Journal of Membrane Science 564 (2018) 682–690

Contents lists available at ScienceDirect

Journal of Membrane Science

journal homepage: www.elsevier.com/locate/memsci

Continuous production of drug nanocrystals by porous hollow fiber-based T

anti-solvent crystallization
Xinyi Zhoua,1, Xuan Zhua,1, Bing Wangb, Jingchao Lic, Qiuhong Liua, Xuemin Gaoa,
⁎ ⁎
Kamalesh K. Sirkard, , Dengyue Chena,
a
School of Pharmaceutical Sciences, Xiamen University, Xiamen, Fujian 361102, China
b
Key Laboratory of Urban Pollutant Conversion, Institute of Urban Environment, Chinese Academy of Sciences, Xiamen, Fujian 361021, China
c
College of Chemistry and Chemical Engineering, Xiamen University, Xiamen, Fujian 361005, China
d
New Jersey Institute of Technology, Otto York Department of Chemical, Biological and Pharmaceutical Engineering, Newark, NJ, USA

A R T I C LE I N FO A B S T R A C T

Keywords: Nanotechnology is being utilized to develop advanced concepts in drug delivery systems including production of
Drug nanocrystals nanodrugs such as, nanocrystals and nanosuspensions. Continuous crystallization methods are of increasing
Griseofulvin interest and porous hollow fiber membrane (HFM) based modules have been recently utilized as an anti-solvent
Anti-solvent crystallization crystallizer. The porous hollow fiber anti-solvent crystallizer (PHFAC) based method has been modified here to
Porous hollow fiber membrane
produce continuously nanocrystals of Griseofulvin (GF). In the PHFAC device, deionized water introduced from
the HFM bore into the shell side is used as the anti-solvent; acetone containing dissolved GF is introduced into
the shell side of the HFM module as the drug solution. Water, the anti-solvent, mixed vigorously with the drug
solution leading to drug crystallization and the drug crystals were separated by vacuum filtration and freeze-
dried. The experimental conditions were varied to control the particle size, size distribution and the appearance
of the nanoparticles. The properties of the drug nanocrystals were characterized via Transmission Electron
Microscopy (TEM), Scanning Electron Microscope (SEM), Dynamic Light Scattering (DLS), Raman Spectroscopy,
Energy Dispersive X-rays Spectroscopy (EDX), Differential Scanning Calorimetry (DSC), FT-IR Spectrometer, X-ray
Diffraction (XRD); drug dissolution tests were also implemented. Drug nanocrystals as small as 86.4 nm were
produced under modest pressure and temperature conditions in a controllable and continuous manner.

1. Introduction
Due to their small size, nanoparticles possess unique capabilities,
such as volume effect, quantum size effect, surface effect, macroscopic
quantum tunneling effect and so on and are therefore widely used in
various fields. In pharmaceutical area, drug nanoparticles have at-
tracted increasing interest from researchers and industry for their dis-
tinct advantages. For instance, they can be passively targeted on liver,
spleen, bone marrow and other organs and tissues to cure the disease.
Nanoparticle drugs sized under 100 nm have biological permeability
through tumor vascular walls and hence are able to achieve a curative
effect. Moreover, nanoparticles can increase the efficacy of antibiotics
and enhance the potency of antifungal and antiviral agents in the
treatment of intracellular bacterial infections.
For drugs with poor water solubility and dissolution rate, the sta-
bility and bioavailability of oral preparations can be drastically
improved with drug nanoparticles for the purpose of protecting pep-
tides, vaccines and other drugs from clearance in the digestive tract [1].
Due to rapid development of nanoscience and nanotechnology, synth-
esis of drug nanocrystals has been widely applied to drug delivery
system (DDS) [2]. Nanonization of the drug is a desirable method to
improve bioavailability and efficacy in order to enhance their ther-
apeutic effect [3]. A number of size reduction techniques were devel-
oped or optimized to produce stable drug nanocrystals [4].
Current approaches in production of drug nanoparticles involve
either top-down or bottom-up methods [5]. For top-down methods,
drug particles are broken down to lower size particles by methods, such
as pearl milling, high pressure homogenization etc.; these have greater
utility for drugs having a high degree of crystallinity [6–8]. For bottom-
up methods of direct particle formation, a classical precipitation process
of ‘via humida paratum’ (VHP) is developed where drug dissolved in a
solvent is precipitated by mixing with a non-solvent. This method

⁎
Corresponding authors.
E-mail addresses: sirkar@njit.edu (K.K. Sirkar), dchen@xmu.edu.cn (D. Chen).
1
X. Zhou and Dr. X. Zhu contributed equally to this work and should be considered as co-first authors.

