1. Introduction

Difference frequency generation (DFG), owning the merits of simplicity, no threshold, room-temperature operation and wide tunability, is one of the most promising approaches for monochromatic terahertz (THz) waves, which are important in high-resolution spectroscopy, imaging, etc [1,2]. Other than high-quality nonlinear crystals for optical frequency conversion, a suitable dual-wavelength laser is regarded as the most important part in a DFG process, determining the efficiency, size and applicability of the THz source. To enhance the conversion efficiency, most solutions were based on high-peak-power lasers or intracavity DFG schemes, available from costly and bulky dual-channel laser systems or optical parametric oscillators (OPOs) [3–7].

For certain applications such as THz imaging and chemical sensing, μW-level THz power with high repetition rate is sufficiently high and the critical problem is to develop portable and costless THz sources for extensive application fields. Compact dual-wavelength OPOs intracavity pumped by 1.06-μm neodymium (Nd) doped lasers have been investigated to achieve this goal [8–10], however, the redundant OPO cavity could be furtherly eliminated if wavelength tuning is unnecessary. In this case the best solution is solid-state lasers directly and simultaneously emitting two wavelengths. The most famous one was the Nd:YAG laser which has two resonating lines at 1319 nm and 1338 nm attributed to similar stimulated emission cross sections, with which THz waves at 3.28 THz was realized from DFG [11,12]. Huang et al. also observed THz beating in a self-mode-locked monolithic Nd:YAG laser operating at 1061 and 1064 nm at cryogenic temperatures [13]. The main shortcoming for a dual-wavelength Nd:YAG laser was the uncontrollable gain competition induced power instability because both wavelengths were from the same upper level and in the same polarization. It was shown in our recent study that the root mean square (RMS) instability for each wavelength reached around 10% [12], which greatly limited the stability of the THz source. The other commonly used laser gain medium was the a-cut Nd:YLF crystals, which could generate two wavelengths at 1047 and 1053 nm with orthogonal polarizations. Although there was obvious difference between their gains, dual-wavelength operation was feasible with a slightly complicated configuration by introducing mode-selecting polarization elements [14]. Nevertheless, gain competition was still inevitable with a single laser gain medium shared by two wavelengths. This problem could be solved with two independent pumping laser diodes (LDs) and laser crystals [15], but the significantly increased complexity deviated from our intention for portable and costless THz sources.

In this paper, we reported a compact and stable terahertz (THz) source based on DFG from an efficient dual-wavelength laser in the coaxially diode-end-pumping configuration. The dynamics of coaxial pumping with two combined laser crystals was firstly demonstrated in [16]. It could enable continuous-wave (CW) or Q-switched power-ratio-tunable dual-wavelength generation without gain competition in the simplest structure, needless of any additional mode-selecting elements. Here we chose two Nd:YLF crystals (a-cut and c-cut respectively) to compose a composite laser gain medium in a coaxially LD-end-pumped acousto-optical (AO) Q-switched laser to realize simultaneous 1047/1053 nm generation. Although there was obvious difference in the stimulated emission cross sections, synchronized orthogonal 1047/1053 nm laser pulses were obtained with comparable average output power and peak power. With a GaSe crystal, μW-level THz wave at 1.64 THz (182 μm) was generated at 5-kHz repetition rate. The success of such a compact and stable THz source can be extended by adopting various neodymium doped laser media to produce different THz wavelengths.

2. Experimental setup

The experimental setup is shown in Fig. 1. The dual-wavelength laser cavity was marked by the dashed line, appearing in the form of the simplest LD-end-pumped AO-Q switched laser configuration except the gain medium. The gain medium was a composite one made up of two 1-at% doped Nd:YLF crystals (CASTECH Inc.) close to each other, with an air gap of 0.2–0.3 mm in between. The first one (LC1) was 4-mm-long, a-cut and used for 1047-nm generation, while the second one (LC2) was 10-mm-long, c-cut and used for 1053-nm generation. Both crystals were anti-reflection (AR) coated around 800 nm and 1.05 μm at all end faces for LD end pumping. The LD (BWT Beijing Ltd.) had a central wavelength at 805 nm with the maximum output power of 30 W, coupled with a 400-μm-core-diameter fiber. A 1:2 imaging system, which could keep a longer Rayleigh range, was used to focus the fiber-coupled pump beam into the laser gain medium. The input mirror (M1, plano-concave, curvature radius 300 mm) had AR coatings for the pump wavelength and high-reflection (HR) coatings for the resonant wavelengths around 1.05 μm, and the output mirror (M2, plane parallel) had partial transmission (T = 10%) for laser out-coupling. A 35-mm-long AO Q-switch (Gooch and Housego), which was AR coated at 1.06 μm and driven at 41 MHz with the maximum RF power of 20 W, was used for active Q-switching. Both the gain medium and the AO Q-switch were cooled by circulating water at 20 °C.

