Fluid flow, heat transfer, and chemical processes are important in many applications including thermal convection; the heating, cooling, and ventilation of interior spaces; manufacturing; thermal management of microelectronics; and biological systems. Study of these coupled processes requires highly sensitive nonintrusive temperature measurements that reveal the temperature distribution and flow dynamics, usually around key temperatures that are determined by the system of interest, for example, the specific physiology or fluid/material properties. There are many optical techniques for fluid temperature imaging, including Rayleigh/Raman scattering, thermochromic liquid crystals (TLCs), and luminescence thermometry, based on optically active tracers including organic compounds (e.g.,  dyes), metal-ligand complexes, quantum dots, phosphors, etc. (see review [1]). Each method has various advantages and drawbacks in specific applications. Among luminescence thermometries, phosphor thermometry is based on inorganic materials usually doped with transition metals or lanthanides. Phosphor particles are used for fluid thermometry by seeding the particles into the flow as a tracer and probing their temperature-dependent luminescence properties using light sources and cameras [2]. It is a versatile approach because it can be applied in both liquid and gas flows; the temperature range can be adjusted from cryogenic temperatures to over ${\sim}{{700^{\circ} {\rm C}}}$ using different phosphor particles of which there are innumerable types; and the luminescence is generally insensitive to the fluid composition and pressure. However, the imaging measurement precision has been limited by low temperature sensitivities ${\lt}{{1}}\% {{/^{\circ} {\rm C}}}$ (e.g.,  for ZnO [3,4]) or low signal (e.g.,  for ${\rm{L}}{{\rm{a}}_2}{{\rm{O}}_2}{\rm{S}}:{\rm{E}}{{\rm{u}}^{3 +}}$ [5]), resulting in a choice between excessive particle seeding (${\gt}{\text{g}}/{\text{L}}$ [3,6]) or otherwise a single-shot precision restricted to ${{\pm 2 {-} 3^\circ {\rm C}}}$ (${{1}}\sigma$) [4].

The aim of this study is therefore to improve the measurement precision to reach the sub-°C range. One way to do this is to use a different phosphor with an inherently higher temperature sensitivity, particularly by using changes in the luminescence signal or lifetime that are directly driven by highly sensitive thermal quenching processes. Near the onset of thermal quenching, lifetime sensitivities of several %/°C are typical [7]. Also, in comparison to spectral-based methods, lifetime approaches are less sensitive to variations in transmittance profile of surrounding media or optics. To exploit the lifetime for temperature imaging, it is necessary to meet two opposing requirements: the decay has to be resolved in time, but also in a fluid flow the total integration time has to be short, preferably on the microsecond (µs) scale, so the particles do not move during the measurement.

For lifetimes in the nanosecond range, fluorescence lifetime imaging uses high repetition rate or modulated excitation sources and time-correlated photon-counting or phase-modulated detectors [8]. In the microsecond range, rapid lifetime determination can be implemented with low frame rate gated cameras [interline transfer CCDs or global shutter complimentary metal oxide semiconductor (CMOS)] and low repetition rate pulsed excitation. Using various gating schemes [9–11], the ratio of signals from exposures covering different portions of the decay in subsequent pulses yields the lifetime. For lifetime imaging with single excitation pulses, kilohertz (kHz) frame rate CMOS cameras can resolve decay times of 10’s µs and longer, as often employed for surface phosphor thermometry [12]. These approaches require relatively long overall measurement times and are therefore difficult to implement for instantaneous fluid temperature imaging.