https://doi.org/10.1016/j.memsci.2018.07.082
Received 10 April 2018; Received in revised form 12 July 2018; Accepted 27 July 2018
Available online 29 July 2018
0376-7388/ © 2018 Elsevier B.V. All rights reserved.
X. Zhou et al.
succeeds in producing crystalline drug nanoparticles while requiring
strict control of the process [9]. One of the common methods, super-
critical fluid (SCF) technology, solubilizes the solute in the supercritical
fluid first and subsequently precipitates it via supersaturation through
rapid expansion to obtain virtually contaminant-free particles [10,11].
In another technique, the spray freeze drying (SFD), the method in-
volves the steps of droplet generation, freezing and sublimation drying
[2,12,13]. However, high expenditure, intensive labor, demand of
material and risks of dust explosion are disadvantages of these techni-
ques [14]. Confined liquid impinging jets (CLIJ) has been used to
produce particles in the submicron to nano range where precipitation
takes place when a jet of drug solution impinges a jet of anti-solvent
from two opposing nozzles [15]. But secondary crystallization and
further growth may be induced by incomplete depletion of solute
during nucleation and crystal growth. A precipitation technique named
as in situ micronization has been developed to produce the drug na-
noparticles in a relatively simple and modest way [16]; however, it may
result in larger particles and poor control of crystal size. The frequently
utilized method of rapid expansion of supercritical solutions (RESS)
enables particle formation due to rapid supersaturation by applying
rapid phase change of supercritical fluids. Supercritical Anti-Solvent
(SAS) method [17–19] induces precipitation/crystallization of solutes
by saturating the polar liquid solvent with carbon dioxide under su-
percritical conditions [11]; such a process would require critical ex-
perimental conditions such as the generation of high pressure condition
for supercritical CO2. Moreover, this method is a batch process with low
output levels and questionable stability of the drug quality, which are
against the requirements for modern pharmaceutical production. Nano
spray drying invented by Büchi® employs a vibration mesh spray
technology for fine droplet generation and is ideal for heat-sensitive
biopharmaceutical products [20].
Meanwhile, porous hollow fiber membrane-based techniques were
being developed to implement continuous phase-change processes. A
continuously flowing solution of BaCl2 on the shell-side permeated into
a solution of K2SO4 flowing in the tube-side of a hollow fiber membrane
device to achieve precipitation of BaSO4 particles via chemical reaction
[21–23]. This configuration is prone to the possibility of plugging of the
hollow fiber bore unless special designs are made and precautions are
taken. Instead of a chemical reaction-based precipitation, Zarkadas
[24], Zarkadas and Sirkar [25] utilized hollow fiber devices to imple-
ment anti-solvent crystallization. They found that it was more useful to
implement crystallization in the shell-side of the module so that chance
of flow-passage blockage was much less due to many directional free-
doms of flow. The specific configuration was adopted in which drug-
containing feed solution flowing in the hollow fiber bore permeated
into the shell side where the anti-solvent was flowing; thus crystal-
lization took place in the shell-side.
Recently, Chen et al. [26] reported a convenient anti-solvent crys-
tallization technique to synthesize continuously polymer-coated drug
particles using a porous hollow fiber membrane device. Interestingly,
the first part of this study [26] described PHFAC-based continuous
crystallization of micron size GF crystals depending on various condi-
tions that produced GF crystals with the median size from 1.61 µm to
11.83 µm. They had used an alternate arrangement for shell-side crys-
tallization: the anti-solvent water was introduced from the hollow fiber
bore into the shell-side having the flowing GF-in-acetone solution
(Fig. 1(a)). The introduction of the anti-solvent water through tiny
membrane pores into the acetone-based drug solution in the shell-side
of the HFM module ensured a high-mixing zone in the microenviron-
ment, which further provided a suitable condition for the crystallization
of small drug particles.
This PHFAC technique and the HFM device were also used suc-
cessfully to produce polymer-coated submicron as well as nanoparticles
of silica in a controllable way [27]. These successes suggested the
possibility of production of even smaller drug nanocrystals by con-
trolling the nucleation and growth condition of crystals as well as the
683
Journal of Membrane Science 564 (2018) 682–690
hydrodynamic optimization of the module [28].
To downsize the production of micron size GF drug crystals into
nanocrystals of GF, we have redesigned the membrane module and
investigated the hydrodynamic and crystallization conditions to
achieve controlled and continuous synthesis of drug nanocrystals of GF
at ambient pressure and temperature. The GF nanocrystals produced
have been characterized extensively and the enhancement of their
dissolution characteristics was also explored.
2. Experimental methods
2.1. Materials
Griseofulvin (purity 99.2%) was obtained from
Yuanchenggongchuang Technology Co. Ltd (Wuhan, Hubei, China).
Acetone (analytical purity) was purchased from Xilong Chemical Co.
Ltd (Guangzhou, Guangdong, China). Deionized water was produced by
multi-function ultrapure water system (Unique-R20, Ruisijie Water-
purification Technology Co. Ltd, Xiamen, Fujian, China). All reagents
were of analytical grade and used as received.
2.2. Apparatus and procedures
The experimental apparatus schematically shown in Fig. 1(b) was
designed for anti-solvent induced precipitation. The PHFAC system,
fabricated to perform the experiment, consisted of three sections: inlet
system; crystallization system; vacuum filtration system. The inlet
system had two parts, with the first part generating the drug solution
stream and the other part generating the anti-solvent stream. The so-
lution containing GF was prepared by stirring a suspension of the as-
received GF crystals in acetone in a conical flask on a magnetic stirring
apparatus until the drug was fully dissolved in the solvent. Upon
complete dissolution, the GF solution was passed into the shell side of
the crystallization system by a peristaltic pump (YZ1515x, Shenchen
Pump Industry Co. Ltd, Baoding, Hebei, China). Meanwhile, water, the
anti-solvent, was pumped into the lumen side of the PHFAC system by
another peristaltic pump at a fixed flow rate at a slightly earlier time. In
