Quantification of different water species in acetone using a NIR-triple-wavelength fiber laser

A fiber laser using a thulium-doped ZBLAN gain medium was used to generate laser radiation simultaneously at 1461, 1505 and 1874 nm, with > 5 mW output power at each of the wavelengths. The laser was used to quantify the near-infrared absorption of liquid water in acetone. Additionally, near-infrared spectra were recorded using a broad band source and were interpreted using parallel factor (PARAFAC) analysis to rationalize the concentration-dependent peak shifts. ©2014 Optical Society of America OCIS codes: (300.1030) Absorption; (140.3510) Lasers, fiber. References and links 1. U. Willer, M. Saraji, A. Khorsandi, P. Geiser, and W. Schade, “Nearand mid-infrared laser monitoring of industrial processes, environment and security applications,” Opt. Lasers Eng. 44(7), 699–710 (2006). 2. S. D. Jackson, “Towards high-power mid-infrared emission from a fibre laser,” Nat. Photon. 6(7), 423–431 (2012). 3. N. J. Scott, C. M. Cilip, and N. M. Fried, “Thulium fiber laser ablation of urinary stones through small-core optical fibers,” IEEE J. Sel. Top. Quantum Electron. 15(2), 435–440 (2009). 4. M. Güney, B. Tunc, and M. Gulsoy, “Incisional effects of 1940 nm thulium fiber laser on oral soft tissues,” Proc. SPIE 8584, 848408 (2013). 5. B. Tunc and M. Gulsoy, “Tm:fiber laser ablation with real-time temperature monitoring for minimizing collateral thermal damage: ex vivo dosimetry for ovine brain,” Lasers Surg. Med. 45(1), 48–56 (2013). 6. N. M. Fried, “Thulium fiber laser lithotripsy: An in vitro analysis of stone fragmentation using a modulated 110watt Thulium fiber laser at 1.94 mu m,” Lasers Surg. Med. 37(1), 53–58 (2005). 7. N. M. Fried and K. E. Murray, “High-power thulium fiber laser ablation of urinary tissues at 1.94 μm,” J. Endourol. 19(1), 25–31 (2005). 8. M. C. Pierce, S. D. Jackson, M. R. Dickinson, and T. A. King, “Laser-tissue interaction with a high-power 2microm fiber laser: Preliminary studies with soft tissue,” Lasers Surg. Med. 25(5), 407–413 (1999). 9. P. Peterka, I. Kasik, A. Dhar, B. Dussardier, and W. Blanc, “Theoretical modeling of fiber laser at 810 nm based on thulium-doped silica fibers with enhanced H4 level lifetime,” Opt. Express 19(3), 2773–2781 (2011). 10. C. Xia, “Mid-infrared supercontinuum laser system and its biomedical applications,” Ph.D. Dissertation (University of Michigan, Ann Arbor, 2009). 11. G. Qin, S. Huang, Y. Feng, A. Shirakawa, and K.-I. Ueda, “784-nm amplified spontaneous emission from Tmdoped fluoride glass fiber pumped by an 1120-nm fiber laser,” Opt. Lett. 30(3), 269–271 (2005). 12. G. Qin, S. Huang, Y. Feng, A. Shirakawa, and K.-I. Ueda, “Multiple-wavelength up-conversion laser in Tmdoped ZBLAN glass fiber,” IEEE Photon. Technol. Lett. 17(9), 1818–1820 (2005). 13. G. Androz, D. Faucher, D. Gingras, and R. Vallée, “Self-pulsing dynamics of a dual-wavelength Tm:ZBLAN upconversion fiber laser emitting around 800 nm,” J. Opt. Soc. Am. B 24(11), 2907–2913 (2007). 14. B. Frison, A. R. Sarmani, L. R. Chen, X. Gu, S. Thomas, P. Long, and M. Saad, “Dual-wavelength lasing around 800 nm in a Tm:ZBLAN fibre laser,” in IEEE Photonics Conference, (2012), pp. 668–669. 15. B. Frison, A. R. Sarmani, L. R. Chen, X. Gu, and M. Saad, “Dual-wavelength S-band Tm:ZBLAN fibre laser with 0.6 nm wavelength spacing,” Electron. Lett. 49(1), 60–62 (2013). #213247 $15.00 USD Received 2 Jun 2014; revised 18 Jul 2014; accepted 25 Jul 2014; published 4 Aug 2014 (C) 2014 OSA 11 August 2014 | Vol. 22, No. 16 | DOI:10.1364/OE.22.019337 | OPTICS EXPRESS 19337 16. K. Ramaswamy, C. Jia, M. Dastmalchi, L. R. Chen, and M. Saad, “Dual-band 810/1480 nm Tm:ZBLAN fiber laser,” in IEEE Photonics Conference (2013), pp. 273–274. 17. W. J. Peng, F. P. Yan, Q. Li, S. Liu, T. Feng, S. Y. Tan, and S. C. Feng, “1.94 μm switchable dual-wavelength Tm fiber laser employing high-birefringence fiber Bragg grating,” Appl. Opt. 52(19), 4601–4607 (2013). 18. J. J. Max and C. Chapados, “Infrared spectroscopy of acetone-water liquid mixtures. II. Molecular model,” J. Chem. Phys. 120(14), 6625–6641 (2004). 19. J. J. Max and C. Chapados, “Infrared spectroscopy of acetone-water liquid mixtures. I. Factor analysis,” J. Chem. Phys. 119(11), 5632–5643 (2003). 20. Y. Koga, F. Sebe, T. Minami, K. Otake, K. Saitow, and K. Nishikawa, “Spectrum of excess partial molar absorptivity. I. Near infrared spectroscopic study of aqueous acetonitrile and acetone,” J. Phys. Chem. B 113(35), 11928–11935 (2009). 21. A. Estrada-Baltazar, A. De Leon-Rodriguez, K. R. Hall, M. Ramos-Estrada, and G. A. Iglesias-Silva, “Experimental densities and excess volumes for binary mixtures containing propionic acid, acetone, and water from 283.15 K to 323.15 K at atmospheric pressure,” J. Chem. Eng. Data 48(6), 1425–1431 (2003). 22. L. Bøje and A. Hvidt, “Densities of aqueous mixtures of non-electrolytes,” J. Chem. Thermodyn. 3(5), 663–673 (1971). 23. K. Noda, M. Ohashi, and K. Ishida, “Viscosities and densities at 298.15 K for mixtures of methanol, acetone, and water,” J. Chem. Eng. Data 27(3), 326–328 (1982). 24. S. L. Pan, C. Y. Lou, and Y. Z. Gao, “Multiwavelength erbium-doped fiber laser based on inhomogeneous loss mechanism by use of a highly nonlinear fiber and a Fabry-Perot filter,” Opt. Express 14(3), 1113–1118 (2006). 25. J. E. Bertie and Z. D. Lan, “Infrared intensities of liquids. 20. The intensity of the OH stretching band of liquid water revisited, and the best current values of the optical constants of H2O at 25 degrees °C between 15,000 and 1 cm,” Appl. Spectrosc. 50, 1047–1057 (1996). 26. B. Dickens and S. H. Dickens, “Estimation of concentration and bonding environment of water dissolved in common solvents using near infrared absorptivity,” J. Res. Natl. Inst. Stand. 104(2), 173–183 (1999). 27. B. Czarnik-Matusewicz and S. Pilorz, “Study of the temperature-dependent near-infrared spectra of water by two-dimensional correlation spectroscopy and principal components analysis,” Vib. Spectrosc. 40(2), 235–245 (2006). 28. B. Czarnik-Matusewicz, S. Pilorz, and J. P. Hawranek, “Temperature-dependent water structural transitions examined by near-IR and mid-IR spectra analyzed by multivariate curve resolution and two-dimensional correlation spectroscopy,” Anal. Chim. Acta 544(1–2), 15–25 (2005). 29. K. R. Murphy, C. A. Stedmon, D. Graeber, and R. Bro, “Fluorescence spectroscopy and multi-way techniques. PARAFAC,” Anal. Methods 5(23), 6557–6566 (2013). 30. R. Bro and H. A. L. Kiers, “A new efficient method for determining the number of components in PARAFAC models,” J. Chemometr. 17(5), 274–286 (2003). 31. H. P. Loock and P. D. Wentzell, “Detection limits of chemical sensors: Applications and misapplications,” Sens. Act., Biol. Chem. 173, 157–163 (2012).


