Supercontinuum broadband laser absorption spectroscopy (SCLAS) is a fairly recent development in the field of broadband spectroscopy. Based on a theoretical proposal by Orofino and Unterleitner in 1976 [1], only the combination of recent developments in the field of high-speed high-bandwidth optical detection and supercontinuum light sources have allowed for application of this technique in general and specifically in the field of gas phase combustion research [2–6]. Due to its broad spectral coverage and high temporal resolution, SCLAS is a promising technique for diagnostics in the context of newly developed combinatory combustion technologies, such as homogeneous charge compression ignition or spark ignition direct injection. All these approaches to pollutant reduction share the common need for extensive simultaneous measurements of multiple combustion parameters with high temporal resolution in high-pressure and high-temperature environments, which SCLAS is able to provide.

SCLAS employs recently available supercontinuum laser light sources (SCL), which emit pulsed laser light, covering the spectrum from 400 to 2400 nm at repetition rates of up to 80 MHz. The emitted pulses of SCL sources are fairly short pulses ($\sim 100\mathrm{ps}$), which are afterward dispersed in time to allow for frequency-resolved intensity detection. The recorded absorption spectra can be fitted against spectral databases for determination of species concentrations and other combustion parameters such as temperature. So far, SCLAS in the combustion context has been applied almost exclusively in multipass or long-pass setups. Sanders [2] used an 1840 mm absorption cell with pressures up to 10 bar to resolve spectral feature of ${\mathrm{H}}_2\mathrm{O}$, ${\mathrm{CO}}_2$, and ${\mathrm{C}}_2{\mathrm{H}}_2$ with a resolution of up to $1.5{\mathrm{cm}}^{-1}$. Werblinski et al. [3] used SCLAS in a McKenna-type flat flame burner with a multipass setup and an effective absorption path length of at least 600 mm. Previous experiments of the same group [4] were performed using a pressurized flow cell with an absorption length of 300 mm and temperatures up to 900 K. For the measurements in Kaminski et al. [5], an absorption path length of 2800 mm was used to measure ambient vapor ${\mathrm{H}}_2\mathrm{O}$ species concentrations. In Hult et al. [6], a multipass setup of 105 mm was used for an unburned mixture of 20% ${\mathrm{CH}}_4$ and air at room temperature. Previous research has been so far focused on basic studies, including very long absorption paths and unburned gas mixtures at ambient temperature and slightly elevated pressures, but is still lacking validation for in-flame measurements. Since, especially in high-pressure and high-temperature environments, the broad spectral width of SCLAS is beneficial for the fitting process and the quality of the results [7–9], extending the SCLAS diagnostic principle toward these applications is most promising for in situ diagnostics in enclosed combustion. A necessary step in this way is the validation of the SCLAS for quantitative in-flame measurements with single-pass absorption paths and path lengths on the order of cylinder diameters ($\sim 80\mathrm{mm}$) and shorter.

In this Letter, we demonstrate the applicability of SCLAS in the context of single-pass spatially resolved absolute mole fraction measurements of combustion species within a flame, which are validated against reference measurements by multiple other diagnostic techniques. To achieve this, we utilized an atmospheric, laminar, nonpremixed ${\mathrm{CH}}_4/\mathrm{air}$ flame consisting of two flame sheets on a Wolfhard–Parker burner (WHP) [10]. The WHP burner enables the quantitative validation of new measurement methods for combustion diagnostics, as it has been extensively characterized with spatially resolved measurements of more than 17 individual species, temperatures, and flow-field with a high number of independent measurement techniques. Measurements, which include, for example, the species ${\mathrm{CH}}_4$, CO, and ${\mathrm{CO}}_2$ as well as temperatures and flow velocities, can be found in [10]. We recorded broadband methane spectra ranging over $110{\mathrm{cm}}^{-1}$ with an absorption path of 41 mm, which were, consequently, fitted against spectral database HITRAN 2012 [11] and evaluated against reference measurements for this burner [10,12,13].

For the experiments, the setup shown in Fig. 1 was used. As a laser light source, a SuperK EXTREME EXW-12 by NKT Photonics A/S was utilized. This source uses a pump wavelength of 1064 nm and generates the supercontinuum laser light in a photonic crystal fiber with a spectrum of 400–2400 nm. To select a specific region of this spectrum for the experiment, spectral filtering was performed using the SuperK SELECT by NKT Photonics A/S, which provides acousto-optical tunable filters (AOTFs) to select up to eight spectral windows (FWHM 10–18 nm). The emitted light is coupled into a single-mode fiber and chromatically dispersed in a dispersion compensating module (OFS Fitel Denmark ApS, LLMicroDK:1500, $-1500\mathrm{ps}/\mathrm{nm}$ at 1550 nm). The dispersed pulses are then modulated by the absorption features of the gases present in the flame of the WHP burner.

