Moisture Behavior of Powder During Tableting

pharmaceutics
Article
Moisture Behavior of Pharmaceutical Powder during the
Tableting Process
Komlan Koumbogle, Ryan Gosselin , François Gitzhofer * and Nicolas Abatzoglou *
Department of Chemical and Biotechnological Engineering, Université de Sherbrooke, 2500 Boulevard de l’Université,
Sherbrooke, QC J1K 2R1, Canada; [email protected] (K.K.); [email protected] (R.G.)
* Correspondence: [email protected] (F.G.); [email protected] (N.A.)
Abstract: The moisture content of pharmaceutical powder is a key parameter contributing to tablet
sticking during the tableting process. This study investigates powder moisture behavior during the
compaction phase of the tableting process. Finite element analysis software COMSOL Multiphysics®
5.6 was used to simulate the compaction microcrystalline cellulose (VIVAPUR PH101) powder and
predict temperature and moisture content distributions, as well as their evolution over time, during
a single compaction. To validate the simulation, a near-infrared sensor and a thermal infrared
camera were used to measure tablet surface temperature and surface moisture, respectively, just
after ejection. The partial least squares regression (PLS) method was used to predict the surface
moisture content of the ejected tablet. Thermal infrared camera images of the ejected tablet showed
powder bed temperature increasing during compaction and a gradual rise in tablet temperature
along with tableting runs. Simulation results showed that moisture evaporate from the compacted
powder bed to the surrounding environment. The predicted surface moisture content of ejected
tablets after compaction was higher compared to that of loose powder and decreased gradually as
tableting runs increased. These observations suggest that the moisture evaporating from the powder
bed accumulates at the interface between the punch and tablet surface. Evaporated water molecules
can be physiosorbed on the punch surface and cause a capillary condensation locally at the punch
and tablet interface during dwell time. Locally formed capillary bridge may induce a capillary force
between tablet surface particles and the punch surface and cause the sticking.
Citation: Koumbogle, K.; Gosselin,
R.; Gitzhofer, F.; Abatzoglou, N.
Keywords: tableting; COMSOL Multiphysics; finite element analysis; compaction; sticking; moisture;
Moisture Behavior of Pharmaceutical
Pharmaceutics; Drucker–Prager cap (DPC); heat; air and moisture (HAM); temperature field; relative
Powder during the Tableting Process.
density; NIR penetration depth
Pharmaceutics 2023, 15, 1652.
https://doi.org/10.3390/
pharmaceutics15061652
Academic Editors: Peter Timmins 1. Introduction

and Juan José Torrado
Tableting is the mechanical process adopted by the pharmaceutical industry to produce
Received: 14 April 2023 medicinal tablets. It involves direct compression of a powder mixture in a die using three
Revised: 25 May 2023 steps: die filling, during which the formulation is delivered to the die cavity; compaction,
Accepted: 31 May 2023 during which pressure is applied to the formulation; and ejection, when the compacted
Published: 4 June 2023 tablet is ejected from the die cavity [1]. Up to 1 million tablets can be pressed per hour with
one multi-die rotating press. The formulation compacted during the process is a mixture of
active principal ingredients and excipients, such as ligands, binders, and lubricants, with a
given moisture content.
Copyright: © 2023 by the authors.
The study of the physics of the tableting process is of great interest in order to un-
Licensee MDPI, Basel, Switzerland.
derstand the mechanism of consolidation of the granules during compaction, the punch-
This article is an open access article
distributed under the terms and
sticking phenomena, and the quality management of the pressed product. Several math-
conditions of the Creative Commons
ematical equations, such as Hekel’s [2], Kawakita’s [3], and Leuenberger’s [4] equations,
Attribution (CC BY) license (https:// have been developed to investigate the ability of formulations to flow, deform, and con-
creativecommons.org/licenses/by/ solidate under pressure. These equations are usually used to predict how a formulation
4.0/). performs during the tableting process and minimize defects related to the pressing process.
Pharmaceutics 2023, 15, 1652. https://doi.org/10.3390/pharmaceutics15061652 https://www.mdpi.com/journal/pharmaceutics

Pharmaceutics 2023, 15, 1652 2 of 33
Although these equations are useful for preparing the best formulations for the tableting
process, they do not provide a detailed insight of the phenomena occurring in the whole
volume of the powder bed during the compaction process. Hence, simulation models
based on the finite element method (FEM) have been developed to analyze in greater detail
the powder bed behavior during compaction [5–8]. These phenomenological constituent
methods are based on experimental calibrations using the variable physical properties
of the compressed material. Several constitutive models have been developed, of which
the DiMaggio–Sandler model [9], the Cam-Clay model [10], and the Drucker–Prager Cap
(DPC) model [5,6] are well known. The DPC model is the most used for the pharmaceutical
tableting process due to its ability to describe physical phenomena taking place during
compression of a granular medium. It was first developed to investigate irreversible defor-
mations of soils during compaction [5]. The DPC model was subsequently modified and
adapted to pharmaceutical powder, and a calibration method was proposed to define all
the involved parameters [11,12]. This modified DPC model was then used to study the
mechanical behavior of powders, such as strain, stress, and density distribution, in the
compact during compaction [13–17]. Thermomechanical analysis using the modified DPC
model described how the heat generated during compaction progressively increased the
powder bed temperature in all its volume [18–20] and was conclusive, along with results
from previous experimental work [21–23].
During the tableting process, defects such as cracking, stretching, capping, lamination,
chipping, restricting, and sticking may be present in the finally ejected tablets [1]. Coupling
the FEM with the modified DPC allows determination of how the defects developed during
compaction and helps to improve the understanding of the tableting process accordingly.
Defects such as capping, cracking, lamination, and chipping have been intensively investi-
gated and have been shown to originate from the inhomogeneous distribution of stress,
relative density, and buildup of air pressure in the powder bed during compaction [24–27].
Furthermore, defects such as sticking, which is a persistent problem during the tableting
process, were investigated to discover how the tableting process parameters, compressed
material, compression tool properties, and ambient conditions contributed to the sticking
defects [28–36]. Multivariate analysis, adhesion force measurement, push-off force measure-
ments, material accumulation quantification, and the discrete element method have often
been used for testing the propensity to stick [37–43]. (1) These studies identified the van
der Waals forces and capillary forces to play a dominating role in the observed adhesion.
The interplay between the interparticle cohesive force and wall–powder adhesion force
results in localized attachment of the surface granules or layer of the tablet to the surface
of the compression tool. Tableting force and speed, moisture level of the formulation,
and properties of the compressed material govern the cohesivity of the tablet, and hence
the tendency to stick. (2) From a phenomenological point of view, an increase in granule
temperature during compaction may lead to the melting of some ingredients, such as
ibuprofen and magnesium stearate, or of eutectic components present in the formulation,
which may contribute to sticking. However, other studies have invalidated the melting
hypothesis by showing that sticking still occurs when high-melting-point material, such
as ibuprofen sodium, are compressed [44]. (3) Recent multivariate analysis of granulation
and tableting process parameters [41,43] showed that the moisture level of the formulation
plays a key role in the occurrence of sticking. While the results obtained from these studies
provide good data to understand sticking, it is of interest to identify the phenomena that
can trigger the development of sticking during compaction.
Near-infrared (NIR) spectroscopy is often used in the pharmaceutical industry as a
faster, cheaper, and nondestructive way of analyzing pharmaceutical products. Its use is
encouraged in process analytical technology (PAT) for on- and inline measurements and
control of pharmaceutical products. The overall objective of its use is to probe a sample
to acquire qualitative and/or quantitative information about the interaction between
NIR electromagnetic waves and the sample’s constituents. NIR spectroscopy has been
employed to investigate H-bonds and hydration of various molecules, such as alcohols,
fatty acids, amides, polymers, and water [45,46]. In the NIR region, (1) bands arising from
overtones and combinations of O-H, N-H and C-H vibrations appear strongly due to their
larger anharmonicity, and (2) large band shifts are induced by the formation of hydrogen
bonds and hydration. Within the NIR wavelength range, i.e., 700 to 2500 nm, prominent
absorption bands of liquid water are found at 760, 970, 1190, 1450, and 1940 nm [47,48].
These absorption bands are due to the second overtone of the OH stretching band (3ν1,3),
the combination of the first overtone of the O-H stretching band and OH-bending band
(2ν1,3 + ν2), the first overtone of the OH-stretching band (2ν1,3), and the combination of the
OH-stretching band and O-H bending band (2ν1,3 + ν2) [49]. Hence, NIR spectroscopy has
been used to measure the moisture content of pharmaceutical tablets and powders during
granulation, drying, and the tableting process [46,50,51]. The quantification of moisture
content is often based on an experimental calibration model obtained with partial least
squares (PLS) regression, multiple linear regression, or principal component regression.
The calibration model thus established is then used to predict the inline or offline moisture
content of powder or tablets during such pharmaceutical processes as wet granulation
or tableting [52–54].
This study aimed to investigate phenomenologically the buildup of the sticking phe-
nomenon during the compaction of pharmaceutical powder through the simulation of
powder moisture behavior under compression. The AFEM method, based on the modified
DPC model, was used to investigate moisture transport and thermomechanical behavior of
powder during the tableting process, particularly during the compaction phase. The evolu-
tion and distribution of powder bed density, temperature, and moisture changes during the
compaction process were studied. We hypothesized that the migration of powder moisture
toward compression tools due to temperature and local humidity gradients contributes
to the development of sticking. The results of this study show how thermomechanical
phenomena can lead to moisture migration and potentially further induce sticking through
capillary forces. To validate the simulation results, the temperature of the tablet surface and
sides and the surface moisture content of the tablet a second after ejection were measured
using PAT tools, such as a thermal infrared camera and NIR sensor. The results of this
study provide new and original insights into the phenomenology of the sticking defect
observed during the tableting process.
2. Materials and Methods

2.1. Materials
Microcrystalline cellulose (MCC) grade VIVAPUR PH101 (JRS Pharma, LP., Patterson,
New York, NY, USA) was used as the model powder material. The bulk density of the loose
powder was measured with a 100 mL beaker and was found to be 279 ± 2.5 kg/m3 . The
true density of the dry, solid particles of MCC ranges between 1512 and 1668 kg/m3 [55],
and hence a value of 1570 kg/m3 [20] was adopted in this study. The powder was stored
at a controlled humidity level of 33.5 ± 0.1% and a temperature of 21.1 ± 0.5 ◦ C. The
water content of the powder was measured with an analytical transmitter (Mettler M100,
Mettler-Toledo Ltd., Leicester, UK), which dries the powder with an infrared source and
calculates the water content by mass difference. The mean of three measurements was
calculated, and the value obtained for water content was 4.5 ± 0.3%. Water content density
of the powder in the die before compaction was obtained by the following formula:
ρ app_w = ρ app ∗ w (1)
where ρ app_w is the apparent moisture content density (kg/m3 ), ρ app is the apparent powder
density in the die before compaction and w is the moisture content (%) in the powder before
compaction. With a powder mass of 500 mg at the maximum capacity of the compression
die (height = 12 mm, diameter = 11 mm), the apparent density of the powder before
compaction was 438.6 kg/m3 . The mean value of apparent moisture density obtained from
the three measurements of moisture content was 20.85 ± 1.2% kg/m3 .
Pharmaceutics 2023, 15, x FOR PEER REVIEW 4 of 34
powder before compaction was 438.6 kg/m3. The mean value of apparent moisture density
obtained from the three measurements of moisture content was 20.85 ± 1.2% kg/m3.
2.2. Methods
2.2. Methods
2.2.1. Heat Capacity of the Material
2.2.1. Heat Capacity of the Material
The heat capacity (Cp) of the powder was determined with a differential scanning
The heat
calorimetry capacityNETZSCH
apparatus (Cp) of theDSC
powder was(NETZSCH-Gerätebau
404F3 determined with a differential scanning
GmbH, Wittels-
calorimetry apparatus NETZSCH DSC 404F3 (NETZSCH-Gerätebau GmbH, Wittelsbach-
bacherstraße 42, Selb, Germany) with platinum as the reference material. The test powder
erstraße
sample 42, Selb,
weighed 23.7Germany) with
mg, and the platinum as were
measurements the reference material.
taken under The test powder
a N2 atmosphere with
sample weighed 23.7 mg, and
the temperature ranging from 29 °C the measurements
to 130 °C andwere
a taken
heating under
rate ofa N
5 2 atmosphere
°C/min. The with
heat
the temperature ranging ◦
from 29 during
C to 130 ◦ C and ◦
a heating rate of 5 C/min. The heat
capacity in this study was constant powder compression.
capacity in this study was constant during powder compression.
2.2.2. Thermal Diffusivity and Conductivity
2.2.2. Thermal Diffusivity and Conductivity
The thermal diffusivity of MCC PH101 powder compacted at 200 MPa pressure with
The thermal diffusivity of MCC PH101 powder compacted at 200 MPa pressure with a
a 12-ton manual hydraulic press (Carver) was measured with a laser flash analysis appa-
12-ton manual hydraulic press (Carver) was measured with a laser flash analysis apparatus
ratus NETZSCH
NETZSCH LFA(NETZSCH-Gerätebau
LFA 457 457 (NETZSCH-Gerätebau GmbH,GmbH, Wittelsbacherstraße
Wittelsbacherstraße 42, Selb,
42, Selb, Germany).
Germany). The test sample thickness was 1.71 mm and the diameter
The test sample thickness was 1.71 mm and the diameter was 25 mm. Measurements was 25 mm. Meas-
were
urements werea taken
taken under under a Nat2 atmosphere
N2 atmosphere a temperature at ranging
a temperature
from 29ranging from
◦ C to 130 ◦ C 29
and°Ca to 130
heating
°Crate
andofa5heating
◦ C/min.rate of 5 °C/min.glass
A neodymium A neodymium
laser with aglass laser with
wavelength a wavelength
of 1054 nm was usedof 1054
with
nm was used with an energy of 10 J per pulse each 0.3 ms and a voltage
an energy of 10 J per pulse each 0.3 ms and a voltage of 2978 V. The measurement principleof 2978 V. The
measurement
is illustrated principle
on Figureis1.illustrated on Figure
The front surface 1. The
of the front surface
plane-parallel of thewas
sample plane-parallel
heated by a
sample
laser pulse, and the resulting temperature increase at the sample’s rear facethe
was heated by a laser pulse, and the resulting temperature increase at wassample’s
recorded
rear
as aface was recorded
function of time. asIn aa function of time. In
one-dimensional a one-dimensional
heat heat flow,was
flow, thermal diffusivity thermal dif-
calculated
fusivity was calculated with this temperature
with this temperature rise as follows [56]: rise as follows [56]:
𝑑 2
𝛼 = 0.1338 d (2)
𝑡.
α = 0.1338 (2)
t0.5
where α was thermal diffusivity in m2/s, d was sample thickness in m, and t0.5 was the half-
where α was thermal diffusivity in m 2 d was sample thickness in m, and t0.5 was the
time (time value at half-signal height) in /s,
seconds.
half-time (time value at half-signal height) in seconds.
Diagram
Figure1.1.Diagram
Figure describing
describing the
the laser
laser flashprinciple.
flash principle.
From the measured thermal diffusivity and heat capacity, thermal conductivity (k)
From the measured thermal diffusivity and heat capacity, thermal conductivity (k)
was calculated with the following equation:
was calculated with the following equation:
λ =λ 𝛼𝜌 𝐶 Cp
= αρ app (3)(3)
where
whereλλwas
wasthermal
thermalconductivity
conductivity(W/(m·K)),
(W/(m·K)),α αwas
wasthermal
thermaldiffusivity
diffusivity(m(m/s),
2 𝜌app
2 /s), was
ρapp was
the apparent density of the tablet at maximum compression (kg/m 3), and
3 C p was the heat
the apparent density of the tablet at maximum compression (kg/m ), and Cp was the
capacity (J/kg·K)(J/kg
heat capacity of the
·K) powder. The thermal
of the powder. conductivity
The thermal die tooldie
conductivity was obtained
tool from
was obtained
COMSOL Multiphysics ® 5.6 software
® (provided by CMC Microssytem,
from COMSOL Multiphysics 5.6 software (provided by CMC Microssytem, Pavillon Pavillon 1, 3000 1,
3000 boul. de l’Université, Sherbrooke, QC, J1K 0A5, Canada) [57]. Like heat capacity, the
thermal conductivity of the punches and die is a function of the local temperature during
powder compression.
2.2.3. Powder Retention/Sorption Curve

