G Model

CHROMA-359040; No. of Pages 9 ARTICLE IN PRESS

Journal of Chromatography A, xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma

A study on the onset of turbulent conditions with SFC mobile-phases

Abhijit Tarafder ∗
Waters Corporation, 34 Maple Street, Milford, MA 01757, USA

a r t i c l e i n f o a b s t r a c t

Article history: Following a recent publication [1], the topic of turbulent flow in SFC has generated both interest and
Received 12 September 2017 questions. Liquid-like density, coupled with significantly low viscosity of CO2 -based mobile-phases may
Received in revised form 27 October 2017 result in high Reynolds number (Re) – higher than what represents laminar flow conditions, reaching
Accepted 24 November 2017
the so-called turbulent regions. Although such turbulent flows can form only in the connecting tubings,
Available online xxx
thus not directly affecting the chromatographic process, it is important to know under many situations,
whether the flow inside the tubing is laminar or turbulent.
Keywords:
In this report a comprehensive guideline to identify the possibilities of turbulent flow conditions is
SFC
Turbulent
provided through a series of charts. Flow properties depend on state conditions (composition, pressure
Laminar and temperature) and also on the tubing material and geometry. Here guidelines to detect the onset
Pressure-drop of turbulent conditions is provided for cylindrical stainless-steel tubings of different internal diameters
Extra-column (i.d.) under a wide range of SFC mobile-phase conditions.
Tubing © 2017 Elsevier B.V. All rights reserved.

1. Introduction
1.1. Background
Like in liquid chromatography, the main source of pressure-drop
in an SFC system was considered to be the packed-bed column till
Rajendran et al. [2] demonstrated the important contribution of
pressure-drop in connecting tubings. Later Pauw et al. [1] demon-
strated that tubing pressure-drops are not only non-negligible,
under certain conditions it can be really high – even higher than
that offered by the columns. Beyond certain flow velocity, pressure-
drop rises non-linearly with flowrate increment. Through their
report Pauw et al. demonstrated that this unusual rise in pressure-
drop is linked to the occurrence of turbulent flow inside the
connecting tubings of an SFC system. Although not directly affecting
the chromatography, any loss of usable pressure-drop to the sys-
tem is a concern, especially when the contribution of the system
pressure-drop is higher than that of the column.
1.2. Causes of turbulent flow in SFC
Reynolds number (Re = du
 , where d is the diameter of the con-
duit,  is the fluid density,  is the dynamic viscosity and u is the
flow velocity) of a flow determines whether the flow is laminar
∗ Corresponding author.
E-mail address: abhijit tarafder@waters.com
or turbulent. Flows with a value of Re > 4000 are typically turbu-
lent and flows with Re < 2100 are laminar flow [3]. The region in
between is considered a transition or buffer region where the flow
could be either laminar or turbulent depending on conditions at
the conduit entrance, the surface roughness of the conduit and also
the distance from the entrance where this phenomenon is taking
place. A discussion on the effect of surface roughness on the onset
of turbulence in SFC is available in Ref. [1].
Although it is rare in liquid chromatography except for extreme
conditions (see Ref. [4]), turbulent flow can occur in SFC system
under common method conditions. The main reason behind is the
much lower viscosity of CO2 based mobile-phases, although the
densities remain similar to that of the LC mobile-phases. See in
Fig. 1 that viscosity of pure CO2 , even when its density is as high
as water’s, is almost 3 times smaller than the least viscous solvents
used in LC – e.g. hexane, acetonitrile, etc. At lower densities, vis-
cosity is even lower. Note that the lower the  is, the higher is
the Reynolds number. In addition, because of lower , SFC is typ-
ically run at higher flowrates, which also increases the Re values
because u increases. These two factors together increase the chances
of SFC mobile-phase entering fluidic conditions beyond the lami-
nar limit. The remaining factor that determines whether the flow
will be laminar or not, is the diameter of the conduit (d). Even with
very favorable solvent properties Re inside SFC columns are too
low to be turbulent. Inside connecting tubings, especially in the
low-dispersion tubings that are typically used in ultra-performance
chromatographic systems, however, flow can become turbulent
depending on the mobile-phase flowrate, tubing i.d., mobile-phase

https://doi.org/10.1016/j.chroma.2017.11.056
0021-9673/© 2017 Elsevier B.V. All rights reserved.