 

Fig. 1 Experimental layout of the THz source based on a coaxial pumping dual-frequency solid-state laser. The dual-frequency laser is marked by the dashed border line. The inset is the detailed description of the focused pump beam and the composite gain medium inside the laser cavity.

Download Full Size | PDF

A parameter z was defined to indicate the position of the pump focal point, demonstrated in the inset of Fig. 1. z = 0 was located at the interface, so positive and negative values of z denoted the pump beams were focused in LC2 and LC1, respectively. As the gains in both laser crystals could be tuned by changing z and the pump absorption coefficients (implemented by varying the LD working temperature, thus the pump wavelength could be tuned), instead of balancing by cavity loss at both resonant wavelengths, specially designed coating parameters for the cavity mirrors were needless in this scheme.

The dual-wavelength pulses could be directly used for THz generation. However, here we introduced beam splitting and combining elements, used only to measure the powers and spectra for the two orthogonally polarized beams, enabled by two polarizing beam splitters (PBS) and two 45° HR mirrors shown in Fig. 1. They were actually unnecessary in both dual-wavelength laser generation and DFG for THz waves, thus all these elements could be eliminated for a practical compact THz source. The combined laser beam without focusing was incident into an uncoated 8-mm GaSe crystal to generate THz wave at 182 μm (1.64 THz) through the DFG process. The phase-matching (PM) condition was fulfilled by rotating the incident angle of the GaSe crystal. A 4.2-K Si bolometer (Infrared Laboratories Inc., calibrated to be 2.89 × 10⁵ V/W) was used for detecting the THz wave after blocking the near infrared (IR) laser with filters. An optical chopper (Stanford Research Systems, SR540) was also applied to modulate the high-repetition-rate terahertz wave into slowly varied pulses considering the detector was not fast enough.

3. Experimental results and discussion

As stated in [16], both the total output power and the power ratio were dependent on the pump focal point for a coaxially diode-end-pumped dual-wavelength laser. Firstly, the pump beam was focused at the location of z = −2 mm, where it was supposed comparable power could be obtained at both wavelengths of 1047 nm and 1053 nm. In order to enhance the pump absorption in Nd:YLF crystals, the LD wavelength was tuned to 803 nm via its temperature controlling system. Figure 2 shows the total and polarization separated powers at the Q-switching repetition rate of 5 kHz versus the input LD power, with the help of the polarizing beam splitters and a Newport 818P-100-25 power meter. The laser threshold was around 1.30 W, where only the vertically polarized beam was found resonant. The horizontally polarized beam began to emerge when the pump power was increased to 2 W. The slope efficiency of the vertically polarized laser demonstrated a declining trend with the increase of pump power, while the horizontally polarized laser power kept going up. The total power, however, increased linearly with the pump power and the slope efficiency was around 33.6%. The maximum output power was 2.92 W (1.50 W and 1.42 W at vertical and horizontal polarizations, respectively) when the input pump power was 10 W, corresponding to the optical-optical conversion efficiency of 29.2%, significantly higher than the results in [14] which used independent cavities for both wavelengths. The beam quality was measured with the knife-edge method with the M²-factor lower than 1.3 in both orthogonal directions.

 

Fig. 2 Total laser output power and the powers at each polarization of the AO-Q-switched Nd:YLF laser versus input LD pump power. The pump focal position was z = −2 mm and the pulse repetition rate was 5 kHz.

