TRANSPORT PHENOMENA AND FLUID MECHANICS

Heat Transfer Performance of NETmix—A Novel Micro-Meso

Structured Mixer and Reactor
Marcelo F. Costa, Carlos M. Fonte, Madalena M. Dias, and Jose C. B. Lopes
Laboratory of Separation and Reaction Engineering–Laboratory of Catalysis and Materials (LSRE-LCM),
Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias,
4200-465 Porto, Portugal

DOI 10.1002/aic.15728
Published online in Wiley Online Library (wileyonlinelibrary.com)

NETmix is a novel static mixing technology consisting on a network of unit cells, comprising chambers interconnected by
channels. To assess the heat transfer capacity of NETmix, the NUB model was implemented to perform hydrodynamics and
heat transfer simulations. Due to the periodic nature of the NETmix structure, two central chambers and six half-chambers
were found to be sufficient to be representative of the whole network. The Nusselt numbers were determined based on the
CFD simulations, and when compared with theoretical results for laminar flow between parallel plates, 3–5 times higher
Nusselt number values were obtained with NETmix. This observed heat transfer rate enhancement, makes it suitable for fast
reactions where heat transfer is crucial. Finally, results obtained from this study show that NETmix presents a heat transfer
capacity one order of magnitude greater than microreactors, and 2–5 orders of magnitude greater than the most commonly
C 2017 American Institute of Chemical Engineers AIChE J, 00: 000–000, 2017
used devices in industry. V
Keywords: NETmix, microstructured devices, mixing, heat transfer, CFD

Introduction temperatures. This is particularly critical for highly exother-

mic or endothermic reactions, where the heat transfer rates
Static mixers have been used in industrial applications to
become crucial. Most heat exchange devices are designed so
perform continuous operations and have become standard
as to operate in the turbulent regime, as it is commonly consid-
equipment since the 1970s, as an alternative to mechanical
ered that laminar regime is not efficient from a heat transfer
mixers, for mixing of miscible or immiscible fluids, homoge- point of view.4 Some authors have proposed chaotic advection
nization of solid particles, and for heat and mass transfer as a means to enhance heat transfer in laminar flows, as it hap-
enhancement.1,2 This type of mixers is applied in a wide range pens in static mixers.5,6 Cross-section heat transfer enhance-
of different industrial processes in the pharmaceutical, petro- ment is due to the appearance of secondary transverse flows,
chemical, food, cosmetics, biotechnology, water/wastewater, improving transverse mixing coupled to flow division, pro-
paper, and polymer industries.2,3 The effectiveness of static moted by the geometry of the static mixer.7 This results in the
mixers to deliver good mixing of miscible fluids or to enhance virtual elimination of boundary layers that form the dominant
heat and mass transfer rates comes from their capacity to resistance to heat transfer in pipe flow.
induce transverse mixing and to bring fluid elements into close Heat transfer enhancement caused by the use of static
proximity.2 One of the most interesting characteristics of these mixers is not a new topic, and several authors have already
mixers is that the products are mixed by flow energy alone, described the benefits of the use of different type of static
thus they do not require external power except the required mixers, by reporting the dimensionless heat transfer coeffi-
power for pumping the fluids through the mixer; furthermore, cient, the Nusselt number, in the form of correlations useful
static mixers usually require small space, low equipment cost, for the design of heat exchange equipment.8,9 Some authors
and no moving parts.2 However, the use of static mixers usu- have shown that the heat transfer coefficients can be enhanced
ally originates an increase of the overall system pressure drop 2–3 times in Kenics static mixers operating in laminar regime,
as well as greater potential for fouling, relative difficulty for while for Sulzer SMX static mixers, the heat transfer rate can
cleaning and greater cost.2 be enhanced by a factor of five when compared with that in an
In many chemical processes/reactions, it is a main requisite empty tube.10 A first compilation on correlations for the Nus-
an effective and precise control of the operational fluid selt number obtained experimentally in tubes containing com-
mercially available static mixers is given in Thakur et al.2
This contribution was identified by Suzanne Kresta (University of Alberta) as Heat exchangers that include static mixing elements have
the Best Presentation in the session “Novel Mixer and Mixed Reactor Design” of
the 2012 AIChE Annual Meeting in Pittsburgh, PA. become commercially available and are currently used in the
industry.
Correspondence concerning this article should be addressed to J. C. B. Lopes at
[email protected].
The NETmix is a novel mixing technology11 that consists
on a network of mixing chambers interconnected by transport
C 2017 American Institute of Chemical Engineers
V channels (Figure 1). The network is built from the repetition

AIChE Journal 2017 Vol. 00, No. 00 1

 Figure 1. (a) Front and side-view drawing of a NETmix network; (b) Lab-scale NETmix.

of NETmix unit cells organized into consecutive pairs of nx NETmix is expected, however, it was never evaluated. The
rows, with ny and ny 21 number of cells, and the NETmix main objective of this work is to assess the NETmix’s heat
device is open to flow through equal number, ny , of inlets and transfer capacity, in terms of the flow regime impact in the
outlets. Each unit cell consists of one chamber and two chan- heat transfer resistances at the fluid level, by determination of
nels (two inlet and two outlet half channels) oriented at a 458 Nusselt numbers, and comparison between NETmix and other
angle from the main flow direction and consists of cylindrical technologies used for heat transfer.
chambers and rectangular channels. In the NETmix, proper
mixing can be achieved only above a critical channel’s Reyn-
Three-Dimentional CFD Modeling
olds number, where the flow inside the chambers evolves from
a fully developed laminar flow to a self-sustained dynamic and The 3-D nature of the flow inside the chambers and chan-
chaotic oscillatory laminar flow regime inducing strong lami- nels of the NETmix network makes problematic any geometri-
nar mixing. This phenomenon is well documented and cal simplification for the modeling strategy. Furthermore, the
described in previous works12,13 and is attributed to the local nonstationarity of the flow regime for certain ranges of the
hydrodynamic instabilities induced by the geometric charac- domain of interest for the operation of this device, obliges the
teristics and by the impinging jets interactions within the characterization of the heat transfer properties in terms of spa-
chambers. tial and time averaging, as discussed and introduced later on.
Mixing in the NETmix has been evaluated (both experimen- Laranjeira et al.12 observed that significant hydrodynamic
tally and using CFD modeling)12,14 showing that mixing can changes occur when a critical channel’s Reynolds number
be effectively and efficiently controlled. In terms of its energy (defined as Re5qt channel dh =l, where q is the fluid density, l is
performance, the power number and the Z factor of NETmix the fluid viscosity and t channel is the mean velocity in the chan-
were determined based on experimental measurements of the nels) of 150 (verified experimentally by Laranjeira et al.12) is
pressure drop, as well as CFD simulations, and proved to be achieved. At this Reynolds number, the flow evolves from a
competitive with existing mixers,15 delivering considerably fully developed laminar regime to a dynamic oscillatory flow
larger mixing power than stirred tanks using impellers. The regime, where mixing is enhanced due to chaotic advection.
NETmix technology has also already been industrially estab- Therefore, the influence of the hydrodynamics of NETmix is
lished when applied as a chemical reactor for the continuous of great importance to determine NETmix’s heat transfer per-
production of nanoparticles of hydroxyapatite, a calcium phos- formance. Fonte, et al.15 also observed this phenomenon using
phate that is predominantly found in teeth and bones within a hydrodynamic 3-D CFD model, that was used for the deter-
the human body, with controlled particle size and morphology, mination of the NETmix’s pressure drop, as well as for the
high surface areas, purity and nano-crystallinity.16,17 determination of parameters (that are a function of the geome-
NETmix can also be used as a heat exchange device, or as a try of the NETmix) to be used in a mechanistic model for pres-
reactor in processes where highly exothermic/endothermic sure drop prediction.
reactions occur and large amounts of heat have to be supplied/ The NUB model (NETmix Unit Block) developed by Fonte
removed to/from the system. As in other static mixers where et al.15 emerged from normal computational limitations, as it
chaotic advection is promoted, heat transfer enhancement in is impractical to simulate the whole control volume of the

