100 W, single frequency, diffraction-limited beam and quantum noise measurements in a continuous-wave laser-diode-pumped Nd:YAG MOPA system

Transcription

1 100 W, single frequency, diffraction-limited beam and quantum noise measurements in a continuous-wave laser-diode-pumped Nd:YAG MOPA system Shailendhar Saraf*, Arun K. Sridharan, Supriyo Sinha, Karel Urbanek, Robert L. Byer E.L. Ginzton Laboratory, Stanford University, CA We describe a Nd:YAG slab laser master oscillator power amplifier (MOPA) system that has been scaled to 104 W of output power at 1064 nm, single transverse mode. The amplifier system consists of rod amplifiers followed by edge- and end-pumped slab amplifiers. The MOPA output of 104 W was 89% fundamental TEM 00 spatial mode with less than 3% depolarization. These results are a key milestone in the development of a high power laser system for the next generation LIGO ground-based gravitational wave detectors. We also present measurements of the power noise due to the optical amplification in a Nd:YAG free-space traveling wave amplifier as the amplifier transitions from the linear regime into the heavily saturated regime. The quantum noise behavior is demonstrated by saturating the gain of the amplifier using a high power beam and measuring the power noise detected by a single spatial mode probe beam traversing the same optical path through a zigzag slab amplifier. The experimental results are in excellent agreement with theory. OCIS codes: ( ) Lasers, diode pumped, ( ) Laser amplifier, ( ) Optical amplifier, (70.590) Photon Statistics High power, TEM 00 mode, low-noise lasers are required for the Laser Interferometer Gravitational Wave Observatory (LIGO) [1]. The LIGO interferometer is designed to detect strains of the order of 10 - for sensing gravitational waves emanating from astronomical sources such as neutron stars and black hole binary inspirals. Various technical noise sources such as seismic noise, laser amplitude and phase noise, and detector shot noise detract from the ability to detect such small strains. Since detector shot noise scales inversely with laser power, there is a strong incentive to scale the existing 10 W diffraction-limited, single frequency LIGO laser source to the 00 W level and beyond. Master oscillator power amplifier (MOPA) based systems are well suited for power scaling while preserving the spatial and temporal coherence of the master oscillator. In addition, the MOPA architecture allows for a robust laser design with the capability of repair or replacement of individual modules without major disruption to the system alignment. As an example of the robustness of the MOPA design, the first generation 10 W laser diode pumped Nd:YAG rod-based MOPA systems have operated to date for more than 0,000 hours at the LIGO sites []. Other applications as well, such as remote sensing and coherent laser radar, require lasers with operating powers more than an order of magnitude greater than the 10 W rod-based MOPA lasers currently used in LIGO. However, thermally induced distortions and birefringence present a fundamental limit to the power scaling of rod-based gain media. The zigzag slab, invented by Chernoch in 197 [3] is an elegant solution to power scaling while maintaining beam quality because the slab s nearly one-dimensional thermal gradients and the zigzagging of the signal beam significantly reduce the thermal phase distortions and birefringence [4]. Scaling to high average power levels while maintaining beam quality and minimizing phase distortions and depolarization has been demonstrated with zigzag-slab geometry systems [5]. Practical use of slab lasers was limited initially by the low efficiencies and phase distortions of the prior designs, and by the complexity of pumping and cooling the laser head. Most of these engineering problems have been solved and along with the availability of higher quality YAG crystals, kw level output from slabs has been demonstrated [6]. Several other approaches to scaling to high powers are being developed including the thin disk laser [7] and recent work in highly overmoded fiber amplifiers [8]. 1

