material-resolved inspection and for the spectroscopic study of irreversible, second-timescale processes such as hydration changes in biologically relevant systems. Methods Power measurements The THz average powers reported in the Results were measured by replacing the EO detection crystal with a calibrated pyroelectric THz power meter (SLT GmbH, THz20, calibrated by the German Metrology Institute PTB).

00744#pg2>. The measured signal dropped to the noise floor when either the optical pump or the electrical bias was switched off, confirming that the recorded THz power was free of detectable pump leakage and spurious electrical heating contributions from the emitter <ref:2610. p. 9
Where the paper says it

Their transmission (88% THz power transmission for each paper and 92% for transparent protective cover) was measured independently and considered in the reported power values, with a combined transmission factor of 0. When either the optical pump or the electrical bias was switched off, the measured signal dropped to the noise floor, confirming that the recorded THz power was

00744#pg2>.

Data Acquisition and Processing

The data acquisition, image registration, and signal processing involved calibrating image registration using a rigid transformation determined from a 1951 USAF resolution target placed at the crystal plane <ref:2610.00744#pg2> The corrected EO signal was then evaluated as I L − α I R (scalar) or I L − α(x, y)I R (pixel-wise) <ref:2610. p. 8

Where the paper says it

These effects, together with the lower DRof the whole-area camera signal (Methods), provide a plausible explanation for its upper-frequency limit of ~2 THz, compared with ~3 THz in Fig. 5 THz limit of a single-pixel measurement is also influenced by its lower signal-to-noise ratio.

00744#pg2>. This procedure cancels the spatial probe envelope, so that the recovered signal represents the THz-induced EO modulation <ref:2610. p. 9
Where the paper says it

Second, the inherent sparsity of the THz signal can be exploited: sparse or adaptive delay sampling reduces the number of temporal samples when only a restricted spectral range or a known waveform structure is needed, as recently demonstrated by compressed-sensing reconstruction of field-resolved traces beyond the Nyquist frequency40, while spatial compressive schemes reduce the number of points sampled across the field of view12 , together offering substantial reductions in both acquired data and processing load. Finally, synchronous detection schemes such as dynamic-subtraction or lock-in-type camera readout, synchronized to a modulation of the THz beam, are known to suppress technical noise in camera-based EO imaging25,41 and would narrow the remaining gap between the whole-area camera signal and conventional lock-in EOS, while also reducing the averaging, and hence the data, required to reach a given dynamic range.

00744#pg2>. The large data volume generated necessitates technical routes to address overhead, such as direct streaming to a binary format like HDF5 <ref:2610. p. 9
Where the paper says it

Several technical routes can address this overhead and accelerate the platform further. First, direct streaming to a binary format such as HDF5 could combine region-of-interest cropping, calibrated registration, balanced subtraction and real-time waveform accumulation.

00744#pg2>.

Conclusion

In conclusion, we have demonstrated fast, wide-field terahertz imaging in which the full THz waveform — amplitude and phase —is recorded in parallel at every pixel, enabled by a record-high average power photoconductive large-area emitter <ref:2610.00744#pg2>. Driven by a commercial, industrially robust Yb-based femtosecond laser, the source delivers a record-high 38. p. 9

Where the paper says it

In conclusion, we have demonstrated fast, wide-field terahertz imaging in which the full THz waveform — amplitude and phase —is recorded in parallel at every pixel, enabled by a record-high average power photoconductive large-area emitter. Driven by a commercial, industrially robust Yb-based femtosecond laser, the source delivers a record-high 38.

3 ± 0.2 mW of THz average power at a 400 kHz repetition rate, sufficient to feed a parallel electrooptic detection scheme across a wide field of view <ref:2610. p. 9
Where the paper says it

Driven by a commercial, industrially robust Yb-based femtosecond laser, the source delivers a record-high 38. 2 mW of THz average power at a 400 kHz repetition rate, sufficient to feed a parallel electrooptic detection scheme across a wide field of view.