the PHFAC system, the left end of the tube side (shown in Fig. 1(b)) was
blocked by epoxy resin. This prevented water from flowing out from the
tube side end and created the pressure difference between the tube and
the shell side to push the drug solution out to the shell side bulk from
the outside of the hollow fiber membrane surface. The PHFAC module
shell side was formed by a fluorinated ethylene propylene (FEP) tubing
with an outer diameter (OD) of 19 mm and an inner diameter (ID) of
17 mm. The effective tube length was 26 cm, which was also the length
of each of the 14 polyvinylidene fluoride (PVDF) porous hollow fiber
membranes (pore size 0.1–0.15 µm; porosity, 0.75) having an OD of
2.5 mm and ID of 1.8 mm. As shown in Fig. 1(b), PHFAC module was
placed at an angle of 15° to the horizontal to maximize the contacting
area for the drug solution and anti-solvent in the shell side along the
fiber length direction.
The perspective view of a single porous hollow fiber membrane
shown in Fig. 1(a) illustrates the precipitation process in the drug so-
lution flowing on the shell-side induced by the anti-solvent. In the shell
side, the drug solution contacted uniformly with the anti-solvent which
penetrated out from the pores (0.1–0.15 µm) in the hollow fiber walls.
The drug crystals along with the excess solution flowed out to the outlet
of the device continuously and subsequently to the vacuum filtration
system (1000 mL, Tianjin Jinteng Experimental Equipment Co. Ltd,
Tianjin, China); this system allowed collection of the particles through
0.1 µm pore-sized membrane filters (VVLP04700, Merck Millipore Ltd,
County Cork, Ireland), which were freeze-dried prior to various char-
acterization steps.
X. Zhou et al.
Fig. 1. (a) Perspective view of single porous hollow fiber in the tube side. (b) Schematic of experimental installation of the porous hollow fiber anti-solvent
crystallizer (PHFAC).
2.3. Experimental operating conditions
Since the feed drug solution exerted a significant influence on a
variety of results, a number of items were varied. The ratios of the flow
rates of the anti-solvent and the drug solution contacting in the shell-
side mixing zone played a significant role in the procedure. The drug
concentration in the solution was also a critical parameter that had
impact on the formation and size of the drug particles. An amount of
0.8 g of as-received GF was placed in a flask containing 22 mL of the
solvent mixture. The solvent mixture was prepared with acetone and
deionized water in the proportion of 10:1. Under continuous stirring on
the magnetic stirring apparatus, the drug solution changed from a
suspension to a solution. In order to explore the formation of the finest
droplets, the flow rates of the anti-solvent and the drug solution were
set at different ratios and the drug concentrations were designed in a
gradient fashion.
2.4. Particle characterization
2.4.1. TEM
The morphology and appearance of GF particles was characterized
by transmission electron microscopy (TEM, Tecnai G, Spirit, FEI, Hong
Kong). After sonication, the samples were well dispersed and subse-
quently deposited on carbon coated copper grids for characterization.
2.4.2. SEM
The surface morphology of drug nanocrystals was observed using a
scanning electron microscope (SEM, Supra 55, Zeiss, Oberkochen,
Baden-wurttemberg, Germany). The drug nanocrystals were first dis-
persed in deionized water; they were subsequently dropped onto the
surface of a conducting resin and dried in air. Finally, the samples were
sputtered with gold for 30 s under vacuum for observation.
2.4.3. DLS
The particle size distribution was measured by a dynamic light
scattering (DLS) instrument (Nano-ZS, Malvern Instruments Ltd,
Worcestershire, UK). The samples were prepared by diluting in deio-
nized water and analyzed at 25 °C, with a dispersant refractive index of
1.33.
2.4.4. Raman spectroscopy
The molecular structure of the GF particles was determined with a
confocal Raman system (Xplora, Horiba, Japan). The intensity of laser
power was set at 10 mW, with the incident laser wavelength at 780 nm.
Raman spectra were recorded over the range of 0–3500 cm−1.
684
Journal of Membrane Science 564 (2018) 682–690
2.4.5. Energy dispersive X-rays spectroscopy
EDX spectroscopy analysis (S-4800, Hitachi, Japan) was applied to
confirm the composition of the elements for as-received drug powder
and the crystallized drug nanocrystals.
2.4.6. XRD
X-ray diffraction (XRD) measurements were conducted on an Ultima
IV X-ray diffractometer (Rigaku, Matsubara-cho, Tokyo, Japan) with
a Cu-Kα radiation source (λ = 0.15418 nm) at room temperature.
The voltage and current were set at 40 kV and 30 mA. Placed in a
glass sample holder, samples were scanned over the 2θ range of
5–40° at a scanning rate of 15°/min with a step size of 0.02°.
2.4.7. DSC
Differential scanning calorimetric (DSC) analysis was performed
using a Synchronous TG-DSC Thermal Analyzer (STA 449 F5
Jupiter, Netzsch, Bavaria, Germany). The sample typically weighing
about 10 mg was placed in an alumina crucible and heated from
75 °C to 250 °C at a heating rate of 10 °C/min under nitrogen at-
mosphere with a constant flow rate of 20 mL/min. An empty cru-
cible was utilized as a reference.
2.4.8. FT-IR
A FT-IR spectrometer (Bruker, Germany) was used to test whether
the chemical structure of the obtained drug crystals had changed. The
mixture of the samples and potassium bromide were compressed into
thin pellets and scanned over the range of 400–4000 cm−1.
2.4.9. In vitro dissolution testing
Dissolution tests were conducted in triplicate by the paddle method
using a RC806 dissolution tester (Tianjin Tianda Tianfa Technology Co.,
Ltd., Tianjin, China) to investigate whether there could be distinction
between the dissolution rate of drug nanocrystals and as-received drug
powder. The drug nanocrystals as well as the as-received GF powder
were carefully weighed and prepared for tests both having a weight of
15 mg. Glass vessels with a maximum nominal volume of 1 L were used
with 500 mL pH 6.8 phosphate buffer (37 ± 0.5 °C) as the dissolution
medium; the paddles were rotated at a speed of 100 rpm. At 5, 10, 20,
30, 40, 50, 70, 90, 110, 130 min time intervals, 5 mL samples were
taken for the UV test and replaced with another 5 mL fresh medium.
Passed through a 0.45 µm filter (Tianjin Jinteng Experimental
Equipment Co. Ltd, Tianjin, China), the absorbance value of samples
was determined by a UV-1750 ultraviolet spectrophotometer
(Shimadzu Suzhou Instruments Mfg. Co., Ltd, Japan) at a wavelength of
292 nm.
X. Zhou et al. Journal of Membrane Science 564 (2018) 682–690

