1. Introduction

Thermographic particle image velocimetry (thermographic PIV) is a laser-based technique for simultaneous temperature and velocity imaging in fluid flows, permitting the fundamental and applied study of turbulence, convective heat transfer, and chemically reacting flows. It is based on thermographic phosphors, which are solid materials with luminescence properties that can be used for remote temperature sensing. For this technique, particles made of a suitable thermographic phosphor material are seeded into the flow of interest. Visible laser light scattered by the particles is imaged to determine the particle velocity using an ordinary PIV approach. By exciting the particles using a UV laser and detecting their temperature-dependent luminescence emission, the particle temperature can be determined. Micrometre-size particles are used, which rapidly assume the temperature and velocity of the surrounding fluid [1]. This approach offers the possibility of simultaneous temperature and velocity imaging using relatively simple instrumentation and a single, chemically inert, durable tracer. At low repetition rates (~1-10 Hz) the technique has been used for average [2] and single shot [1] measurements of temperature and velocity in turbulent heated jets, and motored IC engines [3,4]. The technique has also been demonstrated at kHz repetition rates, for time-resolved measurements in a turbulent heated jet, and in the flow behind a heated cylinder [5].

The primary aim of this work is to improve the precision of the temperature measurement. While a high precision is almost always desirable, this is particularly important for applications involving subtle temperature differences. Examples include natural convection, where the flow is driven by buoyancy forces induced by small temperature differences; heating and ventilation in buildings and vehicles; flows in compression machines; and the cooling of electronic components. These examples involve the investigation of turbulent heat transfer at moderate temperatures (e.g. in the range 300-500 K), demanding precise temperature and velocity field measurements.

For temperature measurements using thermographic PIV, the majority of applications have used the intensity ratio method [1,2,4–11], where the temperature-dependence of the luminescence spectrum is exploited by simultaneously imaging the emission in two spectral regions. This is achieved by using bandpass interference filters, chosen so that the ratio of the two filtered signals is a function of temperature. The two simultaneously-acquired, spectrally-filtered images are divided to produce an intensity ratio image, and converted to temperature using calibration data. This approach is relatively straightforward to apply to planar measurements covering a wide temperature range, using a single exposure that permits the short integration times required for measurements in turbulent flows.

In principle, the intrinsic precision of ratio-based thermometry using phosphors is a function of the signal intensity, which determines the precision in the intensity ratio field; and the sensitivity, defined as the change in intensity ratio with temperature. Signal levels can be improved by using a different laser excitation scheme, employing different phosphor particles with a high absorption cross-section and quantum efficiency, or improving the luminescence collection. For example, previous work has sought to optimise some of these parameters, using a bright phosphor with a short luminescence lifetime (BAM:Eu²⁺) [1,5,9,10], and large-aperture lenses with high-transmission filters and a dichroic beamsplitter [5].

For the same experimental parameters that govern the signal-to-noise ratio (and the intensity ratio precision), the statistical temperature error is inversely proportional to the sensitivity. For a given phosphor, the sensitivity can to some extent be controlled by choosing appropriate filters, but there is a fundamental compromise between sensitivity and signal. Ultimately, finding an equally bright phosphor with a more pronounced temperature-dependence of the emission spectrum is the most effective way to improve the technique precision.

Toward this aim, phosphors reported in the literature used for sensitive ratio-based surface temperature measurements were considered. Suitable candidates are zinc oxide phosphors containing excess zinc (ZnO:Zn) and gallium (ZnO:Ga). The strong redshift of their luminescence with temperature was previously investigated for temperature measurements [12,13]. ZnO:Ga was used to probe the temperature of burning methanol droplets [12]. ZnO:Zn was applied to surfaces to assess the temperature uniformity of a heat flux burner, where temperature differences as low as 60 mK could be measured [14]. These materials are therefore potential tracers for sensitive ratio-based thermometry in fluid flows.

ZnO is a direct band gap semiconductor (3.37 eV) [15], emitting luminescence in two bands: a fast (< 1 ns) UV emission and a slower (1 µs), broad emission in the green spectral region. At room temperature, the UV “edge” luminescence occurring in the near-visible spectral region at ~385 nm is due to excitons [15,16]. The luminescence redshifts with temperature due to the decrease in the band gap width [15]. Pure ZnO emits UV luminescence [16] but the spectral position, linewidth, emission intensity and lifetime depend on the type and concentration of naturally present or deliberately added (“doped”) impurities, which are typically on the ppm level [15]. For example, the lifetime of ZnO is ~700 ps but the addition of e.g. gallium or iron can reduce the lifetime by around one order of magnitude [15].