Please cite this article in press as: A. Tarafder, A study on the onset of turbulent conditions with SFC mobile-phases, J. Chromatogr. A
(2017), https://doi.org/10.1016/j.chroma.2017.11.056
G Model
CHROMA-359040; No. of Pages 9 ARTICLE IN PRESS
2 A. Tarafder / J. Chromatogr. A xxx (2017) xxx–xxx
Fig. 1. Variation of viscosity with density of neat CO2 (from Ref. [9]). More than one
viscosity values at certain density points are due to the influence of temperature [9].
At lower densities, CO2 viscosity can be as low as 0.02 cP, which is 50 times lower
than that of water. Even when CO2 density is equal to that of water its viscosity is
0.122 cP [8], which is about three times lesser than the least viscous solvents of LC,
e.g. acetonitrile (=0.34 cP at 27 ◦ C), hexane (=0.3 cP at 25 ◦ C), etc. [10].
temperature and pressure, co-solvent used in the mobile-phase and
co-solvent percentage.
1.3. Objective of the current report
In practical situation, it is challenging to identify experi-
mental conditions that maximize experimental objective(s) (e.g.
fast analysis) while avoiding turbulent flow – causing excessive
pressure-drop in the tubings. For CO2 -based solvents the contour
of physical property variation and hence identification of the onset
of turbulent conditions, is difficult to describe with words or by
presenting tabular set of data. Charts, on the other hand, depicts
a global picture of the scenario, which is very important to realize
and appreciate the true possibilities of SFC methods.
The objective of the current report is to provide a global pic-
ture of the conditions that may lead to the onset of turbulence in
SFC system tubing. A series of charts is provided here in that direc-
tion. Such charts should be helpful to diagnose, conveniently and
instantly, whether an unexpected pressure rise in an SFC system
is caused by a turbulent flow or not. The charts may also provide
insights regarding the trade-offs in selecting method conditions for
better separation while avoiding excessive pressure-drop in tubing.
2. Theory
This section is divided in two parts. The first part describes fac-
tors that may lead to non-linear variation of pressure-drop across
a system. It will be shown here that turbulent flow is not the only
phenomenon that can lead to non-linear variations of pressure-
drop with flowrate in SFC. The second part describes construction
of the turbulent-flow-identification charts.
2.1. Non-linear variation of pressure-drop with flowrate
In related literature (see Refs. [1,5]), the criterion of a deviation
from linear increment of pressure-drop as a function of increasing
flowrate has been considered as an indicator of transitioning to
turbulent region. While this indeed is the traditional (see Ref. [3])
and also a convenient experimental procedure to detect onset of
turbulent conditions and is employed in this report as well, it is
important to understand if there are any other factor(s) that may
lead to non-linear increment of pressure-drop in a flow system.
Please cite this article in press as: A. Tarafder, A study on the onset of turbulent conditions with SFC mobile-phases, J. Chromatogr. A
(2017), https://doi.org/10.1016/j.chroma.2017.11.056
The reason behind the linear increment of pressure-drop with
flowrate can be understood from the Hagen–Poiseuille equation
[3]:
P = Ct · Q ·  (1)
where P is the pressure-drop, Ct is a constant that characterize
the conduit geometry (= 128L , where L is length and d is conduit
d4
i.d.), Q is the volumetric flowrate and  is the dynamic viscosity.
Note from Eq. (1), which is valid for laminar flow of a New-
tonian fluid, that with liquid (incompressible) mobile-phase, P
shall increase linearly with Q for an isocratic and isothermal system
because  or any other physical properties will remain constant. At
high enough Q, when the flow enters turbulent regime, there is a
rapid increase in pressure-drop – approximately as the square of
the velocity [3]. Eq. (1) is no longer valid in the turbulent region.
For SFC mobile-phases, which are more compressible than liq-
uid, there will be non-negligible variation of physical properties
with pressure even when high concentration of co-solvents are
present. This indicates, in SFC P will increase non-linearly even
under laminar conditions. The extent of this non-linearity, however,
will depend on the state conditions and the extent of pressure-drop.
For example, neat CO2 , which is the most compressible among SFC
mobile-phases, will demonstrate highest non-linearity. Increasing
co-solvent percent, on the other hand, will decrease the non-
linearity. A brief explanation of this issue is presented below.
The first point is, to apply Eq. (1) in SFC we need modifica-
tions. The Q delivered by the SFC pump(s) do not remain constant
throughout the system – it varies depending on local fluid density.
On the other hand, the mass flowrate(s) delivered by the pump(s)
remain constant. Hence, for SFC Eq. (1) can be re-written as:
P = Ct · G ·  (2)
where G is the mass flowrate anywhere in the conduit and  is the
average kinematic viscosity.
The second point is, when SFC pumps deliver a set volumetric
flowrate (Q) independent to the pump-outlet pressure (constant-
flow mode), which is normally done by most pumps of commercial
SFC systems, G increases non-linearly with Q. Note that for liquid
solvents  remains constant, hence mass flowrate varies with the
same multiples as the volumetric flowrate (G = Q). For CO2 (com-
pressible), however, increasing Q leads to increased pump-outlet
pressure (caused by increased pressure-drop in the system) and
hence increased CO2 density at pump-outlet (CO2 ,PO ) and vice-
versa. So mass flowrate variation depends also on how CO2 ,PO
varies with Q (see Ref. [6] for more details). If CO2 ,PO variation
is considerable enough, variation of Q leads to non-linear variation
of G. Similarly, with rising average pressure of the system caused
by increased Q,  will also vary. Variation of both factors with Q
result in non-linear variation of P with increasing Q (see Eq. (2)).
To emphasize again, the extent of this variation will depend on how
CO2 ,PO and  vary during particular experimental runs.
Fig. 2 demonstrates the extent of possible variation of CO2 ,PO at
two different pump-head temperatures and variation of  of neat
CO2 as function of temperature and pressure. Temperatures 4 and