2 Our work is motivated by the need for a high power, single axial mode, single polarization laser source for the proposed Advanced LIGO detector [9]. Three approaches are being investigated to meet the advanced LIGO laser source requirements. The approaches are: injection locked rod laser (LZH, Germany) [10], a stableunstable slab resonator (Adelaide University, Australia) [11] and the end pumped slab amplifier (Stanford, USA). Prior work on a laser source specifically tailored to meet LIGO requirements led to the injection seeding studies [1] and to a Master Oscillator Power Amplifier approach. The latter was selected by initial LIGO and was engineered and delivered to LIGO by Lightwave Electronics Corporation. (LWE) and has operated for the past five years []. Motivated by the LIGO requirements, a novel pumping geometry called edge-pumping which incorporates the zigzag-slab and conduction cooling was demonstrated at Stanford [13]. This novel edge-pumped laser concept was first demonstrated by our group using laser-diode-pumped Nd:YAG slabs [14][15]. In this paper we report on the progress of an end pumped Nd:YAG slab amplifier approach [16] to meet the Advanced LIGO laser source requirements. This approach offers the advantages of utilizing the existing LIGO 10 W prestabilized laser (PSL) source [] and scaling to higher power levels as needed in a series of amplifier stages that be easily repaired or replaced. The MOPA approach also offers long-term operational reliability and the potential for power scaling beyond the 00 W requirements to meet future LIGO requirements. Figure 1 shows a schematic of the experimental set up. The LIGO prestabilized oscillator/amplifier [] consists of a Nd:YAG non-planar ring oscillator (NPRO) with 300 mw output power [17]. The oscillator power is amplified to 10 W TEM 00 mode in a double-passed diode pumped Nd:YAG rod amplifier. The 10 W output from the PSL is then amplified by a second LWE rod amplifier and an edge pumped Nd:YAG slab amplifier [18]. This output is the input to the end pumped Nd:YAG slab amplifier that is the key element for power scaling with a high degree of spatial and spectral coherence. The 0 W amplifier module following the 10 W LIGO laser is identical to the amplifier module inside the 10 W LIGO laser and has near diffraction-limited beam quality and Figure 1. Experimental setup for a 100-W MOPA laser excellent power and pointing stability and a high degree of reliability []. The edge-pumped slab amplifier was used in the MOPA chain since it was well-characterized as a pre-amplifier and with the available pump power could scale to an output power level that would efficiently drive the end-pumped slab at several times the saturation intensity of Nd:YAG. An important feature of our design is the use of fiber-coupled laser diode arrays as the optical pumping source. This scheme has several engineering advantages. Fiber coupling of the pump diodes allows the laser diodes and power supplies to be remotely located, separating the tasks of heat removal from the laser diodes and

3 heat removal from the amplifier head. Hundreds of watts per fiber of laser-diode pump radiation can be delivered to the solid state laser head with high efficiency and high brightness because the CW, broadband laser diode radiation does not damage or excite nonlinear effects in the optical fibers. Figure shows the end pumped slab geometry with undoped end pieces and a doped region in the center of the slab similar to the TRW design [16]. The conduction-cooled slab has a small cross-sectional area (1.1 mm X 0.9 mm) to maximize gain. The aspect ratio of approximately one combined with end pumping results in efficient coupling to the TEM 00 mode and less depolarization and thermal focussing [19]. Additionally, the aspect ratio is designed such that the low-loss transverse parasitic paths do not form a TIR loop. The slab has a 3.33 cm long 0.6 % doped Nd:YAG center with 1.53 cm long undoped ends. The undoped ends are diffusion bonded to Figure. The end-pumped Nd:YAG slab amplifier the doped region (by Onyx corporation) to reduce thermal focusing due to end effects. The top and bottom surfaces are coated with 3 µm of SiO to create the TIR surfaces for the signal and pump and provide a rugged cooling interface. In our setup, the ends of two 600 micron, NA = 0. fibers are reimaged to a 400 micron spot and coupled into the slab by TIR of the end faces. The pump light is guided by TIR in the diffusion-bonded, undoped