1. Introduction

For the optimization of internal combustion engine combustion processes, a detailed knowledge about the process of mixture formation is necessary. Therefore, precise information about local temperature and concentration distribution are of utmost interest. One technique which is commonly used to determine these quantities is laser-induced fluorescence (LIF), as it offers the possibility of 2D measurements with a high signal-to-noise ratio. To determine the concentration and temperature profiles inside combustion engines, fluorescence-tracers like acetone, 3-pentanone or toluene are usually added to the air or fuel supply [1–8]. For reproducible quantitative measurements, calibration data in a cell at defined temperature and pressure referring to the condition of the respective application are required. During the compression stroke of a spark ignition (SI) engine temperatures up to 700 K and pressures up to 30 bar are achieved within milliseconds, before the mixture is ignited. However, in a LIF calibration cell the tracer and the air are supplied and heated continuously leading to residence times of several seconds in the heated area [7,9]. Thus, tracer decomposition has to be taken into account for LIF calibration measurements when aiming at quantitative concentration measurements. A first indication of possible tracer consumption is given by the ignition temperature. For example, the auto-ignition of (CH₃)₂CO occurs at 773 K at 1 bar and at 498 K at 11 bar [10], respectively. However, the auto-ignition temperatures may vary about 70 K depending on the flow conditions, the cell size and geometry as well as the material of the hot surfaces [11]. Furthermore, the corresponding ignition delay times are unknown at these conditions. Consequently, there is a strong need for data of tracer stability in a LIF calibration cell in dependence on temperature and pressure at engine relevant conditions. In addition, information about the maximum possible residence time in the heated area without decomposition of the tracer is required for accurate LIF-calibration measurements.

The thermal decomposition of (CH₃)₂CO was analyzed by several groups [12–18], but their work was mainly focused on the chemical reaction mechanism and reaction constants and less on the tracer stability. Consequently, pressure and temperature conditions were not corresponding to the conditions of mixture formation during the engine intake and compression stroke or gas turbine conditions. Thus, no knowledge about acetone decomposition at engine relevant conditions is available in literature.

Up to now, (CH₃)₂CO decomposition was investigated in two different devices. First, (CH₃)₂CO and air were added continuously or discontinuously to a heated glass cell. Measurement of (CH₃)₂CO decomposition by different analytic methods ranging from pressure increase to concentration measurements with gas chromatography are reported at temperatures from 483 K to 783 K [13], 773K to 873 K [12] and 373 K to 1273 K [14]. Secondly, shock tubes can be used, in which the dissociation is induced by a sudden pressure increase caused by a shock wave. (CH₃)₂CO decomposition was analyzed in a heated cell at 1090 K to 1566 K [15] and 1350 K to 1650 K [16]. Although relevant temperatures were investigated in case of the measurements in the heated cell, the corresponding pressure was in all cases low compared to internal combustion engine conditions.

In these cases the degree of (CH₃)₂CO decomposition was determined by different analytical methods. First approaches by using pressure increase [12] were already proven to be strongly inaccurate because of the different side reactions that occur [17]. Volumetric methods [17,18] give high accurate concentration results, but they are rather time consuming and only restricted to one species. Gas chromatography, a standard tool for gas analysis, has been favored for these investigations because of the simultaneous detection of several species [13,14,18]. However, product gas is analyzed in a column after the cell at a different temperature and pressure condition, which could lead to back reactions or recombinations. Therefore, optical techniques are of special interest for in situ measurements. They allow a fast determination of species concentration during decomposition without influencing the process. For example, absorption spectroscopy was already successfully applied to determine the concentration of (CH₃)₂CO and CO₂ [15,16]. However, such a system is rather complex because in most cases one laser source is needed for each species.

In this work, a new approach to investigate (CH₃)₂CO decomposition based on detection of tracer, air and product species in dependence of pressure, temperature and residence time is proposed. The decomposition process was analyzed directly inside the calibration cell by Raman spectroscopy. Compared to other optical techniques, it enables simultaneous in situ measurements of different species with a comparatively simple setup [19]. Hence, the simultaneous investigation of tracer stability is possible during LIF calibration measurements. In addition, this technique can be adapted to other tracers without any change in the optical arrangement.

2. Measurement principle

The basic principle of Raman scattering is described elsewhere [20,21] and therefore only a short description will be given here. When light passes a gas volume, different interactions between light and molecules occur. One of these interactions is linear Raman scattering, a spontaneous inelastically scattering process. The frequency of the Raman scattered light ν _R is given by

To find suitable spectral ranges for the decomposition measurements it is necessary to look at the species which are involved in the process. The global reaction equation of thermal decomposition of (CH₃)₂CO is

Intermediate species like CO, formaldehyde or CH will not be considered. Due to the low concentration of (CH₃)₂CO in LIF-calibration measurements of few percent, their concentration is expected to be in the range of few hundred ppm. The Raman shifts and the resulting wavelengths of all major species are summarized in Table 1 .

