Principal Investigators:Prof. Dr. Clara Saraceno, RUB; Prof. Dr. Aydin Sezgin, RUB

Achieved results and methods

Saraceno (M01 in 2^nd Phase): High-power transceivers and static imaging

During the 2^nd phase, we continued the development of unique sources with ultra-high bandwidth and high average power and achieved several milestones in the direction of more compact sources. We demonstrated the highest power lab-based THz source so far achieved with 643 mW, realized with the nonlinear crystal lithium niobate in the tilted pulse front geometry [3]. This source operates at 40 kHz repetition rate and is driven by a large 500 W laser amplifier system. We also worked on extending the bandwidth of high-power THz sources beyond 5 THz and could show up to 8 THz bandwidth with multi-mW average powers [6]. These sources were used within MARIE for benchmarking with extreme powers, testing detectors and characterizing materials. However, their rather moderate repetition rate of <1MHz makes them typically unsuitable for fast imaging as targeted here; and the rather large driving energies required by nonlinear crystals makes the laser systems large and cumbersome. For much higher repetition rates and correspondingly more compact driving lasers, photoconductive (PC) emitters and receivers are a more suitable alternative, which we explored in the 2^nd phase: (1) we explored for the first time the average power handling capabilities of GaAs large-area emitters LAEs and demonstrated they could be excited damage-free with up to 18W of average power in transmission and without heatsinking (Fig. 1). Lower conversion efficiency could be attributed to residual thermal effects, providing a clear path forward towards future optimization at even higher powers and efficiency [4], (2) together with C07, we demonstrated the first 1030nm compatible PC receiver and tested its response and capabilities with high THz average power [7]. This resulted in a state-of-the-art peak dynamic range of 120dB in minute-long measurement times. Finally, we also studied a novel path towards more compact broadband high-power THz sources and demonstrated the first few-cycle THz source driven intracavity of a mode-locked thin-disk laser at 45MHz repetition rate [8].

With respect to THz imaging, we demonstrated the potential of our high-power sources compared to commercially available THz-TDS in a first proof-of-principle experiment, by imaging a low reflectivity, 3D printed object using a time-of-flight lensless imaging modality [5], [9]. When comparing images obtained with identical acquisition time, the images showed major advantages in terms of SNR as compared to low-power commercial THz-TDS, setting the stage for the proposed work in the 3^rd phase. However, despite great promise, the potential of the unique high-power pulsed THz sources for THz imaging could so far not fully be exploited. In fact, in these preliminary experiments, the slow acquisition of traces, spatial scanning of the object and reconstruction time of the image was prohibitively long (>30min), even only in 2D, which is incompatible with dynamic imaging scenarios as targeted here. To bring these new high power pulsed sources to their full potential, we propose in the 3^rd phase to explore fast time-domain acquisition techniques including equivalent time sampling and single-shot time-frequency mapping techniques combined with scanless single-detector imaging and fast and resilient image reconstruction algorithms to bring the advantages of high-power THz-TDS to highest speed imaging. 

Sezgin (S02 in 2^nd Phase): Algorithms for image reconstruction

The achievements of S02 in the 2^nd phase are multifold. In [10], we addressed the compensation of phase noise, one of the limiting factors in traditional radar THz sensing, by utilizing the concept of reflecting intelligent surfaces (RIS). We investigated the compensation of clutter effects in [2] and [11].  In [2], the strong clutter is due to the reflection of the surface of composite materials resulting in poor detection of material defects, which are inside a layered material structure. In contrast, in [11], reflected signals from unwanted objects in self-localization based on passive RFIDs represent the clutter with clutter echoes that are much stronger than the backscattered signals of the passive tag landmarks.

Thus, sophisticated signal separation methods were required for improved defect detection. In many scenarios, the number of defects that we are interested in is limited and the signalling response of the layered structure can be modelled as a low-rank structure. Therefore, we propose joint rank and sparsity minimization for defect detection using compressive sensing based multiple-input and multiple-output (MIMO) wireless radar. We proposed a non-convex approach based on the iteratively reweighted nuclear and ℓ1−norm (a double-reweighted approach) to obtain a higher accuracy compared to the conventional nuclear norm and ℓ1−norm minimization. To this end, an iterative algorithm is designed to estimate the low-rank and sparse contributions. Further, we propose deep learning to learn the parameters of the algorithm (i.e., algorithm unfolding) to improve the accuracy and the speed of convergence of the algorithm. Altogether, our approach significantly improved the identification of passive indoor self-localization tag landmarks and defect detection. These novel techniques are of conceptual nature, thus not limited to specific applications, for static environments and can be naturally generalized to time-of-flight imaging with broadband beams in the case of compressed sensing in single-pixel imaging for dynamic recovery we want to address in the 3^rd phase.

