The range of metrology applications for optical frequency combs [1] has been growing since their first demonstration, with their use in space applications now becoming a major focus. Future global navigation satellite systems such as Kepler [2] and applications in satellite time-frequency transfer [3] and formation-flying [4] are driving the development of space-qualified frequency combs. The benefits of obtaining flexible yet precise optical frequency references are not only key for future optical clock systems in space [5] but also key for the laser-based precision interferometers employed, for example, in the Laser Interferometer Space Antenna (LISA) [6] or the Next Generation Gravity Mission (NGGM), as well as for Doppler or absorption lidar [7] applications. Also, the direct generation of ultra-low phase-noise microwave signals [8] for radar or high-speed analog-to-digital (ADC) timing is of increasing interest. To date, only two femtosecond lasers have been tested in spaceflight, both using Er:fiber-laser technology [9–11], which can withstand the extreme accelerations and vibrations of a rocket launch. Despite these exceptional examples of Er:fiber technology, concerns about its vulnerability to long-term radiation-induced fiber damage [12] have led to the consideration of diode-pumped solid-state (DPSS) lasers as an alternative space technology [13]. Specifically, Kerr-lens mode-locked (KLM) DPSS femtosecond lasers are attractive for several reasons. Their short, high-gain crystals allow repetition rates of many GHz from a compact cavity [14,15], and their low dispersion and roundtrip loss offer lower phase-noise than fiber oscillators [16]. Furthermore, the absence of a real saturable absorber eliminates another component that could be vulnerable to aging, leading to Q-switching instabilities [17].

Typically, a KLM laser comprises discrete optical components, co-aligned to high precision by using opto-mechanical mounts, providing the degrees of freedom needed to acquire and maintain mode locking. This flexibility comes at a cost of size, weight, and mechanical stability. Each presents a barrier to utilizing such a laser in a space-borne environment. One solution is to use direct adhesive bonding to fix the components onto a common baseplate, which is a proven integration approach developed for space-borne instruments [18,19] over many years. Unlike solder bonding, using adhesive bonding with a two-component epoxy resin while in thermal equilibrium, and with sufficiently thin epoxy layers, eliminates the need for active optical alignment during the curing phase. Here, we introduce a 1.5-GHz femtosecond laser constructed by this method and compare its performance to that of an identical laser constructed using conventional high-stability optomechanical mounts. While the performance of the two lasers is very similar, the bonded system exhibits higher stability due to the elimination of acousto-mechanical resonances from the components.

The laser oscillator followed the design concept presented in [20]. It was based on a 20-cm-long bow-tie ring cavity, with two $R = - {{20}}\;{\rm{mm}}$ focusing mirrors surrounding the nonlinear gain medium, which was a 2-mm-thick 3%-Yb-doped glass ceramic introduced at a Brewster angle (62.5°). The cavity astigmatism caused by the Brewster-angled gain medium was compensated by a combination of mirror curvature and mirror tilt, and the cavity was optimized to generate a symmetrically round output beam close to the diffraction limit. A plane mirror mounted on a piezo transducer (PZT) allowed the cavity length to be adjusted, tuning the repetition frequency of 1.5 GHz by around 15 kHz. A second plane mirror coated with a Gires–Tournois interferometer (GTI) coating provided a ${-}{\rm{550 }}\;{{\rm{fs}}^2}$ group-delay dispersion compensation for the entire cavity, while also serving as the output-coupling mirror, with a transmission of ${\sim}{0.5}\%$ around 1075–1080 nm. The calculated roundtrip group-delay dispersion of the whole cavity was ${-}{{300}}\;{{\rm{fs}}^2}$. The pump laser was a 980-nm telecom-qualified laser diode, terminated in an FC/PC-connectorized polarization-maintaining single-mode fiber. The laser diode produced a maximum output power of 900 mW. The FC/PC fiber connector facet was re-imaged through one of the curved cavity mirrors and into the ceramic gain medium by using a 2:1 telescope. In Fig. 1(a), a layout of the pump arrangement and the cavity components is presented.

 

Fig. 1. (a) Cavity design. FC, FC/PC fiber connector; L1/L2, collimation/focusing lenses; X, Yb:${{\rm{Y}}_2}{{\rm{O}}_3}$ ceramic; M1, M2, plano-concave pump mirrors; M3, plane mirror; OC, GTI-coated output coupler; PZT, piezo transducer. (b) Prototype built using conventional optomechanics.

Download Full Size | PDF

In order to compare the bonded laser performance with that of the conventional laser design, a prototype was first constructed using regular laboratory optomechanics. The stability of this laser was proven by several months of continuous operation when operated in a clean, boxed environment free of air currents.

