1. Introduction

Frequency up-conversion of established high-performance ultrafast sources in the near-infrared represents an attractive method to access the technologically important blue and UV spectral regions. The high peak intensities of ultrafast pulses allow for better utilization of the available nonlinearity, and the short duration of the resulting up-converted pulses gives increased functionality in such areas as time-resolved spectroscopy. Significantly, a simplified, compact and portable femtosecond blue laser source represents an excellent tool in the field of biomedicine.

Normally, relatively complex and inefficient ultrafast, titanium-sapphire lasers are used as the pump sources, but their typical electrical-to-optical efficiencies are less than 0.05% for the generation of near-infrared pulses [1]. We report here a compact, directly diode-pumped Cr:LiSAF laser as an alternative pump source. Requiring less than 1.2W of electrical diode drive power, this laser has an impressive electrical-to-optical efficiency in excess of 1%, and produces transform-limited, 150fs infrared pulses with peak powers of around 700W. At these modest power levels, efficient frequency doubling usually requires complex arrangements, especially if the pulse duration is to be preserved. However, in this paper we describe the use of a relatively thick doubling crystal in a single-pass, extra-cavity arrangement. Although such thick-crystal approaches allow a highly efficient blue generation, one might expect serious temporal broadening of the second harmonic (SH) pulses. Nonetheless, we measure blue pulses of 543fs in duration, demonstrating that we do in fact remain well within the femtosecond regime, and that this approach represents an effective way to achieve highly-efficient frequency conversion in a simple and practical configuration.

The initial work on this SHG scheme [2] used a Kerr-lens modelocked Ti:sapphire laser that produced 120fs pulses at 860nm, with an average power of 300mW and pulse energy of 3.75nJ. A conversion efficiency of 57% was achieved for up to 170mW of average power at 430nm. With considerably less (27 times) pulse energy and only 45mW average power, we still obtain a 30% conversion efficiency and demonstrate an electrical-blue efficiency of 1% using a diode-pumped femtosecond Cr:LiSAF laser. In addition, we achieve these high efficiencies while operating at a high repetition-rate of 330MHz. The size, cost and overall efficiency of these Cr:LiSAF lasers are superior to any Ti:sapphire system. We generate as much as 11.8mW of blue light at 429nm for only 44.6mW of incident fundamental (860nm). This corresponds to an optical-optical SHG conversion efficiency of 26.5%, but efficiencies as high as 30% have been observed (10.5mW of blue light for 35mW of incident fundamental). We investigated a variety of focusing conditions and their effects on SHG performance, overall slope efficiencies, and blue spectral width. We measured the fundamental and second harmonic pulse durations by autocorrelation techniques, and characterized the wavelength and temperature phasematching acceptance bandwidths of the potassium niobate (KNbO₃) doubling crystal under conditions of ultrashort fundamental pulses.

2. Compact, low-threshold, tunable Cr:LiSAF source laser

Rather than utilising the impressive power characteristics of broad-stripe laser diodes as a pump source for our Cr:LiSAF laser crystal, we have devised a compact and low-threshold cavity design, pumped with inexpensive single-narrow-stripe laser diodes [3, 4], and have demonstrated such a system with modelocking thresholds as low as 22mW [5]. Two pairs of AlGaInP diodes were incorporated in the pump geometry; two provided an output power of 50mW at 660nm and two provided 60mW at 685nm. Our resonator has been designed with very low thresholds and simplified dispersion compensation in mind. The cavity includes only a 3mm (5.5 atm % doped) Cr:LiSAF laser crystal, two dichroic high-reflectors, a semiconductor saturable absorber mirror (SESAM) modelocking element [6] identical to that described in [7], an output coupler, and a single prism for group velocity dispersion compensation [3, 8] as well as a facility for tuning. By tilting the output coupler in the horizontal plane, smooth tuning of the femtosecond-pulse output was possible over the full 50nm (~826-876nm) set by the reflectivity bandwidth of the SESAM used. The set-up is illustrated in fig.1.