2 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal
Figure 2. NETmix Unit Block (NUB) CFD model and unit cell. A unit cell drawing is shown in the top-left corner,
with the relevant dimensions represented. Unit cell dimensions of NUB model are: D 5 6.5 mm,
d 5 1 mm, x 5 3 mm, and L 5 8.5 mm.

NETmix device. However, as both the geometry and the continuity equation (conservation of mass) and the equation of
expected flow patterns have a periodically repetitive nature,12 motion (conservation of momentum) assume the form
it is conceivable to simplify the simulation of the larger system @q
by modelling only a small representative part of its flow 1rðq~
t Þ50 (1)
@t
domain. The NUB model (Figure 2) comprises five rows and
three columns, where only two complete central chambers and @ ðq~
tÞ
t  rðq~
1~ t Þ52rp2r  s 1q~
g 52rp2r
six lateral half-chambers are represented. Transversal symme- @t  
T
try surfaces (see Figure 2), normal to the xy plane, delimit the  2l r~ t1ðr~ tÞ 1q~
g (2)
half-chambers by crossing the complete lateral chambers in
the centre. A translational periodic boundary condition is where ~t is the velocity field, t is the time, p is the fluid pres-
applied to the resulting symmetry surfaces. The NUB transport sure, q and l are the local values of the fluid’s density and vis-
model was validated15 using reported experimental data, cosity, respectively. The energy conservation equation,
enabling the computation of transport properties without the neglecting viscous dissipation, is given by
need for experimental data.
As the original NUB model is already validated regarding @ ðqcP T Þ
1r  ðqcP~
tT Þ5r  ðjrT Þ (3)
its hydrodynamics, simulating particularly well the chaotic @t
behavior observed above the critical Reynolds number, an
adapted NUB model is applied in this work for the study of where T is the temperature, cP is the local fluid heat capacity
heat transfer in NETmix. To compare with laminar flow and j is the local fluid thermal conductivity. Viscous dissipa-
between parallel plates, it is assumed that heat transfer to the tion is neglected due to NETmix’s high energy efficiency
flowing fluid results mainly from conduction through the top demonstrated previously.15 From the observations made by
and bottom solid wall surfaces (identified as “heat transfer Fonte et al.,15 in the operational simulation range of this study,
surface” in Figure 2), neglecting the heat conduction through an energy dissipation of less than 1 mW is expected, in the
the curved surfaces of the chambers and the lateral side surfa- chaotic regime.
ces of the channels, of minor area, and driven by lesser ther- These equations are solved for the following boundary condi-
mal gradients between adjacent chambers and channels. In tions: no-slip condition is assumed in the walls of the domain,
t wall 50; simplified uniform velocity profile is applied to the
~
spite of the fact that heat transfer by conduction through the
network inlet channels; parallel flow condition is applied at the
NETmix walls in the flow direction can play a relevant trans-
network’s outlet channels by imposing a simplified constant
fer mechanism in certain circumstances, such as for large tem- and uniform pressure profile, pout 5p0 ; periodic boundary condi-
perature differences between hot and cold fluid, as in the tions are imposed at the symmetry surfaces of the half-
instances for highly exo/endothermic reacting processes, the chambers of the NUB model; either constant-wall temperature,
heat transfer by conduction along the solid will not be taken Twall 5T0 or constant-wall heat flux, q_ wall 5q_ 0 , is imposed at
into account in this modeling strategy. both top and bottom surfaces. These two last boundary condi-
tions are the two limiting cases consistent with the condition of
Governing equations and boundary conditions no conduction at the wall, as previously discussed.
The flow field simulation is achieved by integration of 3-D The flow is incompressible and the working fluid is water
forms of the continuity and motion equations while the tem- with variable local physical properties with temperature. The
perature field is obtained by solving the energy equation. The physical properties of water were fitted (valid in the range of