The origin of the green “defect” luminescence at around 510 nm has been attributed to localised composition variations including zinc vacancies, oxygen vacancies, and the presence of copper ions [15]. Preparation or heat treatment of ZnO in low oxygen concentration environments (for example, to promote oxygen vacancies and therefore excess zinc, as in ZnO:Zn), increases the green emission intensity [17,18]. As the excitation irradiance increases, the defect luminescence emission intensity decreases relative to the edge luminescence, so that for high excitation irradiance using for example pulsed lasers, the green emission is relatively weak [17]. The previous studies employing doped ZnO used only the redshifted edge luminescence for sensitive thermometry.

Therefore, since pure ZnO and that containing various deliberately added impurities emit UV luminescence, and noting that most commercially available ZnO contains some level of trace impurities, ZnO which is not deliberately doped could be considered as a material useful for thermometry.

Accordingly, in section 2 of this article, spectroscopic investigations of ZnO particles containing no deliberately added dopants reveal a luminescence emission with a strongly temperature-dependent emission spectrum. In [19], Fond et al. reason the need for the characterisation of phosphor particles in the gas phase, where the nature of the particle-laser interaction matches that of the intended application. Another important conclusion of that work was that new phosphors must be characterised for a known particle number density, to provide a meaningful comparison of the performance of different phosphors for practical applications. BAM:Eu²⁺ is a thermographic phosphor currently employed for thermometry in gas flows [1,5,9,10] that has been extensively characterised for flow thermometry [20]. Therefore, in order to directly assess the benefits of using ZnO particles for flow temperature measurements, in section 3 the signal level and temperature precision using ZnO are directly compared to that achievable using BAM:Eu. These are evaluated for a known particle number density, which is measured using a particle counting system [19]. The luminescence and intensity ratio response of ZnO to temperature and laser fluence are measured, and the results are interpreted in the context of ratio-based temperature imaging. A dependence of the intensity ratio on the laser fluence is observed, but a simple corrective method is proposed and implemented in section 4. Precise temperature imaging using ZnO particles is demonstrated in an electrically-heated airflow, and sources of error are analysed in detail. In section 5, additional measurements are performed to examine the origin of the intensity ratio dependence on excitation fluence and irradiance.

2. Spectroscopy

2.1 ZnO phosphor particles

In this work, ZnO particles (96479, Sigma-Aldrich) were investigated. ZnO powder is used in a wide range of applications such as the cosmetics (note that ZnO is not considered harmful to humans) and materials processing industries, so it is relatively inexpensive compared to deliberately-doped ZnO. The provided assay guarantees a chemical purity >99%, listing over a dozen trace elements present in the batch. No specific size distribution was available for this powder, and so samples were imaged using a scanning electron microscope. The images shown in Fig. 1 indicate that the primary particle size is 200 nm, though the particles appear to form agglomerates with a projected size of approximately 1-2 µm.

 

Fig. 1 SEM images of ZnO powder (96479, Sigma-Aldrich).

Download Full Size | PDF

A numerical heat conduction model was used to determine the temperature response time of these particles in air [1]. The 95% relaxation time of primary 200 nm ZnO particles is just over 1 µs at 293 K. For larger ZnO agglomerates seeded into the flow with a projected size of 2 µm, the response time is 100 µs. These results are based on a spherical particle, but the SEM images show that the ratio of surface area to volume of the actual particles is much larger. Therefore this 100 µs response time is a significant overestimate.

In addition, the luminescence emission spectra of two different production lots of the same ZnO particles (96479, Sigma-Aldrich) and also ZnO:Zn particles (GK30/UF-F2, Phosphor Technology) were measured for comparison in section 2.3.

2.2 Experimental setup

To obtain luminescence spectra of ZnO at different temperatures, a sample of the powder was contained in a ceramic crucible and placed in an optically accessible temperature-controlled oven (Lenton Furnaces). The temperature of the phosphor was measured using an N-type thermocouple positioned in the powder. Samples were excited at 355 nm using the third harmonic of a 5 Hz pulsed Nd:YAG laser (Quanta-Ray GCR-150, Spectra-Physics), with a nominal pulse duration of 10 ns. The laser energy was measured using a pyroelectric energy detector (PEM 45K, Radiant Dyes).

Luminescence was collected using an f/4 lens and spectrally dispersed using a 300 mm focal length f/4 spectrometer (Acton SP-2300i, Princeton Instruments) with a grating groove density of 300 g/mm. Spectra were recorded using an interline transfer CCD camera (Imager Intense, LaVision) with an exposure time of 5 µs. The entrance slit width was 100 µm, providing a spectral resolution of 1 nm, as measured using a mercury vapour lamp. The transmittance of the complete detection system was calibrated using the reference spectrum of a tungsten halogen lamp (LS-1, Ocean Optics).

