Open Access
Add to BibSonomyAbstract
Linear Raman scattering has been applied for the determination of the temperature of the liquid phase in water sprays under normal and superheated conditions. The envelope of the Raman OH-stretching vibration band of water is deconvoluted into five Gaussian peaks which can be assigned to five different intermolecular interactions (hydrogen bonding). The intensity of each of the peaks is a function of the temperature and the phase of the water under investigation. The interference of the Raman signals originating from the water vapor is eliminated from the Raman signals originating from the liquid water. Consequently the temperature of the liquid water droplets surrounded by water vapor is accessible which is favorable for the investigation of non-equilibrium sprays where the droplet temperature is different to the vapor temperature.
© 2014 Optical Society of America
1. Introduction
Superheated sprays are widely used in chemical engineering processes and are therefore under investigation in different research fields [1–9].The superheating promotes evaporation of fluid in the nozzle capillary resulting in a biphasic flow. Bubbles contained in the flow burst at the nozzle outlet and enhance spray disintegration. Thermal energy therefore acts as additional driving force for atomization and allows the use of simple nozzle geometries and low pressures although a comparably fine spray results. Especially the spray atomization is a rather complex process which is influenced by different parameters [10]. Temperature of the fluid matter is one of the most influential parameters in respect to rates of evaporation and conversion, e.g., in polymerization, crystallization, or reactions in general. Therefore the knowledge of the liquid phase temperature at different locations in the spray is essential for the understanding of the processes taking place in sprays.
State of the art for the determination of the temperature in a spray is the application of thermocouples or resistant thermometers. The most obvious drawback of these sensors is their invasive measurement principle, which means that they influence the flow and thus the heat and mass transport. Furthermore they cannot measure the temperature of the liquid or the vapor phase separately. This is of particular disadvantage, if the liquid and the vapor phase temperatures differ, or in the case that the vapor phase is not saturated with the liquid phase. In the latter case, during the measurement the liquid droplets impinging on the invasive temperature sensor wet its surface. Subsequently the evaporation of the liquid layer on the surface cools the sensor leading to the measurement of lower temperatures.
Therefore there is a demand for an optical non-invasive working temperature measurement technique. In earlier work, for example laser-induced fluorescence (LIF) was used for the temperature determination in spray investigations inside the nozzle and directly at the nozzle’s exit [11]. The results indicated a strong dependence of the droplet size on the temperature, in particular in the case of dense sprays and the presence of small droplets. The authors proposed an empirical method in order to correct the effect of the droplet size on the temperature evaluation. Besides, the challenging signal evaluation of LIF suffers from other drawbacks, like the need to add LIF tracer molecules if the spray fluid, like water, does not fluoresce by its own or the superposition of different fluorescent signals in a mixture containing different fluorescent molecules.
To avoid the afore mentioned drawbacks of LIF thermometers, Müller et al. [12] introduced a high precision optical temperature measurement technique which is based on linear Raman scattering. They exploited the temperature sensitive envelope of the OH-stretching vibration Raman band of liquid methanol to extract the temperature of the dispersed liquid phase from the detected Raman spectra. As the amount of vapor surrounding the liquid droplets was small in the spray, the Raman spectrum of the vapor phase did not interfere with the Raman spectra originating from the liquid droplets. Therefore they were able to apply the calibration of the envelope of the OH-stretching vibration Raman band measured as a function of temperature in bulk liquid to evaluate the temperature of the dispersed liquid methanol droplets in a spray. For the here reported temperature measurements in water sprays the working principle of the strategy introduced by Müller et al. [12] cannot be transferred to water sprays as the calibrations resulting from measurements in bulk liquid water cannot be applied to measure the temperature of a dispersed liquid water phase being surrounded by water vapor. Therefore, for the first time to the best of our knowledge, we report here a totally different strategy to determine the temperature of water droplets, which can be used even if the droplets are surrounded by water vapor.