00744#pg2>. With this approach we reconstruct complete single-cycle THz waveforms across 145,000 pixels spanning a 14. p. 9
Where the paper says it

2 mW of THz average power at a 400 kHz repetition rate, sufficient to feed a parallel electrooptic detection scheme across a wide field of view. With this approach we reconstruct complete single -cycle THz waveforms across ~145,000 pixels spanning a ~14.

5 mm2 field of view in 500 ms, corresponding to 2. p. 9
Where the paper says it

With this approach we reconstruct complete single -cycle THz waveforms across ~145,000 pixels spanning a ~14. 5 mm² field of view in 500 ms, corresponding to ~2.

9 × 105 spatially and spectrally resolved pixels per second at the 10-µm camera sampling pitch during active acquisition <ref:2610. p. 9
Where the paper says it

5 mm² field of view in 500 ms, corresponding to ~2. 9 × 10⁵ spatially and spectrally resolved pixels per second at the 10-µm camera sampling pitch during active acquisition.

00744#pg2>.

Supplementary Information

The performance of the state-of-the-art THz imaging systems discussed in the introduction is summarized in supplementary Table S1, which lists representative demonstrations of the different approaches together with their key performance parameters <ref:2610.00744#pg2>. The figure of merit (FOM) used in Table S1 quantifies hyperspectral imaging throughput as FOM = Npixels/Tacq, where Tacq is the time required to record a complete THz time-domain waveform at every spatial pixel <ref:2610. p. 6

Where the paper says it

4 reconstructs the complete THz field E(x, y,τ) from the imaging dataset. Because every camera pixel records a full temporal waveform simultaneously, the data recorded forms a three-dimensional field map that can be viewed either as a sequence of two-dimensional spatial snapshots at fixed delay or as spatiotemporal sections through the beam.

00744#pg2>. The results establish that high THz average power is the enabling resource for practical parallel electro-optic imaging, eliminating the need for specialized kHz-laser <ref:2610. p. 9
Where the paper says it

This rate counts sampled pixel waveforms and is not normalized to the independent spatial resolution. These results establish that high THz average power is the enabling resource for practical parallel electro-optic imaging, eliminating the need for specialized kHz-laser.

00744#pg2>.

Reference Method

The comparison between the camera and photodetector measurements confirms that camera-based detection, down to a single 10-µm pixel, reproduces the same waveform shape and spectral distribution as conventional EOS <ref:2610.00744#pg2>. The usable perpixel bandwidth (1. p. 5

Where the paper says it

The whole-area camera measurement reaches approximately 73 dB, compared with approximately 90 dB for balanced-photodetector EOS under the same imaging geometry. The waveforms and spectral features agree over their common measurable frequency range, while their usable bandwidths depend on the respective noise floors (Methods).

5 THz for a single trace) is instead set by the higher per-pixel noise floor, not by spectral distortion of the detection <ref:2610. p. 9
Where the paper says it

Data acquisition and acceleration Beyond speed and DR, a practical consequence of recording the full THz waveform in parallel at every camera pixel is the large data volume generated. Acquiring all frames at the full frame rate and bit depth, including the true noise floor via two dark measurements39, produced raw data on the order of 700 GB for a single experiment (56 s of averaging), creating bottlenecks in data transfer (here set by the camera-toPC Ethernet link), storage, and pixel-by-pixel post-processing.

00744#pg2>.

Spatiotemporal Reconstruction

Fig. 4 reconstructs the complete THz field E(x, y,τ) from the imaging dataset <ref:2610. p. 6

Where the paper says it

Spatiotemporal reconstruction of the THz field To illustrate the spatiotemporal capability of the parallel detection scheme, Fig. 4 reconstructs the complete THz field E(x, y,τ) from the imaging dataset.

00744#pg2> The transverse field distribution therefore does not simply scale up and down with delay; its spatial shape evolves through the cycle, which is captured simultaneously across the field of view during the delay sweep <ref:2610.00744#pg2>.

