Heterodyne Phase-Sensitive Dispersion ... - ACS Publications

May 8, 2017 - Economy and Competitiveness for supporting the projects under the ... (21) Rieker, G. B.; Jeffries, J. B.; Hanson, R. K. Appl. Opt. 2009...
0 downloads 0 Views 983KB Size
Download PDF
Article pubs.acs.org/ac

Heterodyne Phase-Sensitive Dispersion Spectroscopy in the MidInfrared with a Quantum Cascade Laser Pedro Martín-Mateos,*,† Jakob Hayden,‡ Pablo Acedo,† and Bernhard Lendl‡ †

Electronics Technology Department, Universidad Carlos III de Madrid, C/Butarque 15, 28911 Leganés, Madrid, Spain Institute of Chemical Technologies and Analytics, Technische Universität Wien, Getreidemarkt 9/164-UPA, 1060 Vienna, Austria

‡

ABSTRACT: Molecular dispersion spectroscopy encompasses a group of spectroscopic techniques for gas analysis that retrieve the characteristics of the sample from the measurement of the profile of its refractive index in the vicinity of molecular resonances. This approach, which is in clear contrast to traditional methods based on the detection of absorption, provides inherent immunity to power fluctuations, calibration-free operation, and an output that is linearly dependent on gas concentration. Heterodyne phase-sensitive dispersion spectroscopy (HPSDS) is a very recently proposed technique for molecular dispersion spectroscopy based on tunable lasers that is characterized by a very simple architecture in which data processing and concentration retrieval are straightforward. Different HPSDS implementations have been experimentally validated in the near-IR. Here, we present the first demonstration of HPSDS in the mid-IR using a directly modulated quantum cascade laser for the measurement of CO. The setup is put under test to characterize its response to changing concentrations, pressures, and levels of optical intensity on the detector, and the limit of detection is estimated. Besides this, an experimental comparison with wavelength modulation spectroscopy with second-harmonic detection (2f-WMS) is performed and discussed in detail in order to offer a clear view of the benefits and drawbacks that HPSDS can provide over what we could consider the reference method for gas analysis based on tunable laser spectroscopy.

I

new solutions that eliminate the need for a chirped laser, like heterodyne phase-sensitive dispersion spectroscopy (HPSDS)2 or even completely different ideas like the use of electro-optic dual-comb sources11 or Fabry-Pérot lasers12 for multiheterodyne molecular dispersion spectroscopy have also been proposed and demonstrated. Because of the simplicity of its architecture and the ease of operation and data processing, HPSDS is one of the most promising new alternatives in the field of spectroscopic gas sensing. HPSDS has been validated in the near-IR,2 nonetheless, the interest in implementing HPSDS systems working in the mid-IR, where the strongest fundamental ro-vibrational bands of molecules are present, is clear. In this article, we report on the implementation of a HPSDS setup in the mid-IR, using a commercial distributed feedback quantum cascade laser (DFB-QCL). We also present an experimental study of the dependence of the HPSDS signal on gas concentration, pressure, and light intensity, which, to the best of our knowledge, has not been reported elsewhere. Furthermore, the performance of the dispersion spectroscopic setup is quantitatively compared to wavelength modulation

n the past few years, several demonstrations have proved the many advantages of new molecular dispersion spectroscopic methods over traditional absorption measurement techniques for gas detection and analysis based on tunable lasers.1,2 Detection approaches based on molecular dispersion spectroscopy not only overcome baseline and normalization problems3,4 but are capable of providing an output linearly dependent on gas concentration.5 A direct consequence of this is an enhanced dynamic range.6 Furthermore, the inherent immunity of the measurement of dispersion to fluctuations in the optical power reaching the detector enables robust operation. Also, since the recorded phase is a direct measure of molecular dispersion, dispersion spectroscopic sensors are suitable for calibration-free operation.2 Molecular dispersion spectroscopic methods obtain the concentration of a given analyte from a measurement of the profile of the refractive index of the sample under study. The anomalous refractive index accompanying rotational−vibrational absorption lines induces optical dispersion (i.e., a spectrally characteristic change in the phase velocity of electromagnetic radiation) in amounts that are directly proportional to gas concentration. Optical dispersion can hence be detected for the estimation of concentration, and a few different methods can be employed to that end. The technique best established is chirped laser dispersion spectroscopy (CLaDS),1 which was demonstrated in different spectral regions7,8 and improved through several revisions.9,10 However, © XXXX American Chemical Society

