Chemical Engineering Journal 371 (2019) 554–564

Contents lists available at ScienceDirect

Chemical Engineering Journal

journal homepage: [Link]/locate/cej

Fluid mechanics and process design of high-pressure antisolvent T

precipitation of fenofibrate nanoparticles using a customized microsystem
⁎
S. Melziga,c, , J.H. Finkea,c, C. Schildea,c, A. Vierhellerb, A. Dietzelb,c, A. Kwadea,c
a
Institute for Particle Technology, Technische Universität Braunschweig, Volkmaroder Str. 5, 38104 Braunschweig, Germany
b
Institute of Microtechnology, Technische Universität Braunschweig, Alte Salzdahlumer Str. 203, 38124 Braunschweig, Germany
c
PVZ – Center of Pharmaceutical Engineering, Technische Universität Braunschweig, Germany

H I GH L IG H T S

• Visual characterization of the mixing behavior in the microsystem via µPIV and LIF.
• Nanoparticle production via high-pressure antisolvent precipitation.
• Systematic evaluation of formulation and process parameters.
• Pressures till 1000 bar enable product volume flow rates up to 229 ml/min.

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

Keywords: The production of drug nanoparticles is a promising method to enhance dissolution behavior and, by that,
Fenofibrate bioavailability of poorly soluble drugs. In this study, fenofibrate nanoparticles were produced via high-pressure
Nanoparticles antisolvent precipitation (HPAP) using a customized microsystem. The benefit of microfluidic systems is the
Precipitation defined streaming behavior inside the microchannels resulting in homogeneous products produced via a con-
Microsystem
tinuous process. Formulation strategies (e.g. fenofibrate mass fraction) as well as process parameters (e.g.
µPIV
pressure) were studied to evaluate their influence on particle size and size distribution. Furthermore, the
LIF
High-pressure streaming and mixing behavior in the microchannels were visualized via micro particle image velocimetry
(µPIV) and laser induced fluorescence (LIF) in order to elucidate the mechanisms inside the microchannels as
well as to support the results obtained from HPAP experiments.

1. Introduction storage of nanosuspensions often brings along challenges in stability,

nanoparticles are usually transferred into a long-term stable powder
It is estimated that nearly 40% of marketed drugs and 70–90% of form shortly after generation [4]. In particular, bottom up processes
drug candidates in pipeline are poorly water-soluble [1]. Bearing in such as precipitation are getting more attention from the pharmaceu-
mind that the solubility of drugs is an important requirement for their tical industry, as they allow both the synthesis of amorphous drug na-
bioavailability, the development and establishment of suitable galenic noparticles [5], by which the solubility is further improved, as well as
methods to improve the solubility of these drugs takes top priority, the avoidance of impurities (e.g. grinding media wear) [6]. Further-
especially for orally administered medication. In this context, the pro- more, precipitation processes can be implemented in microfluidic sys-
duction of drug nanoparticles is a common method to improve the tems to produce nanoparticles for example composed of lipids [7] or
bioavailability of those drugs. Due to the significantly increased specific polymers [8]. On the one hand, microfluidics enables the production of
particle surface area, the dissolution rate of those active pharmaceutical nanoparticles via a continuous process resulting in homogenous pro-
ingredients (API) can be increased. This phenomenon is described by ducts and, on the other hand, it enables the production of small product
the Noyes–Whitney equation, which implies that the dissolution rate is quantities with low consumption of educts. For these reasons, the
proportional to the surface area of the substance [2,3]. In principle, precipitation of API nanoparticles is also carried out in microfluidic
drug nanoparticles can be produced either via top-down (e.g. commi- systems.
nution) or via bottom-up (e.g. precipitation) approaches. Since the Ali et al. [9] used a Y-shaped mixing device with a microchannel

⁎
Corresponding author.
E-mail address: [Link]@[Link] (S. Melzig).

[Link]
Received 9 October 2018; Received in revised form 2 April 2019; Accepted 9 April 2019
Available online 10 April 2019
1385-8947/ © 2019 Elsevier B.V. All rights reserved.
S. Melzig, et al. Chemical Engineering Journal 371 (2019) 554–564