13 ◦ C represent settings of two different commercial instruments.
Although both figures (Fig. 2a and b) demonstrate considerable
variation of CO2 ,PO and  with pressure, their net effect on P varia-
tion in SFC may not be significant. The main reason being – to impart
a considerable variation in CO2 ,PO and  during an experimental
run, the pressure-drop has to be very high. For example, at 13 ◦ C,
to increase CO2 ,PO from 0.9 to 1.0 g/mL (i.e. by 11%) P has to be
increased from 10 to 200 bar (i.e. by 2000%). Similar observation
can be made regarding the variation of  too.
A detailed discussion on this topic is outside the scope of cur-
rent report. However, to provide an approximate idea regarding the
extent of possible non-linearity imparted by varying CO2 proper-
G Model
CHROMA-359040; No. of Pages 9 ARTICLE IN PRESS
A. Tarafder / J. Chromatogr. A xxx (2017) xxx–xxx
Fig. 2. (a) Variation of neat CO2 density at the pump-outlet as a function of outlet pressure, at pump-head temperatures 4 and 13 ◦ C. (b) Variation of kinematic viscosity and
density with pressure and temperature (from Ref. [7]). Solid curves in (b) represent constant kinematic viscosity () in Stokes or cm2 /s ×104 and dashed curves represent
constant density in g/mL. To provide an idea on the relationship between the  values plotted here and the dynamic viscosity plotted in Fig. 1, refer to the point marked by
a star on the plot. The point represents  = 9.1 Stokes ×104 at pressure = 165 bar and temperature = 37 ◦ C. The density and the dynamic viscosity at the same conditions are
0.82 g/mL and 0.075 cP, respectively. See Ref. [7] for more details.
Table 1
Column specification to demonstrate role of physical property variations in SFC in
imparting non-linear variation of pressure-drop with flowrate.
Column length (cm) 15
Column diameter (cm) 0.46
Particle size (␮m) 1.7
Bed porosity (–) 0.4
ties, three example situations are presented here. At three different
temperatures, P vs. Q was calculated for a column, employing
the simulation program reported in Ref. [7]. A column instead of
a tubing was considered here to ensure laminar condition even at
high flowrates. Note that the nature of property variation depends
only on the net pressure-drop, not the means to achieve it. For
each specified Q, G was calculated iteratively based on the resul-
tant pump-outlet pressure imparted by the increment in Q. To
calculate column pressure-drop, it was assumed that temperature
remained constant at the set temperature throughout the column.
Note that temperature may vary along the column in SFC depend-
ing on the mobile-phase composition, temperature and pressure.
State properties were calculated as functions of pressure only along
the column length.
Neat CO2 was considered for this study to maximize the effect of
compressibility on pressure-drop. Temperatures selected were 30,
40 and 60 ◦ C. Column outlet pressure was fixed at 100 bar. Column
specifications (see Table 1) were chosen to impart high pressure
drop across the column. Fig. 3 plots the calculated pressure-drop
vs. flowrate data employing the conditions described above. Note
that all the conditions chosen for this example were to cause high
variation of CO2 ,PO and of  along the column so that the effect of
non-linear pressure-drop variation can be clearly demonstrated.
Following observations can be made from Fig. 3:
(1) P starts varying non-linearly under the influence of vary-
ing physical properties, only when the pressure-drop is high
enough – e.g. more than 100 bar.
(2) For lower pressure-drop ranges (below 100 bar), its variation
with flowrate is approximately linear.
(3) For packed-bed, flow may transition to turbulent conditions at
Re > 10 (see Ref. [3]), but for all conditions presented here Re is
below 1.0 (see Fig. 3) – so the flow was laminar.
Fig. 3 demonstrates that non-linear variations in pressure-drop
with varying volumetric flowrate does not automatically mean
Please cite this article in press as: A. Tarafder, A study on the onset of turbulent conditions with SFC mobile-phases, J. Chromatogr. A
(2017), https://doi.org/10.1016/j.chroma.2017.11.056
3
transitioning to turbulent regime. With increasing flowrate, if the
resultant pressure-drop is high enough, property changes of CO2
based solvents can lead to P increase non-linearly, even under
laminar flow conditions.
2.2. Construction of turbulent flow identification chart
Construction of a simple chart to guide a practitioner in iden-
tifying turbulent flow conditions can be challenging because of
the presence of multiple factors that can affect flow pattern. Re
is a function of five different variables – the state conditions
(composition–pressure–temperature), the flowrate and the con-
duit diameter. And there can be ten different ways the value of Re
can be presented as a chart. The challenge is to identify the format
or formats that can be more useful and meaningful in a practical
situation.
Among possible alternatives the format presented in Fig. 4, i.e.
co-solvent composition (v/v, %) vs. volumetric flowrate (mL/min)
was adopted to present pictorial guidelines in identifying the onset
of turbulent conditions in SFC. Both factors are typical variables in
analytical method design and may lead to pressure rise in an SFC
system. In an event of unexpected rise in system pressure a prac-
titioner can refer to these charts to check if this is due to turbulent
condition or not. And if turbulent, what can be the optimal way
to avoid the condition. The general method of construction of the
laminar-turbulent boundary in this chart (see Fig. 4) and similar
charts presented later in this report, is described below.
2.2.1. Description of main equations
To construct a chart, e.g. that in Fig. 4, volumetric flowrates from
a commercial SFC pump (see Section 3 for more details) that will
result in a predefined Reynolds number (e.g. Re = 2100 for Fig. 4)
were calculated with various compositions of the mobile phase, at
fixed temperatures and pressures, and also with fixed tubing i.d.,
following Eq. (3):
G
Q = (3)
vCO2 · CO2 ,PO + vCoS · CoS,PO
where G is expressed as:
dRe
G= (4)
4
In Eq. (3), vCO2 and vCoS are the volume fractions of CO2 and
co-solvent respectively. The value of Q was calculated as above in
recognition to the fact that the Q delivered by pumps are not the
G Model
CHROMA-359040; No. of Pages 9 ARTICLE IN PRESS
4 A. Tarafder / J. Chromatogr. A xxx (2017) xxx–xxx