Received: January 24, 2017 Accepted: May 8, 2017 Published: May 8, 2017 A

DOI: 10.1021/acs.analchem.7b00303 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry spectroscopy with second-harmonic detection (2f-WMS),13,14 a widely employed technique of absorption spectroscopy.

■

EXPERIMENTAL SECTION Basis of Heterodyne Phase-Sensitive Dispersion Spectroscopy. HPSDS bases its operation on the measurement of the refractive index (instead of the absorption) of the sample in the vicinity of the spectral feature of interest. Since this change in the refractive index is directly proportional to the amount of the target analyte in the gas, its concentration can be estimated. In HPSDS, a three-tone optical beam, generated by intensity-modulating a monochromatic laser light source, is sent through the gas sample, see Figure 1. It should be clarified that

Figure 2. Different refractive indices found by each tone induce dissimilar phase velocities through the gas that generate optical phase shifts between the three signals, as seen from the different positions of the maxima of the tones after traveling through the sample. These phase shifts are directly proportional to the concentration of gas. It has to be mentioned that the three beams actually co-propagate and have only been vertically shifted in this figure for viewing purposes.

imposed on the laser beam) is generated. The optical phase shifts induced by the gas sample determine the phase of the RF beat note according to the equations given below. Hence, the phase of the beat note (at the center of the spectral feature) is directly proportional to the concentration of gas, providing the means of HPSDS for gas concentration retrieval. The mathematics relating the change in the refractive index of the sample to the phase of the beat note was derived previously.2 The phase φ0 is given by ⎛ a sin(φ ) − b sin(φ ) ⎞ 12 13 ⎟⎟ φ0 = a tan⎜⎜ a cos( φ ) + b cos( φ ) ⎝ 12 13 ⎠

(1)

where φ12 =

ω0L (n(ω0) − n(ω0 − Ω)) c

(2)

φ13 =

ω0L (n(ω0 + Ω) − n(ω0)) c

(3)

describe dispersion and

Figure 1. Profiles of the absorption (top) and the refractive index (center) in the vicinity of a molecular resonance together with the HPSDS three-tone interrogation signal (bottom). The modulation of the laser intensity generates a laser spectrum that is composed of three wavelengths, the laser center wavelength and two sidebands. The three optical tones experience different refractive indices, resulting in a phase-shifted beat note.

a = 10−(A(ω0) + A(ω0 −Ω))/2

(4)

(5) b = 10−(A(ω0) + A(ω0 +Ω))/2 describe the absorption A of the optical tones. Herein, ω0 is the center optical frequency, L is the path length through the sample, c is the speed of light in vacuum, n(ω) is the refractive index at ω, Ω is the modulation frequency, and A(ω) is the absorbance in base 10 at ω. It is instructive to consider the case when absorption is negligible (a = b = 1) and the phase shifts (φ12 and φ 13) are small enough to allow for linear approximation of the trigonometric functions (conditions that are very often fulfilled in practical scenarios). This simplifies eq 1 to ωL φo = 0 (n(ω0 + Ω) − n(ω0 − Ω)) (6) 2c

by sinusoidally modulating the intensity of any monochromatic source, two extra tones, named sidebands, appear in the spectrum at lower and higher wavelengths with a spacing from the central tone equal to the modulation frequency (for large modulation more than two sidebands appear). In the vicinity of the molecular spectral feature, the different refractive indices that each one of the tones experiences (Figure 1) induce different phase velocities that generate optical phase shifts between the three beams (Figure 2). These optical phase shifts are directly dependent on the analyte’s concentration in the gas under investigation. Due to the quadratic response of photodetectors, when laser beams of different optical frequencies strike simultaneously on its sensitive area, a radio frequency (RF) tone with a frequency equal to the difference of the optical frequencies, known as a beat note, is generated. Therefore, when the three-tone HPSDS signal impinges on a photodetector, a beat note with a frequency equal to the modulation frequency (i.e., the intensity modulation super-