The isothermal moisture sorption or retention curve of MCC PH101 was obtained by
stocking 30 g of powder overnight under 25 ± 2%, 35 ± 2%, 50 ± 2%, 65 ± 2%, 80 ± 2%,
and 97 ± 2% relative humidity in a hermetically sealed box. The powder was collected after
12 h, when powder humidity was assumed to have reached an equilibrium state with the
surroundings. Powder moisture content was measured one second after it was collected
from the box. Three replicates of the powder sample were stored at each relative humidity.
2.2.4. Relative Tablet Density

The relative density of the tablet was calculated with the dimensions of the out-of-die
tablet at different rates of axial upper punch pressure:
ρtablet mtablet
RD calc = = (4)
ρ particle Vtablet ∗ ρ particle
where ρtablet was tablet density (kg/m3 ), ρparticle was the true powder particle density
(kg/m3 ), and mtablet and Vtablet were the tablet weight (kg) and volume (m3 ), respectively.
2.2.5. Tableting Process

The powder was compressed with a single-station tablet press (Manesty F3, Federal
Equipment Company, Cleveland, OH, USA). The press was operated at a compression
speed of 32 mm/s, measured with a proximity sensor (NPN 10 mm, Hall effect 3 wires
normally open, 5-24VCC, 200 mA, 320 kHz) coupled with a tachymeter. Compaction
pressure was measured with a calibrated load cell (MLC-10K, Transducer Techniques,
Temecula, CA, USA) fixed in a loge at the top of the upper punch. The set of punches and
die (Natoli Engineering Company, Inc., St. Charles, MI, USA) used were made of ERS
S7 steel and A7 steel, respectively, with a diameter of 10.97 ± 0.01 mm and an inner die
diameter of 11 ± 0.01 mm.
The maximum compression force of the tablet press was set at 410 MPa. The radial
pressure developed on the die wall during compaction was measured indirectly with a
linear strain gage (Micro Measurements, A VPG Brand, Raleigh, NC, USA) with a resistance
of 350 ± 0.3% ohm and a gage factor of 2.09. The strain gage was fixed at the lower middle of
the exterior wall of the die. Radial pressure was calculated from the Hoop stress developed
on the exterior surface of the die, which was obtained through the measured strain and
Young’s modulus of the die material. The data were acquired with a logger (DI-718B-US,
DataQ Instruments, Akron, OH, USA) coupled with amplification modules (8B38) for
strain gage and load cell. The sensors’ output was processed with free Windaq software in
Windows 10.
2.2.6. Tablet Surface and Peripheral

During compression, the temperature of the top and sides of the tablet was measured
one second after its ejection from the die cavity. An FLIR E4 thermal camera (Teledyne
FLIR LLC, Wilsonville, OR, USA) with automatic adjustable emissivity was used to detect
tablet temperature. The FLIR was manually held to focalize the pointer on the region of
interest. The distance between the FLIR and the ejected tablet was approximatively 20 cm,
which allowed more precise recording of the temperature, measured for different levels
of compaction, i.e., relative densities, for validation with the simulation results. For each
compaction level, an average temperature was calculated from 10 samples.
2.2.7. NIR-Penetration Depth

To estimate the NIR-radiation penetration depth in the tablet during scanning, an
MCC disk and regular paper were placed at the top of an acetaminophen (N-acetyl-para-
aminophenol or APAP) tablet. The regular paper disks were used because of their chemical
similarity with MCCs and to simulate thinner thickness. All the samples had the same
2.2.7. NIR-Penetration Depth
To estimate the NIR-radiation penetration depth in the tablet during scanning, an
Pharmaceutics 2023, 15, 1652 MCC disk and regular paper were placed at the top of an acetaminophen (N-acetyl-para- 6 of 33
aminophenol or APAP) tablet. The regular paper disks were used because of their chem-
ical similarity with MCCs and to simulate thinner thickness. All the samples had the same
diameter—20
diameter—20 mm.mm. The
The regular
regular paper
paper hashas aa thickness
thickness of of 74
74 microns
microns and
and porosity
porosity of
of 57%.
57%.
The MCC disk was obtained from pressed MCC powder at a pressure of 220 MPa and had
The MCC disk was obtained from pressed MCC powder at a pressure of 220 MPa and had
aa thickness
thickness of
of 0.45
0.45 mm
mm and
and aa porosity
porosity of
of 38%.
38%. The
The porosity
porosity (ε)
(ε) of
of the
the paper
paper and
and MCC
MCC diskdisk
was estimated
was estimated from
from Equation
equation (4)
4 (ε(ε==11 −−RD
RDcalc). APAP powder was supplied by Sigma-
calc ). APAP powder was supplied by Sigma-
Aldrich Canada Co. (2149 Winston Park Drive Oakville, ON L6H 6J8, Canada). Figure 2
shows the MCC disk (Figure 2a) and one (Figure 2b) and five paper sheets (Figure 2c) at
the top of the APAP
APAP tablet.
tablet.
(a) (b) (c)

Figure
Figure 2.2. NIR-radiation
NIR-radiation penetration
penetration depth measurement setup.
depth measurement setup. (a)
(a) MCC
MCC disk on acetaminophen
disk on acetaminophen
tablet, (b) one sheet and (c) 5 sheets of regular paper on acetaminophen tablet.
tablet, (b) one sheet and (c) 5 sheets of regular paper on acetaminophen tablet.
Primarily, the APAP

Primarily, the APAP tablet,
tablet, MCC
MCC disk,
disk, and
and regular
regular paper
paper were
were scanned
scanned individually
individually
to
to register
register their
their respective spectra. Then,
respective spectra. Then, disks
disks ofof increasing
increasing thickness
thickness were
were stacked
stacked at
at the
the
top of the APAP tablet
tablet until
until the
the corresponding
corresponding fingerprint (absorbance from functional
group in a specific region) of acetaminophen vanished completely. completely. The thickness at which
APAP was not detected was chosen as the maximum
APAP was not detected was chosen as the maximum penetration depth penetration depth of
of NIR
NIR radiation.
radiation.
The regular
regular paper
paperwaswasused
usedatatfirst
firsttotofind thethe
find maximum
maximum penetration
penetrationdepth andand
depth re-
placed
replacedbybythethe
MCC MCC disks to verify
disks the results.
to verify The setup
the results. for thefor
The setup measurement was main-
the measurement was
tained to avoid
maintained variation
to avoid in the
variation height
in the heightat which
at which the sample
the samplewaswasscanned.
scanned.The
The humidity
humidity
and temperature in the lab were also controlled.
controlled.
2.2.8. Determination
2.2.8. Determination of of Relative
Relative Tablet
Tablet Moisture
Moisture Content
Content Density
Density
The tablet
The tablet surface
surface moisture
moisturewas wasmeasured
measuredwith withananNIR
NIRsensor
sensorin in reflection
reflection modemode at
at a wavelength of 1100–1700 nm. A calibration model was developed
a wavelength of 1100–1700 nm. A calibration model was developed with manually com- with manually
compacted
pacted tablets.
tablets. Then, Then, the tablets
the tablets pressed
pressed withwith the Manesty
the Manesty single
single punchpunch
werewere scanned
scanned in-
inline to predict the tablet surface moisture
line to predict the tablet surface moisture content. content.
To create
To createthe
thecalibration
calibrationmodel,
model,thethe
powder
powder was spread
was in an
spread inaluminum
an aluminum foil container
foil con-
and equilibrated overnight at 25 ± 2%, 35 ± 2%, 50 ± 2%, 65 ± 2%, 80
tainer and equilibrated overnight at 25 ± 2%., 35 ± 2%, 50 ± 2%, 65 ± 2%, 80 ± 2%, ± 2%, 97 ±972%
and and ±
relative humidity and room temperature in a humidity-controlled chamber;
2% relative humidity and room temperature in a humidity-controlled chamber; 30 g of the 30 g of the
powder
powder was was used
used for
for each
each relative
relative humidity
humidity level. After 12
level. After 12 h,
h, the
the powder
powder was was removed
removed
from the humidity-controlled chamber, poured into a small, closed plastic
from the humidity-controlled chamber, poured into a small, closed plastic box, and box, and stirred
stirred
manually for 30 s. Powder (3 g) was then sampled from the plastic box for water content
measurement, and the rest of the batch was used to manually produce 5 tablets with
a lower compaction level (thickness: 6.5 ± 0.8 mm), each weighing 0.5 g. Each tablet
was individually scanned five times (25 scans for each relative humidity level). Scans of
30 tablets in total were used for the calibration. The tablets were made in an 11 mm die and
pressed with a punch by hand to apply very low pressure, since high pressure can induce a
rise in powder particle temperature. All steps in the scan collection phase were performed
on the compression plate to mimic the conditions during the tableting process.
The measured water content from the conditioned powder used to make the tablets
manually was chosen as a reference value of these tablets’ surface moisture content. The
moisture content of the manually produced tablet was considered to be homogeneous
throughout the tablet and equal to the moisture content of powder from which it was made.
The time between powder sampling from the plastic box and tablet manual compression
followed by the scan with NIR sensor was less than 5 s, and hence the moisture content
was expected to be preserved. The powder moisture measurement and the tablet manual
compression were performed simultaneously. For the calibration model, the moisture
content was selected to range between 4.2% and 14%. This range of moisture content was
chosen to build a robust model that could predict a high moisture content with minimum
error. In fact, the temperature of the tablet was shown to increase during the compression;
therefore, it induced the migration of water molecules in the powder bed to the external
surface of the tablet.
The raw spectra collected from the tablet presented noise due to interference and
baseline shift due to multiplicative and additive effects caused by the inhomogeneity of
the tablet microstructure. All spectra (Figure 3a) are plotted versus the spectral mean,
as shown in Figure 3b. The red line represents the diagonal introduced for comparison,
the slope of the lines (formed by scatter points) and the intersection with the vertical axis
(absorbance) represent the multiplicative and additive effects [58,59]. The lines that have
the same slope and same intersection (with the absorbance axis) with the diagonal red have
has insignificant multiplicative and additive effects, respectively. in Figure 3b, most lines
(scatter points) have the same slope as the diagonal and a non-zero intersection with the
absorbance axis. This suggests that some spectra have been affected by the multiplicative
effect and most of the spectra are affected by the additive effect. To remove these artefacts,
different mathematical preprocess functions, such as multiplicative scatter correction (MSC)
and its extended version (EMSC), standard normal variance (SNV), Savitzky–Golay (SG)
filter, detrend, and first and second derivative, can be used to treat the spectra [60]. To select
the most appropriate preprocessing method for the present data, the pretreated spectra
were plotted against the mean spectra and compared with the diagonal line. When the
multiplicative and additive effects are removed from the raw spectra, the plotted lines (line
formed by scatter points) will align with the diagonal line. The preprocess technic that
allows the plotted lines to fit best with the diagonal line was selected as the best-performing
technic. In this study, the combination of EMSC and SG (17, 1, 0) was selected, and the
results are shown in Figure 3c. The pretreated spectra are shown in Figure 3d.
After pretreatment, Hotelling’s T2 in principal component analysis (PCA) [61] with
a 95% confidence interval was applied to the data to visualize them and remove outliers.
The PLS regression method [62] was used to establish the calibration model with an
optimal three principal components. K-fold cross-validation was performed on the data.
K-fold cross-validation is an external validation or prediction technic that divides the data
into six groups, with each group containing the data of each moisture content level. A
calibration model is established with five groups systematically selected and used to predict
the sixth group. The process is repeated in a loop until all the groups are predicted by the
respective calibration models. At the end of the process, the model performance [54] in
calibration and prediction was measured with the regression coefficient (R2 ) and root-mean-
square error (RMSE) [54]. A mean R2 and RMSE in calibration and prediction is calculated
from the six results.
After the calibration model had been computed, the tablets were pressed with com-
pression pressure of 37.5 MPa (thickness of 4.83 ± 0.2 mm) and scanned directly on the
tablet press table. Figure 4 shows the steps in the process of inline scan collection on
the press. First, the pressed and ejected tablet from the die was pushed by the scraper
(Figure 4a) and then the tablet was manually guided (Figure 4b) under the sensor to take

measurements (Figure 4c). The entire process took less than 2 s; therefore, the tablet surface
moisture was assumed to have been preserved.
(a) (b)
(c) (d)
Figure 3. NIR (a) raw spectra vs. wavelength, (b) individual raw spectra vs. the mean absorbance of
Figure 3. NIR (a) raw spectra vs. wavelength, (b) individual raw spectra vs. the mean absorbance
Pharmaceutics 2023, 15, x FOR PEER REVIEW 9 of
all spectra before and (c) after pretreatment, and (d) spectra after pretreatment vs. wavelength of 34
of all spectra before and (c) after pretreatment, and (d) spectra after pretreatment vs. wavelength of
MCC VIVAPUR PH101 tablets.
After pretreatment, Hotelling’s T2 in principal component analysis (PCA) [61] with a

95% confidence interval was applied to the data to visualize them and remove outliers.
The PLS regression method [62] was used to establish the calibration model with an opti-
mal three principal components. K-fold cross-validation was performed on the data. K-
fold cross-validation is an external validation or prediction technic that divides the data
into six groups, with each group containing the data of each moisture content level. A
calibration model is established with five groups systematically selected and used to pre-
dict the sixth group. The process is repeated in a loop until all the groups are predicted by
the respective calibration models. At the end of the process, the model performance [54]
in calibration and prediction was measured with the regression coefficient (R2) and root-
mean-square error (RMSE) [54]. A mean R2 and RMSE in calibration and prediction is
calculated from the six results.
After the calibration model had been computed, the tablets were pressed with com-
pression pressure of 37.5 MPa (thickness of 4.83 ± 0.2 mm) and scanned directly on the
tablet press table. Figure 4 shows the steps in the process of inline scan collection on the
Figure
Figure4.
press. 4. The
First, steps
Thethe
steps in tablet
pressed
in tablet
andsurface
ejected
surface moisture content
tabletcontent
moisture from themeasurement.
die was pushed
measurement. (a) by
(a) The The tablet
the
tablet is ejected
isscraper
ejected from
(Figure
from the
the
4a) die,
die, and (b)
then
(b) the the operator
the tablet
operator pushedpushed
was the
themanually tablet manually
guided
tablet manually under
(Figure
under the NIR
4b) sensor
the NIR under and sensor
the (c)
sensorand (c) the
to take
the tablet meas- is
tablet
is scanned
scanned with the NIR sensor. A video of the measurement can be seen in Supplementary Material
urements
with the NIR (Figure
sensor.4c). The of
A video entire process took
the measurement less
can thanin2 Supplementary
be seen s; therefore, the tabletS1.
Material surface
S1.
moisture was assumed to have been preserved.
2.2.9. Tablet Cohesion and Internal Friction Angle Density
Tablet cohesion and internal friction angle were determined with diametral and axial
compression tests. The tablet’s diametral and axial breaking forces were measured with a
tablet hardness tester (C50, Whitehall International, New Lane, Havant, UK) and a man-
ual press (Carver Laboratory Press, model 12 Ton, Fred S. Carver Inc., Wabash, IN, USA).
2.2.9. Tablet Cohesion and Internal Friction Angle Density

