2022 Volume 70 Issue 5 Pages 362-368
We prepared two kinds of fine particles by treating lactose monohydrate (Lac) with the same formulation in a fluidized-bed granulator, which differed in the spraying air pressure. Raman intensities of treated Lac during processing were measured using a handheld-type Raman spectrometer and plotted against particle size. As particle size increased, Raman intensity decreased in both operating conditions. A linear relationship between them was observed, and regression lines were obtained with good correlation coefficients as calibration curves in both operating conditions. We also prepared two other types of fine particles by treating Lac with the same formulation in a fluidized-bed granulator, which differed in the scale or processing mechanism. The particle size could be successfully predicted using the calibration curve obtained taking powder porosity into consideration. Based on this study, we present a new at-line approach in process analytical technology to monitor and predict particle size using a handheld-type Raman spectrometer.
Recently, several techniques for preparing fine particles by treating drug substances directly have been developed in the pharmaceutical field, e.g., spray-drying process,1) Wurster process,2) and treatment with a multifunctional rotor processor.3,4) We reported the preparation of fine particles with improved drug solubility5) and a scaling-up study of fine particles with bitter taste-masking for orally disintegrating tablets (ODTs)6) using a complex fluidized-bed granulator.
Generally, devices for manufacturing investigational or commercial medicinal products are located in restricted clean areas. When operators measure particle size during processing for the preparation of fine particles using a fluidized-bed granulator, they must move to a separate analytical laboratory and change their clothing. Particle size may increase during measurement, which leads to failure to stop at the process endpoint. In addition, spraying problems may occur once the process is stopped because powder in the device may adhere to the spray nozzle or additives in the binder solution may become solidified in the nozzle or tube, preventing the constant supply of liquid or airflow to the powder. Particularly in the preparation of fine particles, the processing time is long, and it is critical to control particle size strictly to prevent grittiness in the mouth when patients receive ODTs,1) maintain content uniformity for solid dosage forms,6) etc. Accordingly, tools to monitor particle size in and around the device are needed.
Recently, analysis using near-IR (NIR) and Raman spectroscopic techniques have received increasing attention. Raman spectroscopy can be applied for the quantitative analysis of assay results or polymorphic ratios in solid dosage forms.7–11) Both NIR and Raman spectroscopy are powerful tools for monitoring pharmaceutical processes as a form of process analytical technology (PAT). Applications to monitor changes in crystal forms during processing using NIR12–14) or Raman spectroscopy15,16) were reported. The effects of particle size on NIR spectra17–19) and the monitoring of changes in particle size during processing using NIR spectroscopy were also investigated, for example, via wet-milling,20) high-shear granulation,21) fluidized-bed granulation,22,23) and drying processes.24) The measurements are performed in-line, on-line, or at-line. In-line means that the process is monitored inside the device, while on-line monitoring occurs in the outer tube through which samples flow. At-line means that samples in the device are removed and measured rapidly in the vicinity of the process,25,26) while the conventional off-line method measures samples in a separate analytical laboratory. However, there are few reports on particle size monitoring using Raman spectroscopy as a PAT tool.
Real-time particle size measurement using a spatial filtering technique (SFT) was reported as a promising initial step for process control.27) In this paper, real-time and in-line feedback control for a fluidized-bed granulation process is discussed with the evaluation of robustness by changing the concentration of binder solution as a formulation factor and spraying rate as a process factor. In the R&D or commercial stage, changes of scale from a laboratory or pilot scale to a commercial scale or changes in the manufacturing site including global sites must be dealt with. The size and geometry of fluidized-bed granulators generally differ by site. Even if a probe for in-line monitoring can be set in the device, the position or angle for a probe setting, which may affect sampling, is different among devices. Moreover, powder attached to the probe, especially at the initial stage during the granulation process, should be cleaned for precise measurement. For these reasons, another particle size-monitoring method using a handheld-type Raman spectrometer is proposed in this study.
In this study, lactose monohydrate (Lac) was selected as a model material and processed in a conventional fluidized-bed granulator to prepare fine particles with a target size in the range of 150 to 200 µm. Changes in particle size during processing were monitored using a handheld-type Raman spectrometer equipped with a 1064-nm laser beam. The goals were to investigate the relationship between particle size of treated Lac and Raman intensity for calibration purposes and to predict the particle size of treated Lac processed on a different scale or using a different mechanism based on the calibration curves obtained for verification purposes.