Fig. 3. Variation of P and Re with volumetric flowrate at three different temperatures. The figure demonstrate that for all temperatures P increase non-linearly with
flowrates even under laminar conditions, provided the P differences are high enough to impart non-linear variations of density and viscosity. The dashed arrows presented
in the plots represent projected P with increasing flowrates based on the first two P points of the respective plots.

Fig. 4. Chart depicting laminar and turbulent flow regions on a methanol composition (v/v, %) vs. flowrate (mL/min) plane at 30 ◦ C, between 172 and 200 bar. Experimental
results demonstrating the onset of turbulent conditions at different flowrates under different methanol compositions are also presented. Note that the points on the chart
corresponds only to the circled flowrate in the pressure-drop vs. flowrate plots presented in the insets. The dashed arrows presented in the inset plots represent projected
P with increasing flowrates based on the first two P points of the respective plots.

same as the Q experienced by the tubing or the column because
of the difference in respective state conditions. The G, however,
remains constant everywhere. Note that Eq.(4) is a rearranged form
of the Reynolds equation and Eq. (3) calculates the required vol-
umetric flowrate set by a commercial instrument to generate a
specified G. CO2 ,PO and CoS,PO can be calculated from the pump-
head temperature and pressure of the respective pumps.
2.2.2. Construction of the chart in Fig. 4
To construct the turbulent regime’s boundary in Fig. 4, volumet-
ric flowrates were calculated from Eq. (3) to generate Re = 2100 at
various compositions of the mobile phase. To calculate the vari-
able G for Eq. (3), viscosities were estimated at 30 ◦ C and 172 bar at
various mobile-phase compositions. An additional boundary was
plotted with 30 ◦ C and 200 bar, with all the other conditions keep-
ing the same, for the reason described later in Section 4.1. Tubing
i.d. was fixed at 178 ␮m (0.007 in.).
To calculate solvent density at the pump-heads (CO2 ,PO and
CoS,PO ), the pressure assumed was 621 bar, the temperature of CO2
pump-head as 13 ◦ C (temperature of an Acquity UPC2 CO2 pump-
head) and that of the co-solvent (methanol) pump as 23 ◦ C, which
is the room temperature. 621 bar, which is the highest allowable
pressure in a commercial SFC instrument, was assumed to over-
estimate mass flowrate, which is better than to underestimate in
constructing such charts. Based on the above pressure-temperature
conditions, densities of CO2 and methanol were estimated to be
1.095 and 0.836 g/mL, respectively (from Ref. [8]), at the respective
pump-outlets. Mass compositions of the mobile-phase were calcu-
lated from volumetric compositions based on these two densities.
The errors made in the estimation of CO2 density for pressures up
to 300 bar and temperatures up to 250 ◦ C, range between 0.03% and
0.05% [8].  in Eq. (4) was calculated from REFPROP ([8]) at 30 ◦ C
and at 172 and 200 bar, for all mass compositions.
3. Experimental setup and methods
To verify the laminar boundary plotted in the charts described in
this report a set of experimental studies were carried out (see Fig. 5
for a schematic diagram of the experimental setup) to measure
pressure-drop across a single tubing. The Binary Solvent Man-
ager (UPC2 -BSM) and the Convergence Manager (UPC2 -CCM) of an
Acquity UPC2 was employed to pump CO2 and methanol mixture
and to maintain the pressure inside the tubing, respectively. CO2
mass flowrate was monitored with a Bronkhorst Cori Flow Mass-