2.3 Results

The left plot in Fig. 2 shows the normalised luminescence emission spectra of the different ZnO samples. The doped ZnO:Zn and both production lots of the ZnO particles emit UV luminescence at ~390 nm at room temperature, but there are clear differences between the doped and undoped phosphors. Though the width of the emission band is similar, the sample with excess zinc shows a pronounced heel on the UV side of the spectrum and the peak is shifted 2.5 nm toward the visible. Variation between different production lots of the same ZnO is barely detectable, despite the low tolerance on impurities. Within the repeatability of the measurement (50%), the absolute signal levels of all samples were very similar (not shown).

 

Fig. 2 Left: Normalised spectra of two different production lots of the same ZnO (1) (2), and ZnO:Zn, recorded at 296 K using a laser fluence of 2.5 mJ/cm². Right: Normalised ZnO spectra recorded using a fluence of 5 mJ/cm². The temperature interval between curves is 15 K. The transmission profiles (provided by the manufacturers) of the filters used in the gas-phase characterisation study are shown in colour: blue: 387-11 nm (notation CWL-FWHM) and red: 425-50 nm. These transmission profiles have been convoluted with the CCD quantum efficiency and reflection/transmission ratio of the beamsplitter, and normalised to the peak transmission of the 425-50 channel. The corrected profiles are used in section 3 to evaluate the gas-phase data.

Download Full Size | PDF

Normalised luminescence spectra of ZnO at temperatures between 293 and 488 K are shown in the right plot in Fig. 2. Following the decrease in the band gap width, the luminescence emission features a pronounced redshift with increasing temperature, indicating good prospects for sensitive ratio-based thermometry. Broadening of the emission line is also evident. The absolute change of the emission peak position for a given temperature difference is approximately constant with increasing temperature (~0.09 nm/K), which demonstrates a high level of temperature sensitivity across the tested temperature range.

3. Gas-phase characterisation

3.1 Experimental setup

A 21 mm diameter jet of air was used for the phosphor characterisation, which was seeded with particles using a seeder containing a magnetic stirrer to continuously agitate the particles within. The seeded gas stream could be electrically heated up to 500 K using an inline heater (AHP-7562, Omega Engineering), controlled using a K-type thermocouple placed at the jet exit.

For this technique, the luminescence signal is proportional to the particle number density and therefore it is essential to evaluate and compare the performance of different phosphors for a known particle number density. However, in practice it is not possible to directly control this, and so for these experiments the particle number density was directly measured using a recently-developed particle counting system [19]. Only a brief overview is given here. Particles were illuminated using a frequency-doubled Nd:YAG laser, the beam of which was formed into a 160 µm sheet in the measurement plane as shown in Fig. 3. High-resolution Mie scattering images, recorded using an interline-transfer CCD camera, were processed using a program that determines the number of local maxima in each image, serving as a measure of the particle number density.

 

Fig. 3 Setup for phosphor characterisation and temperature imaging, including the particle counting system.

Download Full Size | PDF

Phosphor particles were excited at 355 nm using the same frequency-tripled pulsed Nd:YAG laser used for the spectrally resolved measurements. The 355 nm beam and the 532 nm beam for the particle counting were overlapped in the measurement plane using a long-pass dichroic mirror. The 355 nm sheet thickness (600 µm) was set independently of the particle counting system by controlling the beam size before the sheet forming optics (f = + 500 mm cylindrical lens) using a telescope.

The luminescence emission from the particles was detected by two 4 x 4 hardware-binned interline transfer CCD cameras (Imager Intense, LaVision) with 50 mm f/1.4 Nikon lenses. Two high transmission (>90%) interference filters at 387-11 (84094, Edmund Optics) and 425-50 nm (86961, Edmund Optics) were chosen to filter the luminescence emission, the transmission curves of which are shown in Fig. 2. A 50:50 plate beamsplitter (46642, Edmund Optics), anti-reflection-coated on the reverse side, was used to separate the two detection channels. The camera exposure time was set to 5 µs, beginning 1 µs before the laser pulse. The field of view of the cameras was 28 x 21 mm.

The camera/beamsplitter system was manually adjusted to overlap the two images, and no software mapping was used. Using this procedure, the mean residual displacement between the two recorded images was 0.15 binned pixels, as determined by cross-correlation. Background images were subtracted before applying a cutoff filter at 15 counts and a 7 x 7 moving average filter, for a final in-plane resolution of 600 µm, measured using equivalently processed images of resolution target. Further processing used to correct and extract information from the images is described where appropriate.

The particle counting and luminescence imaging systems, operating at a repetition rate of 5 Hz, were synchronised. Therefore, each instantaneously acquired pair of luminescence images has an accompanying measure of the instantaneous particle number density in the probe volume.

To accurately size the probe volumes in order to determine the particle number density and the excitation fluence, the laser sheet profiles were measured using a sheet profiling system also described in [19]. Briefly, the laser sheets were reflected onto a sheet of white paper using a glass window, and the paper fluorescence was imaged using a CCD camera as shown in Fig. 3. A photodiode-based energy-monitoring unit (Energy monitor V9, LaVision) was used to measure the energy of the UV laser on a shot-to-shot basis. This system was calibrated using a pyroelectric energy detector (PEM 45K, Radiant Dyes).