Fig. 2. TEM images of (a) synthesized drug nanocrystals by PHFAC under the condition of flow rate at 6.8 mL/min in tube side and 13.6 mL/min in shell side with
drug concentration at 0.040 g/mL, (b) as-received pure GF powder.

3. Results and discussion
3.1. Continuous synthesis of drug nanoparticles
Griseofulvin crystals were first precipitated in the PHFAC device by
anti-solvent induced crystallization in the shell side. During this pro-
cess, DI water in the tube side penetrated through numerous tiny pores
(0.1–0.15 µm) on every hollow fiber wall; this caused rapid solubility
reduction of the drug in every tiny mixing zone on the shell side which
rapidly led to the formation of drug nanocrystals.
3.1.1. TEM, SEM and EDX
TEM results of the produced drug nanocrystals are shown in
Fig. 2(a). These drug crystals were produced using 4.33% (by wt.) so-
lution of GF in acetone for the following flow rates of the acetone so-
lution and the anti-solvent water: 13.6 mL/min and 6.8 mL/min. The
TEM image of Fig. 2(a) shows that relatively spherical drug particles
were synthesized successfully with a size of around 200 nm. The TEM
image of Fig. 2(b) shows that the original particle size of the as-received
pure GF powder was about 3–4 µm, which is more than 10 times the
size of the synthesized drug nanoparticles obtained continuously.
SEM images of Fig. 3(a) and 3(b) identify the uniform distribution of
drug particles with some crystalline structure. Besides, the morphology
of the GF powder shown in Fig. 3(c) seemed to have some agglom-
eration. SEM-EDX spectroscopy analysis in Fig. 4(a) and 4(b) further
identify the composition elements for the as-received sample and the
crystallized drug product, respectively. In both images, the composi-
tions of particles were identical vis-a-vis carbon, hydrogen, oxygen and
chlorine, which demonstrated that the successfully prepared particles
were nano-sized GF crystals.
3.1.2. Particle size characterization
As shown in Fig. 5, the peak of particle size distribution (PSD) of the
obtained nanocrystals was at 295.6 nm by DLS analysis (we will later
show that it can go as low as 86.4 nm), which was considerably smaller
than that of the as-received drug GF powder whose peak was at
3403.0 nm. The smaller particle size and PSD for synthesized drug
particles could be due to the following reasons. Firstly, the well dis-
tributed submicron-sized pores (0.1–0.15 µm) on the hollow fiber
membrane walls (with 75% porosity) created numerous tiny and uni-
form microenvironments in the shell-side when the penetrated anti-
solvent contacted the drug solution. The small and numerous mem-
brane pores generated tiny mixing zones that provided the required
conditions for nucleation and growth of drug nanocrystals. Moreover,
the short effective length of the PHFAC and continuous flow in the
device could decrease the residence time that further helped shorten the
crystal growth in the crystallization process, eventually leading to a
small crystal size of the drug.
3.1.3. Composition analysis
The results of Raman spectroscopy in Fig. 6 show that the main
peaks of the gained drug nanocrystals were at 2950.62, 1622.55,
1343.43, 648.689 and 375.38 cm−1, respectively which are in agree-
ment with the characteristic peaks of the as-received crude drug. These
results indicated that the synthesized nanocrystals presented no trans-
formation in the peak position over the shift range from 0 to

Fig. 3. SEM images of (a) synthesized drug nanocrystals by PHFAC under the condition of flow rate at 6.8 mL/min in both tube side and shell side with drug
concentration at 0.040 g/mL on a small scale, (b) synthesized drug nanocrystals by PHFAC under the condition of flow rate at 6.8 mL/min in both tube side and shell
side with drug concentration at 0.040 g/mL on a large scale, and (c) as-received pure GF powder.