Sieved Lac (Pharmatose 200, 100, 50 M, DMV, the Netherlands), granulated Lac (Dilactose S, Freund Sangyo, Japan), talc (Fuji Talc Industrial Co., Ltd., Japan), and hydroxypropyl methylcellulose (TC-5R, Shin-Etsu Chemical Co., Ltd., Japan) were used in this study. Lac (Pharmatose 200 M) was used for the preparation of treated Lac. All materials are listed in the Japanese Pharmacopoeia (JP).
Preparation of Treated Lac for CalibrationTwo kinds of treated Lac, LAC-1 and LAC-2, with the same formulation but with different spraying air pressures were prepared in a conventional fluidized-bed granulator (Flow Coater NFLO-5, Freund Corporation, Japan). The formulations and operating conditions of treated Lac are listed in Tables 1 and 2, respectively.
| Formulation | LAC-1 | LAC-2 | LAC-A | LAC-B |
|---|---|---|---|---|
| For calibration | For verification | |||
| Lactose monohydrate (g) | 3000 | 3000 | 600 | 600 |
| Talc (g) | 15 | 15 | 3 | 3 |
| Hydroxypropyl methylcellulose (g) | 300 | 300 | 60 | 60 |
| Total batch size (g) | 3315 | 3315 | 663 | 663 |
| Operating conditions | LAC-1 | LAC-2 | LAC-A | LAC-B |
|---|---|---|---|---|
| For calibration | For verification | |||
| Machine type | NFLO-5 | NFLO-5 | FL-LABO | SFC-MINI |
| Inlet air temperature (°C) | 80 | 80 | 80 | 80 |
| Exhaust air temperature (°C) | 45–54 | 44 | 40–44 | — |
| Product air temperature (°C) | — | — | — | 42–45 |
| Inlet airflow rate (m3/h) | 240 | 180 | 30 | — |
| Inlet damper (%) | — | — | — | 5.0 |
| Outlet damper (%) | — | — | — | 8.0 |
| Slit damper (%) | — | — | — | 5.0 |
| Rotor speed (rpm) | — | — | — | 400 |
| Agitator speed (rpm) | — | — | — | 600 |
| Spraying rate (g/min) | 30 | 30 | 6.0 | 8.0 |
| Spraying air pressure (MPa) | 0.3 | 0.5 | 0.1 | 0.1 |
A mixture of Lac (3000 g) and a small amount (15 g) of talc was fed into the device and 5% TC-5R aqueous solution (6000 g) was sprayed on the mixture. The operating inlet airflow was set at 240 and 180 m3/h for LAC-1 and LAC-2, respectively, and inlet temperature was set at 80 °C. The spraying rate was 30 g/min in both operating conditions, and spraying air pressure was set at 0.3 and 0.5 MPa for LAC-1 and LAC-2, respectively. After spraying, the particles were dried at 80 °C for 3 min in the same device. During and after processing, particles were extracted for sampling and placed into a polyethylene bag for the determination of particle size, powder porosity, morphology, and Raman intensity. All treated Lac samples extracted from the device during processing were completely dried due to the sufficient airflow in the fluidized-bed device. Accordingly, all samples extracted from the device during processing were used for measurements without further drying.
Preparation of Treated Lac for VerificationLAC-A and LAC-B were prepared as described below. Due to the laboratory scale, large amounts of total sampling during processing may affect the dynamic motion of powder in a fluidized-bed device by reducing the amount of powder. Accordingly, particles were not extracted for sampling during processing.
Preparation in a Conventional Fluidized-Bed GranulatorLAC-A was prepared in a conventional fluidized-bed granulator (Flow Coater). A mixture of Lac (600 g) and a small amount (3 g) of talc was fed into the device and 5% TC-5R aqueous solution (1200 g) was sprayed on the mixture. The operating inlet airflow was set at 30 m3/h, and inlet temperature was set at 80 °C. The spraying rate and spraying air pressure were set at 6.0 g/min and 0.1 MPa, respectively. The particles were then dried at 80 °C for 5 min in the same device. After processing, particles as a final product were extracted for sampling and placed into a polyethylene bag for the determination of particle size, powder porosity, morphology, and Raman intensity.
