1. Introduction

Since the early 1990s passively mode-locked fiber lasers have become an essential part of laser systems used for industrial, medical, and research applications [1]. A large selection of fiber components is now commercially available, facilitating the rapid evolution of research on fiber lasers and the expansion of their applications. One of the most recent developments is passively mode-locked dual-color and dual-comb fiber lasers. These devices combine the robustness of fiber laser technology with the high passive mutual coherence that is achieved due to common mode noise suppression when both combs are generated within the same cavity. Such systems are ideal for applications such as dual-comb (DC) spectroscopy; a versatile technique combining the advantages of conventional broadband spectroscopy and tunable laser spectroscopy [2]. DC spectroscopy eliminates the need for complicated and expensive detector assemblies such as virtually imaged phase arrays [3] or Fourier transform spectrometers [4], since the optical spectra are down-converted to the radio-frequency range and can hence be detected using simple photodiodes and a radio frequency (RF) analyzer. DC systems traditionally consist of two separate mode-locked lasers delivering pulse trains with slightly different pulse repetition rates. In the spectral domain, those pulse trains correspond to two frequency combs with slightly different comb line spacing, which leads to the generation of a beating comb in the RF range. In order to ensure mutual coherence between the two optical combs, the lasers need to be actively stabilized to each other, which can quickly become a complex and cumbersome task. Much effort has been spent to reduce the complexity of DC setups, for example by phase-locking the two frequency combs to an external cavity diode laser [5] and using adaptive sampling techniques [6–9].

One of the most promising approaches to significantly simplify DC setups is the generation of two pulse trains using a single laser cavity. The passive mutual coherence leads to common-mode noise cancellation in the down converted RF-comb [10], which represents a clear advantage over actively locked systems [11]. Examples of such single-cavity DC schemes involve the separation of the pulse trains by direction of travel [12,13] polarization [14,15] or branched optical paths using a birefringent crystal [10,11] and most recently dual-color fiber lasers [16–18]. Due to their robustness, all-fiber single cavity dual-color/dual-comb lasers are promising candidates to enable spectroscopic measurements outside the laboratory, thus making them the subject of ongoing studies. In the last few years, many different ways to implement a dual-color fiber laser were reported [19–22]. Zhao et al. described a carbon nanotube mode locked dual-color erbium doped fiber laser [21]. In their case, dual-color operation was induced by using polarization dependent filtering and wavelength dependent birefringence while exploiting the gain profile of erbium doped fibers, which exhibits two spectrally separated peaks under certain pumping conditions. The approach used by Zhao et al. to generate a dual-comb out of a dual-color laser was to split the two pulse trains outside of the laser cavity, then to amplify them, broaden them inside a nonlinear fiber to generate a spectral overlap and spatially overlap them again, hence obtaining a fully coherent down-converted comb [16]. Subsequently, Liao et al. have presented a thulium-doped dual-color dual-comb laser extending this technique deeper into the mid-infrared spectral region [17]. The thulium laser was mode-locked using a nonlinear amplifying loop mirror (NALM). Dual-wavelength operation was implemented by using the artificial gain filter effect induced by the NALM. Spectral overlap was generated similarly to [16] by nonlinear broadening outside of the cavity [17]. More recently, a further step towards a system for fieldable spectroscopy has been taken by Li et al. [23]. They developed an all-polarization-maintaining dual-wavelength mode-locked erbium fiber laser, where dual-wavelength operation was ensured using a Sagnac loop filter. All dual-color laser schemes mentioned above were designed for specific types of laser oscillators. A more general way to generate a dual-color laser is the implementation of spectral filtering in the laser cavity by using fiber Bragg gratings with different transmission wavelengths. Reported results achieved with this technique reached up to triple-wavelength operation centered around 1550 nm with a full width at half maximum (FWHM) of about 0.5 nm with a repetition rate < 7 MHz [24]. Although this method seems very promising, the spectral coverage as well as the repetition rate would have to be increased in order to be useful for dual-comb spectroscopy. To get a broader FWHM and a higher repetition rate, the spectral filtering would have to be perfectly matched to the mode-locking dynamics of the laser cavity.

