1. Introduction

Terahertz waves, due to their long wavelengths and low photon energies, enable unique imaging and spectroscopic capabilities for various security, medical, and biological applications. Terahertz waves have the ability to penetrate clothing, plastics, leather, and other common materials without harming human tissue like typical X-rays [1–4]. This enables the safe detection of concealed objects, which may provide complete security screening capability in a single system. For these security applications in particular, the imaging of concealed objects and the detection of various chemical, biological, and explosive substances are critical. Stand-off detection and imaging requires high average and/or peak terahertz powers for added range [5]. Security imaging at airports requires high power to achieve enough penetration through clothing and luggage to be able to identify weapons and possibly explosives in real-time. In addition, longer terahertz wavelengths are required for increased signal-to-noise and to overcome the high atmospheric attenuation in the terahertz regime [5].

Several methods exist for generating terahertz waves by optical means. One method is by optical rectification of femtosecond pulses [6,7]. This is highly effective for time-domain spectroscopy and lab-based use, but it does not generate enough average power for security applications due to low average powers and conversion efficiencies [6,7]. Another method is by quantum cascade lasers (QCL). QCLs produce very high average power terahertz radiation, but their powers drop off significantly below 2THz due to the structure of the semiconductor itself [3,4]. They also require cryogenic cooling. Difference frequency mixing of near-IR lasers has generated significant peak-power terahertz signals in nonlinear optical crystals, but with low average power output due to low repetition rates. These DFM-based terahertz sources, however, require complicated optical systems to generate the two mixing beams [8–11]. Accelerator-based sources are capable of producing very high average powers (~20W) at terahertz wavelengths, although they are rather large in size [12]. All of these sources are highly effective in generating tunable terahertz radiation, but they do not produce significant powers for stand-off applications in the appropriate frequency ranges or form factor for security applications. We propose a compact, fiber-pumped terahertz source producing high average and peak-power pulses capable of imaging concealed objects in realtime with an uncooled microbolometer array.

Our terahertz laser source uses an Yb-doped master oscillator fiber amplifier (MOFA) architecture to produce two near-IR pump beams which are difference frequency mixed in a ZGP crystal to generate terahertz radiation. Previously, ZGP has been successful in generating terahertz radiation by DFM [8,9]. In addition, it has a high nonlinear coefficient, relatively low terahertz loss, and high birefringence which allows for phase-matching to terahertz frequencies. Similarly, pulsed Yb-doped fiber systems have been successful in generating diffraction-limited, high peak-power, 1064nm pulses at various pulse widths and repetition rates [13,14]. This flexibility makes pulsed Yb-doped fiber sources ideal for driving nonlinear processes in crystals and provides capabilities that diode-pumped solid-state (DPSS) systems cannot. They have a large gain-bandwidth which allows for amplification over a large spectral range and enables the amplification of two different wavelengths (whose difference frequency is in the THz regime) simultaneously in a single gain fiber, eliminating the need for multiple lasers. This reduces the complexity and cost of typical DFM based terahertz sources which utilize YAG lasers and OPOs to generate the two mixing beams. Additionally, the fiber promotes both spatial and temporal overlap of the two mixing signals and does not involve complex optical assemblies or alignment techniques. This allows the fiber output to be focused directly into a nonlinear crystal to generate terahertz radiation.

In this paper, our experimental data and imaging results are presented and discussed.

2. System architecture and setup

For our terahertz source, we use an all-fiber MOFA with no free-space coupling of pump light. Two semiconductor seed diodes (centered at 1064.2nm and 1059nm) are driven at 100kHz with 0.7ns pulses and are pre-amplified in separate polarization maintaining (PM) single-mode amplifier stages. Those signals are then polarization combined and amplified in a series of PM-LMA Yb-doped fiber amplifiers. The 1064.2nm seed diode is fixed in wavelength, but the 1059nm diode is tunable, which allows for controlled tuning of the terahertz output signal. The sub-nanosecond pulse width and high repetition rate allows us to generate significant average and peak-power pulses (>25kW) at 1 micron with relatively low pulse energy (<20μJ). This promotes high intensities in the crystal for high average and peak-power THz generation and allows us to keep the pump fluence on the ZGP crystal below the surface damage threshold.

