You are on page 1of 67

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/332091415

Scale-up of pharmaceutical spray drying using scale-up rules: A review

Article  in  International Journal of Pharmaceutics · May 2019


DOI: 10.1016/j.ijpharm.2019.03.047

CITATIONS READS

0 107

2 authors:

Sadegh Poozesh Ecevit Bilgili


Tuskegee University New Jersey Institute of Technology
22 PUBLICATIONS   469 CITATIONS    107 PUBLICATIONS   1,602 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Modelling View project

Special Issue of Pharmaceutics: Invitation from Guest Editors View project

All content following this page was uploaded by Sadegh Poozesh on 30 March 2019.

The user has requested enhancement of the downloaded file.


Accepted Manuscript

Review

Scale-up of Pharmaceutical Spray Drying Using Scale-up Rules: A Review

Sadegh Poozesh, Ecevit Bilgili

PII: S0378-5173(19)30233-9
DOI: https://doi.org/10.1016/j.ijpharm.2019.03.047
Reference: IJP 18231

To appear in: International Journal of Pharmaceutics

Received Date: 24 January 2019


Revised Date: 20 March 2019
Accepted Date: 21 March 2019

Please cite this article as: S. Poozesh, E. Bilgili, Scale-up of Pharmaceutical Spray Drying Using Scale-up Rules:
A Review, International Journal of Pharmaceutics (2019), doi: https://doi.org/10.1016/j.ijpharm.2019.03.047

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Scale-up of Pharmaceutical Spray Drying Using Scale-up Rules: A Review

Sadegh Poozesh1* and Ecevit Bilgili2


1
Mechanical Engineering Department, Tuskegee University, Tuskegee, AL 36088
2
Chemical and Materials Engineering Department, New Jersey Institute of Technology,
Newark, NJ 07102

*Corresponding author, Email: sadegh.poozesh@gmail.com, Tel: (646) 283-8782

Abstract
Spray drying is one of the widely used manufacturing processes in pharmaceutical industry.
While there are voluminous experimental studies pertaining to the impact of various process–
formulation parameters on the quality attributes of spray dried powders such as particle size,
morphology, density, and crystallinity, there is scant information available in the literature
regarding process scale-up. Here, we first analyze salient features of scale-up attempts in
literature. Then, spray drying process is analyzed considering the fundamental physical
transformations involved, i.e., atomization, drying, and gas–solid separation. Each
transformation is scrutinized from a scale-up perspective with non-dimensional parameters &
multi-scale analysis, and comprehensively discussed in engineering context. Successful scale-
up entails similar key response variables from each transformation across various scales.
These variables are identified as droplet size distribution, outlet temperature, relative
humidity, separator pressure loss coefficient, and collection efficiency. Instead of trial-and-
error-based approaches, this review paper advocates the use of mechanistic models and scale-
up rules for establishing design spaces for the process variables involved in each
transformation of spray drying. While presenting a roadmap for process development and
scale-up, the paper demonstrates how to bridge the current gap in spray drying scale-up via a
rational understanding of the fundamental transformations.

Keywords: spray drying, scale-up, atomization, droplet size, process modeling, non-
dimensional parameters, particle engineering

1 Introduction

Spray drying has been a robust and widely used manufacturing process especially in food and
pharmaceutical industries since 1940s (Wang, 2015). Its applications for pharmaceutical
products include amorphous solid dispersions (ASDs) (Gu et al., 2015; Sawicki et al., 2016)
and spray-dried nanocomposites (Azad et al., 2015; Li et al., 2018) for bioavailability of
poorly soluble drugs, controlled-release (Hacene et al., 2016; Lu et al., 2016), taste masking
(Taki et al., 2017), inhalation products (Chen et al., 2016; Mönckedieck et al., 2016), and
aseptic production of biopharmaceuticals (Chen et al., 2016; Kemp, 2017). Owing to its
reproducibility, continuous mode of operation, high capacity, fast drying capability, and short
exposure of product to high processing temperatures associated with fast evaporative cooling,
this process has become very popular in the pharmaceutical and biotechnology industry (Al-
Khattawi et al., 2018; Lowinger et al., 2015; Maas et al., 2011). These advantages make it
very competitive to freeze drying, which is currently a dominant unit operation in bio-
pharmaceuticals (Bhakay et al., 2018; Sosnik and Seremeta, 2015). Similarly, amongst all the
techniques available for bioavailability enhancement, spray drying and hot melt extrusion
(HME) appear to be the two most widely used platform approaches (Singh and Van den
Mooter, 2016). Although the choice of the process depends on many factors, spray drying has
various advantages over HME, especially for thermolabile and high melting temperature
drugs (Guns et al., 2011).

The current spray drying market is driven, to a large extent (over 80%), by pharmaceutical
products such as active pharmaceutical ingredients (APIs) and intermediate–final products
(Smith, 2014). Based on a recent report on market share of spray drying in pharmaceutical
industry, it is predicted that this market will experience a steady 17% growth between 2018–
2028 (Tzvetanov, 2018). The future advances will be jointly led by the steady rise of the
pharmaceuticals’ market and emergence of new technologies/applications that can protect
sensitive APIs through their processing into final drug products.

Pharmaceutical spray dryers are available in a wide range of scales: from lab units where
milligrams of material can be produced to large commercial units capable of handling
multiple tons of powder per day. As can be seen from Figure 1, if necessary, a spray drying
process may be scaled up directly from the laboratory to a final production scale. However,
the quantities required for clinical trials are more efficiently produced in pilot or small
production scales, where product losses and scale-up risk are noticeably lower. Therefore,
these intermediate production scales are commonly used during the development of the
process (Faure et al., 2001). Although, there are limited options to control drying space across
these scales, differences arise between scales in the evaporation capacity, the throughput,
particle physical and chemical characteristics (e.g. morphology, size, phase behavior) and
powder properties (e.g. flowability, tabletability). For example, spherical and denser particles
with better powder flowability are rendered in larger scales compared to the lab ones (Al-
Khattawi et al., 2018). These changes, in turn, impact downstream properties of the final
dosages such as tablet hardness, friability, disintegration time, and dissolution rate. In view of
these facts, improper process scale-up may lead to considerable losses of expensive materials
and ultimately jeopardize the timelines of a clinical program.

Regardless of numerous advantages displayed by spray drying and its promise for the future,
discrepancies in quality of products across different scales have been the major obstacle in
product development and manufacture (Gil et al., 2010). This is largely due to the fact that
spray drying involves a combination of complex physical transformations that are multi-scale
and multi-physics in nature. They include mixing of API and excipients in an appropriate
solvent, liquid atomization, evaporation of solvent (drying), gas–particle separation, which is
usually followed by a secondary drying via either fluidized bed or vacuum tray dryers
(Masters, 1994). In the first step, feed solution is prepared using a solvent or solvent system
that can readily dissolve the drug, polymer and other functional excipients. In case of
suspensions, drug particles are suspended in a liquid with additional excipients. Then, the
solution is sprayed into the drying chamber, by increasing surface to volume ratio, and in turn
augmenting mass and heat transfer. Meanwhile, due to differences in vapor pressure of
droplet and gas, solvent component of droplets is gradually transferred into gas stream.
Particles formed during evaporation either collide with the wall of the spray drying chamber
or find their ways to cyclone or baghouse filters, where they are separated from gas. Cyclone
separators function based on differences in hydrodynamically induced forces between drying
gas and particles, whereas bag filters use the concept of impaction onto a filter or electrostatic
precipitation. Collected powder may go through a secondary drying step to remove solvent
levels down to target levels depending on both dosing considerations for human use and
physical stability (Abraham, 2009). The complexity of these physical transformations renders
scale-up of a spray drying process challenging and entails a fundamental understanding of
fluid mechanics, heat and mass transfer, and thermodynamics. Unfortunately, current routine
scale-up practice involves first setting a feed capacity based on target batch size and then
adjusting atomization and drying capacity (mostly in terms of processing gas flow rate and
the drying column height) (Paudel et al., 2013). After rigging up the initial setup, rigorous,
costly empirical tests, followed by statistical analysis must be conducted.

Despite the complexity of the physical transformations rendering scale-up a challenge, there
are only a few review papers that scrutinized some aspects of spray drying scale-up. Sosnik
and Seremeta (2015) discussed two main challenges in technology transfer regarding
differences in product yield and particle size distributions across scales. The authors
suggested to remedy these differences by using B-290 spray dryer as the lab-scale
manufacturing unit. This was linked, in part, to the successful scale-up reports using it as the
small-scale dryer. Furthermore, this was due to its capability in capturing very fine particles
to minimize throughput loss during the separation process. Singh and Mooter (2016)
reviewed various formulation and process variables influencing the critical quality attributes
of the ASDs and briefly discussed main challenges in scale-up process. They suggested that
shifting from one scale to another mandates adjustments in: (1) feeding system; (2)
atomization device type, location and conditions; (3) drying gas dispensing system; (4)
chamber dimensions; (5) exhaust gas duct and; (6) shift from usually an open-loop system
(seen in lab-versions) to a closed-loop one (seen in pilot and production scales). A recent
review by Bellinghausen (2018) qualitatively suggested that a successful scale-up is relied on
controlling flow configuration (current or counter-current flow), atomization, particle
separation, and drying process. To facilitate scale-up, the author recommended to: (1) control
droplet size by using rotary atomizer, ultrasonic or mono-dispenser in order to produce same
size droplets; (2) have more control over drying by either creating laminar or counter-current
flow; and (3) use “flow curtains” to avoid particles flowing into wall regions. Al-Khattawi et
al. (2018), with a systematic overview of the particle delivery systems produced via spray
drying between 1990 and 2016, highlighted industrial scale-up approaches. It is offered by
the authors to match the key parameters across scales including: 1) feed properties and feed
solvent content; 2) atomized droplet size distribution; 3) the droplet drying history and
history distribution; 4) the desired particle/droplet collision history to form agglomerates; and
5) avoid wall contacts and build-up at all scales. They further articulated several basic scale-
up approaches such as thermodynamic space approach (see Dobry et al. (2009), statistically-
based design of experiments (see Lebrun et al. (2012), fundamental models including
computational approaches (Li and Zbiciński, 2005), and a hybrid approach, i.e., combination
of these approaches. In general, all these reviews are somewhat scattered in terms of their
coverage of the scale-up issues; rather, they mostly overviewed key formulation and process
variables, and then succinctly proposed ways to remedy scale up challenges. Thanks to these
works, arguably, there seems to be consensus on key response variables that govern the
product attributes, still there is no systematic and standard approach to guide users as to how
rationally lay the groundwork for upscaling. Unfortunately, the literature lacks the
fundamental engineering principles underpinning the robust process scale-up.
It is therefore the aim of this review paper to present a detailed, systematic scale-up approach
based on engineering analysis of individual transformations, i.e., atomization, drying, and
gas–solid separation in spray drying. This paper first covers the attempts made to scale up
spray drying in various industries, especially the ones concerning pharmaceutical
applications. Then, each transformation is scrutinized from a scale-up perspective, using non-
dimensional parameters and multi-scale analysis in engineering context. Successful scale-up
entails similar key response variables from each transformation across various scales. Based
on the aforementioned review papers and experimental literature, these key response
variables are identified. A general strategy is proposed to keep these response variables
similar across scales by making use of mechanistic models and empirical scale-up rules.
Furthermore, upscaling considerations pertinent to each transformation, atomizing, drying,
and finally separation will be offered and discussed in detail.

2 Attempts in spray drying scale-up

In pharmaceutical arena, only few attempts have been reported to use at least two spray driers
for production of solid-state formulations. They all aimed to increase throughput for a given
formulation, without compromising product quality. Here we summarize the findings from
these attempts.

2.1 Drying of biologics

Dry vaccine formulations, one of the recent solid-state formulations, were produced over two
scales (Kanojia et al., 2017). These formulations, compared to liquid-based counterparts, are
generally less sensitive to temperature induced degradation and attain extended shelf-life;
thereby improve cost effectiveness of vaccination programs by reducing vaccine wastage. In
lieu of freeze drying (lyophilization), which is a typical unit operation for solid-state vaccine
formulation, spray drying has been recently employed; and then scaled-up. For example, Zhu
et al. (2014) stabilized a plant-produced recombinant hemagglutinin (rHA) influenza antigen,
HAC1 via spray drying. They used a Buchi 290 mini spray dryer (Buchi Corporation, New
Castle, Delaware) at the lab-scale, and BLD-1 pilot-scale dryer (custom-built dryer, Bend
Research, Inc., Bend, Oregon) to produce powders from a feed solution of API and
excipients. Complete information on formulation and process conditions are listed in Table 1.
On the lab-scale dryer, eight formulations, each containing 360 μg/mL API, with different
excipients and different weight ratios in deionized water were spray dried. Based on antigen
stability and powder analyses of the eight formulations produced at lab-scale, only one
formulation was chosen, and used for scale-up at similar inlet temperature, while rendering
almost similar outlet temperature conditions for a 100-fold larger batch size. This was the
only criteria at which process operation at pilot spray dryer is based on. They concluded that
all batches-maintained antigen content and powder properties throughout the 6-month
stability study at all storage temperatures, suggesting that the pilot-scale spray-drying process
did not alter the immunogenicity of the HAC1 antigen.

2.2 Microencapsulation

In pharmaceutical industry, encapsulation can confer many novel functionalities such as


taste-masking, control release, physicochemical protection of fragile APIs, etc. (Hacene et al.,
2016; Lu et al., 2016; Taki et al., 2017). Excipients used to prepare such particles include
polymers, lipids or insoluble metal salts and oxides. Spray drying has been recognized as a
paramount mean to provide microencapsulation. Yet still, process scale-up, just like in other
applications of spray dryers, has not been explored and needs more attention. In an attempt to
scale-up microencapsulation through spray drying, Raffin et al. (2006b) aimed to produce
enteric microparticles containing sodium pantoprazole in both laboratory and pilot scales.
They prepared a solution out of sodium pantoprazole, as the prodrug candidate, and other
excipients in water on a lab-scale Buchi 290, as well as a pilot scale spray drier (Model S52
APV® Anhydro). In the pilot version, two different types of atomizers: a rotary and a two-
fluid were employed. Additionally, on the same pilot version, counter-flow regime where
processing gas flowed in the opposite direction of droplet flow, was also examined with two-
fluid atomizer. The procedure undertaken by the authors was to first find the best formulation
and process conditions and then set the pilot conditions accordingly. They concluded that the
powders prepared with rotating disc atomizer and counter-current dryer presented either
crystals on the particle surface or very high polydispersity. Product obtained from two-fluid
nozzle and low air pressure presented strings in the powder. Using the same atomizer but
higher air pressure, the microparticles presented high encapsulation efficiency and the highest
stabilization of formulation in acid medium.

2.3 Other industry sectors

Due to scarcity of spray drying scale-up attempts reported in literature, similar attempts in
other industries such as food engineering, where sensitivity of active compounds might not be
as critical as APIs, are reported in Table 1. This table presents information about the attempts
to scale up spray drying either from lab to pilot scale, pilot to production scale or even
directly from lab to production scale in food and pharmaceutical industries. Below are some
of the most important takeaways from the literature on this matter.