3.2 Signal comparison and temperature measurement precision

In this section, measurements of the luminescence intensity, and intensity ratio and temperature precision of ZnO particles are presented. The filtered luminescence signal of seeded particles was evaluated from the luminescence images. Then, the emission spectra and the corrected filter transmission curves (Fig. 2) were used to calculate the total unfiltered luminescence signal from the filtered signal. Simultaneously, the particle number density was measured.

The results shown in Fig. 4 indicate that, as expected, the luminescence signal is a linear function of the particle number density. It should be noted that these particle number densities are in the same range as that required for PIV measurements with a similar spatial resolution, and also that this level of seeding has a negligible effect on the gas heat capacity and thermal conductivity [19].

 

Fig. 4 Unfiltered luminescence signal of seeded ZnO and BAM:Eu particles recorded using an excitation fluence of 50 mJ/cm² at 296 K. Each datapoint represents the average intensity of a single sampled image.

Download Full Size | PDF

To identify the relative benefit of using the new ZnO phosphor, the results were compared to measurements of BAM:Eu²⁺ particles (KEMK63/UF-P2, Phosphor Technology), with a median diameter (based on volume) of 2.4 µm. In a separate experiment using a different set of filters, these BAM:Eu particles were seeded into the flow, and luminescence images were recorded and corrected (using luminescence spectra from [5]) as described above to obtain the unfiltered luminescence signal. These results are also plotted in Fig. 4, indicating that the total emission intensity per particle of BAM:Eu and ZnO is very similar.

Both the particle shape and size distribution, as well as the absorption coefficient and quantum efficiency may be very different for the two investigated materials. Therefore, that equivalent signals were observed is likely to the result of a combination of effects. However, it means that for practical measurement purposes, equivalent signal levels can be achieved with each phosphor, for similar experimental conditions.

The spectroscopic measurements already indicate that ZnO will have an intensity ratio response to temperature. This was investigated in the heated jet over a range of temperatures between 295 and 473 K. Pairs of recorded luminescence images were processed as detailed above, divided to produce intensity ratio images, and corrected for differences in collection using a time-averaged intensity ratio field compiled from the same recording sequence. For each jet temperature, average intensity ratios were calculated to produce the ZnO calibration curve shown in Fig. 5. For BAM:Eu, intensity ratios were determined from digitally-integrated luminescence spectra from [5].

 

Fig. 5 Temperature calibration curves and sensitivity for ZnO and BAM:Eu. Calibration data for ZnO is derived from flow measurements in the heated gas; for BAM:Eu, intensity ratios were extracted from digitally-integrated luminescence spectra (from [5]). A fit (of the form a + bI_R^c, where a, b and c are constants and I_R is the intensity ratio) to the datapoints is shown. The evaluated sensitivity (%/K), based on the fit of intensity ratio to temperature, is plotted in dashed lines and can be read on the right axis.

Download Full Size | PDF

The sensitivity curves indicate that at room temperature, the change in intensity ratio with temperature is approximately three times larger for ZnO. The sensitivity increases with temperature, so that at 500 K the sensitivity is more than ten times that of BAM:Eu.

For each phosphor, the normalised standard deviation of the intensity ratio was calculated from the single shot intensity ratio images. The left plot of Fig. 6 shows intensity ratio precision against particle number density for ZnO and BAM:Eu. The signal is proportional to the number of particles and so the precision improves with particle number density. Since the emission per particle is similar for ZnO and BAM:Eu, for similar experimental conditions using similar filtering schemes the single-pixel single-shot precision of the normalised intensity ratio is the same.

 

Fig. 6 Left: Normalised single shot intensity ratio standard deviation against particle number density for ZnO and BAM:Eu for an excitation fluence of 50 mJ/cm² at 296 K. Each datapoint is the intensity ratio standard deviation of a single sampled image. Right: Temperature precision. Results were evaluated from data on the left using the calibration curves for each phosphor from Fig. 5.

Download Full Size | PDF

Using the calibration curves in Fig. 5, the intensity ratio precision data can be evaluated in terms of temperature precision, as shown in the right plot in Fig. 6. Because the sensitivity of ZnO is higher, the temperature precision is significantly improved. Using these filter combinations, at room temperature a threefold improvement in temperature precision can be achieved for the same particle number density, laser fluence, luminescence collection efficiency and final spatial resolution. This comparison clearly shows the benefit of using a phosphor with increased sensitivity.

3.3 Temperature quenching

The flow was heated to investigate the influence of temperature on the luminescence intensity. The particle number density was also simultaneously measured to determine signal levels on a per particle basis. The effect of temperature on the unfiltered luminescence intensity of ZnO particles is shown in Fig. 7. At 473 K the luminescence signal drops to 25% of the signal at room temperature. However, using this filter combination the spectral detection efficiency increases by 50% over this temperature range and so reasonable signal levels can still be achieved.