Tablet cohesion and internal friction angle were determined with diametral and axial
compression tests. The tablet’s diametral and axial breaking forces were measured with
a tablet hardness tester (C50, Whitehall International, New Lane, Havant, UK) and a
manual press (Carver Laboratory Press, model 12 Ton, Fred S. Carver Inc., Wabash, IN,
USA). Tablets compacted at different relative densities between 40% and 95% were used
for the measurements. Tablets with a relative density < 40% were not hard enough for
the tests. The hardness tester crushed the tablets along their diametral/radial line and
displayed the maximum breaking diametral force (Fd ). On the manual press, the tablets
were pressed between two horizontal plates until they were crushed. The breaking axial
compressive force (Fc) was obtained from the pressure applied and the tablet surface.
Equations
(5) and (6) were used to calculate
the breaking radial/diametral tensile strength
f f
σd and axial compressive strength σc of the tablets [12,16]:
f 2Fd
σd = (5)
πDt
4F c f
σc = (6)
πD2
where D was tablet diameter (mm) and t was tablet thickness (mm).
The breaking force thus determined was used to calculate tablet cohesion (d) and
internal friction angle (β) with Equations (7) and (8) [12,16]:
f f √

σc σd 13 − 2
d= f f
(7)
σc + 2σd

f
σc + d
β = tan−1 [3 f
] (8)
σc
3. Numerical Method
The aim of the numerical analysis was to simulate the mechanical, heat, and moisture
behaviors of the powder bed during the compaction phase. The simulation was based on
coupling a solid mechanical module with the heat and moisture transfer in the building ma-
terials module implemented in the finite element analysis solver COMSOL Multiphysics® .
3.1. Simulation Model for the Mechanical Behavior of the Powder

3.1.1. Density-Dependent Drucker–Prager Cap Model
The density-dependent DPC model [12] has been used in several studies, which found
high agreement between simulated and experimental measurements. The model consists
of three surfaces (Figure 5a):
- The Mohr–Coulomb shear failure surface (Fs ), representing shear flow, defined as
Fs = q − p tan β − d (9)
- The transition surface (Ft ), representing a mathematical smoothing surface, defined as
q
Ft = ( p − Pa )2 + [q − (1 − cosα( β) ((d + Pa ∗ tan β))]2
(10)
−α(d + Pa ∗ tan β) = 0
E and Poisson’s ratio ν can be determined [11,64]:
𝐸= , (26)
3𝐾 − 2𝐺
Pharmaceutics 2023, 15, 1652 𝜈= 10 of(27)
33
2(3𝐾 + 𝐺)
(a) (b)
Figure 5. The Drucker–Prager cap model. Yield surface in (p, q) plane: (a) from [12]; (b) implemented
Figure 5. The Drucker–Prager cap model. Yield surface in (p, q) plane: (a) from [12]; (b) implemented
in COMSOL Multiphysics ® [63].
in COMSOL Multiphysics® [63].
- 3.2. The
Moisture and Heat
cap yield Transport
surface Simulation
(Fc ) defined as Model
During the tableting
v process, heat is generated due to irreversible deformation and
friction phenomena.u The heat diffuses through
Rq the
2 particles and increases the temperature
Pa )2 +the
u
of the entire powder (p −
Fc = tbed during [ compaction − R(dThe
α ] phase. ∗ tan β) of
+ Pabehavior = the
0 inherent mois-
(11)
1 + α − cos β
ture of the powder particles remains unknown. Moisture content distribution and evolu-
tion can
where β isbethe
investigated using the
material friction heat(◦and
angle ), d moisture transport
is its cohesion (Pa),inbuilding
and p andmaterials mod-
q are hydro-
ule inpressure
static COMSOL. stress (Pa) and von Mises equivalent stress (Pa), respectively, which are
The as
expressed heat, air, and moisture (HAM) model is widely used for porous materials [65,66]
and granular materials [67]. It combines 1 I
heat transfer, air diffusion, and moisture
p = (σz + 2σr ) = 1 (12)
transport through porous materials. The 3 heat generated 3 diffuses in the powder by a com-
bination of conduction (through the bulk particles) r and convection (through interparticle
air) and through the tableting tools 3 powder bed is a granular medium
q = |σz − σr | = − The
by conduction. J2 (13)
with interparticle pores. Moisture can exist in pores2in two thermodynamic states of mat-
ter—vapor
where σr andand liquid
σz are [66].
punch The and
stress maindie mechanisms associated
stress, respectively, andtoI1moisture transfer
and J2 are are
the first
stress invariant and second deviatoric stress invariant, respectively.
The parameters needed for the elliptic cap are:
- The cap evolution parameter, Pa, that represents the volumetric plastic strain-driven
hardening/softening, defined as
[3q B + 4d tan β∗ f 2 ]
Pa = −
4 [ ( f )2 ]
q (14)
9q2B + 24dq B (f)2 tan β + 8(3p B q B + 2q2B )f2
+ 4[( f )2]
where f = (1 + α − α/cosβ) and α is a small number (typically 0.01–0.05) used to define a

smooth transition surface between the shear failure surface and the elliptic cap. PB is the
pressure at maximum compression.
- The cap eccentricity, R, is a material parameter between 0.0001 and 1000 that controls
the shape of the cap and is defined as
s
2(f)2
R= ( PB − Pa ) (15)
3q B
- The hardening/softening law is a user-defined piecewise linear function relating the
hydrostatic compression yield stress, Pb , and the corresponding volumetric inelas-
p
tic/plastic strain, ε v :

p
Pb = f ε v = Pa + R(d + Pa tan( β)) (16)
p RD ρ
Volumetric plastic strain can be expressed as ε v = ln( RD 0
), where RD = ρbulk true
is the
current relative density and RD0 is the initial relative density. ρbulk is the tablet’s apparent
density (kg/m3 ) and ρtrue is particle true density (kg/m3 ).
3.1.2. Density-Dependent Drucker–Prager Cap Model Implemented in COMSOL Multiphyisics®

In the present study, the DPC model version of COMSOL Multiphysics® , which uses
two yield surfaces, was used [63]. In COMSOL, the Drucker–Prager cone (shear failure
surface) always meets the yield cap surface at the point of tangency. Hence, no transition
zone is required between the shear failure cone and the elliptical cap, so there is a unique
and smooth transition between the two surfaces (Figure 5b). The surfaces were defined
based on the first stress invariant I1 and the second deviatoric stress invariant J2 as follows.
- The Drucker–Prager yield function, Fs:
Fs = J2 − α0 I1 − k
p
(17)
√ √
where α0 = tan(β)/(3 3) and k = d/( 3) are the Drucker–Prager parameters, β is the
internal friction angle, and d is cohesion.
- The elliptical cap surface, Fc :
√ 2
I1 − Ia 2 J
Fc = ( ) +( 2) −1 (18)
Ib − Ia Ja
√
where Ja is the ordinate on the J2 axis at I1 = Ia .
Here, the hardening law is defined as:
ε pvol
p0 b = pb0 − Kiso log(1 + ) (19)
ε pvol,max
where pb0 is the initial location of the cap, Kiso is the isotropic hardening modulus, and
εpvol,max is the maximum volumetric plastic strain.
The cap eccentricity for R0 i given by:
Ia − Ib √
R0 = = 3 3 ∗ R, (20)
Ja
where Ib = −3p0 b .
The hardening law in the density-dependent DPC model and that in COMSOL are
different; therefore, the parameters Kiso and εpvol,max were chosen in such a way that they
approximatively fitted the hardening law curve obtained from the calibrated DPC model.
The important permanent deformation of the particles during compression was considered
and a large plastic strain plasticity model was used. Hence, the relationship between
current relative density (RD) and volumetric plastic strain is given by:
RD = RD0 ∗ JP−1 (21)
where Jp is the plastic volume ratio.
3.1.3. Nonlinear Elastic Law

The unloading portion of the compression curve of the non-lubricated pharmaceutical
powder shows nonlinear behavior, which can be attributed to the dilatation of the tablet
during the unloading phase and is described using a nonlinear elasticity law. The density-
dependent elastic modules that characterize elastic behavior, such as the bulk modulus K
and the shear modulus G, are expressed in such a way as to avoid path dependence [12]:
K = K ( p, ρ) = K ( I1 , ρ) (22)
p
G = G (q, ρ) = G ( J2 , ρ) (23)
√
where I1 is the first stress invariant, J2 is the stress invariant, and ρ is the compact apparent
density. When considering the linear portions of the unloading curve, the portion between
the maximum compression stress at point B localized on the cap surface and point C where
the hydrostatic state is reached, i.e., σr = σz , can be used to determine the elastic modulus
using two equations [11].
σzB + σzC 4
B C
=K+ G (24)
εz + εz 3
qB 2G
B C
= √ (25)
P +P 3K
With the elastic modulus thus determined, the density-dependent Young’s modulus E
and Poisson’s ratio ν can be determined [11,64]:
9KG
E= , (26)
3K + G
3K − 2G
ν= (27)
2(3K + G )
3.2. Moisture and Heat Transport Simulation Model

During the tableting process, heat is generated due to irreversible deformation and
friction phenomena. The heat diffuses through the particles and increases the temperature
of the entire powder bed during the compaction phase. The behavior of the inherent
moisture of the powder particles remains unknown. Moisture content distribution and
evolution can be investigated using the heat and moisture transport inbuilding materials
module in COMSOL.
The heat, air, and moisture (HAM) model is widely used for porous materials [65,66]
and granular materials [67]. It combines heat transfer, air diffusion, and moisture transport
through porous materials. The heat generated diffuses in the powder by a combination
of conduction (through the bulk particles) and convection (through interparticle air) and
through the tableting tools by conduction. The powder bed is a granular medium with in-
terparticle pores. Moisture can exist in pores in two thermodynamic states of matter—vapor
and liquid [66]. The main mechanisms associated to moisture transfer are vapor diffusion
or capillary suction, or both. The material has a unique equilibrium moisture content
characteristic curve that covers the hygroscopic and capillary water regions. These regions
are commonly referred to as sorption isotherm and retention curves. Inside the hygroscopic
region, the pores are mainly filled with water vapor, and consequently, moisture transport
is mainly achieved by vapor diffusion. Liquid water transport is possible when pores are
filled with liquid water. This flow mechanism is very active in the capillary water region,
where the relative humidity is over 95%. Both vapor and liquid transport can coexist at
the higher end of the hygroscopic region, where vapor diffusion and capillary suction are
active in large and small pores, respectively. Water vapor diffuses into open pores and
condenses on the capillary meniscus, whereas water evaporates into the next open pore
space at the other end of the meniscus. This implies that the diffusion path is reduced,
thereby increasing the moisture evaporation rate.
The moisture balance can be expressed using several thermodynamic potentials de-
pendent on transport–diffusion partial differential equations. Relative humidity and tem-
perature are the dependent variables in the mathematical model. The equation that governs
moisture transfer is as follows [57,66]:
ξ∂ϕ
+ ∇ · ξDw ∇ ϕ − δp ∇( ϕpsat ( T )) = G (28)
∂t
It represents the sum of vapor diffusion flux and capillary moisture flux. Their
respective representative equations are:
g1 = −δp ∇( ϕpsat ( T )) (29)
g2 = ξDw ∇ ϕ (30)
where ϕ is relative humidity (%), δp is vapor permeability (m2 /s), psat is saturation pressure
(Pa), w is moisture capacity (kg/kg), Dw is moisture diffusivity (m/s2 ), G is the generic
moisture source (kg/m3 ), and T is temperature (K).
The moisture source is water molecules adsorbed on the powder surface and pores
that evaporated during the tableting process. The coupled thermal analysis is solved by
the following form of the energy balance equation [20,66]:
∂T
ρC p ef f
+ ∇ · −λe f f ∇ T − Lv δ p ∇( ϕpsat ( T )) = Q (31)
∂t
where ρ is effective density (kg/m3 ), Cp is specific heat at constant pressure (J/(kg·K)), λ is

thermal conductivity (W/(m·K)), Lv (J/kg) is heat of evaporation, and Q is the heat source
(W/m3 ). The energy balance is applied under the assumption of local thermodynamic
equilibrium (LTE). In the first term of Equation (31), effective thermal capacity is defined
as follows [20]:
ρC p e f f = ρs C p,s + wC p,w (32)

where ρC p e f f is effective volumetric heat capacity at constant pressure, defined to account
for both solid matrix and moisture properties; C p,s and ρs are the density and specific heat
capacity of the dry solid; w is water content given by the moisture storage function, i.e.,
the moisture sorption curve for the chosen material; C p,w is water heat capacity at constant
pressure; and λe f f is the effective thermal conductivity function of the solid matrix and
moisture properties [20,66]:
bw
λ e f f = λ s (1 + ) (33)
ρs
where λs is the thermal conductivity W/(m·K) of the dry solid and b (dimensionless) is a
thermal conductivity supplement due to the water content. Equation (31) is composed of
two heat sources: sensible heat due to moisture content variation in the powder, expressed
as vapor diffusion flow multiplied by latent heat of evaporation (Lv ); and the heat source
caused by irreversible deformation, labeled as Q. The heat source Q relates to plastic strains,
stress, interparticle friction, and friction with die walls, and is given as [18,20]:
Q = q1 + q2 (34)
where q1 is the heat generated by interparticle friction and permanent deformation, and
q2 is the heat generated by particle–wall friction. Ninety percent of the energy needed to
permanently deform the particles is assumed to dissipate as heat in the powder bed [18,20].
q1 is defined as
3
∑
pl
q1 = ξ σij dεij (35)
i,j=1
where ξ = 0.9 represents the irreversibly deformed heat fraction [18,20].