Please cite this article in press as: A. Tarafder, A study on the onset of turbulent conditions with SFC mobile-phases, J. Chromatogr. A
(2017), https://doi.org/10.1016/j.chroma.2017.11.056
G Model
CHROMA-359040; No. of Pages 9 ARTICLE IN PRESS
A. Tarafder / J. Chromatogr. A xxx (2017) xxx–xxx
Fig. 5. Schematic diagram of experimental setup described in Section 3. (1) Aqcuity UPC2 BSM to deliver CO2 and methanol streams, (2) Cori-flow mass flowmeter to measure
CO2 mass flowrate, (3) mixer, (4) active pre-heater (APH), (5) tee with pressure-transducer, (6) main system having an empty column (3 × 50 mm) and 100 in. of SS tubing
(0.007 in. i.d.), (7) tee with pressure-transducer, and (8) Aqcuity UPC2 CCM for end pressure regulation.
flow Meter (M13-ABD-99-0-S). No flow meter was installed in the
methanol pump – the flow accuracy was verified at the beginning
of the experiments. A standard 250 ␮L mixer was used to pre-
pare the mixed mobile-phase. The mixture flow was then passed
in sequence through an active-Pre-Heater (APH) of an Acquity
Column Manager (UPC2 -CM) which was installed with an exten-
sion cord to pre-heat the mixed solvent, a high accuracy pressure
transducer (PXM409-350BGV) from Omega Engineering to monitor
tubing inlet pressure, a stand-alone column oven (Systec Met-
alox model 200-C) where an empty stainless-steel (SS) column
(3 mm × 50 mm) and a 100 in. long SS tubing of 0.007 in. i.d was
placed. The empty column was placed before the tubing to increase
residence time of solvent inside the system to ensure that the sol-
vent has reached the set oven temperature. After the oven, another
pressure-transducer was installed to monitor tubing outlet pres-
sure and at the end an Automated Back Pressure Regulator (ABPR),
which is an UPC2 -CCM, to control set pressure inside the tubing.
4. Results and discussion
In SFC, onset of turbulent conditions can be identified with an
increase in pressure-drop that deviates from the expected linear
path, provided the difference in pressure-drops are not high enough
to create non-linearity as a result of property changes (see Section
2.1). In this section first we will present experimental verification
of the chart presented in Fig. 4. And then, based on a set of charts
depicting the influence of pressure, temperature and tubing i.d.,
discuss the effects of these parameters on the onset of turbulence
in SFC.
4.1. Experimental verification of turbulent flow chart
To experimentally verify the boundary depicting the onset of
turbulent conditions in Fig. 4, experiments at 30 ◦ C were carried out
with the setup described in Section 3 at different methanol com-
positions – 0, 15 and 30 (v/v, %) methanol in CO22 . The ABPR was
set at 172 bar and the pump-outlet pressures went up to 200 bar.
Fig. 4 shows that the experimental points were close enough to
the boundary predicted for 30 ◦ C at 172 and 200 bar. Experimen-
tal data, plotting pressure-drop variation as a function of flowrates,
corresponding to these experimental points, are also presented in
Fig 4 as insets. For all the experiments, onset of turbulent condi-
tion was determined by the sudden rise in P at certain flowrate.
It can be noted that the nature of increment in pressure-drop
Please cite this article in press as: A. Tarafder, A study on the onset of turbulent conditions with SFC mobile-phases, J. Chromatogr. A
(2017), https://doi.org/10.1016/j.chroma.2017.11.056
5
demonstrated by the P vs. flowrate insets in Fig. 4, is very differ-
ent from the variations caused by property changes demonstrated
in Fig. 3. In Fig. 3 P variation with flowrate deviated gradually
from the linear-increment path. In Fig. 4, on the other hand, P
deviations from linear path is more abrupt and pronounced. Addi-
tionally, for all cases, such deviations from linear path occurred
at low pressure-drop conditions. This indicates that the deviations
here were triggered mainly by the onset of turbulent condition.
To verify the predictive capability of the chart in other method
conditions too, experiments with two more temperatures were car-
ried out – at 60 and 23 ◦ C (see Fig. 6). Temperature 60 ◦ C was chosen
to represent a high temperature condition used in SFC. Temperature
23 ◦ C, on the other hand, was chosen to represent room tempera-
ture. Fig. 7 presents experimental pressure-drop vs. flowrate data
corresponding to the experimental points shown in Fig. 6. Note that
even for 23 and 60 ◦ C, experimental results were close to the bound-
ary predicted by the calculations presented here. Errors are not
more than 5%, which can accepted as reasonably accurate, given the
total number of factors that can affect the onset of turbulence. Over-
all, experimental results demonstrate that the charts presented
here can provide a reasonable estimate of the onset of turbulent
condition in SFC, which can be expected to be sufficient in practical
conditions.
4.2. Effect of method conditions on onset of turbulent flow
A general description on the effect of temperature and pres-
sure in setting a turbulent flow is presented in this section. In an
SFC system, although the exact pressures can be difficult to esti-
mate without installing additional pressure-transducers, pressure
should progressively decrease along the tubing with values bound
between the pump-outlet pressure and the ABPR pressure. Regard-
ing temperatures, however, there can be different zones where
mobile-phase temperatures have different values inside the tub-
ing. For example, inside the part of the tubing immediately after
the CO2 pump, mobile-phase temperature will be the same as the
CO2 pump-head temperature, which is 13 ◦ C for UPC2 . The fluid
should quickly equilibrate with the ambient condition and by the
time mobile-phase exits solvent mixer, temperature will be very
close to that of the ambient. Temperature changes again inside
any pre-heater that is generally installed before the column to pre-
heat incoming mobile-phase to the set column-oven temperature.
Post-column, temperature in tubing equilibrate with the ambient
at different rates depending on the exposure of the tubing out-
G Model
CHROMA-359040; No. of Pages 9 ARTICLE IN PRESS
6 A. Tarafder / J. Chromatogr. A xxx (2017) xxx–xxx