YAG ends before being absorbed with >90% efficiency in the Nd-doped region of the slab. Likewise, the incident TEM 00 mode 1064 nm beam is transmitted by TIR along the zig-zag path with 4 bounces, reimaged and reflected back into the slab with 40 bounces. The signal beam focussing optics are designed for a waist of 300 microns in the center of the slab. The use of angular multiplexing avoids the need for a Faraday rotator that is difficult to operate at high power levels [0]. Additionally, image inversion for the double pass eliminates the prism effect and beam deflection due to slab flexure. For efficient amplifier operation, the end pumped slab amplifier must be driven by input laser power near the saturation power level. In our experiment the end pumped slab is driven with 30 W at four times the saturation intensity for Nd:YAG. The key to the performance of the end pumped slab is to pump with adequate power and to suppress parasitic oscillations that pin the gain at an arbitrarily low value. The slab is pumped with up to 300 W of 808 nm power from each end for a total available pump power of 600 W. The parasitic modes are suppressed by having rough sides (non zigzag TIR surfaces) on the slabs. The use of rough sides has the drawback of pump light leakage due to scatter, which results in lower gain in the slab. Another technique for 3

4 suppressing parasitic modes was the application of bevels and tailored index of refraction coatings on the non zigzag TIR surfaces. The index of the coating is selected to transmit the parasitic modes and trap the pump light. The rough non zigzag surfaces were polished and a commonly used optical-grade epoxy with a refractive index of about 1.56 was applied on the doped region of the slab. The coating provides the index differential with YAG, which refracts out the parasitic modes incident on the non-zigzag surfaces for incident angles up to 60 degrees but still allows confinement and guiding to the pump light. Additionally a 6-degree transverse bevel and a 0.1-degree longitudinal bevel ensured that a transverse parasitic mode rapidly worked its way out of the slab. The cladding and bevel was successful in increasing the small signal gain and the threshold for parasitics. Fig. 3a shows the small signal gain measurements for slabs with rough sides and with the epoxy coating applied. The ASE-limited small signal gain with parasitic suppression through use of rough sides is limited to 3.9 at a pump power of 430 W and with parasitic suppression using tailored index of refraction coatings is 4.5 at a pump power of 340 W. The no-ase extrapolated small signal gain at a pump power of 430 W was 6.93 for the slab with rough sides and improved to 10.3 on the slab with the cladding. Based on the no-ase extrapolated gain, it can be estimated that about 4% of the pump light is lost to scatter from the two rough surfaces of the slab. This number drops to 17 % with 1 Parasitic control 10 Power Extraction gol Output Power (W) Pump Power (W) RoughSides (a) Coating Figure 3. a) The ASE-limited small signal gain measurements and extrapolated no-ase small signal gain in slabs with two methods of parasitic suppression. b) Power output as a function of pump power for the two slabs. The slope efficiency is improved by 75% using slabs with parasitic suppression coatings. parasitic-suppression coating applied. However, parasitic-suppression coatings require further development to be reliable. The single pass power output of the slab is 65 W for an incident power of 30 W with a slope efficiency of almost 9% as shown in Fig. 3b. The slope efficiency of the slab with the coating as the parasitic suppression was substantially better at almost 15%. Beam quality is near diffraction limited with M < However, M measurements are not adequate to accurately quantify the mode content of the beam. The TEM 00 content of the amplifier output is measured using a Fabry-Perot interferometer called a mode cleaner. The mode cleaner is a three-mirror high finesse ring cavity filter for both spatial modes and temporal noise. The cavity has a finesse of 4400, a FSR of 713 MHz and a FWHM bandwidth of 174 KHz [1]. The LIGO mode cleaners are presently not designed to handle 100 W of CW power operating at a finesse of 4400 since the thermal loading of the mirrors causes a significant loss of TEM 00 power into higher order modes. Hence, the mode content measurement is done by sending the high power beam into a 0.3 % output coupler, dumping the high power and mode matching the transmitted low-power beam into the mode cleaner. From the reflection spectrum of the laser output scanned in Pump Power (W) SinglePass_RoughSides DoublePass_RoughSides SinglePass_Coating (b)