Bias-Dependent Saturation

The roll-off is governed by pulse energy alone, and not by the repetition rate, as shown in Fig. S5b <ref:2610.00744#pg2>. The operational onset of the rollover shifts to higher pump power as the acceleration field increases <ref:2610. p. 3

Where the paper says it

The bias was held at 84 kV/cm rather than at the 95 kV/cm used for the power record, as a conservative margin against breakdown over the long acquisition; at this bias the emitter delivers 34. The normalized waveform is already close to its high-excitation limit at this point, since both the pump power and the acceleration field lie above the values beyond which the normalized waveform no longer changes, as shown in Supplementary Note 3.

00744#pg2>. The practical consequence is that at a fixed average power the emitter is more efficient at a lower repetition rate, because the same energy is then delivered in fewer but stronger pulses <ref:2610. p. 2
Where the paper says it

The optimal conditions were reached using a unipolar pulse-width modulation (PWM) bias with 650 ns voltage pulses, corresponding to a duty cycle of 26%, synchronized to the 400 kHz laser repetition rate. Under these conditions, a record-high THz average power of 38.

00744#pg2>.

Optical Characterization of Pulses

The optical spectra and intensity autocorrelations of the fundamental and second-harmonic pulses were fitted independently by unweighted least squares to a Gaussian with free amplitude, center, FWHM and constant background <ref:2610.00744#pg2>. The pulse duration is the autocorrelation FWHM divided by √2 <ref:2610. p. 2

Where the paper says it

6 W of average power at 515 nm (photon energy 2. 41 eV), with a Gaussian-equivalent pulse duration of approximately 253 fs obtained from the intensityautocorrelation FWHM usingaGaussian deconvolution.

00744#pg2>.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Parallel electro-optic detection systems can be utilized for inline material-resolved inspection by simultaneously recording full THz waveforms in amplitude and phase across a 145,000 pixel field of view at an acquisition rate of 2 Hz.

  2. The system can achieve a throughput of approximately 2.9 x 10 5 sampled pixel waveforms per second at the camera's 10 µm sampling pitch during active acquisition, demonstrating high-speed spectroscopic imaging capability.

  3. High power average THz emitters, such as the record high average power photoconductive large area emitter delivering 38.3 mW, can drive parallel EO imaging to reconstruct pixel-resolved THz waveforms with a ~1.5 THz bandwidth across a 14.5 mm squared field of view in 500 ms.

  4. The system can map the spatiotemporal evolution of the THz field E(x, y, tau) by recording full temporal waveforms at every camera pixel, allowing for evaluation of the overall performance and understanding of limitations in future applications.

  5. AI systems can be trained on the reconstructed 3D field maps to analyze spatiotemporal structures such as how local waveform shapes evolve across the field of view during a single cycle, revealing spatial variations in THz arrival times and amplitude decay toward beam edges.

Abstract

The terahertz (THz) spectral region is highly attractive for imaging because it penetrates many optically opaque materials and can distinguish them by composition. THz time-domain spectroscopy (THz-TDS) is ideally suited for spectroscopic imaging, since it records the full electric-field waveform and thus broadband amplitude and phase information. Translating this capability into fast, wide-field imaging has remained a major challenge, largely because the low average power of conventional TDS sources forces a trade-off between acquisition speed, field of view and sensitivity. Here we demonstrate a fast hyperspectral THz camera based on parallel electro-optic sampling, driven by a large-area photoconductive emitter delivering a record-high 38.3 mW of average power at 400 kHz. A numerical model reproduces the measured emitter scaling and shows that, under 515 nm excitation, electron transfer into low-mobility satellite valleys delays radiation screening, allowing higher pump fluences before its effect dominates. The camera records the full THz waveform, with 1.5 THz bandwidth, simultaneously at 145,000 pixels over a 14.5 mm2 field of view at 2 Hz. These results establish high-average-power THz-TDS as a route to broadband, high-speed imaging, for inline material-resolved inspection and for spectroscopic imaging of irreversible, second-timescale processes such as hydration in biologically relevant systems.

Related papers