The heat generated by friction is defined as
q2 = υµS A σnn dA (36)
where µ is the wall friction coefficient (-), A represents the interacting area between the
powder and die wall (m2 ), σnn is the local normal stress (Pa), and υ is the norm of the
local slip velocity at the interface between the powder and the die (m/s). Friction heat
is an interface heat between the die wall and the powder compact. The portion of heat
transferred to the powder compact can be calculated with the weighting factor proposed
in [68–70] and adopted by [20]. Hence, friction heat can be separated into two parts:
q2die = η ∗ q2 , (37)
q2tablet = (1 − η) ∗ q2 , with θ = (1 − η) (38)

where η is the weighting factor, which is estimated by Equation (39):
1
η= 3∗ λ q , (39)
powder
1+ 2∗λdie ∗ α αdie
powder
where λpowder , λdie , αdie , and αpowder are powder and die thermal conductivities and powder
and die thermal diffusivities, respectively.
3.3. Method of Calculation of Elastic Parameters and Drucker–Prager Cap Model Parameters
The calibration procedure used in this study to determine the DPC model parameters
and elastic parameters is described by [12]. Elastic parameters, such as Young’s modulus
and Poisson coefficient, were determined from the unloading curve. The DPC model
parameters, cohesion (d) and internal friction angle (β) were first obtained with diametral
and axial compressive tests, then cap eccentricity (R) and hydrostatic compression yield
stress (Pb ) were calculated with the d and β obtained.
3.4. Finite Element Model

The FEM simulations of the compaction phase of the tableting process were performed
on the commercial COMSOL Multiphysics® 5.6 finite element analysis solver, which imple-
mented the DPC plasticity model and HAM model. MCC VIVAPUR PH101 was used as
the model powder. A flat-faced, cylindrical compact geometry was modeled with the com-
pact/powder represented by a four-node, 2D axisymmetric model, as shown in Figure 6.
The powder was modeled as continuum, isotropic, poroplastic, and deformable media
in which the pores were uniformly distributed. The flat-faced upper/lower punches and
the cylindrical die were modeled as analytical rigid bodies without any deformation. The
slip velocity boundary condition between the powder and the die wall was set to simulate
the rate at which compression occurred. The finite element simulations were modeled as
quasi-static; therefore, punch velocity and dwell time were disregarded. Only the upper
punch was set to move downward to simulate one side compression. The initial powder fill
height was 12 mm in a die with a radius of 5.5 mm, and the powder was compacted with
the upper punch to a final height of 4.5 mm during the compaction stage. In the analysis,
only the compaction phase of the quasi-static compaction process was simulated. The wall
friction effect was considered through the adoption of a Coulombic boundary condition at
the powder–die wall and powder–punch interfaces. Since MCC PH101 was not lubricated,
a static Coulombic friction of 0.21 was adopted from [12].
was set to move downward to simulate one side compression. The initial powder fill height
was 12 mm in a die with a radius of 5.5 mm, and the powder was compacted with the upper
punch to a final height of 4.5 mm during the compaction stage. In the analysis, only the
compaction phase of the quasi-static compaction process was simulated. The wall friction
effect was considered through the adoption of a Coulombic boundary condition at the pow-
Pharmaceutics 2023, 15, 1652 der–die wall and powder–punch interfaces. Since MCC PH101 was not lubricated, a 15 of 33
static
Coulombic friction of 0.21 was adopted from [12].
Figure 6. The finite element model of die compaction using flat-faced punches.
Figure 6. The finite element model of die compaction using flat-faced punches.
Our simulations used a constant cap eccentricity (R0 ) and a density-dependent constitu-
Our simulations used a constant cap eccentricity (R′) and a density-dependent con-
tive model parameter (cohesion [d], internal friction angle [β], Drucker–Prager parameters
stitutive model parameter (cohesion [d], internal friction angle [β], Drucker–Prager pa-
[α0 , k], or hydrostatic yield compression stress [Pb 0 ]). The mechanical properties of the
rameters
powder as [α′,a k], or hydrostatic
function of relativeyield compression
density stress [Pinb′]).
were expressed theThe mechanical
range of 40% toproperties
80% com-
of the powder as a function of relative density were expressed
paction and are summarized in Supplementary Material S2. A piecewise function in the range of 40% to 80%
was used
compaction and are summarized in Supplementary
in COMSOL Multiphysics® to implement all mechanical parameters. Material S2. A piecewise function was
used in COMSOL
Yield Multiphysics
compression stress (Pto
® 0 implement all mechanical parameters. 0
b ), i.e., the hardening law, and cap eccentricity (R ) in
Yield compression stress (P b′), i.e., the hardening law, and cap eccentricity (R′) in the
the COMSOL DPC model were calculated with Equations (20) and (21), respectively.
COMSOL DPChardening
The isotropic model were calculated
modulus Kisowith
wasEquations
determined (20)
(a)and (21), respectively.
to relatively The iso-
approximate the
tropic hardening modulus Kiso was determined (a) to relatively approximate the experi-
experimental calibrated DPC model’s hydrostatic yield compression stress (Pb ) (Figure 7);
mental
and (b)calibrated DPC model’s
to ensure consistent hydrostatic
powder hardening yield compression
during stress (P
the compaction b) (Figure
phase. 7); and
The point at
(b)
P b
to
= ensure
367 MPa consistent
was powder
separated hardening
from the rest during
of the the compaction
points (Figure 7) phase.
because The
thepoint at
powder’s P b
=incompressible
367 MPa was separated from the For
state was reached. restPof
b
the
lowerpoints
than (Figure
367 MPa, 7) because
relatively the powder’s
good agreementin-
compressible state wasthe
was reached between reached. For Pb lower
experimentally than 367
calibrated pointsMPa,andrelatively good points.
the calculated agreement
was reached between the experimentally calibrated points and the calculated points.
Figure
Figure7.
7.Hydrostatic
Hydrostaticyield
yieldstress,
stress, PPbbvs.
vs.plastic
plasticvolumetric
volumetricstrain.
strain.
The heat was set to dissipate from the system through a heat flux at the interface be-
tween the die wall and the upper and lower punches. At the powder–die wall interface, a
boundary heat source was set with a weighting factor function of the relative density of the
powder bed. The weighting factor was calculated with equation 39, assuming that the ther-
mal conductivity coefficient of the powder bed and the thermal conductivity and diffusivity
Figure 7. Hydrostatic yield stress, Pb vs. plastic volumetric strain.
The
The heat was set
heat was set to
todissipate
dissipatefrom
fromthethesystem
system through
through a heat
a heat flux
flux at the
at the interface
interface be-
between the die wall and the upper and lower punches. At the powder–die
tween the die wall and the upper and lower punches. At the powder–die wall interface, a wall interface,
aboundary
boundary heat
heat source
source was was
set set with
with a weighting
a weighting factor
factor function
function of theofrelative
the relative
density density
of the
of the powder bed. The weighting factor was calculated with Equation
powder bed. The weighting factor was calculated with equation 39, assuming that the ther- (39), assuming
that the thermal conductivity
mal conductivity coefficient of coefficient
the powderofbed theand
powder bed and
the thermal the thermaland
conductivity conductivity
diffusivity
and diffusivity of the die were constant during compaction. From the weighting factor
of the die were constant during compaction. From the weighting factor obtained, the portion
obtained, the portion (θ) of the interfacial frictional heat that transfers to the powder
(θ) of the interfacial frictional heat that transfers to the powder compact by conduction at
compact by conduction at each level of compaction is shown in Figure 8. As the relative
each level of compaction is shown in Figure 8. As the relative density of the powder bed
density of the powder bed increases, the portion of heat that diffuses into the powder bed
increases, the portion of heat that diffuses into theλ powder bed also increases. The thermal
also increases. The thermal diffusivity, α powder = powder ρC p , of the powder bed (where λ ,
diffusivity, 𝛼 = , of the powder bed (where λpowder, 𝜌, and Cp are thermalpowder con-
ρ, and Cp are thermal conductivity, apparent density, and heat capacity, respectively, of
ductivity, apparent density, and heat capacity, respectively, of the powder bed) decreases
the powder bed) decreases when apparent density increases. Hence, the weighting factor,
when
η, whichapparent density increases.
is proportional Hence,
to the square theofweighting
root the thermal factor, Ƞ, which
diffusivity ofisthe
proportional
powder bed, to
the square root of the thermal diffusivity of the powder bed, decreases
decreases during compaction and θ increases (Figure 8). The fraction of irreversible work during compaction
and θ increases
converted (Figure
into heat has 8).
beenTheestimated
fraction of
to irreversible
be in the rangeworkofconverted
80–100% into heatHence,
[20,69]. has been
in
this study, 20% of the input mechanical energy inside the powder bed was assumed tome-
estimated to be in the range of 80–100% [20,69]. Hence, in this study, 20% of the input be
chanical
stored as energy inside
irreversible the powder
particle bed was
deformation andassumed
80% wastoassumed
be stored as converted
to be irreversibleintoparticle
heat.
deformation and 80% was assumed to be converted − 4into
◦ heat.
−1 The
The thermal expansion coefficient, α T = 1.0 × 10 [ C ], of the powder was adopted thermal expansion coef-
ficient,
from 𝛼 = 1.0 × 10 [°C ], of the powder was adopted from [18].
[18].
Figure 8.
Figure 8. Change in weighting factor,
factor, θ,
θ,during
duringcompaction.
compaction.
The coefficient, U, for heat transfer between the compression tools and powder com-
pact was determined with the thermal resistance method [71]. The heat transfer between the
powder and punch, Upowder-die , and between the powder and die wall, Upowder-punch , were
equal to 32 W/(m2 .K) and 18 W/(m2 .K), respectively (See Supplementary Material S3).
In the proposed model, the moisture was set to flow toward the external environment
from the powder compact through the upper punch interface. The powder compact–die
wall and powder compact–lower punch interfaces were considered closed, so no moisture
could evaporate from those interfaces to the surrounding environment. The experimental
powder moisture sorption or moisture retention curve was used in the model, as the
moisture content of the powder bed is a function of the relative humidity of the powder
during compaction. The moisture transport coefficients of MCC Vivapur PH101 used in
the HAM model are given in our recent article [72].
The initial temperature of the powder was set to ambient—21.4 ± 0.2 ◦ C. The moisture
content of the powder was set at 4.5%.
4. Results and Discussion

4.1. NIR Penetration Depth
This work aimed to study the surface moisture of pharmaceutical tablets; therefore,
the penetration depth of the NIR radiation into a tablet was estimated. The NIR sensor
was used in diffuse reflectance measurement mode. Figure 9 shows the SNV + SG (17, 2, 0)
transformed NIR spectra of the APAP tablet, MCC disk, regular paper, and NIR spectra
of the MCC disk and sheet of regular paper superimposed on the APAP tablet surface.
Pretreatment was applied to remove additive and multiplicative effects and noise from the
spectra. The aim of adding several thin layers at the top of the APAP tablet was to determine
the critical thickness at which the APAP tablet cannot be detected by NIR radiation.
APAP absorbed NIR radiation at a wavelength of 1135 nm due to the C-H stretching
second overtone in the aromatic ring [73]. Cellulose (MCC and regular paper) absorbed
NIR radiation at 1220 nm due to hydrogen bond bending and at 1450 nm due to the
first overtone of O-H stretching [73]. The absorption peak of APAP at 1135 nm vanished
when five sheets of regular paper was superimposed at the top of APAP tablet, which
corresponds to a thickness of 0.33 mm or 330 µm (Figure 9a, pink spectrum). Therefore,
the NIR-radiation maximum penetration depth can be estimated to be less than 330 µm
when a pressed cellulose such as regular paper is scanned. To confirm the results, an MCC
disk of 0.45 mm (38% of porosity and water content of 4.25%) was placed at the top of the
APAP tablet and the APAP absorption peak at 1135 nm was not detected (Figure 9b, green
spectrum). The small air gap between the APAP tablet and the paper sheets or the MCC
disk did not affect the NIR signal because the tests were done in dry air (RH = 27.4%).
The increase in moisture content in granular material was found to decrease the
reflectance intensity of NIR light [74], because as the moisture content increases, more
light is absorbed by the water molecules. The absorption of the incident NIR light by
the water molecules at the surface of the scanned sample reduces its incident intensity,
thereby decreasing its penetration depth into the sample according to the Beer–Lambert
law [75]. Based on these results, the use of NIR in diffuse reflectance mode in this study
Pharmaceutics 2023, 15, x FOR PEERcan be considered to penetrate less than 450 µm into the MCC tablet, and the penetration
REVIEW 18 of 34
depth will further decrease as the surface moisture content increases. Hence, only the tablet
surface moisture will be captured during the analysis.
(a)
Figure 9. Cont.
(a)
(b)
Figure 9. Figure
SNV +9.SG SNV
(17,+ 2,
SG0)(17, 2, 0) transformed
transformed NIR spectra
NIR spectra obtained
obtained from
from (a) (a) APAP,
APAP, paperpaper
sheetsheet
and and
paper sheet superimposed on APAP tablet surface, (b) APAP, MCC disk and MCC disk with 0.45
paper sheet superimposed on APAP tablet surface, (b) APAP, MCC disk and MCC disk with 0.45 mm
mm of thickness superimposed on APAP tablet surface. 1: one sheet (thickness = 74 μm), 2: two
of thickness superimposed
sheets on APAP
(thickness = 148 tablet
μm), three surface.
sheets 1: one=sheet
(thickness (thickness
222 μm), = 74(thickness
four sheets µm), 2: two sheets
= 296 μm), five
(thicknesssheets
= 148 (thickness
µm), three=sheets (thickness
370 μm) of regular= paper
222 µm), fouron
stacked sheets
APAP (thickness = 296 µm), five sheets
tablet surface.
(thickness = 370 µm) of regular paper stacked on APAP tablet surface.
4.2. Tablet Compression Curve
4.2. Tablet Compression Curve
The compression stress–strain curve obtained for MCC VIVAPUR PH101 in this
Thestudy
compression stress–strain
at different densities ofcurve obtained
compaction for MCC
is shown VIVAPUR
in Figure PH101
10a. The curve in this is
obtained
study at characteristic
different densities of compaction is shown in Figure 10a. The curve obtained
of the behavior of the MCC powder under pressure (Figure 10b) [12,16]. is The
characteristic of the behavior of the MCC powder under pressure (Figure 10b) [12,16]. The
stress–strain curve can be split into two parts—the loading and unloading phases. The
stress–strain curve can be split into two parts—the loading and unloading phases. The
loading phase, line AB, describes the volume reduction of the powder bed. During volume
reduction, two main phenomena take place for soft particles such as MCC—the escape of
excess air from the powder bed interstices and densification. The latter can be explained
by three complex mechanisms [16,76]. The first mechanism takes place at the early stage
of compaction, during which the particles rearrange due to their inherent movement. On
contact, the tangential and normal components of the contact force between particles take
part in the organization and elastic–plastic deformation of the particles, respectively. The
second mechanism is the elastic–plastic deformation that occurs due to contact interactions
between neighboring particles, leading to hardening. The third mechanism is described
by the sharp increase in material flow resistance due to material strain hardening. The
unloading phase, line BE in Figure 10, is where the upper punch is retracted, and the tablet
tends to revert to expand due to its elastic recovery. During the first step, line BD, the
material tends to release excess energy stored in the compact that does not participate in
the microstructure of the compact. At the end of the unloading phase, line DE, the powder
exhibits strong nonlinear behavior due to dilatation of the particles in the compact. The
stress–strain curve obtained is consistent with previous results [77].
scribed by the sharp increase in material flow resistance due to material strain hardening.
The unloading phase, line BE in Figure 10, is where the upper punch is retracted, and the
tablet tends to revert to expand due to its elastic recovery. During the first step, line BD,
the material tends to release excess energy stored in the compact that does not participate
in the microstructure of the compact. At the end of the unloading phase, line DE, the pow-
der exhibits strong nonlinear behavior due to dilatation of the particles in the compact.
The stress–strain curve obtained is consistent with previous results [77].
(a) (b)
Figure 10. (a) Axial stress–axial strain compaction curve of the microcrystalline cellulose (MCC),
Figure 10. (a) Axial stress–axial strain compaction curve of the microcrystalline cellulose (MCC),
VIVAPUR PH101. (b) Typical axial stress–axial strain curve of MCC. (M is the compression modu-
VIVAPUR PH101.
lus; G is the tablet(b) Typical
shear axial stress–axial
modulus and K is thestrain
bulk curve of MCC.
modulus). A-B:(M is the compression
Loading modulus;
phase; B-E: Unloading
Gphase;
is the tablet shear modulus and K is the bulk modulus). A-B: Loading phase; B-E:
B-C: Linear elastic, C-D and D-E non-linear elastic behaviour of the compact. Unloading phase;
B-C: Linear elastic, C-D and D-E non-linear elastic behaviour of the compact.
4.3. Mechanical, Heat, and Moisture Simulation during Compaction
4.3. Mechanical, Heat, and Moisture Simulation during Compaction
4.3.1.Relative
4.3.1. RelativeDensity
DensityEvolution
Evolutionand andDistribution
Distribution
Mechanicalsimulation
Mechanical simulationofofthe thecompaction
compactionphase phaseofofthe
thetableting
tabletingprocess
processresulted
resultedinin
the evolution of the average relative density of the compact and
the evolution of the average relative density of the compact and its distribution on its distribution onthe
the3D
3D
volume of the die. Figure 11a shows the onset of compaction with a
volume of the die. Figure 11a shows the onset of compaction with a uniform relative powder uniform relative pow-
der density
density of During
of 0.26. 0.26. During compaction,
compaction, the upper
the upper edgesedges
of theof the powder
powder beddenser
bed were were denser
than
than
the restthe
of rest of the powder
the powder bed andbed theand
lowertheedges
lowerwere
edges thewere
leastthe least
dense dense11b–d);
(Figure (Figuresimilar
11b–d);
similarwere
results results were obtained
obtained in other[20,29,41].
in other studies studies [20,29,41].
This can be This can be explained
explained byfriction,
by die wall die wall
friction,
which whicha induces
induces a stressinside
stress gradient gradienttheinside
powder thebed
powder bedprevents
[38] and [38] and prevents
the powderthefrom
pow-
der from
sliding alongsliding along the [78].
the interfaces interfaces
At the[78].
endAtof the
the end of the compaction,
compaction, the relativethe relative
density of den-
the
sity of the powder was somewhat uniformly distributed, and the
powder was somewhat uniformly distributed, and the top-right corner of the tablet showed top-right corner of the
atablet showed
relatively densera relatively denser
region, which wasregion, whichwith
confirmed wasX-ray
confirmed with X-ray
tomography tomography
analysis [27,79].
analysis
This highly [27,79].
denseThis highly
region wasdense
foundregion
to be was foundof
the point to origin
be the of
point
the of origin
crack ofinitiated
that the crack
tablet lamination
that initiated [27,79].
tablet lamination [27,79].
To validate the mechanical simulation, the average relative density measured for the
out-of-die tablet and the relative density obtained from the simulation were compared
(Figure 12). The variation in relative density of the tablet with upper punch displacement
is presented. Good agreement between simulated and measured results was observed
for lower compaction, i.e., low upper punch penetration. For higher compaction i.e.,
high upper punch penetration (from 4 mm), the simulated relative density was slightly
higher than the measured values. These observations are due to the fact that at lower
punch penetration, i.e., lower compaction force, particle deformation is not high enough
to cause high relaxation of the out-of-die tablet; hence the density calculated from tablet
dimensions is relatively close to that obtained for the in-die tablet in the simulation. At
higher penetration, i.e., higher compaction force, particle deformation becomes important
such that relaxation of the out-of-die tablet also becomes more important, highly impacting
the out-of-die tablet dimensions and consequently its relative density.
Pharmaceutics 2023,
Pharmaceutics 15,15,
2023, 1652
x FOR PEER REVIEW 20 20
of of
3433
(a) (b)
(c) (d)
Figure 11. A 3D view of the simulated relative density evolution and distribution during compac-
Figure 11. A 3D view of the simulated relative density evolution and distribution during compaction
tion of a model powder (MCC VIVAPUR PH101) with flat-faced punches. The compaction process
of a model powder (MCC VIVAPUR PH101) with flat-faced punches. The compaction process
progresses from (a–d).
To validate the mechanical simulation, the average relative density measured for the
out-of-die tablet and the relative density obtained from the simulation were compared
(Figure 12). The variation in relative density of the tablet with upper punch displacement
is presented. Good agreement between simulated and measured results was observed for
lower compaction, i.e., low upper punch penetration. For higher compaction i.e., high up-
per punch penetration (from 4 mm), the simulated relative density was slightly higher
than the measured values. These observations are due to the fact that at lower punch pen-
etration, i.e., lower compaction force, particle deformation is not high enough to cause
high relaxation of the out-of-die tablet; hence the density calculated from tablet dimen-
sions is relatively close to that obtained for the in-die tablet in the simulation. At higher
penetration, i.e., higher compaction force, particle deformation becomes important such
that relaxation of the out-of-die tablet also becomes more important, highly impacting the
out-of-die tablet dimensions and consequently its relative density.
Figure
Figure12.12.Comparison
Comparisonbetween the the
between in-die tablet’s
in-die simulated
tablet’s relative
simulated densitydensity
relative and theand
ejected
the tablet’s
ejected
relative density evolution during compaction of MCC VIVAPUR PH101 powder
tablet’s relative density evolution during compaction of MCC VIVAPUR PH101 powder with flat-faced
with
punches.
flat-faced punches.
4.3.2. Temperature Evolution and Distribution