 In all the studies except two (Raffin et al., 2006a; Raffin et al., 2006b), authors tried
to use similar nozzle types, usually a two-fluid nozzle, as it is frequently used in
Buchi 290, the most well-known lab-scale spray dryer. Despite the widespread usage
of this nozzle in the literature, it is well-known that most of industrial spray dryers
are equipped with pressure spray technology (Singh and Van den Mooter, 2016).
Additionally, this type of nozzle results in small particle size (in the order of < 10 μm)
and low bulk density, which have a deleterious effect on the flowability of bulk
powder.
 A shift has been observed from smaller droplets to bigger ones, when going from
smaller scale to bigger scales. This has, in turn resulted in changes in powder residual
solvent content, due to changes in residence time and evaporation rate.
 In some of the studies, the lab-scale was practiced first to determine the most
stable/feasible formulation, and then this formulation was used in larger scale. In
these studies, the authors implicitly considered product stability as a success metric
for proper scale-up (Merlos et al., 2017; Zhu et al., 2014).
 In almost all these studies, outlet temperature, Tout, was kept similar across scales by
altering the inlet temperature, Tin, by up to 30%. This signifies the criticality of
maintaining Tout so as to ensure similarity of various quality attributes such as
chemical and physical homogeneity, from one scale to another.
 Some efforts have been made to match droplet size distribution to achieve similar
particles across scales; however, they have acknowledged dissimilarities between
particles, both chemically and physically across scales (Thybo et al., 2008a; Thybo et
al., 2008b).
 Some studies tried to match inlet and outlet temperatures throughout scales (Thybo et
al., 2008b; Zhu et al., 2014), by adjusting drying gas and feed flow rates.
 In almost all the studies, scale-up process is driven by trial and error approach through
varying process conditions and subsequent checking of product quality. The only
partially successful transfer has been reported by Kemp et al. (2016a), where they
could atomize similar droplets and subsequently similar particles sizes (1.5-2.5µm)
across scales. To some extent, they matched non-dimensional parameters for
atomization stage where they employed external mixing two-fluid nozzles for both lab
and production scales. Their results showed the impact of atomization gas flow rate to
feed flow rate ration (GLR) across the scales. This parameter is especially important
during atomization; and its impact will be discussed later.

The foremost purpose of the aforementioned studies was to match particles’ physical and
chemical attributes across different scales. As can be perceived from the above-mentioned
highlights drawn from the limited attempts on spray drying scale-up, more work is needed to
choose the right nozzle design, atomization operation conditions, drying gas conditions,
drying chamber design, etc. when handling a given formulation. We maintain in this review
paper that mechanistic modeling of the underlying transformations, i.e., atomization, drying,
and gas–solid separation (detailed in Section 3) could provide significant help in addressing
scale-up criteria. Modeling of atomization could aid in predicting the droplet size and process
throughput estimation, while considering any change in the atomizer type. Modeling of
drying and particle formation could account for drying kinetics and final morphology of the
particles; and finally, segregation analysis provides effective information on optimizing the
separation process.

3 Fundamental considerations in spray drying scale-up

The selection of the right scale involves several considerations, but ultimately, it is primarily
driven by the targeted process throughput and batch size requirements. For each scale, there
are input (process) variables that are strictly dependent on the scale. For a given formulation,
the input variables and equipment configuration (e.g. design, mode, etc.) determine the
product characteristics. The principal goal during scale-up is to maintain product quality and
normalized yield. While this goal can be achieved directly by varying the input variables at
the larger scale in a set of statistically-based design of experiments and testing the resultant
product attributes, such an empirical approach is extremely costly and does not yield reusable
fundamental knowledge. An alternative approach is to recognize the impact of underlying
fundamental transformations, i.e., atomization, drying, and gas–solid separation on some key
response variables. Based on an analysis of the literature (Al-Khattawi et al., 2018;
Bellinghausen, 2018; Singh and Van den Mooter, 2016; Sosnik and Seremeta, 2015), as well
as the mechanistic models of each transformation (see below), we identify these key response
variables as droplet size distribution, outlet temperature, relative humidity, separator pressure
coefficient loss, and collection efficiency. It is proposed that a successful scale-up entails
similar key response variables from each transformation across various scales (see Figure 2).
Such an approach has been successfully demonstrated for a pharmaceutical fluidized bed
granulation, where product attributes were maintained similar across scales when key
response variables were maintained similar by using few scale-up rules (Bilgili et al., 2007;
Bilgili et al., 2011). Below, we present analysis of atomization, drying, and gas–solid
separation, and the mechanisms by which they affect the key response variables from a scale-
up perspective.

3.1 Atomization

Atomization is the first and foremost important stage in spray drying process, as it has pivotal
role in determining both physical and chemical properties of final particles (Poozesh et al.,
2018c). The purpose of atomization is to vastly increase the specific surface area of the liquid
through generation of droplets and augment heat and mass transfer from liquid to processing
gas by enhancing feed solution exposure. This high surface exposure to a free to low-solvent
gas allows atomized droplets to dry very quickly; in lab-scale, in the order of milliseconds
and in commercial scale in the order of seconds.

Sprays may be produced in various ways, but essentially, a high relative velocity between the
liquid to be atomized and the surrounding gas is needed. Some atomizers such as pressure
and rotary types achieve this by discharging the liquid at high velocity into a nearly stagnant
gas. An alternative approach is to expose the relatively slow-moving liquid to a high-velocity
gas stream. The latter method is generally known as two-fluid atomization. Independent of
the device, atomization is a complex cascading phenomenon of breakup events where balance
of inertia or electrostatic force, rooted in the type of atomization, and viscous and surface
tension forces, modulated by the formulation of the feed solution, determine the angle and
penetration of the spray, as well as droplet size and velocity distributions (Durli et al., 2014;
Lefebvre and McDonell, 2017; Poozesh et al., 2018a).

A ubiquitous breakup behavior is observed for almost all conventional atomizers where
sheets are formed immediately adjacent to the nozzle tip; then by gradually going downward,
the ligaments are more easily visualized (Moon et al., 2014; Poozesh et al., 2018a). Finally,
droplets are more easily identified as a result of peeling off from the ligaments. The position
and timescale of these breakup processes (i.e. sheet formation, ligament formation and
droplet generation) are functions of nozzle design, operating condition, viscosity, density, and
surface tension of the feed solution, among other variables. Obtaining meaningful droplet size
distribution measurements must include consideration of the location of each of these
transition stages. Specifically, if measurements are focused on the region where sheet
formation or ligaments dominate, the meaning of “droplet size” is unclear. Thus,
measurements of droplet size should be done where the breakup is finished and little to no
change is observed further downstream of jet (although droplets may experience coalescence
downstream) (Poozesh et al., 2018c). Different measurement tools are available to monitor
droplet size distribution. Due to the development of modern technology such as powerful
computers and lasers, quantitative optical non-imaging light scattering spray characterization
techniques have been developed for non-intrusive spray characterization: Phase Doppler
Particle Analyzers (PDPA), laser diffraction analyzers, e.g., Malvern analyzers (Poozesh et
al., 2018c); imaging techniques, and the recent improvements in digital image processing
(Poozesh et al., 2018a). Among them, the PDPA has widely been tested and its usefulness for
spray characterization is recognized (Chen et al., 2018). Both PDPA and high-speed imaging
can render both droplet size distribution and velocity distribution.

In the following section, conventional as well as niche atomizers for spray drying
applications will be reviewed. Various models that account for the impact of atomization
conditions on the Sauter-mean diameter (SMD) of droplet size distribution, a major response
variable, are presented. SMD or D[3, 2] is the diameter of a sphere with the same volume to
surface area ratio of the entire spray and may be considered as a measure of the fineness of
the spray (Filkova and Mujumdar, 1995; Schick and Brown, 2005). This diameter is used to
describe the atomization process since it captures the total area for heat and mass transfer –
important for understanding the particle formation process. The order by which the
discussion is followed is based on the prevalence of each sprayer in pharmaceutical
applications. Driving forces and principal operation of each atomizer differ based on the
design and configuration. Unfortunately, because of intricacies involved in atomization, no
reliable simulation modeling has been proposed for finding SMD. Instead, this key response
factor in spray drying process, has been reported extensively by semi-empirical correlations.
In the following, the most common SMD relationships for spray drying nozzles are given
based on non-dimensional parameters. Since these relations do not rely on the dimension of
atomizers, they could be applied across any dryer.

3.1.1 Operational aspects and SMD correlations for various nozzles


a. Two-fluid nozzle
As can be seen from Table 1, two-fluid nozzles (also known as air-assisted atomizers and
pneumatic atomizers in other industries) are overwhelmingly used in pharmaceutics from lab,
all the way to commercial scales. This spray technology relies on the breakup of liquid
through impact with high-speed gas at the orifice. Shear forces induced by relative velocity
between high speed gas and intact flowing liquid create instabilities on the bulk liquid
surface, which results in sheet, ligament and finally droplets formation. For the most
commonly used designs, the contact between the liquid and gas phase takes place at the
nozzle exit. This principle is known as external mixing (Figure 3(a1)). While in the other
type, the liquid is injected into the high-velocity airstream in the form of one or more discrete
jets. This principle is commonly known as plain jet or internal mixing atomization (Figure 3
(a2)). Since in internal case, both liquid and gas are mixed before atomization, and as such,
number of variables are less compared to external nozzle; making scale-up of internal mixing
type more advantageous. Nonetheless, external mixing type is more prevalent in
pharmaceutical industry because of corrosion liability and less controllability of internal
mixing nozzles (Masters, 1994).

Due to the large body of literature on two-fluid nozzles, droplet size correlations for these
nozzles are relatively satisfactory to report. Often authors have developed empirical
expressions for droplet size distribution based on process parameters, geometrical dimensions
and feed solution properties based on their own type of nozzle; and developed an empirical
correlation for the mean droplet size. Therefore, as far as scale-up is concerned, only
frequently reported/used correlations associated with standard designs, based on non-
dimensional parameters are considered here.

In two-fluid nozzles, establishing an expression for SMD may be accomplished via


describing and capturing the various forces in the atomization process in terms of reduced or
dimensionless parameters which are needed during scale-up. This approach is meant to
compare the relative magnitude of various forces in terms of fundamental properties. The
Weber number, Weg  vg2  g dl /  , compares the dynamic gas pressure acting on the liquid

sheet to the liquid capillary pressure of the liquid sheet (Hede et al., 2008). Here, vg, ρg, dl and
σ are gas velocity, gas density, liquid orifice diameter, and liquid surface tension,
respectively. Another dimensionless number that relates gas inertia to the liquid viscous force
is the Reynolds number (Re) (Poozesh et al., 2016a; Saito and Williams, 2015). However,
while working with both We and Re number, to isolate inertia effects, another pi-number, the
Ohnesorge number, Oh   / l dl , can be derived via combination of We and Re, Oh =

We0.5/Re. This dimensionless number is purely dependent on formulation properties and


geometry of liquid orifice. The liquid density is represented by the variable ρl and other
variables were defined previously. In addition to the involved forces, the information on
phases mass flow rates are needed to analyze the dynamics of the atomization system. The
gas-to-liquid ratio, GLR  m g / m l , is defined as the ratio between gas and liquid mass flow

rates and has often been cited as a dimensionless quantity which captures the atomization
operating conditions (Lefebvre, 1980). It is worth mentioning that GLR, also known as ALR
(air-to-liquid ratio) or ALM (air-to-liquid mass ratio), makes use of the atomization gas
flowrate, not the processing gas flowrate, which largely determines the drying rate. Also,
noteworthy that the discussed non-dimensional parameters are utilized for other types of
nozzles.

For internal mixing nozzles, Rizk and Lefebvre (1984) studied spray characteristics for
simple plain-jet nozzles and offered a correlation according to the following equation based
on experimental data determined by light-scattering technique (the equations are modified
based on the non-dimensional variables in this paper).

C2
 1  GLR  1
SMD / dl  C1    C3Oh(1  ) (1)
Weg GLR  GLR

All variables together with their units are listed in the nomenclature section. Cs are the
constants which vary between studies. As an example, C1, C2, and C3 in (Rizk and Lefebvre,
1984) are found out to be 0.48, 0.4, and 0.15 respectively.

On the other hand, for external mixing, where larger body of literature exists, considering the
ratio of the energy present in the atomizing gas to the surface energy as the main factor
determining the droplet sizes, and having in mind also the liquid loading and the liquid
viscosity, Groom et al. (2005) obtained a similar semi-empirical relationship to predict SMD:

C
 Weg 
2

SMD / d L  C1  2
(1  C3Oh) (2)
 (1  GLR ) 

C1, C2, and C3 are constants that partially depend on the atomizer design configuration.
Nonetheless, typical values that may apply to a wide design configurations are: C1=0.3-0.4,
C2= (0.6)-(0.8), and C3= 0.2-0.4.
In scale-up, normally the same solvent and formulation is used; hence, feed physical
properties do not change. The only factors that change are gas velocity, design configuration
(e.g. liquid or gas orifice diameter) and GLR.

b. Pressure nozzle

When a liquid is discharged through a small aperture under high applied pressure, the
pressure energy is converted into the kinetic energy to overcome surface tension force and
disintegrate bulk liquid (Lefebvre and McDonell, 2017). Contrary to two-fluid, the air is
stagnant for this nozzle. Figure 3(b1) shows a plain-orifice type pressure atomizer which
renders sprays with a cone angle that usually lies between 5° and 15°. This cone angle is a
weak function of the diameter and length/diameter ratio of the orifice and rather mainly
dependent on physical properties of feed liquid such as the viscosity and surface tension, and
the turbulence of the issuing jet. An increase in turbulence increases the ratio of the radial to
the axial component of velocity in the jet and thereby increases the cone angle (Lefebvre and
McDonell, 2017). The advantage of pressure atomizers is their simplicity in design and ease
of operation, yet they are typically unable to produce fine sprays with viscous liquids; and
they are usually eroded in time, as the continuous exerted shear force by the liquid is high.

Instead of a plain orifice, for most applications, pressure swirl atomizers (Figure 3(b2)) are
used as this orifice renders superior control over cone angle, and in turn, enhances droplets
exposure to drying gas. Pressure and pressure swirl atomizers are similar in a sense that both
simply inject liquid at a high pressure through an orifice. On the other hand, pressure-swirl
atomizers impart a radial acceleration to the liquid inside the nozzle to facilitate the liquid
spread radially upon exit, rendering finer droplets with minimized pressure, as a result of
superposing pressure and centrifugal forces. Typically they produce a hollow spray cone with
a wide angle, 15-175o (Pierpont and Reitz, 1995). Adding swirl to the liquid feed is a
technique that is used in other nozzle types such as two-fluid and rotary atomizers. For
instance, usually two-fluid nozzles employed in Buchi 290 benefit from an additional internal
vane (swirl insert).

Although in this atomizer, in view of scale-up, droplet size dependency on fewer parameters
makes it advantageous compared to counterparts, in order to change droplet size at constant
feed flow, it is necessary to change the nozzle dimensions or design. Furthermore, to handle
more viscous feeds, higher hydraulic pressure is required, which may not be realizable for a
given nozzle design. Nevertheless, this atomizer favors production scale drying with more
than 5 kg/h feed flow rate. It renders medium to large particles (30–200 μm) with favorable
powder flowability and with narrower span (1.4-1.8) compared to two-fluid nozzles.

The primary variables in the design of hydraulic nozzles are the geometry, shape, and
arrangement of the orifice through which the fluid flows before exiting the nozzle. In plain-
orifice pressure atomizers, disintegration of the jet into drops is promoted by an increase in
flow velocity, which increases both the level of turbulence in the issuing jet and the
aerodynamic drag forces exerted by the surrounding medium; it is opposed by surface
tension, which delays the onset of atomization by resisting breakup of the ligaments. So, We
number, which weighs these two effects, plays an important role in determining droplet size,
among other spray characteristics. One ubiquitous non-dimensional equation for describing
plain orifice nozzle is as following (Lee et al., 2007; Sallam et al., 1999; Sallam and Faeth,
2003):

C2
 x/ 
SMD /   C1  0.5  (3)
Wel 

Where x is the distance from nozzle, Ʌ is the radial spatial integral scale of turbulence, and
Wel is Weber number at nozzle exit based on liquid properties. One good estimation for Ʌ is
one-eigth of nozzle diameter (Lee et al., 2007).

