1. Introduction

In recent decades ultra-narrow-linewidth lasers attracted growing interest of the opto-electronic industry and of both, the applied and the fundamental scientific community. The application fields include coherent optical free-space communication protocols [1], high precision metrology like optical frequency standards [2] or light- [3, 4] and matter-wave interferometers [5].

Modern semiconductor lasers cover nearly the entire optical spectrum and are available at low cost and in compact packages. However, their spectral properties are not sufficient for many applications. A common approach to reduce the frequency noise of a semiconductor laser is to use resonant optical feedback from an external optical element e.g. a Bragg grating [6] or a Fabry-Pérot (FP) cavity [7, 8]. The first technique allows for reduction of the optical linewidth (FWHM) down to 47 kHz [9] compared to tens of MHz of a free running, single mode diode laser. Due to narrow spectral bandwidth of the feedback from a FP-cavity, the second concept reduces the laser linewidth even further, however, achieving a stable single mode emission is challenging. A combination of the two methods, also known as grating-enhanced external cavity diode laser, has been demonstrated [10–12]. In these cases the two tasks of wavelength selection and generation of optical feedback are carried-out by two different optical elements, i.e. by a grating and a FP-cavity, respectively. To avoid the complexity of the corresponding setup one can utilize the internal grating structure of a DFB-diode-laser for wavelength selection. Two implementations of this concept have been demonstrated so far. The first is based on a whispering gallery-mode (WGM) micro-ring resonator [13] and the other on a macroscopic high-Q, ultra-low expansion (ULE) cavity [14].

In this paper we present and characterize a DFB-diode-laser with resonant optical feedback from a mesoscopic, monolithic, confocal Fabry-Pérot resonator with a finesse of approx. 200. The choice of relatively moderate finesse together with confocal geometry greatly simplifies the coupling of light into the cavity and makes it insensitive to mechanically or thermally induced misalignments. The simplicity and small size of the confocal cavity also allows assembling our laser module much more compact than the setup presented in [14]. Nevertheless, we were able to maintain high reduction of the laser frequency noise. Although the linewidth of the external cavity applied in this work is larger by almost three orders of magnitude than the linewidth of the WGM-based resonator [13], the laser features a white noise floor that is smaller than the noise floor of the WGM-based laser.

2. Setup

The experimental setup is shown in Fig. 1. We use a 1.5 mm long, single quantum well AlGaAs distributed feedback (DFB) diode laser [18] designed and processed at the Ferdinand-Braun-Institute. The design wavelength is 780 nm. Both facets of the laser diode are anti-reflection (AR) coated, which provides two advantages. First, the feedback level, defined as the ratio of the electric field amplitude of the light reflected back from an external cavity to the field amplitude reflected internally from the laser facet, is increased. Second, approximately the same amount of optical power is emitted on either side of the diode. This ensures that sufficient amount of optical power is available for optical feedback, measurements and applications.

 

Fig. 1 Laser schematic and measurement setup (a); transmission spectrum of the confocal cavity in the absence of coupling to the laser diode (b).

Download Full Size | PDF

An LDC-3724C (ILX Lightwave) diode laser controller is used to drive the laser current and to stabilize the temperature of the laser mount. The laser’s rear output is focused with an aspheric lens (Thorlabs C150TME-B, f = 2 mm) and coupled into a mesoscopic, monolithic, confocal Fabry-Pérot cavity with a mirror curvature of 10 mm and power reflectivities of 0.97 and 0.9997. The linewidth (FWHM) of the transmission spectrum of the cavity is 70 MHz. The cavity is tilted by 15° with respect to its optical axis such that only the resonant feedback is re-injected into the DFB-laser, while the non-resonant light is reflected [7]. If the frequency of the free running DFB-diode is close enough to a resonance frequency of the confocal cavity, the laser locks its frequency to the cavity resonance. The locking range of the laser strongly depends on the feedback level and varies from 400 MHz up to several GHz. An optical attenuator can be inserted between the laser chip and the external cavity to adjust the feedback level. The optical power transmitted through the cavity is monitored with a photo-detector (PD1).

We employ a self-delayed homodyne beat note measurement to determine the frequency noise power spectral density (PSD) and the laser linewidth. The output beam of the DFB-laser is collimated and guided through a Faraday isolator (Qioptiq DLI, 60 dB) to suppress parasitic optical feedback. An acousto-optic modulator (Intra-Action ATM-804DA2B) shifts the laser frequency by 78 MHz. After passing an optical delay line (Fibercore SM-800, single mode, 2 km physical length) the frequency shifted beam is superimposed with the unshifted zeroth order beam in a fiber X-coupler. Both beams interfere on a fast photo-detector (New Focus, 1554-B, 12 GHz). We record the resulting beat note signal with a RF-spectrum analyzer (Rohde und Schwarz, FSW) and determine the frequency noise PSD with the method described by Schiemangk et al. [15].