685
X. Zhou et al. Journal of Membrane Science 564 (2018) 682–690

Fig. 4. EDX spectroscopy analysis results of (a) obtained GF nanocrystals under the condition of flow rate at 6.8 mL/min in both tube side and shell side with drug
concentration at 0.040 g/mL, (b) as-received pure GF powder.

Fig. 5. Particle size distribution (PSD) of synthesized drug nanocrystals by PHFAC under the condition of flow rate at 6.8 mL/min in both tube side and shell side with
drug concentration at 0.040 g/mL and as-received pure GF powder.

3500 cm−1; the nanocrystals precipitated from the drug solution re-
tained the original structures of the molecular groups.
FT-IR analysis was utilized to examine the chemical structure of the
obtained crystals as well. As shown in Fig. 7, the two characteristic
peaks at 1665 and 1707 cm−1 were derived from the stretching vibra-
tion of the carbonyl groups (vC˭O). Meanwhile, the peaks at 1463, 1502,
1584, 1614 cm−1 were derived from the aromatic C˭C stretching vi-
bration (vC˭C) and C-H stretching vibration (vC-H) contributed to the
peaks close to 2995 cm−1. It indicated that the chemical structures of
the gained nanocrystals were stable as all the characteristic peaks were
relatively identical to those of the as-received powder.
3.1.4. Crystallinity analysis
XRD patterns for drug nanocrystals along with as-received drug
powder are shown in Fig. 8. The sharp peaks in the patterns confirmed
the crystalline forms of both as-received drug powder and drug nano-
crystals. It is obvious that the crystalline peaks of drug nanocrystals
(86.4 nm) were in agreement with those of as-received drug powder.
Meanwhile, the reduced intensity of crystalline peaks for drug nano-
crystals may be attributed to the reduction in the particle size.
DSC thermal curves for as-received drug powder and obtained drug
nanocrystals are presented in Fig. 9. As can be seen, a melting tem-
perature of drug nanocrystals (86.4 nm) was located at 219.2 °C,
slightly lower than that of as-received drug powder at 222.1 °C.

Fig. 6. Raman spectroscopy results for the obtained GF nanoparticles under the condition of flow rate at 6.8 mL/min in both tube side and shell side with drug
concentration at 0.040 g/mL and as-received GF powder.