In this work, we present a particularly flexible solution to generate a dual-color laser out of a single laser cavity using a mechanical spectral filter. Mechanical spectral filtering as a method to generate dual-color lasers was first implemented in the early 1990s [25] in titanium sapphire lasers, which have an extremely broad gain bandwidth (~450 nm). Knox et al. [26] furthermore used the spatial dispersion in the grating compressor to measure the intra-cavity dispersion by introducing a slit as a spectral filter and forcing the laser to operate in various narrow wavelength bands.

Here, we implement the concept of mechanical filtering in a fiber laser and show its potential to generate a single cavity dual-color laser. The method offers two key features: dynamical adjustment of the spectral separation between the two pulse trains and tuning of the difference in repetition rates. To the best of our knowledge, these features have not been available in previously reported single-cavity dual-color laser schemes. The spectral filtering technique is demonstrated here in a NPE-mode-locked ytterbium (Yb) fiber laser. The outlined technique is however independent of polarization evolution dynamics in the laser and can potentially be implemented in a large variety of different mode-locked lasers, especially also in polarization-maintaining fiber lasers. Our method does not rely on exploiting specific transmission peaks in the laser gain profile and is also independent from other mode-locking dynamics.

2. Experimental setup

We have implemented our mechanical filtering concept using a blade-shaped beam block in a NPE-mode-locked ytterbium (Yb)-fiber laser. The physics of NPE-mode-locking is well understood and has many implementations; we refer to Chong et al. [27] for an extensive review. The basic principle consists of exploiting intensity-dependent polarization changes inside a non-polarization-maintaining fiber. By manipulating the polarization of the intra-cavity light using wave plates and a polarizer, the laser is set up in a state in which low peak power light (continuous-wave operation) experiences higher losses than high peak power light (pulsed operation), leading to stable ultrafast laser pulse generation. We have built a laser similar to the approach described in [28], consisting of a free-space and fiber section (Fig. 1). The fiber section of the laser consists of a wavelength division multiplexer (WDM) for coupling the pump light into the cavity, a 30 cm-long ytterbium-doped gain fiber (LIEKKI Yb1200-4/125) and 8 m of non-polarization maintaining single mode fiber (SMF) (Thorlabs SM980-5.8-125) enabling the nonlinear polarization evolution. The calculated group delay dispersion caused by the fiber section of the cavity amounts to ~0.16 ps². The light is coupled to the free-space section with two collimators. We ensured full control over the polarization by using a quarter wave plate (QWP) and a half wave plate (HWP) in front of the collimators and by inserting a polarization beam splitter cube (PBS), which acts as the primary output coupler of the laser (output A in Fig. 1).

 

Fig. 1 Setup of the single-laser source dual-color cavity. Due to the mechanical spectral filter the laser oscillator emits two pulse trains with different repetition rates and center wavelengths. The optical spectra of the two pulse trains coupled out at the PBS are broadened by self-phase modulation in the fiber portion of the laser cavity.

Download Full Size | PDF

A Faraday isolator ensures unidirectional operation. A transmission grating compressor is used for dispersion compensation (Wasatch Photonics Gratings, 800 l/mm, angle of incidence 24.3° @ 1030 nm). Inside the grating compressor, the spectral components of the intra-cavity light are spatially dispersed (see Fig. 1). We block a narrow part of the spectrum by introducing a small razor blade into the collimated beam section of the grating compressor, thus dividing the spectrum into two broadband regions. We show that it is possible to independently mode-lock the laser within these two spectral regions, hence creating two mode-locked lasers at slightly different repetition rates. Due to the spatial distribution of the spectral components, it is straightforward to manipulate the spectrum by simply translating the beam block. In addition to the position, also the width of the spectral filter can be tuned by rotating the razor blade. The effective width of the beam block can be tuned in this manner from 0.2 mm and 1.5 mm.