3. Results

Figure 2 shows the measured frequency noise PSD of the DFB-diode laser with resonant optical feedback, of the DFB-laser without feedback, and of a narrow linewidth ECDL [9]. Resonant optical feedback from the cavity reduces the frequency noise PSD of the DFB-laser by more than five orders of magnitude at high Fourier frequencies (> 10 kHz). The noise level at these frequencies is three orders of magnitude smaller than that of the ECDL.

 

Fig. 2 Measured frequency noise power spectral density: free running DFB laser (red), ECDL acc. to [9] (blue), and DFB laser with resonant optical feedback (black). Peaks at multiples of 100 kHz are measurement artefacts of the delayed self-homodyne measurement.

Download Full Size | PDF

The white frequency noise level is S₀ = 5 Hz²/Hz which corresponds to a Lorentzian linewidth of Δν = S₀π = 15.7 Hz. According to [13], for lasers suffering from generic 1/f frequency noise the effective linewidth Δν_eff can be estimated from ${\displaystyle {\int }_{\Delta {\nu }_{eff}}^{\infty }ℒ\left(f\right)df=\frac{1}{2\pi }}$, where $ℒ\left(f\right)$ is the phase noise spectral density [19]. Numerical integration yields Δν_eff = 1.4 kHz for the laser presented here. The white noise begins already at 20 kHz, which eases cancellation of noise at lower frequencies by means of an active frequency stabilization, for instance, to an atomic or molecular reference. The contributions to noise at low Fourier frequencies have various origins. First, noise results from acoustic vibrations. By assembling the setup we neither did pay particular attention to its mechanical stability nor to isolation from air turbulences. Integration of the laser on an optical micro-bench [9] should provide improvement of spectral stability at acoustic frequencies. Second, there is a contribution to frequency noise from a network-powered laser current driver, manifested as peaks at 50 Hz and higher harmonics (not resolved in Fig. 2). Last, frequency noise is also induced by acoustic and thermal disturbances to the 2 km long optical fiber delay line [17]. To quantify the contribution to frequency noise from the delay line we plan to realize a second laser and to directly measure the frequency noise PSD of the beat note signal of the two lasers.

To verify the influence of feedback level on the spectrum of the laser we placed two different attenuators (single-pass transmission factors 0.69 and 0.3) between the diode and the external cavity. The resulting RF-power spectra of the beat note signal recorded by the fast photo-detector are shown in Fig. 3. One recognizes a dramatic reduction of the linewidth of the laser as the feedback strength increases. We mention that due to the uncertainty in the coupling efficiency back to the diode it is difficult to estimate the absolute feedback level. However, assuming a coupling efficiency of 80% and relying on the reflectivities of the cavity facets, we can provide an upper limit for the amount of light injected back into the diode to be 1, 0.5, and 0.1% of the power emitted towards the cavity for the transmission factors of 1, 0.69, and 0.3 respectively.

 

Fig. 3 RF-Power spectra of the beat note signal for the free running DFB-laser (black) and with resonant optical feedback at different feedback levels. The green and blue curves are recorded with an attenuator placed between the DFB-laser and the external cavity (single-pass transmission factor 0.3 and 0.69 respectively). The red curve is recorded without an attenuator.

Download Full Size | PDF

In the strong feedback regime we can no longer directly rely on beat note spectra – if the coherence length of a laser exceeds the length of delay line, the beating electrical fields are correlated and a delta-peak appears at the center of the spectrum. Moreover, a modulation pattern with a periodicity equal to the reciprocal delay time (100 kHz in our case) appears [16, 17] which, together with the delta-peak, makes it difficult to estimate the linewidth directly from the spectrum.

Finally, we mention that we carried out all measurements at 781.6 nm and at an operating temperature of 31°C. Reducing the temperature of the diode down to 5°C enables us to operate the laser at the wavelength of Rb D2 transition, i.e. 780.2 nm. Moreover the method presented here is extendable to any wavelength at which DFB-diodes are available.

4. Discussion and conclusion

In summary, we presented a laser system based on a DFB laser with resonant feedback from an external, compact, cavity. We analyzed its spectral stability and compared it to the stability of similar systems described in the literature (Table 1). The laser with resonant optical feedback from a macroscopic ultra-high-Q ULE-cavity [14] provides the narrowest Lorentzian (intrinsic) and FWHM linewidth. We emphasize however that the Q-factor of the external cavity is not the only critical factor determining the laser spectrum. The laser presented here features a moderate-Q cavity, however it provides an effective linewidth that is similar to that of the WGM-cavity laser [13], and a Lorentzian linewidth smaller by one order of magnitude. We attribute this fact to the high feedback level realized in the present setup due to the AR-coating of the laser facet. In contrast to the ULE laser [14], the laser presented in this paper can be assembled on a footprint 1 × 5 cm² (diode, coupling lenses and external cavity). Last, with the present setup we provide more than 50 mW optical output power.

Table 1. Basic properties of the lasers and cavities described in this paper and in [13,14]

View Table