Sensing chlorinated hydrocarbons via miniaturized GaAs/AlGaAs thin-film waveguide flow cells coupled to quantum cascade lasers

Mid-infrared (MIR) sensors based on attenuated total reflection (ATR) spectroscopy provide robust, rapid and sensitive platforms for the detection of low levels of organic molecules and pollutants. Nowadays, MIR (3–15 μm) spectroscopy has evolved into a versatile sensing technique providing inherent molecular selectivity for the detection of organic and inorganic molecules. The excitation of vibrational and rotational transitions enables the qualitative and quantitative analysis of molecular constituents in solid, liquid, and vapor phases, which facilitates the application of MIR chem/bio sensors for on-site environmental analysis in scenarios such as trace pollutant monitoring or spill detection. This report presents the first integration of thin-film gallium arsenide/aluminum gallium arsenide (GaAs/AlGaAs) into a miniaturized liquid flow cell designed for continuous trace analysis of chlorinated hydrocarbons (CHCs) in water coupled to a broadly tunable quantum cascade laser (QCL), which facilitates in-field deployment of QCL-based sensing devices ensuring water quality and water safety.

In water quality monitoring, recent advancements in analytical technologies and methodologies have enabled the detection, identication, and quantication of an increasing number of relevant constituents at ever decreasing concentration levels. 3,4hile a majority of these strategiesand especially chromatographic methods including GC-MS, GC-MS/MS, and HPLC [5][6][7] deliver the demanded qualitative and quantitative information, the associated instrumentation frequently remains rather bulky and conned to operation within a wellcontrolled laboratory environment.In addition, it is common that a series of sample processing steps (i.e., sampling, separation/extraction, etc.) prior to the actual analysis are required.These considerations lead to the associated cost, time, and potential uncertainties, in particular during the analysis of volatile organic compounds (VOCs) such as chlorinated hydrocarbons (CHCs).Based on the need for improved monitoring and sensing systems ideally requiring minimal sample preparation while facilitating compact, robust, rapidly responding, and sensitive devices providing inherent molecular selectivity, IR sensor technologies have emerged as a viable concept in the past two decades, and have in part already successfully been applied in water monitoring scenarios. 8ecent advancements in the miniaturization of MIR sensing techniques should facilitate their more ubiquitous presence in environmental monitoring situations. 9,10Correspondingly, MIR sensing techniques benet in particular from the rapid development of state-of-the-art light source technologies.In particular, quantum cascade lasers (QCLs) and interband cascade lasers (ICLs) are nowadays available throughout almost the entire 3-15 mm spectral window providing broad wavelength tunability (up to 400 cm À1 ), compact dimensions, long life times, and sufficient optical power.2][13][14][15] Even more recently, the rst accessory for routinely utilizing thin-lm waveguides in combination with QCLs has been reported providing a methodology complementary to conventional attenuated total reection (ATR) crystals and sampling accessories coupled to Fourier transform infrared (FT-IR) spectrometers. 15Considering the miniaturization potential of QCLs and such matching sampling accessories, it is anticipated that MIR sensing techniques will become signicantly more prevalent and utilized in environmental monitoring and in-eld sensing scenarios.
Evanescent eld sensing via ATR spectroscopy has evolved into the most frequently deployed sampling technique for probing opaque or strongly absorbing media, as the evanescent eld penetrates only a few micrometers into the adjacent medium, thereby minimizing background absorptions of the matrix, e.g. in aqueous solutions.The penetration depth (d p ) of the evanescent eld at a certain wavelength l may be approximated via with the refractive index of the waveguiding structure represented as n 1 , the refractive index of the adjacent medium (i.e., usually the sample) as n 2 , and the angle of incidence at the waveguide/sample interface as q. 16outinely used internal reection elements (IREs) are usually fabricated at a thickness of several hundreds of mm up to the millimeter scale from a limited set of MIR transparent materials dominated by bulk crystals composed of zinc selenide (ZnSe), zinc sulde (ZnS), silicon (Si), germanium (Ge) and diamond in a wide variety of lateral dimensions (a few centimeters to a few millimeters in length, width or diameter).Alternatively, optical bers made from polycrystalline silver halides or chalcogenide glassesjust to name a fewhave been replacing conventional IREs in a variety of sensing congurations. 17,18However, further miniaturization of the active sensing element, i.e. the development of ultra-thin sections of beroptic waveguides with a thickness on the order of magnitude of the propagating wavelength (few micrometers) was limited due to the mechanical stability of such coreonly structures.
Nowadays, single-or few-mode propagating thin-lm waveguides covering the entire MIR spectral range are readily available based on semiconductors and diamond materials, and are anticipated to largely replace conventional IREs and optical bers as MIR transducers in combination with frequency-matched light sources such as QCLs and ICLs. 19][21] Next to advances in thin-lm waveguide technologies, appropriate surface modication of the transducing interface enables capturing, preselecting or enriching target molecules within the evanescent eld for analytical signal amplication.In addition, hydrophobic coatings further reduce water background absorptions.In particular, enrichment membranes based on aliphatic polymers such as ethylene/propylene copolymer (EPCo) or polyisobutadiene, as well as sol-gel based coatings were studied as a convenient surface modication approach readily enriching VOCs and semi-volatile organic constituents (SVOCs), and enabling detection limits at low ppb levels in aqueous matrices including surface waters, aquifers and seawater.][24][25][26][27][28][29][30][31] In the present study, GaAs thin-lm waveguides for analyzing organic constituents in water were introduced via the development of a miniaturized ow cell enabling efficient coupling of a miniaturized liquid phase sensing accessory to QCLs and QCL spectrometers.This MIR sensing platform was applied for the detection of chlorinated hydrocarbons in water at low ppm concentration levels.