diameter of 500 µm and product volume flow rates of 0.5–3 ml/min. biopharmaceutical system (BCS) and possesses good permeability
They synthesized hydrocortisone nanoparticles using PVP, HPMC and properties but poor solubility [22]. Therefore, the goal was to establish
SDS as stabilizers producing particles with sizes down to 80 nm. Zhao a method for enhancing solubility and, by that, bioavailability of those
et al. [10] and Wang et al. [11] used both the same Y-shaped mixing APIs. It was aimed to investigate the influence of formulation (e.g. fe-
device in order to precipitate danazol (sizes down to 364 nm) and ce- nofibrate mass fraction) and process (e.g. pressure) parameters on
furoxime axetil nanoparticles (sizes down to 300 nm), respectively, product properties (particle size and polydispersity) using HPAP. Fur-
without any stabilizing agents. The dimension of the microchannels thermore, it was of great interest to visualize the streaming behavior in
ranged from 300 µm up to 800 µm and the product volume flow rate the microchannels at pressures up to 200 bar in order to better under-
reached values up to 84 ml/min. Another commonly used type of stand the influence of process parameters on the mixing behavior and,
mixing devices is a T-shaped microsystem. Xu et al. [12] produced thus, on product properties. Finally, this knowledge helped to improve
beclomethasone dipropionate elongated particles with width of 250 nm existing processes and accelerate the design of new ones since the basic
and length of several micrometers in such a microsystem. After a sub- influences could fully be understood. The main difference between the
sequent high pressure homogenization, they obtained the desired na- HPAP and other precipitation methods is the high product volume flow
noparticles with enhanced dissolution behavior. They could achieve rate, which enables the production of a sufficient quantity for the
product volume flow rates up to 98 ml/min in the microsystem with continuous production of tailor-made medications. The overall aim of
dimensions of 400 µm and 500 µm. Odetade et al. [13] pursued a this work pursues the economic production of fenofibrate nanosus-
clearly different approach in order to precipitate API nanoparticles in a pensions with preferably small particles, narrow particle size distribu-
microfluidic device. They designed a microsystem with an inner and tion and high mass fraction at high product volume flow rates.
outer capillary to generate a three-dimensional mixing zone. The dia-
meter of the inner capillary ranged from 50 µm to 400 µm and the 2. Materials and methods
diameter of the outer capillary was 1150 µm. They synthesized hydro-
cortisone nanoparticles with sizes down to 210 nm using HPMC, PVP 2.1. Materials
and SDS as stabilizers, under product volume flow rates of
0.15–0.75 ml/min. Two fluids were used for the synthesis of fenofibrate nanoparticles
The production of fenofibrate particles is generally realized via namely antisolvent and solvent. The antisolvent consists of deionized
three methods: melt emulsification, comminution and antisolvent pre- water, hydroxypropyl methylcellulose (HPMC) and sodium dodecyl
cipitation. During melt emulsification, the drug is molten in heated sulfate (SDS). HPMC was obtained from JRS Pharma (Rosenberg,
water and dispersed for a few minutes with an ultrasonic probe [14]. Germany) and SDS from Sigma-Aldrich (Steinheim, Germany). The
Different surfactants and polymers are used as stabilizing agents in solvent consists of ethanol and fenofibrate (Novartis AG, Basel,
order to yield nanoparticles with sizes of 150 nm in a batch process. Switzerland). Concentrations of both fluids were varied for several
Furthermore, it has also been reported that a fenofibrate emulsion experiments in order to study the influence of various parameters like
prepared under mild stirring conditions and quickly poured into cold mass fraction of fenofibrate (0.01–0.35%) or ratio of stabilizers to fe-
water under vigorous conditions can be used to produce nanoparticles nofibrate (0–0.39).
smaller than 500 nm [15]. Comminution was implemented via wet
media milling. Both Knieke et al. [16] and Bitterlich et al. [17] could 2.2. High-pressure antisolvent precipitation
produce nanoparticles with sizes of 100 nm using different kinds of
stabilizers. However, the already mentioned processes were operated Fenofibrate nanoparticles were synthesized via high-pressure anti-
discontinuously. For that reason, antisolvent precipitation of fenofi- solvent precipitation (HPAP). The precipitation process was im-
brate nanoparticles was implemented in microfluidic devices. Lorenz plemented with two pressure intensifiers and a customized microsystem
et al. [18] designed a three-dimensional flow-focusing microsystem as the mixing device. The microchips were fabricated out of polished
which produces single ethyl acetate droplets containing dissolved fe- stainless steel substrates via micro electrical discharge machining, de-
nofibrate in a aqueous polysorbate 20 solution. They could synthesize scribed in detail by Richter et al. [23,24]. The basic idea of the modular
well defined nanoparticles (128 nm) since every droplet could form concept of the microsystem was developed and used by Gothsch et al.
exactly one nanoparticle due to diffusion of ethyl acetate in water. [25] for dispersing and Finke et al. [26,27] for emulsification. In this
Product volume flow rates were lower than 1 µl/min caused by a mi- study, the microsystem was customized in order to mix two fluid flows.
crochannel dimension of 8–20 µm. Dong et al. [19] used a Y-shaped In fact, it consists of two inlets, a mixing device and one outlet. The
microsystem to produce narrowly distributed nanoparticles (196 nm) mixing zone possesses two rectangular nozzles (width: 100 µm; height:
using PVPK30 as stabilizing agent. Dimensions of the channels were 50 µm) which are arranged symmetrically on both sides of the main
100 µm × 205 µm resulting in product volume flow rates of 0.4–4 ml/ channel (width: 1000 µm; height: 50 µm) at an angle of 45°. In this
min. Beck et al. [20] developed a T-shaped microsystem with an ul- study, the antisolvent was injected into the solvent flow via these two
trasound probe in the mixing zone. The size of the produced nano- nozzles. At first, the pressure of the solvent flow was set constant at
particles was 882 nm and the product volume flow rate reached values 20 bar and the antisolvent flow was adjusted up to pressures of 200 bar.
up to 150 ml/min in channels with minimal dimensions of 1.67 mm Furthermore, investigations regarding the overall pressure of the
(outer diameter). The highest product volume flow rate of 550 ml/min system were done by increasing the solvent as well as the antisolvent
was reached by Hu et al. [21] using a static mixer (inner diameter: pressure at the same time. For such experiments, the pressure of the
25 mm) in which the solvent and the antisolvent were pumped through antisolvent flow was increased up to 1000 bar while the ratio of anti-
a nozzle (500 µm and 1500 µm). They could produce nanoparticles of solvent to solvent pressure was kept constant at a value of 10. Every
328 nm, but with a relatively broad size distribution using HPMC and experiment was conducted once unless otherwise stated. The overall
SDS as stabilizer. design of the experimental setup for the HPAP of fenofibrate nano-
Nevertheless, there are still challenges in the precipitation of nar- particles is shown in Fig. 1.
rowly distributed API nanoparticles, especially with high product vo-
lume flow rates using microfluidic systems. For that reason, the present 2.3. Process characterization
study is focused on the production of fenofibrate nanoparticles with
narrow size distributions via high-pressure antisolvent precipitation 2.3.1. Volume flow rate and volume fraction
(HPAP) using a customized microsystem. Fenofibrate was chosen as a In order to determine the volume flow rates and volume fractions of
model drug, which represents class 2 molecules of the the solvent and the antisolvent in the microsystem, the solvent stream

555
S. Melzig, et al. Chemical Engineering Journal 371 (2019) 554–564

Fig. 1. Experimental setup for high-pressure antisolvent precipitation (HPAP) of fenofibrate nanoparticles.

