Enhanced Detection of Gas Absorption Using an Erbium-Doped Fiber Ring Laser

Jpn. J. Appl. Phys. Vol. 41 (2002) pp. 5458 5462 Part 1, No. 8, August 2002 #2002 The Japan Society of Applied Physics Enhanced Detection of Gas Absorption Using an Erbium-Doped Fiber Ring Laser Kazuyou MIZUNO 1;2,AkiraMUGINO 1,YujiSEKIGAMI 2,YotsumiYOSHII 2,HiroakiKUZE 2 and Nobuo TAKEUCHI 2 1 The Furukawa Electric Co., Ltd., Fitel Photonics Laboratory, 6 Yawata-Kaigandori, Ichihara, Chiba 290-8555, Japan 2 Center for Environmental Remote Sensing, Chiba University, 1-33 Yayoi-cho, Inage-ku, Chiba 263-8522, Japan (Received February 22, 2002; accepted for publication April 26, 2002) We demonstrate gas sensing using a fiber ring laser composed of an erbium-doped fiber as the laser medium and a compact gas cell module. Acetylene absorption at 1.5 m is enhanced by an intracavity configuration, in which multiple band-pass filters are used to control the wavelength of laser oscillation. A maximum enhancement factor of about 80 is achieved with an 18.3 mm cell under a pressure range below 100 Pa. Experimental results can reasonably be explained by a ring laser theory that describes the influence of acetylene pressure on the threshold and slope efficiency of the laser output characteristics. [DOI: 10.1143/JJAP.41.5458] KEYWORDS: erbium-doped fiber amplifier, fiber ring laser, gas detection 1. Introduction A recent increase in interest in atmospheric environment has motivated the development of optical sensors that are capable of detecting atmospheric trace gases. In view of compactness and ease of operation, diode lasers 1 4) and fiber lasers 5 7) are quite promising as light sources. In this paper, we propose and demonstrate the application of an intracavity erbium-doped fiber ring laser to the detection of molecular absorption. Intracavity laser spectroscopy is one of the laser spectroscopic methods of high sensitivity, leading to qualitative as well as quantitative analyses of gas molecules. 4,5) Due to the laser cavity, the effective absorption length can be increased, and consequently the apparent absorption of gas molecules is considerably enhanced. A few works have so far been reported on the application of fiber lasers to gas sensing. Hünkemeier et al. 5) applied diode-ed Nd 3þ and Yb 3þ -doped multimode fiber lasers to sensitive measurements of intracavity absorption in spectral ranges around 0.93 m and 1.1 m. McAleavey et al. 6) used the cw output of a tunable Tm 3þ -doped fluoride fiber laser to detect CH 4 absorption at 2.31 m. Matsuoka et al. 7) employed an Yb fiber laser (1.038 m) and a distributed feedback laser (1.577 m) to generate mid-ir light (3 m) used for NH 3 detection. In these studies, fiber lasers worked as light sources, without a need for ring cavity configurations. Fiber ring lasers have widely been studied in such applications as the generation of ultra-short pulses 8) and that of the cw laser output with narrow linewidth. 9,10) In contrast, here we describe the application of an Er 3þ -doped fiber ring laser to observe gas absorption with an enhanced optical path length. Since all the optical paths are confined in the fiber without free-space transmission, the present approach provides a trace gas sensor more compact and more suitable for environmental applications than conventional systems based on diode laser transmitters. At the telecommunications band of 1.5 m, erbium ions (Er 3þ ) are used as the amplifying medium. When ed by a laser diode oscillating at 980 nm or 1480 nm, the erbium-doped fiber amplifier (EDFA) gives rise to amplification as wide as 60 nm around 1540 nm. This is an obvious advantage of the EDFA for trace gas detection over other light sources with E-mail: hkuze@ceres.cr.chiba-u.ac.jp limited tuning ranges. Moreover, this wavelength region ensures the high transmittance of optical fibers. To the best of our knowledge, this work presents the first case in which enhanced detection of trace gas is achieved using an EDFA ring laser. In order to demonstrate the enhancement of gas absorption, we use vibrational-rotational transitions of acetylene ( 12 C 2 H 2 ) molecule in the 1 þ 3 band, centered at 1525.2 nm. 11) The experimental results are compared with the theoretical analysis of ring laser operation based on rate equations. 