Introduction
Chemical detection, along with many biological and medical techniques, can be improved through the use of dual-and multi-wavelength fiber lasers.By simultaneously measuring the fluorescence spectrum, the reflectance, or the optical attenuation at several wavelengths, it is possible to generate a "response pattern" that can improve the specificity and sensitivity of the chemical detection.Related techniques such as optical coherency tomography (OCT) can also make use of multi-wavelength fiber lasers to provide rapid imaging of biological samples.
The near-infrared (NIR) region is particularly well suited for chemical detection, as most analytes show well resolved and structured overtone or combination bands which are comparable to bands in the mid-infrared (MIR) spectrum.Even though the transitions are weaker, the development of brighter and sometimes tunable laser sources can lead to a similar spectral response.
Thulium ions (Tm 3+ ) provide a very effective means for developing fiber lasers operating at a wide range of wavelengths, including the NIR region.The 3 H 4 → 3 F 4 and 3 F 4 → 3 H 6 transitions can provide lasing around 1480 nm and 1900 nm, respectively.The 1480 nm wavelength lies near an overtone absorption peak of liquid water and can provide a convenient means for water detection in various liquids.Spectral fingerprinting at different NIR wavelengths can be used to detect a variety of chemical species that absorb in this range [1].Light sources around the eye-safe 1900 nm region are also useful for chemical sensing, as well as for biological and medical applications.Soft-tissue medicine has made use of fiber lasers emitting in this range, exploiting the absorption of the OH overtone of water near 1940 nm [2].Finally, tissue and urinary stone ablation can be facilitated through the use of intense ~2000 nm light, owing again to the comparably strong absorption of water at this wavelength [3][4][5][6][7][8].
When thulium is doped in ZBLAN (a fluoride glass host), the lifetime of its excited 3 H 4 state is almost a hundred-fold longer (1350 µs) than in silica (14.2 µs) [9], due to the lower phonon energy of ZBLAN glass.The increased transparency of ZBLAN in the MIR also makes it appealing for any lasing transitions longer than about 2000 nm [10].Several multiwavelength Tm 3+ fiber lasers (in both silica and ZBLAN) operating at around 800 nm, 1480 nm, or 1900 nm have been reported.In [11][12][13], a Tm 3+ :ZBLAN gain medium was pumped at around 1110 nm to achieve lasing at 785/810 nm through the 3 H 4 → 3 H 6 and 1 G 4 → 3 H 5 transitions, respectively.Dual-wavelength emission using a single transition has also been demonstrated around 800 nm using the 3 H 4 → 3 H 6 transition [14] and around 1480 nm using the 3 H 4 → 3 F 4 transition [15].In both cases, up-conversion pumping at 1064 nm in a Tm 3+ :ZBLAN fiber was used.Additionally, dual-band operation at 810/1480 nm in a single piece of Tm 3+ :ZBLAN fiber has been achieved, with both lasing transitions sharing the same upper energy level ( 3 H 4 → 3 H 6 , 3 H 4 → 3 F 4 ) [16].Finally, dual-wavelength lasing at 1942 nm was achieved in a Tm 3+ :silica fiber by pumping at around 790 nm and using two highly birefringent fiber Bragg gratings (FBGs) [17].However, in none of the previous reports were multi-wavelength lasers used for chemical detection and quantification.
In this paper, we demonstrate the use of a fiber laser that emits at three NIR wavelengths to quantify the water content in liquid acetone.Detection is accomplished through the ν 2 + ν 3 combination band near 1930 nm (5180 cm −1 ) and the first overtone of the ν 1 symmetric stretch vibration near 1450 nm (7000 cm −1 ).This chemical system was selected as a case study for several reasons: the partial molar volume corrections are well understood, the related MIR bands have been extensively studied and interpreted using multivariate analysis methods, and acetone itself has a well-defined and sparse background spectrum in this region.As a similar understanding of other matrices becomes available, we envision that the water content of other liquids, such as lubricant oils and fuels, can similarly be determined.
Of particular interest in this study is the system's deviation from the Beer-Lambert absorption law.While these deviations are well documented [18][19][20], it is often overlooked that vibrational absorption spectra -including overtone and combination band spectra -of small molecules can change dramatically depending on their solvent environments or the first solvation cage structure.Here, we explicitly consider the spectral signatures of three distinct water "species" known to exist in acetone at concentrations ranging from 0.55 to 11.4 M (approx.1.0 vol % -20 vol %) of water.