was dyed with methylene blue (30 mg/l). Subsequently, the solvent was
mixed with the antisolvent in the microsystem at different pressures
and the product mass flow ṁ Product was calculated gravimetrically. The
experiments were carried out 5 times for every antisolvent pressure.
Afterwards, the mixture was measured in a polystyrene cuvette at a
wavelength of 666 nm in a UV–VIS spectrophotometer (UV-3100PC,
VWR, Darmstadt, Germany). The respective volume fraction of the
solvent φSolvent and of the antisolvent φAntisolvent in the mixture was de-
termined by means of a calibration line. In order to ascertain the solvent
volume flow rate VSolvenṫ (Eq. (2)) and the antisolvent volume flow rate
̇
VAntisolvent (Eq. (3)), the ideal product volume flow rate VProduct
̇ , ideal (Eq.
(1)) is required. In this context, ideal means that the volume contraction
during mixing of the solvent and antisolvent must be neglected in this
particular case. For the determination of the ideal product volume flow Fig. 2. Experimental setup for the determination of the flow behavior in the
rate, the solvent density ρSolvent and the antisolvent density ρAntisolvent high-pressure microsystem.
were estimated with the density of ethanol at 25 °C and the density of
water at 25 °C, respectively.
only the light emitted by the tracer particles reached the CCD camera
̇ ṁ Product ṁ Product (Imager intense, LaVision, Göttingen, Germany). This made it possible
VProduct , ideal = =
ρProduct , ideal φSolvent . ρSolvent + φAntisolvent . ρAntisolvent (1) to record two associated images of the tracer particles at a defined time
interval Δt of 500 ns with the camera. The tracer particles (PS-FluoRed-
̇
VSolvent ̇
= VProduct , ideal. φSolvent (2) 2.0, microParticles GmbH, Berlin, Germany) had a size of 2.09 μm with
̇ ̇ a standard deviation of 0.03 μm and were dispersed in the fluids. The
VAntisolvent = VProduct , ideal. φAntisolvent (3)
concentration of particles in the fluids was 80 mg/l. The maximum
In order to determine the real product volume flow rate VProduct
̇ (Eq. absorption of the tracer particles happens at a wavelength of 518 nm
(4)), the measured product mass flow ṁ Product was divided by the den- and the emission peak at 547 nm. In order to ensure optical access,
sity ρProduct of the resultant ethanol-water-mixture at 25 °C. microsystems with the same geometry, but made of a combination of
ṁ Product glass and silicon were used instead of the stainless steel microchips as in
̇
VProduct = the precipitation process (compare Section 2.2). The production of the
ρProduct (4)
special microchip took place at the Institute of Microtechnology of the
Technical University of Braunschweig using deep reactive ion etching,
2.3.2. Micro particle image velocimetry which has already been described in detail by Büttgenbach et al. [30].
Flow behavior within the microsystem was observed by micro Furthermore, the mounting device of the microsystem had to be
particle image velocimetry (µPIV). For this purpose, the experimental adapted in order to enable optical access (compare Fig. 1), resulting in a
setup of Gothsch et al. [28,29] was adapted as shown in Fig. 2. The device which is leak-proof up to pressures of 250 bar. In comparison, it
pulsed Nd:YAG-laser (Nano S PIV, Litron Lasers, Warwickshire, Eng- was possible to work with pressures up to 2000 bar using the stainless
land) has a pulse duration of 6 ns. In addition, the frequency of the laser steel microsystem in combination with the corresponding mounting
was doubled, so that green light having a wavelength of 532 nm was device.
generated from the original infrared radiation having a wavelength of In order to evaluate the recorded images, the software DaVis 7.2
1064 nm. The laser light passed through an epifluorescence filter to the (LaVision, Göttingen, Germany) was used. During the evaluation, the
microsystem placed on a fluorescence microscope (Axio Observer Z1, square interrogation area (64 × 64 px) had to be adapted, since the
Zeiss, Göttingen, Germany). Both fluid streams were enriched with minimum time interval Δt was limited to 500 ns. Otherwise, the tracer
fluorescent polystyrene particles, which can absorb the light and emit it particles would have moved too far within an interrogation area of an
at a higher wavelength. Subsequently, both the emitted long-wave and image pair so that it would not be possible to determine valid velocity
the reflected short-wave laser light reached the epifluorescence filter. vectors. The adjusted interrogation areas (shape and aspect ratio) are
Since the filter is only transparent to light with a wavelength > 560 nm, shown in Table 1 and exhibit also 4096 px. In addition, the nozzles had

556
S. Melzig, et al. Chemical Engineering Journal 371 (2019) 554–564

Table 1 approximately 2 ml of the adequately diluted nanosuspension was

Evaluation settings for µPIV measurements. placed in a polystyrene cuvette. The measurement angle was 173°
Antisolvent pressure Main channel Nozzle backscatter with automatic measurement. All tests were carried out at
20 °C with an equilibration time of 2 min. The analysis was performed
50 bar 1:1 square 1:1 square in triplicate for each sample and the z-average (z-avg) was recorded
100 bar 2:1 ellipse (90°) 2:1 ellipse (45° and 135°)
along with the polydispersity index (PdI) values.
150 bar 4:1 ellipse (90°) 4:1 ellipse (45° and 135°)
200 bar 4:1 ellipse (90°) 4:1 ellipse (45° and 135°)
2.4.2. Solubility
Solubility of fenofibrate in different ethanol-water-mixtures was
to be evaluated separately for antisolvent pressures of 100 bar and determined with a UV–VIS spectrophotometer (UV-3100PC, VWR,
above, with an angle of 45° and 135° of the interrogation area de- Darmstadt, Germany). For that purpose, saturated solutions were pre-
pending on the flow direction of the nozzle. In contrast, the main pared by stirring an excess of fenofibrate in the various mixtures for at
channel was evaluated at all antisolvent pressures with an angle of 90°, least 48 h at 25 °C. Subsequently, these saturated solutions were filtered
which also corresponded to the main flow direction. Subsequently, the (pore size 0.22 μm), diluted with the corresponding ethanol-water-
velocity vectors derived from 100 image pairs were averaged in order to mixture and placed in a glass cuvette for the measurements. Samples
calculate valid values. During the setup and the execution of the µPIV were measured at wavelengths from 283 to 292 nm depending on the
experiments as well as during the subsequent evaluation, the generally ethanol and water content, respectively. Finally, drug amounts were
known guidelines for the execution of µPIV measurements [31] were calculated by means of the corresponding calibrations.
taken into account so that it could be ensured that valid values were
obtained. 3. Results and discussion

2.3.3. Cavitation It was aimed to produce fenofibrate nanoparticles as small as pos-

The investigations regarding the evolution of cavitation bubbles was sible with a narrow particle size distribution at reasonable solids con-
performed with the experimental setup described in Section 2.3.2. centrations via HPAP. To achieve this goal, fenofibrate mass fraction,
Antisolvent and solvent were dyed with rhodamine 6G perchlorate ratio of stabilizers to fenofibrate, antisolvent pressure, antisolvent
(Sigma-Aldrich, St. Louis, USA) which possesses fluorescent character- temperature and overall pressure were varied in order to determine the
istics without any temperature dependency. Concentration of rhoda- most appropriate formulation and process parameters (Section 3.2).
mine 6G perchlorate was 15 mg/l. The maximum absorption of the dye However, the process itself was characterized first in order to under-
happens at a wavelength of 528 nm and the emission peak at 551 nm. stand subprocesses in the microsystem and, thus, to support and explain
the results obtained from HPAP. During the process characterization
2.3.4. Laser induced fluorescence volume flow rate, volume fraction and mixing properties were under
Mixing behavior within the microsystem was investigated via laser study (Section 3.1).
induced fluorescence (LIF). The experimental setup is described in
Section 2.3.2. In comparison to the cavitation study in Section 2.3.3, 3.1. Process characterization
only the solvent was dyed with rhodamine 6G perchlorate for LIF ex-
periments. The concentration of rhodamine 6G perchlorate was also 3.1.1. Volume flow rate and volume fraction
15 mg/l. Finally, solvent concentration was calculated by means of the The volume fractions of the solvent and the antisolvent in the pro-
calibration with the software DaVis 7.2. duct determined as a function of the antisolvent pressure are shown in
Fig. 3 (right). It can be seen that the proportion of the solvent decreases
2.4. Product characterization and the proportion of antisolvent increases with raising antisolvent
pressure. Thus, the ratio of antisolvent to solvent, which is the major
2.4.1. Particle size distribution driving force for antisolvent precipitation, increases while increasing
Particle size distributions of nanoparticles were measured via dy- the antisolvent pressure. However, it can also be seen that with anti-
namic light scattering using the Zetasizer Nano ZS (Malvern solvent pressures > 250 bar barely any solvent is contained in the
Instruments, Worcestershire, UK). First of all, nanosuspensions were product and this leads to an exponential increase of the antisolvent to
diluted with a saturated fenofibrate solution. Afterwards, solvent ratio. Since the active ingredient is initially dissolved in the

Fig. 3. Influence of the antisolvent pressure on volume flow rate (left) as well as on volume fraction and ratio of antisolvent to solvent (right).