• Simulated temperature of tablets
The evolution of the 3D powder bed temperature and its distribution during com-
paction are shown in Figure 13a–d. At the beginning of the compaction phase (Figure 13b),
4.3.2. Temperature Evolution and Distribution

• Simulated temperature of tablets
The evolution of the 3D powder bed temperature and its distribution during com-
paction are shown in Figure 13a–d. At the beginning of the compaction phase (Figure 13b),
the temperature increases from 21.4 ◦ C (initial loose powder temperature) to around 24.5 ◦ C
at the powder bed–die wall interface and around 22.4 ◦ C at the surface in contact with the
punches. As compaction continues (Figure 13c), the temperature continues to increase and
the temperature at the punch–powder bed interfaces remains lower than the temperatures
at the center of the powder bed and at the die wall–powder bed interface. At the end of
compaction (Figure 13d), the same temperature distribution was observed in all interfaces,
i.e., punch–tablet and die wall–tablet (about 40 ◦ C), but the temperature at the center of the
tablet remained higher (about 46 ◦ C). However, the edge of the tablet showed the lowest
temperature (about 36 ◦ C). The simulation of the entire compaction phase showed that
heat was generated mostly at the tablet–die wall interface and the center of the tablet, as
shown in other simulations [18,20]. The heat generated in the tablet flowed from the center
to the outer layers by conduction within the tablet and was dissipated from the tablet by
conduction through the die wall and punches. The low temperature at the edges of the
tablet may be correlated with the higher density of these regions (Figure 11). High density
induces higher thermal effusivity at the edges of the tablet and thus lower temperatures
Pharmaceutics 2023, 15, x FOR PEER REVIEW 22 of in
34
these regions. The overall temperature distribution in the tablet at the end of compaction
was consistent with previous studies [18,20,80].
(a) (b)
(c) (d)
Figure 13. A 3D view of the simulated temperature evolution and distribution during compaction
Figure 13. A 3D view of the simulated temperature evolution and distribution during compaction of
of a model MCC PH101 VIVAPUR powder with flat-faced punches. The compaction process pro-
a model MCC PH101 VIVAPUR powder with flat-faced punches. The compaction process progresses
gresses from (a–d).
from (a–d).
• Measured surface and peripheral temperatures of the tablets
The out-of-die tablet surface temperature and peripheral temperature were meas-
ured with a thermal infrared camera one second after ejection. Figure 14 shows the ther-
mal infrared camera images of the compacted tablet at increasing relative density. The
• Measured surface and peripheral temperatures of the tablets