2. Experimental set-up
Figure 1 shows schematically the experimental setup. For spray generation deionized water is pumped into a vessel which is equipped with a stirrer, driven by a motor M and temperature controlled by means of a heating jacket (Wel). The absolute pressure inside the vessel is set to values between 1 and 6 bars and the temperature to values between 298K and 393K. A cylindrical nozzle (l/d = 20; d = 1.8 mm) is used to form a full cone spray of water. Downstream the nozzle the spray is removed by means of a fan. A more precise description of the spray plant can be found in [13]. The Raman scattering excitation and the detection parts are combined in a compact cage system which allows an easy movement of the laser beam focus to different locations in the spray. For the excitation of the Raman signals a frequency doubled continuous wave Nd:YAG Laser with an output power of 250 mW was used. The laser beam has been focused by a spherical lens with a focal length of 200 mm providing a local resolution of ca. 100 µm in the measuring volume. Signal detection is done in a backscattering configuration using an achromatic lens as focusing lens which additionally collects and collimates the generated broadband red shifted Raman signals. The signal detection in backscattering configuration avoids the dependency of the Raman signal intensity from the polarization axis of the laser. In order to get a well-defined measurement volume the laser beam is widened via a Galilean telescope system. The stray light and the elastically scattered light are blocked by a long pass filter and a dichroic mirror with an optical density of 6 and 4, respectively, leading to a total optical density of more than 10. The signal is focused with an achromatic lens (f = 100) into an optical fiber connected to a spectrometer (ocean optics, QE 65000). For each measurement point inside the spray at least 20 single spectra were acquired from which the mean spectrum was computed. The acquisition time of each single spectra was set to values from 500 ms to 4 s depending on the water concentration in the probe volume and thus on the signal level. Following the chosen integration time was a trade-off between signal level and measuring time. Since the amount of heated water in the vessel was rather limited the integration time was set as short as possible to achieve a sufficient signal level for the deconvolution of the OH-band.
3. Results and discussion
For system calibration the described Raman-setup was used to measure the liquid water temperature inside a stirred glass cuvette which was placed on a heating plate collecting Raman signals over the whole relevant temperature range from 293K to 373K. For the application of the measurement strategy in superheated or supercooled water the temperature range should be expanded during the calibration.
In order to measure under different conditions and to test the measurement strategy, the water inside the vessel in Fig. 1 was heated to various temperatures. Special attention was directed to the measurement of the temperature under superheated conditions of the spray (initial water conditions inside the vessel at 393K and 6 bar absolute) to achieve a high concentration of the gaseous water vapor surrounding the water spray droplets. The locations for the temperature determination inside the spray were chosen directly at the nozzle’s exit and 10 mm downstream the nozzle in z- and y-direction (see Fig. 1). At each of these locations the temperature was measured with the laser Raman probe and additionally with a Pt100 resistant thermometer in order to compare the results of both sensors.
Figure 2 shows in the background plane the Raman spectrum of the OH-stretching vibration band taken in the bulk water (calibration measurements) for different temperatures between 293K and 373K. As the spectra are all normalized to their integrated intensity (area under the spectrum envelope), they intersect at the isosbestic point [14], which separates the Raman spectrum of the OH-stretching vibration band into two parts. The part left of the isosbestic point is dominated by hydrogen bonded water molecules, while the part right of the isosbestic point is dominated by free water molecules [15, 16]. The arrows in Fig. 2 indicate the tendency how the signal intensities will change with increasing temperature. The intensity decrease in the Raman OH-band left of the isosbestic point and the intensity increase right of the isosbestic point with increasing temperature is assigned to a weakening of the hydrogen bonding with increasing temperature, which is the result of an increasing molecular motion with temperature. It is known [16], that the OH-stretching vibration band can be deconvoluted into five single peaks, which are represented in Fig. 2 as Gaussian peaks in five foreground planes. The central wavenumbers of the peaks were taken from literature [15]. Their intensity was adjusted to get a match between the Raman spectrum of the OH- stretching vibration band (shown in the background) and the sum of the five single Gaussian peaks. Just the amplitudes of the peaks were varied, since these parameters represent the strengths of the hydrogen bonds. In order to reduce the amount of fit parameters the FWHM was held constant. Again the arrows indicate the tendency how the intensity of the five single Gaussian peaks change with increasing temperature. The central wavenumbers of peaks I and II are located left of the isosbestic point of the Raman OH-stretching vibration band and can therefore be assigned to the hydrogen bonded water molecules. The central wavenumbers of the peaks III, IV and V are located right of the isosbestic point of the OH-stretching vibration band and can therefore be assigned to free water molecules [17]. While the intensities of the peaks I, II and IV decrease with increasing temperature, the intensities of the peaks III and V increase with increasing temperature. In this context it is important to mention that contrary to our findings Crupi et al. [17] reported an intensity increase of peak IV with increasing temperature. If the water OH-stretching vibration band is deconvoluted with a constant FWHM for each of the five Gaussian peaks, the peak IV intensity decreases with temperature. If the band is deconvoluted with variable FWHM, the peak IV intensity increases with increasing temperature. If the investigated spray would contain also solutes which change the shape of the OH-band a separate calibration has to be carried out taking the solute’s influence on the OH-band into account.