Thus, the HPSDS phase is proportional to the difference of the refractive indices found at the frequencies of the two outer lines (E2 and E3 in Figure 1) of the interrogation signal (and, hence, directly proportional to the concentration of the analyte in the gas). The resulting HPSDS phase derived when sweeping the laser center frequency ωc across the spectral feature centered at ωc is shown in Figure 3. For modulation frequencies small as compared to the fwhm of the resonance, B

DOI: 10.1021/acs.analchem.7b00303 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry the HPSDS signal resembles the group velocity spectrum of the line.

Figure 4. Setup of the HPSDS system based on the direct modulation of a QCL. Black lines represent electrical connections, and the red line illustrates an optical beam. Figure 3. Output HPSDS phase signal for a complete spectral line sweep. The dotted line represents the phase for zero concentration. The HPSDS spectrum resembles the spectrum of group velocity in the vicinity of an absorption line.

the laser frequency across the spectral line and the modulation current for the laser are combined in a bias-tee that is directly connected to the laser. After passing the gas cell, the modulated light is detected by a photodetector where, due to its square-law behavior and as previously presented, an electrical beat note signal is generated, with a frequency equal to the modulation frequency. The phase of the beat note, i.e., the HPSDS signal, can be obtained by a lock-in amplifier (IQ phase detectors or phase meters can be used as well). In previous HPSDS studies, intensity modulation was achieved using dedicated intensity modulators. Since these devices are not commercially available in the mid-IR the laser was directly modulated in this work. However, modulating the current of a semiconductor laser not only leads to output power modulation, but also to changes in wavelength. The combination of frequency modulation (FM) and intensity modulation (IM) associated with semiconductor lasers is a well-known issue in the wavelength and frequency modulation spectroscopy community, where, contrarily to the molecular dispersion spectroscopy field, intensity modulation is unwanted. A comprehensive theoretical and experimental study on direct modulation and mixed IM and FM can be found in ref 16. In short, mixed FM and IM yields spectra similar to those discussed above (three-tone signal), but with an asymmetry between the sidebands in both amplitude and phase. This introduces deviations from the theory previously derived, distorted line shapes and leads to a nonlinear relationship between the recorded phase and concentration. The inequality of sidebands can be included in models to still retrieve the molecular line profile and concentration by means of least-squares fitting procedures, as was demonstrated for CLaDS.7 However, besides the simplicity of calculating concentration by multiplying the HPSDS signal by a constant factor, the desirable linear relationship between output phase and gas concentration is partially lost (as illustrated in the following sections of the paper). In this contribution, instead of the least-squares fitting procedure, we minimize frequency modulation by selecting the right operation points of the laser. In general, frequency modulation is minimized as the modulation frequency is increased and as the DC bias current is decreased close to the laser threshold. For a more detailed discussion, we point to an excellent study of the tuning characteristics of DFB-QCLs reported by Hangauer et al.,17 including the identification of operation points advantageous for dispersion spectroscopy. They experimentally confirmed the theoretically predicted exceptionally small frequency tuning of QCLs at high

The so recorded phase spectrum can be used to retrieve the concentration of an analyte directly from the phase at ωc: the refractive index spectrum, and hence the relative phases φ12 and φ13 of the three lines (eq 2 and eq 3), scale linearly with concentration. Nonetheless, their amplitude scales exponentially with gas concentration, as is well known from the Lambert−Beer law (c1 and c2 represent two different concentrations): φ12(c1) φ12(c 2)

=

φ13(c1) φ13(c 2)

=

Δn(ω0 , c1) a /b(ω0 , c1) c = 1 ≠ Δn(ω0 , c 2) c2 a /b(ω0 , c 2) (7)