AIChE Journal 2017 Vol. 00, No. 00 Published on behalf of the AIChE DOI 10.1002/aic 3
 Table 1. Temperature Dependence of Water Physical NUB. A thermal gradient of DTin=out 560 K was chosen in
Properties.18 Valid for a Range of Temperatures from 273 K order to ensure that all local temperatures were within the
to 373 K validity range of the water’s physical properties relationships
(Table 1).
Property Expression
Density q52250111:5T23:4531022 T 2 13:2831025 T 3 Numerical discretization and solution methods
ðkg=m3 Þ 21 23 26 2 R
l51:57310 21:65310 T16:60310 T 2 The computational tool ANSYSV 13, an engineering simu-
Viscosity
ðPa  sÞ
28 3
21:18310 T 17:95310 212 4
T lation technology suite, was used for the geometrical design of
the NUB model (using Design Modeler) and for the simulation
cP 51:533104 2116T14:5131021 T 2 2 grid generation (using Meshing). A grid dependency test was
Specific Heat
Capacity 27:8431024 T 3 15:2031027 T 4 performed by sequentially increasing the grid refinement in
ðJ=kg KÞ different directions. In each grid, both the heat flow rate and
j524:3231021 15:7331023 T28:0831026 T 2 1
Thermal the average temperature at the middle of the channels were
Conductivity 11:8631029 T 3
monitored and compared with the correspondent values
ðw=m KÞ
obtained with the previous grid refinement. This process
stopped when the relative variation of these values between
273–373 K) as a function of the temperature to polynomial meshes with consecutive refinement levels was lower than
expressions from published experimental data18 (Table 1). 1%. This analysis was performed for different Reynolds num-
bers. From this test, it was observed that the grid discretization
Numerical simulation strategy in the flow direction used in previous works15 was not suffi-
As the hydrodynamics of NETmix is crucial for the determi- ciently refined to accurately simulate the temperature distribu-
nation of the Nusselt numbers in the NUB model, it is neces- tion inside the NUB domain. In this work, a grid with
sary to adopt a strategy that takes into consideration the hexahedral finite volumes with maximum volume edge length
presence or not of the oscillatory behavior inside the chambers of Dx50:05 mm in the flow direction is used. Successively
if the critical Reynolds number is exceeded. refined discretizations in the z direction was also performed in
When the Reynolds number is below the critical Reynolds the grid dependency test. It was observed that applying a size
number (channel’s Reynolds number at which the flow growth algorithm in this direction, so that the size of the ele-
evolves from a fully developed regime to a chaotic regime), ments grew from 0.05 mm near the walls (where the tempera-
that is, Re < 150, no oscillation patterns are observed and ture gradients are larger) to 0.1 mm in the centre position of
therefore only a steady-state simulation is performed, applying the NUB model, resulted in an enough refined grid. The num-
the boundary conditions described before. ber of elements in this mesh is lower than the number of ele-
When Reynolds number reaches a value large enough so that ments for a grid with constant sized elements with volume
oscillations are observed, a two-step simulation procedure is edge length of 0.05 mm, thus reducing the computational
required. First, to obtain an initial and symmetric solution, effort and the simulation time. The final computational grid
steady-state simulations are performed where symmetry condi- presented 2.5 million volume elements.
tions are applied by imposing shear-free surface condition to The flow equations solutions were obtained using the finite-
both the symmetry surfaces of the half-chambers of the NUB volume commercial CFD software ANSYS FluentTM 13.0. For
model and to the surfaces that divide the two complete central steady-state simulations (to obtain symmetric flow solutions),
chambers in the middle. Then, once a symmetric solution is the SIMPLEC scheme was used for the pressure–velocity cou-
obtained, the shear-free surfaces at the complete chambers in pling, the Standard and third-order MUSCL schemes were
the NUB model are removed, while the shear-free surface con- used for pressure and momentum discretization, respectively,
dition from the half-chambers is switched to periodic boundary and third-order MUSCL scheme was used for the discretiza-
conditions. With these conditions, a transient simulation is per- tion of the convective term of the energy conservation equa-
formed for a total flow time large enough for the flow inside the tion. For transient simulations, a second-order implicit scheme
NUB to fully develop the oscillatory patterns. A total flow time was used for time discretization. The simulation was consid-
of 15suc (unit cell residence time) is considered, being the Nus- ered to have converged when no significant variation of the
selt number computed in each simulation time step, determined outlet temperature and velocity values, averaged by the mass
using Dt5Dx=tmax;channel where Dx is the minimum dimension flowrate, was lower than 1% relatively to the previous itera-
of a fluid element in the channel, where the highest velocities tion, and when the normalized residues of the governing equa-
occur, and tmax;channel 53=2t channel . tions were lower than 1026 for steady state simulations, or
Furthermore, a uniform temperature of 293 K is imposed at 1024 for transient simulations, in each time step.
the network inlet channels. For the constant wall temperature Heat transfer simulation in NUB model were run on an
simulations, a constant temperature of 373 K is imposed to R R
IntelV XeonV 3GHz 8-core machine with 24 gigabytes of
both the top and bottom surfaces. For simulations with con- installed RAM memory resulting in simulation times of
stant wall heat flux boundary condition, a constant average approximately 30 min and 60 h for the steady and transient
heat flux is imposed. The wall heat flux is estimated by impos- simulations, respectively.
ing a maximum value to the temperature difference, DTin=out ,
between the NUB inlet and outlet, and is given by NETmix heat transfer performance
2wchannel cP DTin=out To evaluate the NETmix heat transfer performance, two
q_ wall  (4) distinct sets of simulations with different wall boundary condi-
Atransf;tot
tions were performed: constant-wall temperature and constant-
where q_ wall is the wall heat flux, wchannel is the channel’s mass wall heat flux. Both sets of simulations were performed at
flowrate, and Atransf;tot is the total heat exchange area of the different Reynolds numbers, varying from 5 to 600 at the

4 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal
Figure 3. Temperature contour maps at the middle plane of the domain for Rein 5 100 and Rein 5 300 for: (a) con-
stant wall temperature simulations; (b) constant wall heat flux simulations.
[Color figure can be viewed at wileyonlinelibrary.com]

NUB inlet channels, to completely observe the transition in interactions at the chambers, thus, promoting the reduction of het-
the hydrodynamic regime as described before, while maintain- erogeneities in the temperature field of the mixer (Figure 4).
ing a dynamic laminar regime. A proper way to assess the heat transfer operational capacity of
CFD simulations have shown that with the increase of the Reyn- NETmix in the dynamic flow regime can be obtained by determin-
olds number, the characteristics of the flow patterns inside the unit ing the time-average unit cell Nusselt number Nuuc , defined by
cells changes. Above the critical Reynolds number, chaotic quasi- tð
int
periodic flow oscillations start to occur inside the chambers, lead-
ing to a more uniform fluid temperature distribution. This is Nuuc ðtÞdt tð
int  
illustrated in Figure 3 that shows the temperature contour maps at 0 dh h ðtÞ h
Nuuc 5hNuit 5 tð
5 dt5dh (5)
the middle plane of the NUB geometry for two different values of int tint  ðtÞ
j j
 t
0
the Reynolds number evaluated at the inlet channels conditions dt
(density and viscosity evaluated at the inlet temperature), Rein : 0
Rein 5100, where no oscillations are observed; and Rein 5300,
above the critical Reynolds number, where the flow inside the mix- where h is the instantaneous unit cell heat transfer coefficient
ing chambers presents an oscillatory behavior that induces strong (evaluated at each instant), j
 is the thermal conductivity eval-
local laminar mixing. This uniformization on the temperature dis- uated at the unit cell average temperature, tint is the total
tribution pattern was observed for both sets of boundary conditions, dynamic simulation integration time, and the operator hit rep-
for constant-wall temperature and constant-wall heat flux. Perma- resents the time average of the ratio of the two physical param-
nent local hot spots are observed when no local laminar mixing eters. Defining Nuuc as the time-average unit cell Nusselt
occurs in the mixing chamber, leading to a highly heterogeneous number is helpful since it can be applied for both hydrody-
temperature field. Although hot spots are still observed for namic patterns above and below critical Reynolds number.
Rein 5300, their position changes randomly with time, due to the When the flow inside the chambers is chaotic, no steady-state
local hydrodynamic instabilities induced by the immerging jets solution is found, therefore, a time-dependent solution is

Figure 4. Temperature contour maps at the middle plane of the domain of the chamber in the fourth row for Rein 5 300
and constant wall heat flux simulation, for different flow times.
[Color figure can be viewed at wileyonlinelibrary.com]

AIChE Journal 2017 Vol. 00, No. 00 Published on behalf of the AIChE DOI 10.1002/aic 5
required. That problem is solved if the Nusselt number is an 1 x
x 5 ; (12)
average of the values at each instant, as both the heat transfer RePr x
coefficient and the thermal conductivity (due to temperature
where Pr5cP l=j is the fluid Prandtl number, x is the distance
changes) are varying with time. The values of Nuuc from Eq. 5
from the NUB inlet measured along the channels axis, x is the
were obtained using a shortcut, by computing the time-
clearance between the top and bottom plates and the Reynolds
average of h and j
 separately, that is
Number Re evaluated at the local conditions. The variable x ,
hhit huc the inverse of the Graetz number, is normally used in studies
Nuuc  dh 5 dh (6)
h
j it juc of entry length effects on laminar flow heat transfer prob-
lems.20 For any unit cell, a convective heat transfer coefficient,
where huc is the time-average unit cell convective heat transfer huc , can be computed by averaging between the dimensionless
coefficient and juc is the time-average unit cell thermal con- thermal entrance points, xin (at the centre of the inlet channel)
ductivity. Finally, to determine Nuuc , it is necessary to inde- and xout (at the centre of the outlet channel) as
pendently determine huc and juc , calculated performing the ð xout
average of the instantaneous values of the heat transfer coeffi- hloc ðx Þdx
cient and thermal conductivity: xin
huc 5 (13)
xout 2xin
tð
int
1 Q_ uc ðtÞ where hloc is the local heat transfer coefficient of the unit cell.
huc 5 dt (7)
Atransf;uc tint DTuc ðtÞ For predominant laminar parallel flow, the local heat transfer
0
coefficient depends on the geometry and configuration of the
tð
int flow, and is a function of x , expressed as19
1
juc 5  ðtÞdt
j (8)
tint hloc / ðx Þ21=3 (14)
0
Substituting Eq. 14 in Eq. 13 and integrating the latter, is pos-
where Q_ uc is the instantaneous p heat flow rate through the unit
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi sible to calculate an average convective heat transfer coeffi-
cell, Atransf;uc 5pD2 =412dðL2 D2 2d 2 Þ is the active heat cient as
exchange area of the unit cell, D is the chamber’s diameter, L  2=3   2=3
is the distance from the centre of two adjacent chambers, and 3 0 xout 2 xin
huc 5 C (15)
DTuc is a characteristic temperature difference. The total trans- 2 xout 2xin

