(W1030-08-47) Influence of the Temperature on Near Infrared (NIR) Spectra Determined with a Novel High Speed Sensor used as a PAT Tool for Monitoring the Tableting Process

Purpose: During a tableting process the temperature may vary, which can lead to changes of the physical and chemical properties of the produced tablets. Therefore, the aim of this study was to investigate the extent of the temperature influence on near infrared (NIR) spectra acquired with a novel high speed in-line NIR sensor. Moreover, a data processing method shall be proposed, how to deal with this influence during the calibration of the sensor in order to reduce the influence of temperature changes on the NIR spectra.

Methods: Tablets consisting of 29.4 % Ibuprofen DC 85 W, 69.6 % Tablettose^® 80, and 1 % magnesium were heated up to tem­peratures of approximately 80 °C for 5 min using a heat gun. During the subsequent cooling phase, the tablets were placed below the novel NIR sensor head for spectral measurements to simulate the conditions during monitoring of the tableting process in the context with PAT, i.e. the spectral data acquisition in the tablet ejection zone after compression. The spectral data acquisition only started at a temperature of 38 ± 4.2 °C because of the time delay caused by the sample transfer from the heat gun to the NIR sensor head. A multiplicative scatter correction (MSC) was applied to the spectral data. Subsequently, spectra smoothing was performed either by using a Savitzky-Golay filter or by first order derivation using the Savitzky-Golay algorithm. The preprocessed spectra were subjected to principal component analysis (PCA).

Results: Heating of the tablets led to an irreversible absorption increase of the initial NIR spectra (Figure 1). Subsequent cooling down to room temperature resulted in an only slight decrease of the absorption spectra (Figure 1); a return to the initial NIR spectra was not observed. The absorption increase caused by heating may be a conse­quence of lower reflectance values detected by the NIR spectrometer. The lower reflectance results from an increase in tablet translucency caused by molten ibuprofen which remained in a viscous state during the cooling phase, as no recrystallization occurred. This phenomenon may be explained by a decrease of the interface between the remaining solid state components of the tablet and air, which leads to a lower area of reflection. MSC of the spectral data revealed an increase of the absorption in the second overtone region and a simulta­neous decrease of the absorption in the first overtone region of the spectra during the cooling phase (Figure 2A). A decrease of the temperature resulted in an increasing H-bond strength and subsequent weakening of the OH-bonds, which led to a red shift of multiple bands observed in the first overtone region (Figure 2B). Both, H-bonds and OH-bonds are present in lactose (Tablettose^® 80) and in adsorbed water. The performed PCA allowed to highlight the spectral alterations caused by a temperature change. These spectral alterations may be explained by the third principal component (Figure 3A), the scores of the spectra increasing along the third principal component axis while lowering the temperature. The PCA of the first order differentiated spectra resulted in an explained variation between the samples in the first component attributed to the contents of ibuprofen and lactose in the samples, while variations caused by changes of the temperature could be observed in neither of the components (Figure 3C), the total explained variation amounting to 26 %. An influence of the red shift observed in the wavenumber range of 7000 – 6500 cm^-1 (Figure 2B) could be reduced by first order differentiation. The separation of the scores of the second principal component could be attributed to the isosbestic point at 7000 cm^-1 (Figure 3D, solid curve). No separation of the scores of the third principal component was observed (Figure 3D, dotted curve). To reduce this influence, the wavenumbers attributed to the isosbestic point may be eliminated from the calibration model as long as they do not contain relevant information for the model. Integration of the differentiated spectra and subsequent PCA showed the original absorption spectra similar to the spectra in Figure 2. This approach confirmed that the variation in the NIR spectra caused by temperature changes is reduced but not eliminated through differentiation of the spectra. Thus, differentiation of the spectra and careful selection of the variables may reduce the influence of a temperature change on the calibration models.

Conclusion: There may be several processes causing a change in the temperature during tablet production: elapsed tableting time, altered tableting speed, or varied lubricant concentration. The novel high speed NIR sensor in combina­tion with chemometric methods was shown to detect the effect of the temperature decrease during the cooling phase. The observed temperature effect could be considerably reduced by first order differentiation of the NIR spectra. This approach should be considered in the calibration phase of the sensor, if the sensor is used as a tool for in-line monitoring of the tableting process.


Figure 1: Raw NIR absorption spectra of tablets at room temperature before heating (black) (means ± 95 % CI, n = 3), and at the highest temperature registered while placed below the NIR sensor (red) as well as after cooling down to room temperature (blue) (means ± 95 % CI, n = 3).


Figure 2: Exemplary NIR spectra after Savitzky-Golay smoothing and MSC of one tablet in the cooling phase (A) and a close-up of the wavenumber range 7500 – 5900 cm-1 (B).


Figure 3: Scores plot (A) and Loadings plot (B) resulting from the PCA of the preprocessed NIR spectra, Scores plot (C) and Loadings plots (D) resulting from the PCA of the first order differentiated NIR spectra of three tablets heated up to 80 °C and subsequently cooled down to room temperature.