686
X. Zhou et al.
Fig. 7. FT-IR results of the obtained GF nanocrystals under the condition of flow rate at 6.8 mL/min in both tube side and shell side with drug concentration at
0.040 g/mL and as-received GF powder.
Fig. 8. XRD results of the obtained GF nanocrystals under the condition of flow
rate at 6.8 mL/min in both tube side and shell side with drug concentration at
0.030 g/mL and the as-received GF powder.
Fig. 9. DSC results of the obtained GF nanocrystals under the condition of flow
rate at 6.8 mL/min in both tube side and shell side with drug concentration at
0.030 g/mL and the as-received GF powder.
Regarding the drug nanocrystals, the endothermic peak did not notably
alter, indicating high crystallinity consistent with that of as-received
drug powder.
3.1.5. Dissolution testing
Smaller particles have a higher surface area. Since particle dis-
solution rate is proportional to the particle surface area, smaller parti-
cles should have a higher dissolution rate. Equilibrium solubility is also
increased as the particle size is reduced. Therefore, dissolution rates of
the nanoparticles formed should be higher than those of the larger
particles of as-received drug. The dissolution profiles for the as-received
drug powder and the synthesized drug nanocrystals are shown in
Fig. 10. The results suggest that over 12.4% of the as-received drug
(3–4 µm) got dissolved in the medium in around 130 min while the
687
Journal of Membrane Science 564 (2018) 682–690
Fig. 10. Dissolution profiles of as-received powder and synthesized drug na-
nocrystals under the condition of flow rate at 6.8 mL/min in both tube side and
shell side with drug concentration at 0.030 g/mL in PBS (phosphate buffer
saline) at pH 6.8.
dissolution rate of the nanocrystals (86.4 nm) increased to 54.0% in the
same period. The faster dissolution rate of drug nanoparticles was due
primarily to the increased specific surface area of obtained nano-
particles leading to much larger contact area with dissolution medium
compared with the crude micron-sized drug powders. Smaller nano-
crystals would get dissolved at even higher rates.
3.2. Variation of flow rate combinations
A series of flow rate combinations for the drug-containing solvent
and the anti-solvent (Table 1) were selected to study their effect on the
drug crystallization process and particle size in PHFAC. Shell-side
Reynolds number was calculated using the formula of Re = ρvd/μ. Here
ρ is the fluid density, v is the bulk velocity in the module, d is the
characteristic dimension of the shell side channel and μ is the liquid
viscosity. Reynolds number under the experimental conditions of F1, F2
and F3 was at 4.08, 7.00 and 10.10 respectively, showing a low Rey-
nolds number laminar flow in the module. The results displayed in
Table 1 and Fig. 11 show that the mean size of the attained drug na-
noparticles was 190.3 nm under experimental condition F2, which was
also in accordance with the result shown in Fig. 12(b). Under condition
F3, the nanoparticle size decreased to around 140 nm, as revealed in
Table 1 and Fig. 12(c).
As the total flow rate of the drug solution in the shell side (sum of
the flow rates of the drug solution on the shell-side and the tube-side
anti-solvent) increased, the particle size decreased due to the following
reasons. The increasing total flow rate due to an increased drug solution
flow rate for a constant feed solution flow rate introduced an increasing
level of supersaturation for drug, which resulted in high rates of nu-
cleation and correspondingly less amount of material available for
further growth. In addition, the shell-side residence time decreased
X. Zhou et al.
Table 1
Particle size under variation of the ratio of the flow rates of the anti-solvent and the drug solution.
Experimental Tube-side flow rate (mL/min) Shell-side flow rate (mL/min)
F1 6.8 6.8
F2 6.8 13.6
F3 6.8 20.4
Fig. 11. Particle size distribution (PSD) under conditions of F1, F2, and F3.
which reduced the opportunity for growth of the formed crystals.
A comparison of the particle size distribution and the TEM images
under various conditions show that the drug nanoparticles seemed to
remain crystalline with the same appearance and morphology.
Consequently, it can be stated that the quality of drug nanoparticles
does not change when synthesized under different flow rate conditions.
3.3. Variation of pore size of the hollow fiber membranes
Pore size in the hollow fiber membranes plays a critical role in the
anti-solvent induced drug crystallization process since each area of tiny
mixing zone was generated and controlled by water ejected from the
pore. The smaller the pore size on the hollow fiber wall are, the smaller
the dimensions of the sub-micromixing zone will be generated when the
penetrating water encounters the supersaturated drug solution in the
shell side, resulting in a better control over the synthesized crystal size.
Compared with the previous study [23] in which micron-sized drug
crystals were successfully produced using the PHFAC method with
larger pore size membranes, the current method with smaller pore size
membranes was able to produce nano-sized drug crystals in a con-
tinuous and controllable manner. The specifications of two types of
hollow fiber membranes for the two studies are provided in Table 2.
The DLS results of the particle size distribution in Table 2 show that
under optimized conditions with identical flow rate combination of
Fig. 12. TEM images of nanocrystals gained under conditions of (a) F1, (b) F2, and (c) F3.
Journal of Membrane Science 564 (2018) 682–690
Residence time (s) Mean (nm) Standard Deviation (nm) d10 (nm) d90 (nm)
14.9 295.6 31.6 263.0 333.0
9.3 190.3 46.7 169.0 214.0
5.0 144.6 24.0 127.0 163.0
6 mL/min in tube side and 11 mL/min in shell side for large module and
6.8 mL/min in tube side and 13.6 mL/min in shell side for small
module, the particle size could be significantly reduced as discussed
above. The membrane flux values were figured out with the formula of
J = V/(T × A). Here V is the volume of the anti-solvent in L, A is the
internal membrane surface area in m2, and T is the time in h. The flux
values of large pore hollow fiber membranes module and small pore
hollow fiber membranes module were respectively 415.19 L/m2 h and
79.29 L/m2 h. Fig. 13 demonstrates that the variation of pore size is an
effective way to reduce the particle size and thus to have some control
over crystallization for the precipitation of nanoparticles.
3.4. Variation of drug concentration in feed solution
In order to study the influence of drug concentration on the size of
obtained drug particles, different concentrations of drug were used in
the feed solution. As shown in Table 3, in experiment F1, the mean
particle size was 295.6 nm, which decreased to 141.8 nm when the drug
concentration was reduced under experimental condition of F4. Fur-
ther, with the concentration decreasing to 0.030 g/mL in experiment
F5, the particle size dropped to 86.4 nm.
The reduced crystal size following a reduction in drug concentration
in the feed solution can be understood in the context of the processes of
nucleation and crystal growth. First, owing to a lower concentration,
the drug dissolved in the solution was less sensitive to the anti-solvent
resulting in a slower progress to supersaturation. In this situation, nu-
cleation took place at a slow rate allowing only a reduced time for
growth if any. Further, there was less material available for growth of
the nuclei due to lower bulk concentration. Due to the local depletion of
concentration, the lack in material also contributed to reduction on
particle size. Consequently, the drug nanocrystals developed at a slow
pace and were swept out of the PHFAC crystallizer before much growth.
3.5. Further considerations
A few general comments on the membrane-based anti-solvent
crystallization device are useful. First, such a device has exceptional
fluid mixing capabilities via tiny anti-solvent fluid jets injected into the
688
X. Zhou et al. Journal of Membrane Science 564 (2018) 682–690

Table 2
Particle size distributions for different hollow fiber membrane modules.
Material Membrane Pore size Fiber OD Fiber ID PSD Mean Standard Deviation PSD d10 PSD d90
(μm) (mm) (mm) (μm) (μm) (μm) (μm)

Large pore hollow fiber membranes Nylon-6 0.2–1.5 1.0 0.6 1.61 0.26 1.00 2.21
[26]
Small pore hollow fiber membranes PVDF 0.1–0.15 2.5 1.8 0.19 0.05 0.17 0.21
[this study]

