F. Zhang et al.
Measurement 231 (2024) 114672
shape, and size (distribution), determine the final function of the crystal
product and affect the downstream process [6]. To obtain a crystalline
product with certain functions, it is necessary to predict, design, and
control the crystal properties through the crystallization processes,
which requires accurate crystal measurement technologies (CMTs).
Over the past two decades, many CMTs have been developed,
including in situ CMTs developed using process analysis technologies
(PAT). Instead of sampling, the crystals are measured directly inside a
crystallizer [7–10]. CMTs extract direct information such as the form,
shape, and size of the solid phase, or indirect information such as spectra
and images, while providing kinetic information related to nucleation
and crystal growth. This information indicates whether the tested crystal
satisfies the production requirements and can be further processed to
determine the impact of the operating conditions on the crystal prop­
erties. This enables the acquisition of crystals with specific properties by
adjusting the operating conditions based on the relationship between
the twos. For this reason, CMT is frequently combined with data analysis
technology [11]. Moreover, this information can be used to further
understand the crystallization process by estimating the parameters of
existing crystallization models [12] or establishing new crystallization
models. Therefore, CMT is of great significance for assessing crystal
quality, manufacturing crystalline products, and studying the crystalli­
zation processes.
This paper briefly describes the development of CMT and expands on
the principles, benefits, and drawbacks of various commonly used CMTs.
Then, by introducing specific applications of CMT, some problems and
challenges in the field are discussed, and development prospects are
proposed.
2. Overview of CMT
As mentioned above, various CMTs have been developed in recent
years to adapt to different crystal quality requirements. These CMTs can
be roughly split into off-line, on-line, and in situ according to the mea­
surement mode, as shown in Fig. 1. In situ CMTs can also be classified as
Fig. 1. Schematic diagrams of different types of measurement technologies.
2
invasive or non-invasive, depending on whether the instrument is
inserted into the crystallizer or measured by a window. Table 1 sum­
marizes the various CMTs and displays their measurement modes,
principles, measuring objects, advantages, disadvantages, whether
sample preparation is required, and special measurement conditions.
These methods can also be categorized according to the measurement
scale illustrated in Fig. 2. In fact, advances in CMT are mainly due to the
advancements in the measurement mode and scale, where the former is
more obvious. For clarity, more details on these technologies are sys­
tematically discussed according to the measurement mode (i.e., off-line,
on-line, and in situ).
2.1. Off-line CMT
Off-line CMT refers to the method of manually measuring a sample in
the laboratory or near an industrial site. There is a significant time delay
in this measurement mode [13], and the measurement is not performed
in the location where crystallization occurs. Therefore, off-line CMT is
generally used to measure the properties of crystalline products instead
of measuring the properties of crystals during crystallization. Early
CMTs were essentially off-line. They are widely used for the measure­
ment of single-crystal and powder samples owing to their unique ad­
vantages: (i) high resolution and accuracy, where atomic force
microscopy (AFM), X-ray diffraction (XRD), and electron microscopy are
used to study the crystal structure, surface, and crystallization kinetics
[14–16]; (ii) extraction of morphological information, such as tomog­
raphy and optical sectioning [17–20]; and (iii) data calibration, where
the results of off-line measurements are often used as a comparison
standard for on-line/in situ measurement results. Owing to these ad­
vantages, off-line CMT is widely used.
However, owing to its time delay, in practice, off-line CMT neither be
used to adapt to the demands of real-time monitoring of the dynamic
behavior of crystals nor introduced into the closed-loop control system.
Several commonly used off-line measurement technologies are
described below.
 F. Zhang et al.
Table 1
Overview of commonly used crystal measurement technologies.