Experimental
A stainless steel waveguide accessory was equipped with a GaAs/ Al 0.2 Ga 0.8 As thin-lm waveguide with dimensions of 5 Â 10 mm.Aer optimization via nite-element (FEM) simulations, GaAs waveguide structures were epitaxially grown onto a Sidoped GaAs substrate at a thickness of 10 mm on top of a previously deposited 10 mm AlGaAs optical buffer layer.Advancing a previously described chip assembly, 15 a miniaturized ow cell was designed and fabricated.As shown in Fig. 1, the ow cell was connected via PTFE tubing to a peristaltic pump.To avoid and sample carryover, the peristaltic pump was operated in reverse mode at a pump rate of 1 mL min À1 20 mL of a 1%wt EPCo solution (in n-hexane) was deposited onto the waveguide surface.Aer evaporation of the solvent, a uniform polymer membrane was obtained with a thickness of approximately 10 mm. 23s previously described, 15 the broadly tunable QCL (MIRcat, Daylight Solutions, San Diego, CA, USA) along with a mercurycadmium-telluride (MCT) detector system was coupled to a custom-built gated-integrating amplier (GIA) for converting laser pulses to a continuous signal.Spectra acquisition was performed via a LabView script (LabView 2014, SP1, V 14.0.1f3,National Instruments GmbH, München, Germany).A schematic of the experimental setup is shown in Fig. 1.