One of the difficulties of the low-rank plus sparse recovery approaches (RPCA) utilized in [2] and [11] is to identify the corresponding domains (dictionaries) in which the signal components are sparse/low-rank in nature. This difficulty is addressed in our work in [12] in which we address parameter estimation (like the dielectric constant) of the material under test (MUT) by wireless sensing. Note that the precision of the estimation depends on the accurate estimation of the properties of the reflected signals from the MUT (e.g., number of reflections, their amplitudes, and time delays). Here, instead of a fixed dictionary for the decomposition of the reflected signals, an iterative dictionary-update technique is proposed to improve the estimation of the reflected signals. To validate the proposed method, a vector network analyzer (VNA)-based measurement setup is used. It turns out that the estimated dielectric constants of the MUTs are in close agreement with those reported in the literature. Further, the proposed approach outperforms the state-of-the-art model-based curve-fitting approaches in thickness estimation.

We also utilized learning concepts in [13] within the direction of arrival estimation problems, in particular when one has to compensate sensor failures. In more details, we presented a MUSIC-based DOA estimation strategy using small antenna arrays, via employing deep learning for reconstructing the signals of a virtual large antenna array. Interestingly enough, the proposed strategy delivered significantly better performance than simply plugging the incoming signals into MUSIC, and it was also better than directly using an actual large antenna array with MUSIC for high angle ranges and low-test SNR values. These concepts can all be expanded to contribute to the more challenging case of broadband, time-domain THz illumination. In fact, many of the conceptual problems are similar in radar high-frequency imaging reconstruction, yet the broad bandwidth up to very high frequencies and the temporal evolutions of the object due to dynamic imaging will result in low sample size leading to novel artifacts, which have not been addressed yet in the literature.

Selected project-related publications

1. B. Schäfer, U. S. K. P. Miriya Thanthrige, and A. Sezgin, “Iteratively Reweighted Nuclear Norm based Distortion Compensation for THz-TDS,” in 2022 Fifth International Workshop on Mobile Terahertz Systems (IWMTS), Jul. 2022, pp. 1–5. doi: https://doi.org/10.1109/IWMTS54901.2022.9832448.
2. S. K. P. Miriya Thanthrige, P. Jung, and A. Sezgin, “Deep Unfolding of Iteratively Reweighted ADMM for Wireless RF Sensing,” Sensors, 22(8), Art. no. 8, Jan. 2022, doi: https://doi.org/10.3390/s22083065.
3. Vogel and C. J. Saraceno, “643 mW Average Power Lithium Niobate THz Source,” in CLEO 2023, paper SF3I.1, Optica Publishing Group, May 2023, p. SF3I.1. doi: https://doi.org/10.1364/CLEO_SI.2023.SF3I.1.
4. Khalili, T. Vogel, Y. Wang, S. Mansourzadeh, A. Singh, S. Winnerl, C. J. Saraceno, “Microstructured large-area photoconductive terahertz emitters driven at high average power.” arXiv, Feb. 21, 2024. doi: https://doi.org/10.48550/arXiv.2402.13940.
5. Mansourzadeh, D. Damyanov, T. Vogel, F. Wulf, R. Kohlhaas, B. Globisch, T. Schultze, M. Hoffmann, J. C. Balzer, C. J. Saraceno, “High-Power Lensless THz Imaging of Hidden Objects,” IEEE Access, vol. 9, pp. 6268–6276, 2021, doi: 10.1109/ACCESS.2020.3048781.
6. Mansourzadeh, T. Vogel, A. Omar, M. Shalaby, M. Cinchetti, and C. J. Saraceno, “Broadband, high power THz source at 540 kHz using organic crystal BNA,” APL Photonics, vol. 8, no. 1, p. 011301, Jan. 2023, doi: 10.1063/5.0126367.
7. Vogel, S. Mansourzadeh, U. Nandi, J. Norman, S. Preu, and C. J. Saraceno, “Performance of Photoconductive Receivers at 1030 nm Excited by High Average Power THz Pulses,” IEEE Trans. on THz Science & Technology, v14(2), pp. 139–151, Mar. 2024, doi: https://doi.org/10.1109/TTHZ.2024.3358616.
8. Wang, T. Vogel, M. Khalili, S. Mansourzadeh, K. Hasse, S. Suntsov, D. Kip, C. J. Saraceno, “High-power intracavity single-cycle THz pulse generation using thin lithium niobate,” Optica, vol. 10, no. 12, p. 1719, Dec. 2023, doi: 10.1364/OPTICA.504513.
9. Sánchez-Pastor, U. S. K. P. M. Thanthrige, F. Ilgac, A. Jimenenz-Saez, P. Jung, A. Sezgin, R. Jakoby, “Clutter Suppression for Indoor Self-Localization Systems by Iteratively Reweighted Low-Rank Plus Sparse Recovery,” Sensors, 21(20), Jan. 2021, doi: https://doi.org/10.3390/s21206842.
10. S. K. P. M. Thanthrige, J. Barowski, I. Rolfes, D. Erni, T. Kaiser, and A. Sezgin, “Characterization of Dielectric Materials by Sparse Signal Processing With Iterative Dictionary Updates,” IEEE Sensors Letters, vol. 4, no. 9, pp. 1–4, Sep. 2020, doi: https://doi.org/10.1109/LSENS.2020.3019924.