We observe the generation of near-transform-limited pulses of 150fs duration centered at 860nm and at a repetition-rate of 330MHz. Average output powers of up to 45mW have been achieved for ~180mW of incident diode pump power, giving an optical-optical efficiency of 25%. With the spectral bandwidth of ~6nm this implies a time-bandwidth product of 0.36. The prism provided a continuous tuning range of 50nm (826-876nm) under modelocked conditions, which was limited by the inherent mirror reflectivity of the SESAM [7]. Modelocking itself was robust and entirely self-starting. Although the laser was not enclosed, nor located in a lab with any temperature stabilization, the amplitude stability of the laser output was observed to be excellent.

We have reported a miniature version of this laser design on a 22 × 28cm² breadboard, which requires only six penlight (AA) batteries as a diode power source for over 12h of stable modelocked operation [4]. The laser is entirely portable and has an impressive electrical-optical efficiency of 4%.

3. The thick-crystal frequency-doubling approach

Within the femtosecond regime, an upper limit is usually set on the nonlinear crystal length by the group velocity mismatch (GVM), which leads to a temporal walk-off between the generating (fundamental) and generated (SH) beams. When this delay becomes longer than the duration of the fundamental pulse, the beams have propagated along a GVM length, L_t, and this length typically defines the crystal thickness. For 120fs pulses at 858nm, L_t for KNbO₃ is ~100μm, but we have used crystal thicknesses of 3mm and 5mm.

This thick-crystal frequency-doubling technique was first reported by Weiner et al [2] who used a 3mm thick KNbO₃ crystal to achieve very high SHG conversion efficiencies of fundamental pulses, despite high GVM. In using KNbO₃, we are able to utilize birefringent, type I, non-critical phasematching (NCPM) at 858nm (central Cr:LiSAF wavelength) at room temperature. The KNbO₃ crystal also boasts a high nonlinear optical coefficient, d₃₂ ~18.3 pm/V [9], which can be fully exploited with this approach. In utilizing NCPM, propagation is along a crystal axis, and this thus avoids any spatial walk-off between the fundamental and second harmonic beams, as well as maximizing the angular acceptance of the phasematching process.

4. Extracavity, single-pass blue generation

This is an extremely simple configuration, requiring the output of the Cr:LiSAF laser to merely be focused into the SHG crystal with a suitable convex lens (fig. 1). The nonlinear crystal of potassium niobate (KNbO₃) was cut for non-critical phase matching at 858nm and 22°C. For optimal SHG performance, the output of the Cr:LiSAF laser was focussed, using a 15mm focal length lens, into a 3mm length of potassium niobate which was anti-reflection coated for the fundamental and second harmonic wavelengths, and mounted on a small thermo-electric cooler for temperature tuning. The laser was tuned to 858nm, and the crystal temperature-tuned to ~22°C for maximum conversion.

In the optimization of the blue-light generation, the performance of the Cr:LiSAF laser was altered in such a way that it produced pulses approaching 200fs at the selected central wavelength of 858nm. These pulses are longer than we would normally expect from this system, but as the second harmonic process requires a high fundamental pulse energy [1], rather than the high peak powers associated with shorter pulses, this is not surprising.

 

Fig. 1. Schematic of the single-pass, frequency-doubled modelocked Cr:LiSAF laser and photograph of the extracavity, single-pass SHG set-up. (PC: polarisation cube; HWP: half-wave plate; HR: high reflector; 1.5%: output coupler; TEC: thermoelectric cooler.)