The out-of-die tablet surface temperature and peripheral temperature were measured
with a thermal infrared camera one second after ejection. Figure 14 shows the thermal
infrared camera images of the compacted tablet at increasing relative density. The images
show a clear increase in temperature during compaction. The rise in temperature is
explained by the heat generated by the irreversible deformation of powder particles,
interparticle friction, and particle–die wall friction [20,77,81]. Furthermore, the out-of-
die tablet’s surface temperature was lower than its peripheral temperature. The higher
temperature at the periphery may be due to the additional heat generated at the tablet–die
wall interface through friction during the ejection phase [18]. By increasing the level of
compaction of the powder bed, the temperature of all parts of the tablet increased because
more friction and irreversible deformation phenomena occurred as the tableting pressure
increased. Similar results were obtained by other studies [18,77].
(a)
(b)
Figure 14. Thermal infrared camera images of (a) the tablet surface and (b) tablet periphery showing
Figure 14. Thermal infrared camera images of (a) the tablet surface and (b) tablet periphery show-
temperature distribution and evolution during compaction of MCC VIVAPUR PH101 powder with
ing temperature distribution and evolution during compaction of MCC VIVAPUR PH101 powder
flat-faced punches. The relative density increases from left to right (RD1 = 0.26%, RD2 = 0.5%, and
with
RD3 flat-faced punches.
= 0.75%) pressed Thearelative
with density
compression increases
pressure of from left KPa,
80 ± 3.5 to right
18.8(RD1
± 1.3= MPa
0.26%,
andRD2 = 0.5%,
68.5 ± 5.4
and RD3 = 0.75%)
MPa, respectively. pressed with a compression pressure of 80 ± 3.5 KPa, 18.8 ± 1.3 MPa and
68.5 ± 5.4 MPa, respectively.
To investigate the temperature evolution during a single compaction phase, the tem-
To investigate the temperature evolution during a single compaction phase, the tem-
peratures of out-of-die tablets obtained at increasing levels of compaction, i.e., increasing
peratures of out-of-die tablets obtained at increasing levels of compaction, i.e., increasing
relative density, were measured. Figure 15 shows the temperature evolution with increas-
relative density, were measured. Figure 15 shows the temperature evolution with increas-
ing relative density. Both peripheral and surface temperatures increased with increasing
ing relative density. Both peripheral and surface temperatures increased with increasing
compaction. As the level of compaction increases, the contact pressure between particles
compaction. As the level of compaction increases, the contact pressure between particles
increases [76,81]; hence, more heat is generated and causes the temperature in parts of the
increases [76,81]; hence, more heat is generated and causes the temperature in parts of the
tablet to increase. The peripheral tablet temperature is higher than the surface tempera-
tablet to increase. The peripheral tablet temperature is higher than the surface temperature
ture at all levels of compaction, which may be explained by (1) the additional frictional
at all levels of compaction, which may be explained by (1) the additional frictional heat
heat generated at the tablet–die wall interface during ejection and (2) the difference in the
generated at the tablet–die wall interface during ejection and (2) the difference in the rate
rate of heat loss between the surface and periphery of the tablet. The temperature evolu-
of heat loss between the surface and periphery of the tablet. The temperature evolution
tion
curvecurve
can becan be separated
separated intodistinctive
into two two distinctive
phases.phases. (1) A low-compression
(1) A low-compression pressure
pressure phase
(21.4% < relative density < 35%) in which a small temperature change, from 21.421.4
phase (21.4% < relative density < 35%) in which a small temperature change, from ◦ C °C
to
to 24
◦ °C, is observed. During this phase, particle reorganization and interparticle
24 C, is observed. During this phase, particle reorganization and interparticle friction friction
dominate
dominate [16], which may
[16], which may explain
explain the
the low
low rise
rise in
in temperature.
temperature. (2)
(2) A
A high-compression
high-compression
pressure phase (35% < relative density < 80%) with a high temperature change, from 24 °C
to 38 °C, is observed. During this phase, particles are close enough to deform irreversibly
on contact and the consolidation between particles induces more radial pressure, which
may cause stronger wall friction [76,81]. Therefore, a higher temperature change is ob-
served during the second phase. These two phases observed during temperature evolu-
pressure phase (35% < relative density < 80%) with a high temperature change, from 24 ◦ C
to 38 ◦ C, is observed. During this phase, particles are close enough to deform irreversibly on
contact and the consolidation between particles induces more radial pressure, which may
cause stronger wall friction [76,81]. Therefore, a higher temperature change is observed
during the second phase. These two phases observed during temperature evolution were
consistent with the loading phase curve (Figure 10a), where axial pressure is low in the
region of 21.4% < relative density < 35%, followed by a sharp increase in pressure from
35% to about 80% relative density. Furthermore, as the number of tableting runs increased,
Pharmaceutics 2023, 15, x FOR PEER tablet temperature increased gradually (Figure 16). As the number of compressions
REVIEW
Pharmaceutics 2023, 15, x FOR PEER REVIEW
runs
2424ofof3434
increases, compression tools become hotter and transfer additional heat to powder particles,
leading to gradual temperature rise [77,80].
Figure
Figure15.
Figure 15.Temperature
15. Temperatureevolution
Temperature evolution
evolutionmeasured
measured
measuredduring
duringcompaction
duringcompactionofofMCC
compaction MCC
of PH101
PH101
MCC VIVAPUR
VIVAPUR
PH101 powder
powder
VIVAPUR powder
with
withflat-faced
flat-facedpunches.
punches.
with flat-faced punches.
Figure
Figure16.
16.Temperature
Temperatureevolution
evolutionof MCC PH101 VIVAPUR tablets asasaafunction ofofthe duration ofof
Figure 16. Temperature evolutionofofMCC
MCCPH101
PH101VIVAPUR
VIVAPURtablets
tablets asfunction
a function the duration
of the duration of
tablet
tabletpress
pressoperation.
operation.The
Thetablets
tabletswere
weremade
madeby bycompressing
compressingMCC
MCCpowder
powderwith withflat-faced
flat-faced
tablet press
punches. operation. The tablets were made by compressing MCC powder with flat-faced punches.
punches.600600tablets
tabletswere
wereproduced
producedinin1010min.
min.
600 tablets were produced in 10 min.
The
Thesimulation
simulationresults
resultswere
werecompared
comparedtotothethemeasured
measuredtemperature
temperaturefor forvalidation
validation
purposes.
purposes.Figure
Figure17
17shows
showsthe
themeasured
measuredtemperature
temperatureofofthetheejected
ejectedtablet
tabletcompared
comparedtoto
the
thesimulated
simulatedtemperature
temperatureofofthe
thein-die
in-dietablet
tabletasasaafunction
functionofofthe
thelevel
levelofofcompaction,
compaction,
which
whichisisrepresented
representedby
bypunch
punchdisplacement
displacementininthethedie.
die.The
Themeasured
measuredand andsimulated
simulatedtem-
tem-
The simulation results were compared to the measured temperature for validation
purposes. Figure 17 shows the measured temperature of the ejected tablet compared to
the simulated temperature of the in-die tablet as a function of the level of compaction,
which is represented by punch displacement in the die. The measured and simulated
temperatures increased with the level of compaction. In the cases of both measured
Pharmaceutics 2023, 15, x FOR PEER REVIEW
and
25 of 34
simulated temperatures, the tablet’s peripheral temperature was higher than its surface
temperature. In the simulation results, the difference in temperature between the surface
and the periphery may be explained by the rate of heat generation and dissipation at each
interface during
interface during compaction.
compaction. In
In the
the experimental
experimental results,
results, the
the additional
additional heat
heat generated
generated
during ejection may be responsible for the difference in temperature.
during ejection may be responsible for the difference in temperature.
Figure 17.
Figure 17. The
The evolution
evolution of
of the
the measured
measured and
and simulated
simulated temperatures
temperatures of
of MCC
MCC PH101
PH101 VIVAPUR
VIVAPUR
tablets as
tablets as aa function
function of
of upper
upper punch
punchpenetration.
penetration.
4.3.3. Tablet
4.3.3. Tablet Moisture
MoistureContent
ContentEvolution
Evolutionand
andDistribution
Distributionduring
duringCompaction
Compaction
At
At low
lowupper
upperpunch
punchpenetration
penetration (upper
(upper punch
punchdisplacement
displacement of <2
ofmm), the simulated
<2 mm), the simu-
results and the
lated results measured
and resultsresults
the measured matched. However,
matched. as the punch
However, as thewent
punch deeper
went into
deeperthe into
die,
the
the simulated temperature
die, the simulated of the in-die
temperature tablets
of the in-diebecame
tabletsgradually higher than
became gradually the measured
higher than the
temperature of the ejected
measured temperature oftablets. Because
the ejected the measured
tablets. Because the temperatures
measured were obtained from
temperatures were
ejected
obtained tablets, duringtablets,
from ejected the ejection
during phase, water phase,
the ejection molecules watercould have evaporated
molecules could have evap-from
the tablet
orated fromsurface beforesurface
the tablet the measurements were taken. The
before the measurements werelatent heat
taken. Theused to evaporate
latent heat used
the
to evaporate the water molecules could have produced a cooling effect on theinstance,
water molecules could have produced a cooling effect on the tablet. For tablet. Forat
maximum punch penetration (about 7.83 mm in the die), the difference
instance, at maximum punch penetration (about 7.83 mm in the die), the difference be- between the average
simulated
tween thetemperature and the temperature
average simulated average measured and thetemperature 15 ◦ C. Assuming
was about temperature
average measured was
that
about water
15 °C.evaporated
Assuming from
thatthe top evaporated
water 2 mm of thefrom tablet,the
thetop
corresponding
2 mm of the amount
tablet, theof water
corre-
evaporated
sponding amount is aboutof1water
mg. With the initial
evaporated amount
is about of With
1 mg. moisture being amount
the initial 43 mg ofofwater per
moisture
1being
g of powder, the evaporation of 1 mg during tablet ejection is plausible.
43 mg of water per 1 g of powder, the evaporation of 1 mg during tablet ejection is The agreement
between
plausible. theThe
simulated
agreementand measured
between the temperatures
simulatedatand low measured
punch penetration distance
temperatures at may
low
be explained by the low water evaporation rate due to the low temperature
punch penetration distance may be explained by the low water evaporation rate due to rise. A decrease
in
thetablet temperature rise.
low temperature during ejection was
A decrease also temperature
in tablet found in other studies
during [18,77].
ejection was also found
Figure 18 shows
in other studies [18,77]. the moisture sorption isotherm curve of the MCC VIVAPUR PH101
used Figure
in this 18study. The curve has a type II and/or III shape according
shows the moisture sorption isotherm curve of the MCC VIVAPUR PH101 to the Brunauer
used in this study. The curve has a type II and/or III shape according to the Brunauer
classification [82], which is obtained for highly porous material, such as the MCC PH101
used in this study, which is well known to have high intraparticle porosity. The type II
shape indicates that adsorption begins with the first monolayer between the powder and
classification [82], which is obtained for highly porous material, such as the MCC PH101
used in
Pharmaceutics 2023, 15, x FOR PEER REVIEW this study, which is well known to have high intraparticle porosity. The type
26 of II
34
shape indicates that adsorption begins with the first monolayer between the powder and
water molecules at lower humidity, followed by multilayer adsorption between water
molecules, and finally leading to capillary condensation at® higher humidity [47]. The
moisture sorption curve was used in COMSOL Multiphysics software to follow the water
moisture sorption curve was used in COMSOL Multiphysics® software to follow the water
content in the powder bed as the compaction level increased.
content in the powder bed as the compaction level increased.
Figure18.
Figure 18.The
Thesorption
sorptioncurve
curveof
ofMCC
MCCVIVAPUR
VIVAPURPH101
PH101powder.
powder.
Figure
Figure19 19shows
showsthe themoisture
moisturedistribution
distributionin in the
the tablet
tablet during
during the the compaction
compaction phase.
During
During compaction,
compaction, the the heat
heat generated
generated can can bebe transmitted
transmitted as as thermal
thermal energy
energy to to water
water
molecules
molecules in in the
the form
formof ofkinetic
kineticenergy.
energy. TheThewater
watermolecules
molecules that thathave
haveenough
enoughkinetic
kinetic
energy
energycan candiffuse
diffusebetween
betweenparticles
particlesand
andintraparticle
intraparticleporesporesininthe
thepowder
powderbed bed[27,72,83]
[27,72,83]
due
due to the water vapor pressure gradient between the powder bed and thesurrounding
to the water vapor pressure gradient between the powder bed and the surrounding
environment.
environment.The Theair
aircontained
containedin inthe
thepowder
powderbed beddiffuses
diffusesradially
radiallyfromfromthethecenter
centerto tothe
the
periphery
periphery of ofthe
thetablet
tabletandandescapes
escapes from
from thethe powder
powder bed bed atatthe
theedges
edges of
ofthe
thetablets
tablets[27]
[27]
during
during compaction.
compaction. The The outflowing
outflowing airair carrying
carrying water
water vapor
vapor largely
largely contributes
contributes to to the
the
path of water vapor diffusion. Consequently, the water content of the
path of water vapor diffusion. Consequently, the water content of the tablet is higher at tablet is higher at the
periphery
the peripherythanthan
at theatcenter (Figure
the center 19b–d).
(Figure Furthermore,
19b–d). Furthermore, the higher temperature
the higher temperature of theof
center of the tablet can lead to higher kinetic energy in the water molecules
the center of the tablet can lead to higher kinetic energy in the water molecules in this in this region.
Hence,
region. the highthe
Hence, heat coupled
high with thewith
heat coupled air path may
the air pathexplain the lower
may explain the water
lower content
water con-in
the
tentcenter
in theof the tablet
center of the(Figure 19d). The
tablet (Figure 19d).tablet edgesedges
The tablet appeared
appearedmoister because
moister of the
because of
higher powder density, which can slow down the rate of diffusion
the higher powder density, which can slow down the rate of diffusion of water molecules of water molecules to
the surrounding
to the surrounding environment.
environment.
During
During compaction,3the
compaction, the overall
overall water
water content
content density
density ofofthe
the powder
powderbed beddecreased
decreased
from 20.85 kg of water/m of powder to about 8 kg of water/m 3 of powder. The decrease
from 20.85 kg of water/m3 of powder to about 8 kg of water/m3 of powder. The decrease
in
inwater
watercontent
contentmaymaybebeexplained
explainedbybythe evaporation
the evaporation of of
water molecules
water molecules fromfromthethe
powder
pow-
compact or tablet to the surrounding environment, particularly the interfaces.
der compact or tablet to the surrounding environment, particularly the interfaces. Because Because the
simulation in this study allowed water molecules to evaporate from the tablet surface, no
the simulation in this study allowed water molecules to evaporate from the tablet surface,
accumulation was observed on the tablet surface.
no accumulation was observed on the tablet surface.
Pharmaceutics 2023,
Pharmaceutics 15, x15,
2023, FOR1652PEER REVIEW 2627 of 34
of 33
(a) (b)
(c) (d)
Figure 19. A 3D view of the simulated moisture content evolution and distribution during compac-
Figure 19. A 3D view of the simulated moisture content evolution and distribution during com-
tion of model powder (MCC VIVAPUR PH101) with flat-faced punches. The compaction process
paction of model powder (MCC VIVAPUR PH101) with flat-faced punches. The compaction process
Water
Watermolecules
moleculesthat
thatdiffuse
diffuse from
from thethe powder
powderbed bedcan canbebetrapped
trapped at at
thethe interfaces
interfaces
between
betweenthe thecompression toolsand
compression tools and powder
powder bedbed
duringduring compaction,
compaction, which
which may may
cause cause
water
water
vaporvapor to accumulate
to accumulate in regions.
in these these regions.
To verifyTothis
verify this assumption,
assumption, an NIR
an NIR sensor wassensor
used towas
used to measure
measure the amount
the amount of moisture
of moisture at thesurface
at the tablet tablet just
surface
afterjust after Figure
ejection. ejection.18aFigure
shows18a
the raw
shows the noisy NIR spectra
raw noisy collected
NIR spectra for tablets
collected with awith
for tablets watera content of 4–14.5%.
water content The raw
of 4–14.5%. The
spectra
raw without
spectra withoutpretreatment show show
pretreatment a watera absorption band in the
water absorption bandranges of 1190–1220
in the ranges of nm 1190–
andnm
1220 1400–1690 nm. After
and 1400–1690 pretreatment,
nm. with the combination
After pretreatment, of EMSC and
with the combination SG (17, and
of EMSC 1, 0),SG
weak and strong water absorption at 1202 nm
(17, 1, 0), weak and strong water absorption at 1202 nm and and 1450 nm, respectively, was observed
1450 nm, respectively, was
(Figure 20b). The PCA score plot coupled with the Hotelling T2 ellipse at a 95% confidence
observed (Figure 20b). The PCA score plot coupled with the Hotelling T2 ellipse at a 95%
interval created with the pretreated data does not show outliers (Figure 21a). The model
confidence interval created with the pretreated data does not show outliers (Figure 21a).
with three principal components explained 98% of the total variance of X (wavelength)
The
andmodel
98.5%with
of thethree principalofcomponents
total variance explained
Y (water content). 98% ofmodel
A calibrated the total
(Figurevariance
21b) wasof X
(wavelength)
obtained with and 98.5% of coefficient
a regression the total variance of Y (R
in calibration (water content).
2 ) of 97% A calibrated model
and a root-mean-square
(Figure
error in calibration (RMSEC) of 0.14%. In the K-folds cross-validation, an (R
21b) was obtained with a regression coefficient in calibration R2 )ofof0.96%
2 97% and
and a
root-mean-square error in
RMSECV of 0.20% were obtained. calibration (RMSEC) of 0.14%. In the K-folds cross-validation,
an R2 of 0.96% and RMSECV of 0.20% were obtained.
Pharmaceutics2023,
15,1652
x FOR PEER REVIEW 28 of
27 of3334
(a) (b)
(a) (b)
Figure 20. (a) Raw near-infrared (NIR) and (b) pretreated NIR spectra collected on the surface of
Figure 20. (a) Raw near-infrared (NIR) and (b) pretreated NIR spectra collected on the surface of
MCC VIVAPUR
MCC VIVAPUR PH101
PH101 tablets.
tablets.
(a) (b)
(a) (b)
Figure 21. (a) PCA Hotelling score plot of pretreated data and (b) PLS regression model of actual
Figure 21.
21. (a)
(a) PCA
PCAHotelling
Hotellingscore
scoreplot
plot of pretreated data and
(b)(b)
PLSPLS regression model of actual
Figure
and predicted values of tablet surface of pretreated
water data
content. and regression model of actual and
and predicted values of tablet surface water content.
predicted values of tablet surface water content.
Figure 22 shows the predicted tablet surface moisture content for 10 min of tablet
Figure 22 shows the predicted tablet surface moisture content for 10 min of tablet
press operation, representing 600 tablets pressed. The tablets had the same compaction
press operation, representing 600 tablets tablets pressed.
pressed. The tablets had the same compaction compaction
relative density of about 65%. All the tablets were compressed from the same batch of
relative density of about 65%. All All the tablets were compressed from the same batch of
powder. Each point on Figure 22 represents the average moisture content of five tablets,
powder. Each point on Figure 22 represents the average moisture content of five tablets,
with the measurements made every two minutes. Three replicate tests were performed,
measurements made every two minutes. Three replicate tests were performed,
with the measurements
with two hours between each replicate to allow enough time for the tableting tools to cool
with two hours
hoursbetween
betweeneach eachreplicate
replicatetoto allow
allowenough
enough time for for
time the the
tableting toolstools
tableting to cool
to
down. ANOVA showed that the three replicates were not significantly different (p = 0.96).
down.
cool ANOVA
down. ANOVAshowed that thethat
showed threethereplicates were notwere
three replicates significantly different (p
not significantly = 0.96).
different
Figure 22 shows that at 0 min, the moisture level of the powder was the same as the am-
(p = 0.96).
Figure Figure that
22 shows 22 shows
at 0 min,thattheat 0moisture
min, thelevel
moisture
of thelevel
powderof thewaspowder
the same wasasthethesame
am-
bient condition of 4.25%. After a half-minute run, the moisture content of the tablet surface
as thecondition
bient ambient condition of 4.25%.
of 4.25%. After After a half-minute
a half-minute run, the
run, the moisture moisture
content of thecontent of the
tablet surface
increased from the ambient to 9% for the first replicate and 7.5% for the second and third
tablet surface
increased fromincreased
the ambient from to the ambient
9% for to replicate
the first 9% for the andfirst
7.5%replicate
for the and
second7.5%and forthird
the
replicates.
second The
andThethird difference
replicates. in moisture content between
The difference the replicates may be explained by
replicates. difference in moisture content in moisture
between thecontent
replicates between
may bethe replicates
explained by
the
may rate of moisture
be explained evaporation
byevaporation from
the rate of moisture the tablet surface after the ejection phase, which
the rate of moisture from the evaporation
tablet surfacefromafterthethetablet
ejectionsurface
phase,after the
which
may vary
ejection due to
phase, differences
which may in tablet
vary due tosurface temperature
differences in tablet between
surface replicates.
temperature Thebetween
increase
may vary due to differences in tablet surface temperature between replicates. The increase
observed after
replicates. Theaaincrease
half minute
observedof tableting amay
after may be
halfbe attributed
minute to the abrupt
of tableting increase
may increase
be in tablet
attributed to
observed after half minute of tableting attributed to the abrupt in tablet
temperature
the abrupt (Figure
increase in 17),
tabletwhich led
temperatureto the transfer
(Figure of
17), enough
which kinetic
led to the energy
transferto water
of enoughmol-
temperature (Figure 17), which led to the transfer of enough kinetic energy to water mol-
ecules energy
kinetic in the tablet’s
to waterpore channels. Between
molecules thepore
half minute and 10 min of tableting runs,
ecules in the tablet’s pore channels.inBetweenthe tablet’s
the half channels.
minute andBetween
10 min ofthe half minute
tableting runs,
the
and moisture content
10 min ofcontent
tableting of the tablet surface decreased for the three replicates (Figure 22).
the moisture ofruns, the moisture
the tablet content of for
surface decreased the the
tablet surface
three decreased
replicates (Figurefor 22).
the
Pharmaceutics2023, 15,1652
x FOR PEER REVIEW 28 of
29 of 33
34
three replicates
The gradual rise(Figure 22).
in tablet The gradual
temperature rise in
shown in Figure
tablet temperature
17 may explain shown in Figurede-
the observed 17
may explain the observed decrease in moisture content, because as the
crease in moisture content, because as the tableting runs increases the evaporation from tableting runs
increases
the tabletthe evaporation
surface may becomefrom the
moretablet surfaceHowever,
important. may become the more
tabletimportant. However,
surface temperature
the tablet surface temperature increased at a slower rate from the first tablet
increased at a slower rate from the first tablet compression, which consequently led to a compression,
which
slowerconsequently led toofa the
rate of decrease slower ratesurface
tablet of decrease of thecontent.
moisture tablet surface
After 10moisture
min ofcontent.
runs, a
After 10 min of runs, a steady state was not reached because the
steady state was not reached because the tablet temperature was still rising. tablet temperature was
still rising.
Figure 22.
Figure 22. The
The predicted
predicted surface
surface moisture
moisture content
content of
of ejected
ejected tablets
tablets during
during 10
10 min
min of
of tableting
tableting runs.
runs.
Based on
Based on these
these observations,
observations, we we may
may assume
assume thatthat the
the moisture
moisture evaporating
evaporating from from the
the
tablet at
tablet at the
the punch
punch and and tablet
tablet interface
interface accumulates.
accumulates. The water molecules molecules in the the water
water
vapor can be
vapor be physiosorbed
physiosorbedthrough throughvan van der Waals
der Waals forces
forceson on
the the
upper punch
upper surface
punch dur-
surface
ing the dwell time [84–86]. Due to the confined space
during the dwell time [84–86]. Due to the confined space between the powder bed andbetween the powder bed and the
upper
the upperpunch,
punch, i.e.,i.e.,
thetheinterface,
interface, thethenumber
numberofofvan vander derWaals
Waals interactions between the the
vapor-phase
vapor-phasemolecules
moleculesininthat thatregion
region cancanincrease.
increase.Therefore,
Therefore, thethe
firstfirst
layer adsorbed
layer adsorbed on the
on
punch
the punchsurface maymay
surface attract moremore
attract water molecules,
water and multiple
molecules, and multiple layers of water
layers molecules
of water mole-
may
culesdevelop
may developon theon punch surface.
the punch The higher
surface. amount
The higher of water
amount vaporvapor
of water evaporating from
evaporating
the
fromtablet
the may
tabletresult
mayin waterinvapor
result watercondensation below water
vapor condensation below vapor
water saturation pressure,
vapor saturation
which is known
pressure, whichas iscapillary
known as condensation [87–89]. Capillary
capillary condensation [87–89]. condensation may occur atmay
Capillary condensation the
interface between the surface and the tablet during the dwell
occur at the interface between the surface and the tablet during the dwell time. On the time. On the punch surface,
capillary condensation
punch surface, capillary can be more pronounced
condensation can be more in cavities,
pronouncedsuch as inthe valleyssuch
cavities, in surface
as the
roughness, logos cavities,
valleys in surface roughness, andlogos
eventual cracks
cavities, and[87–89].
eventual cracks [87–89].
Capillary
Capillary condensation
condensation at at the
the punch
punch surface
surface during
during tableting
tableting can can lead
lead to
to capillary
capillary
adhesion
adhesion of the tablet surface particles to the upper punch surface [37,90,91]. The tablet’s
of the tablet surface particles to the upper punch surface [37,90,91]. The tablet’s
surface
surface particles
particles can can adhere
adhere to to the
the punch
punch surface
surface when
when the the capillary
capillary forces
forces between
between thethe
punch
punch surface
surface andand thethe particles
particles are are stronger
stronger thanthan the
the cohesive
cohesive forces
forces between
between the the tablet’s
tablet’s
particles
particles atat the
the surface
surface and and the
the core
core [92].
[92]. Moisture
Moisture is is an
an important
important factorfactor that
that contributes
contributes
to
to the sticking of tablets observed during tableting runs [37,41,43]. Granules with higher
the sticking of tablets observed during tableting runs [37,41,43]. Granules with higher
moisture
moisture content
content led led totoaahigher
higherfrequency
frequencyofofsticking
sticking [37].
[37]. Similarly,
Similarly, tableting
tableting parame-
parameters
ters
suchsuch as compression
as compression speed speed
increaseincrease the occurrence
the occurrence of sticking
of sticking [93,94].
[93,94]. In addition,
In addition, com-
compression speed increases the heat generated during tableting
pression speed increases the heat generated during tableting [18,23]. Consequently, the [18,23]. Consequently,
the greater
greater occurrence
occurrence of of sticking
sticking could
could bebeattributed
attributedtotothe theincreased
increasedwater water vapor
vapor at at the
the
punch–tablet interface due to the higher heat generated with increased speed. Furthermore,
punch–tablet interface due to the higher heat generated with increased speed. Further-
punches with surface cavities, such as logos and cracks, due to surface wear have been
more, punches with surface cavities, such as logos and cracks, due to surface wear have
found to increase sticking [1,35,79], which may be attributed to the capillary condensation
been found to increase sticking [1,35,79], which may be attributed to the capillary conden-
in these cavities, where capillary bridges between the punch and particles cause adherence
sation in these cavities, where capillary bridges between the punch and particles cause
through capillary forces. The use of punches coated with hydrophobic materials has been
shown to minimize or diminish the frequency of sticking [1,29]. This may be due to hy-
drophobic surfaces preventing water molecules from adsorption, leading to the avoidance
of capillary condensation.
5. Conclusions
Finite element modeling of the compaction phase of the tableting process was per-
formed to investigate the density, temperature, and moisture distribution and evolution of
tablets. The DPC and HAM models implemented in COMSOL Multiphysics® 5.6 software
were used. At the end of the compaction phase, (1) relative density tended to be uniform
across the tablet, but was higher at the tablet edges, and (2) the temperature of the powder
bed increased during compaction and was higher at the center of the tablet than on the
surface, which was much cooler due to heat dissipation through the compression tools.
These thermomechanical simulation results are in broad agreement with the experimental
results of this study and corroborate findings reported in the literature. The simulation
results showed that powder moisture evaporated from the powder bed to its surroundings
during the compaction phase. The NIR light (wavelengths: 970 nm to 1700 nm) penetration
depth was found to be less than 450 µm into the MCC tablet. Hence, the NIR sensor was
used to measure the water content of the tablet surface just after ejection, and showed that
moisture content was higher at the tablet surface than the loose powder.
These results suggest that a fraction of the water molecules evaporating from the pow-
der bed at the powder and punch interfaces accumulates on the tablet surface. In addition,
the evaporation rate increased as the tableting runs increased due to the gradual rise in
tablet temperature. Based on these observations, we propose that moisture evaporation
from the powder bed and accumulation at the punch–tablet interfaces can induce capillary
condensation between the tablet and the punch during the dwell time, leading to sticking.
The results of this study bring new insights into punch sticking and may contribute to the
investigation of other tableting problems.
Supplementary Materials: The following supporting information can be downloaded at https://