Fig. 13. SEM images of crystals obtained from (a) hollow fibers having larger pores [23] and (b) hollow fibers having smaller pores.

shell-side axial flow [25]. Computational fluid dynamics-based studies
of the mixing of two fluids in such a device [28] have demonstrated
such capabilities even though the Reynolds number of the shell-side
flow is low and is in the laminar region. Consequently, supersaturation
consumption is quite efficient in such a device. As the shell-side fluid/
suspension moves, the solute concentration will decrease considerably
along the axial direction. If the starting concentration of the solute in
the fluid entering the shell side is low, nanoparticles once formed will
hardly have any chance to grow due to rapid solute concentration de-
pletion along the main flow axis. However, that does not prevent col-
lisions between nanoparticles in two separate streamlines since there is
significant radial mixing due to anti-solvent injection through the
pores. There is an extraordinary opportunity of controlling various
parameters such as membrane pore size, incoming solute concentration,
length of device and the suspension velocity in such a configuration.
4. Concluding remarks
A porous hollow fiber crystallization device (PHFAC) was designed
to successfully achieve continuous production of drug nanocrystals
under room temperature and atmospheric pressure conditions. The as-
received micron-sized drug powder was first fully dissolved in an
acetone solution and pumped into the shell side of the module where
the anti-solvent was already introduced through the lumen of the
porous hollow fibers. The extensive mixing micro-environment gener-
ated by the penetrating water streaming jets further provided con-
siderable supersaturation level that eventually led to nucleation and
crystallization of the drug nanoparticles. This device is flexible in its
design and the interior parameters to allow relatively precise mon-
itoring and control of the drug nanocrystal production in a continuous
manner.
By increasing the flow rates of shell side from 6.8 mL/min to
20.4 mL/min, the residence time could be further reduced. Accordingly,
the size of the drug particles decreased with the decrease of the re-
sidence time since shorter residence time reduces the growth and nu-
cleation time for newly produced GF crystals. With a decrease of drug
concentration in the feed solution, the nanoparticle size could be re-
duced to as low as 86.4 nm in mean size, while still retaining identical
crystalline structure, appearance, and morphology. Moreover, the var-
iation of pore size on the hollow fibers is a critical parameter to directly
influence the particle size and thus to create a more controllable crys-
tallization environment for the production of nanoparticles. The smaller
the pore size on the hollow fiber wall and narrower of the pore size
distribution, the better is the control achieved over synthesized crystal
size. Under relatively similar conditions, the particle size of synthesized
drug nanocrystals was reduced to 190.3 nm when using the small pore
size (0.1–0.15 µm) compared with the mean particle size of 1.61 µm
when applying large pore hollow fiber membrane (0.2–1.5 µm).
In this technique, the porous hollow fibers crystallization device
enabled uniform and relatively controlled synthesis of drug nanocrys-
tals under ambient conditions. Moreover, the hollow fiber membrane
device can be easily scaled up by adjusting the numbers of the hollow
fibers in the shell side while still not impacting the quality of produced
drug nanocrystals [28], which is promising to fit the needs of modern
industrial production. The achieved results in the study have demon-
strated that the PHFAC technique for continuous manufacturing of drug
nanoparticles is feasible and realizable in a relatively straightforward
way.
Acknowledgements
This work was supported by the National Natural Science
Foundation of China [grant number 81772278] and [grant number
21706221] and the Natural Science Foundation of Fujian Province

Table 3
Mean particle size of drug crystals for different drug concentrations in the feed solution.
Experimental run Drug concentration (g/mL) Tube-side flow rate (mL/min) Shell-side flow rate (mL/min) Mean (nm) Standard Deviation (nm) d10 (nm) d90 (nm)