Download Full Size | PDF

5. Results

Our initial observations concerned the efficiency of the second harmonic (SH) generation process, as well as the temporal and spectral characteristics of both the infrared fundamental and blue SH pulses. Six lenses of varying focal lengths (6.2mm to 25.6mm) were investigated with two lengths (3mm and 5mm) of KNbO₃ crystal. As can be seen from fig.2a, the 15mm lens with the 3mm crystal was found to provide the most efficient SH generation. Up to 11.8mW (36pJ pulse energy, 66W peak power) of blue light was generated for an incident 44.6mW (140pJ, 680W) of fundamental power, with an infrared-to-blue conversion efficiency of 26.5%. Efficiencies as high as 30% were observed (with 10.5mW of blue from only 35mW of fundamental), as well as a slope efficiency at low power of 215%/nJ (fig.2b). The electrical-to-second-harmonic conversion efficiency was around 1%, which we believe is the highest reported efficiency for any femtosecond blue source. This is two orders of magnitude higher than has been achieved with a standard femtosecond Ti:sapphire laser, and at significantly higher pulse repetition rates. As can be seen from fig.2, the blue power has the expected parabolic form, whilst the conversion efficiency is relatively linear with fundamental average power. At higher powers, there is evidence of saturation of these responses, arising from the significant (20%-30%) depletion of the fundamental.

 

Fig. 2. (a) Focusing dependence, and (b) the slope efficiency of the SHG process for the 3mm KNbO₃ crystal.

Download Full Size | PDF

The temporal characteristics of both the fundamental and SH pulses were measured with the now well-established two-photon absorption autocorrelation technique, using a GaN laser diode as a two-photon detector for the measurement of the blue pulses at 429nm [10]. Spectral measurements were made with a standard IST-REES spectrometer.

Until now, pulse durations have not been reported for blue pulses generated in this manner, and it has been assumed that they would lie in the picosecond regime, due to the spectral narrowing effects of phasematching and the group velocity mismatch between the fundamental and second harmonic pulses [2, 11–13]. The striking result presented here is that, by contrast, the SH pulses are in fact still well within the femtosecond domain, and exhibit relatively little frequency-chirp (fig.3). Incident fundamental pulses of 192fs with a time-bandwidth product of 0.34 have produced SH blue pulses of 543fs with a time-bandwidth product of 0.39. A sech² transform-limit of 0.32 is assumed. The fundamental pulses had a bandwidth of 4.3nm, centered at 858nm, and the SH pulses had a bandwidth of 0.44nm, centered at 428.8nm.

 

Fig. 3. Measured intensity autocorrelations (above) and spectra (below) of the fundamental and second harmonic (blue) pulses, using the 3mm KNbO₃ crystal.

Download Full Size | PDF

Figure 3 shows the intensity autocorrelations of both the fundamental and SH pulses for the 3mm KNbO₃ crystal. Although the 3:1 ratio is observed in the trace for the fundamental, this was not achievable for that of the second harmonic, due to the proximity of the laser wavelength (429nm) to the band edge in the GaN laser diode (393nm) being used as the two-photon autocorrelator detector. This effect has previously been reported with this diode at these wavelengths [10], where despite the reduction in the autocorrelation contrast ratio, confidence in temporal measurements was asserted to wavelengths around 415nm. Bearing in mind also that the time-bandwidth product of the SH pulses is as low as 0.39, it can be deduced that the SH pulses are 500-600fs in duration.

Temporal and spectral measurements were also made for the 5mm crystal, which again confirmed the presence of blue femtosecond pulses. In this case, 305fs fundamental pulses with a time-bandwidth product of 0.35, generated 643fs SH blue pulses with a time-bandwidth product of 0.73. The spectral widths were 2.8nm and 0.7nm respectively.

Given that the Cr:LiSAF pump laser was wavelength-tunable, the potential of a tunable blue source using this approach becomes an interesting prospect. However, due to the restrictions imposed on the phasematching process by the acceptance bandwidths (discussed in section 6), there would essentially be no tunability of the blue light without adjusting the crystal parameters.

6. Further Analysis

Confirming the effects already reported [2], we observed an optimal SHG focusing condition that contradicts the well-known Boyd-Kleinmann [14] focusing dependence, as well as a tuning of the blue spectral width with focusing. In addition, a measurement of the temperature and wavelength phasematching acceptance bandwidths was made. The generated blue beam was observed to be of high spatial quality and, with a measured M-squared of 1.8, it was observed to reproduce the TEM₀₀ transverse mode quality of the fundamental beam.