Triple-wavelength laser
A schematic drawing of the laser system is shown in Fig. 1.It consists of two independent branches, which generate lasing at 1461/1505 nm and 1874 nm, respectively, as well as a third shared output branch.The gain medium for operation at 1461 nm and 1505 nm (top branch) is a 52 cm length of Tm 3+ :ZBLAN fiber, while the gain medium for operation at 1874 nm (bottom branch) is an 85 cm length of the same Tm 3+ :ZBLAN fiber.Manufactured by IRphotonics/Thorlabs, the double-cladding ZBLAN fiber is doped with 8,000 ppm Tm 3+ , has an 8 µm core diameter, a 125 µm cladding diameter and is coated with 15 µm of mixed fluoroacrylate and acrylate.The 52 cm gain fiber is pumped through a 1064/1480 wavelength division multiplexer (WDM) by up-conversion pumping from a 1064 nm ytterbium-doped fiber laser (YDFL, P 1064 ), whereas the 85 cm gain fiber is pumped directly by a 1560 nm pump (P 1560 ) consisting of an external cavity laser (ECL) and an EDFA.In the top branch, the cavity is formed on one end by two fiber Bragg gratings (FBGs), which also define the lasing wavelengths.FBG 1 and FBG 2 have respective center wavelengths of λ 1 = 1505 nm and λ 2 = 1461 nm.In the bottom branch, the cavity is formed on one end by FBG 3 , with center wavelength λ 3 = 1874 nm.All three gratings are written in SMF-28 fiber and have a peak reflectivity > 99%.The two branches are connected to a coupler and use a common gold-tipped fiber mirror to form the other end of their respective cavities.The second port on the coupler serves as the output for all three wavelengths, where an optical spectrum analyzer (OSA, Yokogawa, AQ6375) measures the power at each wavelength.The coupler itself has a 50/50 splitting ratio at 1461/1505 nm and an 87/13 splitting ratio at 1874 nm (87% to the mirror).A polarization controller (PC) is placed between FBG 1 and FBG 2 in order to equalize the power generated at λ 1 and λ 2 .The Tm 3+ :ZBLAN gain fibers are coupled to the SMF-28 fibers through mechanical splices (represented by × in Fig. 1), which have a loss of approximately 2 dB per pair.