ferred heat flux Q_ uc is determined by the energy balance to the
unit cell, and given by the sum of the enthalpy flow rates at where C0 is the configuration flow constant. For the different
the two inlets and two outlets cells of the NUB, the dimensionless thermal entrance points
Xð Xð can be calculated from
Q_ uc 5 TqcP ð~t ~
n ÞdA1 TqcP ð~
t ~
n ÞdA (9) 8  
>
>  nx 21 
< xin 5 nx 21=2 xuc
>
in Ai out Ai

where i stands for each of the two inlet channels, j stands for   (16)
>
> nx
each of the outlet channels, ~ n represents the normal unitary > 
: xout 5 x 

vector pointing outwards the control surface that delimits the nx 21=2 uc
boundaries of the domain, and A represents the respective being xuc is computed from Eq. 12, at the centre of the unit
cross-sectional flow area. cell’s chamber. The cell center distance xuc is determined by
The characteristic temperature difference is determined  
using an appropriate logarithmic mean temperature 1 L
xuc 5 nx 2 pffiffiffi (17)
difference19 2 2
   
Tb;out 2Twall;out 2 Tb;in 2Twall;in where L represents the linear distance between the in and out
DTuc 5    
(10)
ln Tb;out 2Twall;out = Tb;in 2Twall;in surfaces of each cell, a single constant value in the regular
NUB. In this way, it is now possible to estimate the existing
where Twall;in and Twall;out are the wall temperatures, at the inlet ratio at any cell of the NUB, between the average and local
and outlet of the unit cell, respectively and Tb;in and Tb;out are convective heat transfer coefficient as
the bulk temperatures, estimated from the average temperature
weighted by the mass flow rate of the fluid in the channel.  2=3  2=3
nx x 21
2 nnx 21=2
huc 3 nx 21=2
ðð 5   (18)
hloc x 5xuc 2 1
qð~
t ~
n ÞTdA nx 21=2
A
Tb;i or j 5 ð ð i or j : (11)
qð~
t ~
n ÞdA This ratio tends quickly to one when nx grows along the NUB.
Ai or j Using Eq. 18 and the NUB dimensions in Figure 2, relatively
small differences are obtained between the local and averaged
The unit cell Nusselt numbers for the NUB model, calculated heat transfer coefficient (e.g., 0.86% and 0.15% for unit cells
from Eq. 6, are surface averaged values rather than local val- of row two and four, respectively). Therefore, the Nusselt
ues. To determine the importance of the entry length effect numbers can be considered as effectively constant after a few
along the cells in the NUB, a dimensionless thermal entrance cells, and thus all the reported heat transfer coefficients were
distance variable x is introduced as20 determined at the fourth cell of the NUB.

6 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal
 the lateral regions of the chambers, and in the loss of the sym-
metry within the z direction. This represents the start of the
appearance of instabilities in the chambers, although there is
no significant influence of these instabilities on the overall
hydrodynamics. Nonetheless, this transition results in an
increase of the Nusselt number, as observed from Figure 5. At
this point, the flow is no longer fully parallel. The vortices
intensity, area and asymmetry increase with the increase of the
Reynolds number, resulting in deformation phenomenon, such
as stretching and folding, and in an increase of fluid volume

Figure 5. Nusselt number as a function of the inlet

channel Reynolds number for the unit cell of
the fourth row of the NUB model for two
cases: constant wall temperature and con-
stant wall heat flux.

The computed Nusselt number values, for the two sets of

boundary conditions (constant wall temperature and constant
wall heat flux), are reported in Figure 5 as a function of the
Reynolds number evaluated at the inlet channel’s conditions,
Rein . As can be observed, the Nusselt number value remains
constant for low Reynolds numbers ð<20Þ. The Nusselt num-
bers converge to an asymptotic lower constant value, where
the thermal boundary layer is completely developed, resulting
in a constant heat transfer coefficient along the mixer, regard-
less of the Reynolds number value. For larger Reynolds num-
bers ð>20Þ, a gradual increase of the Nusselt number is
observed, with the largest increase between Rein 550 and
Rein 5150. Flow dynamics change at Rein 550, with a strong
impact in the increase of the Nusselt number. For Rein  150
the oscillatory flow patterns are completely developed, which
induces strong chaotic mixing that increases the heat transfer
coefficient of the NETmix with increasing Reynolds number,
in a similar manner as in the cases of turbulent pipe flows.
The analysis of Figure 5 is more interestingly made if differ-
ent flow regimes are associated to different Reynolds numbers.
In fact, the increase of Nusselt number with the Reynolds num-
ber is explained by the flow regime transitions that occur inside
the NETmix chambers. The contour maps of the sign (positive
or negative, i.e., the direction of the flow) of the z component of
the vorticity xz , determined by x5r3~ t, in the middle plane
of the unit cell at the fourth row for simulations with constant
wall temperature boundary-conditions, for different Reynolds
numbers evaluated at the inlet of the NUB model, is shown in
Figure 6. It is possible to distinguish different operational vor-
ticity regimes from the vorticity map analysis.
For low Reynolds number ðRein  10Þ, the flow is mostly
segregated and steady parallel flow is observed. For Rein 55, Figure 6. z component vorticity contour maps evalu-
the jets present opposite sign vorticity due to the symmetry of ated at the middle plane of the unit cell at
the expansion that occur at the chamber. Small recirculation the fourth row as a function of the Reynolds
zones are only observed in the lower region of the chamber. number evaluated at the inlet for simulations
For Rein 510, the same behavior is observed. However, the with constant temperature boundary-
increase of inertial forces translates the point of velocity curl- condition. Color scale represents the sign of
ing above the centre of the chamber. This turnover point the component: white—positive, counter-
moves towards the wall of the chamber as Reynolds number clockwise; blue/dark—negative, clockwise.
increases. Contour map for Rein 5 150 was obtained at
At Rein 520, a flow regime transition occurs, resulting in a time step of 15suc.
the appearance of strong dynamic vortices near the walls in [Color figure can be viewed at wileyonlinelibrary.com]