557
S. Melzig, et al. Chemical Engineering Journal 371 (2019) 554–564

solvent at the HPAP, the solvent content is crucial for the drug content decreases. Finally, an area at the wall behind the upper nozzle with
in the final product. Thus, antisolvent pressures > 200 bar combined undefined velocity vectors was formed at an antisolvent pressure of
with a solvent pressure of 20 bar are not suitable to produce a sus- 150 bar. In this context, “undefined” means that the orientation and
pension with sufficient active ingredient content. For this reason, only direction of the vectors is random and does not represent the actual
pressures up to 200 bar were investigated with the microsystem except flow. Additionally, streamlines in this area at this particular pressure
for the study regarding the overall pressure (see Section 3.2). As a are no longer oval-shaped but rather asymmetrical and angular. The
consequence, the maximum antisolvent to solvent ratio of 10.5 was reason for this phenomenon is the formation of cavitation bubbles be-
achieved at the investigations with the minimum ethanol content of cause of the high pressure drop behind the nozzles (compare Section
8.7%. 2.3.3). Due to the cavitation, zones are formed where only a few or no
The volume flow rates of the solvent and antisolvent flow were tracer particles pass. This in turn leads to the fact that there are not
calculated, taking the determined volume fractions of solvent and an- enough tracer particles in the interrogation areas of these zones for
tisolvent into account as well as the measured product mass flow. sufficient statistical certainty although 100 image pairs were averaged.
Equivalent to the volume fractions, the antisolvent volume flow rate This renders such random velocity vectors invalid. However, the cavi-
drastically increases and the solvent volume flow rate slightly decreases tation causes a decrease in the residence time of the fluids in the mixing
when the antisolvent pressure is increased (Fig. 3, left). Accordingly, zone of the microsystem and more homogeneous mixture because of
the product volume flow rate increases with higher antisolvent pres- minimized backflow. In addition, the distances are shortened by the
sures being 58.6 ml/min at an antisolvent pressure of 50 bar and narrowing of the mixing zone, so that both the convective and the
112.7 ml/min at 300 bar. Later in this study, it is shown that an anti- diffusive mixing can be intensified. The phenomenon of undefined ve-
solvent pressure of 200 bar is well suited for the precipitation of feno- locity vectors could be found at the walls behind both nozzles at anti-
fibrate nanoparticles. At this particular pressure, the solvent flow rate is solvent pressures of 200 bar. Furthermore, a backflow of the antisolvent
8.3 ml/min and the antisolvent flow rate is 87.3 ml/min. In comparison streams was formed at antisolvent pressures > 175 bar flowing con-
with other microsystems used for precipitation processes, the product trary to the main flow to an area behind the injecting zone. This
volume flow rate is very high within the microsystem used in this work. backflow was enhanced by increasing antisolvent pressure and formed
Typical values of the product volume flow rate found in the literature a homogeneous backflow area at pressures of 200 bar which is shown
are between 1 and 24 ml/min [9,19,32]. Nevertheless, microsystems by the velocity vectors as well as by the streamlines. As a result, this
with higher volume flow rates up to 82 ml/min [11] or even up to backflow can explain the decrease in the solvent flow rate despite
550 ml/min [21] were already described. constant solvent pressure (compare Fig. 3). If the backflow increases,
Due to the small channel geometries in the microsystem used in this the solvent flow rate is suppressed and nearly cut off at antisolvent
study (in particular the two nozzles) and the high volume flow rates, pressures > 300 bar. However, the formation of vortices in this area
very high fluid velocities could be achieved in the microchannels which always introduces some solvent into the product stream. It should be
can lead to turbulence within the microsystem. In order to estimate if underlined that the amount of solvent is negligible for antisolvent
there was a turbulent flow within the microsystem, the corresponding pressures > 300 bar, which results in extraordinarily high ratios of
Reynolds numbers were calculated based on the applied volume flow antisolvent to solvent. In addition, it was observed that the achieved
rates and the microchannel geometries. It was evident that the Reynolds maximum flow velocity increases almost linearly with the increase of
numbers for the antisolvent flow were drastically higher than for the the antisolvent pressure. Maximum flow velocities of 65 m/s were de-
solvent and product flows. This is because the antisolvent was in- termined at an antisolvent pressure of 50 bar, whereas at a pressure of
troduced through two nozzles into the main channel, resulting in very 200 bar flow velocities of almost 170 m/s already occurred. As a result,
high fluid velocities and, thus, in Reynolds numbers of over 10,870 at the outcomes of the µPIV analysis are in good agreement with the vo-
antisolvent pressures of 200 bar and higher. In comparison to that, lume flow rates determined in Section 2.3.1, whose mean velocities are
Reynolds numbers of the solvent flow decreased from 445 to 46 while a bit below the maximum velocities calculated in this section.
increasing the antisolvent pressure from 50 to 300 bar due to the re-
duction of the solvent volume flow rate. However, Reynolds numbers of 3.1.3. Cavitation
the product flow increased from 1714 to 3966 while increasing the As already mentioned, high antisolvent pressures increase the for-
antisolvent pressure over the same pressure range. Considering the mation of cavitation bubbles due to the high pressure drop in the nozzle
relatively high Reynolds numbers of the antisolvent flow in the nozzle area (see Fig. 5). It can be observed that the formation of cavitation
and of the product flow in the main channel, it could be assumed that a bubbles primarily takes place directly at the front edge of the nozzles.
turbulent flow was present in the mixing zone of the microchannel. Due However, it could also be shown that no cavitation bubbles were
to the turbulent flow, in turn, a rapid mixing of the two fluids can be formed at pressures of 50 bar. First formation of individual cavitation
assumed, because it is unlikely that the three fluid streams are lamellar, bubbles was observed behind both nozzle structures at antisolvent
but rather are permeating each other. This rapid mixing is essential for pressures of 75 bar. The cavitation bubble size increases up to pressures
the precipitation of nanoparticles with a narrow size distribution. of 100 bar, whereas at pressures from 125 to 150 bar already one
Nevertheless, the assumption of the turbulent flow was proved by sidewall is almost completely covered with a large and permanent ca-
imaging techniques in the following sections. vitation bubble. At antisolvent pressures of 175 bar, a second large
cavitation bubble was formed behind the other nozzle, which is already
3.1.2. Micro particle image velocimetry (µPIV) fully formed at pressures above 200 bar. As already indicated in the
Fig. 4 shows the results of µPIV measurements at various antisolvent analysis of the flow vectors, the formation of these cavitation bubbles
pressures (50–200 bar) and a constant solvent pressure (20 bar). The leads to the fact that backmixing cannot occur in both wall regions.
velocity vectors are shown on the left side and the streamlines on the Furthermore, it can be seen that the sidewall, at which a larger cavi-
right side. It can be seen that a similar flow profile is formed at all tested tation bubble initially forms, is arbitrary and can be presumably at-
antisolvent pressures. At both nozzles, the antisolvent flows with tributed to surface roughness or defects in the channels, but always
equally high speeds transversely into the middle of the channel, behaves reproducibly and consistently within a channel. This can be
forming a main flow in this region with also high speed. Accordingly, very well recognized in the averaged images, for instance at an anti-
significant lower flow velocities prevail in both wall regions. Here it solvent pressure of 150 bar, at which the cavitation bubble is pro-
even comes to backflow in the direction of the mixing zone, which nounced only on one sidewall. Significant differences between the vo-
reunites again with the antisolvent flows from the nozzles. However, if lume flow rates and the flow velocities of the two antisolvent streams in
the antisolvent pressure increases, the velocity of the return flow the nozzle structures can be ruled out, since the values determined in