Sensor optimization
The quality of the polymer membrane is crucial to ensure reproducible hydrocarbon enrichment.Therefore, the coating of an EPCo layer onto a GaAs thin-lm waveguide surface was optimized via a series of 10, 15, 20, and 25 mL of EPCo solution View Article Online applied onto the waveguide surface.Aer tuning the QCL emission wavelength to 1640 cm À1 (i.e. the H-O-H bending vibration of water), water was pumped across the waveguide surface and the associated IR signal was recorded for 1 h.For a membrane deposited from 10 and 15 mL of EPCo solution, a signicant decrease in the IR signal (i.e., damping) was evident, which indicates that the reconstituted polymer membrane was not of sufficient thickness and homogeneity for preventing the emergence of a signicant water background signal recorded via the evanescent eld.For membranes established from 20 and 25 mL solutions, a signicantly reduced damping of the IR signal at 1640 cm À1 resulted from water, which results from equilibration of the outermost layers of the polymer coating in agreement with previous observations. 32ence, during subsequent studies polymer membranes were reconstituted from 20 mL EPCo solution aliquots yielding a coating thickness of approximately 10 mm.A typical membrane preconditioning experiment is shown in Fig. 2a.During equilibration of the membrane, the IR signal at 1640 cm À1 decreases (within the rst 10 min), and then increases until equilibration of the coating is achieved, and the signal remains constant.

Measurement procedure
For establishing the measurement routine, a diluted perchloroethylene (PCE) solution with a concentration of 10 ppm was prepared in deionized (DI) water.PCE was used as a model analyte for chlorinated hydrocarbons providing discriminatory absorption features within the tuning range of the applied QCL light source (i.e.@910 cm À1 ).Therefore, single wavelength experiments were performed tuning the QCL to emit at 910 cm À1 (i.e.10.99 mm) operating in pulsed mode at a current of 1100 mA, which resulted in an emission power of approx.300 mW.Aer aligning the thin-lm waveguide such that maximal energy throughput was achieved, the ow cell was purged with DI water for 30 min for establishing a stable baseline.Subsequently, the purging solution was switched to an analyte solution containing CHCs, again 30 min until an enrichment equilibrium for PCE between the polymer membrane and PCE in solution was reached. 33Aer each measurement, the polymer membrane was regenerated via purging with DI water for at least 30 min until the initial IR signal was recovered.A typical measurement cycle is illustrated in Fig. 2b.The long-term stability of the polymer membrane was conrmed via numerous measurement cycles executed during the present study.

Sensor calibration
To calibrate the sensor, PCE calibration solutions were prepared at concentrations of 5, 10, 15, and 20 ppm in DI water.For each concentration, three independent measurement cycles were recorded and individually evaluated following A ¼ Àlog(I/I 0 ), as shown in Fig. 3.The detector signal obtained during water purging was used as I 0 , while the analyte signal at equilibrium between the solution and polymer was applied as I. Based on DIN 32645, the limit of detection (LOD) was calculated to be 5 ppm.This rather high LOD is mainly attributed to the low optical emission power of the applied QCL at 910 cm À1 .It is anticipated that QCLs specically tailored to a desired wavelength window will provide emission powers signicantly exceeding the levels currently provided by generic devices.It should also be noted that in comparison to IR sensors demonstrating enrichment-based analysis of CHCs signicantly larger waveguide/sample interface areas have been applied, while the current device based on thin-lm waveguide technology evidently achieves respectable sensitivities using only a few square millimeters of the sensing surface given the dimensions of the GaAs chip (10 Â 5 mm with an actively available surface aer mounting of approx.8 Â 3 mm).