AIChE Journal 2017 Vol. 00, No. 00 Published on behalf of the AIChE DOI 10.1002/aic 7
interfacial area,21 thus enhancing heat transfer, by bringing For constant wall temperature, for the flow region where the
fluid elements with larger temperature gradients closer. A sec- thermal boundary layer is still developing, the local parallel
ond flow regime transition is observed at Rein  50, where plate Nusselt is given by19
vortices start to appear in the upper region of the chamber, 2   21=3
near the walls. From Figure 5, it is possible to observe that for NuPP 5 1=3 xPP (20)
3 Cð4=3Þ
Rein  50, the slope of the curve increases when compared to
the region 20 < Rein < 50, indicating that the appearance of and for the thermally fully developed flow, the Nusselt number
these secondary vortices enhance mixing and heat transfer is found to be NuPP 57:54. Eqs. 19 and 20 were obtained by
even further. When the Reynolds number reaches the critical analytically solving the energy equation, assuming that the
value, Rein  150, the system evolves from a steady flow to a laminar parabolic velocity profile is fully developed and
self-sustained chaotic dynamic laminar flow regime, character- constant fluid properties. But as the used properties for the
ized by its dynamic asymmetric flow patterns. This behavior comparison are the local properties, the comparison can be
induces strong dynamic oscillations, resulting in the formation done.
of vortices. The presence of these dynamic vortices that have Prior to a direct comparison between Nusselt numbers for
the capacity to interact with each other deforms the fluid ele- NUB and parallel plates, it is necessary to take into account
ments even further,22 resulting in an increase of heat transfer the differences in geometry and Reynolds and Nusselt num-
rate through the walls. Observed chaotic oscillation frequency bers definitions for both cases. The following analogy for the
increases with the increase of the Reynolds number.15 thermal entrance length is obtained
These hydrodynamic regime transitions were observed in 2xPP 
simulations for both sets of boundary conditions. xuc 5 x (21)
dh PP
and for the Nusselt number, a similar analogy can be found
Heat transfer equipment benchmarking
dh
To ascertain the impact of the results from Figure 5, a com- Nuuc 5 NuPP (22)
parison of the obtained results from the NUB model with a 2xPP
similar heat transfer device is necessary. Since the flow in the In this way, it is now possible the comparison of the Nusselt
NETmix for low Reynolds numbers, is parallel and segre- numbers computed for the unit cell at the fourth row, with the
gated, a comparison with heat transfer with laminar flow equivalent values obtained for laminar flow between parallel
between parallel plates is the most suitable for this purpose. plates, using the analogue Eqs. 19 through 22. However, direct
Also, NETmix and parallel plates have geometrical similari- comparison is only meaningful if the hydrodynamic conditions
ties, as heat transfer is performed through both bottom and top lead to similar flow patterns for both cases.
walls, when neglecting conduction through lateral walls as this From Figure 6, it is observable that parallel and segregated
transfer area is relatively unimportant. Additionally, the case flow in NETmix occurs only at low Reynolds number
for laminar flow between parallel plates presents analytical ðRein < 10Þ and Figure 7 shows that the Nusselt numbers
solutions for asymptotic regimes, derived for Newtonian fluids obtained from simulations for the NUB model agrees well
with constant physical properties and fully developed laminar with the analytical solutions for laminar flow between parallel
velocity profiles,19 making it directly comparable. plates, in the whole range of working regimes, for constant
For laminar flow between parallel plates, two asymptotic Rein 510. In these NUB simulations, the variations in the
values are considered: a first region where the thermal dimensionless entrance length were obtained by changing the
boundary-layer is being developed; and a second region where Prandtl number of the fluid. Also, the comparison is done by
the thermal boundary layer achieves its maximum thickness of computing the Nusselt numbers only for the constant wall heat
xPP =2, the half plate gap size, where the maximum heat trans-
fer resistance is attained. The use of the dimensionless thermal
entrance length (or its inverse, the Graetz number) is useful to
determine in which regime the system is: for low thermal
entrance lengths, a developing thermal boundary layer is pre-
sent; while for high thermal entrance lengths, the boundary
layer is fully developed.
The local Nusselt number is defined as NuPP 52hPP xPP =j
for parallel plates, where hPP is the parallel plates local con-
vective heat transfer coefficient. For constant wall heat flux in
the flow region where the thermal boundary layer is still in
development, the local Nusselt number is function of the
dimensionless thermal entrance length and is given by19
2Cð2=3Þ   21=3
NuPP 5 xPP (19)
31=3
Figure 7. Nusselt number as a function of the thermal
where C is the gamma function and xPP is the dimensionless entrance length for the constant wall heat
thermal entrance length, calculated through Eq. 12, using flux simulation for Rein 5 10. NUB values
the parallel plates local Reynolds number, defined as were computed for the unit cell in the fourth
RePP 52xPP qt =l. For the thermally fully developed region,19 row. x*uc is altered by changing the heat
the Nusselt number for constant wall heat flux is NuPP 58:24. capacity of the fluid.

8 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal
 number gradually decreases within the reactor. The flow dis-
turbances that occur in the chambers due to the impinging jets
interactions result in flow instabilities and oscillations result-
ing in a constant boundary layer thickness, even in the first
rows of cells. This is a significant aspect as it maximizes and
homogenizes the heat transfer capacity of the NETmix
throughout its whole domain.
NETmix heat transfer performance
The predictions of Nusselt numbers in the NUB model are
crucial for the determination of the heat transfer capacity of
NETmix, and for its comparison with the most common tech-
nologies used at laboratorial or industrial scales used for heat
transfer, such as shell and tube, and plate heat exchangers. The
Figure 8. Nusselt number as a function of the thermal predicted NETmix Nusselt numbers are compared in Figure
entrance length for the wall constant temper- 10 with the Nusselt numbers of fully developed flow in tubes
ature simulation. NUB values were computed and between parallel plates, for both laminar and turbulent
for the unit cell in the fourth row. The dashed regimes. In both, the heat transfer resistance was estimated for
line represents the new asymptotic value for just the inner side of the system, that is, the heat transfer by
NETmix. conduction through lateral solid surfaces and the external heat
transfer resistance were not considered.
In Figure 10, the Nusselt numbers for fully developed flow
flux boundary condition for both laminar flow between parallel
in NETmix, previously computed and analyzed in Figure 5,
plates and the NUB model.
The dependence of the unit cell Nusselt number on the are presented within the laminar regime by fitting the simula-
dimensionless thermal entrance length is shown in Figure 8, tion results by a fitting curve, as well as the points obtained
for the simulations with constant wall temperature for the NUB from CFD simulations. Nusselt numbers for fully developed
model and the analogue solution for laminar flow between par- laminar flow in tubes and between parallel plates are also rep-
allel plates. With the increase of Reynolds number, the Nusselt resented in Figure 10, for a direct comparison with NETmix,
number values start to deviate from the expected value for the being constant and independent of Reynolds number, and are
flow between parallel plates, tending to a considerably higher available in literature19,20 from the analytical solution. Nusselt
asymptotic value, represented by the dashed line in Figure 8. numbers for fully developed turbulent flows in pipes and
This behavior is expected from the evolution of the flow pat- between parallel plates are also shown in Figure 10, and
terns analysis in Figures 5 and 6. This augmented heat transfer strongly depend on the Reynolds and Prandtl numbers, being
capability clearly demonstrates the impact of the vortices gen- estimated from correlations existing in literature. The Dittus-
eration on the Nusselt number, as this increase is not only due Boelter correlation23 is widely used for the determination of
to the reduction of the boundary-layer thickness alone. In fact, Nusselt number in tubes, while for flow between plates, a simi-
the vortices have an important impact in reducing the lar correlation can be found.24 In both cases, the Prandtl num-
boundary-layer thickness, thus enhancing the heat transfer ber used is that of water at temperature of 258C ðPr  7Þ.
rates. This behavior is observed after Rein > 20. However, The different devices and flow regimes are compared in Fig-
when the critical Reynolds number is reached, the Nusselt ure 10, where it can be observed, that within the range of this
number tends to a new asymptotic curve, parallel to the one study, for the typical NETmix operational Reynolds numbers
observed for parallel plates. The effect of the existence of 3-D
flow in the chambers due to the dynamic oscillatory regime is
responsible for the observed enhancement of the heat transfer
rate when compared to the rate obtained for flow between par-
allel plates. In the low Reynolds number region, NETmix
presents the same behavior as parallel plates, in agreement
with the previous observations. The comparison for constant
wall heat flux, as well as the new asymptotic NETmix values
in the dashed line, is shown in Figure 9. The flow behavior pre-
sented by NUB is similar to what is observed for constant wall
temperature simulations, but in this case the NUB Nusselt dif-
ference with parallel plates is even larger.
Furthermore, it also observed that, for large Reynolds num-
ber, Rein > 150, where flow oscillations are fully developed,
the Nusselt number values determined at the other rows of the
NUB model (rows 1, 2, and 3), are constant and equal to the Figure 9. Nusselt number as a function of the thermal
values computed at the fourth row presented in Figures 8 and entrance length for the constant wall heat
9, that is, they become independent of the distance from the flux simulation. NUB values were computed
inlet. This means that for large Reynolds numbers, the average for the unit cell in the fourth row. The dashed
NETmix unit cell Nusselt number is constant and equal for all line represents the new asymptotic value for
chambers, whereas for low Reynolds numbers, the Nusselt NETmix.