Fig. 6. Experimental verification of turbulent flow boundaries at temperatures 23, 30 and 60 ◦ C, between pressures 172 and 200 bar. Results of 30 ◦ C are the same as presented
in Fig. 4. Overall, the deviation of experimental results are within 5% of the values predicted by the boundaries presented here.

Fig. 7. P vs. flowrate plots corresponding to the experimental points shown in Fig 6, for 60 and 23 ◦ C. Similar plots for 30 ◦ C is presented as insets in Fig. 4. The dashed
arrows presented in the plots represent projected P with increasing flowrates based on the first two P points of the respective plots.

side their enclosed modules. Inside UV/vis detectors, temperature
may slightly rise due to heat emission from the deuterium lamp.
Although detector flow-cell diameters are typically large enough
to not experience any turbulence under standard method condi-
tions, the tubings inside a detector that lead the flow in and out
of the flow-cell, may experience turbulent flow. In summary, pres-
sure inside an SFC system tubing will continuously decrease – the
lowest being close to the ABPR pressure. Temperature on the other
hand will be different at different zones.
4.3. Effect of temperature and pressure
Temperature and pressure has opposite effects on the onset of
turbulent condition. Increasing temperature expands the boundary
of turbulent region towards lower flowrates for all mobile-phase
compositions (see Fig 8 ). Increasing pressure, on the other hand,
contracts the boundary (see Fig 9 ).
To study the effect of temperature variation only, the pressure
was fixed at 138 bar. This particular pressure value was chosen to

Please cite this article in press as: A. Tarafder, A study on the onset of turbulent conditions with SFC mobile-phases, J. Chromatogr. A
(2017), https://doi.org/10.1016/j.chroma.2017.11.056
G Model
CHROMA-359040; No. of Pages 9 ARTICLE IN PRESS
A. Tarafder / J. Chromatogr. A xxx (2017) xxx–xxx 7

Fig. 8. Progression of turbulent flow region with temperature – at 138 bar (2000 psi) inside a 0.007 inch (178 ␮m) tubing. With increasing temperature, the boundary of
turbulent region expands towards lower flowrates. Which signifies onset of turbulent flow at increasingly lower flowrates with increasing temperature. Approximately, for
all co-solvent compositions, it will need approximately 0.2 mL/min lesser flowrate to reach turbulent flow, with each 10 ◦ C rise in temperature.