www.mdpi.com/article/10.3390/pharmaceutics15061652/s1: S1: Video of tablet surface moisture con-
tent measurement; S2: Drucker Prager Cap model parameters; S3: Heat transfer coefficients calculation.
Author Contributions: All authors were involved in the conceptualization, methodology, validation,
and formal analysis of this study. Original draft preparation, K.K. All authors were involved in
reviewing, editing, and visualization. Supervision, R.G., N.A. and F.G. All authors have read and
agreed to the published version of the manuscript.
Funding: This work was funded by the Pfizer Chair on Process Analytical Technologies in Pharma-
ceutical Engineering (RDCPJ 522450-17) and CRSNG Natural Sciences and Engineering Research
Council of Canada (RDCPJ 522450-17).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Acknowledgments: . We thank the PRAM technicians at Université de Sherbrooke for their assistance
with the measurement of powder properties. We also thank Yosser Ben Sghaier, Arcène Loìc Dwegeo,
and Erwan Baudillon for their help and input during the data collection phase.
Conflicts of Interest: The authors declare that they have no conflict of interest.
References
1. Chattoraj, S.; Daugherity, P.; McDermott, T.; Olsofsky, A.; Roth, J.W.; Tobyn, M. Sticking and picking in pharmaceutical tablet
compression: An IQ Consortium review. J. Pharm. Sci. 2018, 107, 2267–2282. [CrossRef]
2. Zhoa, J. Compaction of Pharmaceutical Powders on a High-Speed Rotary Press. Master’s Thesis, Faculty of Pharmaceutical
Sciences, University of British Columbia, Vancouver, BC, Canada, 2003. Available online: https://open.library.ubc.ca/media/
download/pdf/831/1.0090907/2 (accessed on 14 December 2022).
3. Denny, P.J. Compaction quations: A compression of the Heckel and Kawakita equations. Powder Technol. 2002, 183, 162–172.
[CrossRef]
4. Leuenberger, H.; Rohera, B.D. Fundamentals of powder compression: 1. The compatibility and compressibility of pharmaceutical
powders. Pharm. Res. 1986, 3, 12–22. [CrossRef]
5. Drucker, D.C.; Prager, W. Soil mechanics and plastic analysis or limit design. Q. Appl. Math. 1952, 10, 157–165. [CrossRef]
6. Baladi, G.Y.; Rohani, B. Soil plasticity. In Theoretical Foundation for Large-Scale Computations for Nonlinear Material Behavior. Mechanics
of Elastic and Inelastic Solids 6; Nemat-Nasser, S., Asaro, R.J., Hegemier, G.A., Eds.; Springer: Dordrecht, The Netherlands, 1984;
Volume 6, pp. 317–352. [CrossRef]
7. Schwer, L.E.; Murray, Y.D. A three-invariant smooth cap model with mixed hardening. Int. J. Numer. Anal. Methods Geomech. 1994,
18, 657–688. [CrossRef]
8. Foster, C.D.; Regueiro, R.A.; Fossum, A.F.; Borja, R.I. Implicit numerical integration of a three-invariant, isotropic/kinematic
hardening cap plasticity model for geomaterials. Comput. Methods Appl. Mech. Eng. 2005, 194, 5109–5138. [CrossRef]
9. DiMaggio, F.L.; Sandler, I.S. Material model for granular soils. J. Eng. Mech. Div. 1971, 97, 935–950. [CrossRef]
10. Schofield, A.N.; Wroth, P. Critical State Soil Mechanics; McGraw-Hill: London, UK, 1968; Volume 310. Available online: http://www.civ.
eng.cam.ac.uk/geotech_new/publications/schofield8wroth81968.pdf (accessed on 15 March 2020).
11. Wu, C.Y.; Ruddy, O.M.; Bentham, A.C.; Hancock, B.C.; Best, S.M.; Elliott, J.A. Modelling the mechanical behaviour of pharmaceu-
tical powders during compaction. Powder Technol. 2005, 152, 107–117. [CrossRef]
12. Han, L.H.; Elliott, J.A.; Bentham, A.C.; Mills, A.; Amidon, G.E.; Hancock, B.C. A modified Drucker-Prager Cap model for die
compaction simulation of pharmaceutical powders. Int. J. Solids Struct. 2008, 45, 3088–3106. [CrossRef]
13. Si, G.N.; Lan, C. Modeling and simulation of pharmaceutical powder compaction. Appl. Mech. Mater. 2012, 217, 1403–1406.
[CrossRef]
14. Hayashi, Y.; Otoguro, S.; Miura, T.; Onuki, Y.; Obata, Y.; Takayama, K. Effect of process variables on the Drucker–Prager cap
model and residual stress distribution of tablets estimated by the finite element method. Chem. Pharm. Bull. 2014, 62, 1062–1072.
[CrossRef]
15. Diarra, H.; Mazel, V.; Busignies, V.; Tchoreloff, P. Investigating the effect of tablet thickness and punch curvature on density
distribution using finite elements method. Int. J. Pharm. 2015, 493, 121–128. [CrossRef] [PubMed]
16. Baroutaji, A.; Lenihan, S.; Bryan, K. Combination of finite element method and Drucker-Prager Cap material model for simulation
of pharmaceutical tableting process. Materialwiss. Werkst. 2017, 48, 1133–1145. [CrossRef]
17. Krok, A.; Peciar, M.; Fekete, R. Numerical investigation into the influence of the punch shape on the mechanical behavior of
pharmaceutical powders during compaction. Particuology 2014, 16, 116–131. [CrossRef]
18. Zavaliangos, A.; Galen, S.; Cunningham, J.; Winstead, D. Temperature evolution during compaction of pharmaceutical powders.
J. Pharm. Sci. 2008, 97, 3291–3304. [CrossRef]
19. Klinzing, G.R.; Zavaliangos, A.; Cunningham, J.; Mascaro, T.; Winstead, D. Temperature and density evolution during compaction
of a capsule shaped tablet. Comput. Chem. Eng. 2010, 34, 1082–1091. [CrossRef]
20. Krok, A.; García-Triñanes, P.; Peciar, M.; Wu, C.Y. Finite element analysis of thermomechanical behaviour of powders during
tabletting. Chem. Eng. Res. Des. 2016, 110, 141–151. [CrossRef]
21. DeCrosta, M.T.; Schwartz, J.B.; Wigent, R.J.; Marshall, K. Thermodynamic analysis of compact formation: Compaction, unloading,
and ejection I. Design and development of a compaction calorimeter and mechanical and thermal energy determinations of
powder compaction. Int. J. Pharm. 2000, 198, 113–134. [CrossRef]
22. Ketolainen, J.; IIkka, J.J.; Paronen, P. Temperature changes during tabletting measured using infrared thermoviewer. Int. J. Pharm.
1993, 92, 157–166. [CrossRef]
23. Krok, A.; Mirtic, A.; Reynolds, G.K.; Schiano, S.; Roberts, R.; Wu, C.-Y. An experimental investigation of temperature rise during
compaction of pharmaceutical powders. Int. J. Pharm. 2016, 513, 97–108. [CrossRef]
24. Wu, C.Y.; Hancock, B.C.; Mills, A.; Bentham, A.C.; Best, S.M.; Elliott, J.A. Numerical and experimental investigation of capping
mechanisms during pharmaceutical tablet compaction. Powder Technol. 2008, 181, 121–129. [CrossRef]
25. Garner, S.; Ruiz, E.; Strong, J.; Zavaliangos, A. Mechanisms of crack formation in die compacted powders during unloading
and ejection: An experimental and modeling comparison between standard straight and tapered dies. Powder Technol. 2014,
264, 114–127. [CrossRef]
26. Mazel, V.; Diarra, H.; Malvestio, J.; Tchoreloff, P. Lamination of biconvex tablets: Numerical and experimental study. Int. J. Pharm.
2018, 542, 66–71. [CrossRef] [PubMed]
27. Klinzing, G.R.; Troup, G.M. Modeling the air pressure increase within a powder bed during compression—A step toward
understanding tablet defects. J. Pharm. Sci. 2019, 108, 1991–2001. [CrossRef] [PubMed]
28. Waimer, F.; Krumme, M.; Danz, P.; Tenter, U.; Schmidt, P.C. The influence of engravings on the sticking of tablets. Investigations
with an instrumented upper punch. Pharm. Dev. Technol. 1999, 4, 369–375. [CrossRef]
29. Al-Karawi, C.; Leopold, C.S. A comparative study on the sticking tendency of ibuprofen and ibuprofen sodium dihydrate to
differently coated tablet punches. Eur. J. Pharm. Biopharm. 2018, 128, 107–118. [CrossRef]
30. Danjo, K.; Kamiya, K.; Otsuka, A. Effect of temperature on the sticking of low melting point materials. Chem. Pharm. Bull. 1993,
41, 1423–1427. [CrossRef]
31. Sun, C.C. Materials science tetrahedron—A useful tool for pharmaceutical research and development. J. Pharm. Sci. 2009,
98, 1671–1687. [CrossRef]
32. Wakins, V.; Chu, E.; Schlam, R.; Sidorenko, A.; Badawy, S.; Yin, S.; Narang, A.S. Molecular basis of crystal morphology dependent
adhesion behavior of mefenamic acid during tableting. Pharm. Res. 2014, 31, 160–172. [CrossRef]
33. Danjo, K.; Kojima, S.; Chen, C.; Sunada, H.; Otsuka, A. Effect of water content on sticking during compression. Chem. Pharm. Bull.
1997, 45, 706–709. [CrossRef]
34. Roberts, M.; Ford, J.L.; MacLeod, G.S.; Fell, J.T.; Smith, G.W.; Rowe, P.H. Effects of surface roughness and chrome plating of punch
tips on the sticking tendencies of model ibuprofen formulations. J. Pharm. Pharmacol. 2003, 55, 1223–1228. [CrossRef] [PubMed]
35. Roberts, M.; Ford, J.L.; MacLeod, G.S.; Fell, J.T.; Smith, G.W.; Rowe, P.H.; Dyas, M.A. Effect of lubricant type and concentration on
the punch tip adherence of model ibuprofen formulations. J. Pharm. Pharmacol. 2004, 56, 299–305. [CrossRef]
36. Roberts, M.; Ford, J.L.; MacLeod, G.S.; Fell, J.T.; Smith, G.W.; Rowe, P.H.; Dyas, M.A. Effect of punch tip geometry and embossment
on the punch tip adherence of a model ibuprofen formulation. J. Pharm. Pharmacol. 2004, 56, 947–950. [CrossRef]
37. Tejedor, M.B.; Nordgren, N.; Schuleit, M.; Millqvist-Fureby, A.; Rutland, M.W. AFM colloidal probe measurements implicate
capillary condensation in punch−particle surface interactions during tableting. Langmuir 2017, 33, 13180–13188. [CrossRef]
38. Pingali, K.C.; Shinbrot, T.; Cuitino, A.; Muzzio, F.J.; Garfunkel, E.; Lifshitz, Y.; Mann, A.B. AFM study of hydrophilicity on
acetaminophen crystals. Int. J. Pharm. 2012, 438, 184–190. [CrossRef]
39. Wang, J.J.; Guillot, M.A.; Bateman, S.D.; Morris, K.R. Modeling of adhesion in tablet compression. II. Compaction studies using a
compaction simulator and an instrumented tablet press. J. Pharm. Sci. 2004, 93, 407–417. [CrossRef] [PubMed]
40. McDermott, T.S.; Farrenkopf, J.; Hlinak, A.; Neilly, J.P.; Sauer, D. A material sparing method for quantitatively measuring tablet
sticking. Powder Technol. 2011, 212, 240–252. [CrossRef]
41. Costa, N.F.; Paulo, M.G.; Diogo, H.P.; Pinto, J.F. Solving a sticking related tablet problem by multivariate statistics and computa-
tional tomographic analysis. Powder Technol. 2020, 367, 456–463. [CrossRef]
42. Freiberg, D.; Zavaliangos, A. Adhesion of powder onto tools during compaction: A numerical study. Powder Technol. 2022,
407, 117644. [CrossRef]
43. Dembele, M.; Hudon, S.; Simard, J.-S.; Abatzoglou, N.; Gosselin, R. Multivariate data analysis approach to tablet sticking on an
industrial scale: A qualitative case study of an ibuprofen-based formulation. Pharm. Dev. Technol. 2022, 27, 1093–1109. [CrossRef]
44. Boudoin, L.N.; Lambeth, G.H.; Llimatta, E.W.; Hu, P.C. High Content Sodium Ibuprofen Granules, Their Preparation and Their
Use in Preparing Non-Effervescent Solid Dosage Forms. U.S. Patent US9629809 B2, 25 April 2017.
45. Siesler, H.W.; Ozaki, Y.; Kawata, S.; Heise, H.M. (Eds.) Near-Infrared Spectroscopy. Principles, Instruments, Applications; Wiley-VCH:
Weinheim, Germany, 2002; 348p, ISBN 3-527-30149-6.
46. Watanabe, A.; Morita, S.; Ozaki, Y. A study on water adsorption onto microcrystalline cellulose by near-infrared spectroscopy
with two-dimensional correlation spectroscopy and principal component analysis. Appl. Spectrosc. 2006, 60, 1054–1061. [CrossRef]
47. Luck, W.A.P. Structure of Water and Aqueous Solutions. In Proceedings of the international Symposium, Marburg, Germany,
18–28 July 1973; Luck, W.A.P., Ed.; Verlag Chemie: Weinheim, Germany, 1974; pp. 248–284.
48. Curcio, J.A.; Petty, C.C. The near infrared absorption spectrum of liquid water. J. Opt. Soc. Am. 1951, 41, 302–304. [CrossRef]
49. Büning-Pfaue, H. Analysis of water in food by near infrared spectroscopy. Food Chem. 2003, 82, 107–115. [CrossRef]
50. Gupta, A.; Peck, G.E.; Miller, R.W.; Morris, K.R. Real-time near-infrared monitoring of content uniformity, moisture content,
compact density, tensile strength, and Young’s modulus of roller compacted powder blends. J. Pharm. Sci. 2005, 94, 1589–1597.
[CrossRef] [PubMed]
51. Mainali, D.; Li, J.; Yehl, P.; Chetwyn, N. Development of a comprehensive near infrared spectroscopy calibration model for rapid
measurements of moisture content in multiple pharmaceutical products. J. Pharm. Biomed. Anal. 2014, 95, 169–175. [CrossRef]
[PubMed]
52. Otto, M. Chemometrics. Statistics and Computer Application in Analytical Chemistry; Wiley-VCH: New York, NY, USA, 1998;
ISBN 978-3-527-34097-2.
53. Naes, T.; Isaksson, T.; Fearn, T.; Davies, T. A User-Friendly Guide to Multivariate Calibration and Classification, 2nd ed.; IM Publications
Open: Chichester, UK, 2017; Chapter 5. [CrossRef]
54. Pasquini, C. Near infrared spectroscopy: Fundamentals, practical aspects and analytical applications. J. Braz. Chem. Soc. 2003, 14,
198–219. [CrossRef]
55. Guy, A. Cellulose Microcrystalline. Available online: https://file.wuxuwang.com/hpe/HPE6/HPE6_59.pdf (accessed on 10
April 2023).
56. NETZSCH Proven Excellence. Laser Flash Technique. Available online: https://www.netzsch-thermal-analysis.com/en/
contract-testing/methods/laser-flash-technique-lfa-1 (accessed on 3 January 2019).
57. COMSOL Documentation. Available online: https://doc.comsol.com/5.6/docserver/#!/com.comsol.help.comsol/helpdesk/