The dual-frame capability of interline transfer CCD cameras allows the recording of two frames separated by a sub-microsecond interframe time [see Fig. 1(a)] permitting rapid lifetime determination from two frames straddling a single luminescence decay waveform. This approach has already been used for phosphor thermometry [13,14]. However, for this method to be a workable strategy specifically for instantaneous temperature imaging in fluid flows, a thermographic phosphor with suitable luminescence properties must be identified: the lifetime must be (1) at the appropriate µs scale and (2) temperature-sensitive in the temperature range of interest. Bismuth-doped vanadates are a possible class of phosphors previously investigated by Boulon et al. [15], who showed that ${\rm{ScV}}{{\rm{O}}_4}{:}{\rm{B}}{{\rm{i}}^{3 +}}$ has a lifetime in the µs range, which decreases strongly with temperature around ambient (${\sim}{\rm{20^{\circ} \rm C}}$) conditions. This is our target temperature range for studies of thermal convection or biological processes. Therefore, the objectives of the research described herein are to (1) synthesize ${\rm{ScV}}{{\rm{O}}_4}{:}{\rm{B}}{{\rm{i}}^{3 +}}$ phosphor particles, (2) characterize their luminescence properties, and (3) demonstrate and evaluate a temperature imaging technique using the dual-frame methodology.

 

Fig. 1. (a) Timing of a dual-frame lifetime thermometry strategy; (b) experimental setup for temperature imaging.

Download Full Size | PDF

Phosphors were produced using a conventional solid-state route. ${\rm{S}}{{\rm{c}}_2}{{\rm{O}}_3}$, ${{\rm{NH}}_4}{{\rm{VO}}_3}$, and ${\rm{B}}{{\rm{i}}_2}{{\rm{O}}_3}$ (99.9% purity) were weighed with stoichiometric compositions and 0.1–5 mol.% Bi replacing Sc (${\rm{S}}{{\rm{c}}_{1 - x}}{{\rm{VO}}_4}{:}{\rm{Bi}}_x^{3 +}$). The mixtures were finely ground using a glass mortar/pestle with a small amount of acetone for 30 min, placed in quartz crucibles, fired in a box furnace at 1100°C for 3 h in air (10°C/min heating rate and natural cooling), ground again, fired again under the same conditions, and ground once more.

The particles were analysed using x-ray diffraction (XRD) and scanning electron microscopy (SEM), from which we determined the crystal structure and particle size distribution as shown in Figs. 2(a) and 2(b), respectively. The average particle size is ${2.1}\;{\rm{\pm}}\;{0.7}\;{\rm{\unicode{x00B5}{\rm m}}}$, which is suitable for flow tracing in fluids (temperature and velocity response times at ${{20^{\circ} \rm C}}\;{\tau _{95\% \:}}\sim{{4}}\;\unicode{x00B5} {\rm{s}}$ and 100 µs in water and air, respectively [2]). Rietveld refinement shows that the particles doped up to 1 mol.% are single-phase ${\rm{ScV}}{{\rm{O}}_4}$ with ${\lt}{0.05}\;{\rm{mol}}.\%$ ${\rm{BiV}}{{\rm{O}}_4}$ residual phase, indicating that the Bi ions are well-incorporated into the host. Note that, using this synthesis procedure, at higher dopant concentrations up to 5 mol.% Bi, the diffraction data indicated that higher fractions of the separate ${\rm{BiV}}{{\rm{O}}_4}$ phase were formed, presumably due to the large difference in ionic radius of ${\rm{B}}{{\rm{i}}^{3 +}}$ (117 pm) compared to ${\rm{S}}{{\rm{c}}^{3 +}}$ (87 pm).

 

Fig. 2. (a) XRD pattern for ${\rm{ScV}}{{\rm{O}}_4}{:}{\rm{Bi}}$ 1 mol.%; (b) SEM image and particle size distribution; (c) quantitative luminescence signal measured for different Bi dopant concentrations, 20°C.

Download Full Size | PDF

Luminescence signal characterization was performed using a calibrated system described fully in Ref. [5] based on liquid dispersions (20 mg/L), which allows quantitative determination of the signal per given mass of phosphor particles (${\rm{\pm 10}}\%$). The absolute luminescence signal of dispersed ${\rm{ScV}}{{\rm{O}}_4}{:}{\rm{B}}{{\rm{i}}^{3 +}}$ particles increases with a sublinear trend between dopant concentrations of 0.1 and 1 mol.% [Fig. 2(c)], presumably because the reduction in quantum efficiency due to concentration quenching counteracts increased absorption. For ${\rm{ScV}}{{\rm{O}}_4}$ doped with 1 mol.% Bi, the total luminescence emission is ${{7}} \times {{1}}{{{0}}^{13}}\;{\rm{photons}}/{\rm{mg}}$ at 20°C using the third harmonic of a pulsed Nd:YAG laser (355 nm, 10 ns) with a fluence of ${{10}}\;{\rm{mJ/c}}{{\rm{m}}^2}$. This is an appropriate luminescence intensity to ensure excessively high mass concentrations are not required in experiments.