558
S. Melzig, et al. Chemical Engineering Journal 371 (2019) 554–564

Fig. 4. Investigation of the flow behavior in the microsystem via µPIV measurements under variation of the antisolvent pressure.

the µPIV analysis are almost identical (see Fig. 4). Moreover, the which this backflow prevails ahead the inject zone is not constant, as it
merged fluid stream in the mixing zone is very centered along the can be seen in the illustration of the average image at an antisolvent
channel, another cause for the preference to form the cavitation bubble pressure of 200 bar. The favored side seems to be the lower side of the
on one specific sidewall that can be excluded. In addition, it can be seen figure, since there is a much lower solvent concentration. However, on
from antisolvent pressures of 150 bar that the product flow is highly the upper side there is also a clear area with slightly reduced solvent
turbulent since the flow is wavelike in the center of the main channel. concentration. The reason for this could be either the surface roughness
Furthermore, there are vortices at the edge area of the centered product or defects or also a not exactly planar alignment of the microsystem.
flow which in principle should lead to a faster mixing of antisolvent and These phenomena could make it energetically more favorable to flow in
solvent. a certain direction through lower flow resistance. However, the influ-
ence on the mixing is probably negligible as it shows a clear flow
profile. Apart from the basic flow behavior in the microsystem, it is
3.1.4. Laser induced fluorescence (LIF) clearly shown in Fig. 6 that the proportion of solvent decreases with
Based on the already analyzed flow characteristics within the mi- increasing antisolvent pressure since the solvent concentration de-
crosystem, the actual mixing process, which is crucially important for creases across the entire cross-section of the microsystem, which is al-
the product quality of the precipitation process, is described in this ready discussed in Section 3.1.1. In addition, it can be clearly shown
section. The concentration of the solvent in the mixture was determined that vortices are formed in the contact region between solvent and
by LIF analysis at various antisolvent pressures to characterize the antisolvent already at antisolvent pressures of 50 bar which indicates
mixing properties (see Fig. 6). In this case, geometrically similar flow that turbulence already prevails there. This turbulence is favored in
profiles are observed in all recordings, as it was already evident in the particular by the wavelike flow in the main channel. Furthermore, it
analysis of cavitation (see Fig. 5). Both antisolvent flows streamed ob- can be seen that on the one hand, the turbulence increases within the
liquely into the main channel and were combined in the mixing zone. main channel in the direction of the flow and, on the other hand, when
However, it can clearly be seen in this approach that the location of the the antisolvent pressure increases. This fact can be of decisive im-
unification of the two antisolvent flows is shifted further towards the portance for the fast mixing of the antisolvent and the solvent within
inject zone at higher antisolvent pressures. At pressures above 175 bar, the microchannel and, thus, for the production of nanoparticles as small
even a backflow ahead the inject zone is formed. However, the side on

559
S. Melzig, et al. Chemical Engineering Journal 371 (2019) 554–564

Fig. 5. Evolution of cavitation bubbles in the microsystem under variation of the antisolvent pressure.

as possible with a narrow particle size distribution. of experiments, the maximum fenofibrate mass fraction was limited. As
displayed in Fig. 7, particle size does not further decrease from a mass
fraction of 0.18% on and reaches a plateau with a value of around
3.2. High-pressure antisolvent precipitation of fenofibrate nanoparticles
250 nm. Moreover, the PdI decreases to values lower than 0.1 assuring
a very narrow particle size distribution. Because of the defined aim of
3.2.1. Mass fraction of fenofibrate in the product
achieving small particle size and high fenofibrate content, further stu-
The product mass fraction of fenofibrate was varied in order to
dies were carried out with a fenofibrate mass fraction of 0.35% re-
examine its influence on particle synthesis. Moreover, it was aimed to
sulting in a particle size of 254 nm and a PdI of 0.037.
find an optimum in formulation to obtain a suspension with high fe-
nofibrate mass fraction and small particle sizes with narrow size dis-
tribution. During this examination, values of antisolvent pressure 3.2.2. Ratio of stabilizers to fenofibrate
(200 bar), solvent pressure (20 bar), ratio of stabilizers (HPMC and SDS) The influence of different stabilizer concentrations on particle size
to fenofibrate (both 0.26) as well as solvent and antisolvent tempera- and polydispersity index is presented in Fig. 8. It is clearly shown, that
ture (25 °C) were kept constant. It can be clearly seen, that particle size the addition of HPMC as a further stabilizing agent led to smaller
and polydispersity index decrease by increasing mass fraction of feno- particles and narrower size distributions. However, it is not the highest
fibrate (Fig. 7). This is because of an increase of the initial super- but the lowest amount of HPMC which results to the best product. It
saturation due to higher mass fraction of fenofibrate which leads to an seems to be a compromise between hindered diffusion of the SDS mo-
enhanced nucleation rate, thus, to smaller particles [19]. However, it lecules because of increased viscosity and the improved steric stabili-
was assumed that there would be again an increase in particle size by zation caused by the polymeric structure of HPMC. Nevertheless, at
increasing fenofibrate mass fraction because the possibility of agglom- high SDS concentrations, the influence of HPMC is negligible. Fur-
eration and aggregation phenomena was enhanced [11]. Furthermore, thermore, higher amounts of SDS basically led to a decreased particle
aging processes like Ostwald ripening are gaining importance at higher size. Unfortunately, the solubility of fenofibrate increases with higher
solids contents. This could not be observed for the performed experi- amounts of SDS [33]. Since it was aimed to produce the maximum
ments. Unfortunately, still higher mass fractions could not be in- amount of solid fenofibrate nanoparticles, it was preferred to use low
vestigated because of the solubility of fenofibrate in ethanol (51.1 mg/ amounts of SDS instead of further increasing the concentration of SDS.
ml). Since solvent to antisolvent ratio was kept constant for these series For that reason, an optimized ratio of SDS and HPMC to fenofibrate was