 

Fig. 1. Experimental setup for SCLAS measurements on an atmospheric, laminar, nonpremixed ${\mathrm{CH}}_4/\text{air}$ flame consisting of two flame sheets stabilized using a wire screen chimney (Wolfhard–Parker burner [10]). Supercontinuum laser pulses, emitted by a supercontinuum laser light source (SCL), are spectrally filtered and dispersed chromatically in a dispersion compensating module (DCM). After passing through the flame, the light is coupled into a multimode fiber (MMF, 50 μm) and detected by a photodiode (PD) and an oscilloscope (DAQ).

Download Full Size | PDF

After the transmission across the flame, the light is coupled into a multimode fiber (step-index fiber, 50 μm core diameter) and detected by a fiber-coupled high-speed high-bandwidth InGaAs photodiode (Newport Inc., New Focus 1544-B-50, 12 GHz bandwidth, $-1000\mathrm{V}/\mathrm{A}$ transimpedance amplifier gain). The electrical output of the photodiode is recorded with an oscilloscope (LeCroy Corporation, SDA 820Zi-A, 20 GHz bandwidth, $40\mathrm{GS}/\mathrm{s}$).

The complete spectroscopic system has been characterized and verified regarding its spectral behavior with a grating spectrometer (AOTF, FWHM, etc.). Furthermore, it is necessary for the determination of the optical resolution to measure the dispersion of the DCM, since the dispersion is directly related to the spectral resolution of the system by the following equation [2]:

In Eq. (1), $\mathrm{\Delta }\nu$ (nm) is the spectral resolution, $\tau$ the FWHM of the shortest optical pulse (ps), resolvable by the photodiode, and $\dot{\nu }$ the scan rate (nm/ps) as a parameter of the dispersion fiber used in the setup. Based on the rise-time of 38 ps of the currently used photodiode, $\tau$ can be estimated to be 76 ps. The dispersion of the DCM was measured using a novel approach using the available AOTF module (described in detail in a future publication). Based on the measured delay (ps) of different wavelengths (nm), the dispersion can be determined as the first derivative of that delay, given that the dispersion (ps/nm) is the slope of the delay for a given length of fiber [14]. For different lengths of dispersion compensating fiber, the dispersion must be scaled according to length. For the DCM used in these experiments (9.382 km of fiber), the measured dispersion curve is shown in Fig. 2. Based on the measurements of the dispersion, the optical resolution of the system can be estimated according to Eq. (1) for the relevant wavelength range of 1650–1700 nm:

This optical resolution is equivalent to $\sim 0.04\mathrm{nm}$ at 1675 nm. To allow for the recording of broadband spectra of methane in the WHP burner, this resolution is more than sufficient and could even be decreased for high-pressure environments. Since the optical resolution is directly related to the applied dispersion, a reduction in resolution is achieved by a reduction in dispersion. Such a reduction, on the other hand, would allow for an increase in repetition rate, since the time duration of a dispersed pulse is reduced as well, and no overlap of pulses is present at higher repetition rates.

 

Fig. 2. Measurement of the dispersion of the DCM used in the experiments. The delay for multiple wavelengths was recorded and, subsequently, fitted with a fourth-order polynomial function, which was thereupon used for calculation of the dispersion as the first derivative of the fourth-order polynomial function.

Download Full Size | PDF

For the experiments, a modified WHP burner design was used employing nitrogen purge slots to suppress end-flame effects [13]. The modified WHP burner consists of a center slot of methane (length: 41 mm) with two adjacent slots of air. The two diffusion flames are stabilized against a wire screen chimney, and the complete setup is mounted on a two-axis motion stage to allow for precise positioning of the flame. The WHP burner was operated with an exit velocity for methane of $11\mathrm{cm}/\mathrm{s}$ and for both air slots of $22\mathrm{cm}/\mathrm{s}$ as well as for nitrogen of $22\mathrm{cm}/\mathrm{s}$. Using the motion stage, measurements were taken at different transversal positions as well as at different heights above the burner surface.

Based on the HITRAN 2012 spectral database [11], multiple multiline Voigt shapes (over 1000 absorption lines) were fitted against the measured spectra using a previously validated Levenberg–Marquardt algorithm [13] to minimize the residual. A higher-order polynomial function was fitted to the background as a reference intensity. The obtained direct absorption spectra were evaluated using the extended Lambert–Beer law, as shown in Eq. (3):

In Eq. (3), $I(\lambda )$ describes the light intensity after absorption, $I_0(\lambda )$ the light intensity before absorption, $S(T)$ the line strength, $g(\lambda -{\lambda }_0)$ the line form function, $N$ the number density, $L$ the absorption path length, $\mathrm{Tr}(t)$ the broadband transmission fluctuation (i.e., particles), and $E(t)$ the broadband background emission. Given the near static conditions of this experiment, $\mathrm{Tr}(t)$ is almost identical to one. Using Eq. (3) and previously performed flame temperature measurements as well as pressure measurements in combination with the tabulated line strength values and the ideal gas law, the concentration of methane was determined and compared to reference measurements found in [10].