Fig. 2 Raman OH-band in the temperature range from 293K to 371K (background) and the five corresponding deconvoluted Gaussian contributions. The plots in the foreground show the difference between the measured and the computed spectra for different temperatures. The inserted arrows indicate the temperature dependency of the different peaks and point towards increasing temperature.
From the temperature sensitive behavior of the OH-stretching vibration band three different strategies can be deduced to extract the temperature of liquid water from the Raman spectrum. The first and most straight forward one is that one, which only considers the ratio of the Raman signal intensities from the left and the right side of the isosbestic point. In general this strategy was applied for example by Vehring et al [18] and Müller et al [12], regarding Müller et al methanol instead of water was used. As mentioned before and explained in more detail below, this strategy cannot be well applied here. The second strategy is shown in Fig. 3(a). Once the Raman intensity normalized OH-stretching vibration band is deconvoluted into five single peaks, the integrated intensities of the five single peaks contain information on the temperature of the liquid water. In general each of the five peaks shows a temperature sensitive behavior. The peaks II and III seem to be most suitable for temperature determination as they are among the most intensive peaks and additionally show high temperature sensitivity. The third strategy is represented in Fig. 3(b). Here the intensity ratio of two temperature sensitive peaks is correlated with temperature. Again peaks II and III are most suitable for the ratio formation as their temperature dependencies are in the opposite direction and as their intensities are similar high. Due to this temperature dependent behavior of the single peak intensities, strategy 3 provides the highest temperature resolution.
Fig. 3 Integrated Raman signal intensity for the five Gaussian peaks as a function of the temperature (a) and the applied calibration line for the temperature determination (b), corresponding to the third strategy.
Table 1 summarizes the polynomial functions representing the intensity temperature relation of the above mentioned temperature evaluation strategies where x is the ratio formed by splitting the OH-band envelope at the isosbestic point for strategy (1), the single peak intensity either for peak II or peak III for strategy (2) and the ratio formed by dividing the intensities of peak II and III for strategy (3) from the spectra detected in the spray.
Table 1. Polynomial Functions of the Different Evaluation Strategies
In the following the applicability of the three different evaluation strategies is discussed in the context of Fig. 4. Figure 4 shows two Raman spectra of the OH-stretching vibration band taken from spray measurements, one for an exemplary cold spray (not superheated spray) and one for a hot spray (superheated spray). The third spectrum corresponds to the calibration measurements carried out in bulk liquid water at 371K. A Raman spectrum of water vapor can be found in [19].