At the output of the fiber amplification system, the pump signals are collimated and focused directly into the ZGP crystal. The generated THz radiation is collimated with an off-axis parabolic mirror and is filtered using germanium and Teflon to block residual pump radiation and any heating effects. The THz signal is then directed onto the focal plane array of a microbolometer with a parabolic mirror for imaging. To measure the terahertz output power, a thermopile detector is used to detect the incident heat change due to the terahertz radiation. The terahertz system architecture is shown in Fig. 1.

 

Fig. 1. Schematic of the terahertz system architecture including the fiber-pump setup, terahertz conversion crystal, optics, and detector.

Download Full Size | PDF

The all-fiber pump is a truly monolithic laser source which requires limited alignment and provides a flexibility that cannot be achieved in DPSS systems [13,14]. The use of fiber also allows for efficient pump pulse amplification and compact packaging of the pump source as well as the terahertz conversion stage, which can be integrated into a small, fiber-coupled package with micro-optics to provide a robust handheld terahertz source.

3. Data and results

With our fiber-pumped ZGP setup, we were able to generate a wide range of continuously-tunable terahertz wavelengths from our crystal ranging from 0.8-2.45THz (122-375μm) by temperature tuning one pump wavelength. The terahertz tuning range was limited by the tuning range of the 1059nm diode. The highest average terahertz output power we achieved from our system was 2mW at 2.45THz. This corresponds to a peak terahertz power of 40W at 100kHz PRF and 0.5ns pulse width. As we tuned to longer terahertz wavelengths, the output power dropped, which was expected based on phase-matching conditions in the ZGP crystal. A plot of our experimental results at 2.45THz is shown in Fig. 2.

 

Fig. 2. Average THz output power vs. total incident pump power for a 2.45THz (122μm) output signal. Solid line: theoretical model based on (1), data points: experimental measurements. Total incident pump power refers to the sum of the average 1055nm and 1064.2nm pump powers incident on the crystal face. In the experiment and the model, these pump powers were set to be equal, each with a 100kHz repetition rate and 700ps pulse width.

Download Full Size | PDF

This plot shows the high conversion efficiencies for our terahertz system as well as good correlation to theory. At low pump powers, the measured data does not show a quadratic dependence on pump power as the model predicts. The data does, however, show a good relation to the theory at higher pump powers. There appears to be threshold effect in the conversion process, which we believe is a result of the crystal absorptive losses at the pump and terahertz wavelengths, which must be overcome for efficient conversion to occur. In addition, the model is based on a plane-wave approximation which neglects the effects of the thermal loading on the crystal (as a result of the absorbed pump power) and the diffraction of the pump beams. Since the ZGP crystal absorbs pump radiation, heating effects may lead to thermal de-phasing in the crystal, resulting in a discrepancy between the theory and the actual results. We are currently developing a fully diffractive nonlinear model to account for these issues which are not addressed in the plane-wave approximation. The plane-wave model does, however, provide a good first-order approximation for the mixing process, especially at high pump powers.

The theoretical model, as shown in (1), is a plane-wave, fixed-field approximation and includes refractive losses as well as the crystal absorption at the pump and terahertz wavelengths [15,16]

where ω_i is the idler frequency; d_eff is the effective nonlinearity (17.7pm/V for generation at 2.45THz); L is the crystal length; I _p2 is the pump intensity for the second mixing beam; T _pl, T _p2, and T_i are the Fresnel coefficients of the first pump, second pump, and the idler, respectively; λ _pl, λ _p2 ,and λ_i are the first pump, second pump, and terahertz idler wavelengths, respectively; n_o(λ _pl) is the ordinary index of pump 1; n_e(λ _p2,θ) and n_e(λ_i,θ) are the angular dependent extraordinary indices of pump 2 and the idler, respectively; θ is the phase-matching angle; ∆k is the momentum mismatch; α_i is the idler absorption in the crystal; and ∆α = |α _pl + α _p2 + α_i| where α _pl and α _p2 are the pump absorptions in the crystal.

Based on (1), the conversion efficiency to terahertz at 2.45THz was calculated to be 0.16%, which corresponds to 18.5% of the quantum defect. According to our measurements, we achieved a conversion efficiency of 0.1% (measured THz power to incident pump power), which is 11.6% of the quantum defect and in good agreement with the predicted value. If the Fresnel losses on crystal faces are reduced (i.e. antireflection coatings are used on each surface), the conversion efficiency and output terahertz power will increase. This essentially allows for the pump to be efficiently coupled into the crystal and the generated THz to be coupled out of the crystal with much less loss. By using antireflection (AR) coatings on the input surface of the ZGP crystal, for the same input condition, the calculated conversion efficiency based on (1) increases to 0.307% (35.5% of quantum defect). If an antireflection coating is implemented on the output face of the crystal for the terahertz radiation, the efficiency may be increased to 0.438% (50% of quantum defect). However, AR coatings for terahertz wavelengths are not standard any may prove to be too expensive for practical use.

