Laser diode linewidth measurement using high-Q spherical microresonators P. Pineda-Vadillo, M. Lynch, C. Charlton, S. Tessarin, J.F. Donegan and V. Weldon A novel technique to measure the emission linewidth of laser diodes (LDs) utilising whispering gallery modes (WGMs) in high-Q micro- spheres (diameter ≃ 200 mm) is proposed and demonstrated. Laser light from a distributed feedback (DFB) LD operating at l ≃ 1537.1 nm is coupled into silica glass microspheres of Q ≃ 10 7 (esti- mated experimentally). WGM spectra are recorded and then used to estimate the DFB LD emission linewidth, yielding a value of 3.8 MHz. The estimated value is in good agreement, within experimen- tal error, with a linewidth value of 6.0 MHz independently measured using the self-homodyne technique, strongly supporting the validity of this novel method. Introduction: Optical microspheres are a special case of optical resona- tors within which light can be guided in a low-loss way in so-called whispering gallery modes (WGMs) [1]. Light coupled into the micro- sphere under appropriate conditions is confined by total internal reflec- tion (TIR) and guided very close to the spherical surface. The quality or Q-factor is determined by the losses which are mainly dependent on the surface quality of the microsphere. In high-Q microspheres (high surface quality, low losses), the measured width of the Lorentzian-shaped WGM resonances becomes sensitive to the linewidth of the source when the latter is greater than a significant fraction of the WGM linewidth. This is essentially the basis of our laser linewidth measurement technique which is similar to that using Fabry-Pe ´rot cavities [2, 3]. Experimental method: The experimental arrangement is illustrated in Fig. 1. The emission from two types of singlemode lasers, a slotted dis- crete mode (DM) [4] and a DFB LD, was coupled, in turn, into micro- sphere resonators fabricated by melting the tip of a singlemode silica glass fibre (refractive index n ¼ 1.47) using a high power CO 2 laser [5]. The DM laser has a much lower linewidth than the DFB and the WGM and so is used to measure the WGM spectra directly. The two LDs were used in conjunction with a polarisation controller (POL CTRL) to select between excited TE/TM WGM bands and optimise their intensity. Wavelength modulation was achieved via current injec- tion with a ramp signal of frequency f MOD and voltage V MOD . A fraction (5%) of each LD output was, in turn, directed to a wavemeter (WAV) and the remaining light was prism-coupled into the microsphere. Piezo con- trollers and high-precision position stages allowed for coupling optimis- ation. Scattered light from the microsphere surface was collected with a multimode (MM) fibre and focused onto a photodetector (PD1), while transmitted light through a prism was collected using a second photode- tector, PD2. The weak signal from PD1, depicting the microsphere modes, was amplified to improve its signal-to-noise ratio (SNR), and WGM profiles recorded using a storage oscilloscope (OSC) or PC. PC OSC microsphere PD1 PD2 out LIA ref V MOD , f MOD in I + LAS WAV POL CTRL AMP MM fibre Fig. 1 Scheme showing experimental setup used for WGM based linewidth determination Frequency locking loop: Thermally induced changes in the micro- sphere dimensions result in a drift in the resonant WGM frequency n 0 , and can be regarded as effectively being low-frequency noise sources. To lock the laser frequency to the resonant value a feedback loop was implemented via the laser current controller (I) according to Fig. 1. A top-of-fringe method [6] was used with the help of a lock-in amplifier (LIA) referenced to the first harmonic of the modulation frequency f MOD . If several WGMs are simultaneously excited in the microsphere this loop will tend to shift and lock the laser frequency to the strongest mode within the band. Frequency locking to n 0 effectively suppresses the low-frequency drift of the system so the spectral shape significantly depends on the laser linewidth (assuming a constant micro- sphere Q factor during the measurement time). In addition, to avoid dis- tortion of the spectra, a high modulation frequency was chosen. If the wavelength is swept too slowly, the absorbed optical power is enough to shift the resonant frequency as the laser scans across the WGM [7]. In that case the WGMs, as seen in the rising and falling sides of the tri- angular wavelength ramp, are modified (Fig. 2a, top inset: undistorted spectra at f MOD ¼ 1.124 kHz; bottom inset: distorted spectra at f MOD ¼ 2.2 Hz). –80 0 0.1 0.2 0.3 0.4 0.5 a b DFB sphere modes, f MOD = 1.12 kHz DM sphere modes, f MOD = 1.12 kHz f MOD 2.2 Hz f MOD 1.124 kHz data fit normalised photodiode voltage, V 0 0.1 0.2 0.3 0.4 0.5 photodiode voltage, V –60 –40 –20 detuning from mode centre, MHz 0 20 40 60 80 –80 –60 –40 –20 detuning from mode centre, MHz 0 20 40 60 80 0 0 150 300 600 1200 frequency interval, MHz data fit Fig. 2 Experimental results for WGMs excited at f MOD ¼ 1.124 kHz using DM and DFB LD including corresponding fits Distortion of intensity profile due to thermal effects at two modulation frequencies is depicted in insets of Fig. 2a over one modulation period a Using DM b Using DFB WGM measurements: The modulation frequency was set at f MOD ¼ 1.124 kHz for both LDs and a high V MOD was first chosen so the wave- length modulation interval exceeded the WGMs’ free spectral range (FSR). The LDs’ current and temperature were optimised to find the strongest WGM band and the locking loop then closed. Polarisation and coupling distance were optimised to give maximum SNR and set at the same values for both lasers to enable accurate comparison. TM modes were selected because the electric field is transverse to the micro- sphere surface, enhancing optical effects related to the Q-factor. Then V MOD was decreased until only the strongest WGM was observed in real time, and multiple traces recorded using the two LDs consecutively. The frequency axis was calibrated by shifting the laser frequency by a known value. WGM spectra for the DM and DFB lasers are presented in Figs. 2a and b, respectively, corresponding to operating laser wavelengths of l SET DM ≃ 1537.102 nm and l SET DFB ≃ 1537.134 nm. The emission wavelength of both LDs can be tuned by either temperature or current control, with typical tuning ranges of approximately 5 and 0.5 nm respectively. In both cases data were fitted to a Lorentzian lineshape and a final averaged value for the FWHM extracted. Final results yield values of Dn DM WGM ≃ 11.2 MHz and Dn DFB WGM ≃ 15.0 MHz. Estimation of the Q-factor for the microspheres from these traces yields a value of Q ¼ n/Dn ≃ 10 7 , using a resonant frequency value n ¼ c/nl corresponding to l CENTRE WGM ≃ 1537.1 nm, n ¼ 1.47 and a measured DM WGM linewidth of Dn ≃ 11.2 MHz (DM WGMs correspond to the unperturbed micro- sphere modes, as discussed later). ELECTRONICS LETTERS 20th January 2011 Vol. 47 No. 2