Fig. 4 Spline corrected Raman spectra of the measurement in the water spray and from the calibration; by means of the inserted arrows the spectra measured at hot and cold spray locations and the spectrum from the calibration are distinguishable
At first the origin of the noise, which can only be found in the spectra of the superheated spray, is described. The droplet atomization within the superheated spray is enhanced compared to the non-superheated spray. Therefore the interfacial area between the dispersed liquid water phase and the bulk gas phase composed of air and water vapor is increased. As the laser beam used for the excitation of the Raman scattering is diffracted and scattered from the interfacial area on its way to the probe volume and as the signals from the probe volume to the detector are also diffracted and scattered when they pass the spray, the Raman signal level is small for the superheated spray compared to the non-superheated spray, for which the droplet atomization and thus the interfacial area is less. After intensity normalization of the intensive cold and the weak hot spray spectrum, the noise is visible only in the hot spray spectrum. Since these random peaks can also be found in spectral regions where no Raman signal from water appears, these narrow peaks cannot be due to morphology dependent resonances [20] of the droplets but must be reasoned in the relatively low detected signal level. In Table 2 the deviation between the normalized experimental and the deconvoluted spectra are quantified as the sum of the differences (positive values) for each wavenumber position between 2700 and 3800 cm−1 for the three cases shown in Fig. 4.
Table 2. Overview of Deviations between Normalized Experimental and Deconvoluted Spectra given in Fig. 4
Due to the noise in the spectrum taken from the hot spray, the deviation between the experimental and the deconvoluted spectrum is increased relatively to the spectra taken from the bulk. Nevertheless the least squares fit method the deconvolution strategy is based on balances out the noise and focuses on the main shape of the OH stretching vibration Raman spectrum.
At a first glance, the first evaluation strategy, which considers the ratio of the signal intensities left and right of the isosbestic point, is the most straight forward one with respect to temperature evaluation, as it is the simplest one. Here it must be pointed out, that this strategy is only applicable, if the Raman signal of the OH-stretching vibration band originates from liquid water alone. In the context of the investigations of sprays, this means that the Raman signals may only origin from the dispersed liquid phase from the droplets, and not from the continuous vapor phase. If there is much gaseous water in the vapor phase, the Raman sensor will detect the signals of the OH-stretching vibration band of the liquid and of the vapor phase simultaneously. As in the vapor phase the mean distance between the water molecules is large, they do not form hydrogen bonds, which means that the Raman spectrum of the OH-stretching vibration band of water vapor is only composed of peak V (totally free water molecules). Consequently the interference of the Raman signals originating from the water vapor only contributes to the intensities on the right side of the isosbestic point. Under these circumstances, the calibration results coming from measurements in bulk liquid water cannot be applied to evaluate the temperature of Raman spectra which combine the OH-stretching vibration band of the liquid and the vapor phase all at once. The hot spray spectrum in Fig. 4 is one example, where the OH-stretching vibration band Raman signals of the liquid and the vapor phase interfere. Obviously, the intensity right of the isosbestic point is significantly more intense than for the calibration spectrum in bulk water at 371K. Accordingly taking into account the intensity ratio of intensities from the left and right sides of the isosbestic point would result in a temperature exceeding 373K. However, at a pressure of 1 bar the temperature of liquid water cannot reach such high temperatures, in thermodynamic equilibrium the highest temperature possible is 373K. Summarizing, whenever there is a significant contribution of the water vapor interfering with the OH- stretching vibration band of water, the first evaluation strategy is not applicable as the evaluated temperatures exhibit too high values.
With respect to the second evaluation strategy, which considers the intensities of the single deconvoluted peaks, the interference of the water vapor OH-stretching vibration causes similar artifacts. Before the entire OH-stretching vibration band is deconvoluted, it is normalized to its integrated signal intensity. One should keep in mind that during calibration there was only liquid water in the measurement volume. This means, that the normalization of the OH-stretch vibration spectrum to its intensity means normalization to the signals originating solely from the liquid phase. If the Raman spectrum of the OH-stretching vibration band contains signals from the liquid as well as from the vapor phase –this is the case in the spray measurements-, the normalization step reduces those peaks too much, which origin only from liquid water. Looking at the calibration results of peaks II and III which are summarized in Fig. 3(a), a too small evaluated peak II would result in too high temperatures, while a too small evaluated peak III would result in too small temperatures.