5 frequency and transmitted through the mode cleaner, it was determined that 84% of the output is in the TEM 00 mode. The TEM 00 content measurement would likely improve with better mode matching into the cavity. The depolarization loss in the slab is < 1.5 % as would be expected from the slab geometry with a uni-directional temperature gradient [19]. The intensity noise spectrum is measured to be 10-6 / Hz at 1 khz and 10-8 / Hz at 10 MHz []. After characterization of the single pass output, we double passed the slab with an input power of 35 W and total pump power of 430 W and achieved 104 W of output power with an M < The slope efficiency was measured at almost 15 % as shown in Fig. 3b. As described before, the transmitted low-power beam through a 0.3% output coupler was carefully mode matched into the mode cleaner cavity to determine the power in the TEM 00 mode. It was determined that 89% of the output power is in the TEM 00 mode. The intensity noise spectrum was measured to be x10-7 / Hz at 1 khz and 10-8 / Hz at 10 MHz. MOPA configurations are fundamentally noisy due to the power noise added by the unavoidable spontaneous emission into the spatial mode being amplified. One way to reduce the power noise added by an amplifier and extract most of the available gain in the medium is to saturate the amplifier. The final power amplifier stages will not only be highly efficient in terms of extracting most of the available power in the medium but also add minimum power noise to the amplifier system. In this study, we confirm the noise performance of a free-space Nd:YAG optical amplifier in the non-linear or saturated regime. Previous work at Stanford has demonstrated the noise performance of unsaturated amplifiers [3] where the quantum noise scales linearly with the power gain of the amplifier. Prior work on the noise performance of saturated amplifiers has focussed on fiber amplifiers and has shown that the effect of gain saturation is to increase the signal to noise ratio relative to the case of an unsaturated amplifier [4]. To our knowledge, we present the first quantitative measurement of the quantum noise of a free-space Nd:YAG optical amplifier as it transitions from the linear regime into the heavily saturated regime. The starting point of noise analysis is the photon statistics master equation from the fundamental paper by Shimoda, Takahasi and Townes [5] describing the evolution of a photon stream in an amplifier based on the stochastic process of birth, death and immigration (BDI). Solving the probability equation yields the following expression for the for the noise power σ d at the detector for an unsaturated amplifier [4] σ d = σ shot 1+ η G 1 (1) ( ( )) where G is the power gain of the amplifier, η is the efficiency of the photodetection process and noise power of the output photon stream and the spontaneous emission factor f sp σ shot is the shot f sp is only a function of the α transmission lossα in the gain medium and the power gain as defined by the equation f sp = 1+. A ln( G) similar analysis of the photon statistics master equation after including a saturation function yields an expression for the noise added to a probe beam in a amplifier whose gain has been saturated by an uncorrelated high power beam yields an identical expression for the noise power σ d of the probe beam for a saturated amplifier σ = σ 1+ f η G 1 () g 0l 1+ sn ( ( )) d shot sp sat ext where G sat = e is the saturated power gain of the amplifier which is a function of the small signal gain g 0 l and the uncorrelated saturating beam photon flux n ext andσ shot is the shot noise power of the probe beam at the output of the saturated amplifier. 5