Generally, this leads to nonlinearity of the HPSDS output phase signal with concentration. The important exception of this case is when a = b; i.e., A(ω0 − Ω) = A(ω0 + Ω). For an isolated ro-vibrational line, this is the case at the absorption maximum ω0 = ωC (at the exact center frequency of the spectral line). It is easy to show from eq 1 that in this case the HPSDS phase is independent of absorption and is given by eq 6, even for large phase differences. Therefore, when an ideally intensity modulated laser is assumed, the trough of the HPSDS spectrum (minimum value of phase in Figure 3) is, within the range of validity of Beer−Lambert law, a linear measure of the analyte’s concentration in a gas and provides constant sensitivity over the full dynamic range (in contrast, WMS signals scale with the derivative of transmission profiles. Linearity with concentration is hence given only for optically thin samples, for which T = 10−A ≈ 1 − ln(10)A).15 From here on, the trough of the HPSDS spectrum will be referred to as the HPSDS signal, and any value of concentration provided in this article will be estimated from this quantity. It is worth noting that, for optimum performance (maximization of the HPSDS signal), the modulation frequency has to be 0.58 times the full width at half-maximum (fwhm) of the target spectral absorption profile, given the profile is a Lorentzian.2 Lower modulation frequencies will reduce the HPSDS signal.2,5 Heterodyne Phase-Sensitive Dispersion Spectroscopy with Directly Modulated Quantum Cascade Lasers. The architecture of the HPSDS setup based on direct laser modulation is shown in Figure 4. The three-tone optical signal is produced by a small sinusoidal modulation of the laser current. The DC bias with a superimposed ramp for sweeping C

DOI: 10.1021/acs.analchem.7b00303 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

implemented using the same components and architecture described in the previous section. Only the wiring of the QCL and the detector was changed owing to the reduced modulation frequency. The modulation signal, together with the ramp signal, was connected to the analog set point connector of the QCL driver which was directly connected to the QCL (the bias tee was removed). A modulation frequency of 30 kHz was selected, and the modulation depth was adjusted for optimum WMS performance.21 The detector signal was directly fed into the lock-in amplifier with the modulation signal as a reference (no frequency mixers were used). The amplitude of the second harmonic was output and digitized. It is worthwhile noting that, provided the bandwidth of all components is sufficient, a WMS setup is transformed into a HPSDS setup by simply increasing the modulation frequency and recording the phase of the first harmonic, rather than the amplitude of higher harmonics.

frequencies, which is related to their small line width enhancement factor and comes in handy when aiming for pure IM.18 Besides pure IM, single sideband modulation, in which the down-shifted sideband is suppressed, is an exceptionally advantageous modulation for HPSDS. Since only two tones are present in a single sideband modulated beam, the amplitudes of the tones are irrelevant for the recorded phase, making the measurement independent of absorption and the recorded phase linear with concentration (refer to eq 1 for a = 0). Single sideband modulation was demonstrated recently for directly modulated QCLs, circumventing the need for post-laser modulators to achieve single sideband modulation.19 Experimental Setup for HPSDS in the Mid-IR. The HPSDS setup used for validation is based on a Distributed Feedback QCL (HHL-14-15, Ad Tech Optics Inc., California, USA), with emission wavelength of 4.59 μm, operated in continuous-wave. The driver used was a QCL500 OEM Quantum Cascade Laser Driver (Wavelength Electronics Inc., Montana, USA) powered by Lead-acid batteries. The ramp signal for the sweeping of the spectral feature was obtained from an Agilent 33120A (Agilent Technologies Inc., California, USA). The modulation signals were generated by a twochannel, 160 MHz synthesizer (DG4162, RIGOL Technologies, Inc., Beijing, China). The bias and modulation signals were combined and taken to the laser by a self-designed biastee in parallel with the QCL protection electronics. The laser temperature was stabilized by a PTC2.5K TEC-controller (Wavelength Electronics Inc., Montana, USA). The laser radiation was directed through a 100 m multipass Herriot gas cell (Aeodyne Research Inc., Massachusetts, USA) onto a PVMI-4TE-8 IR photovoltaic detector with a MIPACv2F-250 preamplifier (both VIGO Systems S.A., Ożarów Mazowiecki, Poland) providing an output bandwidth of 250 MHz. The gas cell was evacuated continuously using a membrane pump while a constant flow of sample gas was fed into the cell from a custom built gas mixer system that allowed for setting gas concentrations.14,20 Pressure was adjusted via the gas flow and a manometer. Given the restricted bandwidth of the available lock-in amplifier (SR850, Stanford Research Systems Inc., California, USA) a frequency mixer was utilized to downshift the RF beat note to below 100 kHz. The reference for the lock-in amplifier had to be downshifted in the same manner. The analog phasesignal from the lock-in amplifier was digitized by a 16 bits acquisition card (USB-6003, National Instruments Corp., Texas, USA). Two points of operation were chosen to illustrate the influence of the direct current IDC and modulation frequency Ω on HPSDS: IDC = 220 mA, Ω = 2π*160 MHz, TQCL = 14.95 °C and IDC = 240 mA, Ω = 2π*100 MHz, TQCL = 13.65 °C. As demonstrated below, these points represent an IM-dominated regime (high modulation frequency, low bias current) and a mixed IM and FM regime (lower modulation frequency, higher bias current). The threshold current of the QCL was 175 mA. The QCL temperature TQCL was adjusted to tune the laser to the CO line under study. All experiments presented below are measurements of the ro-vibrational line at 2179.77 cm−1 of CO in N2. Implementation of a 2f-WMS Setup for Comparison of Performance. Besides the experimental validation of the HPSDS setup in the mid-IR, a comparison of its performance with 2f-WMS was performed. The 2f-WMS setup was