F1 0.040 6.8 6.8 295.6 31.6 263.0 333.0

F4 0.035 6.8 6.8 141.8 13.5 126.0 160.0
F5 0.030 6.8 6.8 86.4 14.1 73.5 101.0

689
X. Zhou et al.
[grant number 2018J05143].
References
[1] F.D. Cui, X.Y. Long, Pharmacology, seventh ed., People's Medical Publishing House,
Beijing, 2011, p. 395.
[2] Y. Gao, J. Wang, Y. Wang, Q. Yin, B. Glennon, J. Zhong, J. Ouyang, X. Huang,
H. Hao, Crystallization methods for preparation of nanocrystals for drug delivery
system, Curr. Pharm. Des. 21 (2015) 3131–3139.
[3] P. Sheth, H. Sandhu, D. Singhal, W. Malick, N. Shah, M.S. Kislalioglu, Nanoparticles
in the pharmaceutical industry and the use of supercritical fluid technologies for
nanoparticle production, Curr. Drug. Deliv. 9 (2012) 269–284.
[4] W.E. Bawarski, E. Chidlowsky, D.J. Bharali, S.A. Mousa, Emerging nanopharma-
ceuticals, Nanomed. Nanotechnol. Biol. Med. 4 (2008) 273–282.
[5] B. Sinha, R.H. Mueller, J.P. Moeschwitzer, Bottom-up approaches for preparing
drug nanocrystals: formulations and factors affecting particle size, Int. J. Pharm.
453 (2013) 126–141.
[6] H.B. Chen, C. Khemtong, X.L. Yang, X.L. Chang, J.M. Gao, Nanonization strategies
for poorly water-soluble drugs, Drug. Discov. Today 16 (2011) 354–360.
[7] E. Merisko-Liversidge, G.G. Liversidge, Nanosizing for oral and parenteral drug
delivery: a perspective on formulating poorly-water soluble compounds using wet
media milling technology, Adv. Drug. Deliv. Rev. 63 (2011) 427–440.
[8] B. Van Eerdenbrugh, G. Van den Mooter, P. Augustijns, Top-down production of
drug nanocrystals: nanosuspension stabilization, miniaturization and transforma-
tion into solid products, Int. J. Pharm. 364 (2008) 64–75.
[9] R. Shegokar, R.H. Müller, Nanocrystals: industrially feasible multifunctional for-
mulation technology for poorly soluble actives, Int. J. Pharm. 399 (2010) 129–139.
[10] B. Subramaniam, R.A. Rajewski, K. Snavely, Pharmaceutical processing with su-
percritical carbon dioxide, J. Pharm. Sci. 86 (1997) 885–890.
[11] J. Jung, M. Perrut, Particle design using supercritical fluids: literature and patent
survey, J. Supercrit. Fluids 20 (2001) 179–219.
[12] S.A. Desobry, F.M. Netto, T.P. Labuza, Comparison of spray-drying, drum-drying
and freeze-drying for beta-carotene encapsulation and preservation, J. Food Sci. 62
(1997) 1158–1162.
[13] P. Teixeira, H. Castro, R. Kirby, Spray-drying as a method for preparing con-
centrated cultures of lactobacillus-bulgaricus, J. Appl. Bacteriol. 78 (1995)
456–462.
690
Journal of Membrane Science 564 (2018) 682–690
[14] K.R. Vandana, Y.P. Raju, V.H. Chowdary, M. Sushma, N.V. Kumar, An overview on
in situ micronization technique – an emerging novel concept in advanced drug
delivery, Saudi Pharm. J. 22 (2014) 283–289.
[15] H.K. Chan, P.C.L. Kwok, Production methods for nanodrug particles using the
bottom-up approach, Adv. Drug. Deliv. Rev. 63 (2011) 406–416.
[16] N. Rasenack, B.W. Muller, Dissolution rate enhancement by in situ micronization of
poorly water-soluble drugs, Pharm. Res. 19 (2002) 1894–1900.
[17] J.W. Tom, P.G. Debenedetti, Particle formation with supercritical fluids–a review, J.
Aerosol Sci. 22 (1991) 555–584.
[18] S.D. Yeo, E. Kiran, Formation of polymer particles with supercritical fluids: a re-
view, J. Supercrit. Fluids 34 (2005) 287–308.
[19] D.W. Matson, J.L. Fulton, R.C. Petersen, R.D. Smith, Rapid expansion of super-
critical fluid solutions - solute formation of powders, thin-films, and fibers, Ind. Eng.
Chem. Res. 26 (1987) 2298–2306.
[20] S.H. Lee, D. Heng, W.K. Ng, H.K. Chan, R.B. Tan, Nano spray drying: a novel
method for preparing protein nanoparticles for protein therapy, Int. J. Pharm. 403
(2011) 192–200.
[21] G.G. Chen, G.S. Luo, J.H. Xu, J.D. Wang, Membrane dispersion precipitation
method to prepare nanoparticles, Powder Technol. 139 (2004) 180–185.
[22] R. Kieffer, D. Mangin, F. Puel, C. Charcosset, Precipitation of barium sulphate in a
hollow fiber membrane contactor: Part I Investigation of particulate fouling, Chem.
Eng. Sci. 64 (8) (2009) 1759–1767.
[23] R. Kieffer, C. Charcosset, F. Puel, D. Mangin, Numerical simulation of mass transfer
in a liquid-liquid membrane contactor for laminar flow conditions, Comput. Chem.
Eng. 32 (2008) 1333–1341.
[24] D.M. Zarkadas, Crystallization Studies in Hollow Fiber Devices (Ph.d. thesis), New
Jersey Institute of Technology, New Jersey, 2004.
[25] D.M. Zarkadas, K.K. Sirkar, Anti-solvent crystallization in porous hollow fiber de-
vices, Chem. Eng. Sci. 61 (2006) 5030–5048.
[26] D. Chen, D. Singh, K.K. Sirkar, R. Pfeffer, Continuous synthesis of polymer-coated
drug particles by porous hollow fiber membrane-based anti-solvent crystallization,
Langmuir 31 (2015) 432–441.
[27] D. Chen, D. Singh, K.K. Sirkar, R. Pfeffer, Porous hollow fiber membrane- based
continuous technique of polymer coating on submicron and nanoparticles via anti-
solvent crystallization, Ind. Eng. Chem. Res. 54 (19) (2015) 5237–5245.
[28] D. Chen, B. Wang, K.K. Sirkar, Hydrodynamic modeling of porous hollow fiber anti-
solvent crystallizer for continuous production of drug crystals, J. Membr. Sci. 556
(2018) 185–195.