Figure 3 shows a raw measurement of methane in the WHP burner. For this measurement, 9870 single pulses (2 MHz repetition rate) of the SCL were averaged, resulting in a spectral sampling rate approximately 200 Hz (5 ms).

 

Fig. 3. Raw measurement using the SCLAS setup of methane gas concentration in WHP burner at transversal position 0.5 mm and height 7 mm above burner surface.

Download Full Size | PDF

After application of time-wavelength correlation (nonlinear dispersion of DCM), application of background correction and fitting of appropriate spectral features, an exemplary result spectrum is shown in Fig. 4. The estimated fractional absorption from the local residual of $1.133\times {10}^{-2}$ gives a signal to noise ratio of 24 at 489 K (3 %_Vol. or 1210 ppm·m, [15]). In advanced processing algorithms, advanced combustion parameters such as temperature and species concentration of additional combustion educts and products (i.e., ${\mathrm{H}}_2\mathrm{O}$) will be derived from single SCLAS broadband spectra as opposed to the need to combine multiple narrowly tuned laser sources for multiparameter measurements. SCLAS might suffer from slightly less optical resolution but can overcome this drawback by providing robust and versatile broadband measurements that allow for a unified approach to multiparameter combustion measurements.

 

Fig. 4. Exemplary methane spectrum as recorded with the SCLAS setup at lateral WHP position of $+0.5\mathrm{mm}$ with a temperature of 489 K. For the fit 1001 methane lines from HITRAN 2012 [11] (Voigt line-shape functions) were fitted simultaneously with a higher-order polynomial reference intensity $I_0$, as described in [13].

Download Full Size | PDF

By combining the recorded spectra of multiple positions, a spatially resolved methane profile across the flame sheets can be measured. In Fig. 5, the recorded spectra at the lateral positions $+0.5\mathrm{mm}$ (489 K), $+1.0\mathrm{mm}$ (571 K), $+1.5\mathrm{mm}$ (685 K), and $+2.0\mathrm{mm}$ (822 K) are shown at their corresponding temperature.

 

Fig. 5. Methane mole fraction measurements at 7 mm above burner surface and different transversal positions. Different lateral positions are plotted at their corresponding temperature, which increases from the center outward from 489 to 822 K.

Download Full Size | PDF

To evaluate the performance of SCLAS, the measured spatial concentrations for methane according to [10] were compared against our SCLAS measurements, as shown in Fig. 6. It can be derived from the comparison that the species concentrations of methane obtained with SCLAS are consistent with the concentrations of Norton et al. [10], thus validating SCLAS as a combustion diagnostics tool for absolute species concentration measurements in flames.

 

Fig. 6. Comparison of spatially resolved methane concentration comparing data from Norton et al. [10] and our recent SCLAS measurements.

Download Full Size | PDF

Overall, in this Letter we were able to validate SCLAS as a versatile and feasible tool for quantitative in situ combustion diagnostics. It was shown that SCLAS is able to correctly determine species concentrations inside a laminar flame. Additionally, the setup presented here allowed for spatially resolved quantitative in-flame gas concentration measurements with an absorption path length of only 41 mm in a single-pass setup. Based on these results, SCLAS can be considered as a validated quantitative diagnostic system for in-flame combustion diagnostics. Given the fully fiber-coupled nature of SCLAS, this diagnostic technique is, furthermore, well suited for use in harsh combustion environments as a broadband diagnostic tool.

In the future, we will continue to apply SCLAS to the WHP burner setup and record, as well as evaluate, complete spatial profiles to further characterize the behavior and applicability of SCLAS in combustion environments. Additionally, we currently investigate the applicability of this technique in more complex combustion scenarios, including high-pressure chemical reactors or internal combustion engines. Such environments provide new challenges for the SCLAS system, since the pressure levels are significantly higher as well as optical accessibility being limited. Regarding the data evaluation, we see great potential for specifically designed SCLAS fitting algorithms that fully utilize the spectral information contained in SCLAS spectra. As mentioned earlier, the temporal resolution can be increased in high-pressure environments as well, since reduced requirements for optical resolution allow for an increase in pulse repetition rate.

Furthermore, an optimization of the SCL source itself for the nIR range would be beneficial to the signal quality and the overall spectrometer performance, since the noise level inherent to the supercontinuum generation could be reduced when reducing the spectral range to the nIR [16].