Preparation in a Hybrid Fluidized-Bed GranulatorLAC-B was prepared in a hybrid fluidized-bed granulator (Spir-A-Flow SFC-MINI, Freund Corporation, Japan). A mixture of Lac (600 g) and a small amount (3 g) of talc was fed into the device and 5% TC-5R aqueous solution (1200 g) was sprayed on the mixture. The inlet airflow was adjusted to generate moderate powder flow using inlet, outlet, and slit dampers, with the inlet temperature set at 80 °C. Rotor and agitator speeds were set at 400 and 600 rpm, respectively. The spraying rate and spraying air pressure were set at 8.0 g/min and 0.1 MPa, respectively. The particles were then dried at 76 °C for 5 min in the same device. After processing, particles were extracted for sampling and placed into a polyethylene bag for the determination of particle size, powder porosity, morphology, and Raman intensity.
Particle Size of Treated and Raw LacThe particle size of treated Lac during processing, LAC-1 and LAC-2, and final products LAC-1, LAC-2, LAC-A, and LAC-B was examined using a laser diffraction method (LS 13 320, Beckman Coulter, Inc., U.S.A.). About 1 g of powder was placed in a dedicated sample cup and the samples were measured in the dry mode after suctioning them following optical axis adjustment in the auto mode and background adjustment in the default mode. Measurements were performed twice for each sample, and particle size was expressed as the average of a volume mean diameter (D50). The particle size of several grades of raw Lac was also measured using the method described above.
Density of Treated and Raw LacBulk density was measured because sampled powder from a fluidized-bed granulator was loosely packed in the polyethylene bag for Raman measurements. The bulk density of treated Lac during processing, LAC-1 and LAC-2, and final products LAC-1, LAC-2, LAC-A, and LAC-B was measured by calculating the volume occupied by a known mass of treated Lac as the samples were passed through a 16-mesh sieve and poured into a graduated cylinder through a funnel according to the procedure described in the JP, 17th edition. The true density of treated Lac was determined using an air-comparison pycnometer (1000, Tokyo-Science, Co., Ltd., Japan). Measurements of bulk and true densities were performed three times and expressed as the averaged values. Bulk and true densities of raw Lac of several grades were also measured using the method described above.
Porosity of Treated and Raw LacPowder porosity (ε) was calculated using the following equation28):
 | (1) |
where ρb and ρt are bulk density and true density, respectively.
Observation of Treated LacThe morphology of the final products LAC-1, LAC-2, LAC-A, and LAC-B was examined using scanning electron microscopy (SEM) (Miniscope TM3030Plus, Hitachi High-Tech Corporation, Japan) with magnification of ×500. Prior to the examination, the samples were sputter coated (MSP-mini, Vacuum Device Inc., Japan) with a gold layer under a vacuum for 30 s.
Raman Spectroscopy of Treated and Raw LacA handheld-type Raman spectrometer (Progeny, Rigaku Corporation, Japan) equipped with a 1064-nm laser beam was used for spectra recording. Measurements were performed in auto mode with an analysis range of 200–2400 cm−1 by attaching the handheld-type Raman spectrometer to the polyethylene bag filled with samples. Raman intensities at shifts of 355, 474, and 1090 cm−1 were investigated as representative Raman shifts. Samples were measured five times, and Raman intensity was expressed as the averaged value.
Preparation of Calibration CurveRaman intensity of LAC-1 and LAC-2 during processing was plotted against particle size around 60, 90, 110, 130, 150, 175, and 180 µm for LAC-1 and 60, 90, 110, 130, 150, and 160 µm for LAC-2. The regression lines were fitted to the plots as calibration curves for both forms of treated Lac at each Raman shift of 355, 474, and 1090 cm−1.