The laser operates with a repetition rate around 23 MHz. This rather low repetition rate allows us to maintain sufficient peak power for simultaneous mode-locking of the two different pulse trains. Starting dual-color operation at a higher repetition rate would require more than the currently available 300 mW of pump power. However, once started, dual-wavelength operation in the current configuration can be sustained with 80-100 mW of pump power. Since the gain maximum is situated around 1030 nm, any spectral overlap generated in this spectral region by self-phase modulation prior to the gain fiber will be strongly amplified, leading to competitive behavior between the two spectrally separated pulse trains. To avoid this issue, we kept the length of the single mode fiber between collimator 2 and the WDM (see Fig. 1) as short as possible. Adjustment of the waveplates [28] as well as the spectral filter allowed us to achieve independent and simultaneous mode-locking of both spectral regions.

3. Results

In Fig. 2, we show the optical and the RF spectrum of the light coupled out of the cavity. The optical spectra shown in Fig. 2(a) were recorded using a Fourier-transform infrared spectrometer with a resolution of 0.3 nm. The spectra were measured at two different locations in the setup (see Fig. 1): the solid line corresponds to output A (after the single mode fiber), while the dashed spectrum was measured at output B, i.e. directly after the spectral filter. At output B, the p-polarized light is coupled out using a pellicle beam splitter (Thorlabs BP145B3). The pellicle has a reflectivity that increases from 22% at 1010 nm to 31% at 1050 nm. The remaining light then propagates through the Faraday isolator as well as a HWP and QWP, changing the polarization state of the light from linear (p-polarized) to elliptical before the gain fiber. At output A, the light exits the cavity via a polarizing beam splitter, with an output coupling rate that depends on the exact settings of the waveplates. Compared to the spectrum obtained at output B, the FWHM of the spectrum recorded at output A is broadened by 3 nm and 2 nm for the short-(red) and long (black)-wavelength section respectively. The broadening is induced by self-phase modulation occurring in the fiber section of the cavity.

 

Fig. 2 Optical and radio-frequency spectra of the dual-color laser. (a) The dashed line shows the spectrum directly after spectral filtering (output B, see Fig. 1), the solid line shows the spectrum exiting the cavity via the polarizing beam splitter (output A, see Fig. 1). (b) Radio frequency spectrum around the repetition rates with a span of 2 kHz and a resolution bandwidth (rbw) of 3 Hz. The strong signal (> 80 dB) and absence of side-peaks indicate clean mode-locking for both pulse trains. (c) Schematic sketch for measuring the single pulse trains independently from each other by spectrally separating the two pulse trains using a grating after the cavity output and focusing each pulse train individually on the photo diode while blocking the other one. (d),(e) 500 MHz span showing the repetition rate of the single pulse trains and their respective harmonics.

Download Full Size | PDF

Figure 2(b) shows an RF measurement obtained by simultaneously focusing both pulse trains onto a photodiode. The two repetition rate signals are 1.25 kHz apart in this example (see Fig. 4 for tuning results) and show strong signal-to-noise ratio of > 80 dB at 3 Hz resolution bandwidth. In order to verify that each spectral region corresponds to an independent, cleanly mode-locked pulse train, we used an additional grating to spatially separate the two spectral regions outside of the laser cavity (see Fig. 2(c)). Figure 2(d) and (e) show the repetition rates of the single pulse trains and their respective harmonics. The measurement was done with a fast (5 GHz) photodiode.

In Fig. 3 we show the temporal drift of the two repetition rates and their difference over 30 min of continuous dual-color mode-locking operation. Note that the system is completely free-running and is neither boxed nor optimized for mechanically stability. The data was recorded using a microwave spectrum analyzer set to 1 Hz resolution bandwidth.

 

Fig. 3 Temporal drift of ∆f_rep, f_rep,1 and f_rep,2 over 30 min. The measurement was recorded with a microwave spectrum analyzer set to a frequency resolution of 1 Hz (1 read-out every 20 s).