CMTs Mode Principle/theory Advantage Disadvantage Properties Pretreatment Special condition Reference
measured

Sieve analysis Off-line The crystals pass through Low cost, simple operation and direct Small measuring range, time- Crystal size Drying Dry screen [21,22]
sieve mesh of varying sizes determination of size range consuming, and damaging effect on (distribution)
fragile particles
Coulter counter Off-line/ Different sizes of crystals Good repeatability Small measuring range and easy Crystal size Suspended in Electrolyte solution [23]
on-line cause different changes in blockage of small holes (distribution) electrolyte
resistance when they pass solution
through small holes
Laser diffraction Off-line, Scattering of light, Mie light High speed, good reliability, high Affected by crystal shape, when used as Crystal size Dispersion / [22,29]
on-line/ scattering theory or reproducibility, and wide an on-line/in situ measurement (distribution)
in situ Fraunhofer diffraction theory measurement range technique, the accuracy is affected by
the concentration of the suspension
X-ray diffraction Off-line, Bragg’s Law and Kinematic Fast measurement speed, high Requirements for the crystal size Crystal size and Grinding and / [33]
on-line/ diffraction theory precision, and non-destructive analysis polymorph sifting
in situ
Microscope Off-line, Direct observation Intuitiveness and ability to measure Not quantitative measurement Crystal size Drying, freezing, / [38]
on-line/ size and shape simultaneously (distribution) and staining, etc.
in situ crystal shape
Solid-state nuclear Off-line/ Nuclear analysis Distinguishes multi-component Expensive Polymorph Grinding / [39,40]
magnetic on-line samples
resonance
3

Infrared In situ/ Absorption of light Rich spectral peak information Used for surface analysis and Polymorph / Specific solvents, the /
spectroscopy on-line qualitative analysis, slower analysis solution concentration is
speed and a more complex operation generally 3–5 %
Near-infrared In situ/ Absorption of light High acquisition speed and non- Difficult to extract information Polymorph / A solute containing a [48]
spectroscopy on-line destructive, weak absorption based on specific functional group
overtones and combined bands, and
little or no sample preparation
Raman In situ/ Scattering of light High acquisition speed, non- Susceptible to fluorescence and Polymorph / Special solvent, [51]
spectroscopy on-line destructive, and no sample organic solvents preferably water
preparation; monitoring of both the
solid and liquid phases; easy
to interpret
Ultrasonic In situ/ Scattering or absorption of Can be applied to high-solid- Ultrasonic waves affect crystal Crystal size / / [55,56]
attenuation on-line ultrasonic waves concentration suspensions, allows both properties, the poor real-time (distribution)
method the solid and liquid phases to be performance
monitored
Focused beam In situ/ Laser backscattering Effective measurement of chord length The true size distribution cannot be Chord length / / [57]
reflectivity on-line and detecting the onset of nucleation, obtained directly distribution
measurement growth, and dissolution

Measurement 231 (2024) 114672

In situ imaging In situ/ Image segmentation, shape Intuitiveness, simultaneous Direct measurement of 3D crystal Crystal size / The solution [54,67–71]
and on-line on-line descriptor measurement of size and shape, fast information is rare and image quality is (distribution) and concentration should
image analysis and accurate segmentation, and poor owing to high solid concentration crystal shape not be too high (less
quantitative description and crystal movement than 8–10 %)
F. Zhang et al. Measurement 231 (2024) 114672

Fig. 2. The measurement ranges of various CMTs.

2.1.1. Sieve analysis
In the sieve analysis method, the crystalline sample is passed through
a series of standard sieves with varying sizes [21], obviously, the crystals
that pass through the sieve are smaller than the size of the sieve, and the
crystals remaining on the sieve are larger. Weighing the samples on each
sieve separately yield the crystal size distribution (CSD), which is
expressed as a percentage of the mass. This is the earliest particle size
distribution measurement method and can also be used to measure CSD.
The advantages of sieve analysis are its low cost and ease of opera­
tion [22]. Simultaneously, because the size of the sieves is known, the
CSD is directly estimated; that is, it is a CMT that directly acquires
crystal information. However, its disadvantages make its widespread use

Fig. 3. The structure diagram of the Coulter counter.

4
F. Zhang et al.
difficult. First, the minimum crystal size that can be measured by sieve
analysis is ≈50 μm [21]. Second, wear and collision during the sieving
process often significantly reduce the size of the crystal and thus affect
the results. Finally, even for off-line CMTs, sieve analysis is relatively
time-consuming.
2.1.2. Coulter counter
A Coulter counter, also known as the Coulter’s principle or resistance
method, is an off-line CMT for measuring particle size. A structural di­
agram is shown in Fig. 3. When particles in a solution travel through a
micro-channel, they replace the same volume of the solution and cause
instantaneous changes in the resistance between the two electrodes in
the constant-current circuit, resulting in a potential pulse. Based on the
relationship between the pulse signal, particle size, and number of
particles, the particle size distribution can be obtained indirectly. The
method was first used for blood cell counting and was then applied to
measuring size and counting other particles.