AIChE Journal 2017 Vol. 00, No. 00 Published on behalf of the AIChE DOI 10.1002/aic 9
 literature.25,26 For tubular reactors and microreactors, geomet-
rical definitions are used in order to determine the specific sur-
face area, as they can be conceptualized as cylinders/tubes in
direct contact with the cooling fluid.
Due to the fact that reactors/exchangers may present differ-
ent sizes, the specific areas of equipment area are determined
within a specific range, considering a minimum and a maxi-
mum characteristic dimension, in agreement with typical val-
ues used in these technologies, for active area determination.
Specific surface areas of different heat exchange equipment
are available in Table 2, as well as the characteristic dimen-
sions for each type of equipment, if applicable. Table 2 shows
that NETmix presents 2–3 orders of magnitude larger specific
surface area for heat transfer than the most commonly used
technologies, such as stirred tanks and tubular reactors. Due to
the NETmix geometry and small size, it presents a specific
surface area comparable with that of microreactors, which are
known to have large surface/volume ratios.27,28 This result
shows that NETmix has the potential to remove large amounts
Figure 10. Thermally fully developed Nusselt number
of heat per volume of fluid.
as a function of the Reynolds number for
Finally, the specific heat transfer capacity for different tech-
fully developed laminar flow in pipes and nologies can be determined by multiplying the heat transfer
between parallel plates,19,20 turbulent flow coefficient and the specific surface area
between pipes and parallel plates23 and for ^ Atransf
h5hhi t (24)
NETmix device (obtained in this work). V
where h^ is the specific heat transfer capacity.
between 100 and 1000, the NETmix presents Nusselt numbers The specific heat transfer capacity range for each equipment
2–3 times larger than those of laminar flow in tubes or is computed using the lower/upper bound values of the specific
between parallel plates. For low Reynolds numbers, the surface area, from Table 2, and the corresponding characteris-
absence of significant flow disturbances (such as vortices or tic dimensions, to determine the heat transfer coefficient from
oscillations) in NETmix results in a constant Nusselt number. Nusselt number. For stirred tanks, heat transfer coefficient
However, the presence of vortices and oscillations causes a range boundaries are obtained from literature.26,29 No infor-
significant increase in the Nusselt numbers, and at or above mation regarding Reynolds number is given by Trambouze
the critical Reynolds number, the Nusselt number is larger for et al.26 and Ferrouillat et al.29 However, these devices usually
the NETmix than for tubes and plates. operate in turbulent regimes, to maximize heat fluxes. In this
However, the overall heat transfer performance of different devi- case, the lower bound of the specific heat transfer capacity is
ces cannot be solely compared based in Nusselt number compari- determined using the lower bound value of specific surface
sons (and hence, heat transfer coefficients). The heat transfer area and the smaller value of heat transfer coefficient, while
capacity of heat exchange equipment, is also determined by the the upper bound value is computed using the larger values of
availability of area for heat transfer. Therefore, a comparison of specific surface area and heat transfer coefficients. For tubular
active specific surface area for heat transfer, for different heat reactors, the Nusselt number is determined using correlations
exchange equipment cannot be neglected. For that, the most com- for turbulent flow in tubes,19,23 assuming a Reynolds number
monly used equipment for reaction coupled with heat exchange are of 104 for the lower bound value of specific heat transfer
considered here for comparison purposes, namely: stirred tanks capacity, and 105 for the upper bound value, as this is the typi-
with external jacket or internal serpentines; jacketed tubular reac- cal range of operation of liquid tubular reactors. For the spe-
tors; stirred tanks with external heat exchangers; microreactors. cific case of microfluidic devices, due to its characteristic
The specific area of NETmix, Atransf =V, is determined from sizes, operation Reynolds number is very small, typically in
geometrical considerations, using the unit cell dimensions.
NETmix has a similar geometry when compared to parallel
Table 2. Surface to Volume Ratio of Typical Heat Exchange
plates. However, there is a reduction of the effective exchange
Equipment
area of NETmix when compared with parallel plates with the
same plate clearance x as NETmix. For the unit cell dimen- Heat Transfer
sions used in the NUB model (see Figure 2), an area reduction Atransf =V Reynolds  Coefficient 
of 27% relatively to parallel plates is verified, even if the lat- Device ðm2  m23 Þ Number W  m22  K21
eral areas are considered. The specific surface area of NETmix Stirred tank— 2–925,26 103 2105 60–200025
considering the lateral areas is estimated from Jacket
 pffiffiffiffiffiffiffiffiffiffi  Stirred tank— 0.5–2025,26 103 2105 700–300025
  112x1 8 L2 D2 2d2 d 1 x d Serpentine
Atransf 1 D p D D D D Stirred tank— 5–1029 103 2105 100029
  pffiffiffiffiffiffiffiffiffiffi  (23) External
V x 1 2
1 4 L2 D 2d d
2
2 p D D heat exchanger
Tubular reactor 15–40029 104 2105 150–8000
For stirred tanks with cooling jackets, internal serpentine or Microreactors 2000–4000035 1–5032 60–160034
external heat exchangers, specific areas were obtained from NETmix 1500–11000 150–60012,15 10000–60000