Finally, only the third evaluation strategy, which considers the intensity ratios of peak II and III is not influenced by the interference of the water vapor OH-stretching vibration band. This band only affects the very right wing of the OH-stretching vibration band of liquid water, which is dominated by peak V (absolutely free water molecules not undergoing any hydrogen bonding). Therefore the intensity ratio of the peaks II and III is not affected by this interference and finally is the only evaluation strategy which is applicable also to sprays carrying a significant portion of water vapor in the continuous gas phase. If the amount of water in the continuous gas phase around the dispersed liquid droplet phase is small enough to cause no serious interference with the OH-stretching vibration band Raman signal the evaluation strategies 1 and 2 are also applicable.
Finally Fig. 5 exactly underlines this behavior. In Fig. 5 on the y-axis the deviation of the different temperature evaluation strategies from the third evaluation strategy (ratio peaks II and III) is given. On the x-axis all the evaluated temperatures from the third evaluation strategy are given forming the “zero-line” in the diagram. The cold temperatures on the x-axis indicate that a non-superheated spray has been analyzed. Therefore the amount of vapor in the continuous gas phase surrounding the dispersed liquid droplets is small. Consequently the deviation of the different evaluation strategies is small. With increasing temperature on the x-axis, also the amount of water vapor in the continuous gas phase surrounding the dispersed liquid droplets increases. Therefore also the interference of the OH-stretching vibration band of water vapor with the OH-stretching vibration band of the liquid water dominates more and more the intensities on the right side of the isosbestic point. As described in the context of Fig. 4, the evaluation strategy 2 based on the consideration of the peak II intensity evaluates too high temperatures, while the evaluation strategy 2 based on considering the peak III intensity evaluates too small temperatures. The deviations increase with increasing temperature on the x-axis, as the water vapor content in the gas phase increases, too. Additionally, temperatures evaluated by using a conventional Pt100 thermometer are provided, which are all smaller than the temperature evaluated using the third evaluation strategy. For cold temperatures this can be due to the evaporation of the liquid film surrounding the thermometer. For hot temperatures, for which the vapor phase can be considered to be saturated with water vapor and the evaporation of the water film is not possible anymore. Then the temperature deviation can be explained by a gas and a liquid phase of the spray, which both own different temperatures. And also the entrainment of fresh cold air into the spray can cool the gas phase temperature below the liquid phase temperature. If the continuous gas phase temperature is below the liquid phase temperature, again the Pt100 would indicate too small temperatures.
Fig. 5 Deviation of the temperatures obtained by the different evaluation strategies; the crosses on the x-axis indicate the temperatures obtained by the third strategy in forming the ratio of peak II and peak III.
4. Conclusion
Three different Raman based temperature evaluation strategies have been identified which in principle can be used for temperature determination in water sprays. These strategies have been compared with the result that only one of them is suitable for the determination of the liquid phase temperature of superheated sprays or of sprays for which the amount of water vapor in the continuous gas phase contributes significantly to the Raman spectrum of the OH-stretching vibration band of water. This non-invasive technique can be used to probe temperatures locally resolved at many locations inside a spray. The deconvolution method introduced here reduces the impact of noise on the temperature evaluation strategy which means that measurements are possible also in dense sprays with a large interfacial area between the gas and liquid phases in the spray.
Acknowledgments
The authors gratefully acknowledge financial support for parts of this work by the German Research Foundation (DFG) in the priority program “Process Spray” (SPP 1423). The DFG additionally funds the Erlangen Graduate School in Advanced Optical Technologies (SAOT) within the framework of the Excellence Initiative of the German Federal and State Governments to Promote Science and Research at German Universities.
References and links
1. G. Lamanna, H. Kamoun, B. Weigand, and J. Steelant, “Towards a unified treatment of fully flashing sprays,” Int. J. Multiph. Flow 58, 168–184 (2013).