Fig. 9. Progression of turbulent flow region with pressure – at 40 ◦ C inside a 0.007 in. (178 ␮m) tubing. With increasing pressure, the boundary of turbulent region contracts
towards higher flowrates. Which signifies onset of turbulent flow at increasingly higher flowrates with increasing pressure. Approximately, for all co-solvent compositions,
it will need approximately 0.16 mL/min higher flowrate to reach turbulent flow, with each 1000 psi (69 bar) rise in pressure.

represent a common ABPR pressure setting employed during SFC
method development. From Fig 8 we can note that at low tem-
peratures, e.g. 20 ◦ C, flow can be turbulent inside a 178 ␮m i.d.
tubing at 1.5 mL/min with neat CO2 when the pressure is 138 bar.
As the temperature increases, e.g. at 30 ◦ C, onset of turbulence can
occur at even lesser flowrates – 1.25 mL/min in this case. At the
same temperature, turbulence onsets at 1.65 and 2 mL/min, respec-
tively, if the mobile-phase has 10 and 20 (v/v, %) of methanol in it.
With increasing percent of methanol, the flowrate boundary of the
onset of turbulent condition increased almost logarithmically. If the
motivation is to decrease the analysis time – it can either be done
through increasing flowrate and/or increasing co-solvent concen-
tration. If the flowrate is already at the verge of being turbulent or is
already turbulent, it can be a better strategy to increase co-solvent
concentration rather than increasing the flowrate, assuming selec-
tivities do not change noticeably.
At high temperatures, e.g. 70 ◦ C (Fig 8), flow can turn turbulent
inside a 178 ␮m i.d. tubing at 0.5 mL/min with neat CO2 . Even when
15 (v/v, %) of methanol is added, the flow is turbulent at 1 mL/min.
Note that at high temperatures solvent viscosity reduces – leading
to lesser pressure-drop across the column but potentially leading to
higher pressure-drop across the tubing if the flow enters turbulent
regime.
To study the effect of pressure variation only, the temperature
was fixed at 40 ◦ C, which is a temperature commonly employed
during SFC method development. From Fig 9 we can see that at
higher pressures possibility of turbulent flow reduces. The impact
of varying pressure by 1000 psi is almost equivalent to varying