The excitation spectrum of ${\rm{ScV}}{{\rm{O}}_4}{:}{\rm{B}}{{\rm{i}}^{3 +}}$ 1 mol.% powder measured using a fluorescence spectrometer is shown in Fig. 3(a). The spectrum exhibits features corresponding to excitation of ${\rm{VO}}_4^3$-groups at shorter wavelengths (265 and 330 nm) [16]. The excitation band is extended toward the visible range due to direct absorption of ${\rm{B}}{{\rm{i}}^{3 +}}$ ions. These particles can be efficiently excited using 355 nm, which is convenient from a laser diagnostic perspective. The ${\rm{B}}{{\rm{i}}^{3 +}}$ emission is attributed to the ${^3{{\rm{P}}_1} \to {^1}{{\rm{S}}_0}}$ and $^3{{\rm{P}}_0}\to {^1}{{\rm{S}}_0}$ transitions, appearing as a broad band centred at 610 nm covering most of the visible region with negligible overlap with the excitation spectrum [16]. The emission broadens toward the blue with increasing temperature.

 

Fig. 3. (a) Excitation and (b) emission spectra.

Download Full Size | PDF

The temperature dependence of the lifetime was measured using a photodiode and oscilloscope (measurements of repeat dispersions at 33.5°C ${\lt}\;{{2}}\% \;{\rm{error}}$). Figure 4(a) shows that the lifetime strongly decreases in the temperature range of interest. The lifetime sensitivity ${\psi _\tau}$ is 2.2%/K at 37°C, which is among the highest reported for lifetime-based luminescence thermometers [1]. For reference, the luminescence signal decreases exactly in accordance with the lifetime, i.e., from ${{8}} \times {{1}}{{{0}}^{13}}\;{\rm{photons}}/{\rm{mg}}$ at 20°C to ${2.5} \times {{1}}{{{0}}^{13}}\;{\rm{photons}}/{\rm{mg}}$ at 70°C.

 

Fig. 4. (a) Luminescence lifetime versus temperature; (b) dual-frame intensity ratio and resultant temperature sensitivity.

Download Full Size | PDF

For temperature imaging, the third-harmonic output of a pulsed Nd:YAG laser (355 nm, 10 ns, 10 Hz) was formed into a 400 µm thick light sheet (${{10}}\;{\rm{mJ/c}}{{\rm{m}}^2}$) and directed through a particle-water dispersion contained in a fused silica cuvette (30 ml volume) [see Fig. 1(b)]. The particles were imaged using a ${{4}}\;{\times}\;{{4}}$ hardware-binned interline transfer CCD (1600, PCO) with a 50 mm ${{f}}/{1.4}$ objective and Schott OG570 filter to block both 355 and 532 nm light. The laser pulse and camera exposure were timed as in Fig. 1(a). The laboratory lights were turned off to ensure the background was negligible in the long (11 ms) second camera frame. Note, that the measurement duration is actually determined by the luminescence lifetime (2 µs).

To calibrate the dual-frame intensity ratio with temperature, the particle dispersion was slowly heated on a heating/stirring plate, and camera images were recorded at different constant liquid temperatures measured using calibrated type ${{K}}$ thermocouples (${{\pm 0}.{1^\circ \rm C}}$ error). Dual-frame image pairs were (i) background subtracted, (ii) smoothed to a geometric in-plane resolution of 400 µm, (iii) divided by one another (frame 1/frame 2, see Fig. 1), and (iv) divided by a time-average intensity ratio field recorded at uniform room temperature. The resultant data are shown in Fig. 4(b). Owing to the strong decrease in the lifetime with temperature, the dual-frame ratio has a pronounced dependence on the temperature, with a measurement sensitivity of 3 to 6 %/°C in the 20 to 60°C range. This is ${\gt}{{5}}$ times higher than that of the previously cited ratiometric measurements using the phosphor ZnO [4].