Verification of Calibration CurvesThe two final products LAC-A and LAC-B were used for the verification of calibration curves obtained using the above method. Predicted particle sizes of LAC-A and LAC-B were calculated by Raman intensity and equation of the regression line at each Raman shift. Regression lines were expressed by the following equation.
 | (2) |
where x and y are particle size and Raman intensity, respectively, and a and b are slope and intercept of the straight lines, respectively. Equations of regression lines obtained from LAC-1 at three Raman shifts were defined as Eqs. 2a and those obtained from LAC-2 at three Raman shifts were defined as Eqs. 2b. Predicted particle sizes were compared with the observed sizes. We evaluated whether LAC-A and LAC-B particle size could be accurately predicted using Eqs. 2a or Eqs. 2b.
Raman Intensity of Tapped LacA mixture of Lac and a small amount of talc used for processing was poured into a graduated cylinder using the method described above. Bulk density and Raman intensity were measured before and after tapping the graduated cylinder. Measurements were performed by attaching the handheld-type Raman spectrometer to the graduated cylinder filled with samples. Tapping was performed once, and measurement of Raman intensity was performed three times. Raman intensity was expressed as the average value. Other measurement conditions were same as in the method described above.
One of the critical process parameters for controlling particle size in fluidized-bed granulation is spraying air pressure.3,4,6) Accordingly, two kinds of particles were prepared with different spraying air pressure using a conventional fluidized-bed granulator.
NIR spectra are affected by powder properties. In addition to size, particle density and surface texture affect the back-reflected light.29) Powder porosity and surface texture may also affect Raman intensity, and these properties were also examined. Progressive changes in particle size and powder porosity as a function of processing time are shown in Fig. 1.
○: Preparation of LAC-1; ●: preparation of LAC-2.
As shown in Fig. 1-A, in both operating conditions, particle size rapidly increased at the earlier stage and gradually increased at the later stage. However, during all processing times, particle size of LAC-1 was always greater than that of LAC-2 due to the lower spraying air pressure, which may provide a broader spraying mist as previously reported for the preparation of fine particles.3,4,6) As shown in Table 3, particle size of final products LAC-1 and LAC-2 after processing was 206 and 147 µm, respectively. As shown in Fig. 1-B, in both operating conditions, powder porosity first rapidly decreased to the minimum and gradually increased toward a constant level. However, during all processing times, powder porosity of LAC-1 was always greater than that of LAC-2 due to the larger particle size, which may be filled less tightly. As shown in Table 3, powder porosity of final products LAC-1 and LAC-2 after processing was 66 and 61%, respectively. Two kinds of particles with different profiles of powder porosity were obtained in this study. As discussed below, the slight differences in porosity profiles may be important in the prediction of particle size based on Raman intensity.
| Lot | Particle size (µm) | Bulk density (g/mL) | True density (g/mL) | Powder porosity (%) |
|---|---|---|---|---|
| LAC-1 | 205.9 | 0.49 | 1.42 | 65.5 |
| LAC-2 | 146.7 | 0.57 | 1.46 | 61.0 |
| LAC-A | 176.6 | 0.42 | 1.38 | 69.6 |
| LAC-B | 173.9 | 0.47 | 1.43 | 67.1 |
The difference in surface texture, for example, between the smooth surface of spherical particles and rough surface of formless granules, may affect Raman intensity because Raman scattering may change. The morphology of final products LAC-1, LAC-2, LAC-A, and LAC-B was examined using SEM (Fig. 2). They appeared formless, and their surface textures did not differ markedly.
NIR spectroscopy is a promising technique for process monitoring because the measurements are nondestructive and can be rapidly collected on-line. However, there are several disadvantages in the use of NIR in process monitoring. One of the most obvious is that complex analysis is necessary to isolate features of interest.30) In addition, when discussing the effect of particle size on NIR spectra during wet granulation, the presence of water used as a binder solution must be considered. This effect cannot be solely ascribed to increasing particle size, and contributions from other factors such as changes in the optical properties of granules due to added water, distribution of water within granules, and changes in granule bed density must be considered.21,30) Moreover, handheld-type NIR spectrometers for use in production areas are not yet commercially available. On the other hand, Raman spectroscopy has advantages over NIR in monitoring changes in drug substances during processing. Raman measurement can be performed without complex analysis. Water used as a binder solution is a poor Raman scatterer, whereas drug substances are typically good Raman scatterers.30) Handheld-type Raman spectrometers for use in production areas are also commercially available.