12,13) Our method deals with the cw output of a fiber ring laser, and we can attain a cavity-length enhancement of 10 80, depending upon the ing power of a 1480 nm diode laser. The cell length used in the experiment is 18.3 mm; this is much more compact than the open-path space length of 190 cm employed in the case of the intracavity experiment of Hünkemeier et al., 5) in which time-resolved spectra of broad-band laser emission were observed. Our scheme is similar to an experiment of absorption enhancement (of a factor of about 44), conducted with an external-cavity diode laser (1.5 m) and a 25-cmlong Fabry Perot cavity of a moderate finesse (100). 3) Both this Fabry Perot and the present fiber ring laser configurations are characterized as the direct absorption spectroscopy (as opposed to the modulation spectroscopy); this fact sets a limit to the detection sensitivity, yet these schemes are in principle more suitable for the observation of pressure-broadened molecular absorption under atmospheric conditions than the modulation spectroscopy. 2. Experimental A schematic diagram of our erbium-doped fiber ring laser is shown in Fig. 1(a). The ring laser consists of a coupler, an EDFA, three band-pass filters (BPFs), an isolator, and a gas cell. The output of the EDFA first passes through the BPFs and then through a gas cell module. Then a fraction is taken from a coupler (20 db coupling ratio with an insertion loss of less than 0.3 db), and the main portion of the laser power is returned to the EDFA. The filter specifications are listed in Table I. BPF1 is based on dielectric coating, while the other two are of the Fabry Perot etalon configuration. The combination of these two etalons results in a reduced bandwidth of 1.4 GHz for the full-width at the 80% transmission (see Fig. 5 below). For each filter, the peak wavelength is manually tuned to the absorption peak of the 5458

Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 8 K. MIZUNO et al. 5459 Fig. 1. (a) Schematic diagram of the fiber ring laser. EDFA and BPF stand for erbium-doped fiber amplifier and band-pass filter, respectively. (b) Details of the EDFA. EDF and WDM stand for erbium-doped fiber and wavelength division multiplexer, respectively. (c) Gas cell used for the present experiment. The optical pass length is 18.3 mm. Table I. Specifications of optical filters. FWHM and FSR stand for fullwidth at half maximum and free spectral range, respectively. Filter Loss (db) FWHM (nm) FSR (GHz) BPF1 2.23 1.39 BPF2 5.58 0.028 (3.50 GHz) 271 BPF3 0.93 0.043 (5.37 GHz) 51.3 acetylene P(13) absorption line. In the process of actual measurement, acetylene gas of more than 1000 Pa is introduced into the cell, and the amplified spontaneous emission from the erbium-doped fiber (EDF) is monitored (without the ring configuration) to tune the wavelengths of the BPFs to the acetylene absorption peak. The output spectrum is monitored using a spectrum analyzer (Advantest, Q8381) with a wavelength resolution of 0.1 nm. Figure 1(b) illustrates the detailed configuration of the EDFA. The EDF is ed by a 1480 nm laser diode with a maximum power of 35 mw. A wavelength division multiplexer (WDM) directs the 1480 nm ing light to a 30-mlong EDF. Amplified spontaneous emission of the fiber amplifier is obtained in a wavelength range of 1510 1570 nm. Polarization-independent optical isolators are used to determine the direction of light propagation. The EDF has Er 3þ and Al 3þ dopant molar concentrations of 250 ppm and 5000 ppm, respectively, resulting in a gain of 18 db for small signals (less than 10 W) and 12 db for large signals (more than 1 mw). Controlling the ing power with an accuracy of 0.1 mw, we obtained a stability of EDFA output power of better than 0.7% during a test operation lasting two days. When the cell is evacuated, the laser oscillation takes place above a threshold of P th ¼ 11:7 mw, having a linewidth of about 100 MHz (full-width at half maximum). Since the total laser cavity length is 40 m, it is expected that the laser oscillates in multimodes, with a mode spacing of about 5 MHz. The feature of laser oscillation has been checked by using a high-resolution spectrometer (Burleigh SA, 27 MHz spectral resolution), showing a small fluctuation presumably associated with this multimode operation. This fluctuation, however, does not influence the measurement with the 0.1 nm resolution spectrum analyzer. When the acetylene pressure is relatively high (e.g., 1000 Pa), a double-peak feature has been occasionally observed. This