Sample preparation
Acetone (ACP Chemicals, ≥ 99.9%, reagent grade) and deionized water (Thermo Scientific, Type 1 reagent grade) were used to create solution samples without additional purification.Twelve mixtures of water in acetone between 0.55 M and 11.4 M were prepared using Eppendorf micropipettes to measure volumes of both neat components.As the solution volume differs from the sum of component volumes, molar concentrations were calculated using an average of the values for the concentration-dependent density provided in the literature [21][22][23].

Absorption spectroscopy
The absorption spectra of water-acetone samples were obtained using both the triplewavelength Tm 3+ :ZBLAN fiber laser and a conventional Fourier-transform (FT) NIR absorption spectrometer (Perkin-Elmer Spectrum 400).FT-NIR spectra over the 1.0-2.5 μm (10,000 -4000 cm −1 ) spectral range were obtained at 1 cm −1 increments and 4 cm −1 resolution, using a 2 mm/s mirror scan rate and a triglycine sulfate pyroelectric detector.Samples were placed in a 1.0 mm quartz cuvette and each transmittance spectrum was taken from an average of 8 scans.Background measurements were taken without a cuvette between spectra, with pure acetone serving as a blank.
NIR transmittance measurements on the same samples were also obtained using the triplewavelength Tm 3+ :ZBLAN fiber laser.GRIN lenses were used to couple the fiber laser output to a 1 cm quartz cuvette containing the samples.Sample transmittance was detected using an OSA with a 55 dB dynamic range and 0.5 nm bandwidth from 1450 nm to 1900 nm.Each transmittance measurement was taken from an average of 30 sequential OSA scans.Background measurements were taken using neat acetone before and after each sample.The reference was taken as the average of the two background measurements.