2. A. Günther and K.-E. Wirth, “Evaporation phenomena in superheated atomization and its impact on the generated spray,” Int. J. Heat Mass Transfer 64, 952–965 (2013). [CrossRef]
3. C. Desnous, A. Cartellier, and N. Meyers, “Experimental investigation of explosive vaporization of C6F14,” C. R. Mec. 341(1–2), 88–99 (2013). [CrossRef]
4. S. Mutair and Y. Ikegami, “Experimental investigation on the characteristics of flash evaporation from superheated water jets for desalination,” Desalination 251(1–3), 103–111 (2010). [CrossRef]
5. S. Mutair and Y. Ikegami, “On the evaporation of superheated water drops formed by flashing of liquid jets,” Int. J. Therm. Sci. 57, 37–44 (2012). [CrossRef]
6. Y. Ra and R. D. Reitz, “A vaporization model for discrete multi-component fuel sprays,” Int. J. Multiph. Flow 35(2), 101–117 (2009). [CrossRef]
7. J. Kim, “Spray cooling heat transfer: the state of the art,” Int. J. Heat Fluid Flow 28(4), 753–767 (2007). [CrossRef]
8. W. Zeng, M. Xu, G. Zhang, Y. Zhang, and D. J. Cleary, “Atomization and vaporization for flash-boiling multi-hole sprays with alcohol fuels,” Fuel 95, 287–297 (2012). [CrossRef]
9. J. Senda, Y. Hojyo, and H. Fujimoto, “Modeling on atomization and vaporization process in flash boiling spray,” JSAE Rev. 15(4), 291–296 (1994). [CrossRef]
10. V. Cleary, P. Bowen, and H. Witlox, “Flashing liquid jets and two-phase droplet dispersion I. Experiments for derivation of droplet atomisation correlations,” J. Hazard. Mater. 142(3), 786–796 (2007). [CrossRef] [PubMed]
11. A. Labergue, A. Delconte, G. Castanet, and F. Lemoine, “Study of the droplet size effect coupled with the laser light scattering in sprays for two-color LIF thermometry measurements,” Exp. Fluids 52(5), 1121–1132 (2012). [CrossRef]
12. T. Müller, G. Grünefeld, and V. Beushausen, “High-precision measurement of the temperature of methanol and ethanol droplets using spontaneous Raman scattering,” Appl. Phys. B 70(1), 155–158 (2000). [CrossRef]
13. A. Günther, M. Rossmeissl, and K. Wirth, “Discharge characteristics of the atomization of superheated liquids.”
14. P. L. Geissler, “Temperature dependence of inhomogeneous broadening: On the meaning of isosbestic points,” J. Am. Chem. Soc. 127(42), 14930–14935 (2005). [CrossRef] [PubMed]
15. D. M. Carey and G. M. Korenowski, “Measurement of the Raman spectrum of liquid water,” J. Chem. Phys. 108(7), 2669–2675 (1998). [CrossRef]
16. R. Li, Z. Jiang, Y. Guan, H. Yang, and B. Liu, “Effects of metal ion on the water structure studied by the Raman OH stretching spectrum,” J. Raman Spectrosc. 40(9), 1200–1204 (2009). [CrossRef]
17. V. Crupi, S. Interdonato, F. Longo, D. Majolino, P. Migliardo, and V. Venuti, “A new insight on the hydrogen bonding structures of nanoconfined water: a Raman study,” J. Raman Spectrosc. 39(2), 244–249 (2008). [CrossRef]
18. R. Vehring and G. Schweiger, “Optical determination of the temperature of transparent microparticles,” Appl. Spectrosc. 46(1), 25–27 (1992). [CrossRef]
19. W. F. Murphy, “The rovibrational Raman spectrum of water vapour v 1 and v 3,” Mol. Phys. 36(3), 727–732 (1978). [CrossRef]
20. R. Symes, R. M. Sayer, and J. P. Reid, “Cavity enhanced droplet spectroscopy: Principles, perspectives and prospects,” Phys. Chem. Chem. Phys. 6(3), 474–487 (2004). [CrossRef]