Because we have a scaleable fiber pump, we can also increase the output terahertz power by increasing the pump power. This can be accomplished by adding additional amplification stages onto our fiber pump system. The result would be higher peak and average power terahertz pulses. The fiber scalability is limited by the fiber core size, which will allow for high peak-power amplification without introducing nonlinear effects while still maintaining near single-mode operation [13,14], as well as the surface damage threshold of the ZGP. If the pump power is increased, the pump spot in the ZGP will have to increase as well to accommodate the higher fluence on the crystal surface. We believe that a practical pump source generating 10W of average power in each pump beam (20W of total combined pump power) could be built by adding another power amplification stage to the current fiber amplifier chain, resulting in terahertz powers of >10mW at 2.45THz (by only scaling the pump power). If antireflection coatings on the ZGP crystal are used along with the increased pump power, the terahertz power and conversion efficiency will increase significantly.

With this terahertz setup shown in Fig. 1, we are able to conduct real-time transmission imaging experiments. To test this system, items were placed inside DuPont Tyvek shipping envelope and were completely concealed to the naked eye. The envelopes (with the objects inside) were then passed through the collimated portion of the THz beam 1m from the camera. The terahertz radiation transmits through the envelope but is blocked by the objects inside, resulting in an image. Results of these imaging experiments are shown in Fig. 3.

The camera used in this setup is an uncooled vanadium oxide MIM500H microbolometer array manufactured by BAE Systems, Lexington, MA. It has a 320×240 focal plane array with a 46μm detector pitch and is optimized for the far-IR wavelength range (7.5-14μm) with an electrical NEP of ~1.0pW/√Hz and an NETD of <60mK with f/1 optics; however, it remains sensitive enough in the THz region to allow for the detection and imaging of THz radiation [3,4]. An RS-170 output from the camera is connected to a frame-grabber, which is used to capture and process the analog image data. The background was subtracted, and the data was integrated over 130ms (8 frames). No additional signal processing was used.

We were successfully able to image a razor blade, a common pocket knife with a serrated blade, and a black fiberglass knife which is undetectable to metal detectors. These images demonstrate the capability of real-time THz imaging systems for use in security screening. The ability to detect metallic objects as well as non-metallic weapons provides a distinct advantage to current security screening methods.

 

Fig. 3. Pictures of objects used in THz imaging and false-color images of those objects concealed in a shipping envelope using the terahertz system; (a) razor blade, (b) imaged razor blade inside envelope, (c) knife, (d) imaged knife inside envelope, (e) fiberglass knife, (f) imaged fiberglass knife inside envelope.

Download Full Size | PDF

With this system, we can not only image concealed objects, but we can examine the absorption, reflectance, and transparencies of different materials in the THz region to help us more accurately identify concealed objects and weapons. The terahertz imaging capability coupled with the spectroscopic detection of explosive and hazardous materials may provide a robust and complete real-time security screening solution without the need for multiple systems and detection methods.

4. Conclusion

We have developed a compact, efficient, and scaleable fiber pumped terahertz source based on DFM in a ZGP crystal. The system is capable of producing high peak and average power terahertz pulses for the real-time imaging of concealed objects without the need for cryogenic cooling. Higher terahertz output powers would allow for longer stand-off distances for remote sensing and imaging. By implementing antireflection coatings to reduce losses and by increasing the pump power, the terahertz power, conversion efficiency, and imaging range can be increased significantly. Signal processing algorithms may also be implemented to improve the images produced from the system.

Future imaging experiments will focus on increasing the range of our system and using a reflection-based imaging scheme to demonstrate the capability for practical concealed object detection in real-time. This will focus on stand-off imaging at various THz wavelengths using the average and peak-power generated by our THz source with uncooled microbolometer focal plane arrays as well as RF-based receiver technologies. We also intend to examine improving conversion efficiencies by using backward-wave mixing, running the terahertz converter as an oscillator, using different nonlinear materials, invoking pump recycling, out-coupling the terahertz signal, and optimizing the mixing wavelengths.