effect is ascribed to the loss profile of the BPF-gas system (see Fig. 5 below), but does not affect the operation under the low pressure regime. We have also tested the influence of an additional polarization controller inserted into the ring configuration. It was found, however, that the behavior of the ring laser sensor is independent of the polarization effects. The gas cell design is illustrated in Fig. 1(c). The laser light emitted from a single-mode optical fiber is collimated by a ball lens and passed through an optical path length of 18.3 mm, where the sample gas is introduced with an appropriate pressure below 1013 hpa. The light is then collected with another ball lens and coupled to the exit fiber path. Fiber faces are cut slantwise to avoid reflection at both entrance and exit connections. The insertion loss of the cell is 2.16 db. The acetylene gas is supplied from a gas cylinder (purity 97.5%) and used without further purification. The gas pressure is measured with a standard diaphragm pressure meter and capacitance manometers (Ulvac, CCMT-10 and CCMT-100). 3. Theory We consider the output power of the fiber ring laser that is dependent on the intracavity gas absorption. It is shown that under appropriate conditions, we can expect enhancement of the effective absorption path length. The output characteristics of erbium-doped fiber ring lasers have been investigated by Pfeiffer et al. 12) and Pfeiffer and Bülow, 13) and here their analytic expressions are modified to include the molecular absorption. The propagation of and signal waves is described by a set of equations that only take into account the populations of the upper laser level ( 4 I 13=2 ) and the lower laser level ( 4 I 15=2 ) of the erbium ions.

5460 Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 8 K. MIZUNO et al. We consider the ring laser configuration as depicted in Fig. 1(a). Because of the presence of an optical isolator, the laser operates in a traveling wave mode. An EDFA of length L is ed at a wavelength ¼ 1480 nm with a ing power of P. The intracavity power at the laser wavelength at the entrance of the amplifier [reference point 1 in Fig. 1(a)] can be analytically expressed by the gain G between points 1 and 2: P in ðgþ ¼ 1 ðg 1Þ ( " G # P 1 G max ) P sat ðl þ ln GÞ : ð1þ Here G max stands for the maximum available gain at, given by G max ¼ exp L : ð2þ This maximum gain is realized only for sufficiently high ing. The small signal absorption coefficients at and are denoted as and, respectively: ¼ abs N ¼ abs N ; ð3þ where abs and abs are the absorption cross section at each wavelength, N is the erbium concentration, and and are the overlap factors. These factors are required to consider the transverse distribution of the light intensity (at and ) in the fiber as well as the radial distribution of the erbium ions. In eqs. (1) and (2) the parameter is defined as P sat ¼ ; P sat ð4þ where P sat and P sat are saturation powers at and, given by ha eff P sat ¼ em þ abs P sat ¼ h A eff : ð5þ em Here, em and em are the emission cross sections, A eff is the effective dope section in the fiber, and is the spontaneous lifetime of the upper level. We use the following values for the parameters in eqs. (1) (5): A eff ¼ 5:31 10 12 m 2, ¼ 10 ms, N ¼ 1:98 10 24 m 3, abs ¼ 5:91 10 25 m 2, em ¼ 3:72 10 25 m 2, abs ¼ 1:67 10 25 m 2, em ¼ 3:69 10 26 m 2, ¼ ¼ 0:360, P sat ¼ 0:199 mw, P sat ¼ 0:971 mw and ¼ 0:212. These parameters, except ¼, are based on our measurement of the EDF characteristics; the value of ¼ is obtained by fitting the theoretical result to the experimental result (see 4). The laser wavelength is ¼ 1532:84 nm, in agreement with the absorption wavelength of the P(13) transition in the 1 þ 3 band of acetylene. When the ing power is above the threshold, the laser output power P obs at point 3 in Fig. 1(a) is given by P obs ¼ P in ðgþgt 23 ð1 RÞ; where P in ðgþ is the intracavity power of the EDFA (at point 1). The gain G between point 1 and point 2 is determined by the oscillation condition of G ¼ 1= ðrt 21 Þ: ð7þ In eqs. (6) and (7), T 21 and T 23 are, respectively, the transmittance from point 2 to 1 (the transmittance in the passive fiber cavity) and the transmittance from point 2 to 3 (between the end of the EDFA and the output connector). The reflectivity of the output coupler (the power fraction returned to the ring) is denoted as R (R ¼ 0:990 in the present