Characterization of the laser light source
We first characterize laser operation at λ 1 and λ 2 .Figure 2(a) shows the output power at λ 1 and λ 2 as a function of pump power, P 1064 .The laser begins emitting light at λ 2 when the pump exceeds a threshold of P 1064 > 522 mW, and the slope efficiency is 2.2% up to a saturated output power of 15 mW.Only at a threshold of 1144 mW does lasing at λ 1 start, increasing with 3.4% slope efficiency.The laser emits at both wavelengths (with a higher power at λ 2 ) until P 1064 reaches 1580 mW, when λ 1 and λ 2 are approximately equal in power.As P 1064 increases to 1859 mW, the power at λ 1 increases to a maximum output of 21 mW, while the power at λ 2 saturates at 15 mW.Second, we characterize the laser operation at 1874 nm only.Figure 2(b) shows the output power at λ 3 as a function of P 1560 .Lasing at λ 3 occurs at a threshold of 126 mW and continues to increase with P 1560 at 2.5% slope efficiency.The maximum output power is 13 mW for a pump power of 637 mW.
Following this, we re-measure the output power at λ 1 and λ 2 with λ 3 set to operate at 6.8 mW, as well as the output power at λ 3 with λ 1 and λ 2 set to operate at 13.17 mW and 13.95 mW, respectively.Similar output power characteristics are obtained, indicating that the laser cavities of the two branches work independently.Figure 3(a) shows the output spectrum of the three lasing lines at λ 1 , λ 2 and λ 3 .Note that the peak at 2128 nm is an artifact of the OSA due to higher-order diffraction from P 1064 .The peak fluctuations of the three wavelengths over 30 minutes are shown in Fig. 3(b).We observe that the fluctuations of the three peaks are all less than 1.5 dB, resulting in three relatively stable output peak powers.The power fluctuations are mainly induced by gain competition, environmental variations, and mechanical splices between the ZBLAN and silica fibers.Through implementation of fusion splices over mechanical splices, we predict the laser stability could be greatly improved.In addition, further stability can be achieved by packaging the laser to reduce environmental influences.Finally, gain competition between wavelengths sharing the same fiber gain medium can be reduced using cascaded cavities [15] or inhomogeneous loss mechanisms [24].
For the top branch in Fig. 1, single-wavelength operation can be obtained by appropriately adjusting the polarization state of the PC.As shown in Fig. 4 with the pump power set to 1.7 W, the Tm 3+ :ZBLAN fiber laser can be switched to operate at either 1461 nm or 1505 nm or at both wavelengths.

