1. Introduction

Broadband free-space terahertz time-domain spectroscopy (FS-THz-TDS) is a rapidly evolving field due its relevance for the optical characterization of a wide variety of material systems [1]. The setup for THz-TDS consists mainly of two parts: terahertz generator and detector. It is desirable for the setup to have high signal-noise ratio (SNR) but cost-effective design. Many efforts have been reported to improve the SNR, e.g. by minimizing the background, maximizing the optical modulation depth of the EO detection, and scaling up the energy of terahertz pulses [2-7]. Usually, in this kind of setup lock-in amplification technology is still used due to the weak terahertz radiation thus both high power and high repetition rate of the pumping laser pulses are preferred in order for high SNR. As the pumping sources of the THz-TDS, the femtosecond laser pulse oscillators have much higher pulse repetition rate, more cost-effectiveness and more compactness but much lower pumping pulse power compared with the chirped pulse amplification systems. Intracavity optical method is a hopeful way to avoid somewhat the disadvantage of the output power of the laser oscillators. Optical rectification is one of the main mechanisms used to generate few cycle terahertz pulses pumped by fs lasers due to its broad bandwidth, simple alignment, compact configuration, and high commercialization of the used nonlinear crystals [8,9]. This paper, to our knowledge, reports firstly to generate broadband terahertz radiation through intracavity nonlinear optical rectification. According to our measurements, our compact, cost-effective design scales up the pumping power of the terahertz emitter up to 11.3 W by a femtosecond laser pulse oscillator thus improves the SNR by about 29 times compared with the counterpart through extracavity nonlinear optical rectification.

2. Experimental results and discussions

Our setup is illustrated schematically in Fig. 1 . Similar to the designs of many femtosecond Ti:sapphire oscillators, a continue-wave 532 nm light beam from a Nd:YVO₄ frequency-doubled laser is used to pump a 5 mm-thick Brewster-angled Ti:sapphire crystal through a pumping mirror CR1 so that about 90% pumping power is absorbed by the crystal. The CR1 with a radius of curvature of 10.0 cm is coated highly reflective (HR) over 750~920 nm spectral region and highly antireflective (AR) at 532 nm wavelength. The laser cavity extends from the mirror CR4 to output coupling mirror OC with a total length of 2.05 m. The curved mirrors CR1 and CR2 provide the laser crystal with a focus for Kerr mode-locking and fold the laser beam by a angle of ~16° for astigmatism compensation due to the Brewster angle-cut gain crystal, while CR3 and CR4 offer another focus for strong pumping intensity of a 1mm-thick <110> ZnTe crystal used as terahertz emitter (TC). The separation between CR1 and CR2 shall be carefully aligned for stable mode-locking while that between CR3 and CR4 is designed for an appropriate waist size of the 800 nm beam (a diameter of ~220 μm here) in order for the TC to generate terahertz radiation with high conversion efficiency but weak diffraction effect [10,11]. In order to minimize the induced loss, the emitter is pumped by Brewster incidence. The curved mirrors (CR2~CR4) with a radius of curvature of 10.0 cm are coated HR over 750~920 nm spectral region. The output coupler (OC) is a piece of plan mirror which has about 2.5% or 1.0% transmission at 800 nm without or with the terahertz emitter inside the cavity. Four BK7 Brewster prisms (P1~P4) are used for the compensation of intracavity dispersion and for the output laser pulses from OC without spatial chirps. At the absence of the terahertz emitter inside the cavity, the oscillator can output a mode-locked pulse chain with 356 mW average power, ~38 nm bandwidth and 38.5 fs auto-correlation pulse duration under 3.9 W pumping power. In spite of Brewster alignment, the insertion of the ZnTe crystal induces about 1.0% single-pass loss due to the linear, nonlinear absorption and scattering etc. of the crystal to the NIR laser. Correspondingly, the output power decreases to typically 113 mW at the presence of the terahertz emitter inside the cavity with 4.2W pumping power, so the intracavity laser power can be estimated about 11.3 W. As shown in Fig. 2(a) and Fig. 2(b), the bandwidth (FWHM) and the auto-correlation width of the output pulses are measured as 32 nm and 52 fs separately.