When it comes to pressure-swirl nozzle (Figure 3(b2)), as the process of atomization gets
highly complex, it is convenient to subdivide it into two main stages. The first stage
represents the generation of surface instabilities due to the combined effects of hydrodynamic
and aerodynamic forces. The second stage is the conversion of surface protuberances into
ligaments and then drops. This simplified approch leads to the following final SMD which
results in summation over both of these statges (Lefebvre and McDonell, 2017).

SMD / ts  SMD1 / ts  SMD2 / ts  C1 (Re We )C2  C3WeC4 (4)

where ts is the initial thickness of the liquid sheet at the nozzle exit, which along with Re and
We can be obtained as following:

 g vl2ts
0.25
d m   l vl ts
ts  2.7  l l  cos ( ), Re  and We=
 pl l  l 

Δp is feed pressure differential; other symbols are explained in the nomenclature section.
c. Rotary nozzles

Rotary nozzles exploit centrifugal force created by rotating disk or plates. The liquid is fed
onto a rotating surface, and the rotating disk or plate spins at very high velocities, where
centrifugal forces overcome surface tension to produce droplets. After atomizing and leaving
the nozzle surface, the spray contacts with the surrounding gas. The main advantages of these
nozzles are absence of clogging issues even with viscous suspension and solutions and high
uniformity of produced droplets in terms of both velocity and size. Usually, gas is
accompanied with these nozzles for facilitating atomization and creating desired spray shape
(even though this type has not been used in pharmaceutical industry). These nozzle types are
called twin fluid rotary nozzles, which include well-known rotary bells in automotive coating
industry and combustion.

Under rotary atomizers, there are four main classes of nozzle, namely, flat disk (Figure 3(c1))
which in turn is divided into vaneless, and vanned where vanes/grooves are carved on the
disk to prevent liquid slippage, rotary cup (Figure 3(c2)) with the walls of the cup usually
inclined by 5–6o for the purpose of handling viscous feeds, and finally, previously-mentioned
twin-fluid rotary nozzles.

Droplet formation in rotary nozzles typically takes place around the rim of the rotating disk,
in one of three modes: (1) direct droplet formation with droplets peeled off directly from disk
edge, (2) ligament disintegration, in which droplets are formed by breaking unstable
ligaments, and (3) sheet disintegration, in which a free liquid sheet is formed beyond the disk
rim, then due to sheet instability, ligaments are formed, and finally the ligaments break up
into droplets. SMDs for each type of rotary bell with specified droplet formation modes are
listed in Table 2.

d. Ultrasonic

The principle of the ultrasonic nozzles is based on the usage of high-frequency sound waves
through piezoelectric materials to atomize the feed and produce very narrow droplet size
distributions with low velocities. The exponential growth of surface disturbances results in
formation of droplets whose diameters are proportional to the wavelength of the most rapidly
growing initial disturbance. Produced uniform droplets, with mean size spans from 20 to 100
μm is imperative to precisely program particles on demand; yet in order to change droplet
size, user needs to employ specific nozzle, operating with specified frequencies. Another
drawback of this atomizer is limited produced yield (typically up to 50 ml/min) which limits
these nozzles’ applicability to laboratory units. The major types of generating methods for
ultrasonic nozzles are the transducer and horn, vibrating capillary, and ultrasonic twin fluid
(Omer and Ashgriz, 2011).

In the capillary wave design (Figure 3(d1)), the bulk fluid flow arising from viscous
absorption of the sound energy leads to vibration of the entire liquid body at its natural
resonant frequency. Given vibration of liquid body, capillary waves are excited at the free
surface, and destabilize beyond a threshold amplitude, resulting in break-up to form aerosol
droplets. The Nano-spray dryer B-90, where the vibrating mesh spray technology is used to
generate the aerosol, is based on this design principle (Heng et al., 2011).

In the transducer and horn design (Figure 3(d2)), also known as the standing wave design,
two piezoelectric transducers are placed around a pair of horns. Each transducer, in-sync with
each other, either contracts or expands, at the same frequency and amplitude. This repeated
action results in waves being created. Usually these nozzles contain a wave amplifier right
before the discharge orifice because otherwise ultrasonic atomization would consume a lot of
energy. Figure 3(d2) shows the case where the ejected bulk liquid is exposed to a high-
intensity sound field and atomized in the antinodes of the standing waves.

Ultrasonic twin-fluid nozzles are the last major category of ultrasonic nozzles. Their design is
similar to that of a general two-fluid nozzle, except that the gas is transported through the
conduit into the liquid stream ultrasonically. This creates a fluctuating gas flow, both in terms
of velocity and pressure.

In general, compared to other nozzles, predicting ultrasonic performance is very


straightforward, and compared to the other atomizers, less variation in SMD has been
reported in the literature. A universal correlation was proposed for ultrasonic nozzles based
on dimensionless parameters by Rajan and Pandit (2001):

 fV 
 We 
 
 Oh  
 fAm 2 
 1/3 
SMD / ( )  C1  A(We)C (Oh)C ( I N )C 
2 3 4
where  (5)
f 2
 Am  1 2I
 2 f  c

 fAm 4
 I N  cV
This correlation was further supported by other investigations, but with different exponents
and constants (Avvaru et al., 2006; Barba et al., 2009; Dalmoro et al., 2013; Muñoz et al.,
2018).

e. Emerging nozzles
 Inkjet

The commonly used atomizers mentioned above yield a rather broad size and velocity
distribution of droplets, which could lead to formation of particles with a broad variation in
chemical and physical attributes. For example, droplets with different sizes are dried at
different rates and different drying histories which in turn changes the attributes of the final
dried particles (Poozesh et al., 2017). To remedy this issue, novel inkjet printing as a mono-
dispersing atomizing unit is proposed and employed in pharmaceutical industry for a while
now for certain bio-pharmaceuticals (van Deventer et al., 2013). Inkjet technology is usually
classified as either Continuous Inkjet printing (CIJ) or Drop on Demand (DoD) printing; the
two are distinguished based on the physical process by which the drops are generated (Thabet
and Breitkreutz, 2018). In CIJ, a stream of droplets is generated via forcing the precursor
through an orifice in a thin plate. It is well known that such a free jet is unstable to
perturbations and will disintegrate into a series of droplets based on the Rayleigh-Plateau
instability (Clarke and Rieubland, 2012). Generated droplets are conventionally charged via
an electrode placed near the point of jet breakup and subsequently deflected by an
electrostatic field; so as to return undesirable droplets to a liquid tank. In DoD, however,
droplets with desired size are formed individually, driven by either a piezoelectric or heater
actuator.

Application of inkjet technology in pharmaceuticals is in early phases, yet, researchers found


it capable in revolutionizing spray drying from a probabilistic mechanism, with countless
process control variables, to a more under-controlled one by producing droplets/particles on
command for various applications (Hutchings and Martin, 2012). A comprehensive review of
applications of this technology as atomizer is presented by Daly et al. (2015). Despite
dramatic progress in this area, two main issues still exist: high shear experienced by liquid
during passing through nozzle which denaturalize bio-sensitive compounds; and low
throughput of drug-loaded particles (for a single nozzle, production capacity can be as low as
0.01 mg/min) which is the main hurdle for scale-up. However, there are extensive efforts in
spray drying specialized companies such as Bend Research to assemble hundreds or
thousands of inkjet guns, similar to what is found in printers, to augment throughput rate and
facilitate scale-up.

For inkjet printers, in terms of droplet size we have (Gan et al., 2009; Poozesh et al., 2016b):

DoD: SMD / dl  C1 where 0.25  C1  2.1


(6)
CIJ: SMD / dl  C2 where C2 2.1

 Aerosol-assisted atomizer

Aerosol-assisted atomizers turn precursor solution into micron or submicron droplets with
subsequent augmented drying of droplets that can be achieved in room temperature with no
need for processing gas. Although it has not been demonstrated yet, this atomizer can be
employed for submicron and nanoparticles production of pharmaceuticals. Quite recently,
this method with a novel approach based on disintegration of thin liquid films formed as
bubbles on the liquid surface, was developed and tested by Mezhericher et al. (2018) to
produce particles with about hundred nanometer mean diameter, and throughput rate in the
order of 1 g/min. Since atomization should produce thin liquid filaments and disintegrate
them into droplets, then one might expect viscous effects, surface tension of liquid and
inertial effects to be significant when the liquid filament breaks. This is to say SMD for this
atomizer is primarily a function of We and Re numbers. Although this method seemed to be a
promising approach for spray drying thermolabile APIs, the production capacity is a problem;
and even by augmenting droplet generation through enhancing bubble formation, using
processing gas to dry up wet particles can diminish its primary advantage.

 Electrostatic

In electrostatic atomization, the energy causing the surface to disrupt comes from the mutual
repulsion of charges that have accumulated on the surface. An electrical pressure is created
that tends to expand the surface area. This pressure is opposed by surface tension forces,
which tend to contract or minimize surface area. When the electrical pressure exceeds the
surface tension forces, the surface becomes unstable and droplet formation begins. When the
two forces are equal, the so-called Taylor cone (Figure 3(e1)) is formed at the capillary
interface, and if the electrical force is maintained above the critical value consistent with the
liquid flow rate, then atomization is continuous. Electrostatic nozzles are particularly
advantageous because their droplet sizes are more uniform than any other type of
conventional nozzles, and they offer a greater degree of control on the droplet size. However,
their relatively low liquid flow rates have prevented these nozzles from widespread use. Not
enough literature exists to predict the SMD of this family of nozzles. For an ultra-fine liquid
atomization procedure combining the advantages of electrospray, and considering the spray
to be roughly monodisperse, Gañán-Calvo et al. (2006) modeled the droplet diameter
(confirmed by experiments) as follows:

 
 d 0  

SMD / d 0  C1 (V / V0 )1/2  C2  Where  (7)
V  (  )1/2
4

 0  3

Here, Δψ is the total available potential energy per unit volume of the liquid.

3.1.2 Further remarks


Atomization attributes are critical in spray dryer scale-up as they have direct influence on the
particle size, morphology, etc., which in turn impact flowability, compressibility and
dissolution performance (Singh and Van den Mooter, 2016). There is a direct relationship
between droplet size distribution to particle size distribution of dried product, which can be
established by simple mass conservation of the feed solid fraction (Kemp et al., 2016b).
Although in the atomization section, we only discussed SMD for various types of atomizers;
there are other atomizers’ characteristics that bear importance; amongst others, velocity
distribution and the width of droplet size distribution are pivotal to fulfill dynamic similarity
for flying particles across scales. Just to mention a few of their impacts on product quality,
drop size distribution is paramount in determining the span of final particle size distribution
and drop velocity distribution gives insightful information on particle residence time, and
subsequent moisture content. Despite their importance, there is limited literature on these
aspects. Just to give a few examples, Poozesh et al. (2018c) related distribution width (span)
of a two-fluid nozzle to non-dimensional factors across different formulations and design
configurations. For the same nozzle, same research group presented a technique based on
Particle Imaging Velocimetry (PIV) to shed light on axial and radial velocity distribution,
spray angle, etc. as functions of atomization driving forces (Poozesh et al., 2018a). After all,
in order to fulfill similarity between two scales, first to a minimum, we propose to have
similar SMD and span across scales. Also, ideally residence time of particles, across scales
must be similar to avoid underheating and overheating. A very rough estimation of residence
time can be obtained by dividing volume of dryer to feed flow rate. A more precise one can
be calculated by dividing mean velocity of atomized droplet to spray’s height. Therefore,
non-dimensional ratio of droplet mean velocity to dryer height/residence time, can be used as
a PI number to assure similar residence time of particles in chamber. There is no report,
looking at this factor in spray drying scale-up process; and more is needed to be done is this
area.

3.2 Drying and particle formation

Despite multi-faceted nature of spray drying process, there are fundamental modeling
approaches for process characterization in the literature that can help users to estimate outlet
conditions of dryer and physical and chemical properties of resultant particles and subsequent
powder. Here, mathematical, numerical, and experimental approaches are reviewed to give
insight into spray drying output in terms of both powder quality and gas outlet conditions
across scales. Herein, the goal is to reach similar drying conditions in terms of key response
variables, i.e., powder residual solvent and outlet temperature, when going from one scale to
another. To attain this goal, pharmaceutical engineers can use various modeling tools, which
vary in implementation difficulty and computational resources, and offer different predictive
capabilities and accuracies as well. Macroscopically, the most important drying outputs are
powder solvent content and outlet temperature. Yet microscopically, particle drying history,
such as particle solvent content, residence time, density and morphology, and diffusion of
species are of high significance. Here, we begin with simple models that can be used to
macroscopically predict outlet conditions, and then we overview CFD tools to monitor
individual particles’ drying history.

3.2.1 Spray dryer’s outlet conditions


A thermodynamic model of the spray drying permits the characterization of the process in
terms of drying gas flow; relative humidity at the outlet of the drying chamber; inlet, outlet,
and condenser temperatures (Ti, To, Tcond); and feed flow rates. Various simplified or
mathematical models that usually entail numerical analysis can be used to define the process
operation range over which, desired outlet conditions are attainable (Dobry et al., 2009). The
determination of such variables plays a role in the development of the spray drying process in
several distinct ways. To is one of the most important process parameters since it has an
impact on particle and powder properties such as density, surface area, mechanical strength,
and physical stability (Maas et al., 2011). Relative humidity is the main driving force in
drying process and, therefore, has direct impact on the level of residual solvents in final
powder. The latter is of particular importance when producing amorphous materials since the
level of residual solvent in the solids strongly affects its glass transition temperature (Hsieh et
al., 2015).

a) Simple model

The simplified model is based on overall (macroscopic) mass and enthalpy balances for the
whole spray dryer control volume. One can consider the dryer as a “black box” and find
rough estimations on the gas outlet temperature and humidity, together with the final
moisture content of the powders, while assuming outlet gas and the outlet particles being
virtually in thermodynamic equilibrium with each other. This means that the gas and powder
outlet temperatures are almost equal. Furthermore, the outlet powder solvent content is also
assumed to be in equilibrium with the gas humidity. Worthy of note, the equilibrium
assumption is only of any use where equilibrium is attained. Intuitively, this is more likely to
be valid for small droplets with long residence times, which mostly happens in pilot or
commercial spray dryers. Ozmen and Langrish (2003) provide experimental evidence that
this is indeed the case for dried-particles with a final size of less than 30 µm.

Figure 4 shows the schematic of the problem with appropriate assumptions and mass
conservation for each species. As a result of the mass balance equation one can first obtain an
expression for the outlet solvent content of gas, Yog, and another for the gas relative humidity,
RH, at the outlet as following:

(1  x)mso ln. pv M sol . p Yo M g


Yog   and RH  v  ( p / psat ) (8)
mg  m p ( p  pv )M g psat M sol .  Yo M g

where psat can be found from Antoine equation at To for specified solvent (Rodgers and Hill,
1978).

Now, we take a closer look at the energy equation. Before that, in order to calculate the heat
loss from the dryer, assuming given thermal efficiency of the spray dryer, ηth, defined as
evaporation capacity to the total heat provided by processing gas, one can determine heat-loss
based on the following:

1
Qloss  (1  th )(msol . ) th2 (mg  mso ln. )c p ,so ln. (Ti  Tam )  (9)
th
Finally, by equalizing the total input enthalpy, stemming from processing gas, and enthalpy
due to phase change, heat loss, and elevated gas and particles temperatures, one can estimate
To as following:

1
To  Ti  Qloss  msol . 
(mg  msoln. )c p 
(10)

Theses equations can be arranged to solve for spray-dryer outlet conditions as a function of
given information based on solution, processing gas, and drying chamber data.