The benefits of Coulter counters include simple durability, low
power consumption, and the ability to detect individual particles.
Simultaneously, they not only accurately measure the particle size dis­
tribution of the material but also the absolute number and concentration
of particles [23]. However, owing to the special microchannel structure
of the Coulter counter, it can be directly used as an external circulation
loop; therefore, the Coulter counter is highly promising for the on-line
measurement of CSD.
However, the application of the Coulter counters in crystal mea­
surement has been concentrated in the last century owing to several
disadvantages. First, compared with biological particles, crystals are
more likely to clog small holes. Second, the lower measurement limit is
not sufficiently low; in reality, the minimum aperture of the tube is ≈60
μm, and the lower measurement limit is ≈1.2 μm. Therefore, Coulter
counters are far less widely used for crystal measurement than laser
diffraction (LD).
2.1.3. LD
LD is the most commonly used off-line CMT and is also the core
technology of the laser particle size analyzer widely used in the market
today. LD is a CMT based on the principle of light scattering. When a
laser beam passes through the dispersed crystal sample, the scattered
light intensity distribution is measured at a forward small angle to form
a scattering spectrum. The size distribution of the crystal sample is
Fig. 4. Schematic diagram of the laser diffraction.
5
Measurement 231 (2024) 114672
obtained indirectly by analyzing the scattering spectrum. Either Mie
light scattering theory or Fraunhofer diffraction theory is used in the
data analysis process to obtain the CSD from the scattering spectrum
[22,24]. When the crystal size is large, the scattering in the forward
small angle range is mainly diffraction, hence giving rise to the name of
the technology.
Although LD is more often used as an off-line technology for
measuring the size distribution of crystal samples in the laboratory, it
can also be used as an on-line measurement/in situ measurement tech­
nology to monitor the crystallization process [25]. A schematic of LD is
shown in Fig. 4.
LD technology has the following advantages: high speed, good reli­
ability, good reproducibility, and a wide measurement range [24–26].
These advantages make LD an important method for measuring CSD.
However, the CSD determined using LD is affected by the crystal shape.
Mie theory is based on the assumption that the crystal is spherical, which
induces a large error when the actual shape of the crystal is not spher­
ical. The size distributions of crystals with different shapes measured by
LD are different, and researchers [27] have studied the correlation be­
tween particle shape and particle size distribution [28–30]. Therefore,
when the crystal shape is known, ideally, it should be incorporated into
the optical model to obtain the correct CSD. In addition, when LD is used
as an on-line/in situ technology to measure the CSD in solid suspensions,
requirements for the concentration of solids exist [31]. At higher con­
centrations, the CSD measurement becomes less accurate. In industrial
crystallization processes, LD is more commonly used for the off-line
measurement of dry crystal samples because the solid concentration in
industrial crystallizers is usually an order of magnitude higher than the
required value.
2.1.4. XRD
XRD is a powerful method for studying crystal structure in the field of
crystallography and is mainly divided into single crystal XRD (SXRD)
and powder XRD (PXRD). The former is mainly used to determine the
position of atoms in crystals as well as bond lengths, bond angles and
other three-dimensional (3D) structural information [32], while the
latter is more often applied to phase analysis, that is, to distinguish
different substances or polymorphs in a sample. As the products ob­
tained from industrial crystallization are often in powder forms, more
widely used is PXRD [33]. XRD is used to obtain information about the
samples by analyzing their diffraction patterns. For crystals, the atomic
F. Zhang et al.
arrangement is long-range-ordered in 3D space, and the XRD pattern
only shows the diffraction peaks at specific positions. Therefore, the
diffraction patterns produced by the crystals reflect the distribution of
atoms inside the crystal. By identifying the location and intensity of
diffraction peaks in space, a qualitative and quantitative relationship
between XRD and the crystal structure can be established. Bragg’s law
describes the basic principle of diffraction and reveals the intrinsic
relationship between diffraction and crystal structure [34]. Using
Bragg’s law, the crystal plane spacing can be solved using X-rays of a
known wavelength, and crystal structure information can be obtained
indirectly. Meanwhile, the presence of a polymorph can be ascertained
based on anomalies in both the number and position of the diffraction
peaks within the diffraction pattern.
