THz Time-Domain Spectroscopy & Imaging

The question is field strength, not preference. A photoconductive antenna is fibre-friendly, repeatable and gives better than 75 dB of dynamic range out to about 5 THz — the right answer for spectroscopy, imaging and inspection. An organic crystal under optical rectification is what you use when the measurement itself depends on field strength: BNA at 0.5 – 3 THz pumped at 1030 nm, DSTMS at 0.3 – 15 THz and DAST at 0.1 – 17 THz pumped at 1.3 – 1.5 µm. A fully integrated chain reaches above 10 MV/cm at the focus.

On a commissioned turnkey system: above 10 MV/cm in everyday operation with a guaranteed floor of 8 MV/cm, at a 0.14 mm focus, from roughly 0.7 mJ of pump at 0.6 – 0.9 % conversion — about 4 µJ per THz pulse. Those numbers come from commissioning measurements with the beam path in open air; purging improves them. Field is calibrated by Kerr rotation in diamond and cross-checked against an independent source to within about 8 %. Published work with DSTMS pumped by a 34 mJ OPA has reached 85 MV/cm, so the ceiling is set by your pump, not the crystal.

GaP (110) at 0.3 mm is the default for a 1030 nm Yb probe — its optical and THz indices nearly match at 1 µm, giving roughly 0.1 – 7 THz. ZnTe is more sensitive but rolls off near 3 THz. Thickness is a direct trade: 0.1 – 0.3 mm for detection bandwidth, 0.3 – 1 mm when you want signal. In every case the detection arm is crystal → quarter-wave plate → Wollaston prism → balanced photodetector.

Because EOS measures the field, not the power. You get amplitude and phase from one time-domain scan, which means the complex refractive index follows directly without a Kramers-Kronig inversion. A pyroelectric sensor measures total incident power incoherently — useful for alignment, calibration and daily power checks at below 1 nW/√Hz noise-equivalent power, but it cannot give you a spectrum.

Frequency resolution is roughly the inverse of your delay range. 600 ps is the standard configuration; extending to 3 ns is what buys better than 0.5 GHz. If you need both long range and fine steps, a fibre-coupled optical delay line covers up to 50 ns at 5 fs resolution. On an intense-field bench the motorised stage steps down to 0.015 ps with about 1 ps time resolution.

Yes, with an uncooled microbolometer array — no cryogens. A 160 × 120 array at 25 µm pitch covers below 1 THz to 18 THz at 10, 30 or 60 Hz over USB 3.0. Paired with an f/0.7 objective covering 0.1 – 30 THz, 3 mm features have been resolved in a single short exposure without correction. For quantitative imaging you can also raster a photoconductive spectrometer, which trades speed for spectral information per pixel.

Above roughly 0.5 THz, water vapour rotational lines cut visible holes in the spectrum. For survey work in a dry room you can live with it; for quantitative absorption measurements you cannot. Dry nitrogen purge is the usual answer. If your sample sits in a cryostat, high-resistivity silicon (n ≈ 3.42, nearly flat across the band, rigid enough for vacuum) or TPX (n ≈ 1.46, near-zero dispersion, transparent in the visible so you can align through it) are the two window materials to choose between.

Handle them as consumables with a defined operating envelope. BNA and the stilbazolium salts (DSTMS, DAST) are humidity and damage sensitive and want dry-purged storage. Damage threshold differs between them — DAST tolerates more than DSTMS — so the pump fluence budget is part of the system design. Apertures up to 65 mm are available, and larger apertures let you spread the same pulse energy over more area, which is often the cheapest way to stay under threshold.