Table 1. Raman Transitions and Signal Wavelength of Different Molecular Transitions-Relevant for (CH₃)₂CO Decomposition

View Table

The detection system was chosen so that the Raman transitions of all molecules of interest can be monitored. Therewith the stability can be analyzed by Raman signals of the educts (CH₃)₂CO and oxygen or by Raman signals of the products H₂O and CO₂. Using the Raman transitions of the educts to determine their decomposition is rather complex as their initial concentrations must be known. This requires quantitative measurements at the inlet of the mixing duct. In addition, spectral overlap of Raman signals of intermediate products like formaldehyde or CH radicals with the Raman signal of acetone has to be considered. It seems easier to use the information of the products CO₂ and H₂O. Here a qualitative evaluation is sufficient, because the occurrence of these molecules indicate the decomposition process. In case of CO₂, this is rather complicated as the Raman transitions at 1285 cm⁻¹ and 1428 cm⁻¹ are overlapping with the Raman transitions of (CH₃)₂CO at 1220 cm⁻¹ and 1428 cm⁻¹. However, the situation is different for H₂O as its Raman lines are clearly separated from all other molecules. Additionally, the scattering cross section of H₂O is significantly higher than the scattering cross section of CO₂, which makes the technique much more sensitive for H₂O.

3. Experimental setup

The experimental setup including the calibration cell is shown in Fig. 1 . A diode pumped, frequency doubled, continuous wave Nd:YVO₄ laser with a maximum power of 5 W at 532 nm is used for the experiments. The laser beam is focused by a lens with focal length of 100 mm in the probe volume located inside the calibration cell. The Raman signal is collected perpendicularly to the incident laser beam by a convex lens with focal length of 100 mm, separated from elastically scattered light by a colored glass filter with cut-off wavelength of 540 nm. Afterwards it is focused by a lens with the focal length of 100 mm to a glass fiber, which is connected to the spectrometer with an implemented air cooled CCD-chip. The resolution of the spectrometer enables the detection of Raman transitions from 300 cm⁻¹ to 4300 cm⁻¹ with a spectral resolution of 7 cm⁻¹. Different flows of air and liquid (CH₃)₂CO are prepared by using high precision mass flow controllers (MFC) with a relative deviation of less than 1%. (CH₃)₂CO is evaporated and added to the air flow, which is preheated to 333 K to avoid condensation. The mixtures are heated and pass continuously through a gas cell that is heated additionally by two heating jackets. Due to the small cross section of the pipe inside the heater the flow is turbulent; inside the cell it is laminar. Flow velocities do not exceed 5 cm/s with an inner diameter of the cell of 2.5 cm. The chamber is equipped with four fused silica windows. Hence, the cell is designed for simultaneous LIF- and Raman measurements. Gas temperature inside the chamber is measured with a resistance thermometer of Pt-100 type. LIF calibration measurements of earlier works show that the temperature profile inside the cell is homogeneous [9]. The pressure is controlled by a pressure transducer in combination with a throttle valve at the outlet of the chamber. The continuous flow has two advantages compared to a discontinuous filling of the cell which is proposed in the literature. First, the residence time can easily be varied by changing the volume flow and comparatively small residence times can be achieved. It is calculated by the division of the volume in the heated area, which is the volume of the gas inside the heater plus the volume of half of the cell (as the measurement is conducted in the middle of the cell), by the volume flow rate. The minimum residence time depending on pressure is 1 s for 10 bar, 2 s for 20 bar and 3 s for 30 bar. The second advantage is that steady state conditions are realized in the cell. Consequently, the exposure time of the detector can be chosen independently and is not limited by the residence time in the cell.

 

Fig. 1 Optical setup.

Download Full Size | PDF

The measurements were performed at different operating points referring to engine relevant conditions. Raman spectra were taken two times with an exposure time of 60 s. First measurements are presented demonstrating the influence of pressure and temperature on the stability of (CH₃)₂CO.

4. Results and Discussion

In Fig. 2 the influence of temperature is demonstrated. Figure 2a shows a spectrum taken at a pressure of 30 bar, a temperature of 598 K, a residence time of 5 s and with an absolute (CH₃)₂CO volume concentration of 1.9%. No decomposition occurred as only the Raman signal of the educts N₂ (${\tilde{\nu }}_R$ = 2331 cm⁻¹), O₂ (${\tilde{\nu }}_R$ = 1556 cm⁻¹) and (CH₃)₂CO (.. = 2925 cm⁻¹) are clearly visible, even though the chamber temperature seems to be well above the ignition temperature (at 11 bar: 498 K). Probably, the residence time in the heated cell volume is still uncritical. In contrast to that, a Raman signal of H₂O (${\tilde{\nu }}_R$ = 3658 cm⁻¹) can be clearly recognized at a temperature of 748 K (see Fig. 2b) and the CH-Peak of (CH₃)₂CO vanishes which is a sign for complete fuel conversion. This indicates a too long residence time in this case.