10 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal
 Figure 11. Specific heat transfer capacity of typical heat exchangers equipment.

the order of 1 or even lower,30,31 in conditions where the flow where the flow is fully dynamic and chaotic. Therefore, the
is fully laminar and parallel. Microfluidic devices can be seen Nusselt number is determined at Re5600, considering con-
as a tube, or a rectangular duct with aspect ratio close to 1, hav- stant wall temperature.
ing micrometric dimensions. Therefore, assuming a Re51 and The specific heat transfer capacity range of each of the con-
a tube diameter of 2.5 mm (the larger characteristic dimension sidered equipment is shown in Figure 11. The results from
considered in the analysis in Table 2), it is possible to verify, Figure 11 highlight the great potential of the use of NETmix
using the analogue for tubes of Eq. 12, that it is necessary only as a reactor with an outstanding heat transfer rate when com-
1.5 mm to achieve the region where the inverse of the Graetz pared with the most commonly used technologies for reaction
number is large enough to consider a fully developed thermal coupled with heat exchange. The difference from stirred tanks
boundary-layer.19 As microfluidics typically have lengths of a and NETmix is further accentuated coupling the surface area
few centimetres, it is assumed that microfluidics mostly oper- and the heat transfer coefficient, resulting in five orders of
ate in the fully developed region. Few studies using larger magnitude larger in some conditions. It is possible to verify
Reynolds numbers are described. However, due to the increase that NETmix is capable to remove heat at a rate that is almost
of the pressure drop, microreactors are usually operated at rela- one order of magnitude larger than microreactors, due to mix-
tively low Reynolds numbers. For this analysis, it is considered ing induced by oscillations. In microreactors, the flow is lami-
that the Reynolds numbers in the microchannels varies from 1 nar and mixing is mostly due to molecular diffusion, therefore,
to 50 (however, it is typically observed lower values).32 low convection rates are observed, leading to small heat trans-
It is important however to note, that when the channel’s fer coefficients.
dimensions are within the micrometer range, usually lower These results validate the capability of NETmix to outper-
Nusselt numbers than the ones obtained from laminar flow in form the most common reaction devices coupled with heat
pipes or parallel plates are obtained.33 Peng et al.34 determined exchange, being a reaction device with low power require-
the Nusselt numbers for several microchannels geometries and ments, high mixing intensity, high throughput potential and
Reynolds numbers within the laminar flow regime, and pro- enhanced heat transfer, essential for highly exo/endothermic
posed an empirical correlation similar to the Dittus-Boelter reactions.
correlation, by changing the constant for laminar flow in
microchannels. Peng et al.34 focused their study on the Reyn- Conclusions
olds number range of 50 to 4000, and observed that the lami- In this study, the heat transfer performance of the static
nar transition never occurred at Reynolds numbers lower than mixer NETmix was assessed and quantified. The study is
200. Therefore, the range of Reynolds number of interest in based on CFD simulations of a previously developed15 3-D
the present study (1 to 50) is always within the laminar flow model of NETmix.
regime, thus enabling the use of the correlations presented by The NETmix average unit cell Nusselt number was deter-
Peng et al.34 The heat transfer coefficient in microchannel is mined for different Reynolds numbers, and for two different
determined using the information provided by Peng et al.34 boundary conditions: constant wall temperature and constant
For NETmix, it is considered that the Reynolds number is wall heat flux. In both cases, it was observed that the Nusselt
equal to 600, within the laminar regime but in conditions number increases with increasing Reynolds number. It is

AIChE Journal 2017 Vol. 00, No. 00 Published on behalf of the AIChE DOI 10.1002/aic 11
observed that the existence of hot-spots is reduced when the nx = number of NETmix rows
flow patterns present an oscillatory behavior, and the remain- ny = number of NETmix columns
ing are renewed with time, leading to a more efficient mixing p= pressure, Pa
Pr = Prandtl number
and heat transfer. q_ = boundary condition heat flux, W
This study suggests that for large Reynolds numbers, 3–5 Q_ = heat flux, W
times higher heat transfer rates can be achieved when com- Re = Reynolds number
pared to flow between parallel plates. Maximum convective t= time, s
T= temperature, K
heat transfer coefficients are achieved when the flow inside the V= volume, m3
mixing chambers evolves to a self-sustained oscillatory lami- w= mass flow rate
nar flow regime. An important result obtained from these sim- x= entrance length, m
ulations is that, above the critical Reynolds number, the x = dimensionless entrance length
thermal boundary layer is renewed along the network of cham-
bers, enhancing the global heat transfer capacity of NETmix. Greek letters
From this study and from the flow patterns observed in C= gamma function
NETmix, a similar behavior and mass transfer rate enhance- Dt = simulation time step, s
ment is expected. Therefore, NETmix can be a mixing device DT = temperature difference, K
Dx = grid element size, m
particularly suited to handle reactions where fast interfacial j= thermal conductivity, W  m21  K21
mass transfer is required, such as heterogeneous catalytic and j
= spatial average thermal conductivity, W  m21  K21
gas–liquid reactions. l= viscosity, Pa  s
The specific heat transfer capacity of NETmix was com- q= density, kg  m23
pared to other commercially available technologies. It is s= residence time, s

s= stress tensor, Pa
observed that NETmix presents 2–5 orders of magnitude ~t= velocity vector, m  s21
higher specific heat transfer capacity than most of the technol- t = mean velocity, m  s21
ogies used industrially, such as stirred tanks with jackets or x= height/gap distance between plates, m
tubular reactors, and nearly one order of magnitude larger spe- ~=
x vorticity vector, s21
cific heat transfer capacity than microreactors. This NETmix Subscripts
performance is due to its very large surface to volume ratio
b= bulk
together with the heat transfer coefficient enhancement exhib- channel = NETmix channel
ited from the CFD simulations and Nusselt number in = inlet
computations. int = interval
This work has shown that the NETmix mixer/reactor is a loc = local
max = maximum
proficient technology to remove/supply heat from/to a fluid, out = outlet
making it suitable for fast reactions where heat transfer is the PP = parallel plates
kinetically limiting step, and for highly exo/endothermic reac- uc = NETmix unit cell
tions, increasing the overall production capacity of the wall = wall
process. x= x direction
z= z direction

