Corrections Required for IR Thermography

Before IR thermographic methods can be applied successfully to screen heterogeneous catalyst libraries in gas-phase reactions, a few problems have to be addressed. First of all, the different catalyst materials on the library may emit different amounts of IR radiation even if their surface temperatures are identical. Such emissivity differences can be misinterpreted as temperature differences, and therefore have to be corrected. To be able to image the temperature changes in real time during a catalytic experiment, suitable calibration data which account for these emissivity differences have to be recorded prior to the experiment.

Below we describe the technique of emissivity-corrected IR thermography (ECIRT) as it is applied in our laboratory and give an example where time resolved monitoring of catalytic activity on a model combinatorial library is demonstrated. A schematic drawing of the IR-camera with a PtSi-Focal Plane Array (FPA) detector can be seen in Figure 1 (9). The main component of the camera is the detector with 256 × 256 detector elements on a FPA-chip which is sensitive to radiation in the range of 3 to 5 μm. The detector is cooled to liquid nitrogen temperature by a Stirling heat exchanger. Analog data from the camera are digitised in the analog/digital converter and correction to the detector response is performed in real time in a Digital Signal Processing (DSP) unit which also reduces the IR data from a 14-bit to an 8-bit format. In the frame-grabber, colours are assigned to the 8-bit values from the DSP so that the data can be visualised on a colour monitor.

A common problem of all high resolution FPA-IR cameras is the inhomogeneity of the detector response (10, 11). It is not possible to fabricate the detector elements with sufficient precision so that all detector elements have the same response to IR radiation. To reach sufficient temperature resolution the differences of the individual pixels of the detector have to be electronically corrected. This task is performed by the DSP unit in real time on the basis of correction data which have to be calculated from calibration images taken with the camera prior to the experiment. The PtSi-FPA detector used in our experiments shows a very stable response with time so that the calibration data retain their validity for more than 24 hours. In the case of other commonly used FPA-detector materials, which employ the compound semiconductors indium antimonide and cadmium mercury telluride, the detector response is not stable with time so that frequent recalibradon of the detector is necessary. The advantage of compound semiconductor FPA detectors is their higher sensitivity to IR radiation. High sensitivity is an advantage if fast moving objects have to be imaged as high frame rates can then be applied. It may also be useful for imaging objects at low temperatures, such as room temperature or below. At the elevated temperatures used in our experiments, the amount of emitted IR radiation is quite high, so enough photons are available for detection. Also, as no moving objects are observed, signal averaging can be employed to improve the signal to noise ratio further. In fact, the higher sensitivity of the compound semiconductor detectors is not needed for our experiments.

The correction algorithm used in the camera is based on work by Schulz and Caldwell (11). Basically, calibration images of the library have to be taken at three different temperatures from which a linear correction (gain correction) and an offset correction (image subtraction) are calculated. The image for the offset correction is ideally taken at the final reaction temperature – just before the start of the catalytic experiments. The data are loaded into the DSP unit which then performs the correction in real time. The effect of the correction is that the response of each individual pixel in the detector becomes the same as the average response over the whole detector. Any differences in emissivity on the library surface as well as any temperature differences will be evened out. For imaging of the catalyst library, this means that when no temperature differences occur on the library due to catalytic activity then a structureless image will be recorded. Temperature differences due to catalytic activity can then be easily recognised. Afterwards a temperature calibration can be performed which assigns temperature values to the different colours of the corrected images. Image correction (offset and gain correction) and temperature calibration together make a correction to the emissivity of each spot on the library surface, which allows the photon intensities to be detected as temperature values by the IR camera. Our approach has the advantage that the emissivity correction can be performed in real time by the DSP unit. Other approaches, such as using the Planck formula, require extensive computation and can therefore not be performed in real time. It is essential to have as little differences in temperature as possible on the library since otherwise the correction may become inaccurate.

It is also of general importance to avoid the IR radiation being reflected, especially on the library surface since this can cause artifacts unrelated to catalyst temperature. The IR thermographic method works accurately only if the intensity of emitted IR radiation from each spot on the library surface can be detected, so that the surface temperature values can be calculated and assigned to the location from which the IR radiation was originally emitted. If the emitted radiation is reflected, then the detected information about the location of the emission is obviously not accurate. Thus, it is very important to select a support material for the library which has low reflectivity for IR radiation.