 

Fig. 1 Experimental setup. L: Lens; P1~P4: Brewster prisms; CR1~CR4: concave mirrors; PM1~PM4: golden coated off-axis parabolic mirror; M1~M5: mirrors coated with 45° HR @ 800 nm; OC: output coupler; TC: crystal used as terahertz emitter; EO: electro-optics crystal; C: phase compensator; PL1, PL2: polarizers; D1, D2: detectors.

Download Full Size | PDF

 

Fig. 2 The output spectrum (a) and pulse duration (b) of the femtosecond Ti:sapphire oscillator with intrcavity optical rectification crystal.

Download Full Size | PDF

The terahertz radiation from the emitter overflows the CR4, and then is collimated by a gold-coated 90° off-axis parabolic mirror (PM1). The out diameter of CR4 shall be chosen large enough for 800 nm laser but as small as possible (1.0 cm here) for coupling effectively the terahertz radiation out of the laser cavity. In order to filter out the noises from the laser and environments, a piece of silicon wafer is inserted between PM1 and PM2. After focused by another 90° off-axis parabolic mirror (PM2), the radiation is chopped optically in order for optical detection with Lock-in amplifier. The 90° off-axis parabolic mirrors (PM3 and PM4) are used to image the focused beam into another 1mm-thick <110> ZnTe crystal for electro-optic detection (EO). Here, all the off-axis parabolic mirrors (PM1~PM4) have a diameter of 62.5 mm and an effective focal length of 12.5 cm. The output beam from OC is used as probe beam for electro-optic detection as described by reference 12, which overlaps with the terahertz beam inside the EO after through a temporal delay line and a linear polarizer (PL1).

As is well known, the <110> ZnTe crystal as terahertz emitter usually operates under normal incidence. In our experiment, the crystal is firstly aligned optimally for terahertz generation pumped by normal incidence based on extracavity nonlinear optical rectification. This work has been carried out without the emitter inside the cavity and with the 2.5% transmission to couple out the IR pulse chain. About 320 mW laser power is split to work as pumping beam of the terahertz emitter and the rest (35 mW) as the probing for terahertz EO detection. For the convenience of comparison with the intracavity case, the focal length of the pumping lens is chosen so that the pumping size on the terahertz emitter is approximately equal to the waist size between CR3 and CR4 in Fig. 1. The crystal is then rotated horizontally so as to be pumped by Brewster angle. During all our measurements for THz signals, we use the same setting of the Lock-in amplifier, including the time constant, the optical chopped frequency, the dynamic reserve and so on. Figure 3(a) presents the measured terahertz field signals by EO detection under normal incidence pumping configuration (dot line) and with Brewster angle pumping design (solid line) while Fig. 3(b) shows their corresponding Fourier spectra. Obviously, the dot lines is more smooth than the solid one in Fig. 3(a), which means the former has better SNR than the latter. According to the definition in reference 13, SNR is the ratio of the peak-to-peak signal amplitude to the root-mean-square noise level, so the calculated SNRs is estimated as 1030:1 for the dot curve and 326:1 for the solid in Fig. 3(a). The oscillating temporal period of the dot seems shorter than that of the solid thus the generated THz pulse has shorter temporal duration pumped by normal incidence than by Brewster angle incidence. Correspondingly, as shown in Fig. 3(b), the former has broader spectral distribution than the latter, especially in the spectral region lower than 0.6 THz. All these differences from the signals, we guess, are originated from the optimal design of the emitter for the normal incidence rather than the Brewster angle pumping configuration. When the crystal is pumped by normal incidence, the conversion efficiency can be maximized by rotating the crystal around optical axis so that θ, the angle between the crystallographic axis [0, 0, 1] and the polarization of the NIR pumping laser is about 55° [14]. Interestingly, the generated terahertz field has almost the same polarization direction as the pumping beam. If the pumping beam deviates away from the normal incidence, the available maximum of the conversion efficiency will decrease. Correspondingly, the polarization of the terahertz field will be different from that of the pumping beam, which means that the s-components of the generated terahertz field will be filtered mostly out by Brewster window. Obviously, the transmission of the Brewster window has stronger wavelength-dependent than that of the optical boundary to normal incident beam. The analyses above imply that the Brewster window here can make some contributions to the wavelength/polarization-independent amplitude shaping /filtering of the terahertz field.