Download Full Size | PDF

The output spectra at both polarizations were monitored with a Yokogawa AQ6370D optical spectrum analyzer. It was observed that the vertical polarization was pure 1047-nm laser and the horizontal polarization was pure 1053-nm laser. That is, the output powers at 1047 nm and 1053 nm were 1.50 W and 1.42 W, respectively, with the maximum input pump power of 10 W. It should be noted that a free-running a-cut Nd:YLF (LC1) usually gives polarized laser at 1047 nm only because of a larger gain, but a free-runing c-cut Nd:YLF (LC2) laser gives unpolarized 1053-nm output. Following this concept a portion of the vertically polarized beam should be at 1053 nm, which contradicted with the experimental results. This abnormality could be explained by the gain-guiding effect at 1053 nm in the a-cut Nd:YLF crystal. Although the 1053-nm laser was mostly generated from the population inversion provided in LC2, the stored energy in LC1 could affect the generation of every 1053-nm laser pulse. However, LC1 could only give 1053-nm laser in horizontal polarization, making it preponderant in the 1053-nm pulse building process in both crystals and guiding the gain in the c-cut Nd:YLF crystal consumed by the same polarization. As a result, linearly polarized 1053-nm laser was possible. At a relatively high LD pumping level of 10 W, the 1047-nm laser generation approaching saturation would leave more population inversion to generate polarized 1053-nm laser. Such a coupling effect between a- and c-cut Nd:YLF crystals was a reliable approach to realize a stable linearly and orthogonally polarized 1047/1053 nm dual-wavelength laser. The output spectrum for the combined beam and the power fluctuation of the output powers at both wavelengths are shown in Fig. 3. The root-mean-square (RMS) instabilities of the 1047-nm and 1053-nm laser powers were 1.93% and 1.23%, respectively, while the RMS instability of the total power was merely 0.25%, much better than a common dual-wavelength laser sharing the same gain medium [12] and excellent to be used as the pump source for stable THz generation.

 

Fig. 3 Output spectra (a) and power fluctuation (b) of the coaxially pumped dual-wavelength Nd:YLF laser (z = −2 mm). The LD pump power was 10 W and the pulse repetition rate was 5 kHz.

Download Full Size | PDF

According to the theoretical model for a coaxially pumped dual-wavelength laser [16], the power ratio between two oscillating wavelengths was tunable by changing the pump focal position z, as shown by the solid curves in Fig. 4. Here the LD pump power was set as 10 W and the absorption coefficient of 1.8 cm⁻¹ was adopted for both the a-cut and c-cut Nd:YLF crystals, and two laser crystals were supposed to generate pure 1047-nm and 1053-nm lasers, respectively. Comparing with the experimental results shown as discrete symbols in Fig. 4, the 1047-nm laser power was higher and the 1053-nm laser power was lower in theoretical calculation. Equivalent power for both wavelengths happened at z around −2 mm in the experiment while it should be at z = 0 in calculation. The discrepancy was caused by the participation of the a-cut crystal in generating the 1053-nm laser. Once the inverted population in the a-cut crystal was partly consumed by the 1053-nm laser, the gain provided for the 1047-nm laser became less, resulting in the deviation at both wavelengths. The deviation became more distinct with the increase of z where the net gain at 1053 nm was even higher. The total power, however, fitted quite well between experimental and theoretical results. Considering the DFG process fulfilled the small-signal approximation condition, equivalent powers (assuming their repetition rate and pulse width were identical) at both wavelengths should be good for enhancing conversion efficiency, thus z = −2 mm was optimal with the LD pump power of 10 W.

 

Fig. 4 Experimental (discrete symbols) and theoretical (curves) output powers versus the pump focal position z for the coaxially pumped dual-wavelength Nd:YLF laser. The LD pump power was 10 W and the pulse repetition rate was 5 kHz.

Download Full Size | PDF

With 10-W LD pump power focused at z = −2 mm and Q-switched at the repetition rate of 5 kHz, the temporal behavior of the Q-switched dual-wavelength laser pulses was detected with a fast response InGaAs detector (Thorlabs DET08C) and shown in Fig. 5. The pulse train of the dual-wavelength laser is given in Fig. 5(a), demonstrating good pulse-to-pulse stability. Figures 5(b)-5(d) are the temporal profiles for the synthetic dual-wavelength pulse (29.44 ns, full wave half maximum, FWHM), pulse at 1053 nm (26.02 ns) and 1047 nm (24.06 ns), respectively. The consistency of three curves indicated good pulse synchronization at both wavelengths, favorable to further nonlinear frequency conversion to the THz range. The 1047-nm pulse was a bit narrower than that at 1053 nm because the 1047-nm laser had larger stimulated cross section, higher gain and output power, leading to faster building and decaying processes for each pulse. The single pulse energies at 1053 nm and 1047 nm were 284 μJ and 300 μJ, corresponding to the peak power of 10.9 kW and 12.5 kW, respectively. The timing jitter for the dual-wavelength pulses was found inconspicuous attributed to actively Q-switching and elegantly alleviated gain competition. When the incident LD pump power or the pump focal position was changed, the temporal overlapping of the pulses could be changed, which was useful in the case of monitoring a transient process.

 

Fig. 5 The temporal behavior of the coaxially pumped Q-switched dual-wavelength Nd:YLF laser (z = −2 mm) recorded by a Tektronix DPO 2024B oscilloscope: (a) Pulse train; (b) Synthetic pulse shape for both wavelengths; (c) Pulse shape at 1053 nm; (d) Pulse shape at 1047 nm. The LD pump power was 10 W and the pulse repetition rate was 5 kHz.