Interpretation of the absorption spectra
The NIR spectra of the 20 samples plus one acetone blank are given in Fig. 5. Linear offset corrections were first applied to all spectra to compensate for stray light contributions at regions with no absorption features, i.e. at 1250-1300 nm, 1660-1670 nm, 1820 nm and 2190 nm.Then a second common linear offset was applied to align the baseline with the spectrum of neat water in [25] for ease of comparison.For water mole fractions above 0.6, the spectra are saturated in the 1900-2000 nm region.The absorption spectrum of neat water as reported in [25] is also included in the figure.In all spectral regions, the peaks attributed to water absorption show a clear shift to longer wavelengths as the water concentration increases.The peak shifts in Fig. 5 are attributed to the emergence and decline of water species that interact with their solvent environments in different ways, depending on the number of hydrogen bonds formed with either surrounding water molecules or with acetone.The average hydrogen-bonding network around any water molecule is predicted to change with water concentration; several species with distinct MIR spectral features have been reported [18][19][20]26].Assuming solution homogeneity, we expect to find water in 3-4 distinct hydrogen bonding environments in the concentration range of 0.0 -11.4 M, and at least five distinct water species in the complete concentration range of 0.0 -55.6 M.These species can be interpreted in a first approximation as: water molecules surrounded exclusively by acetone (W1); water forming a hydrogen bond to one adjacent water molecule (W2); to two adjacent water molecules (W3); to three adjacent water molecules (W4); and four adjacent water molecules (W5).Since these species cannot be isolated from one another, deconvolution of the spectrum is difficult.We therefore used factor analysis to identify the number of components, their absorption spectra, and their relative contributions to the triple-wavelength laser spectra.Previously, Max and Chapados (MC) also used factor analysis to distinguish between the bonding environments in water and acetone over the entire miscibility regime [18,19].They recorded and analyzed MIR spectra of the O-H and C-H stretch regions, as well as strong overtone and combination bands.Nine water environments were identified, correlating to five factors (W1-W5), as factor analysis cannot distinguish between species that evolve concurrently [18,19].The presence of distinct water species was also discussed by Koga et al. [20] and by Czarnik-Matusewicz and associates [27,28], who analyzed water at different temperatures using multivariate analysis to rationalize the peak shifts.To our knowledge, MC provides the only previous characterization of mixtures using factor analysis [18,19].These workers provided a persuasive model that describes the emergence and decline of each species as a function of water mole fraction.From their factor analysis, MC derived equilibrium constants that were used as parameters to fit their data, in addition to a comprehensive model describing each species' concentration as a function of water mole fraction.Here, we adopt their parameterized fit, recognizing that (a) their MIR spectral band analysis is likely more reliable than our NIR analysis, (b) MC used anhydrous acetone to prepare their samples, and (c) MC used a larger number of spectra.
Since our data set was not large enough to analyze the different acetone species, we assumed that the acetone spectra are very similar and subtracted from the spectra in Fig. 5 the contribution of neat acetone absorption weighted by its mole fraction (Fig. 6(a)).Parallel Factor (PARAFAC) analysis on the processed NIR spectra was performed using drEEM software [29], which is derived from DOMFLUOR software [30].The program was run 10 times using random initialization, a non-negativity constraint and a convergence criterion of 10 −8 .The spectra were fitted to 2 to 5 components, and it was found that 3 components provided a large improvement in core consistency over a 2-component fit.An increase to 4 or 5 components did not improve the fit results.The three spectral component loadings are shown in Fig. 6(b).Figure 7 shows the relative contribution of each component to the overall water absorption spectrum, compared to the theoretical curves calculated by MC [18].The completeness of the spectral analysis is supported by the observation of an isosbestic point around 1440 nm (Fig. 6(b)), which is consistent with the three components being stoichiometrically related.Component 3 is related to the completely water-solvated species (W5 in MC), and thus its score increases as water is added (Fig. 7).Component 1 falls as the water concentration increases, and the data follows the sum of W1 + W2 described by MC, i.e., water species that form either zero or one hydrogen bond(s) with other water molecules.Similarly, component 2 rises and falls as water concentration increases and follows closely the sum of species W3 + W4, corresponding to water molecules surrounded predominantly by either 2 or 3 water molecules.Given a larger number of samples and a stronger absorption feature, it is conceivable that components 1 and 2 could be further deconvoluted into the water species described by MC.In any case, it is apparent that with the addition of water, more hydrogen bonds are formed and the O-H bond weakens, resulting in the observed red shift [18].
In our analysis, we were guided by this previous work, but our spectra were much weaker and we did not desiccate the supplied acetone.Despite these sources of error, our analysis is nevertheless adequate to guide the chemical analysis performed with our triple-wavelength laser.

Quantification of water content in acetone using multi-wavelength laser absorption
The absorbance measurements of water in acetone, which are taken with the triplewavelength fiber laser, show approximately linear trends (Fig. 8).The detection limits are lowest (about 2.8 M) when absorbance was measured at the 1461 nm laser line, presumably due to higher laser stability and a larger absorption cross section.The molar absorptivity, Ɛ H2O (λ), is calculated from the slope of the linear fit at each wavelength λ, and the respective detection limits are calculated from the 99% confidence interval as described previously [31] (Table 1).The difference between experimental and literature Ɛ H2O (λ) values is due to variations in sample composition.There is little reason to assume that the hydrogen-bonded water species in tetrahedrally coordinated bulk water have the same absorption cross section compared to water species surrounded largely by acetone.Unfortunately we cannot extend the laser based absorption measurements to pure water due to total attenuation of the laser light through a Fig. 8. Absorption of solutions of water in acetone measured using the fiber laser at (a) 1461 nm, (b) 1505 nm and (c) 1874 nm.The 99% confidence intervals are shown as two blue lines and used to calculate the limit of detection (dashed line).The molar absorptivity at each wavelength was calculated from the slope of the linear fit (red line).The data were analyzed as in reference 20.The solid green symbols are absorbance values extracted from Fig. 5 and are connected to guide the eye.
1 cm cuvette.The discrepancy between our measured Ɛ H2O (λ) at low water concentrations and the literature Ɛ H2O (λ) for pure water is therefore understandable.Consider that both the transition dipole moment of the overtone absorption and its resonance wavelength are related to the anharmonicity of the O-H stretching vibration, which increases with the strength and number of hydrogen bonds.This explains both the shift to longer wavelength and the increased overtone absorption cross section of bulk water.A similar argument for the bathochromic shift was made by Dickens and Dickens [26] and later quantified by Max and Chapados [19].The data in Figs. 6 and 8 can be combined to show the absorption coefficients as a function of water concentration.In principle, measurements taken at two wavelengths -one corresponding to an absorption peak and one in a transparency window -would be enough to determine the analyte concentration.Measurements at more than two wavelengths help increase the reliability of the measurement.Here, we intend to highlight the nonlinearity of the absorption at these three wavelengths.Using background subtracted absorption signals at each of the three wavelengths, the absorption at each wavelength is found in relation to the total absorption, A(c), at all three wavelengths (Eq.( 1)).For a single absorber following the Beer-Lambert law, it is expected that these fractions depend linearly on the analyte concentration, c.However, as is shown in Fig. 9, the fractions at 1461 nm and 1874 nm exhibit nonlinear behavior, due to a pronounced red shift of the absorption peaks with increasing water concentration.We attribute this to the concurrent absorption of multiple coexisting water species.Of course, the relative concentration of the water species depends on the mole fraction of water in acetone according to Fig. 7.
More importantly, Figs. 8 and 9 show that absorption experiments with the multiwavelength laser source are able to accurately quantify water, agreeing with independent FT-NIR measurements within the 99% confidence interval.While the absorption measurements at 1874 nm agree well with the FT-NIR measurements in Fig. 9, the relative intensities at 1461 and 1505 nm show larger deviations.The larger uncertainty of these measurements can be attributed to the 10-15% laser power fluctuations at these wavelengths (Fig. 3(b)).