■

RESULTS AND DISCUSSION Experimental Validation of a HPSDS Setup in the MidIR. Figure 5 shows the HPSDS spectra for 3 ppm of CO at 100

Figure 5. HPSDS phase signal for different operation configurations of the QCL for CO at 3 ppm and at 2179.77 cm−1 (4.587 μm).

mbar, corresponding to a peak absorbance of 0.64 (all values of absorbance presented in this paper are given as absorbance in base 10), for the two operation points specified in the previous section, which are representative for a series of measurements performed for a wide range of modulation frequencies and bias currents in our laboratory. Both spectra appear slightly distorted as compared to the calculated spectrum obtained for pure intensity modulation (Figure 3). As frequency modulation becomes more dominant (for higher bias currents and lower modulation frequencies), asymmetry increases, which is explained by the more asymmetric laser spectrum. In fact, fitting of the laser tuning parameters following the procedure described in17 revealed a ratio of the high frequency sideband to low-frequency sideband of approximately 4:1, which is in good agreement with results for a directly modulated DFB-QCL reported elsewhere.17 The experimental conditions can hence be seen as an intermediate situation between intensity modulation and single sideband modulation discussed above. Also, the negative peak phase of the HPSDS spectrum strongly depends on the operation point. The HPSDS signal for D

DOI: 10.1021/acs.analchem.7b00303 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

(corresponding to a peak absorbance of roughly 0.0048) for an integration time on the lock-in amplifier of 10 ms is obtained (corresponding to 225 ppb*m/Hz1/2 for CO at 2179.77 cm−1). At 240 mA and 100 MHz, the limit of detection is improved by a factor of 2 (to 12.15 ppb) at the cost of a slightly deteriorated linearity (as previously shown). For 2f-WMS we found a LOD of 3.6 ppb. Since the sensitivity of WMS decreases toward larger concentration (compare Figure 6), the minimum detectable change in concentration at a given concentration increases for larger concentrations. For 5 ppm, we found a minimum detectable change in concentration of 22.5 ppb, i.e., an increase by a factor of 6 relative to the LOD at low concentration. In contrast, the sensitivity of HPSDS is constant for all concentrations under study. In the estimation of the LODs presented above, the influence of fringes on the measurements has to be discussed. Fringes affect both the amplitude and phase of recorded signals, and hence influence absorptionas well as dispersion measurements. However, as shown in Figure 7, for WMS, the much