XRD has the merits of high speed, high precision, and non-
destructive analysis. The disadvantage of XRD is that it requires a
certain crystal size and usually measures the structure of large crystals
more accurately than that of small crystals.
Recently, XRD research has led to several advances in the field of
crystal measurement: (i) using XRD to determine the grain size (crystal
size) [35] and (ii) applying PXRD in crystallization processes to monitor
polymorphs in situ during crystallization [36,37]. The former is gov­
erned by Scherrer’s formula, which describes the relationship between
the grain size and half-peak width of the diffraction peak; the smaller the
grains are, the wider the peaks of XRD diffraction lines are. By contrast,
it is more concentrated. However, the scope of application of Scherrer’s
formula is still worth discussing. The latter is intended to convert PXRD
to on-line/in situ PXRD; however, the current application of in situ
PXRD are limited to small batches and microfluidic reactors [37].
2.1.5. Microscopy
Microscopes are optical instruments composed of a single lens or a
combination of several lenses and are mainly used to magnify small
objects that cannot be observed directly. In the field of crystal mea­
surements, microscopes are a vital tool for intuitively providing infor­
mation on crystal size and shape, and the technology is mainly divided
into optical and electron microscopes [38]. The measurement range of
an optical microscope is generally 0.4–150 μm, while an electron mi­
croscope can measure the size of a crystal from 1 nm to 1 μm, which also
leads to its high price. In most cases, this imaging method is mainly used
for observation, and measurements should only be regarded as
qualitative.
The greatest advantage of microscopy is its intuitiveness; it can
obtain both the size and shape information of the crystal. However, it
has the following problems: (i) the magnification range of optical mi­
croscopes cannot provide detailed morphological information on crys­
tals, whereas electron microscopes are expensive and require
pretreatment of a sample before the measurement process, which can
potentially cause errors [38]; (ii) microscopy measurements are quali­
tative, and images require further processing to obtain quantitative in­
formation about the crystals, which often leads to time delays.
2.1.6. Solid-state nuclear magnetic resonance spectroscopy
Solid-state nuclear magnetic resonance (ssNMR) spectroscopy is an
analytical technique that characterizes solid samples. In the field of
crystallography, ssNMR is a robust technique for distinguishing poly­
morphs and discerning crystalline and amorphous forms in complex
formulations. Consequently, it effectively addresses the limitations of
numerous CMTs that encounter challenges in accurately characterizing
diverse components within multi-component samples [39].
The techniques used in ssNMR can be classified into two categories:
static and magic angle rotation (MAS). The former has a low resolution
and limited applicability, thus making MAS the dominant ssNMR tech­
nology. The MAS technique involves placing a sample in a rotor (a cy­
lindrical container containing an air-driven turbine) and mechanically
rotating it around an axis tilted at 54.7356◦ (magic angle) relative to the
direction of the static magnetic field. This excites the nuclear spin and
6
Measurement 231 (2024) 114672
generates an ssNMR curve, thereby providing information on the in­
ternal structure of the sample for subsequent analysis [40]. The me­
chanical rotation of the sample along the magic angle effectively
mitigates anisotropic interactions, such as chemical shift anisotropy and
dipole–dipole coupling, enhancing site-specific resolution and sensi­
tivity. The spectral library obtained using ssNMR catalogues the multi­
ple peaks attributed to the distinct chemical shift ranges of various
structures, facilitating the facile discrimination of sample constituents
and enabling the measurement of polymorphs within a sample.
Despite its high cost, ssNMR is a non-destructive and quantitative
measurement method that provides more comprehensive sample infor­
mation, making it a valuable tool for research.
2.2. On-line CMT
Unlike the off-line measurement technology used to measure single
crystals and powders, on-line measurement technology measures the
crystals in the sample solution via automatic sampling and returns the
sample solution to the crystallizer after measurement. Some off-line
measurement technologies, such as LD and the in situ measurement
technologies discussed in the next section, can be used for on-line
measurement.
The most important feature of on-line measurement technology is the
combination of automatic sampling and ability to measure sample so­
lutions. Not only is the sampling time greatly shortened, but the short­
comings of the off-line CMTs are improved, and real-time measurement
is realized.