560
S. Melzig, et al. Chemical Engineering Journal 371 (2019) 554–564

Fig. 6. Concentration of the solvent in the microsystem under variation of the antisolvent pressure via LIF measurements.

Fig. 7. Influence of fenofibrate mass fraction on particle size and polydispersity Fig. 8. Influence of the ratio of stabilizers to fenofibrate (FEN) on particle size
index. and polydispersity index.

found for SDS at 0.26 and for HPMC at 0.13. It was possible to produce reproducibility.
fenofibrate nanoparticles with a particle size of 240 nm ( ± 3 nm) and a
PdI of 0.036 ( ± 0.006) using this formulation. The standard deviations
were determined on the basis of 3 experiments. The low values proof 3.2.3. Antisolvent pressure
that HPAP enables the production of nanosuspensions with very good The variation of the antisolvent pressure affects both the process
and the formulation due to changed mixing behavior and altered ratio

561
S. Melzig, et al. Chemical Engineering Journal 371 (2019) 554–564

pressures lead to less particulate product after precipitation and more

dissolved API, thus, aging processes such as the formation of large
crystals due to Ostwald ripening would be favored. Consequently, an-
other reason could be uncovered to corroborate the need to maintain
the ratio of antisolvent to solvent pressure at 10 for further experi-
ments.

3.2.4. Temperature
It has already been proved that lower antisolvent temperatures lead
to smaller particles [11]. This is because of higher supersaturation
which increases the nucleation rate. As it is shown in Fig. 11, particle
size and polydispersity index stay almost constant while decreasing the
antisolvent temperature from 25 °C down to 5 °C. However, if the
temperature was increased up to 45 °C an increase in particle size and
polydispersity index could be observed. For that reason, an antisolvent
temperature of 25 °C is preferred for the precipitation of fenofibrate
Fig. 9. Influence of antisolvent pressure on particle size and polydispersity nanoparticles via HPAP.
index under constant and variable fenofibrate mass fraction ξ in the final pro- Nevertheless, it could be noted that the product temperature was
duct formulation. within a range of 10 °C despite the fact that antisolvent temperature
was varied over a range of 40 °C. This is caused by the accelerated heat
of antisolvent to solvent, respectively. For this reason, two different transfer due to the small dimensions of the microsystem and by the
formulation strategies were pursued. On the one hand, the concentra- solvent flow. Both are at ambient temperature so that antisolvent flows
tions of fenofibrate, HPMC and SDS were held constant in the solvent at colder temperatures will be heated up and antisolvent flows at
and antisolvent, respectively. On the other hand, fenofibrate con- warmer temperatures will be cooled down. However, the product
centration in the solvent as well HPMC and SDS concentration in the temperature at an antisolvent temperature of 25 °C is also increased.
antisolvent were adjusted to hold API mass fraction and ratio of stabi- This is because of the generated heat during the injection of the anti-
lizers to API constant during the experiments. solvent through the nozzles. For that reason, the temperature has to be
As it is shown in Fig. 9, the first formulation strategy (ξ ≠ const.) controlled directly at the microsystem and not only at the pressure
leads to a very strong increase in particle size and PdI if the antisolvent intensifiers if well-defined process temperatures are important for the
pressure was reduced. Particle sizes > 535 nm and PdI > 0.19 were product properties. In the case of the HPAP, it is sufficient if the anti-
reached at antisolvent pressures of 100 bar. The reason for this fact solvent temperature is not higher than 25 °C resulting in product tem-
could be the lower mixing intensities due to the lower antisolvent peratures lower than 30 °C.
pressures. However, this great impact on product properties is more
likely due to the change in the final formulation. Fenofibrate mass 3.2.5. Overall pressure
fraction increased as well as stabilizer amount decreased because of It has been reported in literature that a short mixing time is the most
changed volume fractions of antisolvent and solvent (compare Fig. 3) important feature to produce nanoparticles as small as possible via
resulting in formation of agglomerates due to increased particle con- antisolvent precipitation [10]. The mixing time of the used device is
centration and insufficient stabilization. The second formulation strongly dependent on the mixing intensity which is directly adjustable
strategy (ξ = const.) enables to investigate the influence of mixing in- via the pressure of the volume flows. For that reason, it was assumed
tensity and of antisolvent to solvent ratio only. Reducing the antisolvent that a simultaneous increase of antisolvent and solvent pressure leads to
pressure leads to a less intensive mixing and a decreased ratio of anti- finer particles. As it is shown in Fig. 12, particle size slightly increases
solvent to solvent. Both cause a lower initial supersaturation and, with higher pressures. The reason for this behavior could be the strong
therefore, a coarser particle size. The increase in particle size at lower increase in the process temperature, which is indirectly presented by
antisolvent pressures is not that significant in comparison with the first the product temperature. It can be observed, that the product tem-
formulation strategy because the influence of the resulting product perature increases from 29.5 °C at an antisolvent pressure of 200 bar to
formulation was eliminated. Also the polydispersity index increases over 42 °C at an antisolvent pressure of 1000 bar. Because of that, initial
with decreased antisolvent pressure because of the longer mixing time supersaturation at the mixing step was reduced due to the higher so-
due to the lower mixing intensity (compare Fig. 6). However, it seems lubility of fenofibrate at higher temperatures [34]. Consequently, nu-
that a further increase of antisolvent pressure (> 200 bar) leads to even cleation rate decreased and led to coarser particles. The particle size
smaller particles. Nevertheless, a further increase of the antisolvent distribution was in all cases relatively narrow and could be presumed
pressure was not reasonable because of the antisolvent backflow which constant. The values of the PdI are in a range of 0.02 and 0.05. The
would lead to cut off from the solvent stream and, therefore, to un- product volume flow rate could be increased up to 229 ml/min and
defined mixing properties (compare Fig. 6). Furthermore, this would even at this high flow rate, it was still possible to produce narrowly
lead to much lower ethanol amounts which would not be applicable to distributed particles with a size of 264 nm in a continuous process.
produce a suspension with a mass fraction of 0.35. Because of the According to literature, product volume flow rates in microsystems are
reasons stated above, a ratio of antisolvent to solvent pressure of 10 was typically 1–24 ml/min [9,19,32]. This fact represents the difference of
selected for further experiments. the system used in this study and the commonly used microfluidic set
In addition, the solubility of fenofibrate dramatically increases at ups for the precipitation of API nanoparticles.
higher proportions of ethanol or lower ratios of antisolvent to solvent,
respectively (Fig. 10). The saturation concentration of fenofibrate could 4. Conclusion
be deduced from the diagram for different antisolvent pressures by
knowing the antisolvent to solvent ratio or the solvent volume fraction Fenofibrate nanoparticles were synthesized via high-pressure anti-
depicted in Fig. 3. For example, the solubility of fenofibrate is 3.5 times solvent precipitation (HPAP) using a customized microsystem. First of
higher in the resulting product flow if the antisolvent pressure was all, the microsystem was characterized regarding its streaming behavior
reduced from 200 to 100 bar. That implies that lower antisolvent and its resulting mixing ratios of antisolvent and solvent flow since
mixing properties during the precipitation process as well as the final