We then made a demonstration experiment as follows. A particle-water dispersion (20 mg/L) at 27°C was injected using a syringe with a tip diameter of 2 mm at ${\sim}{{10}}\;{\rm{ml/min}}$ into a dispersion (also 20 mg/L) heated to 48°C in the cuvette. Dual-frame images were continuously recorded at 10 Hz, processed as above, and converted to temperature using the calibration data [Fig. 4(b)]. A sequence of images corresponding to the first 0.5 s are shown in Fig. 5. Initially at 0 s, there is some transfer of the colder liquid into the cuvette, which forms a denser plume drifting downwards. As the injection starts, cold fluid is forced into the cuvette forming a plume. Fine vortical structures leading to thermal mixing are clearly visible, indicating the high level of precision afforded by this method.

 

Fig. 5. Time series of temperature images recorded at 10 Hz during the injection of cold water into warmer water. Images are displayed at true resolution (1 pixel = 400 µm). The histogram shows independent samples in the indicated uniform region of the single shot at 0.2 s.

Download Full Size | PDF

From instantaneous images recorded at uniform temperature (20°C), we assess that the single-shot pixel-pixel precision is ${{\pm 0}.\rm{4^\circ {\rm C}}}$ (${{1}}\sigma$), 0.1% of the absolute temperature, at a spatial resolution of 400 µm. This level of precision represents a fivefold improvement compared to previous two-color measurements using ZnO phosphor particles [4], and is comparable with that achieved using an optimized dye combination for two-color dye laser-induced fluorescence (LIF) [17]. This method can be used in both liquid and gas flows; particles are already present, allowing combination with simultaneous particle image velocimetry (PIV); and only a single camera is used, which is straightforward compared to color-based imaging thermometries, which require image registration and/or correction for angle dependence of the signal collection, or even in-plane position-dependent calibration as can be the case for TLC temperature imaging [18].

The sensitivity of the dual-frame ratio to temperature can be arbitrarily altered by choosing a different gating scheme. However, it is important to note that the true objective here is to minimize the temperature measurement uncertainty, which is also influenced by the signal captured by each frame. This was evaluated analytically by deriving expressions for how the fractional temperature sensitivity of the dual-frame ratio ${\psi _R}$ and the uncertainty in the dual-frame ratio ${\sigma _R}$ depend on the choice of the first frame exposure time ${t_1}$, here normalized by the luminescence lifetime $\tau :\;x = {t_1}/\tau$. For single-exponential luminescence decay dynamics, and assuming a negligible interframe time and that the second frame is long compared to the luminescence lifetime (which are both features of this interline transfer CCD that cannot be directly controlled), the intensity in the first and second frame are ${I_t}({1 - {e^{- x}}})$ and ${I_t}{e^{- x}}$, respectively. ${I_t}$ is the total collected luminescence signal, which is independent of the gating strategy. The temperature error ${\sigma _T}$ is

The first term describes the sensitivity of the ratio to temperature ${\psi _R}$ and includes the inherent luminescence lifetime sensitivity ${\psi _\tau}$. The second term describes the uncertainty in the ratio, assuming shot-noise-limited operation.

The results shown in Fig. 6(a) confirm that the normalized sensitivity ${\psi _{R}/\psi_{\tau}}$ can indeed be increased by lengthening the first frame exposure. The ratio error shown in Fig. 6(b) is minimized at $x = {\ln}({{2}})$, corresponding to an even distribution of the luminescence signal between the two frames. Beyond this optimum, the ratio error steadily increases. Therefore, in combination, the resultant temperature error ${\sigma _T}$ is minimized using a gating scheme in the range $x\;\sim\;{{1 {-} 3}}$ [Fig. 6(c)]. Similar analyses of different gating schemes [19,20] should guide the choice of dual-frame gating time to optimize the temperature precision, not the temperature sensitivity alone.