Two types of automatic sampling structures are common: flow cell
systems and external circulation loops. Flow cell systems are widely
used for crystal measurement during crystallization; Patience and
Rawlings [41] were among the early researchers to use flow cell systems
for crystal measurement. They combined flow cell systems with Photo­
microscopy and digital image analysis, by periodically stopping the flow
cell to capture a clearer image and then conducted a digital processing
step to extract the relevant information of NaClO3. Their work suc­
cessfully enabled the monitoring of transient changes in the crystal
shape using commercial image analysis software. An external circulation
loop is also important for automatic sampling in on-line crystal mea­
surement. Compared with flow cell systems, the external circulation
loop cannot control the flow speed of the sample solution, however, the
crystals in the external circulation system can move at a constant rate,
unlike the crystals in the mold, which move irregularly. When the solid
concentration of the slurry in the crystallizer is high, the crystals move at
a constant rate. The solid content in the slurry can be effectively reduced
using an external circulation loop, which is conducive to the application
of CMTs such as LD and image analysis techniques. Tung [42] described
a crystallizer with an external recirculation loop and an in-line mixer
that not only measures the size of the crystals, but also seeds them to
control the size of the crystals inside the crystallizer.
However, on-line measurement technologies still cannot completely
eliminate the time delay caused by sampling. Because the measurement
is not performed in the crystallizer, changes in the environment may
alter the properties of the measured crystals, and the sample solution
returned to the crystallizer may have unknown effects on the crystalli­
zation process.
2.3. In situ CMT
In situ CMT can effectively solve the two problems of on-line CMT,
because no sampling is required. The crystals suspended in the crystal­
lizer can be directly measured, whether by invasive or non-invasive in
situ measurement technology. The development of in situ CMT is
inseparable from the extensive application of PAT tools for crystal
measurement. PAT is defined by the U.S. Food and Drug Administration
(FDA) as “a system for designing, analyzing, and controlling
manufacturing through timely measurements (i.e., during processing) of
F. Zhang et al.
critical quality and performance attributes of raw and in-process mate­
rials and processes, to ensure final product quality” [7].
PAT not only focuses on the application of measurement technology
but also combines the processing and analysis of in situ measurement
data. The most representative method is a combination of in situ im­
aging technology and image analysis [43–45]. In addition, chemo­
metrics can be used to process in situ measurement data (usually
spectral data) to derive mathematical relationships between the mea­
surement data and crystal properties and to develop deconvolution or
calibration methods [46]. In situ CMTs commonly used for crystal
measurement are reviewed next.
2.3.1. Spectroscopy technology
Spectroscopy is widely used for the in situ measurement of solution
concentration [13]. Among the technologies, infrared (IR), near-infrared
(NIR) and Raman spectroscopy are used to measure the solid phase
(mainly polymorphs) in situ; therefore, they are widely used in the field
of crystallography.
IR spectroscopy, also known as mid-infrared (MIR) spectroscopy,
involves a different wavelength range than near-infrared spectroscopy,
that is, from 2500 to 25,000 nm. This spectral range provides a wealth of
information on organic matter owing to its ability to capture the peaks of
various functional groups. However, compared to NIR light, the energy
of MIR light is lower, and its penetration is insufficient; therefore, the IR
spectrum can only provide the surface characteristics of the measured
object. In addition, IR spectroscopy is typically used to detect certain
functional groups in molecules and is rarely used for quantitative
analysis. Compared with NIR spectroscopy, MIR spectroscopy has a
slower analysis speed and more complex operation, leading to its
gradual replacement by NIR spectroscopy in the field of crystal analysis.
The NIR region (780–2500 nm) is located between the red band of
visible light and mid-infrared region. NIR signals (spectra) result from
light absorption owing to the molecular vibrations (a combination of
overtones and fundamental vibrations) of hydrogen bonds (e.g., C–H,
N–H, and O–H) [47]. The principle of NIR spectroscopy applied to the
measurement of crystal forms is as follows: the NIR absorption wave­
length and signal intensity of different groups (such as methyl, methy­
lene, and benzene ring) or the same group in different chemical
Fig. 5. A simple schematic of in situ Raman spectroscopy.