case). Combining eqs. (6) and (7) with eq. (1), we obtain the observed power in a form of P obs ¼ ðp P th Þ: ð8þ Here, the slope efficiency,, is given by ¼ ð1 RÞT 23 1 T 21 R 1 1 ðt 21 RG max and the threshold, P th, is given by P th ¼ Þ ð6þ ; ð9þ L lnðt 21 RÞ 1 ðt 21 RG max Þ Psat : ð10þ 4. Results and Discussion When a ring laser is utilized as a gas sensor, the most important aspect is the variation of the observed power P obs according to the pressure p of the absorbing gas. The result is depicted in Fig. 2(a), where the pressure dependence of the output power is shown for various values of the EDFA ing power. The data were obtained after tuning all the BPFs to the peak of the acetylene P(13) transition. Note that in Fig. 2(a), the observed power is normalized to the corresponding value observed at p ¼ 0. This figure indicates that the sensitivity of the ring laser sensor, defined as j@p obs =@pj, tends to be large when the ing power is close to the threshold value (as mentioned above, the threshold for p ¼ 0 is 11.7 mw). For higher ing power, the curves become closer to the single-pass absorption line (dotted curve), which is hypothetically calculated for a single path length of 18.3 mm. The conventional Lambert- Beer law is employed for this calculation, with an absorption cross section of Pð13Þ ¼ 6:84 10 22 m 2 : this value was determined in a separate experiment using a 20 cm gas cell and an external cavity diode laser. 3) For each data in Fig. 2(a), it is possible to calculate the effective path length by means of the same procedure: we define the enhancement factor of the present detection scheme as the ratio of this effective path length to the actual path length of 18.3 mm. Figure 2(b) summarizes the results for P ¼ 11:8, 11:9, and 12:0 mw. When the ing power is 11:8 mw, we obtain an enhancement factor of 70 80 in a pressure range of 20 80 Pa. In this case, higher acetylene pressure terminates the laser oscillation: since the operation is close to the threshold, we have rather unstable behavior of the enhancement factor. When the ing power is 11:9 mw, on the other hand, we obtain a factor of 22, which is nearly constant within a pressure range of 20 400 Pa.

Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 8 K. MIZUNO et al. 5461 Fig. 2. (a) Acetylene pressure dependence of P obs for various values of the ing power. The observed power is normalized to the value at the pressure of 0 Pa. The vertical and horizontal error bars indicate the laser power fluctuation and the resolution of pressure measurement, respectively. (b) Enhancement factor of the ring laser sensor. The factor is calculated from the data in (a) in comparison with a (hypothetical) singlepass absorption with an absorption path length of 18.3 mm. Fig. 3. Output characteristics of the fiber ring laser ( ¼ 1532:84 nm) for various values of acetylene pressure. The theoretical analysis is undertaken as follows. Since the operation of the ring laser is expressed by the slope efficiency and the threshold P th of eqs. (9) and (10), first we examine the ing power dependence of the observed power. Figure 3 depicts the experimental results for the acetylene pressures of 0, 100, 190, 390, 780, and 1000 Pa, obtained by tuning all the BPFs to the acetylene absorption peak, as described in 2. For each pressure, data are reasonably fitted to a straight line: this leads to the experimental values of the slope efficiency and the threshold P th. The pressure dependences of these two Fig. 4. Acetylene pressure dependence of (a) slope efficiency and (b) threshold, on the ring laser output. Dots denote the observed data, while solid and broken curves, respectively, show the theoretical curves calculated using eqs. (9) and (10) with and without consideration of the profile correction (see text). parameters are shown in Fig. 4. The theory of the ring laser yields broken lines in Figs. 4(a) and 4(b), when the parameters quoted below eq. (5) are employed. Actually, the value of (overlap factor) was found to be influential to : the value ¼ ¼ 0:360 was determined so as to obtain agreement between the experimental and theoretical values at p ¼ 0. As seen from Fig. 4(a), the theory predicts that becomes smaller as the pressure increases, whereas the experimental data indicate that the change with the pressure is rather small. Similarly, Fig. 4(b) shows discrepancy between the theory (increases with the pressure) and