(W1030-09-59) A Novel High Speed NIR Sensor as a PAT Tool for Monitoring of the Tableting Process: Determination of the Effective Sampling Depth

Purpose: The purpose of this study was to determine the effective depth of NIR rays registered by a novel high speed in-line near infrared (NIR) reflection sensor. This key parameter allows the calculation of the sampled volume. In contrast to Monte Carlo simulations, which have already been used in other studies, in this study a practical experimental setup was employed. Moreover, a mathematical model which describes the intensity of the light reflected by the top layer of bilayer tablets was developed. This model was subsequently used to compute the effective sampling depth using a definition similar to that described by Berntsson et al. [1].

Methods: Bilayer tablets were produced in single punch mode on a Fette 102i rotary tablet press. The bottom layer consisted of plain polytetrafluoroethylene (PTFE), which was compressed at 29.6 ± 2.95 kN (mean ± SD, n = 59). PTFE was chosen for its known linear NIR spectrum. The top layer contained a mixture of 29.4 % Ibuprofen DC 85W, 69.6 % microcrystalline cellulose 102, 1 % magnesium stearate, and < 1 % Ponceau 4R (E 124) as red coloring agent. This layer was compressed at 4.22 ± 0.68 kN (mean ± SD, n = 59) while its target thickness ranged between 0.08 and 2.0 mm (Fig. 1).
The thickness of the top layer was measured using a Zeiss Discovery.V8 stereo microscope.
The raw spectra were acquired with an integration time of 3 ms and a wavenumber bandwidth of 4,535 to 11,111 cm-1. The bilayer tablets were placed under the NIR sensor maintaining a constant distance of 2.5 mm between the sensor head and the tablet surface (Fig. 1).
The bilayer tablets were then put in a beaker containing deionized water so that the top layer disintegrated, while the bottom layer maintained its form. Spectra were acquired under the same conditions as described above after storing the obtained PTFE tablets in a desiccator for one week (Fig. 2 A, blue line).
Data pre-processing and non-linear fitting were performed using RStudio, comprising signal filtering and calculation of reflection spectra from raw reflection values. Data was then fitted to the following equation:
R(x)=R_∞+(R_interface−R_∞)∙e^−2∙k∙x (Eq. 1)
Equation 1: R(x) – reflection of the bilayer tablets with top layer thicknesses of x mm, R_∞ – reflection of the bilayer tablets with an optically infinite top layer thickness, R_interface – reflection of the interface, k – attenuation coefficient, x – thickness of the top layer of the bilayer tablets.

Results: An increase in reflection was observed with increasing thickness of the tablets top layer (Fig. 2, B – D). This relation may be explained by the stronger spectral influence of the bottom layer in case of thin top layers. With increasing top layer thickness, the reflection value did not change above a specific thickness and an asymptotic behavior was observed. After this specific thickness, which was defined as the sampling depth in this study, a contribution of the bottom layer to the spectral information was not observed. For each wavenumber, an asymptotic function (Eq. 1) was fitted to the reflection values. The deviation of the reflection values from the fitted curve at thick top layers ( > 1 mm) was used to calculate the confidence interval, which was extrapolated to lower thicknesses (Fig. 2, B – D, dotted lines). The intersection point between the fitted curve and the lower confidence limit is assigned to the top layer thickness, below which the reflection changes are significant. The thickness value at this point (Fig. 2, B – D, red arrows) was considered as the true sampling depth of the novel NIR sensor. This definition was applied to each wavenumber. Thus, a sampling depth with values between 0.2 and 0.8 mm was obtained (Fig. 3).

Conclusion: A novel method to physically determine the sampling depth of a high-speed in-line NIR tablet uniformity sensor from 4,535 to 11,111 cm-1 was successfully developed and applied. The knowledge about the sampling depth may be further used to determine the sampling volume and thus to evaluate the monitoring capabilities of this novel NIR sensor.

References: [1] Berntsson O, Danielsson L-G, Folestad S. Estimation of effective sample size when analyzing powders with diffuse reflectance near-infrared spectrometry. Anal. Chim. Acta 1998; 364: 243-251

Acknowledgements: The authors would like to thank Gerhard Waßmann from Lehmann&Voss for providing the PTFE powder used in this study.

Figure 1: Experimental setup: top layer with variable thickness represented as red rectangles; bottom PTFE layer represented as dotted rectangles.

Figure 2: Reflection spectra of the bilayer tablets (n = 59) with top layers of different thicknesses (A) and the corresponding reflection intensity profiles at different wavenumbers (B-D). The blue line and the blue squares, respectively, represent the reflection spectrum or reflection values of the interface. The red arrows point to the abscissa values of the intersection point between the asymptotic curve and the extrapolated confidence interval, which was considered as the sampling depth.

Figure 3: Sampling depth profile of the novel in-line reflection NIR sensor.