6 Figure 4 shows the experimental setup. A shot-noise limited probe beam is generated from a monolitihic, single frequency, single spatial mode non-planar ring oscillator (NPRO). The NPRO is locked to a monolithic, high finesse and impedance matched three-mirror Fabry-Perot ring cavity (mode cleaner) [1]. This cavity operates in s-polarization and at resonance using the Pound-Drever-Hall RF technique. The high finesse mode cleaner reduces the power noise and non-tem 00 modes of the NPRO output through temporal and spatial filtering and results in a shot noise limited probe beam at the input to the amplifier. The high power beam generated in a MOPA chain is also locked to an identical Fabry-Perot cavity but with a much lower finesse of approximately 50 operating in p-polarization. Locking is accomplished by controlling the frequency of the master oscillator in the MOPA chain using the same RF technique. The two cavities have identical constructions except for the reflectivity of the input and output couplers and consequently support identical TEM 00 eigenmodes. Fig 4. Layout of the quantum noise measuring system of a Nd:YAG free space saturated amplifier. This translates into identical spot sizes of the beams at the output of the cavities and permits the beams to be overlapped with a high degree of precision. The temperature of the Nd:YAG crystal in the NPRO is tuned to ensure a frequency offset of approximately 750 MHz between the locked probe and the locked high power saturating beams. This frequency offset is much less than the 00 GHz fluorescence bandwidth of Nd:YAG, therefore the overlapped probe and high power beams have the same power gain. However, the frequency offset enables separation of the beams after the amplification process with a third Fabry-Perot cavity with a large FSR as described later. The two beams are combined on a polarizing beam splitter equidistant from the two Fabry-Perot cavities. This ensures the beams have the same spot size. The overlapped high power beam and the probe beam are then sent through a power amplifier and see the same power gain. The two beams are uncorrelated, since they are derived from two sources that have no spatial or temporal correlation and additionally have a frequency offset of 750 MHz. The power amplifier in the system is an end-pumped Nd:YAG zigzag slab [16]. The slab is pumped at power levels of up to 175 W per fiber. The unsaturated small signal power gain in the experiment at a total pump power of 350 W is measured to be The high power saturating beam is varied from 0 W which corresponds to a linear unsaturated amplifier to 7 W in which case the amplifier is heavily saturated with an input intensity of almost 9 times the saturation intensity for Nd:YAG. The amplified saturating high power beam and probe beam are separated after the end-pumped slab amplifier by a pair of thin film polarizers. The polarizers alone do not provide rejection of the power coupled from the saturating high power beam into the probe beam due to the thermally induced birefringence in the slab at high pump powers. However, further temporal and spatial post-filtering is performed by locking the probe beam to the resonance of a short low finesse mode cleaner operating in s-polarization. The post filtering Fabry-Perot cavity 6

7 has a finesse of 15 and a FSR of 3 GHz. The 750 MHz frequency offset between the probe and high power beams results in an additional 30 db rejection of the high power saturating beam in both polarizations. The output of the short low finesse mode cleaner is then sent though a beam splitter into a balanced detector. The narrow passband of the Fabry-Perot cavity transmits the probe and completely rejects ASE (amplified spontaneous emission) and 808 nm pump light from the amplifier. The balanced detector outputs are measured over the frequency range of MHz to quantify the quantum noise added to the signal through amplification. This range is chosen based on the frequency response of the photodetection system. The spontaneous emission factor f sp is calculated on the basis of the measured transmission loss (1.5%/cm), the transmission efficiency of the photodetector window (95%) and the quantum efficiency of the detectors (86%). Fig. 5 shows in detail the amplifier noise behaviour as it transitions from the unsaturated regime to the heavily saturated regime. The dashed horizontal lines locate the experimental points from fig. 5a on fig. 5b. Fig. 5a shows the measured power noise of the probe relative to the shot noise limit for