We have tested different library support materials, such as metals, graphite, slate and others that are stable at high temperatures of a few hundred degrees Celsius and are stable against the strongly acidic solutions of the sol-gel catalyst preparations (6). Supports which have high electrical conductivity, such as metals, can only be used with an antireflective coating, since their surfaces show very high reflectivity for IR radiation. Certain roughened graphite surfaces did show lower IR reflectivity than metal surfaces but were too porous and therefore not impervious to liquids. Slate was finally chosen as the library material because of its low IR reflectivity.

The application of ECIRT to follow catalytic reactions in a time resolved mode has already been demonstrated for liquid-phase reactions (8). The use of this method for the detection of catalytic activity in libraries of heterogeneous catalysts is described below. For this purpose a catalyst library containing amorphous microporous catalyst samples of powdered mixed oxides was prepared by filling each hole of a slate library plate with ∼ 1 mg of different catalyst samples. The catalysts were prepared by a sol-gel process (6, 12, 13). This procedure allows the production of a microporous oxide which contains isolated catalytically active centres dispersed in the matrix material. Due to the amorphous nature of the mixed oxide powders there are a few restrictions on the type of elements that can be built into the oxide matrix (12). Most elements of the Periodic Table can be dispersed in different oxide frameworks and thus become accessible as potential catalytic reaction centres. Because of the large number of possible combinations of framework oxide and catalytically active centres, combinatorial approaches are ideally employed for the selection of active catalysts.

The reactor was equipped with an IR-transparent BaF₂ window to allow imaging of the library, and the gas-phase oxidation of toluene with air was chosen as a test reaction (6). Figure 2 shows the chemical composition of the library. All catalysts are binary mixed oxides, for example, oxides of palladium, rhodium, ruthenium, iridium, osmium, gold, silver, copper, zinc, nickel, vanadium, cobalt, tantalum and others, dispersed in a matrix of either titania or silica.

The catalysts on the library were activated by heating at 400°C in a stream of synthetic air for 1 hour to remove any organic residues remaining from their sol-gel preparation. The reactor temperature was then lowered to 300°C and the reaction was started with a constant flow of air and toluene (air:toluene = 25:1) at 20.8 ml min^-1. IR images were taken every 2.5 minutes for a total time of 72.5 minutes. Figure 3 shows IR images of the library for different times after toluene was injected into the gas stream.

Active catalysts were found in the first, second and fourth row of the library. Plots showing the time dependency of the temperature profiles along these rows can be seen in Figure 4. In the first row Pd₁Si and Cu₃Ti are very active catalysts: while Cu₃Ti reaches constant activity quite quickly, Pd₁Si needs more time to reach maximum activity and shows slight deactivation. Rh₁Ti and Mn₃Ti show lower but constant activity. Ru₁Ti and Ru₃Ti show some initial activity but fast deactivation. In the second row Ag₂Ti, Zn₃Ti and Ni₃Ti are quite active with constant heat response, while Ir₃Ti, Cr₃Ti and Rh₂Ti again show some initial activity but fast deactivation. In the fourth row Cr₂Si shows high activity. In the same row Ir₂Si and Rh₂Si show very small temperature decreases which indicate an endothermic reaction but it is not clear if these effects are real.

Temperature plots of Pd,Si and Ru₁Ti can be seen in Figure 5; clearly Ru,Ti shows some initial activity but seems to deactivate quickly. Pd₁Si shows high activity after a period of activation which takes about 10 minutes. Its activity decreases slightly during the course of the experiment.

These observed time dependent phenomena indicate that deactivation (probably due to coking or catalyst changes) occurs. However, it has to be kept in mind that extensive coking or changes in the microstructure of the catalysts may lead to changes in emissivity of the catalyst spots and may give rise to erratic temperature readings. It is more likely that this leads to emissivity increases than to emissivity decreases. To avoid misinterpretations, it is recommended that at the end of an experiment, the reactant gases should be replaced by inert gases, which will allow the immediate identification of catalyst spots where the emissivity has changed during the experiment.

A more detailed investigation of these effects can be performed later with selected catalysts using other techniques, such as mass spectrometry, to detect the reaction products.

Conclusions

Emissivity-corrected infrared thermography allows time dependent phenomena to be observed during a catalytic experiment and is thus useful as a high throughput screening technique. It would ideally be employed at an early stage of the discovery process to narrow down the number of candidate catalysts. Other combinatorial or even conventional techniques, which give more detailed or complementary information about the catalytic processes involved, could be then employed to aid catalyst selection. Some of these techniques, such as kinetic studies or mass spectrometric screening, are more time consuming due to the much more detailed information obtained; but as a result of employing the ECIRT screening, there are fewer selected candidate catalysts to examine. ECIRT could therefore be applied to reduce the amount of time and effort involved in new heterogeneous catalyst selection for gas-phase reactions.