Representative Raman spectra before and after treatment of Lac to produce LAC-1 are shown in Fig. 3. Raman shifts in the spectra of both were essentially the same. However, Raman intensity was greater before Lac treatment than after. A similar result was obtained for LAC-2. From these results, it is suggested that the handheld-type Raman spectrometer used in this study could be useful in particle size monitoring during processing. As the spraying operation proceeds, the amount of hydroxypropyl methylcellulose in treated Lac increases. The effect of Raman peaks was confirmed by comparing the Raman peaks of hydroxypropyl methylcellulose with that of the final treated Lac product. In addition, the Raman intensity of the final product of LAC-1 is less than that of LAC-2, and therefore the effect of Raman peaks of hydroxypropyl methylcellulose on the final product of LAC-1 would be maximum in this study. Spectra of hydroxypropyl methylcellulose, which reflect the ratio in the final product of LAC-1, are also shown in Fig. 3.
………: before processing, ———: after processing, – – – –: hydroxypropyl methylcellulose.
As seen from the Raman spectra of hydroxypropyl methylcellulose, significant peaks were not observed at 355, 474, and 1090 cm−1. The effect of Raman peaks of hydroxypropyl methylcellulose on the calibration and verification study would therefore be small. We confirmed that Raman shifts in raw and treated Lac were also essentially the same, indicating that the Raman spectra of treated Lac were not significantly affected by other materials (data not shown). Three representative Raman shifts at 355, 474, and 1090 cm−1 were selected for further examination.
Preparation of Calibration CurveFor the preparation of calibration curves, Raman intensities at the Raman shifts of 355, 474, and 1090 cm−1 were plotted against particle size during processing in both operating conditions, and regression lines were fitted to the plots. The results of the preparation of LAC-1 and LAC-2 are shown in Figs. 4 and 5, respectively. In both operating conditions, Raman intensity decreased as particle size increased in all Raman shifts. Wang et al. reported that a similar tendency was observed in inorganic materials with different particle sizes using Raman spectroscopy.31)
Raman shift at (A) 355, (B) 474, and (C) 1090 cm−1.
Raman shift at (A) 355, (B) 474, and (C) 1090 cm−1.
Higgins et al.20) applied NIR spectroscopy to on-line particle size monitoring during the wet-milling process and found that an increase in the overall intensity of NIR spectra was observed with baseline elevation as the process proceeded and particle size decreased. It was suggested that the reason for this was that NIR light undergoes greater multiple scattering between particles before they are reflected back to the detector as particle size decreases and particle number increases, although less NIR light is directly scattered back to the detector from smaller particles.20) A similar tendency was observed in organic materials with different particle sizes using NIR spectroscopy.17,18) The reason for the increased Raman intensity of smaller particles is suggested to be similar in NIR spectroscopy.
Equations of regression lines and their correlation coefficients are also listed in Table 4. The variables x and y in the equations of regression lines represent particle size and Raman intensity, respectively. Straight lines with good correlation coefficients were obtained for all calibration curves.
| Lot | Raman shift | Equation of regression line | Correlation coefficient |
|---|---|---|---|
| LAC-1 | 355 cm−1 | y = −0.1980x + 59.813 | 0.9550 |
| 474 cm−1 | y = −0.1230x + 37.653 | 0.9553 | |
| 1090 cm−1 | y = −0.1431x + 51.474 | 0.9142 | |
| LAC-2 | 355 cm−1 | y = −0.1457x + 54.911 | 0.9534 |
| 474 cm−1 | y = −0.0816x + 33.716 | 0.9678 | |
| 1090 cm−1 | y = −0.0893x + 46.379 | 0.9659 |
It should be noted that the regression lines obtained differed between operating conditions at all Raman shifts, even in the same formulation, especially in the slope of the lines. For more precise evaluation, we investigated the effects of powder porosity on Raman intensity. Raman intensity with lower powder porosity was larger compared with that with higher powder porosity even in particles of similar size, as shown in Figs. 4 and 5. Accordingly, powder porosity should be considered when evaluating the relationship between particle size and Raman intensity.