the 90 ma photocurrent for an unsaturated amplifier in the absence of a high power beam. Fig. 5b shows the power noise on the probe beam as the high power beam saturates the gain of the amplifier. Fig. 5. Power noise of an unsaturated and saturated amplifier. a) Linear scaling of quantum noise of an unsaturated amplifier as the pump power is increased. b) Quantum noise curves as the gain is saturated down with a high power beam overlapped with the probe beam. The dashed curve shows the extraction efficiency of the power amplifier as it transitions to deep saturation. The theoretical curves in fig. 5a and 5b are plots of equations 1 and respectively based on the measured power gain and loss in the slab and the measured efficiencies of the photodetection process. Additionally, the power extraction efficiency of the amplifier versus the high power beam intensity is plotted in fig. 5b based on the extracted power and the available power in the slab at the highest pump power of 350 W. At an input power of 7 W, which corresponded to I/I sat of 8.6, the extraction efficiency of the amplifier is about 93 % while the quantum noise is.8 times the shot noise limit at 90 ma of photocurrent. The figure clearly shows that as the power is increased, the power extraction efficiency from the amplifier approaches 1 while the quantum noise level is reduced towards the shot noise limit set by losses in the amplifier as predicted by theory. The next step in this study is to power scale to 00 W to meet the Advanced LIGO requirements by using two Nd:YAG end pumped slabs. The two slabs will be double passed to maximize the power extraction from the slabs. Presently a single pass slab provides about 35 W of power. The slabs with the parasitic control claddings should extract about 50 W per pass. Hence two slabs double passed should yield 00 W of power operating in the heavily saturated regime. We plan to confirm that the TEM 00 content measurements, RIN and phase noise measurements, pointing stability of the power amplifier meet LIGO specifications [9]. 7

8 This work was supported in part by the National Science Foundation under grant number PHY and in part by the US Army Research Office under ARO grants DAAD S. Saraf s address is saraf@stanford.edu. References: 1) Barry C. Barish and Rainer Weiss, Physics Today, October ) W. Wiechmann, T.J. Kane, D. Haserot, F. Adams, G. Truong, and J.D. Kmetec, CLEO (1998). 3) W.S. Martin and J.P. Chernoch,U.S. Patent (197). 4) Eggleston, J.M, T.J. Kane, K. Kuhn, J. Unternahrer, and R.L. Byer, IEEE JQE, 0, 89 (1984). 5) R. J. Shine, Jr., A. J. Alfrey, and R. L. Byer, Opt. Lett. 0, 459 (1995). 6) R. J. St. Pierre, D.W. Mordaunt, H. Injeyan, J.G. Berg, R.C. Hilyard, M.E. Weber, M.G. Wickham and G. M. Harpole, High-Power Lasers; Santanu Basu; Ed. (1998) 7) D. Muller, S. Erhard, O. Ronsin, and A. Giesen, OSA Trends in Optics and Photonics 83, 78 (003). 8) J. Limpert, A. Liem, S. Hoffer, H. Zellmer, A. Tunnermann, S. Unger, S. Jetschke, and H.R. Muller, CLEO Technical Digest 590 (00). 9) LIGO internal working note, LIGO document number LIGO-C D. 10) M. Frede, R. Wilheim, M. Brendel, C. Fallnich, F. Seifert, B. Willke, K. Danzmann, Optics Express, 1,15 (004) 11) D. Mudge, et.al,, Classical and Quantum Gravity 19, 7, (00). 1) A.D. Farinas, E.K. Gustafson, and R.L. Byer, Optics Letters 19, 114 (1994). 13) W.M. Tulloch, T.S. Rutherford, and R.L. Byer, US Patent 6,134,58, ) W.M. Tulloch, T.S. Rutherford, E.K. Gustafson and R.L. Byer, OSA TOPS Vol. 6, 9, (1999) 15) T. S. Rutherford et al, IEEE JQE, 36, 05 (000). 16) G. D. Goodno, S. Palese, J. Harkenrider, and H. Injeyan, OSA Trends in Optics and Photonics Vol. 50, (001) 17) T.J. Kane and R.L. Byer, Opt. Lett. 10, 65, ) T. S. Rutherford, W. M. Tulloch, S. Sinha, and R. L. Byer, Opt. Lett. 6, 986 (001). 19) T.J. Kane, J.M. Eggleston and R.L. Byer, QE-1(8), 1195 (1985). 0) T.J. Kane, W.J. Kozlovsky and R.L. Byer, Opt. Lett. 11, 16 (1986). 1) B. Willke, N. Uehara, E.K. Gustafson, R.L. Byer, P. King, S. Seel and R.L. Savage, Jr., Optics letters 3, 1704 (1998) ) B. Willke, 3) W. M. Tulloch, et.al., Opt. Lett. 3, 185, ) E.Desuvivre, Erbium-Doped Fiber Amplifiers, (Wiley and Sons, New York, 1994). 5) K. Shimoda, H. Takahasi, and C.H. Townes, Journal of the Physical Society of Japan, Vol. 1, No.6, p.687 (1957). 8