 

Fig. 7 Normalised unfiltered luminescence signal per particle with increasing temperature. The error bar corresponds to the standard deviation of five repeated measurement sequences at 295 K.

Download Full Size | PDF

3.4 Influence of the laser fluence

The influence of the laser fluence on the signal level was investigated by varying the laser energy during recording sequences, while measuring the luminescence intensity and the pulse energy. The particle number density was simultaneously measured to correct the data for seeding fluctuations. The results are shown in the left plot in Fig. 8.

 

Fig. 8 Left: Evolution of luminescence signal (normalised to the signal in the saturated regime) with laser fluence, at 296 K. Each datapoint represents the average intensity of a single sampled image. Right: Dependence of the intensity ratio on the laser fluence, at 296 K. Each datapoint is the average intensity ratio of a single sampled image.

Download Full Size | PDF

Below 5 mJ/cm² the signal increases linearly with the laser fluence. Above this value the rate of increase of luminescence emission with laser fluence decreases until the signal becomes constant. This is similar to the previously reported saturation behaviour of BAM:Eu [1,19], which was analysed in detail in [20].

The influence of the excitation fluence on the measured intensity ratio was also investigated by simultaneously recording the laser energy while acquiring intensity ratio images. As shown in the right plot of Fig. 8, from the lowest fluence investigated (3 mJ/cm²) the ratio continuously increases with increasing fluence. Beyond 60 mJ/cm² the ratio becomes independent of the fluence. It should be noted that these measurements were not corrected for particle number density, since separate measurements verified that the intensity ratio was independent of particle number density in a tested range up to 6 x 10¹¹ particles/m³. Later, in section 5 of this article, reasons for the dependence of the intensity ratio on laser fluence are explored using additional measurements.

The fact that the intensity ratio depends on both the fluence and the gas temperature is clearly critical when using ZnO as a tracer for temperature measurements, because temporal and spatial variations in laser fluence must be considered, and where appropriate corrected for. Strategies to correct for these are presented in the next section.

4. Temperature imaging experiments

4.1 Laser sheet profile correction

The foregoing measurements identified that the intensity ratio depends on the laser fluence, but in imaging experiments, the laser light sheet may not be uniform and its distribution depends on the spatial profile of the laser beam and the optics used to form the light sheet. A particle will therefore be illuminated by a fluence that varies locally depending on its position within the light sheet.

Ordinarily, a correction for non-uniform light collection efficiency must be applied, sometimes referred to as a “white-” or “flat-field” correction. A simple method for this is to use an average intensity ratio field obtained in a uniform temperature flow, a standard procedure employed in for example [5,7,8]. It is important to note that this average intensity ratio field will also contain the effect of the laser fluence variation across the field of view. This is illustrated by considering the dependence of the intensity ratio field on the spatial coordinates x and y, the temperature T and fluence F:

Following the assumptions outlined below, the dependencies on these variables can be separated:

The first term describes the white field correction, assuming this is independent of temperature. The second term describes the dependence of the intensity ratio on temperature, and the third the dependence of the intensity ratio on the overall and spatial distribution of the fluence field. The separation of these last two terms assumes that the relative change of the intensity ratio with temperature is independent of the laser fluence, an assumption which is explored below.

Dividing by the room temperature intensity ratio field, following the further assumption that the fluence field does not vary in time, results in the normalised ratio:

which is now solely dependent on temperature. Therefore, the corrected, single shot intensity ratio images can then be converted to temperature using a single calibration curve.

Measurements were performed to explore the validity of the assumption that that the relative change of the intensity ratio with temperature is independent of the laser fluence. The effect of laser fluence was investigated in the heated jet over a range of temperatures between 295 and 473 K. At each stable jet temperature the fluence was varied between 5 and 20 mJ/cm². From recorded curves of intensity ratio against fluence (similar to that shown in the right plot of Fig. 8), temperature calibration data was extracted for specific laser fluences.

Figure 9 shows calibration datapoints for different laser fluences. As the laser fluence increases between 5 and 20 mJ/cm² the absolute intensity ratio increases, as shown in the left plot. If each curve is normalised to the intensity ratio at room temperature, as described in Eq. (3), the curves collapse, as shown in the right plot of Fig. 9. The maximum deviation (7.2 K) between curves measured at different fluences is marked on the plot. This source of error is discussed in more detail in section 4.3

 

Fig. 9 Temperature calibration points at fluences of 5, 10, 15 and 20 mJ/cm², through which curves have been fitted to guide the eye. Left: absolute intensity ratio. Right: data normalised at 295 K.

Download Full Size | PDF

This indicates that the change in the normalised intensity ratio with temperature is the same irrespective of laser fluence. The implication of this is that a single calibration curve can be used to convert the intensity ratio images to temperature, regardless of the laser fluence. This greatly simplifies the technique: were this not the case, then the absolute laser fluence would need to be known at every point within the measurement plane.