 

Fig. 2 Raman spectra taken at a pressure of 30 bar with an initial (CH₃)₂CO concentration of 1.9% and a temperature of 598 K (a) or 748 K (b).

Download Full Size | PDF

The pressure sensitivity of the decomposition process is demonstrated in Fig. 3 . After a temperature drop of 50 K, the tracer is stable against thermal decomposition at 10 bar and 698 K (see Fig. 3a) but instable at 30 bar (see Fig. 3b). Beside the pressure, also the (CH₃)₂CO concentration is changed due to the limited adjustable volume flow of the fuel MFC. However, this fact can be neglected as the influence of concentration on ignition temperature is rather weak [22].

 

Fig. 3 Raman spectra taken at a temperature of 698 K and a pressure of 10 bar (a) or 30 bar (b).

Download Full Size | PDF

In order to compare the Raman technique with conventional techniques, like e.g. the detection of heat release during auto-ignition, for all measurements the temperature was recorded simultaneously. This was done by using a Pt-100 resistance thermometer positioned inside the cell. But although (CH₃)₂CO decomposition could be observed within the Raman spectrum in some cases, no temperature increase was recognized. This is probably due to the comparatively small amount of (CH₃)₂CO. Hence, pure temperature measurements are not appropriate for accurate indication of tracer decomposition.

Also the detection limit of H₂O and (CH₃)₂CO is of interest. The (CH₃)₂CO concentration in the given results were already at the lower limit typical for LIF measurements [9] for combustion applications. Therefore, the results presented here could be used to estimate the detection limit of this Raman setup for (CH₃)₂CO and H₂O detection. It is set because of the scattered light background in the LIF-calibration cell, which is not optimized for Raman spectroscopy; an improvement of the signal by increasing the detection time could not be achieved, as above 60 s recording time saturation of the CCD-Chip occurred.

As detection limit, a minimum signal-to-noise ratio of 3 is assumed. This is determined by the ratio of signal intensity to half of the noise amplitude taken from the background noise between the Raman signals. The detection limit has to be regarded separately for every operation point, as the Raman signal intensity is dependent on the temperature and pressure conditions. At a temperature of 598 K and a pressure of 30 bar (see Fig. 2a) a signal-to-noise ratio of 8 is achieved with an (CH₃)₂CO concentration of 1.9%. Hence, the detection limit of (CH₃)₂CO is 0.7%. For the operating point shown at Fig. 3a (698 K, 10 bar), a signal-to-noise ratio of (CH₃)₂CO of 3 is achieved for a (CH₃)₂CO fraction of 2%, which is then also the detection limit.

The detection limit of H₂O is determined from Raman spectra at operation points with complete conversion like shown in Fig. 2b and Fig. 3b. The concentration of H₂O is calculated from Eq. (3) (taking also into account excess non-reacting O₂ and inert N₂) with the initial concentrations of (CH₃)₂CO. This leads to a H₂O concentration of 5.5% for an initial (CH₃)₂CO concentration of 1.9% (see Fig. 2b) and a H₂O concentration of 3.9% for an initial (CH₃)₂CO concentration of 1.3%, respectively (see Fig. 3b). Then, the detection limit is 2.8% for the conditions shown in Fig. 2b (748 K and 30 bar) and 2% for the conditions shown in Fig. 3b (698 K and 30 bar). However, due to the rather weak signal-to-noise ratios between 3 and 6, this is only a rough estimation with absolute deviations up to 0.8% depending on the achieved signal intensities.

5. Conclusion

A novel approach based on Raman spectroscopy is presented for the in situ determination of (CH₃)₂CO decomposition for calibration of LIF-concentration measurements in a flow cell. With this method, the occurrence of decomposition can be analyzed directly inside the LIF-calibration cell at conditions relevant for the mixture formation in an internal combustion engine by simultaneous detection of O₂, N₂, (CH₃)₂CO and H₂O. With the use of this flow cell, temperatures up to 748 K and pressures up to 30 bar can be investigated. Because of a continuous flow, acetone residence times in the heated flow can be varied easily by adjusting different flow conditions.

Although Raman spectroscopy is characteristic for rather weak signal intensities, we have demonstrated that this is a very systematic approach to detect (CH₃)₂CO decomposition. This method is even more sensitive than conventional temperature measurements. Additionally, measurements are possible directly during LIF-calibration. Consequently, the stability of (CH₃)₂CO against decomposition can be controlled during the calibration. This method can also be adapted to other fluorescence tracers like 3-pentanone and toluene without any change in the optical setup, due to the nature of the Raman process.

First measurements at different operating points showed the influence of temperature and pressure. These results offer now the possibility for further systematic measurements with focus on residence time and pressure on the decomposition of (CH₃)₂CO and other fluorescence tracers.