Broadband detection via broadly tunable QCLs
In order to demonstrate the utility of the developed QCL-based infrared sensor concept for the simultaneous determination of several analytes, a mixture of PCE and trichloroethylene (TCE) was investigated.TCE has a characteristic absorption band at 930 cm À1 , and may thus be discriminated from PCE.The ow cell was purged as described before; however, the QCL was now continuously tuned between 950 and 890 cm À1 while the detector signal was recorded.Again, QCL operation was in pulsed mode at a current of 1100 mA, which resulted in an output power between 100 and 300 mW and a laser linewidth < 1 cm À1 corresponding to the emission characteristics of the used QCL diode.Absorption spectra were again calculated via A ¼ Àlog(I/I 0 ) using the scan during water purging as the background spectrum, and the scan aer signal equilibrium has been achieved as the sample scan.An exemplary spectrum comprising both analytes aer a single scan from 950 and 890 cm À1 with a spectral resolution of 0.5 cm À1 is shown in Fig. 4 aer smoothing with a low-pass FFT lter.The spectrum was recorded at a scan rate of 2 cm À1 s À1 , whereby the spectrum was acquired within 30 s with a signal-to-noise ratio of approx.3 dB.Faster scan rates were limited by the custom made GIA system.The applied QCL would allow scan rates up to 50 cm À1 s À1 , however, at the expense of spectral resolution, which indicates that a balance between detector and light source performance has to be found.The spectral features of PCE appear well separated from TCE ensuring that both target analytes may be individually evaluated.In contrast to single wavelength experiments, baseline dris are prevalent, which may be readily taken into account via multivariate data treatment routines utilizing spectral segments without analyte features taken into account for more robust calibration algorithms.While miniaturization of conventional FT-IR spectrometers directly affects the analytical gures-of-merit such as the signal-to-noise ratio (SNR), the intensity of the laser-based signal or the achievable scan rate in such miniaturized devices remains high due to the minute dimensions of the laser chip providing emitted beam diameters of a few tens of micrometers, thus efficiently coupling into thin-lm transducers.Hence, further increasing the output power, the laser stability, and the tuning range of QCLs will directly and positively affect the achievable SNR while maintaining a small device footprint.Clearly, conventional IR-based chem/bio sensing schemes have demonstrated the fundamental feasibility of this optical sensing concept for VOC and SVOC analysis.Yet, QCL-based sensors in combination with highly efficient thin-lm waveguides and appropriately miniaturized sampling accessories such as the ow cell shown herein will be instrumental in enabling performance characteristics suitable for analytical measurement scenarios demanding for particularly rapid and sensitive in situ measurements under in-eld conditions.

Conclusions
The rst prototype of an IR sensing system comprising broadly tunable QCLs or QCL spectrometers in combination with thin-lm GaAs/AlGaAs waveguides serving as evanescent eld transducers mounted within a miniaturized ow cell for the detection of chlorinated hydrocarbons PCE has been shown.An EPCo polymer membrane coated onto the GaAs transducer serving as a hydrophobic enrichment membrane facilitated low ppm detection limits using a remarkably small (i.e. a few square millimeters) sampling interface.The presented sensor system is modularly designed facilitating further system miniaturization/ integration and adaptability to a wide range of analytes (e.g. via tailoring or selecting appropriate enrichment membranes) and sensing scenarios.Hence, next to the analysis of trace pollutants and environmental sensing it is anticipated that a variety of point-of-care scenarios may likewise benet from this sensing platform.

Fig. 1
Fig. 1 Schematic of the developed thin-film waveguide flow cell accessory coupled to a broadly tunable QCL.Abbreviations: Aptpinhole aperture, Lnsconvex ZnSe lens.

Fig. 2
Fig. 2 Typical sensor response of a 10 mm thick EPCo membrane coated onto a GaAs thin-film waveguide during a preconditioning cycle (1 h) prior to analyte application; IR signal damping at 1640 cm À1 resulting from the presence of H 2 O was recorded (a).Preconcentration behaviour visualized by increased damping of the IR signal at 910 cm À1 due the increasing presence of PCE within the EPCo membrane, and thus, within the penetration depth of the evanescent field.The PCE concentration in the water sample was 10 ppm.Final purging is not shown focusing on the preconcentration step (b).

Fig. 3
Fig. 3 Single wavelength calibration for PCE at 910 cm À1 with N ¼ 3 for each concentration using aqueous calibration samples containing 5, 10, 15 and 20 ppm of PCE.

Fig. 4
Fig.4IR single-scan spectrum of TCE and PCE simultaneously preconcentrated via an EPCo membrane coated onto a GaAs thin-film waveguide from an aqueous sample matrix using a broadly tunable QCL covering the wavelength window of 950-890 cm À1 .The spectrum was smoothed via a low-pass FFT filter.