The bonded laser was constructed on an aluminum baseplate and exactly copied the design shown in Fig. 1(b). All optics except for pump optics L1, L2, and FC were bonded. While aluminum has a higher coefficient of thermal expansion compared to steel or Invar, it offers $7 \times$ better thermal conductivity, allowing efficient heat dissipation from the Yb:ceramic gain medium. To implement the bonded laser, all components inside the cavity were first held by custom-designed six-degrees-of-freedom jigs to maintain their positions with a high precision. Once stable Kerr-lens mode-locking was obtained, the entire cavity components were bonded and cured in situ. The jigs were then removed, leaving the laser perfectly mode locked, very close to the position of the optimal performance. The repetition frequencies of the two lasers were nearly identical, with the optomechanical laser operating at 1483.47 MHz and the bonded laser at 1510.84 MHz. Figure 2 shows the resulting laser, with the gain medium (X) and all four cavity mirrors fully bonded.

 

Fig. 2. Bonded and operational mode-locked laser cavity, including PZT-mounted mirror for repetition rate locking. Breadboard dimensions are provided to highlight the small footprint of the system (£2 coin shown to indicate scale). Labels follow the same convention as in Fig. 1.

Download Full Size | PDF

The spectra of the optomechanical and bonded lasers were both centered near 1077 nm, with ${-}{{3}}\;{\rm{dB}}$ bandwidths of around 8 nm [Fig. 3(a)]. Both lasers were mode locked in a soliton regime, which was evident by the presence of Kelly sidebands in the spectrum, shown in Fig. 3(a). The absence of double pulsing is confirmed by the radio frequency spectra shown in Figs. 3(b) and 3(c), which show 80 dB signal:noise and an instrument-limited linewidth at 50 Hz resolution bandwidth. A two-photon autocorrelation of the pulses from the bonded laser (operated at 720 mA pump current) is shown in Fig. 4(a) and agrees well with the envelope calculated from the chirp-free 152-fs pulses obtained by the inverse Fourier transforming the pulse spectrum [Fig. 4(b)]. A ${\sec}{{\rm{h}}^2}$ envelope of the same duration shows perfect overlap, strongly indicating that the laser is operating in a fundamental soliton mode locking state.

 

Fig. 3. (a) Optical and (b) radio frequency (RF) spectra of the bonded and optomechanical lasers. The RF spectrum was recorded using an InGaAs photodiode with a nominal bandwidth of 2.5 GHz. (c) RF spectrum of the bonded laser for a 50-Hz resolution bandwidth.

Download Full Size | PDF

 

Fig. 4. (a) Interferometric autocorrelation of the pulses from the bonded Yb:ceramic laser oscillator, and (red) the envelope calculated from the inverse Fourier transform of the pulse spectrum (no additional phase). (b) Pulse intensity corresponding to the envelope in (a). A ${\sec}{{\rm{h}}^2}$ profile with a full width at half maximum (FWHM) duration of $\Delta {\tau _{{\rm{FWHM}}}} = 152\;{\rm{fs}}$ is shown for comparison.

Download Full Size | PDF

Both systems had a similar laser threshold, with continuous wave (CW) oscillation beginning for pump currents above 90 mA and with stable mode locking observed at currents ${\gt}{{300}}\;{\rm{mA}}$. As Fig. 5 shows, both lasers exhibited hysteretic behavior for currents between 300 and 500 mA and had similar total output powers of up to 15 mW, for a total electrical power of 1.5 W. In this regime, the optomechanics-based laser operated in a bidirectional mode, while the bonded laser performed in a unidirectional mode. This contrasting behavior is explained by small differences in the intracavity focusing, which is sensitive to the position of mirror M2. In the optomechanical laser, perhaps due to slightly tighter focusing, KLM action with a single resonant pulse is overdriven at all values of intracavity power, so only bidirectional operation is observed. For the bonded laser, KLM is not overdriven for currents from 470 to 570 mA, allowing unidirectional operation, which, once initiated, can persist when the current is reduced below 470 mA. In the transition regions (Fig. 5, shaded) the lasers showed bandwidth fluctuations and CW breakthrough, stabilizing when they were pumped with higher power and were operating in a bidirectional mode and consistent with overdriving KLM in each system.

 

Fig. 5. Laser output power as a function of the pump current for (a) the laser constructed using optomechanics and (b) the bonded laser.