Boyd and Kleinmann calculated that the optimal focusing conditions for efficient SHG in the long-pulse (or cw) limit occurs when the ratio of the SHG crystal length (L) to the confocal parameter of the fundamental beam (b) is 2.84. Figure 4 shows that, rather than optimizing at the Boyd-Kleinmann condition of L/b ~2.84, we observe a clear peak at L/b ~10, corresponding to a significantly tighter focusing condition than the long-pulse theory would suggest.

 

Fig. 4. SHG efficiency vs. L/b focusing ratio for the 3mm crystal. The dashed black line illustrates the Boyd-Kleinmann optimal focusing condition (L/b ~ 2.84), and the solid red line shows our experimental peak at L/b ~10.

Download Full Size | PDF

The shape and width of the blue spectra were observed to tune with focusing. As we moved to tighter focusing, we obtained a blue spectral bandwidth which broadened and became asymmetric. As can be seen from fig. 5, Δλ at 429nm increased from 0.32nm (with a 25.6mm lens) to 1.4nm (with a 2.8mm lens). With increasingly tighter focusing, the bandwidth expands mainly on the long wavelength side, and this effect has been attributed to the effects of non-collinear phasematching [2]. At shorter wavelengths these measurements are limited by the resolution of the spectrometer (0.25-0.3nm), and hence the spectral measurements for the smallest waist sizes in the crystal are likely to be resolution limited.

We then characterized the temperature and wavelength acceptance bandwidths of the phasematching process, under conditions of a femtosecond pulse and hence a relatively broad fundamental spectral bandwidth. The FWHM bandwidths were measured to be Δλ ~2.7nm and ΔT ~5°C, for a fundamental pulse duration of ~190fs and a bandwidth of 4.3nm. These relatively broad bandwidths ease the stability requirements on wavelength and temperature.

 

Fig. 5. Change in blue spectral width with focusing lens (left), and this effect for both crystal lengths (right)

Download Full Size | PDF

Interestingly, both measured bandwidths (fig.6) show a distinctly broad Gaussian profile, rather than a narrow sinc² profile that is typical of the response of phasematched systems [15]. This was confirmed by searching for the characteristic 5% secondary maxima of a sinc-squared function at wavelengths and temperatures outwith the scales shown in fig.6. Only smooth, monotonic reductions in conversion efficiency were observed. If the phasematching equations are solved for NCPM in a 3mm KNbO₃ crystal at 22°C, then acceptance bandwidths are found to be 0.6nm and 2.4°C. These phasematching equations do, however, assume a monochromatic fundamental source. The observed broad responses result from a smearing of the response due to the broadband fundamental, coupled to the effects of group velocity mismatch. The significant results obtained for the two KNbO₃ crystal lengths used are summarized in Table 1.

 

Figure 6. Measured wavelength and temperature FWHM acceptance bandwidths of the KNBO₃ crystal

Download Full Size | PDF

Table 1. Summary of the performance for the two KNbO₃ crystal lengths

View Table

7. Conclusions

In this paper we have reported an efficient femtosecond blue light source that offers compactness, portability and robustness. The combination of a highly efficient, yet robust and inexpensive pump laser, combined with a single-pass conversion scheme that requires no complex or expensive components, leads to a system that is simple and practical. As a result, this configuration would be well-suited to applications where high repetition rates and portability are advantageous, such as in biomedical imaging.

We have already demonstrated a two-diode-pumped miniature, entirely portable, self-contained version of the femtosecond Cr:LiSAF laser on a 22 × 28cm² breadboard, which boasts an electrical-optical efficiency of 4%, and requires only six penlight (AA) batteries as a diode power source for over 12h of stable modelocked operation [4]. With such a simple second-harmonic generation scheme, this demonstrates how a stable, compact, battery-powered femtosecond blue source can now be achieved.