Download Full Size | PDF

Monitored with a Spiricon Pyrocam III camera for space overlapping, the synchronized dual-wavelength pulses were combined and directly incident into an 8-mm uncoated GaSe crystal without focusing for THz generation. A 1-mm-thick germanium (Ge) wafer and a 2-mm-thick black polyethylene (PE) wafer were placed close to the detector window to block the residual pump beams at 1 μm. The Ge wafer could filter most of the laser powers, or the black PE would melt if directly exposed to high-power laser radiation. Rotating the angle of the GaSe crystal in the x-z plane it was found the external PM angle was at 11° corresponding to the internal PM angle of around 4° (type o-e→e DFG), which accorded well with the theoretical results calculated by Sellmeier equations in [17]. The effective nonlinear coefficient was $d_{eff}=d_{22}{\cos }^2\theta \cos (3\phi )\approx d_{22}=54pm/V$, where $\phi =0$ was applied for maximum conversion efficiency. The THz output voltage of the bolometer as a function of the input dual-wavelength laser power is shown in Fig. 6(a). The maximum voltage of 269 mV was achieved corresponding to the average power of 0.93 μW, when the pump power was 2.8 W. The DFG efficiency was 3.3 × 10⁻⁷ and the photon conversion efficiency was 1.15 × 10⁻⁴. Such an efficiency was among the mainstream results with GaSe crystal based on 1-μm high-repetition-rate Q-switched lasers [14,18,19], but much lower than pumping with high-peak-power pulsed lasers and lasers at longer wavelength [8,9,20,21]. The peak power was estimated to be around 10 mW assuming the THz pulse width was less than 20 ns. We didn’t focus the pump beam to avoid damage to the GaSe crystal, but it was clear the THz power went up in quadratic relation to the pump intensity. A typical pulse shape is shown in Fig. 6(b) modulated by the chopper at the frequency of 40 Hz. The power instability (RMS) of the THz wave was less than 3.9% in 10 minutes, good enough for practical applications.

 

Fig. 6 THz output characteristics. (a) Dependence of average THz output voltage on the dual-wavelength laser power. Solid curve corresponds to the quadratic fit to data points. (b) A typical THz waveform from the bolometer.

Download Full Size | PDF

Neglecting the beam splitting and combining elements, such a construction was extremely compact with a laser cavity of 80 mm and a whole length of less than 250 mm including the coupling lens, nonlinear DFG crystal and filters, providing a good approach for stable and portable THz sources. It should also be noted that such a compact THz source based on a coaxially pumped dual-wavelength laser is a general method for various Nd-doped laser gain media to obtain different THz frequencies. For example, a Nd:YLF/Nd:YAG composite medium gives 1047/1064 nm or 1053/1064 nm dual-wavelength lasers for THz generation at 4.58 THz and 2.95 THz, provided suitable DFG materials. The advantages of utilizing an a-cut Nd:YLF and a c-cut Nd:YLF in this paper include the same laser-upper-level lifetime for better pulse synchronization and high DFG efficiency for GaSe around 1.5 THz. Nonlinear crystals like quasi-phase-matched (QPM) GaP crystal is more prospective in improving DFG efficiency with this scheme, because of its good physical and optical properties such as high nonlinear gain, no walk off, AR coatings to reduce reflection loss and higher damage threshold allowing higher pump intensity by focusing. To make the THz source more compact, further improvement can be focused on passive Q-switching and organic DFG materials (e.g., DAST) with dual-wavelength lasers in the 1.3-μm range, through which a palm-size THz source is feasible.

4. Conclusion

We demonstrated a novel coaxial pumping configuration for dual-wavelength solid-state lasers to generate THz wave through DFG. With a composite gain medium composed by an a-cut Nd:YLF crystal and a c-cut Nd:YLF crystal, efficient, stable, orthogonally polarized and synchronized high-repetition-rate dual-wavelength laser pulses at 1047 nm and 1053 nm were achieved by tuning the LD pump focal position to balance the gains at both wavelengths. THz wave at 1.64 THz was generated in a GaSe crystal with the maximum average power of 0.93 μW, providing a realistic scheme for compact and stable high-repetition-rate THz sources. By replacing the composite gain medium with different Nd-doped laser crystal pairs, a variety of THz frequencies can be covered with portable devices, which have good prospect in extensive applications such as imaging, chemical sensing and so on.