The calibration curve between particle size and Raman intensity should be prepared in the constant region of powder porosity (after a granulation time of 50 min in Fig. 1-B) because the porosity affects the intensity to a certain extent. However, it is important to conduct process monitoring from the initial stage of granulation due to the drastic changes in particle size. In this study, the calibration curve was prepared from the initial stage of granulation.
Verification of Calibration CurvesWe prepared two forms of treated Lac, LAC-1 and LAC-2, with the same formulation but under different operating conditions for calibration purposes. Equations of regression lines, Eqs. 2a and Eqs. 2b, were obtained for each treated Lac form. We also prepared two other forms of treated Lac, LAC-A and LAC-B, with the same formulation but which differed in scale or processing mechanism to verify whether their particle size could be accurately predicted using Eqs. 2a or Eqs. 2b. The results are shown in Table 5.
| Lot | Observed (µm) | Raman shift | Raman intensity | Calculated by | |
|---|---|---|---|---|---|
| Eqs. 2a (µm) | Eqs. 2b (µm) | ||||
| LAC-A | 176.6 | 355 cm−1 | 23.74 | 182.2 | 213.9 |
| 474 cm−1 | 15.58 | 179.5 | 222.3 | ||
| 1090 cm−1 | 25.32 | 182.8 | 235.8 | ||
| LAC-B | 173.9 | 355 cm−1 | 23.58 | 183.0 | 215.0 |
| 474 cm−1 | 15.56 | 179.6 | 222.5 | ||
| 1090 cm−1 | 24.94 | 185.4 | 240.1 | ||
LAC-A and LAC-B particle size was successfully predicted by Eqs. 2a at all Raman shifts within around 10 µm. However, Eqs. 2b did not predict particle size accurately, likely due to the difference in powder porosity. The powder porosity of LAC-A and LAC-B was 70 and 67%, respectively, similar to that of LAC-1. It should be noted that the prediction of particle size of LAC-A and LAC-B from the calibration curve of LAC-2 was performed using an extrapolation method rather than an interpolation method, which may be less precise. One of the possible reasons for the failure to predict LAC-A and LAC-B particle size from the calibration curve of LAC-2 may be due to the extrapolation method in addition to the difference in powder porosity.
In this study, a calibration model was established on a pilot (not laboratory) scale. Larger total sampling amounts during processing on a laboratory scale may alter the dynamic motion of powder in a fluidized-bed device by reducing the amount of powder. The reason for selecting the pilot scale was to minimize the effect of sampling on the preparation and to establish a more robust calibration model. Although this model was applied on a laboratory scale for the preparation for LAC-A and LAC-B, it can also be applied on a commercial scale. It should be noted, however, that careful consideration is needed when using this model because powder density changes markedly on the commercial scale.
Effect of Powder Porosity on Raman Intensityi) Raman Intensity of Raw LacRaman intensity was compared among several types of commercially available Lac with various powder properties. Pharmatose 200, 100, and 50 M as sieved Lac and Dilactose S as granulated Lac were selected for investigation. In addition, LAC-1 extracted for sampling at 140 min during processing was also used as a reference. Their powder properties are listed in Table 6, and Raman intensities are shown in Fig. 6.
| Product name | Particle size (µm) | Bulk density (g/mL) | True density (g/mL) | Powder porosity (%) |
|---|---|---|---|---|
| Pharmatose 200 M | 44.0 | 0.46 | 1.79 | 74.3 |
| Pharmatose 100 M | 160.9 | 0.66 | 1.47 | 55.1 |
| Pharmatose 50 M | 396.0 | 0.79 | 1.47 | 46.3 |
| Dilactose S | 90.6 | 0.50 | 1.55 | 67.7 |
| LAC-1 (140 min) | 160.6 | 0.50 | 1.41 | 64.5 |
Raman shift at (A) 355, (B) 474 and (C) 1090 cm−1.
When comparing the three types of sieved Lac, Pharmatose 200, 100, and 50 M, as the particle size increased, Raman intensity decreased, although powder porosity also decreased. When comparing Pharmatose 200 M and Dilactose S with similar powder porosity, Pharmatose 200 M particle size was smaller and Raman intensity was greater. These results were also seen in comparisons of LAC-1 and LAC-2, suggesting that the dependency on particle size may be greater than that on powder porosity.