7
Measurement 231 (2024) 114672
environments differ significantly; therefore, the NIR spectra of different
forms of crystals often have large differences and contain rich structure
and composition information. NIR spectroscopy is suitable for deter­
mining the structure and measuring the contents of hydrocarbon organic
compounds to obtain information on the crystal type. In the process of
study, by measuring the NIR spectrum of the crystal, the crystal form,
especially the polymorph information, can be quickly obtained through
the calibration model established by using the calibration sample and its
near-infrared spectral data by using the multivariate partial least-
squares regression algorithm and other methods [48].
The advantages of NIR are as follows [10,49,50]: (i) high acquisition
speed and non-destructive nature; (ii) weak absorption based on over­
tones and combined bands; and (iii) little or no sample preparation
required. The disadvantage is that the NIR spectra are characterized by
wide and overlapping absorption peaks and may have thousands of
wavelength variables. Assigning vibrations to specific functional groups
can be quite difficult, so establishing the relationship between NIR
spectra and crystal form is not easy. Although this shortcoming has been
addressed to a certain extent using chemometrics, the further develop­
ment of more efficient multivariate data analysis methods is required
[10].
Compared to the NIR spectrum, the Raman spectrum is not a light-
absorption spectrum but a light-scattering spectrum [13]. Fig. 5 shows
a simple schematic of in situ Raman spectroscopy. When the high-
intensity incident light of a laser source is scattered by molecules,
most of the scattered light has the same wavelength (color) as the
incident laser and does not provide useful information; this is called
Rayleigh scattering. However, a very small fraction (approximately 10-
9
) of the scattered light has a wavelength (color) that differs from that of
the incident light, where its wavelength varies according to the chemical
structure of the test sample (the so-called “scattered” substance). This is
called Raman scattering. Owing to these characteristics, Raman spec­
troscopy can be used to distinguish crystals of different structures and
thus to measure crystal polymorphism.
Raman spectroscopy has all the advantages of NIR spectroscopy, as
well as others. (i) It can simultaneously measure both solid (polymorph)
and liquid (concentration) phases. (ii) Because Raman spectroscopy
involves fundamental vibrations, it is generally less complex and has
F. Zhang et al.
high-resolution peaks, therefore, it is easier to interpret Raman data and
chemometrics are not necessarily required [51]. Raman spectroscopy
also has the following disadvantages: first, the Raman signal can be
hidden by fluorescence, so fluorescence is not conducive to Raman
measurements; and second, organic solvents can produce strong peaks
that interfere with the Raman spectrum of the solid phase [52], but
water does not produce peaks; therefore, Raman spectroscopy is often
applied to the crystallization process in water.
2.3.2. Ultrasonic attenuation method
Ultrasonic attenuation is a common method for measuring the crystal
size of slurries [53,54]. The measurement is based on the interaction of a
suspended crystal and an ultrasonic wave in a suitable solvent, using the
fundamental equations of mass, momentum, and energy to measure the
CSD. The ultrasonic generator emits ultrasonic waves of a certain fre­
quency and intensity that pass through the test area and reach the signal-
receiving end. When the crystal passes through the test area, the degree
of ultrasonic wave attenuation at the receiving end varies owing to the
varying absorption degrees of crystals of different sizes. The size dis­
tribution of the crystal is obtained according to the relationship between
the crystal size and the attenuation of the ultrasonic intensity, which is
generally given by models such as the ECAH theoretical model [55],
Harker model, and Temple model [56]. A diagram depicting the prin­
ciple behind the ultrasonic attenuation method is shown in Fig. 6.
A significant advantage of ultrasonic attenuation is its ability to
characterize crystals in highly concentrated systems, even when the
volume fraction of the solid phase in the suspension reaches ≈70 %.
Another advantage of ultrasonic attenuation is that it can measure the
concentration of solutions in addition to the size distribution of solids,
enabling the simultaneous monitoring of both solid and liquid phases,
same as Raman spectroscopy.
The disadvantage of the ultrasonic attenuation method in CSD
measurement is that the CSD is calculated by analyzing the information
obtained by a complex mathematical model, which often leads to a long
data processing time; thus, the real-time performance of the ultrasonic
attenuation method is relatively poor. In addition, ultrasound directly
affects the crystal properties, resulting in measurement errors.