experiment (nearly constant). the value at the zero pressure for both and P th These discrepancies in both and P th can be resolved in the following manner. In the theoretical consideration above, we assumed that the laser frequency is fixed exactly at the acetylene absorption peak. Because of the limited bandwidth of the coupled filters, however, it is possible that the ring laser oscillates at a frequency slightly different from the initially prepared value. This situation is schematically illustrated in Fig. 5. This figure shows the curves that include both effects of the coupled filter transmission curves and the absorption profile of the acetylene gas. As for the filter transmission, we assume a flat response of BPF1 within this limited frequency range, and full-widths at half maximum for BPF2 and BPF3 as quoted in Table I. If the laser oscillates at a frequency where the loss is minimal, we have a frequency shift of about 0.5 GHz in a pressure range of 0 2600 Pa, as indicated by the dotted curve in Fig. 5. (In a

5462 Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 8 K. MIZUNO et al. frequency shift problem altogether. 5. Conclusions We have demonstrated the capability of a fiber ring laser as a gas sensor. The enhancement of molecular absorption is observed as a change in the output power of the ring laser. The observed dependences of the slope efficiency and the threshold ing power of the laser oscillation on the acetylene pressure were successfully explained by a ring laser theory. The resulting path-length enhancement of a factor of 10 80 will be useful for future development of a compact fiber sensor for ambient gas detection. Fig. 5. Ring transmittance calculated on the basis of the BPF transmission curves (BPF2 and BPF3) with the absorption spectrum for various values of acetylene pressure (in Pa). separate experiment using the high-resolution spectrometer, we actually observed a shift of 0:44 GHz when the acetylene pressure was increased from 0 to 2500 Pa.) To incorporate this frequency shift into the simulation, it is required to modify the value of gas transmission included in T 21 and T 23. The results are shown as solid curves in Fig. 4(a) for the slope efficiency and in Fig. 4(b) for the threshold. By considering the effect of frequency shift with the acetylene pressure, we attain reasonable agreement between the theory and experiment with respect to the pressure dependence of and P th. The remaining difference between the measurement and calculation can be ascribed to the inaccuracy in the setting of the BPF wavelengths. Basically, this frequency shift of the laser oscillation degrades the performance of the ring laser as a gas sensor, since the effect of gas absorption effectively diminishes. It is noted, however, that this degradation is insignificant when the dip due to the molecular absorption is relatively small, and does not hinder its operation as a trace gas sensor. In addition, installation of an auxiliary wavelength feedback mechanism using BPFs and a gas cell would eliminate the Acknowledgements The authors would like to thank Mr. K. Kondow for his contribution at the initial stage of this work. 1) F. S. Pavone, F. Marin, M. Inguscio, K. Ernst and G. Di Lonardo: Appl. Opt. 32 (1993) 259. 2) S.-Q. Wu, T. Kimishima, H. Kuze and N. Takeuchi: Jpn. J. Appl. Phys. 39 (2000) 4034. 3) R. Okazawa, H. Kuze, H. Masusaki and N. Takeuchi: Jpn. J. Appl. Phys. 38 (1999) 4946. 4) V. M. Baev, J. Eschner, E. Paeth, R. Schüler and P. E. Toschek: Appl. Phys. B 55 (1992) 463. 5) J. Hünkemeier, R. Böhm, V. M. Baev and P. E. Toschek: Opt. Commun. 176 (2000) 417. 6) F. J. McAleavey, J. O Gorman, J. F. Donegan, B. D. MacCraith, J. Hegarty and G. Mazé: IEEE J. Sel. Top. Quantum Electron. 3 (1997) 1103. 7) N. Matsuoka, S. Yamaguchi, K. Nanri, T. Fujioka, D. Richter and F. K. Tittel: Jpn. J. Appl. Phys. 40 (2001) 625. 8) K. Tamura, H. A. Haus and E. P. Ippen: Electron. Lett. 28 (1992) 2226. 9) J. Zhang, J. W. Y. Lit and G. W. Schinn: IEEE Photonics Technol. Lett. 8 (1996) 1621. 10) R. J. Forster, N. Langford, A. Gloag, L. Zhang, J. A. R. Williams and I. Bennion: Opt. Commun. 141 (1997) 283. 11) K. Nakagawa, M. de Labachelerie, Y. Awaji and M. Kourogi: J. Opt. Soc. Am. B 13 (1996) 2708. 12) Th. Pfeiffer, H. Schmuck and H. Bülow: IEEE Photonics Technol. Lett. 4 (1992) 847. 13) Th. Pfeiffer and H. Bülow: IEEE Photonics Technol. Lett. 4 (1992) 449.