Essentially, this simplified modelling consists of a set of equations that relate process
parameters through mass and enthalpy balances and liquid–vapor equilibrium equations. If
some of the marginal assumptions would be relaxed, e.g. variant thermal efficiency instead of
a constant one, more complexity is added to the final equations. Due to resultant implicit
equations, usually numerical tools are used to solve equation system. This simplified
thermodynamic space is mostly appropriate when dealing with well-mixed gas flow patterns
where differences between the two phases are ignored. This is especially the case in lab-scale
spray dryers, which can be regarded as a well-mixed gas flow pattern due to its small aspect
ratio. Also, considering the way in which simple mass and energy balances might be applied
to spray dryers, this model is a simple but useful first step, and could be regarded as a large-
scale approach to modeling the equipment behavior. Due to shortcomings of this modelling
technique, more detailed models which factor in individual droplet drying are succeeded.

b) 1D models

Before proceeding with these models, first, main drying kinetic models for particle formation
process shall be briefly reviewed. For a more comprehensive overview, interested readers are
referred to (Mezhericher et al., 2010b). Theoretical models and experimentation have been
used to determine droplet drying kinetics. These models include: (1) the characteristic drying
curve (CDC) model based on semi-empirical drying curves; (2) models based on the reaction
engineering approach (REA); (3) transport-based drying models that incorporate continuity,
momentum, energy, and species conservation equations; and (4) deterministic models
containing distributions of the solid component described by a population balance. Table 3
summarizes these models and describes some major studies in which they were used.

To verify particle drying kinetic models, and also gain insight into chemical and physical
changes during transformation of single droplets into solid particles, empirical data, based on
experimentation on drying kinetics have been extensively reported (Lewandowski et al.,
2018; Pajander et al., 2015; Sadek et al., 2016; Sadek et al., 2015; Tsotsas and Mujumdar,
2014). Typically, this experimentation has studied free fall or sessile droplets. During free
fall, a single droplet or a series of near mono-dispersed droplets fall freely under the influence
of gravity and gas drag forces (Amelia et al., 2011; Rogers et al., 2012; Sloth et al., 2009;
Vehring et al., 2007; Wu et al., 2011). Some studies used ultrasonic, inkjet technology or
JetCutter capabilities (Sloth et al., 2009) to produce the droplets. Because conditions inside
the free fall drying chamber are precisely selected and all droplets should have nearly
identical sizes, the sizes and properties of the final dried particles are anticipated to be very
uniform. In general, these experimental data on drying of single droplets are extremely useful
to identify the range of favorable droplet size range, based on finial chemical and physical
attributes of produced particles (Sadek et al., 2015). However, the droplets produced via these
approaches are usually too large and/or the air velocities are too small to be representative of
real conditions in a spray dryer.

Although simple overall heat and mass balance-based model has been reported to
successfully predict gas outlet conditions (within 10%), one disadvantage that has been
brought up in the literature, (Ali et al., 2014; Sormoli and Langrish, 2016), is its inability to
capture final powder solvent content. Therefore, more detailed one-dimensional (1D)
transport-based models, discarding outlet equilibrium moisture content assumption have been
proposed. These models are usually 1D simulation approaches, where mass, heat and
momentum balances are executed at an individual droplet level as a function of drying time
or dryer height. Any of the described drying kinetics models (discussed in Table 3) can be
incorporated in this approach to allow predicting drying-rate profiles and identifying fast and
slow drying-rate periods. This simulation approach is widely used to model product
characteristics during operation of spray dryer. A major advantage of the 1D models is the
capability to evaluate the “average” behavior (temperature, solvent concentration and velocity
profiles) of the powder and hot gas during drying and along dryer’s height in a short period of
time with fairly simple calculations. These profiles can assist in predicting various thermo-
physical properties and quality parameters of the product throughout drying. A relatively
simple Excel spreadsheet is usually enough to build effective 1D spray drying simulation.

Another advantage of these 1D models is their simplicity for use in process control and
optimization in industry, as compared with multidimensional transport-based models (will be
discussed later). Specially, when it comes to production scales with large dimensions, having
a comprehensive 3D modeling would be very time consuming and clumsy. Unlike
multidimensional transport-based models that entail solving partial differential equations, the
1D models entail solving ordinary differential equations, which substantially decreases
complexity and computing time. Yet, the disadvantage of this approach is its inability to
capture the complex interactions between droplets/particles and drying gas due to
coalescence, agglomeration, deposition on the wall and re-entrainment of deposited particles
into the gas.

Among 1D models themselves, we can refer to two main categories of approaches: in the first
category (plug-flow assumption), a particle drying kinetic model is integrated with a spray
drying tower simulation based on mass, energy and momentum balances for either mono-
sized particles, or droplet size distribution, uniformly dispersed over the cross-section of
dryer column with no interactions between particles/droplets. The droplet/particle and air
velocities are 1D and parallel to the axis of the dryer. The temperature and humidity of air
could vary axially. The drying model accounts for shrinking of the droplet due to evaporation
of moisture. These models generally consider the overall process to be running at steady
state, although the drying kinetics of the particles are clearly unsteady state. This model is
mostly suitable for tall-form dryers (Figure 5(a)). Both CDC and REA drying kinetic models
have been reported to be integrated with this model (Ali et al., 2014; Montazer-Rahmati and
Ghafele-Bashi, 2007; Patel and Chen, 2005). The second category allows variation of the
droplets/particles parameters along the height and relaxing the assumption of plug-flow for
the entire system (Omer and Ashgriz, 2011). These models are called quasi one-dimensional
models. A numerical model is developed for the drying tower by dividing into two zones,
namely the spray zone and the plug-flow zone. The spray zone considers variations in the
cross-section of the droplets/particle’s parameters, but the drying gas properties remain
constant over the cross-section. In the plug-flow zone, the parameters of the
droplets/particles, as well as the drying gas vary only in the axial direction. This quasi-1D
model is more accurate than the 1D model with plug-flow assumption, especially for lab-
scale spray dryers. Representative research works that employed these models along with
other successive ones are given in Table 4.

3.2.2 Multidimensional transport-based models


Although preceding simplified & 1D models are suitable for rough estimation of design and
operation conditions for an upscaled spray dryer, they cannot predict all experimental
observations (see e.g. results of Hanus and Langrish (2007)). This is, in part, due to the
impact of atomization conditions on the wall deposition behavior. Additionally, although
compared to simple models, 1D model can give more accurate estimation for powder solvent
content, reasonably precise estimations could be only achieved by considering the kinetic
effects (drying rates) that are accounted for in the finer 2D & 3D models. Besides, in 1D
models, turbulence arising from non-uniform, swirling processing gas, and even atomizing
gas, is ignored, which can heavily influence prediction precision of the solvent content in the
final powder. Specially with lab-scale dryers, the complexity of the flow patterns is greater
than that in tall-form dryers, in the sense that many short-form dryers have no plug-flow
zone; and there exists a wide range of gas residence times in these dryers. Hence, 2D and 3D
transport-based models, consisting of a set of highly complex non-linear PDEs, need to be
solved via CFD (Computational Fluid Dynamics). CFD has particularly been applied to spray
dryers, in part due to the complexity of the gas and particle flow patterns and to the many
parameters involved in spray drying process (Masters, 1994).

For the modelling of particle transport and particle interactions, most CFD tools use Euler-
Lagrangian approach, where the spray is represented by numerous discrete droplet parcels in
a continuum of gas, with each parcel comprised of a group of dilute physical droplets. Then,
solutions for the airflow patterns are found by calculating approximate solutions of the
Navier-Stokes and continuity equations on a grid of control volumes. The particle phase is
followed by tracking a number of individual particles through the airflow domain with the
exchange or transfer of mass, energy and momentum within the continuous phase calculated
along particle trajectories (Poozesh et al., 2018b). The grid results are recreated in the form of
graphical visualizations, which is the most common form of analyzing CFD results. Any of
the discussed particle formation drying models can be integrated into simulation to account
for mass and heat transfer over the course of particle drying.

Despite fascinating advancement in CFD capabilities, systematic errors or idealistic


assumptions may render completely unrealistic results. Critical attention should be given to
grid generation, turbulence models, steady steadiness, particle drying kinetic models, and
convergence of solution. For instance, one of the observed problems is fluctuating residuals
when solving for balance equations, while assuming a steady state flow. In one hand, this
means spray drying process at least for the lab-scale dryer is unsteady and should be treated
as such. On the other hand, transient flow will dramatically increase computing time.
Continuous improvements in high performance parallel computing is likely to alleviate
computing time issue. Additionally, as far as convergence is concerned, uncertainty analysis
which has been proposed and implemented to continuously decrease relative error between
measured data and CFD results might be a tangible solution (Poozesh et al., 2018b).
Commercially available software packages such as Fluent, CFX, both in ANSYS software
package are most commonly used for CFD modeling. For example, Mezhericher et al. (2012)
assessed a two-stage drying approach using a 3D model of a pilot spray dryer in ANSYS-
Fluent. Compared to their prior 2D simulation (Mezhericher et al., 2010a), a significant
influence of both drying kinetics and the number of utilized dimensions on the predicted
particle trajectories and transport phenomena in the spray-drying process was observed. Table
4 lists all the discussed modeling tools, 1D, 2D, and 3D simulations along with example
references. Reporting similar works are avoided and only paradigm studies are given in this
table. Also, Figure 5 illustrates examples of configuration/simulation results for all the
discussed models.

3.2.3 Chemical homogeneity and amorphous state


Due to the rapid evaporation of the micron size sprayed droplets (usually most of solvent is
vaporized near atomizer), and in turn, minimal time for crystalline molecular arrangements, it
is most likely to see produced powder in amorphous state after drying. Anyhow, formation of
crystallinity, however, may occur to different degrees depending on the formulation and
process conditions. Chemical homogeneity and stability of system can be a matter of concern
throughout the whole process, beginning from solution preparation to secondary drying stage.
Usually at lab-scale, the interval time is not long- in the order of minutes, while at larger
scales, product may be held in solution for many hours or days, and therefore, the kinetics of
degradation at different feed temperatures should be known. In some cases, the feed
temperature needs to be reduced to prevent impurity growth.

Over the course of drying, for a ternary amorphous solid dispersion comprising of solvent,
drug and polymer, subtle concentration gradient may emerge and expand based either on
molecular weight, solubility differences between the drug and polymer, and solvent drying
rate. Here our chemical heterogeneity analysis is based on two non-dimensional parameters,
Peclet number (Pe), i.e., the ratio of drying rate (K) to mass diffusivity (D), defined in the Eq.
(11), and the ratio of initial concentration to solubility (ICS) in the solvent. Both numbers are
defined separately for polymer and drug.

K
Pe  (11)
8D
We discuss two case scenarios based on solubility differences. In the first case, there is a
relatively high ICS for both polymer and drug in the spray solution, at high Pe (happens for
smaller droplet). In this case, the differential diffusion rates of polymer and drug may lead to
a polymer skin formation at the surface of the drying droplet to produce a particle with a
relatively polymer-enriched surface, leading to desirable high Tglass formulation. In the
second case, at similar ICS and low Pe (happens for larger droplet), both drug and polymer
molecules may have enough time to diffuse uniformly toward the core. If polymer and drug
have tangible difference in ICS in the spray solution, then the lower soluble component
would be expected to precipitate initially and concentrate on the surface due to the slow
diffusion coefficient of the larger precipitated material (Boraey and Vehring, 2014; Poozesh
et al., 2017).

Although using non-dimensional parameters such as Pe and ICS might be implemented to


optimize droplet size (which ties to nozzle design/process) and feed formulation, it can often
require significant understanding or experimental data for a particular system to be applied
effectively (e.g. measurement of diffusion coefficients). Furthermore, usually, Pe and ICS are
not independent, as molecules with higher molecular weight, which in turn have higher mass
diffusivity and small Pe, possess lower solubility (Knopp et al., 2015). Additional
considerations should be given to this area to further shed light on non-dimensional
parameters affecting amorphous phase behavior during drying.

As mentioned earlier, in addition to final distribution of components in the dry particle (i.e.,
particle homogeneity); particle residual solvent is important in determining final physical
stability. If not enough solvent has been driven out during drying process to result in a solid
particle with a Tglass close to or greater than that of To, the process will yield a mobile glass
that risks phase separation (Singh and Van den Mooter, 2016).

There are few references on crystallization during particle formation, which is also referred as
“in-situ crystallization” in spray dryer (Woo et al., 2017). Many different parameters have
been suggested by Woo et al. (2017), in their recent review, to impact in-situ crystallization
such as Ti, To, initial solid concentration, polymorphism of material, solvent type,
incorporating excipients, alternative flow mode (e.g. counter flow), etc. In the review, case by
case report on the literature, given remarkable discrepancies occurred when using different
scales (even with the same formulation), except some of the basics that were discussed
earlier, no conclusive remark can be drawn that could be applied across the scales.
A potential remedy for this shortcoming might be provided by using CFD simulations to
model the spray drying process, while integrating in-situ crystallization. This approach will
provide detailed information on the drying, shrinkage and crystallization behavior of the
droplets throughout their entire flight within the spray drying chamber. The simulation
strategy to address crystallization during drying thus far, have been similar to the earlier ones
in previous sections, except adding degree of crystallization. Two studies report on this area
are given by Woo et al. (2012) and a very recent one by Har et al. (2018). In the former, the
relative crystallization rate was related to the difference between droplet temperature and
Tglass of the dried/semi-dried droplet. In the latter, REA was incorporated both for
crystallization and particle formation. In this case, crystallization was viewed as a chemical
reaction, where activation energy is related to drying temperature, solvent content and
material characteristics.

The challenge in using CFD, however, is in the development and validation of droplet
crystallization sub-models. It is envisaged that more fundamental work on droplet
crystallization kinetics is required to advance this area of CFD on in-situ crystallization
control.

3.2.4 Physical attributes


Chemical heterogeneity and subsequent amorphous state of spray dried products must be
well-controlled, and design space must be conservatively chosen to avoid failure. On the
other hand, particle physical characteristics, including particle size, density and morphology
may also be critical for some applications. For instance, in pulmonary administration, highly
porous particles of sizes less than 5µm are desired (Champion et al., 2007).

Particle size and density can be predicted via mass conservation equation based on initial
droplet size distribution (Boraey and Vehring, 2014). Other properties such as surface area
and roughness or pore volume distribution are all known to affect the performance of spray-
dried powders. Therefore, it is common during the development of a spray-dried product to
vary the process parameters to produce powders with distinct characteristics. It is
recommended that this step of powder optimization would be conducted at pilot scale, where
a good compromise is obtained in view of the range of particle/powders that can be produced
and the ability to target these ranges throughout the remaining of the scale-up process (Singh
and Van den Mooter, 2016).
Particle properties are related to the drying kinetics of the droplets inside the drying chamber
and are primarily dependent on the mechanical and chemical properties of the spray-dried
material. These interactions are very complex to model, and therefore, process development
is typically governed by some general trends, as described in Figure 6. The parameters that
most influence particle morphology are RH and ICS (Littringer et al., 2013). RH can be

modulated through different means, such as changing Tin, solvent type, and msol . / mg . It is

worth noting that by adjusting RH, one can control the morphology in such a way that size
and morphology become almost independent of each other (Nandiyanto and Okuyama,
2011).

3.2.5 Design space


Employing one or more of the discussed modeling tools, including simplified, 1,2 or 3D
simulations, formulator/engineer can calculate at least the relative humidity of the drying gas
and temperature at the outlet. Proposed models can be further verified with a small set of
experiments at lab-scale. Then the data can be mapped on a graph such as the one shown in
Figure 7, to determine failure modes and subsequent borderlines. Figure 7 is based on three
non-dimensional parameters, Ti / To , msol. / mg and RH. There are four main constraints for each

of these parameters which stem from allowable conditions as following:

 Due to thermolability of APIs, there should be a limit in maximum allowable Ti.