220 mA and 160 MHz is almost half of that for 240 mA and 100 MHz. The higher the DC current and the lower the modulation frequency (stronger frequency modulation), the higher the negative peak of the HPSDS signals. This increase is readily explained when considering the phase spectrum obtained in frequency modulation spectroscopy, which exhibits a 180° phase jump at the center of the line.22 As will be discussed in the following section, this apparent advantage of a higher HPSDS signal comes at the cost of linearity. Therefore, an interesting balance between sensitivity and linearity appears. Analysis of the Performance of HPSDS and Comparison with WMS. The performance of the HPSDS setup is discussed in terms of linearity, limit of detection (LOD), influence of pressure, and the effect of power fluctuations. Additionally, a comparison with the performance of a 2f-WMS setup is established. To test linearity, gas samples with increasing concentrations of CO in N2 between 0 and 5 ppm at a pressure of 100 mbar, ranging in peak absorbance from 0 to 1.07, were pumped through the cell. The trough of the HPSDS phase signal and peak amplitude for 2f-WMS were used for a linear calibration (Figure 6). For HPSDS, it is straightforward to see the

Figure 7. Comparison of the effect of interference fringes for HPSDS and 2f-WMS. The lower modulation frequencies that are required for WMS induce a stronger frequency modulation contribution on the laser that amplifies the etalon effect. Both measurements were performed for a concentration of 5 ppm of CO in N2. The 2f-WMS was shifted by 0.09 cm−1 to higher wavenumbers for viewing purposes. Figure 6. Comparison of the linearity in the measurement of concentration for the HPSDS signal and the 2f-WMS peak-to-peak signal. The output signals have been normalized with an equal slope for the smallest range of concentrations.

larger modulation depth causes amplified interference fringes in the recorded spectra (most likely created in the multipass cell, no artificial etalons were introduced). Therefore, to provide the previous LODs, it was necessary to fit the interference fringes for WMS (the LOD is roughly 30 ppb if not compensated). No fitting whatsoever was used for HPSDS, as the modulation depth is much smaller. The third comparison parameter between absorption and dispersion spectroscopic methods is the response to increasing pressure. This is an interesting comparison as if the pressure on the sample changes without notice and/or compensation on the instrument, the output signal will provide an erroneous reading of concentration for the two methods. With both setups adjusted for optimum performance 100 mbar (optimum modulation frequency and optimum modulation depth, respectively), increasing pressure widens the line width of the spectral feature, forcing both setups into the suboptimal operation regime. The results of this test are presented in Figure 8 and show that HPSDS (a similar behavior for the two operation points is found on this test) is more sensitive to pressure than 2f-WMS even though the responses finally converge. Adjusting the measurement to increasing pressures is in general easier for WMS, for which the amplitude of the

influence of the operation point of the QCL on the linearity. As expected, the higher the DC current and the lower the modulation frequency, i.e., the more pronounced frequency modulation becomes, the more nonlinear the phase becomes. For 220 mA and 160 MHz, the R2 value of the linear regression equals 0.9997, indicating the exquisite linearity of HPSDS. For the given range of concentrations, 2f-WMS yields a strongly nonlinear signal. In general, the sensitivity and linearity of absorption-based techniques decreases rapidly toward large absorptions due to the exponential relation with concentration. From the previous set of measurements, the LODs achievable with both methods were estimated as the noise equivalent concentration, i.e., the concentration yielding a signal equal to 3 times the standard deviation of the baseline signal (the ratio between 3 times the noise floor σ recorded for a sample containing no CO and the sensitivity obtained from Figure 6 assuming low concentration). For the QCL operating at 220 mA and 160 MHz, a detection limit of 22.5 ppb E