helpdesk.html (accessed on 10 March 2021).
58. Geladi, P.; Macdougall, D.; Martens, H. Linearization and Scatter Correction for Near Infrared Reflectance Spectra of Meat. Appl.
Spectrosc. 1985, 39, 491–500. [CrossRef]
59. Roger, J.-M.; Mallet, A.; Marini, F. Preprocessing NIR spectra for Aquaphotomics. Molecules 2022, 27, 6795. [CrossRef]
60. Robert, G.; Gosselin, R. Evaluating the impact of NIR pre-processing methods via multiblock partial least-squares. Anal. Chim.
Acta 2022, 1189, 339255. [CrossRef]
61. Jackson, J.E. A User’s Guide to Principal Components, 1st ed.; John Wiley and Sons: New York, NY, USA, 1991. [CrossRef]
62. Geladi, P.; Kowalski, B.R. Partial least-squares regression: A tutorial. Anal. Chim. Acta 1986, 185, 1–17. [CrossRef]
63. Pharmaceutical Tableting Process. Available online: https://www.comsol.com/model/download/943861/models.nsm.
pharmaceutical_tableting_process.pdf (accessed on 10 April 2023).
64. Michrafy, A.; Ringenbacher, D.; Techoreloff, P. Modelling the compaction behaviour of powders: Application to pharmaceutical
powders. Powder Technol. 2002, 127, 257–266. [CrossRef]
65. Dos Santos, G.H.; Mendes, N. Combined heat, air and moisture (HAM) transfer model for porous building materials. J. Build.
Phys. 2009, 32, 203–220. [CrossRef]
66. Tariku, F.; Kumaran, M.K.; Fazio, P. Transient model for coupled heat, air and moisture transfer through multilayered porous
media. Int. J. Heat Mass Transf. 2010, 53, 3035–3044. [CrossRef]
67. Wang, J.; Christakis, N.; Patel, M.K.; Cross, M.; Leaper, M.C.A. A computational model of coupled heat and moisture transfer
with phase change in granular sugar during varying environmental conditions. Numer. Heat Transf. A-Appl. 2004, 45, 751–776.
[CrossRef]
68. Chao, B.T.; Trigger, K.J.; Zystra, L.B. Thermophysical aspects of metal cutting. Trans. ASME 1952, 74, 1039–1049. [CrossRef]
69. Grzesik, W. Advanced Machining Processes of Metallic Materials: Theory, Modelling, and Applications, 2nd ed.; Elsevier Science:
Amsterdam, The Netherlands, 2008. [CrossRef]
70. Bever, M.B.; Holt, D.L.; Titchener, A.L. The stored energy of cold work. In Progress in Materials Science; Chalmers, B., Christian, J.W.,
Massalski, T.B., Eds.; Elsevier: Amsterdam, The Netherlands, 1973; Volume 17, pp. 5–88. [CrossRef]
71. Von Böckh, P.; Wetzel, T. Heat Transfer: Basics and Practice; Springer: Berlin/Heidelberg, Germany, 2012. [CrossRef]
72. Koumbogle, K.; Gitzhofer, F.; Abatzoglou, N. Moisture transport coefficients determination on a model pharmaceutical tablet.
Processes 2022, 10, 254. [CrossRef]
73. Ostuka, M.; Koyama, A.; Hattori, Y. Real-time release monitoring for water content and mean particle size of granules in lab-sized
fluid-bed granulator by near-infrared spectroscopy. RSC Adv. 2014, 4, 17461–17468. [CrossRef]
74. Bogreckci, I.; Lee, W.S. The Effects of Soil Moisture Content on Reflectance Spectra of Soils Using UVVIS-NIR Spectroscopy. In
Proceedings of the 7th International Conference on Precision Agriculture, Minneapolis, MN, USA, 25–28 July 2004.
75. Lambert, J.H. Photometria Sive de Mensura et Gradibus Luminis, Colorum et Umbrae. 1760. Available online: https://www.
amazon.com/Photometrie-Photometria-Mensura-Gradibus-Luminis/dp/1279983108 (accessed on 10 May 2023).
76. Gou, D.; Li, Y.; An, X.; Yang, R. DEM modelling of particle fragmentation during compaction of particles. Powder Technol. 2022,
398, 117073. [CrossRef]
77. Buckner, I.S.; Friedman, R.A.; Wurster, D.E. Using compression calorimetry to characterize powder compaction behavior of
pharmaceutical materials. J. Pharm. Sci. 2010, 99, 861–870. [CrossRef]
78. Kadiri, M.S.; Michrafy, A. The effect of punch’s shape on die compaction of pharmaceutical powders. Powder Technol. 2013,
239, 467–477. [CrossRef]
79. Laity, P.R. Effects of punches with embossed features on compaction behaviour. Powder Technol. 2014, 254, 373–386. [CrossRef]
80. Bechard, S.R.; Down, G.R.B. Infrared imaging of pharmaceutical materials undergoing compaction. Pharm. Res. 1992, 9, 521–528.
[CrossRef]
81. Nezamabadi, S.; Ghadiri, M.; Delenne, J.-Y.; Radjai, F. Modelling the compaction of plastic particle packings. Comput. Part. Mech.
2021, 9, 45–52. [CrossRef]
82. Brunauer, S. The Adsorption of Gases and Vapors; Oxford University Press: Oxford, UK, 1943; Volume 1, p. 49, ISBN 10: 1443727369.
83. Christodoulou, C.; Sorensen, E.; García-Muñoz, S.; Mazzei, L. Mathematical modelling of water absorption and evaporation in a
pharmaceutical tablet during film coating. Chem Eng Sci. 2018, 175, 40–55. [CrossRef]
84. Schiros, T. Water-Metal Surfaces: Insights from Core-Level Spectroscopy and Density Functional Theory; Department of Physics,
Stockholm University: Stockholm, Sweden, 2006; p. 72, ISBN 978-91-7155-607-3.
85. Fernandes, F.W.; Campos, T.M.B.; Cividanes, L.S.; Simonetti, E.A.N.; Thim, G.P. Adsorbed water on iron surface by molecular
dynamics. Appl. Surf. Sci. 2016, 362, 70–78. [CrossRef]
86. Meng, S.; Wang, E.G.; Gao, S. Water adsorption on metal surfaces: A general picture from density functional theory studies. Phys.
Rev. B 2004, 69, 195404. [CrossRef]
87. Charlaix, E.; Ciccotti, M. Capillary Condensation in Confined Media. In Handbook of Nanophysics I: Principles and Methods, 1st ed.;
Sattler, K.D., Ed.; CRC Press, Taylor & Francis Group: Boca Raton, FL, USA, 2010; Volume 1, pp. 194–210.
88. Tanaka, H.; Hiratsuka, T.; Nishiyama, N.; Mori, K.; Miyahara, M.T. Capillary condensation in mesoporous silica with surface
roughness. Adsorption 2013, 19, 631–641. [CrossRef]
89. Philip, J.R. Adsorption and capillary condensation on rough surfaces. J. Phys. Chem. 1978, 82, 1379–1385. [CrossRef]
90. Naidich, Y.V.; Lavrinenko, I.A. Capillary adhesion forces between solid particles with an intermediate liquid layer at the contact.
Powder Metall. Met. Ceram. 1967, 6, 709–710. [CrossRef]
91. De Lazzer, A.; Dreyer, M.; Rath, H.J. Particle-surface capillary forces. Langmuir 1999, 15, 4551–4559. [CrossRef]
92. Paul, S.; Taylor, L.J.; Murphy, B.; Krzyzaniak, J.F.; Dawson, N.; Mullarney, M.P.; Meenam, P.; Sun, C.C. Mechanism and kinetics of
punch sticking of pharmaceuticals. J. Pharm. Sci. 2017, 106, 151–158. [CrossRef] [PubMed]
93. Aoki, S.; Danjo, K. Effect of tableting conditions on the sticking of tablet using ibuprofen. Yakugaku Zasshi 1998, 118, 511–518.
[CrossRef] [PubMed]
94. Kakimi, K.; Niwa, T.; Danjo, K. Influence of compression pressure and velocity on tablet sticking. Chem. Pharm. Bull. 2010,
58, 1565–1568. [CrossRef] [PubMed]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

Moisture Behavior of Powder During Tableting

Uploaded by

Copyright:

Available Formats

Moisture Behavior of Powder During Tableting

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Moisture Behavior of Powder During Tableting

Uploaded by

Copyright:

Available Formats

pharmaceutics

Academic Editors: Peter Timmins 1. Introduction

Pharmaceutics 2023, 15, 1652. https://doi.org/10.3390/pharmaceutics15061652 https://www.mdpi.com/journal/pharmaceutics

2. Materials and Methods

ρ app_w = ρ app ∗ w (1)

2.2.3. Powder Retention/Sorption Curve

2.2.4. Relative Tablet Density

2.2.5. Tableting Process

2.2.6. Tablet Surface and Peripheral

2.2.7. NIR-Penetration Depth

(a) (b) (c)

Primarily, the APAP

Pharmaceutics 2023, 15, x FOR PEER REVIEW 8 of 34

After pretreatment, Hotelling’s T2 in principal component analysis (PCA) [61] with a

2.2.9. Tablet Cohesion and Internal Friction Angle Density

3.1. Simulation Model for the Mechanical Behavior of the Powder

where f = (1 + α − α/cosβ) and α is a small number (typically 0.01–0.05) used to define a

3.1.2. Density-Dependent Drucker–Prager Cap Model Implemented in COMSOL Multiphyisics®

RD = RD0 ∗ JP−1 (21)

where Jp is the plastic volume ratio.

3.1.3. Nonlinear Elastic Law

3.2. Moisture and Heat Transport Simulation Model

g1 = −δp ∇( ϕpsat ( T )) (29)

where ρ is effective density (kg/m3 ), Cp is specific heat at constant pressure (J/(kg·K)), λ is

where ξ = 0.9 represents the irreversibly deformed heat fraction [18,20].

The heat generated by friction is defined as

q2 = υµS A σnn dA (36)

q2tablet = (1 − η) ∗ q2 , with θ = (1 − η) (38)

3.4. Finite Element Model

Figure 7. Hydrostatic yield stress, Pb vs. plastic volumetric strain.

4. Results and Discussion

4.3.2. Temperature Evolution and Distribution

4.3.2. Temperature Evolution and Distribution

• Measured surface and peripheral temperatures of the tablets

Supplementary Materials: The following supporting information can be downloaded at https://

57. COMSOL Documentation. Available online: https://doc.comsol.com/5.6/docserver/#!/com.comsol.help.comsol/helpdesk/

You might also like