Researchers Recover Terahertz Phase Information From Shadows

Every wave has more than brightness. It also has timing: where each oscillation sits in its cycle relative to another. That phase information can reveal a material's thickness or refractive index even when an ordinary intensity image looks nearly featureless. The problem is that phase is harder to measure than power.

Researchers from Chiba University, Tohoku University and the Autonomous University of Barcelona have now demonstrated a terahertz imaging method that reconstructs spatial phase information from shadow-like intensity patterns. The technique combines spiral-phase filtering with a frequency-tunable monochromatic terahertz source and single-pixel imaging.

The work was led by Chiba University graduate student Shusaku Mikami with professor Katsuhiko Miyamoto, Tohoku University assistant professor Seigo Ohno and colleagues. Their paper, “Terahertz phase-sensitive imaging via spiral-phase filtering and single-pixel detection,” was published in Optics Express on September 2, 2026.

The conceptual reversal is the story. Spiral-phase filters are often used to enhance edges, while the accompanying “shadow effect” can look like an unwanted distortion. The team treated the position and polarity of that shadow as data. Because the pattern changes with the incoming wavefront's phase, intensity measurements after filtering can be used to infer the phase distribution that produced them.

Why phase matters when brightness is not enough

Imagine two transparent sheets that absorb almost no terahertz radiation. One is slightly thicker, or has a different refractive index. Their transmitted intensities may be similar, yet the wave takes longer to cross one sheet. That delay appears as a phase shift. Recovering it can distinguish features that amplitude alone misses.

Conventional terahertz time-domain spectroscopy, or THz-TDS, records the electric field as a pulse evolves in time. Fourier analysis then yields both amplitude and phase across frequency. Interferometers can also compare an unknown wave with a reference. Both approaches are powerful, but they require timing control, reference paths or more elaborate detection than a straightforward monochromatic intensity measurement.

The new work addresses that mismatch. Compact or continuously tunable monochromatic terahertz sources are useful, yet their detectors often report power rather than the full electric-field waveform. If a spatial filter can translate phase variations into visible changes in intensity, a simpler detector can carry more information than it appears to measure directly.

What a spiral filter does in a 4f system

A 4f optical arrangement uses two lenses separated so that the first produces the spatial-frequency, or Fourier, representation of the object field and the second forms an output image. A filter placed in the shared Fourier plane can change how different spatial frequencies contribute to that image.

The researchers used a spiral phase plate at this plane. Its phase winds around the optical axis, creating a vortex-like transformation. For an integer topological charge the familiar result is broadly symmetric edge enhancement. With an orientation-dependent discontinuity or equivalent asymmetric condition, the output develops a directional bright-dark relief: the shadow effect.

Crucially, the bright side and dark side depend on the initial phase of the incoming terahertz field. Reversing the handedness of the optical vortex reversed the intensity pattern. Rotating the spiral phase plate changed the measured phase-related distribution in a predictable way. Those controlled reversals helped establish that the signal was carrying phase information rather than merely marking an object's outline.

Why one pixel can still make an image

Terahertz focal-plane arrays are less mature and generally more difficult than visible-light camera sensors. Single-pixel imaging offers another route. The scene is measured repeatedly under known spatial patterns; each exposure produces one integrated number. Because the pattern is known, an algorithm can solve backward for the spatial image.

The cost is time and computation. A single-pixel system does not escape the need for spatial information—it encodes that information across a sequence of measurements. Acquisition speed, source stability, signal-to-noise ratio, calibration and the number of patterns all affect the reconstruction. The new paper demonstrates compatibility between this strategy and spiral-phase filtering, not instantaneous phase video.

That distinction also clarifies the phrase “intensity-only.” The detector records intensity rather than directly measuring phase, but the illumination, filtering and modulation are engineered so phase leaves a recoverable fingerprint in those intensities. It is an indirect phase measurement, not phase created from nothing.

From an imaging artifact to a useful observable

In visible-light optics, spiral phase contrast has been studied for more than two decades. A 2005 microscopy paper demonstrated edge enhancement by placing a spiral phase element in the Fourier plane. Follow-up work showed that the method could quantitatively recover amplitude and phase profiles and that asymmetric filtering could make edge response orientation-selective.

The shadow effect often accompanied that asymmetry. For an operator seeking a clean, uniformly enhanced outline, a lopsided bright-dark pattern could be treated as a defect to suppress. The 2026 study asks a different question: if the direction of the relief changes with phase, why discard it?

Moving this idea into the terahertz band is not a trivial wavelength swap. Optical elements, detectors and spatial modulators behave differently at millimeter and submillimeter scales. The team combined a tunable source, terahertz-compatible spiral phase filtering and a single-pixel architecture previously developed by several of the same researchers. The result converts a known phase sensitivity into a practical reconstruction route for a spectral region where array detection is challenging.

How terahertz imaging reached this point

The terahertz band lies between microwave electronics and infrared photonics. It was once called a technological “gap” because efficient sources and detectors were harder to build there than on either side. Ultrafast lasers and photoconductive antennas began to change that picture in the 1980s. A landmark 1990 study used free-space terahertz time-domain spectroscopy to measure both absorption and dispersion in dielectrics and semiconductors from roughly 0.2 to 2 THz.

Time-domain methods made phase a central strength of terahertz science: measuring the electric field in time preserves information that an intensity detector discards. By the mid-1990s, researchers had demonstrated terahertz imaging. Later systems expanded into security inspection, semiconductor analysis, pharmaceutical testing and other nondestructive measurements.

Another line of development attacked the detector problem computationally. Terahertz single-pixel systems coded a scene and reconstructed it from a detector that had no imaging array. In 2020, a team including Adam Vallès, Seigo Ohno and Katsuhiko Miyamoto reported broadband, high-resolution terahertz single-pixel imaging. The new study adds phase-sensitive spiral filtering to that lineage.

Where it could matter—and what has not been proved

Phase can reveal optical path length: the combined effect of physical thickness and refractive index. In principle, that can help distinguish internal layers, measure subtle material variation or detect defects that absorb little terahertz power. The universities identify nondestructive inspection, materials analysis and biological measurement as future application areas.

Those are directions, not validated products. The public release does not report a factory inspection line, a clinical study, a portable instrument, acquisition speed suitable for moving targets or performance across a wide range of unknown samples. Nor does avoiding a conventional interferometer mean the system is mechanically simple: it still uses aligned optics, a spiral phase plate, coded measurements and calibration.

Terahertz waves are non-ionizing, but “non-ionizing” is not a blanket statement that every biological exposure or medical use is automatically safe or diagnostically effective. Biological samples also contain water, which absorbs strongly in much of the terahertz region and can limit penetration. Any medical application would require exposure assessment, tissue-specific validation and comparison with existing methods.

The immediate value is narrower and scientifically solid: the work shows that phase-sensitive spatial information can be recovered with a monochromatic intensity-detection architecture that does not use a conventional reference-arm interferometer. That expands the design space for researchers building around sources that are tunable and spectrally narrow but not naturally paired with time-domain sampling.

Five tests for the next instrument

Imaging advances often begin by converting an inconvenient physical effect into a signal. Phase-contrast microscopy turned transparent cells into visible contrast. Time-domain terahertz spectroscopy turned pulse timing into refractive-index information. Single-pixel imaging turned a sequence of totals into a picture.

This study makes a similar move. The shadow at an edge is no longer merely a blemish to remove. Under controlled filtering, it is evidence of how the wavefront arrived. The achievement is not magic vision from darkness; it is a carefully engineered translation from phase, to shadow, to numbers, and back to phase.