When comparing Pharmatose 50 M and Dilactose S with similar Raman intensity (355 and 474 cm−1) or greater Raman intensity in Pharmatose 50 M (1090 cm−1), the particle size of Pharmatose 50 M was larger than that of Dilactose S. Raman intensity should decrease with larger particle size, but Raman intensity did not decrease even though Pharmatose 50 M had larger particle size. The reason for this may be the lower powder porosity of Pharmatose 50 M compared with Dilactose S. As described in the previous section, the Raman intensity of LAC-1 with higher powder porosity is less than that of LAC-2. When comparing Pharmatose 100 M and LAC-1 (at 140 min) with similar particle size, Pharmatose 100 M showed greater Raman intensity than LAC-1 (at 140 min). The reason for this may be the lower powder porosity of Pharmatose 100 M. These results suggest that powder porosity affects Raman intensity to some extent.
ii) Raman Intensity of Tapped LacProgressive changes in powder density and Raman intensity at the Raman shift of 355 cm−1 with tapping time are shown in Fig. 7. As seen in Table 6, Pharmatose 200 M is bulky compared with the other Lac grades. Changes in Raman intensity may become obvious because powder density increases after tapping. Accordingly, Pharmatose 200 M was selected for this study.
As shown in Fig. 7-A, bulk density increased gradually until 120 tapping times. On the other hand, as shown in Fig. 7-B, Raman intensity rapidly increased until 40 tapping times, after which it remained almost constant. The profiles of Raman intensity at the Raman shifts of 474 and 1090 cm−1 were also similar. From these results, if bulk density increases in the range of around 0.47 to around 0.62 g/cm3, Raman intensity may increase. If bulk density increases to more than around 0.62 g/cm3, Raman intensity may not increase further.
As shown in Table 3, the powder densities of LAC-1 and LAC-2 were different within the range of around 0.47 to around 0.62 g/cm3, meaning that Raman intensity might be affected by powder density. The calibration curves obtained from LAC-1 and LAC-2 were different, which may indicate that Raman intensity is affected not only by particle size but also by powder density. Table 6 shows that the powder densities of Pharmatose 100 and 50 M were greater than 0.62 g/cm3, indicating that Raman intensity may not be affected by powder density.
For more precise particle size prediction, Raman intensity taking powder porosity/density into consideration will be necessary. There are many points and restrictions to be overcome in this at-line approach. If this model scaled up from the laboratory or pilot scale to a commercial scale, monitoring or prediction of particle size can be performed by measuring Raman intensity although powder porosity/density should be confirmed as a preliminary study. If this model is used for process control on a commercial scale after establishment, monitoring and end-point determination can be performed by measuring Raman intensity in the at-line approach.
We prepared two types of treated Lac with the same formulation under different operating conditions to obtain the calibration curves. Powder properties and Raman intensity during processing were evaluated. There was a relationship between particle size and Raman intensity, and two types of regression lines with good correlation coefficients were obtained for all Raman shifts.
We also prepared two other types of treated Lac with the same formulation, but on different scales or using different processing mechanisms to verify the calibration curves. The particle size of treated Lac was successfully predicted by the equation of the regression line obtained taking powder porosity into consideration.
Furthermore, the effect of powder porosity/density on Raman intensity was investigated using several types of commercially available Lac and by tapping Lac. A dependency on powder porosity as well as particle size was observed in the comparisons among the different Lac grades. Although Raman intensity increased with increasing powder density, there was a limit on the final increase.
Although the range of target particle size was limited in this study and further detailed investigations are necessary, this PAT method using a handheld-type Raman spectrometer is useful for monitoring and predicting particle size.
We gratefully acknowledge Mr. K. Miyata (Freund Corporation, Japan) for technical support in the preparation of treated Lac and Mr. H. Kagoshima (Rigaku Corporation) for technical support in Raman measurements using Progeny. We also thank Dr. H. Hisada (Tekanalysis Inc.) for informative advice on Raman measurement.
The authors declare no conflict of interest.