Download Full Size | PDF

We compared the lasers’ performance in the frequency domain, operating them both in a stable bidirectional mode locking state. A phase-locked loop was used to control the repetition rate, as shown in Fig. 6(a). An InGaAs photodiode detected the pulse-repetition frequency, and its output was amplified and mixed with a 1.5-GHz reference to give an error signal. This was low-pass filtered at 2 MHz before entering a proportional-integral (PI) amplifier, whose output actuated the intracavity PZT to control the repetition rate. After phase locking the repetition rate, the optomechanical laser was locked at 1483.471 MHz and measured to be highly stable with no fluctuations within the detectable range of the equipment, and the repetition rate of the bonded laser was locked at 1510.838 MHz. As can be seen from Figs. 6(b) and 6(c), the integrated residual phase noise (1 Hz–1 MHz) of the optomechanical laser was 5.2 mrad, while for the bonded laser was 3.0 mrad. For both measurements the noise contributed by the detection system was 0.088 mrad. The noise spectrum of the bonded laser had a smooth, single-peaked characteristic, while the optomechanical laser had structure in the acoustic band, which we attribute to unique resonances from the different component mountings. The 3-mrad phase noise of the bonded laser’s repetition rate can be compared with the performance of a contemporary Er:fiber laser comb, which, in a similar measurement, showed 0.34-mrad phase noise at a repetition frequency of 101 MHz [21], implying comparable timing jitters of 300–500 fs (1 Hz–1 MHz range). We expect that the phase noise in the 100–1000-Hz band could be suppressed further by using a piezo driver with a higher current than was available to us.

 

Fig. 6. (a) Phase-locked loop for repetition rate control. REF, reference signal; LPF, low-pass filter; PI, servo controller; KLM, Kerr-lens mode-locked laser; PD, photodiode; RFA, radio frequency amplifier. (b) Residual phase noise (blue) and instrument noise floor (red) of the locked optomechanics-based laser. (c) As (b), but for the bonded laser.

Download Full Size | PDF

The drifts in the repetition rates (${f_{{\rm{rep}}}}$) of the lasers were compared to assess the improvement in passive stability provided by the bonded laser. Both lasers were operated in similar laboratory environments (${\pm}{{1}}^\circ {\rm{C}}$), and the feedback loops were disconnected to allow free-running operation. By monitoring the frequency of the mixer output, the frequency drifts of the two lasers were counted sequentially over an 8-h period with a gate time of 100 ms (Fig. 7). Over the measurement time, the standard deviations of the roundtrip cavity lengths (repetition rates) were 143 nm (1.08 kHz) for the bonded laser and 327 nm (2.45 kHz) for the optomechanical laser, demonstrating that the drift of the optomechanical laser was around ${2.3} \times$ greater than the drift of the bonded laser. While actual temperature differences in the laser environments could explain this, another fundamental performance difference was observed. Occasional jumps occurred in the repetition rate of the optomechanical laser (Fig. 7). We suggest these are caused by the sudden release of internal stress within a mount. When the temperature changes, the mechanical stress accumulated within a mount can be randomly released, leading to a shock, which causes a small but sudden excursion of the repetition rate. Since bonding eliminates the effects of the optomechanical components, no similar features are visible in the bonded-laser performance.

 

Fig. 7. Cavity-length and repetition-rate drifts of the free-running optomechanical and bonded lasers; shading shows ${\pm}\;1\sigma$.

Download Full Size | PDF

Several steps are required to qualify the system for space. The small quantum defect of Yb:${{\rm{Y}}_2}{{\rm{O}}_3}$ means that heating of the gain crystal is small, and the direct contact with the aluminum baseplate allows heat to be conducted away and exchanged convectively with the surrounding air. For operation in a purely conductive environment, as is usual in satellites, active temperature control will be required independent of the breadboard material; its performance would be verified in a thermal vacuum test campaign. After this, a dedicated vibration and shock test would be performed to prove the mechanical resilience to launch accelerations and shocks. Only minor issues for the cavity and pump components are expected due to the proven integration technology. A major milestone will be the radiation test, providing evidence of the expected superior radiation hardness of Kerr media compared to fiber systems.

The system's repetition rate exceeds those of typical fiber frequency combs by roughly ten times, offering easier comb-mode selection for absolute metrology purposes, with routes to higher frequencies already demonstrated [14]. Direct locking to an atomic reference may therefore be possible, for example by frequency doubling to access an ${{\rm{I}}_2}$ transition [22], or even by spectral broadening to access a Rb transition [23]. Carrier-envelope offset frequency control is also expected to be possible through pump-power adjustment. Si₃N₄ waveguides can provide the prerequisite octave-spanning supercontinuum [24], allowing f-to-2f locking of $1{\text{-}\mu{\rm m}}$ femtosecond lasers atcGHz repetition frequencies [25].