Fig. 6. Principal diagram of ultrasonic attenuation method.
8
Measurement 231 (2024) 114672
2.3.3. Focused beam reflection measurement
Focused beam reflection measurement (FBRM) uses a rotating laser
beam to scan a crystal and record the back reflected laser. The intensity
sensor measures the length of the scanning path through the particle and
defines these lengths as chords [57]. Specifically, in the FBRM, a high-
speed spinning laser beam is focused and directed into the slurry
through a sapphire window mounted on the tip of a cylindrical probe.
When the laser beam intersects the edge of the particle, the back­
scattered light is detected by a detector installed within the same probe,
generating a rise signal in the circuit until it reaches the opposite edge of
the particle. The rise time combined with the tangential velocity of the
spinning laser beam determines the chord length. The scanning speed of
the laser beam determines the measurement range of chord lengths,
which is divided into fixed linear channels in the hardware. Each
channel records chord-length counts to generate a chord-length distri­
bution (CLD), which effectively captures 1D information about particles
that are nearly spherical in shape [58]. Fig. 7 shows a schematic of
FBRM.
As a technology based on the principle of laser back reflection, FBRM
can fingerprint the particle system distribution under in situ process
conditions and provide highly sensitive real-time tracking of particle
size, particle number, and shape changes without off-line sampling.
Another advantage of FBRM is that it can qualitatively detect the onsets
of nucleation, growth, and dissolution [11,59]. Therefore, FBRM is also
a powerful tool for monitoring crystal behavior.
However, direct tracking of the crystal size using FBRM is limited to
spherical crystals owing to the equivalence between CLD and CSD for
such crystals. For non-spherical crystals, a mathematical algorithm is
necessary to correct dimensions [51] and convert the raw CLD data
obtained by FBRM into CSD. This limitation is a drawback of FBRM
technology. In recent years, extensive research has been conducted to
establish the relationship between the CLD obtained by FBRM and the
CSD [58,60]. However, the development of the relevant algorithms
poses significant challenges.
2.3.4. In situ imaging and on-line image analysis
In addition to the several in situ measurement technologies described
F. Zhang et al.
Fig. 7. The schematic diagram of focused beam reflection measurement.
above, a key crystal measurement method that has developed signifi­
cantly in recent years is based on in situ imaging and on-line image
analysis. This technology is mainly used in the measurement of crystal
size (distribution) and crystal shape, where the latter is the more com­
mon application. Fig. 8 shows the basic process of measuring the crystal
shape in crystallization processes using in situ imaging equipment and
on-line image analysis technology. In situ crystal images contain abun­
dant crystal shape information, the basic principle of this method is to
segment the crystals in the images through on-line image analysis and
describe them quantitatively with appropriate shape descriptors.
Because the method includes imaging technology, it has all the ad­
vantages of microscopic technology, namely, intuitiveness and simul­
taneous measurement of the crystal size and shape. These advantages
benefit from the development of imaging equipment, mainly including
METTLER TOLEDO’s particle vision measurement (PVM) system, which
is an invasive in situ imaging equipment [61,62] and Easyviewer 400
[63]; Process image analyzer (PIA), Dutch Perdix [64] and in situ
Fig. 8. The basic flow of extracting and processing crystal morphological information during crystallization using in situ imaging equipment and on-line image
analysis technology.
9
Measurement 231 (2024) 114672
particle viewer (ISPV) by Messtechnik Schwartz GmbH; and Glax­
oSmithKline’s on-line microscopy systems [65]. These imaging devices
are described in detail in [54]. Other advantages of this method include
fast, accurate segmentation and the quantitative description of crystals
due to the development of on-line image processing technology. In
recent years, neural networks have become the most popular image
processing technology [66–71]. They have contributed to improving the
real-time and accuracy of crystal image analysis. Lu et al. [72] combined
the background difference method and local threshold method based on
graph cutting and proposed an effective particle segmentation method
that effectively eliminates the influence of shadows caused by particle
motion. Zhang et al. [73] constructed a deep learning framework based
on a deep convolutional neural network (DCNN) to classify the sucrose
crystallization image of the sucrose crystallization process of a sugar
mill. They compared it with the results of other methods, highlighting its
superior performance. Many similar studies exist [74–76], all of which
show the importance of superior image analysis techniques for crystal
F. Zhang et al.
shape measurement.