Download Full Size | PDF

Previously demonstrated single cavity dual-color lasers featured a fixed difference ∆f_rep between the two pulse repetition rates, which is usually defined by the laser design and cannot be tuned. In our case however, changing the grating spacing (and thus the amount of dispersion) allows us to tune the difference in the repetition rate from 3 kHz down to 650 Hz (Fig. 4). The cavity end mirror is mounted on the same stage as the second grating. Changing the grating separation therefore also has an influence on the repetition rates f_rep,1 and f_rep,2. Decreasing the grating separation leads to an increase of the repetition rate and to an increase of ∆f_rep, as it is expected when working in a regime where the net total cavity group delay dispersion (GDD) is positive. The tuning range of the grating separation was limited by the travel range of the translation stage used ( ± 12.5 mm around the nominal separation of 45 mm) corresponding to an intra cavity dispersion change of ± 41,600 fs² around a nominal value of + 41,800 fs² (total GDD fiber + grating), meaning the complete tuning was done without crossing the zero-dispersion point.

 

Fig. 4 Tuning of ∆f_rep = f_rep,1 - f_rep,2 by changing the grating separation, with f_rep,1 being set to zero in all data sets for straight-forward comparison. The offset values corresponding to the values of f_rep,1 for each configuration are listed in the legend.

Download Full Size | PDF

Furthermore, by rotating the beam block, which is equivalent to changing its thickness, we were able to vary the width of the spectral cut between the two individually mode-locked spectra, see Fig. 5. Although the grating separation is kept constant, we observe a slight change in Δf_rep of 120 Hz between the configuration with the narrowest blade width (Fig. 5, blue curve) and the widest (Fig. 5, red curve) caused by the central wavelength shifts. Ultimately, the spectral tuning is limited by the bandwidth of the gain medium, in our case the Yb:fiber, whose emission and absorption spectra have been included in Fig. 5 for comparison.

 

Fig. 5 Tuning of the spectral separation by rotation of the razor blade, which corresponds to changing the effective width of the beam block from 0.2 mm to 1.5 mm.

Download Full Size | PDF

4. Prospects for dual-comb generation

The intra-cavity broadening due to self-phase modulation mentioned in the previous paragraph may be usable to generate a dual-color dual-comb without any external spectral broadening. However, to achieve an aliasing-free down-converted radio frequency comb, the spectral overlap ∆ν needs to be chosen in order to fulfill the following condition [29]:

With a nominal repetition rate f_rep,1 around 23 MHz, a difference between the two repetition rates of ∆f_rep = 1 kHz and a center wavelength of 1025 nm, the non-aliasing dual-comb spectral bandwidth is calculated to be currently about 0.26 THz, corresponding to only 1 nm at a center wavelength of 1025 nm. By increasing the nominal repetition rate (which implies supplying more pump power) this issue can be avoided.

Furthermore, nonlinear interactions between the two pulse trains can occur inside the fiber-part of the cavity. This type of intra-cavity dynamics has been studied recently using dispersive Fourier transform technique in a dual-color soliton mode-locked erbium fiber laser [30]. The intra-cavity pulse collision may cause a spurious signal in the recorded dual-comb interferogram. However, this signal can be separated from the center burst using a delay stage as described by Liao et al. [17].

5. Conclusion and outlook

We have demonstrated a versatile and easily implementable method to generate a dual-color mode-locked laser from a single fiber laser cavity. By manipulating the cavity losses for the center part of the gain spectrum using a mechanical spectral filter, we obtained two pulse trains with slightly different repetition rates originating from the same laser cavity. In our experiment, the spectral filtering was implemented in a NPE laser cavity by introducing a beam block in the collimated beam section of the grating compressor. By changing the grating spacing, the difference of the repetition rates could be tuned in the range of 650 Hz to 3 kHz. Although a NPE laser was used in this demonstration, the polarization evolution is not part of the dual-color laser generation process (in contrast to other schemes using the polarization as a selection criteria to generate two pulse trains [16,17,21]). The spectral filtering itself may be extended to lasers without a grating or prism compressor by using fiber Bragg gratings or dielectric coatings. Hence, we believe that the method presented here can be extended to basically any kind of passively mode-locked fiber laser, including state-of-the-art all-polarization-maintaining configurations, provided that the gain bandwidth is large enough to support two spectrally distinct pulses. The demonstrated scheme is not limited to any specific wavelength region; in particular, it is applicable to the mid-infrared, where compact dual-comb setups for molecular spectroscopy are strongly desired for many applications.