Acknowledgments Abbreviations
This work was financially supported by: Project UID/ CFD = Computational Fluid Dynamics
NUB = NETmix Unit Block
EQU/50020/2013 – POCI-01-0145-FEDER-006984 – Asso- RAM = Random Access Memory
ciate Laboratory LA LSRE-LCM funded by FEDER funds
through COMPETE2020 – Programa Operacional Competiti-
vidade e Internacionalizaç~ao (POCI) – and by national funds
Literature Cited
through FCT – Fundaç~ao para a Ci^encia e Tecnologia. M. F. 1. Szalai ES, Muzzio FJ. Fundamental approach to the design and opti-
Costa and C. M. Fonte gratefully acknowledge their Ph.D. mization of static mixers. AIChE J. 2003;49(11):2687–2699.
2. Thakur RK, Vial C, Nigam KDP, Nauman EB, Djelveh G. Static
scholarship from Fundaç~ao para a Ci^encia e Tecnologia mixers in the process industries—a review. Chem Eng Res Des.
(FCT) SFRH/BD/87817/2012 and SFRH/BD/39040/2007, 2003;81(7):787–826.
respectively. 3. Meijer HEH, Singh MK, Anderson PD. On the performance of static
mixers: a quantitative comparison. Prog Polym Sci. 2012;37(10):
1333–1349.
Notation 4. Rohsenow WM, Hartnett JP, Cho YI. Handbook of Heat Transfer.
A= area, m2 New York, NY, USA: McGraw-Hill, 1998.
Atransf = active area for heat exchange, m2 5. Mokrani A, Castelain C, Peerhossaini H. The effects of chaotic
C0 = configuration flow constant, W  m22  K21 advection on heat transfer. Int J Heat Mass Transfer. 1997;40(13):
cP = specific heat capacity, J  kg21  K21 3089–3104.
d= NETmix channel’s diameter/depth, m 6. Kumar V, Saini S, Sharma M, Nigam KDP. Pressure drop and heat
D= NETmix chamber’s diameter, m transfer study in tube-in-tube helical heat exchanger. Chem Eng Sci.
dh = hydraulic diameter, m 2006;61(13):4403–4416.
~
g= gravitational acceleration vector, m  s22 7. Ghanem A, Lemenand T, Della Valle D, Peerhossaini H. Static
h= convective heat transfer coefficient, W  m22  K21 mixers: mechanisms, applications, and characterization methods – a
h^ = specific heat transfer capacity, W  m23  K21 review. Chem Eng Res Des. 2014;92(2):205–228.
i= inlets 8. Lammers G, Beenackers AACM. Heat transfer and the continuous
j= outlets production of hydroxypropyl starch in a static mixer reactor. Chem
L= distance from centres of two adjacent NETmix chambers, m Eng Sci. 1994;49(24, Part 2):5097–5107.
~
n= unit vector 9. Joshi P, Nigam KDP, Nauman EB. The kenics static mixer: new
Nu = Nusselt number data and proposed correlations. Chem Eng J. 1995;59:265–271.

12 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal
10. Li HZ, Fasol C, Choplin L. Hydrodynamics and heat transfer of rhe- 22. Amon CH, Guzman AM, Morel B. Lagrangian Chaos, Eulerian
ologically complex fluids in a sulzer SMX static mixer. Chem Eng Chaos, and Mixing Enhancement in Converging–Diverging Channel
Sci. 1996;51(10):1947–1955. Flows. Phys Fluids. 1996;8(5):1192–1206.
11. Lopes JCB, Laranjeira PE, Dias MM, Martins AA. Inventors. Net- 23. Dittus FW, Boelter LMK. Heat transfer in automobile radiators of
work Mixer and Related Mixing Process. PCT/IB2005/000647, Feb- the tubular type. Int Commun Heat Mass Transfer. 1985;12(1):3–22.
ruary 2005. European Patent EP172643 B1, October 2008. 2005. €
24. Shibani AA, Ozisik MN. A solution to heat transfer in turbulent flow
R
12. Laranjeira PE, Martins AA, Lopes JCB, Dias MM. NETmixV, a new between parallel plates. Int J Heat Mass Transfer. 1977;20(5):565–573.
type of static mixer: modeling, simulation, macromixing, and micro- 25. Villermaux J. G enie de la R eaction Chimique: Conception et Fonc-
mixing characterization. AIChE J. 2009;55(9):2226–2243. tionnement des R eacteurs. Tec & Doc, 1993.
13. Laranjeira PE, Martins AA, Nunes MI, Lopes JCB, Dias MM. 26. Trambouze P, Euzen JP. Les R eacteurs Chimiques - Conception,
R
NETmixV, a new type of static mixer: experimental characterization Calcul, Mise en Oeuvre. Editions Technip, 1984.
and model validation. AIChE J. 2011;57(4):1020–1032. 27. Kiwi-Minsker L, Renken A. Microstructured reactors for catalytic
14. Gomes PJ, Fonte CP, Santos RJ, Dias MM, Lopes JCB. Experimen- reactions. Catal Today. 2005;110(1-2):2–14.
tal and numerical characterization and quantification of mixing in a 28. Kashid MN, Kiwi-Minsker L. Microstructured reactors for multi-
R
NETmixV reactor. Paper presented at ECCE 7 - 7th European Con- phase reactions: state of the art. Ind Eng Chem Res. 2009;48(14):
gress of Chemical Engineering & CHISA 2010- 19th international 6465–6485.
Congress and Process Engineering, Prague, Czech Republic, Septem- 29. Ferrouillat S, Tochon P, Della Valle D, Peerhossaini H. Open loop
ber 2010. thermal control of exothermal chemical reactions in multifunctional
15. Fonte CM, Leblebici ME, Dias MM, Lopes JCB. The NETmix reac- heat exchangers. Int J Heat Mass Transfer. 2006;49(15):2479–2490.
tor: pressure drop measurements and 3D CFD modeling. Chem Eng 30. Janasek D, Franzke J, Manz A. Scaling and the design of miniatur-
Res Des. 2013;91(11):2250–2258. ized chemical-analysis systems. Nature. 2006;442(7101):374–380.
16. Silva VM, Quadros PA, Laranjeira PE, Dias MM, Lopes JCB. A 31. Tabeling P, Chen S. Introduction to Microfluidics. OUP Oxford, 2005.
novel continuous industrial process for producing hydroxyapatite 32. Sen N, Singh KK, Mukhopadhyay S, Shenoy KT. Comparison of
nanoparticles. J Dispersion Sci Technol. 2008;29(4):542–547. different microreactors for solvent-free, continuous synthesis of
17. Gomes PJ, Silva VM, Quadros PA, Dias MM, Lopes JCB. A highly [EMIM][EtSO4] ionic liquid: an axperimental and CFD study. J Mol
reproducible continuous process for hydroxyapatite nanoparticles Liq. 2016;222:622–631.
synthesis. J Nanosci Nanotechnol. 2009;9(6):3387–3395. 33. Rebrov EV, Schouten JC, de Croon MHJM. Single-phase fluid flow
18. Lide DR. CRC Handbook of Chemistry and Physics: A Ready- distribution and heat transfer in microstructured reactors. Chem Eng
Reference Book of Chemical and Physical Data. Boca Raton, FL, Sci. 2011;66(7):1374–1393.
USA: CRC Press, 2004. 34. Peng XF, Peterson GP, Wang BX. Heat transfer characteristic of
19. Bird RB, Stewart WE, Lightfoot EN. Transport Phenomena, 2nd ed. water flowing through microchannels. Exp Heat Transfer. 1994;7(4):
New York, NY, USA: Wiley, 2002. 265–283.
20. Shah RK, London AL. Laminar Flow Forced Convection in Ducts. 35. Jahnisch K, Hessel V, Lowe H, Baerns M. Chemistry in microstruc-
New York, NY, USA: Academic Press, 1978. tured reactors. Angew Chem Int Ed Engl. 2004;43(4):406–446.
21. Ottino JM. The Kinematics of Mixing: Stretching, Chaos, and Trans-
port. Cambridge University Press, 1989. Manuscript received Oct. 28, 2016, and revision received Mar. 23, 2017.

AIChE Journal 2017 Vol. 00, No. 00 Published on behalf of the AIChE DOI 10.1002/aic 13