DOI: 10.1021/acs.analchem.7b00303 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

dispersion-based methods in terms of linearity and immunity to power fluctuations. We found very good linearity (R2 = 0.9997) for concentrations corresponding to peak absorbances from 0 to 1.07 using the traditional HPSDS architecture. Nonetheless, in a very recently proposed adaptation of HPSDS (2f-HPSDS, based on second harmonic phase detection),5 linear behavior was demonstrated even for an absorbance of 3 at the cost of a more complex setup. In our test, 2f-WMS provided a lower LOD than HPSDS. Due to its linear response, the minimum detectable change in concentration is smaller for HPSDS for small concentrations. Nonetheless, fringes were much more dominant in the WMS spectra than in HPSDS, making it necessary to remove them from the WMS spectra in post processing. Modifications to the HPSDS setup, like the use of a lock-in amplifier with adequate bandwidth, should noticeably improve its LOD. We could also mention that the influence of the operation point of the laser on HPSDS provides the ability of dynamically balancing the performance of a setup for either targeting maximum sensitivity or optimum linearity just by changing the bias point of the laser and the modulation frequency. HPSDS was outperformed by 2f-WMS in the influence of pressure, even though both techniques require for pressure compensation. Finally, unlike 2f-WMS, the demonstrated independence of the HPSDS signal from the power reaching the detector represents a major advantage of dispersion over absorption spectroscopic sensors. In comparison with other spectroscopy methods, HPSDS stands out for the simplicity of its architecture. The setup is in fact the same as for WMS, with the differences of an increased modulation frequency and the detection of the phase of the first harmonic of the modulation signal instead of the amplitude of its second harmonic. In addition, data processing and concentration retrieval are extremely simple and in clear contrast to the normalization (and very often fitting) procedures required by 2f-WMS or the high-speed acquisition and signal processing of CLaDS. HPSDS also provides an additional advantage over CLaDS, which is the direct measurement of the dispersion profile, rather than the profile’s wavelength derivative that is recorded in CLaDS. The same applies to WMS, that, in the same way, does not provide a direct measurement of the absorption profile but a wavelength derivative that is dependent on many contributions and, without significant efforts, does not allow for calibration-free operation. The main drawback of dispersion spectroscopic techniques is the increased modulation frequencies that are required for optimum performance, reaching several GHz at atmospheric pressure. Much faster electronics for the detector and the laser are then required, increasing the cost (exclusively the cost associated with the electronics, that it is very often a small part of the overall budget) of HPSDS sensors beyond that of WMS instruments. Nonetheless, as the price of fast optoelectronics decreases, this cost imbalance is expected to be reduced. In the same way, several groups are working toward overcoming this issue.10,23 Operating HPSDS at lower modulation frequencies is also a feasible option, yielding slightly reduced sensitivities while the advantageous characteristics of HPSDS persist.

Figure 8. Effect of pressure in the output of the setups for a concentration of CO of 1 ppm.

modulating current is adapted, than for HPSDS, for which the modulation frequency is adapted. The final test performed aimed to compare how the level of power reaching the photodetector influences the output signal of both measurement methods. The results are provided in Figure 9. For WMS and most other absorption-based methods,

Figure 9. Influence of the level of power on the detector on the HPSDS and 2fWMS signals for a concentration of CO of 1 ppm.

the output signal is highly dependent on the power level. This issue, known as the normalization problem, requires careful referencing to laser intensity. In WMS, this is typically addressed by referencing the 2f-signal to the 1f-signal.21 In contrast, apart from the effect on the signal-to-noise ratio evaluated in ref 2, the HPSDS signal is virtually immune to the amount of power reaching the receiver, and no referencing is required.

■

CONCLUSIONS In this article, to the best of our knowledge, the first HPSDS setup operating in the mid-IR was discussed and experimentally validated. An experimental characterization of its response to changing concentration, pressure, and optical power level has been presented. Besides this, the performance of the abovementioned HPSDS setup and a 2f-WMS setup in response to various parameters of influence was compared and discussed. For the implementation of the mid-IR HPSDS setup, a directly modulated quantum cascade laser, instead of the optical intensity modulators used in previous publications, has been employed. The consequences of mixed frequency and intensity modulation typical for directly modulated semiconductor lasers were discussed and investigated. These mostly undesirable effects were minimized by selecting the right operation point of the laser. The results of the comparison between HPSDS and 2f-WMS yield strong conclusions like the superior performance of

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Telephone: +34 624 8390. F

DOI: 10.1021/acs.analchem.7b00303 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry ORCID