 

Fig. 3 The measured terahertz signal by EO detection through extracavity nonlinear optical rectification when terahertz emitter pumped by normal incidence and by Brewster angle.

Download Full Size | PDF

After optimization of the emitter’s alignment, one can move the TC to the waist position between CR3 and CR4. Now some appropriate alterations for the separation between the P3 and P4, and the output coupler OC are necessary so that the terahertz radiation is generated from TC pumped by the stable mode-locked laser pulses inside the cavity and probed by the output laser beam from OC as shown in Fig. 1. The measured terahertz field is shown as Fig. 4(a) . For the convenience of our analyses, the probe laser power is also attenuated to 35 mW as that used in Fig. 3. The calculated SNR in Fig. 4(a) is estimated as 9360:1, about 9.1 or 29 times higher than the dot line or the solid line in Fig. 3(a), separately. As we know, the amplitude of the measured terahertz field is proportional to both the pumping and the probing intensities, which is equivalent to the pumping and the probing powers for a given beam size, through optical rectification and EO detection [14]. What is more, according to our experiment before [3], the noise level is also proportional to probe power for large enough probing power at absence of terahertz field. So it seems that the measured ratio of SNRs of the signal in Fig. 4(a) and the solid line in Fig. 3(a) is lower slightly than the value from our theoretical estimation, 35.3, the ratio of the pumping powers. The difference maybe results from the small discrepancy of the two pumping beam sizes, the stabilities of the NIR mode-locked laser pulses of the two pumping sources and the noises induced by the two terahertz fields. The OC also blocks some terahertz radiation under intracavity pumping configuration thus results in the decrease of the measured terahertz signal. The measured ratio of SNRs of the signal in Fig. 4(a) to the dot line in Fig. 3(a) is 9.1 much lower than 35.3, the value of our theoretical prediction, which also results from the decease of the maximal nonlinear conversion efficiency and the optical filtering due to the Brewster incidence as well as the stronger nonlinear absorption operating at intracavity pumping. Figure 4(b) shows, similar to the solid curve in Fig. 3(a), the THz signal from the intracavity optical rectification has longer oscillating temporal period and narrower spectral distribution.

 

Fig. 4 The measured terahertz signal by EO detection through intracavity nonlinear optical rectification when the terahertz emitter pumped by Brewster angle incidence.

Download Full Size | PDF

3. Conclusion

Summarily, we present a novel design for getting high SNR broadband terahertz signal. This setup uses a femtosecond laser oscillator as pumping sources for operation at high pulse repetition rate and cost-effectiveness. In order to scale up the pumping pulse power, the pumping configuration is designed based on intracavity nonlinear optical rectification. A piece of 1-mm-thick <110> ZnTe used as a terahertz emitter is pumped by Brewster angle at a waist of the ultrashort NIR laser beam inside a Ti:sapphire laser cavity. Our experimental results show that our efforts improve the SNR by about a factor of 29 due to the much stronger pumping compared with the counterpart based on extracavity nonlinear optical rectification. Compared with conventional normal pumping design, our intracavity nonlinear optical method also has 9.1 times higher SNR in spite of the optimal design of the emitter pumped by normal incidence. One can concludes from our results that if an emitter designed for Brewster angle pumping configuration is used, it is practicable for our novel design to increase the SNR of the measured THz radiation by ~30 times compared with the counterpart through extracavity nonlinear optical rectification. This work opens a new way to improve the SNR of the broadband FS-THz-TDS pumped by a femtosecond laser oscillator.