562
S. Melzig, et al. Chemical Engineering Journal 371 (2019) 554–564

Fig. 10. Saturation solubility of fenofibrate in a mixture of ethanol (solvent) and deionized water (antisolvent) at 25 °C.

the antisolvent streams through the nozzles into the main channel.
Cavitation zones appear singularly at low pressures (75–125 bar) and
largely and permanently at the sidewall of the microchannels at high
pressures (≥125 bar). With the support of micro particle image velo-
cimetry (µPIV) and laser induced fluorescence (LIF) measurements, it
was shown that the cavitation reinforced the formation of a centered
fluid flow resulting in shorter distances and, therefore, in a faster
mixing process through intensification of the convective and the dif-
fusive mixing properties. Furthermore, it was observed that a backflow
ahead of the inject zone was formed at an antisolvent pressure of
200 bar. This led to the assumption that antisolvent pressures greater
than 200 bar in combination with a solvent pressure of 20 bar are not
applicable to produce well defined products. The backflow could lead to
undefined mixing states because of increased residence time and,
hence, longer mixing times resulting in coarser particles and a broader
Fig. 11. Influence of antisolvent temperature on particle size, polydispersity size distribution. Taking this into consideration, an optimized ratio of
index and product temperature.
antisolvent to solvent pressure of 10 was determined. Moreover, it was
unraveled that a fenofibrate mass fraction of 0.35% as well as a com-
product formulation are of the utmost importance in order to produce bination of HPMC and SDS as stabilizers with optimized ratios of sta-
highly defined drug nanoparticles. It was visualized that the increase of bilizers to fenofibrate (HPMC 0.13; SDS 0.26) enabled the production of
antisolvent pressure leads to the formation of cavitation bubbles at the fenofibrate nanoparticles with sizes of 240 nm and a very narrow size
outlet of the antisolvent stream because of the high pressure drop when distribution (PdI 0.04). In addition, the variation of antisolvent

Fig. 12. Influence of overall pressure on particle size, polydispersity index, product temperature and product volume flow rate while keeping ratio of antisolvent to
solvent pressure constant at 10.