 There is minimum amount of solution that should meet minimal yield.
 Maximum RH should be respected to avoid failure mode and subsequent instabilities.
 Gas outlet temperature, To, should be lower than Tglass (recommended Tglass -To≥10oC)
to avoid physical instability and product stickiness.

Maximum RH can be obtained by establishing a relationship between ASD glass transition


temperature via dynamic vapor sorption (DVS), coupled with common thermal analysis
methods (such as Differential Scanning Calorimetry) used to find Tglass (Carter and Schmidt,
2012). First, the minimum Tglass is determined based on API, and then according to
established relationships between Tglass and RH, one can find maximum value for RH. Having
these restrictions, a confined area is formed in which thermodynamic space should be based
on; irrespective of the scale. Based on discussed parameters involved in physical attributes of
particles, further restrictions can be applied to achieve desirable morphologies.
3.2.6 Further remarks
Although the preceding design space provides a bridge between processes at different scales,
aside from the imposed restrictions, there are other constraints normally due to equipment
limitations (e.g., maximum Ti or minimum condenser temperature), product requirements
(e.g., To limited by the product degradation profile), and other process constraints, depending
on main and supplementary equipment, that may interfere in the right design space.

Presented design space is a rough estimation which except Tglass does not entail delicacies
associated with API. Proper design space also encompasses the target drug profile where the
attributes of the drug are to be kept within appropriate limits in order to ensure the desired
product quality.

3.3 Gas and particle separation

Usually segregation step in spray drying is not as much explored as the other previous
operation steps; nevertheless, it holds significant importance in determining yield and
pressure drop at the end of the drying process. The throughput is especially important in
pharmaceutical industry as APIs are usually expensive and serious consideration should be
given to conserve as much API as possible during operation. This is especially important in
lab-scale spray dryer where the yield may fall below 50% in worst case scenarios. This
change in yield also alters finial particle size distribution unless powder is collected from
filter bags and throughout cyclone. Traditionally, drying performance was considered
independent from cyclone, yet recent studies have shown meaningful correlation between
cyclone design/selection to bio-active retention, which turned out to be due to differences in
ability to capture different size particles (Bögelein and Lee, 2010; Heng et al., 2011). In the
following section, we discuss two of the most common cyclone types, namely vortex-based
(also known as tangential inlet reverse-flow cyclone) and electrostatic particle separator (see
Figure 8).

3.3.1 Vortex-based cyclone


Upon created vortices in the cyclone column, particles with different size/densities
experience different forces which in turn cause their stratification. Bigger particles experience
higher centrifugal forces; thereby get closer to wall where effect of centrifugal forces
becomes small (due to boundary layer) compared to gravity, and consequently move
downward and get trapped into the collector. However, fine particles which move almost in
the middle of column, are in low pressure region (even less than atmosphere pressure);
herewith are often drifted in upward spiral flow due to the slower speed and escape through
the gas outlet. Distribution of particles across cross section of the cyclone is such that the
most heavily populated regions move from the wall toward the center with decreasing
particle size.

The geometrical configuration is probably the most crucial aspect affecting the performance
of a cyclone separator as the cyclone performance is sensitive to the smallest geometrical
changes (Cortes and Gil, 2007). For convenience, the dimensions of various cyclone parts are
usually stated in dimensionless form by normalizing them with the cyclone diameter. This
method allows a comparison between the cyclone designs, without using the actual size of
each individual part. Through meticulous experimentation, optimal designs of cyclone
separators have been proposed (Bögelein and Lee, 2010). One of the most commonly used
designs is the Stairmand high efficiency cyclone, which even after almost 60 years is still
popular. This cyclone happens to retain the velocity profile at every axial point well and
achieves very high collection efficiency. The key for Stairmand high efficiency cyclone
design lies in the optimal geometrical ratio of cone length to diameter (length to diameter of
4). Unfortunately, it is impossible to always accommodate Stairmand cyclone in the industrial
processes, therefore an optimal design, best suited for application at hand, has to be explored.

The main performance metrics of a vortex-based cyclone separator are collection efficiency
and pressure losses. Although the operation of a cyclone seems easy, no generalized model of
the flow pattern is available that allows the prediction of velocity components in the cyclone
(El-Batsh, 2013). Pressure loss mechanism and particle separation in a cyclone can be
examined by determining the cyclone gas flow pattern via CFD for versatile
configurations/processes/designs. Because of the inherent anisotropic turbulence, unless
axisymmetric design is employed, 3D modeling must be used. As a result of both
experiments and simulations, it is suggested to characterize the flow through cyclone
separators by Re number, based on cyclone diameter and inlet velocity; and the total pressure
loss coefficient (represented by Euler number (Eu)) to characterize the pressure loss
mechanism through the cyclone (Utikar et al., 2010).

Cyclone separation efficiency depends on different physical and operational parameters,


which can be arranged into two groups. The first group contains the parameters related to the
gas flow inside the cyclone. These parameters include gas properties such as density and
viscosity, inlet velocity and cyclone diameter. This group is represented by Re number. The
second group includes particle specifications and includes particle size and density and is
represented by Stokes number (St). Therefore, for the standard cyclone, collection efficiency,
η, and pressure drop can be related to the dimensionless parameters Re, St and previously
mentioned Eu number. These numbers are defined as:

  g vi di
 Re 
Pt  
Eu  ,   f (Re, St ) where  (12)
0.5 g vi2  St  (  p   g )d p vi
2

 18 g d

where ΔPt is the mean total pressure difference across the inlet and outlet. Also, vi is inlet
velocity, d is cyclone diameter, dp is the mean size of solid particles and di is the inlet
diameter. An increase in St, up to a certain value (0.01 in Brar and Elsayed (2018)) results in
efficiency improvement, the implication of which is collapse in collection efficiency as
cyclone diameter increases. On the other hand, relationship of Re and cyclone efficiency is
not that straightforward. At constant cyclone body diameter, increase in Re to a certain point
ramps up efficiency; but further increase in Re beyond that reversely impacts efficiency
(Utikar et al., 2010). In addition to the discussed non-dimensional parameters, in relation to
the three discussed parameters, other variables that have been taken into account in the
literature are reviewed below (Cortes and Gil, 2007; El-Batsh, 2013):

 A larger exit pipe diameter (normalized by cyclone diameter) causes lower pressure-
drop through the cyclone. The collection efficiency was found to behave differently.
Increasing the exit pipe diameter increases the collection efficiency and vice versa.
 Increase in inlet particle concentration leads to increased collection efficiency and
smaller pressure-drop coefficient.
 On increasing the inlet gas velocity for a given solid loading, both pressure-drop
coefficient and the collection efficiency increase.
 Decrease in the cone tip diameter (ratioed with cyclone diameter) causes considerable
increase in collection efficiency and pressure loss coefficient.
 Increase in viscosity through either using different types of processing gas or increase
in To results in unfavorable lower collection efficiencies and higher pressure-drop.
 Increase in mean particle size will increase collection efficiency and also raises
pressure-drop.
 Increasing the cylinder length (ratioed with cyclone diameter) showed remarkable
reduction in pressure-drop with slight increase in collection efficiency.
 Increasing the cone length (ratioed with cyclone diameter) resulted in better collection
efficiency, but higher pressure-drop.

3.3.2 Electrostatic particle separator


This type of separator is mostly found in lab-scale spray dryers, where low production rate
(usually when using ultrasonic or mono-dispenser nozzles) hinders sufficient yield, if other
cyclone types are used. Compared to vortex-based separators, this separator operates with
high collection efficiency (more than 85%) with almost no pressure loss. This type of cyclone
is usually used in coal cleaning, mineral processing, solid waste recycling, (Wang et al.,
2014), and recently in dry separation of pharmaceutics and food ingredients (Wang et al.,
2015).

In this separator, micron or submicron particles, which are usually falling in a laminar flow
are separated due to the electric field generated by a high voltage between a grounded and a
collecting electrode. The behavior of the particles depends on their electrical and geometrical
properties. Therefore, separation of different species is possible if the parameters of the
electric field, the separation unit, and the ambient conditions are set up in the right way.
Collection efficiency depends on particle size (ratioed by the distance between electrodes),
charge per particle surface (also can be considered as charge surface resistance (Draheim et
al., 2015)) and velocity of processing gas. Residual solvent may also effect charge per surface
of particle and increases surface electrical resistivity (Draheim et al., 2015). Although these
types of collectors have not been used in pilot or commercial spray dryers, depending on
future drying technologies, may find avenue to appear in upscaling process. Unfortunately,
this collector’s fundamental operation has not been explored yet in details, and more future
work is needed to identify principal parameters affecting efficiency, while improving
throughput, required for scale-up.

3.3.3 Further remarks


We only discussed the major attributes of cyclone including yield and pressure-drop
coefficient, but cyclone has been proven to impact biodegradation as well (Bögelein and Lee,
2010). In contrary to intuition, this is not because of thermodynamically induced effects of
cyclone on particles, but it stems from the capability of cyclone to capture different particles
with different bioactivities. In this sense, more effort is needed to identify the role of particle
size on chemical homogeneity and in turn physical stability. From cyclone side, in addition to
pressure loss and collection efficiency, more work is needed to highlight cyclone capacity to
capture certain particle sizes. This helps in identifying failure modes when transferring from
lab-scale spray driers where finer particles are produced, to higher scales, where both particle
size and yield are different.

3.4 Roadmap for spray-drying development and scale-up based on fundamental


engineering considerations

One objective of spray-drying development is to determine a suitable SMD range for scale-up
(see Figure 9). The SMD at smaller scale should be determined based on powder quality,
employing experimental tests. This could be done by following three approaches: first, drying
experiments are to be used on single droplets in a controlled environment operating based on
input variables at lab-scale spray dryer conditions, to identify a suitable range of droplet size
whose functions, in term of physical attributes and amorphous state are favorable. Relevant
rigs and measurement devices are discussed in section 3.2.1. Meanwhile, discussion around
Pe and ICS in sections 3.2.3 and 3.2.4 can be useful for manipulating formulation/process
conditions to achieve desired chemical and physical attributes, while seeking for the best
range of droplet size. As one alternative to this approach, where single droplet tests are not
available, on smaller scale, atomizer simulator experiments are to be carried out to find the
link between process/design/formulation conditions and SMD. In these experiments, atomizer
is taken out of the drying chamber, and atomization performance is separately investigated
under similar conditions (Bilgili et al., 2011; Poozesh et al., 2018c). Finally, a third method
for identifying the right SMD would be based on back-calculating droplet size from spray
dried particles on small-scale. It is worth mentioning that due to the dependency of produced
particles’ physical attributes on process conditions/formulation, this method is prone to
inaccuracies in predicting SMD. Kemp et al. (2016b) showed that, given different produced
particles’ shapes induced by initial solid concentration to solubility, denoted by ICS earlier,
the deviation between back-calculated and measured droplet size distribution could be
significant, which limits this method’s applicability. After identifying a proper range for
SMD, by applying similar conditions in small-scale spray dryer, and after powder quality
assessments, a link between SMD and product quality can then be made to further identify
the right range for SMD.
To have similar SMD in upscaled spray dryer, depending on atomizer type, discussed
correlations in section 3.1.1 are to be used to set for atomizer operation conditions. Moreover,
as mentioned above, the atomizers can be used in spray simulators to determine the droplet
size distribution (Bilgili et al., 2011; Poozesh et al., 2018c) at the scaled-up, high spray rates.
The atomization air pressure or air flow rate can be adjusted to keep SMD similar across
scales empirically (Bilgili et al., 2011).

Finding thermodynamic design space, discussed in section 3.2, is another objective of spray-
drying development. This can be accomplished by referring to the thermodynamic methods in
section 3.2 through simple, 1D and multi-dimensional models. To find the constraints, as
mentioned in section 3.3, maximum RH and To can be obtained by incorporating established
relations between these variables and Tglass. Noteworthy that the thermodynamic models must
be verified by limited experimental data on lab-scale spray dryer. Based on the SMD and feed
flow rate at large-scale, one can modulate Ti (preferably) and processing gas flowrate to
operate in established "design space".

The process scale-up requires engineers to identify key response variables for each
transformation and use correlations with non-dimensional parameters and possibly some
fundamental dimensional variables to ensure similarity of drying, atomization, and gas–solid
separation across scale. One task is to ensure “similar drying” across various scales. A
standard method of scale-up in industry (by manufacturers and users) is to scale the spray rate
proportionately to the maximum processing gas flow rate (Kemp, 2017). The aim of this
empirical approach is to maintain To between scales upon a modest adjustment of Ti for
different heat losses and atomization gas flow rate (cold) in different equipment, which can
be back-fitted during the plant trials. This approach also allows one to keep the exhaust
humidity and RH similar across scales as well. This method appears to be generally effective
as a rough, first approach for drying. Using the modeling tools of varying complexity,
accuracy, and dimensionality including the thermodynamic design space approach mentioned
above, engineers can fine-tune the empirical scale-up approach while minimizing
experimental effort and better predict product properties because the modeling tools can yield
fundamental information about the formation, thermal history, and structure of particles,
residence time distribution, etc.

Finally, in order to maintain key response variables across separators, i.e., pressure-loss
coefficient, Eu, and collection efficiency through Re and St numbers, adjustments should be
made for outlet and inlet pressure difference, inlet velocity, and inlet diameter. To make these
adjustments, based on small-scale cyclone, and associated non-dimensional variables,
geometrical aspects along with inlet velocity (in view of both magnitude and turbulent kinetic
energy) of the separator at large-scale, especially the inlet and outlet dimeters, are to be
modulated.

4 Conclusion and Final Remarks

In the current work, first, the attempts that have been made to scale up of spray drying,
especially for pharmaceutical products, have been reviewed and discussed. Despite the
importance of this matter, few works have been concerned to accomplish meaningful
transition from lab scale to larger scales. Discrepancies in choosing process conditions and
design configuration corroborate the fact that this area needs a more mechanistic, predictive,
scale-up approach using scale-up rules which encompass modeling/mathematical tools, in
conjunction with limited empirical measurements. This review paper presents a thorough
analysis of literature on each transformation taking place during spray drying: atomization,
drying and gas–solid separation. Then, a roadmap has been presented whose first step is to
identify a suitable range of SMD and thermodynamic design space for preparing a spray-
dried product with acceptable performance using mostly small/pilot scale data. Then, the
roadmap describes how to scale-up a process. Similar drying conditions (To, RH) across
scales can be achieved by maintaining the ratio of spray rate to drying gas flowrate while
adjusting Ti, whereas similar SMD of droplets can be achieved by adjusting the atomization
air flow rate of the same atomizer at the scaled-up spray rate in a spray simulator and using
correlations established in the literature. Finally, cyclone separation types with critical
parameters, identifying their operation and performance were discussed. Pressure-drop
coefficient and collection efficiency have been realized to be the key response variables.
Besides the aforementioned uses of empirical approaches, this review paper advocates the use
of modeling tools for process scale-up. Engineers can fine-tune the empirical scale-up
approach while minimizing experimental effort and better predict product properties because
the modeling tools can yield fundamental information about the formation, thermal history,
and structure of particles, residence time distribution, etc. Other reasons could be based on
changes in atomizer type, required final particle size and residence time in the dryer
(including the downstream pipework and cyclone). The presented roadmap in this paper can
help formulators and engineers to establish a proper SMD range and thermodynamic design
space and then scale-up the process. Future scale-up studies are warranted toward
establishing a more mechanistic and predictive approach with advanced modeling tools for a
robust pharmaceutical spray drying scale-up and assessing the cost/benefit and practicality of
such tools.