Pedro Martín-Mateos: 0000-0003-1656-9456 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The work by P.M.-M. has been performed in the frame of the “Ayudas para la movilidad del programa propio de investigación” from the Universidad Carlos III de Madrid. P.M.-M. and P.A. would also like to thank the Spanish Ministry of Economy and Competitiveness for supporting the projects under the RTC-2014-2661-7 and TEC-2014-52147-R grants. J.H. and B.L. acknowledge financial support received from the Competence Center ASSIC (Austrian Smart Systems Integration Research Center) as part of the Austrian COMET (Competence Centers for Excellent Technologies) program.

■

REFERENCES

(1) Wysocki, G.; Weidmann, D. Opt. Express 2010, 18, 26123− 26140. (2) Martín-Mateos, P.; Acedo, P. Opt. Express 2014, 22, 15143− 15153. (3) Nikodem, M.; Wysocki, G. Sensors 2012, 12, 16466−16481. (4) Nikodem, M.; Plant, G.; Sonnenfroh, D.; Wysocki, G. Appl. Phys. B: Lasers Opt. 2015, 119, 3−9. (5) Ding, W.; Sun, L.; Yi, L.; Ming, X. Appl. Opt. 2016, 55, 8698− 8704. (6) Nikodem, M.; Wysocki, G. Opt. Lett. 2013, 38, 3834−3837. (7) Hangauer, A.; Spinner, G.; Nikodem, M.; Wysocki, G. Appl. Phys. Lett. 2013, 103, 191107. (8) Nikodem, M.; Plant, G.; Wang, Z.; Prucnal, P.; Wysocki, G. Opt. Express 2013, 21, 14649−14655. (9) Nikodem, M.; Weidmann, D.; Wysocki, G. Appl. Phys. B: Lasers Opt. 2012, 109, 477−483. (10) Martín-Mateos, P.; Jerez, B.; Acedo, P. Opt. Lett. 2014, 39, 2611−2613. (11) Martín-Mateos, P.; Jerez, B.; Acedo, P. Opt. Express 2015, 23, 21149−21158. (12) Westberg, J.; Sterczewski, L.; Wysocki, G. Imaging and Applied Optics 2016; OSA: Washington, DC, 2016; p LT1G.4. (13) Bomse, D. S.; Stanton, A. C.; Silver, J. A. Appl. Opt. 1992, 31, 718−731. (14) Moser, H.; Pölz, W.; Waclawek, J. P.; Ofner, J.; Lendl, B. Anal. Bioanal. Chem. 2017, 409, 729−739. (15) Kluczynski, P.; Gustafsson, J.; Lindberg, Å. M.; Axner, O. Spectrochim. Acta, Part B 2001, 56, 1277−1354. (16) Olesen, H.; Jacobsen, G. IEEE J. Quantum Electron. 1982, 18, 2069−2080. (17) Hangauer, A.; Spinner, G.; Nikodem, M.; Wysocki, G. Opt. Express 2014, 22, 23439−23455. (18) Aellen, T.; Maulini, R.; Terazzi, R.; Hoyler, N.; Giovannini, M.; Faist, J.; Blaser, S.; Hvozdara, L. Appl. Phys. Lett. 2006, 89, 091121. (19) Hinkov, B.; Hugi, A.; Beck, M.; Faist, J. Opt. Express 2016, 24, 3294−3312. (20) Genner, A.; Gasser, C.; Moser, H.; Ofner, J.; Schreiber, J.; Lendl, B. Anal. Bioanal. Chem. 2017, 409, 753−761. (21) Rieker, G. B.; Jeffries, J. B.; Hanson, R. K. Appl. Opt. 2009, 48, 5546−5560. (22) Bjorklund, G. C.; Levenson, M. D.; Lenth, W.; Ortiz, C. Appl. Phys. B: Photophys. Laser Chem. 1983, 32, 145−152. (23) Nikodem, M. Opt. Eng. 2016, 55, 044103.

G

DOI: 10.1021/acs.analchem.7b00303 Anal. Chem. XXXX, XXX, XXX−XXX