563
S. Melzig, et al. Chemical Engineering Journal 371 (2019) 554–564

temperature (5–45 °C) shows that under 25 °C a reduction of tempera- ultrafine particles of beclomethasone dipropionate for dry powder inhalation, Int. J.
ture has nearly no influence but an increase leads to coarser nano- Pharm. 436 (2012) 1–9.
[13] D. Odetade, G.T. Vladisavljevic, Microfluidic fabrication of hydrocortisone nano-
particles. Furthermore, the established HPAP was used to produce fe- crystals coated with polymeric stabilisers, Micromachines 7 (2016) 236.
nofibrate nanoparticles with product volume flow rates from 95 up to [14] C. Knieke, A. Rawtani, R.N. Davé, Concentrated fenofibrate nanoparticle suspen-
229 ml/min resulting in very similar product properties. sions from melt emulsification for enhanced drug dissolution, Chem. Eng. Technol.
37 (2014) 157–167.
These results represent a good scalability either through numbering [15] Q.-P. Huang, J.-X. Wang, Z.-B. Zhang, Z.-G. Shen, J.-F. Chen, J. Yun, Preparation of
up or through volume flow adjustment. In addition, the product volume ultrafine fenofibrate powder by solidification process from emulsion, Int. J. Pharm.
flow rate is higher compared to most of other developed microsystems. 368 (2009) 160–164.
[16] C. Knieke, M.A. Azad, R.N. Davé, E. Bilgili, A study of the physical stability of wet
This enables higher production capacities of suspensions for further media-milled fenofibrate suspensions using dynamic equilibrium curves, Chem.
processing and, prospectively, for reasonable market supply. Moreover, Eng. Res. Des. 91 (2013) 1245–1258.
microsystems enable a very high controllability over the precipitation [17] A. Bitterlich, C. Laabs, E. Busmann, A. Grandeury, M. Juhnke, H. Bunjes, A. Kwade,
Challenges in nanogrinding of active pharmaceutical ingredients, Chem. Eng.
process because of the small reaction volume. Additionally, micro-
Technol. 37 (2014) 840–846.
systems enable the production in a continuous way which opens new [18] T. Lorenz, S. Bojko, H. Bunjes, A. Dietzel, An inert 3D emulsification device for
opportunities in process design for pharmaceutical applications. individual precipitation and concentration of amorphous drug nanoparticles, Lab
Finally, it can be summarized that the synthesis of API nanoparticles Chip 18 (2018) 627–638.
[19] Y. Dong, W.K. Ng, J. Hu, S. Shen, R.B.H. Tan, Continuous production of redis-
using HPAP was provided with a sound process characterization. HPAP persible and rapidly-dissolved fenofibrate nanoformulation by combination of mi-
shows great potential for the production of drug formulations with crofluidics and spray drying, Powder Technol. 268 (2014) 424–428.
highly defined properties (very narrow particle size distribution and [20] C. Beck, S.V. Dalvi, R.N. Dave, Controlled liquid antisolvent precipitation using a
rapid mixing device, Chem. Eng. Sci. 65 (2010) 5669–5675.
low particle size) by superior process control. [21] J. Hu, W.K. Ng, Y. Dong, S. Shen, R.B.H. Tan, Continuous and scalable process for
water-redispersible nanoformulation of poorly aqueous soluble APIs by antisolvent
Acknowledgements precipitation and spray-drying, Int. J. Pharm. 404 (2011) 198–204.
[22] G.L. Amidon, H. Lennernäs, V.P. Shah, J.R. Crison, A theoretical basis for a bio-
pharmaceutic drug classification: the correlation of in vitro drug product dissolu-
The authors gratefully acknowledge the financial support by the tion and in vivo bioavailability, Pharm. Res. 12 (1995) 413–420.
Lower Saxony Ministry of Science and Culture within the research [23] C. Richter, D. Stegemann, A. Vierheller, T. Gothsch, J.H. Finke, A. Kwade,
C.C. Müller-Goymann, A. Dietzel, S. Büttgenbach, Innovative process chain for the
project “Processing of poorly soluble drugs at small scale”. The authors development of wear resistant 3D metal microsystems, Microelectron. Eng. 110
would also like to thank Clara Sangrós and Zhaoping Guo for their (2013) 392–397.
assistance as well as Novartis AG for supplying the fenofibrate. [24] C. Richter, T. Krah, S. Büttgenbach, Novel 3D manufacturing method combining
microelectrial discharge machining and electrochemical polishing, Microsyst.
Technol. 18 (2012) 1109–1118.
References [25] T. Gothsch, J.H. Finke, S. Beinert, C. Lesche, J. Schur, S. Büttgenbach, C. Müller-
Goymann, A. Kwade, Effect of microchannel geometry on high-pressure dispersion
[1] A.M. Thayer, Finding solutions, Chem. Eng. News 88 (2010) 13–18. and emulsification, Chem. Eng. Technol. 34 (2011) 335–343.
[2] W. Nernst, E. Brunner, Theorie der Reaktionsgeschwindigkeit in heterogenen [26] J.H. Finke, S. Niemann, C. Richter, T. Gothsch, A. Kwade, S. Büttgenbach,
Systemen, Z. Phys. Chem. 47 (1904) 52–102. C.C. Müller-Goymann, Multiple orifices in customized microsystem high-pressure
[3] A.A. Noyes, W.R. Whitney, The rate of solution of solid substances in their own emulsification: the impact of design and counter pressure on homogenization effi-
solutions, J. Am. Chem. Soc. 19 (1897) 930–934. ciency, Chem. Eng. J. 248 (2014) 107–121.
[4] S. Melzig, D. Niedbalka, C. Schilde, A. Kwade, Spray drying of amorphous ibuprofen [27] J.H. Finke, J. Schur, C. Richter, T. Gothsch, A. Kwade, S. Büttgenbach, C.C. Müller-
nanoparticles for the production of granules with enhanced drug release, Colloids Goymann, The influence of customized geometries and process parameters on na-
Surf., A 536 (2018) 133–141. noemulsion and solid lipid nanoparticle production in microsystems, Chem. Eng. J.
[5] S. Melzig, J.H. Finke, C. Schilde, A. Kwade, Formation of long-term stable amor- 209 (2012) 126–137.
phous ibuprofen nanoparticles via antisolvent melt precipitation (AMP), Eur. J. [28] T. Gothsch, C. Richter, S. Beinert, C. Schilcher, C. Schilde, S. Büttgenbach,
Pharm. Biopharm. 131 (2018) 224–231. A. Kwade, Effect of cavitation on dispersion and emulsification process in high-
[6] F. Flach, C. Konnerth, C. Peppersack, J. Schmidt, C. Damm, S. Breitung-Faes, pressure microsystems (HPMS), Chem. Eng. Sci. 144 (2016) 239–248.
W. Peukert, A. Kwade, Impact of formulation and operating parameters on particle [29] T. Gothsch, C. Schilcher, C. Richter, S. Beinert, A. Dietzel, S. Büttgenbach,
size and grinding media wear in wet media milling of organic compounds – a case A. Kwade, High-pressure microfluidic systems (HPMS): flow and cavitation mea-
study for pyrene, Adv. Powder Technol. 27 (2016) 2507–2519. surements in supported silicon microsystems, Microfluid. Nanofluid. 18 (2015)
[7] P. Erfle, J. Riewe, H. Bunjes, A. Dietzel, Optically monitored segmented flow for 121–130.
controlled ultra-fast mixing and nanoparticle precipitation, Microfluid. Nanofluid. [30] S. Büttgenbach, A. Balck, S. Demming, C. Lesche, M. Michalzik, A.T. Al-Halhouli,
21 (2017) 179. Development of on chip devices for life science applications, Int. J. Eng. 3 (2009)
[8] R. Othman, G.T. Vladisavljević, Z.K. Nagy, Preparation of biodegradable polymeric 148–158.
nanoparticles for pharmaceutical applications using glass capillary microfluidics, [31] R. Lindken, M. Rossi, S. Grosse, J. Westerweel, Micro-particle image velocimetry
Chem. Eng. Sci. 137 (2015) 119–130. (microPIV): recent developments, applications, and guidelines, Lab Chip 9 (2009)
[9] H.S.M. Ali, P. York, N. Blagden, Preparation of hydrocortisone nanosuspension 2551–2567.
through a bottom-up nanoprecipitation technique using microfluidic reactors, Int. [32] P. Valeh-e-Sheyda, M. Rahimi, A. Parsamoghadam, H. Adibi, Effect of microchannel
J. Pharm. 375 (2009) 107–113. confluence angles on size reduction of curcumin nano-suspension via liquid anti-
[10] H. Zhao, J.-X. Wang, Q.-A. Wang, J.-F. Chen, J. Yun, Controlled liquid antisolvent solvent precipitation process, J. Taiwan Inst. Chem. Eng. 46 (2015) 65–73.
precipitation of hydrophobic pharmaceutical nanoparticles in a microchannel re- [33] S. Jamzad, R. Fassihi, Role of surfactant and pH on dissolution properties of feno-
actor, Ind. Eng. Chem. Res. 46 (2007) 8229–8235. fibrate and glipizide–a technical note, AAPS PharmSciTech 7 (2006) E17–E22.
[11] J.-X. Wang, Q.-X. Zhang, Y. Zhou, L. Shao, J.-F. Chen, Microfluidic synthesis of [34] H. Sun, B. Liu, P. Liu, J. Zhang, Y. Wang, Solubility of fenofibrate in different binary
amorphous cefuroxime axetil nanoparticles with size-dependent and enhanced solvents: experimental data and results of thermodynamic modeling, J. Chem. Eng.
dissolution rate, Chem. Eng. J. 162 (2010) 844–851. Data 61 (2016) 3177–3183.
[12] L.-M. Xu, Q.-X. Zhang, Y. Zhou, H. Zhao, J.-X. Wang, J.-F. Chen, Engineering drug

564