Nomenclatures
Normal Greek symbols
A Surface area (m2)  Efficiency
C Constants  Density (kg/m3)
cp Specific heat (J/kg.K)  Latent heat of vaporization (J/kg)
D Diffusion coefficient (m2/s)  Difference in states
d Diameter (m) Ʌ radial spatial integral scale of
turbulence (m)
Eu Pressure drop coefficient Subscripts
H Enthalpy (J) d Droplet
h Heat transfer coefficient (W/m2.K ) eq Equilibrium
k Heat conduction coefficient (W/m.K) ex Exit
2
K Rate of surface area change (m /s) glass Gas phase
m Mass flow rate (kg/s) i Inlet
P Pressure l Liquid phase
Pe Peclet number m Mean
Q Heat transfer rate (W) o Outlet
Re Reynolds number p Particle
RH Relative humidity sat. Saturation
Sh Sherwood number sol. Solvent
St Stokes number soln. Solution
T Temperature (K) t Total, sheet
V (v ) Velocity (m/s) th Thermal
X Solution Solvent content v vapor
Y Gas solvent content

5 References

Abraham, J., 2009. International Conference On Harmonisation Of Technical Requirements For


Registration Of Pharmaceuticals For Human Use, Handbook of transnational economic
governance regimes. Brill, pp. 1041-1054.
Al-Khattawi, A., Bayly, A., Phillips, A., Wilson, D., 2018. The design and scale-up of spray
dried particle delivery systems. Expert Opin Drug Del. 15, 47-63.
Ali, M., Mahmud, T., Heggs, P.J., Ghadiri, M., Djurdjevic, D., Ahmadian, H., de Juan, L.M.,
Amador, C., Bayly, A., 2014. A one-dimensional plug-flow model of a counter-current spray
drying tower. Chem. Eng. Res. Des. 92, 826-841.
Amelia, R., Wu, W.D., Cashion, J., Bao, P., Zheng, R., Chen, X.D., Selomulya, C., 2011.
Microfluidic spray drying as a versatile assembly route of functional particles. Chemical
engineering science 66, 5531-5540.
Avvaru, B., Patil, M.N., Gogate, P.R., Pandit, A.B., 2006. Ultrasonic atomization: effect of
liquid phase properties. Ultrasonics 44, 146-158.
Azad, M., Arteaga, C., Abdelmalek, B., Davé, R., Bilgili, E., 2015. Spray drying of drug-
swellable dispersant suspensions for preparation of fast-dissolving, high drug-loaded,
surfactant-free nanocomposites. Drug Dev. Ind. Pharm. 41, 1617-1631.
Barba, A.A., d’Amore, M., Cascone, S., Lamberti, G., Titomanlio, G., 2009. Intensification of
biopolymeric microparticles production by ultrasonic assisted atomization. Chem. Eng.
Process. 48, 1477-1483.
Bellinghausen, R., 2018. Spray drying from yesterday to tomorrow: An industrial perspective.
Drying Technol., 1-11.
Bhakay, A., Rahman, M., Dave, R., Bilgili, E., 2018. Bioavailability Enhancement of Poorly
Water-Soluble Drugs via Nanocomposites: Formulation–Processing Aspects and Challenges.
Pharmaceutics 10, 86.
Bilgili, E., Ko, J., Chen, A., Smith, E., Rajniak, P., Fliszar, K., Wong, G., Rosen, L., Grubb, D.,
Bika, D., 2007. Controlling the key response variables through scaling rules during the scale-
up of a fluidized bed granulation process, Annual Meeting of the AIChE, Paper No: 553d.
Bilgili, E., Rosen, L.A., Ko, J.S., Chen, A., Smith, E.J., Fliszar, K., Wong, G., 2011.
Experimental study of fluidized bed co-granulation of two active pharmaceutical ingredients:
an industrial scale-up perspective. Particul. Sci. Technol. 29, 285-309.
Bögelein, J., Lee, G., 2010. Cyclone selection influences protein damage during drying in a mini
spray-dryer. Int. J. Pharm. 401, 68-71.
Boraey, M.A., Vehring, R., 2014. Diffusion controlled formation of microparticles. J. Aerosol
Sci 67, 131-143.
Brar, L.S., Elsayed, K., 2018. Analysis and optimization of cyclone separators with eccentric
vortex finders using large eddy simulation and artificial neural network. Separation and
Purification Technology 207, 269-283.
Carter, B.P., Schmidt, S.J., 2012. Developments in glass transition determination in foods using
moisture sorption isotherms. Food Chem. 132, 1693-1698.
Champion, J.A., Katare, Y.K., Mitragotri, S., 2007. Particle shape: a new design parameter for
micro-and nanoscale drug delivery carriers. J. Controlled Release 121, 3-9.
Chen, L., Li, G., Ma, X., Lim, J., Sivathanu, Y., 2018. A method for measuring planar Sauter
mean diameter of multi-component fuel spray based on the combined statistical extinction
tomography and particle imaging velocimetry. Fuel 214, 154-164.
Chen, L., Okuda, T., Lu, X.-Y., Chan, H.-K., 2016. Amorphous powders for inhalation drug
delivery. Adv. Drug Delivery Rev. 100, 102-115.
Clarke, A., Rieubland, S., 2012. Continuous inkjet printing. Google Patents.
Cortes, C., Gil, A., 2007. Modeling the gas and particle flow inside cyclone separators. Progress
in energy and combustion Science 33, 409-452.
Dalmoro, A., d’Amore, M., Barba, A.A., 2013. Droplet size prediction in the production of drug
delivery microsystems by ultrasonic atomization. Translational Medicine@ UniSa 7, 6.
Daly, R., Harrington, T.S., Martin, G.D., Hutchings, I.M., 2015. Inkjet printing for
pharmaceutics–a review of research and manufacturing. Int. J. Pharm. 494, 554-567.
Dobry, D.E., Settell, D.M., Baumann, J.M., Ray, R.J., Graham, L.J., Beyerinck, R.A., 2009. A
model-based methodology for spray-drying process development. J Pharm Innov. 4, 133-142.
Draheim, C., de Crécy, F., Hansen, S., Collnot, E.-M., Lehr, C.-M., 2015. A design of
experiment study of nanoprecipitation and nano spray drying as processes to prepare PLGA
nano-and microparticles with defined sizes and size distributions. Pharm. Res. 32, 2609-2624.
Durli, T., Dimer, F., Fontana, M., Pohlmann, A., Beck, R., Guterres, S., 2014. Innovative
approach to produce submicron drug particles by vibrational atomization spray drying:
influence of the type of solvent and surfactant. Drug Dev. Ind. Pharm. 40, 1011-1020.
El-Batsh, H.M., 2013. Improving cyclone performance by proper selection of the exit pipe.
Appl. Math. Modell. 37, 5286-5303.
Faure, A., York, P., Rowe, R., 2001. Process control and scale-up of pharmaceutical wet
granulation processes: a review. Eur. J. Pharm. Biopharm. 52, 269-277.
Filkova, I., Mujumdar, A.S., 1995. Industrial spray drying systems. Handbook of industrial
drying 1, 263-308.
Gan, H., Shan, X., Eriksson, T., Lok, B., Lam, Y., 2009. Reduction of droplet volume by
controlling actuating waveforms in inkjet printing for micro-pattern formation. J Micromech
Microeng. 19, 055010.
Gañán-Calvo, A.M., López-Herrera, J.M., Riesco-Chueca, P., 2006. The combination of
electrospray and flow focusing. J. Fluid Mech. 566, 421-445.
Gil, M., Vicente, J., Gaspar, F., 2010. Scale-up methodology for pharmaceutical spray drying.
chimica oggi/Chemistry Today 28.
Groom, S., Schaldach, G., Ulmer, M., Walzel, P., Berndt, H., 2005. Adaptation of a new
pneumatic nebulizer for sample introduction in ICP spectrometry. Journal of Analytical
Atomic Spectrometry 20, 169-175.
Gu, B., Linehan, B., Tseng, Y.-C., 2015. Optimization of the Büchi B-90 spray drying process
using central composite design for preparation of solid dispersions. Int. J. Pharm. 491, 208-
217.
Guns, S., Dereymaker, A., Kayaert, P., Mathot, V., Martens, J.A., Van den Mooter, G., 2011.
Comparison between hot-melt extrusion and spray-drying for manufacturing solid dispersions
of the graft copolymer of ethylene glycol and vinylalcohol. Pharm. Res. 28, 673-682.
Hacene, Y.C., Singh, A., Van den Mooter, G., 2016. Drug loaded and ethylcellulose coated
mesoporous silica for controlled drug release prepared using a pilot scale fluid bed system.
Int. J. Pharm. 506, 138-147.
Hanus, M., Langrish, T., 2007. Re‐ entrainment of wall deposits from a laboratory‐ scale spray
dryer. Asia-Pac. J. Chem. Eng. 2, 90-107.
Har, C.L., Fu, N., Chan, E.S., Tey, B.T., Chen, X.D., 2018. In situ crystallization kinetics and
behavior of mannitol during droplet drying. Chem. Eng. J. 354, 314-326.
Hede, P.D., Bach, P., Jensen, A.D., 2008. Two-fluid spray atomisation and pneumatic nozzles
for fluid bed coating/agglomeration purposes: A review. Chem. Eng. Sci. 63, 3821-3842.
Heng, D., Lee, S.H., Ng, W.K., Tan, R.B., 2011. The nano spray dryer B-90. Expert Opin Drug
Del. 8, 965-972.
Hsieh, D.S., Yue, H., Nicholson, S.J., Roberts, D., Schild, R., Gamble, J.F., Lindrud, M., 2015.
The secondary drying and the fate of organic solvents for spray dried dispersion drug
product. Pharm. Res. 32, 1804-1816.
Hutchings, I.M., Martin, G.D., 2012. Inkjet technology for digital fabrication. John Wiley &
Sons.
Kanojia, G., Have, R.t., Soema, P.C., Frijlink, H., Amorij, J.-P., Kersten, G., 2017.
Developments in the formulation and delivery of spray dried vaccines. Human vaccines &
immunotherapeutics 13, 2364-2378.
Kemp, I., Hartwig, T., Hamilton, P., Wadley, R., Bisten, A., 2016a. Production of fine lactose
particles from organic solvent in laboratory and commercial-scale spray dryers. Drying
Technol. 34, 830-842.
Kemp, I., Hartwig, T., Herdman, R., Hamilton, P., Bisten, A., Bermingham, S., 2016b. Spray
drying with a two-fluid nozzle to produce fine particles: atomization, scale-up, and modeling.
Drying Technol. 34, 1243-1252.
Kemp, I.C., 2017. Drying of pharmaceuticals in theory and practice. Drying Technol. 35, 918-
924.
Knopp, M.M., Olesen, N.E., Holm, P., Langguth, P., Holm, R., Rades, T., 2015. Influence of
polymer molecular weight on drug–polymer solubility: a comparison between experimentally
determined solubility in PVP and prediction derived from solubility in monomer. J. Pharm.
Sci. 104, 2905-2912.
Lebrun, P., Krier, F., Mantanus, J., Grohganz, H., Yang, M., Rozet, E., Boulanger, B., Evrard,
B., Rantanen, J., Hubert, P., 2012. Design space approach in the optimization of the spray-
drying process. Eur. J. Pharm. Biopharm. 80, 226-234.
Lee, K., Aalburg, C., Diez, F.J., Faeth, G.M., Sallam, K.A., 2007. Primary breakup of turbulent
round liquid jets in uniform crossflows. AIAA J. 45, 1907-1916.
Lefebvre, A.H., 1980. Airblast atomization. Progress in Energy and Combustion Science 6, 233-
261.
Lefebvre, A.H., McDonell, V.G., 2017. Atomization and sprays. CRC press.
Lewandowski, A., Jaskulski, M., Zbiciński, I., 2018. Experimental analysis of particle breakage
and powder morphology in foam spray drying, IDS 2018. 21st International Drying
Symposium Proceedings. Editorial Universitat Politècnica de València, pp. 1213-1220.
Li, M., Suriel, I., Vekaria, J., Proske, J., Orbe, P., Armani, M., Dave, R., Bilgili, E., 2018.
Impact of dispersants on dissolution of itraconazole from drug-loaded, surfactant-free, spray-
dried nanocomposites. Powder Tech. 339, 281-295.
Li, X., Zbiciński, I., 2005. A sensitivity study on CFD modeling of cocurrent spray-drying
process. Drying Technol. 23, 1681-1691.
Littringer, E., Paus, R., Mescher, A., Schroettner, H., Walzel, P., Urbanetz, N., 2013. The
morphology of spray dried mannitol particles—The vital importance of droplet size. Powder
Tech. 239, 162-174.
Lowinger, M., Baumann, J., Vodak, D.T., Moser, J., 2015. Practical considerations for spray
dried formulation and process development, Discovering and Developing Molecules with
Optimal Drug-Like Properties. Springer, pp. 383-435.
Lu, Z., Yang, Y., Covington, R.-A., Bi, Y.V., Dürig, T., Ilies, M.A., Fassihi, R., 2016.
Supersaturated controlled release matrix using amorphous dispersions of glipizide. Int. J.
Pharm. 511, 957-968.
Maas, S.G., Schaldach, G., Littringer, E.M., Mescher, A., Griesser, U.J., Braun, D.E., Walzel,
P.E., Urbanetz, N.A., 2011. The impact of spray drying outlet temperature on the particle
morphology of mannitol. Powder Tech. 213, 27-35.
Masters, K., 1994. Scale-up of spray dryers. Drying Technol. 12, 235-257.
Merlos, R., Wauthoz, N., Levet, V., Belhassan, L., Sebti, T., Vanderbist, F., Amighi, K., 2017.
Optimization and scaling-up of ITZ-based dry powders for inhalation. J Drug Deliv Sci
Technol 37, 147-157.
Mezhericher, M., Levy, A., Borde, I., 2010a. Spray drying modelling based on advanced droplet
drying kinetics. Chemical Engineering and Processing: Process Intensification 49, 1205-
1213.
Mezhericher, M., Levy, A., Borde, I., 2010b. Theoretical models of single droplet drying
kinetics: a review. Drying Technol. 28, 278-293.
Mezhericher, M., Levy, A., Borde, I., 2012. Three-dimensional spray-drying model based on
comprehensive formulation of drying kinetics. Drying technology 30, 1256-1273.
Mezhericher, M., Nunes, J.K., Guzowski, J.J., Stone, H.A., 2018. Aerosol-assisted synthesis of
submicron particles at room temperature using ultra-fine liquid atomization. Chem. Eng. J.
346, 606-620.
Mönckedieck, M., Kamplade, J., Littringer, E.M., Mescher, A., Gopireddy, S., Hertel, M.,
Gutheil, E., Walzel, P., Urbanetz, N.A., Köster, M., 2016. Spray Drying Tailored Mannitol
Carrier Particles for Dry Powder Inhalation with Differently Shaped Active Pharmaceutical
Ingredients, Process-Spray. Springer, pp. 517-566.
Montazer-Rahmati, M., Ghafele-Bashi, S., 2007. Improved differential modeling and
performance simulation of slurry spray dryers as verified by industrial data. Drying Technol.
25, 1451-1462.
Moon, S., Gao, Y., Wang, J., Fezzaa, K., Tsujimura, T., 2014. Near-field dynamics of high-
speed diesel sprays: Effects of orifice inlet geometry and injection pressure. Fuel 133, 299-
309.
Muñoz, M., Goutier, S., Foucaud, S., Mariaux, G., Poirier, T., 2018. Droplet size prediction in
ultrasonic nebulization for non-oxide ceramic powder synthesis. Ultrasonics 84, 25-33.
Nandiyanto, A.B.D., Okuyama, K., 2011. Progress in developing spray-drying methods for the
production of controlled morphology particles: From the nanometer to submicrometer size
ranges. Adv. Powder Technol. 22, 1-19.
Omer, K., Ashgriz, N., 2011. Spray nozzles, Handbook of Atomization and Sprays. Springer,
pp. 497-579.
Ozmen, L., Langrish, T., 2003. A study of the limitations to spray dryer outlet performance.
Drying Technol. 21, 895-917.
Pajander, J.P., Matero, S., Sloth, J., Wan, F., Rantanen, J., Yang, M., 2015. Raman mapping of
mannitol/lysozyme particles produced via spray drying and single droplet drying. Pharm.
Res. 32, 1993-2002.
Patel, K.C., Chen, X.D., 2005. Prediction of spray‐ dried product quality using two simple
drying kinetics models. J. Food Process Eng. 28, 567-594.
Paudel, A., Worku, Z.A., Meeus, J., Guns, S., Van den Mooter, G., 2013. Manufacturing of solid
dispersions of poorly water soluble drugs by spray drying: formulation and process
considerations. Int. J. Pharm. 453, 253-284.
Pierpont, D., Reitz, R.D., 1995. Effects of injection pressure and nozzle geometry on DI diesel
emissions and performance. SAE transactions, 1041-1050.
Poozesh, S., Akafuah, N., Saito, K., 2016a. New criteria for filament breakup in droplet-on-
demand inkjet printing using volume of fluid (VOF) method. Korean Journal of Chemical
Engineering 33, 775-781.
Poozesh, S., Grib, S.W., Renfro, M.W., Marsac, P.J., 2018a. Near-field dynamics of high-speed
spray dryer coannular two fluid nozzle: Effects of operational conditions and formulations.
Powder Tech. 333, 439-448.
Poozesh, S., Lu, K., Marsac, P.J., 2018b. On the particle formation in spray drying process for
bio-pharmaceutical applications: Interrogating a new model via computational fluid
dynamics. Int. J. Heat Mass Transfer 122, 863-876.
Poozesh, S., Saito, K., Akafuah, N.K., Graña-Otero, J., 2016b. Comprehensive examination of a
new mechanism to produce small droplets in drop-on-demand inkjet technology. Appl. Phys.
A 122, 110.
Poozesh, S., Setiawan, N., Akafuah, N.K., Saito, K., Marsac, P.J., 2018c. Assessment of
predictive models for characterizing the atomization process in a spray dryer’s bi-fluid
nozzle. Chem. Eng. Sci. 180, 42-51.
Poozesh, S., Setiawan, N., Arce, F., Sundararajan, P., Della Rocca, J., Rumondor, A., Wei, D.,
Wenslow, R., Xi, H., Zhang, S., 2017. Understanding the process-product-performance
interplay of spray dried drug-polymer systems through complete structural and chemical
characterization of single spray dried particles. Powder Tech. 320, 685-695.
Raffin, R., Guterres, S., Pohlmann, A., Ré, M., 2006a. Powder characteristics of pantoprazole
delivery systems produced in different spray-dryer scales. Drying Technol. 24, 339-348.
Raffin, R.P., Jornada, D.S., Ré, M.I., Pohlmann, A.R., Guterres, S.S., 2006b. Sodium
pantoprazole-loaded enteric microparticles prepared by spray drying: Effect of the scale of
production and process validation. Int. J. Pharm. 324, 10-18.
Rajan, R., Pandit, A., 2001. Correlations to predict droplet size in ultrasonic atomisation.
Ultrasonics 39, 235-255.
Rizk, N., Lefebvre, A., 1984. Spray characteristics of plain-jet airblast atomizers. Journal of
engineering for gas turbines and power 106, 634-638.
Rodgers, R., Hill, G., 1978. Equations for vapour pressure versus temperature: derivation and
use of the Antoine equation on a hand-held programmable calculator. British J anaesthesia
50, 415-424.
Rogers, S., Wu, W.D., Lin, S.X.Q., Chen, X.D., 2012. Particle shrinkage and morphology of
milk powder made with a monodisperse spray dryer. Biochemical Engineering Journal 62,
92-100.
Sadek, C., Schuck, P., Fallourd, Y., Pradeau, N., Jeantet, R., Le Floch-Fouéré, C., 2016.
Buckling and collapse during drying of a single aqueous dispersion of casein micelle droplet.
Food Hydrocolloids 52, 161-166.
Sadek, C., Schuck, P., Fallourd, Y., Pradeau, N., Le Floch-Fouere, C., Jeantet, R., 2015. Drying
of a single droplet to investigate process–structure–function relationships: a review. Dairy
Sci.& Tech. 95, 771-794.
Saito, K., Williams, F., 2015. Scale Modeling in the Age of High-Speed Computation, Progress
in Scale Modeling, Volume II. Springer, pp. 1-18.
Sallam, K., Dai, Z., Faeth, G., 1999. Drop formation at the surface of plane turbulent liquid jets
in still gases. Int. J. Multiphase Flow 25, 1161-1180.
Sallam, K., Faeth, G., 2003. Surface properties during primary breakup of turbulent liquid jets in
still air. AIAA J. 41, 1514-1524.
Sawicki, E., Beijnen, J.H., Schellens, J.H., Nuijen, B., 2016. Pharmaceutical development of an
oral tablet formulation containing a spray dried amorphous solid dispersion of docetaxel or
paclitaxel. Int. J. Pharm. 511, 765-773.
Schick, R.J., Brown, K., 2005. Spray Dryer Scale-Up: From Laboratory to Production. Spray
Analysis and Research Services, Printed in the USA, Spraying Systems Co.
Singh, A., Van den Mooter, G., 2016. Spray drying formulation of amorphous solid dispersions.
Adv. Drug Delivery Rev. 100, 27-50.
Sloth, J., Jørgensen, K., Bach, P., Jensen, A.D., Kiil, S., Dam-Johansen, K., 2009. Spray drying
of suspensions for pharma and bio products: Drying kinetics and morphology. Industrial &
Engineering Chemistry Research 48, 3657-3664.
Smith, S., 2014. Pharmaceutical Spray Drying Market, 2014-2024.
Sormoli, M.E., Langrish, T.A., 2016. The use of a plug-flow model for scaling-up of spray
drying bioactive orange peel extracts. INNOV FOOD SCI EMERG 37, 27-36.
Sosnik, A., Seremeta, K.P., 2015. Advantages and challenges of the spray-drying technology for
the production of pure drug particles and drug-loaded polymeric carriers. Adv. Colloid
Interface Sci. 223, 40-54.
Taki, M., Tagami, T., Ozeki, T., 2017. Preparation of polymer-blended quinine nanocomposite
particles by spray drying and assessment of their instrumental bitterness-masking effect using
a taste sensor. Drug Dev. Ind. Pharm. 43, 715-722.
Thabet, Y., Breitkreutz, J., 2018. Printing pharmaceuticals by inkjet technology: Proof of
concept for stand-alone and continuous in-line printing on orodispersible films. Journal of
Manufacturing Processes 35, 205-215.
Thybo, P., Hovgaard, L., Andersen, S.K., Lindeløv, J.S., 2008a. Droplet size measurements for
spray dryer scale-up. Pharm Dev Technol. 13, 93-104.
Thybo, P., Hovgaard, L., Lindeløv, J.S., Brask, A., Andersen, S.K., 2008b. Scaling up the spray
drying process from pilot to production scale using an atomized droplet size criterion. Pharm.
Res. 25, 1610-1620.
Tsotsas, E., Mujumdar, A.S., 2014. Modern Drying Technology, Volume 5: Process
Intensification. John Wiley & Sons.
Tzvetanov, G., 2018. Pharmaceutical Spray Drying Market (2nd Edition), 2018-2028, 2nd ed.
Utikar, R., Darmawan, N., Tade, M., Li, Q., Evans, G., Glenny, M., Pareek, V., 2010.
Hydrodynamic simulation of cyclone separators, Computational Fluid Dynamics. InTech, pp.
241-266.
van Deventer, H., Houben, R., Koldeweij, R., 2013. New atomization nozzle for spray drying.
Drying Technol. 31, 891-897.
Vehring, R., Foss, W.R., Lechuga-Ballesteros, D., 2007. Particle formation in spray drying.
Journal of Aerosol Science 38, 728-746.
Wang, B., 2015. Development of spray drying technology for microencapsulation of bioactive
materials. PhD diss. Auburn University, Auburn.
Wang, H., Chen, S., Cai, B., Ge, L., Chen, Q., 2014. Study on the dynamics of tribocharged coal
and mineral particles in free-fall triboelectric separator. Separation Science and Technology
49, 2990-2998.
Wang, J., de Wit, M., Boom, R.M., Schutyser, M.A., 2015. Charging and separation behavior of
gluten–starch mixtures assessed with a custom-built electrostatic separator. Separation and
Purification Technology 152, 164-171.
Woo, M., Fu, N., Moo, F., Chen, X., 2012. Unveiling the Mechanism of in situ Crystallization in
the Spray Drying of Sugars. Ind. Eng. Chem. Res. 51, 11791-11802.
Woo, M.W., Lee, M.G., Shakiba, S., Mansouri, S., 2017. Controlling in situ crystallization of
pharmaceutical particles within the spray dryer. Expert Opin Drug Del. 14, 1315-1324.
Wu, W.D., Amelia, R., Hao, N., Selomulya, C., Zhao, D., Chiu, Y.L., Chen, X.D., 2011.
Assembly of uniform photoluminescent microcomposites using a novel
micro‐ fluidic‐ jet‐ spray‐ dryer. AIChE Journal 57, 2726-2737.
Zhu, C., Shoji, Y., McCray, S., Burke, M., Hartman, C.E., Chichester, J.A., Breit, J., Yusibov,
V., Chen, D., Lal, M., 2014. Stabilization of HAC1 influenza vaccine by spray drying:
formulation development and process scale-up. Pharm. Res. 31, 3006-3018.
Figure captions:

Figure 1: Photos illustrating the changes in equipment and spray rates during the scale-up of
spray drying process

Figure 2: Schematic for spray drying scale-up strategy: keeping key response variables
unchanged across different scales.

Figure 3: Basic configuration of the most commonly used atomizers in spray drying: (a),
internal and external mixing two-fluid, (b) plain and swirl pressure nozzle, (c) spinning disk
and rotary cup, (d) capillary and standing wave ultrasonic, and (e) electrospray atomizer and
inkjet nozzles.

Figure 4: Schematic of the problem for large-scale analysis, along with given data, and mass
conservation equations.

Figure 5: Modelling tools in spray drying; (a), schematic of a 1D model in which droplet are
axially flowing downward, (b), solvent content of particle/droplet for a 2D, and for a 3D, (c),
simulation.

Figure 6: Morphology changes based on ICS and RH (Nuzzo et al., 2015; Vicente et al.,
2013).

Figure 7: Thermodynamic space for spray drying: red solid lines are constraints on Ti/To and
msol. / mg . Color map shows RH, begins from blue as minimum to red as maximum.

Figure 8: Two typical cyclone types in spray drying: left panel, vortex-based, right panel,
electrostatic separator.

Figure 9: Offered approach to handle sale-up process in spray drying of pharmaceutics.


Pilot-scale (x10 kg/hr)
Lab-scale (x100 g/hr)

Production-scale (x100 kg/hr)

Figure 1: Photos illustrating the changes in equipment and spray rates during the scale-up of
spray drying process

Lab

Key
response
variables

Pilot Production

Figure 2: Schematic for spray drying scale-up strategy: keeping key response variables
unchanged across different scales.
(d1)
(a1)
Liquid

Fluctuating
mesh

(a2) Piezoelectric
(d2)
Sound field

Nozzle
(b1)

(b2) (e1)

(c1) Liquid
Taylor cone

(e2) Liquid
(c2)
Piezo

Figure 3: Basic configuration of the most commonly used atomizers in spray drying: (a),
internal and external mixing two-fluid, (b) plain and swirl pressure nozzle, (c) spinning disk
and rotary cup, (d) capillary and standing wave ultrasonic, and (e) electrospray atomizer and
inkjet nozzles.
mig  mog  mg
Mass balance (gas)
Assumptions:

-Ignoring atomizing gas flow


Mass balance (liquid)
rate
Z

z
-Ideal gas law for gas and vapor
Mass balance (solid)

mil (1x)msoln. mYgog mYpop mp  xmso ln.


Y
X
Known: Z

 mso ln .
 x


 mig

 Tig
 Til

  th

Figure 4: Schematic of the problem for large-scale analysis, along with given data, and mass
conservation equations.
Q
l
o
s
s
mog

Top  Tog  To
mp mil
Yop  Yog  Yo
(a) (b) mig (c)

Figure 5: Modelling tools in spray drying; (a), schematic of a 1D model in which droplet are
axially flowing downward, (b), solvent content of particle/droplet for a 2D, and for a 3D, (c),
simulation.
ICS
 Decrease in RH and
increase in ICS promote
production of smooth
spherical particles
RH RH  Increase in ICS promotes
the production of dense
particles
 Decrease in ICS results
in raisin-like particles

ICS

Figure 6: Morphology changes based on ICS and RH (Nuzzo et al., 2015; Vicente et al.,
2013).

Allowable Tin
Min
imu RH
Ti m Design space
yiel =co
To
d nsta
nt

Allowable To

msol. / mg

Figure 7: Thermodynamic space for spray drying: red solid lines are constraints on Ti/To and
msol. / mg . Color map shows RH, begins from blue as minimum to red as maximum.

48
Exit pipe

Cylinder

Cone

Cone tip

Figure 8: Two typical cyclone types in spray drying: left panel, vortex-based, right panel,
electrostatic separator.

49
Input variables Output variables

SMD

RH
Small-scale Large-scale
Tout
spray dryer spray dryer
Eu

St

Re
Ap
pro
ach

Determining SMD from powder chemical and physical requirements


Scale-up criteria:
Keep key response
Establishing thermodynamic design space via various variables similar at
drying models, verifiable with few data on small-scale spray different scales to
maintain product
dryer
Keeping key response variables across separators, mainly by properties
adjusting dimensional aspects of separator

Figure 9: Approach for process development and scale-up of spray-drying of a


pharmaceutical product.

50
Table captions:
Table 1: Attempts to scale up the spray drying process

Table 2: SMD correlations for rotary nozzles

Table 3: Various types of kinetic drying models for particle formation

Table 4: Different modeling approaches for spray drying process

Table 1: Attempts to scale up the spray drying process


Scales Model Inlet Outlet Atomize Formulation Applicatio Findings Referen
conditio conditio r type n ce
ns ns
Lab Closed 30 kPa To=58- Two- HAC1 Vaccines All batches Zhu et

51
loop atomizer 62oC; fluid antigen showed al.
Buchi- pressure yield of nozzle (API) and amorphous (2014)
70-88%
290 and ∼0.6 Trehalose phase,
mL/min Dextran-40, physically
flow gelatin stable after
rate; hydrolysate storage at
processi enzymatic, elevated
ng gas sucrose, temperature
with arginine, and s, no
Ti=100o PVP-10 significant
C and no (excipients) differences
info for in water- in HAI
processi based salt antibody
ng gas solution titers
flow rate (solvent)
Pilot BLD-1 55-205 To=58oC Two- HAC1 (API)
(custom- kPa, ; yield fluid and
built) atomizer of ≥90% nozzle Trehalose
pressure and
and ∼4 hydrolyzed
mL/min gelatin
flow (excipient)
rate, in water-
processi based salt
ng gas solution
with (solvent)
o
Ti=100
C and
rate of
30 kg/h.
Lab NA 5, 8, 13, NA EM Polymer NA Qualitative Schick
17 and two- (excipients evaluation and
21 fluid and ns) and of the Brown
ml/min Acetone impact of (2005)
feed (solvent) nozzles’
flow rate operation
and conditions
varying on SMD
air and drop
pressure velocity.