In situ imaging and on-line image analysis for crystal measurement
have two main disadvantages: first, the most widely used in situ imaging
devices provide 2D crystal images, while 3D imaging devices are rarely
used; and second, poor image quality owing to higher solid concentra­
tions (8 % [1], 10 % [77]) or crystal motion makes it difficult to separate
the crystal the crystal difficult to separate from the background.
3. Applications of CMTs in crystallization processes
In the past two decades, the above-mentioned CMTs have been
widely applied to crystallization processes. The applications are deter­
mined by two key factors: the measuring instrument used to acquire the
data and the data processing method used to convert the acquired data
into crystal information. Among the numerous applications of CMTs, it is
important to determine the crystal form, quantitatively describe the
crystal shape, and measure the crystal size (distribution) during
crystallization.
3.1. Determination of crystal form
The crystal forms that may occur during crystallization processes
include polymorphs, co-crystals, solvates, and clathrate compounds. A
simple diagram of them is shown in Fig. 9. They are determined by the
internal structure of the crystal, and different crystal forms often affect
the external shape of the crystal and the function of the crystal product
(such as drug efficacy and physical and chemical properties). Therefore,
detecting and determining the composition of distinct crystal forms
during the crystallization process is of paramount importance for com­
prehending their nucleation and growth kinetics, establishing and
optimizing models, and controlling the crystallization process to achieve
the desired crystal form.
3.1.1. Polymorphs
Polymorphs are substances that exist as two or more crystal struc­
tures. Polymorphs have the same chemical composition, but different
crystal structures because they adopt different packing arrangements or
conformations in the lattice [78]. The polymorph phenomenon is very
common in organic crystals, and it has been reported that more than 50
% of active APIs have polymorphs [79], so polymorph research has been
Fig. 9. Several different solid phase forms, (a) pure crystals; (b) polymorph of crystal; (c) solvates; (d) co-crystal; (e) clathrates.
10
Measurement 231 (2024) 114672
more applied conducted into the field of drugs polymorphs. As
mentioned earlier, different polymorphs such as CB tend to have
different properties. Various polymorphs may appear during crystalli­
zation without crystal seeds. To select a target polymorph, controlling
the crystallization process is necessary. However, the design of the
control scheme often requires monitoring of the formation and trans­
formation of the desired polymorph.
Over the past two decades, on-line/in situ CMTs have been widely
used to determine which polymorphs are present during the crystalli­
zation of various drugs. According to Ostwald’s rule of stages [80], for a
crystallization system, when a driving force is generated, the system
tends to minimize the free energy, resulting in the most stable crystal
form. It is also possible that metastable polymorphs first nucleate and
grow, and then stable polymorphs nucleate and grow, accompanied by
the dissolution of metastable polymorphs [81]. The nucleation and
growth mechanisms of each substance ultimately lead to the formation
of different polymorphs and the transitions between various polymorphs
[82]. From the above description, the type of polymorphs produced by a
substance during its crystallization process is determined at the nucle­
ation stage. If the nucleation process can be measured on-line/in situ to
determine which polymorphs occur during nucleation, and the nucle­
ation mechanism of these polymorphs can be studied to control nucle­
ation, then polymorph selection can be achieved. However, nucleation is
a phenomenon that occurs on the nanoscale or even molecular level,
using traditional CMTs to directly measure the polymorph during the
nucleation process is difficult; higher-resolution instruments are often
necessary to study the nucleation process. With L-glutamic acid as the
object, Lai et al. [78] used FBRM to on-line monitor the change of in
crystal size in a mixed suspension mixed product removal (MSMPR)
crystallizer under different varying residence times and temperatures
and determined the nucleation and mutual conversion processes of the
two crystal forms of L-glutamic acid. Their results showed that, at a low
temperature (25 ◦ C), a long residence time is needed to dissolve the
metastable crystals and nucleate and growth grow the stable crystals,
which provides a research direction for polymorph selection. Driessche
et al. [83] used time-resolved frozen transmission electron microscopy
(TEM) to image the nucleation of protein glucose isomerase crystals,
revealing the nucleation pathway leading to two polymorphs and one
gel state with molecular resolution; they proved that the selection of
polycrystalline forms occurs in the early stage of nucleation, providing a