These results show that within these assumptions, a simple correction method can be used. This involves dividing by an average room temperature intensity ratio field, which contains the effect of both non-uniform light collection and the laser fluence field (α(x,y)χ(F(x,y))). The limits of this approach and additional corrections for temporal fluctuations in the laser pulse energy are presented in section 4.3.

4.2 Temperature imaging demonstration

Measurements were performed in the heated jet. The flow velocity was 1.9 m/s, corresponding to a Reynolds number Re = 2000. The central jet was surrounded by an 80 mm diameter coflow (T = 296 K) with an exit velocity of 0.5 m/s, independently seeded with the same phosphor using a reverse cyclone seeder to ensure particles were present everywhere in the measurement volume. Flow temperatures were measured using a K-type thermocouple (nominal accuracy 2.2 K) placed at the jet exit.

The laser light sheet, formed using + 500 mm and −50 mm cylindrical lenses, intersected the central axis of the jet. While operating at higher fluences (>60 mJ/cm²) would remove the dependence of the intensity ratio on laser fluence, this regime cannot be reached with high-speed diode-pumped solid-state (DPSS) UV lasers used for time resolved measurements. These lasers are limited to pulse energies of a few mJ, and therefore fluences of ~20 mJ/cm² for typical planar imaging experiments. The authors have previously demonstrated kHz-rate measurements using BAM:Eu [5]. For the measurements presented here, the particles were excited using a fluence of 5 mJ/cm² to demonstrate the possibility of performing kHz measurements with ZnO. Pulse energies were measured on a shot-to-shot basis using the energy monitor system.

Luminescence images were recorded using the thermometry system. The field of view was 5 mm above the nozzle exit, slightly offset from the jet centreline to visualise the shear layer at one side of the jet. Processing was applied as detailed above, using a 7 x 7 moving average filter to obtain a final resolution of 600 µm, matching the light sheet thickness. The measured intensity ratio field requires correction for the non-uniformity in light collection efficiency, and also the spatial variation in fluence as described above. This was accounted for by recording a set of images at room temperature before heating the jet. An average intensity ratio image compiled from this set was used to correct the data. The recorded laser energy was used to correct each individual image for overall pulse-to-pulse energy fluctuations, using fluence-ratio data extracted from Fig. 9. Corrected intensity ratio fields were converted to temperature using normalised calibration data as shown in Fig. 9.

Figure 10 shows a single shot and average temperature field recorded in the jet stabilised at 363 K. The jet is in the laminar-turbulent transition regime, and the single shot shows large vortical structures generated by shear between the jet and surrounding coflow. The resolution of the measurement is sufficient to resolve cooler regions within these eddies, caused by the mixing of hot gas and cooler coflowing air. The average temperature field shows the effect of these turbulent fluctuations, leading to some broadening of the shear layer.

 

Fig. 10 Temperature fields in a turbulent jet (T = 363 K). Left: single shot. Right: average field compiled from 100 single shot images. Image size has been reduced to 20 mm in the x-direction.

Download Full Size | PDF

Average temperatures and single shot temperature statistics were evaluated from a constant temperature region of the flow comprising of 200 independent measurements in each single shot image. The single-shot, single-pixel precision (one standard deviation) at 296 K was 5.4 K (1.8%). Despite the decrease in signal with increasing temperature due to thermal quenching (Fig. 7) and the decrease in particle number density caused by the reduction in gas density, because the sensitivity of ZnO increases with temperature (Fig. 5), the temperature precision improved to 4.0 K at 363 K (1.1%). At this temperature, signal statistics predict an error of 3.4 K, indicating these results are close to the noise limit. The deviation between temperatures measured using the thermocouple and the measurement technique was 3.0 K. The shot-to-shot standard deviation of the average temperature was 2.7 K.

4.3 Error analysis

In this section the sources of error resulting from the dependence of the intensity ratio on the laser fluence are analysed in detail. These potential errors, primarily systematic in nature i.e. leading to reduced accuracy, are quantified in two ways. First, errors arising from fluence variations that are not accounted for can be assessed, such as drift and fluctuation of the laser sheet profile. Secondly, the possible error (termed here “calibration error”) stemming from the assumption that the change in the normalised ratio with increasing temperature is the same for all fluences (in order to separate the variables as in Eq. (3)) is also addressed.

A particle will be illuminated by a different fluence depending on its depth position (z) within the light sheet. While the particle position in the image plane (x,y) is known, the z-position is not. Therefore dividing by an average ratio image acquired at room temperature does not correct for this. However, at 10¹¹ particles/m³, statistically there are approximately 50 particles in each independent probe volume (600 x 600 x 600 µm) and so this potential contribution to the single-shot, single-pixel precision is considered negligible.