52
range of Dissimilar
5-35 psi findings.
(35-240
kPa)
Producti NA 34, 51,
on 68 and
85
ml/min
and
varying
air
pressure
range of
35-240
kPa
Lab Open Ti= To=98±3 EM Sodium Prodrug Increase in Raffin
dm from lab
loop 150±5 ◦C two- pantoprazole et
(8µm) to
Buchi ◦C, fluid (API) and pilot scales al.(2006
(30µm); no
190 processi nozzle Eudragit® b)
other
ng gas with dl = S100, NaOH comparison
; pilot scale
flow 0.7 mm and
powder
rate: Methocel® attributes
were
10L/h, F4M
reproducibl
suspensi (excipients) e.
on feed in and water
flow (solvent)
rate: 4.0
ml/min/,
atomizin
g air
flow
rate: 60
l/min.
Pilot S52 Rotation To= rotating
® al
APV 85±5 ◦C. disc,
velocity
Anhydro of Powder and two-
30,000
humidit fluid
rpm;
suspensi y less with dl
on flow
than 4% =1.5
rate 2
L/h in mm

53
rotating
disc, and
atomizin
g air
with 49
to 196
kPa; Ti=
170±1◦C
.
Pilot Closed Varied To=45 Two- Acetaminop Drug Both render Thybo
Ti; N2
loop and 65 fluid hen (API) amorphous et al.
flow
Mobile rate= 70 ◦C with and PVP- particles (2008b)
kg/h,
GEA dl=1.0 K30 with dm=4–
atomizin
Niro g gas mm (excipient) 10 μm for
flow
do=2.0 in Ethanol pilot scale
rate= 9.0
kg/h; mm (solvent) and 4–30
feed
μm for
flow
rate=1.5- production
2.5 kg/h
scale;
Producti SD-12.5 Varied To=45 Two-
Ti; Flow particles
on closed and 65 fluid
rate of
with similar
cycle N2=1,25 ◦C with
0 kg/h, morphology
plant dl=4.0
atomizin
and
GEA g gas mm and
flow crystallinity
Niro d0= 10.0
rate=73
could be
kg/h; mm
feed produced in
flow
both;
rate=45
kg/h however,
scale-up
based
solely on
matching
droplet size
distribution
s was not
feasible.
Lab Buchi Ti= 103– To=60 Two- Lactose plus Food They Kemp et
290 197◦C, and 85 fluid a small obtained al.
closed processi ◦C nozzle, proportion of similar (2016)
loop ng gas EM, excipients particles
flow dl=0.7 were milled sizes (1.5-
rate: 12– mm, do= in 500 L of 2.5µm)
30 kg/h, 1.4 mm ethyl acetate across lab

54
feed (solvent) scale, and
flow production
rate: scale with
0.3–1.9 EM nozzle.
kg/h, Their
atomizin results
g airflow showed
rate: impact of
1.1–1.8 GLR across
kg/h the scales.
Producti Niro Ti= 90– To=60 Two-
on SD- 152◦C, and 85 fluid
12.5CC processi ◦C nozzle,
ng gas IM and
flow EM
rate:
1250
kg/h,
feed
flow
rate: 50–
100
kg/h;
atomizin
g airflow
rate: 70–
250 kg/h
Lab Open Ti= To=98±3 Two Sodium Prodrug Size and Raffin
loop 150±5 ◦C fluid pantoprazole morphology et
Buchi ◦C, nozzle (API) and of particles al.(2006
®
190 processi with dl = Eudragit were a)
ng gas 0.7 mm S100, NaOH different; as
flow and such the
®
rate: Methocel drug release
10L/h, F4M was
suspensi (excipients) different
on feed in water (⁓10%
flow (solvent) difference)
rate: across
04.0 scales,

55
ml/min/, especially
atomizin for lower
g airflow API
rate: concentrati
500NL/h ons
.
Pilot S52 Rotation To= rotating
® al
APV 85±5 ◦C. disc,
velocity
Anhydro of Powder and two-
atomizer
humidit fluid
30,000
rpm; y less with dl
suspensi
than 4% =1.5
on flow
rate mm
2 L/h in
rotating
disc, and
atomizin
g air
with 49
to 196
kPa; Ti=
170±1
Lab Buchi Ti= To= Two ITZ (API) Pulmonary Lab scale Merlos
o 0
290 90 C, 50 C fluid and mannitol drug showed a et al.
o
Tf=60 C, nozzle (excipient) 10% higher (2017)
air with dl = were API
processi 0.7 mm dissolved in release;
ng gas a solvent particle size
flow mixture of at pilot is
rate: 35 ethanol/ethyl twice; both
3
m /h, acetate/water scales
feed . rendered
flow spherical
rate: particles,
274.2 residual
g/h; solvent
atomizin content for
g airflow both were
rate: 800 similar
L/min (⁓0.3%);

56
Pilot GEA Ti= 80- To= Two both
o 0
Mobil 110 C, 50 C fluid achieved
o
Minor Tf=65 C, Powders nozzle ASD.
SD N2 were with dl =
processi collecte 1.3 mm
ng gas d at both
flow before
rate: 80- and after
90 kg/h, cyclone
feed
flow
rate: 0.5-
6.0 kg/h;
atomizin
g airflow
rate: 4-
25 kg/h

Pilot Niro liquid NA EM Acetaminop Drug By Thybo


A/S flow two- hen (API) measuring and
mobile rate: 0.8, fluid and PVP droplet size Hovgaar
minor 1.5, and nozzle K-30 distribution; d
2.5 kg/h; with dl= (excipient) they found (2008a)
atomizin 0.5, 0.8, in ethanol correlations
g gas 1.0 mm, (solvent) based on
flow do=5.0 non-
rate:5-30 mm dimensional
kg/h; parameters
Producti NA Liquid NA EM water to produce
on flow two- similar
rate: 45 fluid sizes across
kg/h; nozzle the scales
atomizin dl=4.0,
g gas do=9.0,
flow 10.0, or
rate: 45- 11.0
158 kg/h mm
Lab Buchi Ti=125– To=60- two- Phytosterols functional Similar Tolvea
B-191 155 °C, 100 oC fluid (active foods water et al.
liquid with dl= agent), whey encapsulati activity (2018)

57
feed 0.7 mm. protein on (⁓0.3),
flow yield: isolate, residual
rate= 0.5 43.21 ± inulin and moisture
L/h, 10.56% chitosan content
processi (excipients) (⁓4%), and
ng air in encapsulati
flow water(solven ng
rate: t) efficiency
600L/h (⁓20%)
across
Pilot Armfile Ti= 155 To=65- two- scales
d FT 80 o
°C, 75 C fluid
SD
liquid with dl=
feed 0.7 mm.
flow yield:
rate= 0.5 44.25
L/h, ±9.27%
processi
ng air
flow
rate:
600L/h
Lab Niro feed rate To=65, a 50 mm Mannitol pulmonary At lab- Littring
MOBIL 114,
of 0.84 Niro (API) drug scale, er et al.
E 140
MINOR l/h. and and rotary dissolved in delivery droplet with (2013)
™ 150 °C
rotationa having water (15% dm =27µm
l speed 24 [w/w]) produce 10-
of opening 15 µm
23,000 s with particles, at
rpm 3mm pilot, with
Pilot self-built feed rate To=67, LAMR 140 µm
tower 80, and
of 10.0 OT droplet,
with 92 °C
diameter l/h. and rotary particles of
of 2.7 m,
rotationa with a 80 µm were
height of
3.7 m l speed diameter obtained;
chamber
of 7,200 of 100 opposite
rpm mm and morphology
containi and
ng 60 crystallinity
bores of behavior is

58
3 mm seen; in lab
scale: low
To 
smooth
surface,
small
crystals;
rougher
morphology
with bigger
crystals for
higher To.
EM (external mixing), dl (liquid orifice dimeter), d0 (gas passageway diameter), Ti (gas inlet
temperature), To (gas outlet temperature), Tf (feed temperature), dm (mean dimeter), SMD (Sauter
mean diameter), NA (not applicable), GLR (gas to liquid ratio)

Table 2: SMD correlations for rotary nozzles


Reference Drop formation Ligament Sheet
disintegration disintegration
Vaneless rotary SMD / D  1.6We0.523 , We  l D3 2 / (8 ) 3We 1/6
SMD / dl  (1.5 )1/3 (1  )
nozzles Re2
(Matsumoto et Re  l D 2 / (4 )
al., 1974;
Matsumoto and
Takashima,
1969)
Vaned rotary SMD / D  aWeb (a and b contacts)
nozzles (atomizer
wheels) (Willauer
et al., 2006)
Rotary cup Q 0.05 L 0.6 0.65 0.02
SMD / D  C1 ( ) ( ) Re We
nozzles (Ahmed  D3 D
and Youssef,
2014)
Twin-fluid rotary 2  L V 0.41
SMD / D  C1 Re0.056 We0.02 ( ( ) )
nozzles  .D  .D
(Domnick, 2012)

59
Table 3: Various types of kinetic drying models for particle formation
Model type Description Pros Cons Sample
refs.
Characteristic Two drying stages: in the first, Simple and Unrealistic initial Keey
drying curve drying rate is similar to pure liquid, easy to drying rates, does not (1991);
(CDC) during the second stage, the relative implement, consider inflation of Langrish
drying rate is a function of a widely used in particle, and diameter and Kockel
characteristic solvent content many changes after first stage (2001);
parameter which, for example, can industries Woo et al.
be described to depend on the including (2008)
difference of vapor pressures at the detergent, food
liquid-gas interface etc.
Reaction It uses an empirical correlation simplicity and activation energies Chen and
engineering between particle solvent content success in depend on the material Xie (1997);
approach and partial vapor pressure at the modeling and drying conditions Putranto et
(REA) particle surface to estimate the during the employed, so its al. (2011);
drying rate based upon an activation spray drying overall applicability Schmitz-
process in which a material-based, of food and predictive Schug et al.
activation energy barrier must be materials. capabilities are limited (2016)
overcome for evaporation to happen to explored materials
such as skim milks,
whole milks, lactose,
proteins, sucrose,
maltodextrin, etc.
Transport uses coupled heat and mass More computationally Marshall Jr
(phenomena)- diffusion equations to account for comprehensive expensive because of (1955);
based drying transport phenomena with varying than CDC and moving boundary to Mezhericher
models complexities depending on made REA, better account for shrinking et al.
assumptions such as constant adjustment to of the droplet radius (2012);
volume in second drying stage. Eqs. different feed during the first drying Nešić and
are based on pure liquid droplet formulations stage and a receding Vodnik
evaporation theory with corrections and possible interface between the (1991)
that account for vapor pressure assumptions dry skin and wet core
decreases and droplet temperature during the second
increases during drying of the drying stage. A precise
droplet solution. screening of droplet
drying attributes such
as mass diffusivity,
particle porosity and

60
critical solvent content
during drying is
needed.
Deterministic particle formation is described by The only Computationally Handscomb
model population balance eqs. that account model that has expensive because of et al.
for the growth and aggregation of potential to several sub-models, (2009a);
suspended species leading to illustrate only applied for binary Handscomb
droplet solidification over drying morphology solutions so far et al.
durations. The diffusion coefficient and (2009b);
of micro- or nano-particles in the crystallization Handscomb
liquid phase depends on their during drying and Kraft
changing size (2010)

Table 4: Different modeling approaches for spray drying process


Model Description Solution Remarks Ref.
approach
Plug-flow- Pilot SD in finite Provides To, and particles history attributes Ali et al.
single zone detergent difference such as size, temperature, moisture content, (2014)
industry; four method in density, and axial velocity. Relative error up
stages, diffusion- MATLAB to 50% have been seen between data and
based DK simulation.
coupled with
particle inflation;
considered DSD;
considered heat
loss
Quasi-one- 1-D particle gSOLIDS Predicts To, and particles history attributes Pinto et al.
dimensional tracking model, software such as size, temperature, moisture content, (2014)
models the package density, residence time, and axial velocity.
temperature from PSE Qualitative validation with experimental
history and drying data, renders sensitivity analysis, e.g. the
rate of the analysis confirms high impact of initial
particles for droplet size on both particle size and
initial DSD, moisture content
considering heat
loss and ignoring
agglomeration,

61
DK based on
CDC
Plug-flow- 1-D particle finite Screening activity of bioactive agent, T o, Patel and
single zone tracking model difference and particles history attributes such as size, Chen (2005)
for skim milk, method via temperature, moisture content, density, and
model the Microsoft residence time. Predicted initial higher
temperature Excel drying rates when using CDC compared to
history, solvent REA and in turn lower bioactivity for bio-
activity and agent. Models were verified by experiments,
drying rate of the employing a glass-filament method.
particles for
mono-dispersed
droplets,
considering heat
loss and ignoring
agglomeration,
DK based on
CDC and REA
Plug-flow- A 1-D model for fourth-order tracked the changes of droplet diameter, Truong et
single zone a conical based Runge–Kutta density, temperature, moisture content, al. (2005)
and co-current procedure velocity and the difference between particle
pilot spray dryer via Qbasic temperature and its glass transition
with log–normal language temperature. With up to 40% relative error,
DSD, customized the prediction and measurement of Tog and
drying curve moisture content of particles agreed well for
based on CDC, different drying conditions.
but droplet
diameter was
updated based on
balloon shrinkage
(or changes in
particle density)
Plug-flow- The spray drying MATLAB Simple heat and mass balance-based model Sormoli and
single zone of orange-peel programming predicts the Tog and the outlet humidity and Langrish
extracts was relative humidity have been predicted within (2016)
scaled up five- 5% and for final moisture content within 60-
fold in terms of 75% of experimental data on pilot scale.
the feed flow rate, However, plug-flow model renders outlet
from a laboratory- gas temperature and absolute humidity with
scale SD to a less than a 1.5% error, and the overall

62
pilot-scale SD. moisture content of particle with 15% error.
CFD (3D) Modeled moisture Fluent 13 Reported particle and gas parameters; the Wawrzyniak
evaporation in an software 3D CFD calculations showed high et al. (2017)
industrial SD, in package instability of airflow in the dryer (lack of
which CDC DK from axial symmetry and oscillations of the flow
model is ANSYS field) due
integrated; to the specific construction of the air
agglomeration of distribution ring. Verification of CFD model
particles are was solely based on air velocity
considered. measurements in the industrial SD.
CFD (3D) Thin-film drying STAR- presented drying history of each individual Schmitz-
CCM+
data was first particle, then elaborated to optimize the Schug et al.
7.04.011
used in software spray drying process of dairy powders (2016)
from CD-
determining considering the spray drying conditions,
adapco
parameters in particle residence time distribution, droplet
REA, and then dispersion and geometric factors.
REA was Verified their simulation with data from a
implemented in GEA Niro two fluid nozzle with max 4%
CFD simulation errors in finial particle moisture content.
of a pilot-scale
spray dryer. The
simulations
investigated bio-
active agent
degradation
during drying
CFD (2D) A 2D using the Typical results viz. particle moisture Li and
CFD
axisymmetric SD content, temperature, velocity and particle Zbiciński
package
fitted with a FLUENT. histories were presented. It is found that the (2005)
pressure nozzle is stochastic effect of turbulent flow has a
simulated; significant influence on the particle path in
modified CDC as the drying flow. Simulation verification had
DK with different been done only by air temperature velocity
assumptions was measurement.
integrated into
commercial
software.
PSE (Process Systems Enterprise), SD (spray dryer), DK (drying kinetic), DSD (droplet size
distribution)

63
'Conflict of interest: none'

64
Graphical abstract

Inputs Outputs Modeling tools for fine-tuning


empirical scale-up approach

Sauter mean diameter


small-
Relative humidity scale Retaining key outputs
of spray drying for
Outlet temperature each transformation:
I. Atomization
Euler number II. Drying
large- III. Separation
Stokes number scale

Reynolds number

Spray drying
scale-up rules

65

View publication stats

You might also like