Conclusion
We have demonstrated a triple-wavelength Tm 3+ :ZBLAN fiber laser emitting simultaneously at 1461, 1505, and 1874 nm, and its potential for use in chemical detection.The chemometric analyses of water in acetone NIR spectra provide guidelines for the optimization of this laser chemical sensing.To maximize sensitivity, it may seem preferable to measure at a water NIR absorption maximum (1480 nm and/or 1920 nm), though both peaks exhibit a pronounced red shift and nonlinear absorption with increasing water concentration.It is, however, conceivable to measure at the isosbestic point around 1440 nm, where all water species have an identical absorption cross section.At this wavelength, the absorption signal scales linearly with concentration.
In practice, it is preferable to select several wavelengths showing either high sensitivity or a linear dependence on concentration -in addition to at least one wavelength that probes the absorption and scattering background.This method can easily be extended to the detection of impurities in other solvents, such as water in lubricant oils or hydrocarbon fuels, as a better understanding of each of these solvent matrices becomes available.

Fig. 1 .
Fig. 1.Schematic of the tri-wavelength fiber laser.Two Tm 3+ :ZBLAN fibers are incorporated into fiber cavities that are defined by 3 different FBGs and a shared gold mirror.

Fig. 2 .
Fig. 2. (a) Measured output power at λ 1 (red circles) and λ 2 (black squares) as a function of P 1064 when P 1560 is on (solid symbols) and off (open symbols).(b) Measured output power at λ 3 as a function of P 1560 when P 1064 is on (solid) and off (open symbols).

Fig. 5 .
Fig.5.NIR spectra of water in acetone with 21 different concentrations.The dashed line shows the spectrum of neat water from reference[25], whereas the solid lines show water acetone solutions with mole fractions between 1.0 and 0.0 in intervals of 0.05.

Fig. 6 .Fig. 7 .
Fig. 6.(a) Near-infrared spectra of water in acetone with 20 different concentrations as in Fig. 5 after a weighted contribution of the acetone absorption spectrum had been subtracted from all spectra in Fig. 5.(b) The spectra of the three PARAFAC components obtained by decomposition of (a) show an isosbestic point near 1440 nm.The 1461 and 1505 nm laser wavelengths are included as lines.

Table 1 . Absorption Coefficients, Ɛ H2O (λ), in [L mol −1 m −1 ] Calculated Using the Laser Absorption at Low Concentrations in Acetone Compared to the Literature Values for Bulk Water
Fig. 9. Fractional absorption of solutions of water in acetone calculated from Eq. (1).The absorption fractions at 1461 nm (squares), 1505 nm (circles) and 1874 nm (triangles) are shown for both the fiber laser (solid symbols) and a commercial NIR spectrometer (open symbols).The lines are results of a polynomial fit meant to guide the eye.