Using the applied correction procedure, the sheet profile must not drift between the recording of room temperature normalisation sets and the actual experiment. The horizontal stripes faintly visible in the average images of Fig. 10 are attributed to such a drift. In the jet core, where the temperature should be uniform (no mixing), the measured temperature variation i.e. the local temperature accuracy along the centerline is ± 4.6 K. However, vertical profiles of the intensity ratio field were compared on a shot-to-shot basis, and fluctuations were within the measurement noise and could not be distinguished. Shot-to-shot beam profiling is therefore not considered necessary using this particular laser.

For the experiments performed here, fluctuations in the laser pulse energy are approximately 10%. Unaccounted for, the predicted error contribution is 6.2 K, which corresponds well to the uncorrected shot-to-shot standard deviation of the average temperature (5.0 K). In these experiments pulse energies were measured on a shot-to-shot basis and after correction this error is reduced to 2.7 K.

The specifications of this laser indicate a 1% pulse-to-pulse stability at full lamp energy. It is anticipated that running at higher lamp energies and reducing the output energy using a half-wave plate and a polarising beamsplitter would provide the same laser fluence but a significant improvement in pulse-to-pulse stability and drift of the beam profile. Frame referencing can be used to remove these artefacts, by using a region or vertical strip of the image at a known temperature (a procedure that is often employed for Rayleigh imaging e.g [21].) to correct images on a shot-to-shot or average basis.

Next, the calibration error must be considered. Figure 9 indicates that the maximum error stemming from the use of a calibration curve recorded at 20 mJ/cm² to process intensity ratio images recorded at 5 mJ/cm² would be 7.2 K. Also, in this experiment the light sheet optics formed a sheet that was diverging across the field of view (see Fig. 3), corresponding to fluence variations of 50 and 5% in the horizontal (beam focusing towards its waist) and vertical (diverging) directions. Using a similar approach to determining the maximum error as illustrated in Fig. 9, this leads to possible calibration errors of 2.2 and 0.2 K respectively. Using a longer focal length lens or forming a uniform beam waist using a different combination of lenses will effectively remove these calibration errors. Alternatively, a beam homogeniser could be used [22].

5. Laser excitation fluence and irradiance effects

In this section some trends observed in the preceding characterisation study (section 3.4) are explored with additional measurements. Two reasons are proposed for the observed increase of the intensity ratio with excitation fluence.

Firstly, the laser could heat the particles. This would originate from non-radiative transitions, where excess energy is dissipated into the material as heat. This could be caused by the difference between the excitation and Stokes-shifted emission frequencies, which corresponds to an energy difference which will contribute to a heating effect; or non-radiative relaxation between the exciton states and valence band.

The calibration data of Fig. 9 at 5 mJ/cm² was used to interpret the evolution of intensity ratio with fluence in Fig. 8 in terms of temperature. Assuming the ratio increase is only due to a temperature increase of the particles, changing the fluence between 5 and 60 mJ/cm² leads to an observed temperature rise of approximately 100 K. Based on the measurements of the signal per particle for ZnO, and using the known camera specifications and collection efficiency, the number of photons emitted by a particle can be calculated following an approach detailed in [20]. Assuming that, based on the SEM images, a probed particle consists of 100 agglomerated primary particles (200 nm diameter), and that the quantum efficiency of ZnO is 10% [23], using the known specific heat capacity and density of ZnO the temperature rise is predicted to be 34 K. This calculation indicates that laser-induced heating due to non-radiative relaxation is, at least in part, a plausible explanation for the observed trend in intensity ratio with fluence.

A second reason for the observed trend in the intensity ratio could be that the excitation irradiance used in these measurements (and in many diagnostic techniques based on pulsed lasers), of the order MW/cm², affect the luminescence emission of ZnO independently of any particle heating effect. This was investigated using a second pulsed laser with a longer pulse duration. Using two lasers with fixed but different pulse durations allows independent variation of the laser excitation fluence and irradiance.

5.1 Experimental setup

For this purpose, in addition to the flashlamp-pumped laser used in the measurements described previously, a DPSS laser (Hawk HP-355-20M, Quantronix) was used. Triggered at 1 kHz, the laser produced 3 W (3 mJ/pulse), measured using a thermoelectric power meter (HP 25S, Radiant Dyes).

To determine the excitation irradiance, the temporal profiles of the laser pulses were measured using a silicon photodiode (DET10A, Thorlabs). The pulse duration of the flashlamp-pumped laser was approximately 10 ns (FWHM). For the DPSS laser, the pulse energy was adjusted almost independently of the pulse duration by attenuating the radio frequency signal which controls the laser Q-switch. For the fixed drive current used in all measurements the pulse duration was approximately 170 ns.

