THz radiation gets absorbed by polar molecules such as water. These features can be seen in Fig. 1, where the simulated frequency dependent transmission through ambient air with humidity of 30% is plotted for different propagation lengths. Starting with a short distance of 1 m it is obvious that discrete frequencies are cut out of the useable spectrum (e.g. at 557 GHz, 754 GHz and 989 GHz). Reaching to frequencies above 1.6 THz those absorption lines become denser, reducing the accessible frequency range. For distances between 5 m and 20 m, the wings of the narrow lines reduce the transmission even further, so that only frequency bands ("working windows") remain. If the length reaches up to 100 m, no broadband results can be expected. This shows that the maximum working distance is limited to a couple of meters.

Explosives, like most molecular solids, have vibrational modes in the THz range. These phonon modes of the crystal lattice are used to distinguish between different energetic materials. A good spectral resolution of absorption features can be obtained by measuring a diluted pressed pellet of a substance. Using a THz inactive dilutor keeps the overall losses small while still showing the individual absorption features of the substance under investigation. The obtained results (see Fig. 2) clearly show that the measured explosives have distinct absorption lines. All explosives were identified by their THz absorption lines, which were detected in transmission geometry. But the stand-off scenario implies that these features have to be detected in reflection mode.

The mathematic relation between absorption and reflection is given by the Kramers-Kronig relation. In principle, this states that at each absorption dip, the reflectivity has an S-shaped step caused by dispersion. This is exemplarily shown in Fig. 3 for a lactose sample as an available substitute. The transmission (black curve) clearly has two absorption features at the literature values of 0.53 THz and 1.37 THz. The independently measured reflection (red curve) exhibits an S-shaped structure at the identical position. As a first-step approximation of the Kramers-Kronig relation the derivative can be used to relate reflection to absorption features. The corresponding derivative is plotted in the same graph (green curve). Clearly, the detected line positions correspond but the signal quality is reduced compared to the transmission measurement. This is related to the overall reflection coefficient limiting the accessible frequency range. Further on, artifacts such as modulations or residual water lines impair the signal. If now the sample is covered by clothes, the detected amplitude will drop further. But still, the features are seen. The development of more sensitive detectors and powerful emitters will improve the performance of that type of measurement in reflection geometry even further. Current investigations are dedicated to the system characterization and the extension of the data base including explosives, drugs and chemicals.