 

Fig. 6. Dependence of (a) the dual-frame ratio temperature sensitivity normalized by the inherent luminescence lifetime sensitivity; (b) normalized dual-frame ratio uncertainty; and (c) resultant normalized temperature error, on the first frame exposure time ${t_1}$.

Download Full Size | PDF

In further reference to the measurement uncertainty, first, the shot-to-shot standard deviation is ${{\pm 0}.{5^\circ{ \rm C}}}$, which is caused primarily by laser-camera jitter (${\sim}{{50}}\;{\rm{ns}}$). This error could be reduced by recording the timing fluctuation and correcting each image, or using in-frame referencing where a region of uniform liquid temperature is always known. Second, in test cases where strong temperature gradients exist, laser striping caused by refractive index gradients can lead to temperature errors if the measured quantity (here, the intensity ratio) depends on the laser fluence. We tested the effect of changing the laser fluence and found that across an exaggerated range between 2 and ${{60}}\;{\rm{mJ/c}}{{\rm{m}}^2}$, the total temperature difference amounts to only 3°C. A more reasonable 10% variation in the fluence results in a negligible temperature difference of ${\sim}{{0}.{3^\circ {\rm C}}}$. This effect is similar in magnitude to the phosphor ${({\rm{Sr}},{\rm{Mg}})_3}{({{\rm{PO}}_4})_2}{:}{\rm{S}}{{\rm{n}}^{2 +}}$ [21], and 20 times lower than the dependency observed for ZnO [4], which confers a significant improvement in measurement accuracy. Third, the measured jet core temperatures shown in Fig. 5 match the temperature recorded by the thermocouple to within ${\sim}{{3^\circ {\rm C}}}$. Therefore, we are confident that at and below this mass concentration (${\sim}{{20}}\;{\rm{mg/L}}$), there is no significant adverse effect of multiple scattering between particles, a problem that was reported in gas flows seeded with ${\rm{BaMgA}}{{\rm{l}}_{10}}{{\rm{O}}_{17}}{:}{\rm{E}}{{\rm{u}}^{2 +}}$ tracer particles [22]. Fourth, a potential issue is that the particles can move between frames, which will interfere with the necessary division of images, causing a loss in spatial resolution and distorting the decay waveform because different particles are interrogated in each frame. Considering a flow velocity of 10 m/s and extreme (100%) variation in laser illumination or particle number density, this can result in a 10% ratio error at 400 µm spatial resolution. This is a worst-case scenario and realistic variations would be significantly lower, and if necessary, simultaneous PIV could be used to shift the particle images and remove the in-plane displacement.

Two general suggestions can be made regarding this technique. First, the onset of thermal quenching and therefore the temperature-sensitive range can also be adjusted by choosing a different compound, for example, an alternative host for the Bi ions such as Y or Gd vanadates, which quench at higher temperatures (100°C) [15]. Such particles can be synthesized in the laboratory as described here or by specialist luminescent material suppliers. Second, measurements at kHz sampling rates using a frequency-tripled diode-pumped solid-state laser and a frame straddling method as commonly used for high-speed PIV measurements should be possible, for the time-resolved study of faster flow dynamics.

In summary, ${\rm{ScV}}{{\rm{O}}_4}{:}{\rm{B}}{{\rm{i}}^{3 +}}$ thermographic phosphor particles were synthesized and shown to be suitable for single-shot temperature imaging in a near-ambient temperature liquid flow using a dual-frame lifetime method. The high temperature sensitivity results in a single-shot single pixel temperature precision of ${{\pm 0}. {4^{\circ} {\rm C}}}$ (${{1}}\sigma$), with a spatial and temporal resolution of 400 µm and 2 µs, respectively, representing a fivefold improvement in precision compared to previous measurements using thermographic phosphors. This method is applicable in both gas and liquid flows, displays negligible dependence on the laser illumination, and uses a single ordinary interline transfer CCD camera. These are interesting measurement features when compared alongside other fluid temperature imaging techniques, using a highly versatile class of luminescent sensor materials, and therefore we anticipate this technique will prove useful to the thermal and fluid science metrology community.