Slight modifications to the laser sheet formation and camera settings were required. First, to achieve comparable fluences using the two laser systems, different cylindrical lenses were also used to alter the vertical height of the laser sheets. The sheet heights and thicknesses were monitored using the beam profiling system described in section 3.1. Second, the interline-transfer CCD cameras employed here are still sensitive during readout of the nominal exposure. Although this sensitivity is reduced by the extinction ratio (2000:1), the readout time at full frame is long (100 ms). The minimum repetition rate of the DPSS laser was 1 kHz, and therefore the 100 additional laser pulses occurring during the frame readout would lead to a significant background signal contribution. Increasing the hardware binning to 8 x 8 for measurements using the DPSS laser sufficiently reduced the readout time to eliminate this contribution. Consistency between intensity ratios measured with different hardware binning was verified.

5.2 Results

Measurements made with each laser system are shown in Fig. 11. For the same fluence, the difference in the measured intensity ratio using the two systems can be attributed to the difference in excitation irradiance (due to the different pulse duration). This is consistent with an increasing redshift of the emission with excitation irradiance, due to a decrease of the band gap width with increasing exciton density in the material [23].

 

Fig. 11 Effect of the laser fluence on the intensity ratio for excitation using pulse durations of 10 ns and 170 ns.

Download Full Size | PDF

Considering a fixed pulse duration, both the excitation fluence and the irradiance increase with the laser energy. Figure 11 also shows that increasing the excitation irradiance for a fixed fluence increases the intensity ratio. From these two observations, it can be deduced that the real temperature increase of the particles caused by laser excitation must be lower than that assumed in the calculation above, because part of the intensity ratio increase can be attributed to an increased excitation irradiance.

6. Conclusions

ZnO particles were characterised as tracers for temperature imaging in fluid flows. Spectroscopic measurements of ZnO particles with no deliberately added dopants show that there is a strong redshift of the edge luminescence with temperature that can be used for sensitive thermometry. In gas phase measurements, direct comparison with BAM:Eu phosphor particles indicate similar signal levels per particle and intensity ratio calibration showed significantly improved temperature sensitivity of ZnO. At room temperature a threefold gain in temperature precision can be achieved for the same particle number density, laser fluence, luminescence collection and spatial resolution.

Further measurements explored the dependence of the luminescence emission intensity and intensity ratio on the temperature and laser fluence. Similar saturation behavior to that previously identified for BAM:Eu was observed. A dependence of the intensity ratio on the excitation laser fluence was observed. Laser-induced heating of the particles caused by non-radiative heat dissipation was identified as a plausible explanation for this behavior. Further measurements conducted using a laser with a different pulse duration showed that the excitation irradiance also lead to an increase of the intensity ratio, possibly because the high density of electron-hole pairs leads to a redshift of the emission. Therefore, we consider that both phenomena are responsible for the observed increase in the intensity ratio with fluence.

A scheme to correct for the dependence of the intensity ratio on the laser fluence was demonstrated. The procedure is the same as an ordinary white field correction used for a typical intensity ratio imaging experiment and requires no additional experimental effort. The errors arising from the laser are specific to the beam profile and pulse-to-pulse stability, and can be reduced or eliminated using corrective procedures. Calibration measurements at laser fluences in the range 5 to 20 mJ/cm² indicate a residual calibration error of 7.2 K, but shrinking the range of fluences in experiment and calibration reduces these errors to a negligible level. Following a detailed error analysis, we conclude that the intensity ratio-laser fluence dependence has little impact for practical measurements using ZnO tracer particles.

At room temperature, a precision of 5.4 K was achieved. Temperature images in an airflow heated to 363 K were presented, and because the sensitivity of ZnO increases with temperature the precision increased to 4.0 K (1.1%). At 500 K, using a seeding density of 2 x 10¹¹ particles/m³ and after accounting for thermal quenching of the luminescence emission, a temperature precision of 3.0 K (0.6%) is feasible. This is using a fluence of 5 mJ/cm² (also used in the demonstration experiments) which can be easily achieved using commercial DPSS lasers, so this tracer can be used for sensitive measurements at kHz repetition rates. These particles scatter light efficiently and can readily be used as PIV tracers for simultaneous temperature-velocity imaging.

While doped ZnO (ZnO:Zn and ZnO:Ga) was previously used for ratio-based temperature measurements the possibility of using the undoped material was not recognised. Thus one of the novelties of this work is the use of semiconductors as a tracer. It may possible to increase the emission intensity or increase the shift of the edge luminescence emission with temperature by changing the concentration or type of impurities. Using ZnO and perhaps other semiconductors provides exciting prospects for further improvements in the measurement sensitivity, and is a good example of the broad perspectives for development of the technique based on the near-infinite variety of thermographic phosphors, most of which are unexplored for their use as temperature sensors.

Given the high temperature sensitivity of ZnO particles, this work significantly extends the capabilities of the technique toward the study of flows with small temperature variations, essential for the investigation of natural convection and heat transfer.