Please cite this article in press as: A. Tarafder, A study on the onset of turbulent conditions with SFC mobile-phases, J. Chromatogr. A
(2017), https://doi.org/10.1016/j.chroma.2017.11.056
G Model
CHROMA-359040; No. of Pages 9 ARTICLE IN PRESS
8 A. Tarafder / J. Chromatogr. A xxx (2017) xxx–xxx
Fig. 10. Progression of turbulent flow region with varying tubing i.d. – at 40 dgrC and 2000 psi (138 bar). The i.d. values shown in the legend are in ␮m. With increasing i.d.,
the boundary of turbulent region contracts towards higher flowrates. Which signifies onset of turbulent flow at increasingly higher flowrates with increasing tubing i.d.
temperature by 10 ◦ C. A summary of observations, which can be
employed as a guideline to understand the onset of turbulent con-
ditions in SFC tubing, is presented below:
1. Mobile-phase flowrate leading to onset of turbulent condition
increase logarithmically with increasing percent of methanol in
mobile-phase.
2. With neat CO2 flowrates as low as 0.5 mL/min can lead to turbu-
lent flow inside a standard tubing of 178 ␮m i.d. at sufficiently
high temperatures and low pressures.
3. Turbulent flow can happen even with high percent of co-solvent.
E.g. a solvent with 30% methanol can enter turbulent regime
beyond the flowrate of 2.5 mL/min.
4. With all mobile-phase compositions, an increase in 10 ◦ C results
in an expansion of the turbulent flow boundary by approxi-
mately 0.2 mL/min towards lower flowrate.
5. Similarly, with all mobile-phase composition, an increase in
1000 psi results in a contraction of the turbulent flow boundary
by approximately 0.16 mL/min towards higher flowrate.
4.4. Effect of tubing diameter
Internal diameter of connecting tubing can vary from system to
system – and within a system, between different parts. In UPC2 the
tubing i.d. representing the maximum length of tubing is 0.007 in.
or 178 ␮m. Tubing i.d. before the injector is 0.01 in. (254 ␮m).
Decreasing tubing diameter increases flow velocity (assuming no
change in mass-flow) and hence the Re (see Eq. (4)). Which means,
with decreasing tubing i.d. there will be onset of turbulent condi-
tions at lower flowrates. Fig. 10 demonstrates how varying tubing
diameter affects the turbulent zone, assuming the same surface
roughness of the tubing. For example, with 127 ␮m (0.005 in.) tub-
ing i.d. flow can be turbulent with 15 (v/v, %) methanol even at
1.5 mL/min. This highlights the challenges of designing instrument
with ultra-low system volume so that it is more compatible with
sub-2 ␮m particles. With increasing co-solvent concentration the
effect of tubing i.d. in determining the onset of turbulence becomes
more prominent. For example, with neat CO2 increasing tubing
i.d. from 127 to 178 ␮m will push the turbulence onset limit by
0.25 mL/min. With 30 (v/v, %) methanol, on the other hand, the
limit will be pushed by 0.7 mL/min.
Please cite this article in press as: A. Tarafder, A study on the onset of turbulent conditions with SFC mobile-phases, J. Chromatogr. A
(2017), https://doi.org/10.1016/j.chroma.2017.11.056
5. Conclusion
A series of charts that demarcate the onset of turbulent condi-
tions with an SFC mobile-phase is proposed here. Based on these
charts a general guideline to identify the possibilities of having
turbulent flow in an SFC system is provided. It can be noted that
with 178 ␮m (0.007 in.) i.d. tubing one can reach turbulent condi-
tions and hence experience enhanced extra-column pressure-drop
at flowrates as low as 1.5 mL/min and beyond, when neat CO2 is
employed. With increasing modifier concentration the turbulent
zone gradually moves towards higher flowrates. It is demonstrated
here that even with high concentration of co-solvent (e.g. 30%),
flow inside tubing can be turbulent within the permissible flowrate
provided by commercial instruments. With increasing temperature
the turbulent boundary expands towards lower flowrates, whereas
with increasing pressure it recedes towards higher flowrate. These
charts can be employed to reduce unwanted pressure-drop in sys-
tem tubing during the development of an experimental method.
5.1. Cautions in using the charts
The charts provided here supply approximate guidance to iden-
tify turbulent conditions in SFC. While using these charts, however,
the following points should be remembered:
1. Estimation of turbulent region worked satisfactorily with ran-
domly selected experimental points, with the system described
in Section 3. However, that does not necessarily mean that
Re = 2100 can be always considered as the boundary of turbulent
region. Although Re = 2100 is considered as the starting condi-
tion of turbulence, onset of turbulence in tubings is controlled by
several other factors which are difficult to quantify. This is why
there is a wide transient region (2100 < Re < 4000) within which
turbulence can set in, depending on other supporting factors.
2. The data generated by REFPROP are the state-of-the-art;
however, they are estimations based on interpolation of exper-
imental data and there can be glitches. For example while
calculating the mixed-phase viscosity for all the CO2 + methanol
compositions, REFPROP generated unexpected data for certain
conditions. To generate these plots those data were deleted and
G Model
CHROMA-359040; No. of Pages 9 ARTICLE IN PRESS
A. Tarafder / J. Chromatogr. A xxx (2017) xxx–xxx
filled in by interpolated data. An example is provided in the
supplementary document.
3. In a commercial SFC system, along most of the tubing length,
mobile-phase can be considered to be experiencing room tem-
perature. Although laminar flow boundaries of a variety of
temperatures are provided here, the ones close to the ambient
temperature are more important in estimating pressure-drops
inside a commercial system.
Acknowledgements
Jason Hill for experimental ideas.
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at https://doi.org/10.1016/j.chroma.2017.11.
056.
References
[1] R.D. Pauw, K. Choikhet, G. Desmet, K. Broeckhoven, Occurrence of turbulent
flow conditions in supercritical fluid chromatography, J. Chromatogr. A 1361
(2014) 277–285.
Please cite this article in press as: A. Tarafder, A study on the onset of turbulent conditions with SFC mobile-phases, J. Chromatogr. A
(2017), https://doi.org/10.1016/j.chroma.2017.11.056
9
[2] A. Rajendran, O. Kräuchi, M. Mazzotti, M. Morbidelli, Effect of pressure drop
on solute retention and column efficiency in supercritical fluid
chromatography, J. Chromatogr. A 1092 (2005) 149–160.
[3] W.L. McCabe, J.C. Smith, P. Harriot, Unit Operations of Chemical Engineering,
McGraw-Hill Inc, Singapore, 1993.
[4] J. Halvorson, A.M. Lenhoff, M. Dittmann, D.R. Stoll, Implications of turbulent
flow in connecting capillaries used in high performance liquid
chromatography, J. Chromatogr. A (2017), http://dx.doi.org/10.1016/j.chroma.
2016.12.084.
[5] T.A. Berger, Characterizing pressure issues due to turbulent flow in tubing, in
ultra-fast chiral supercritical fluid chromatography at up to 580 bar, J.
Chromatogr. A 1475 (2016) 86–94.
[6] A. Tarafder, P. Vajda, G. Guiochon, Accurate on-line mass flow measurements
in supercritical fluid chromatography, J. Chromatogr. A 1320 (2013) 130–137.
[7] A. Tarafder, K. Kaczmarski, M. Ranger, D.P. Poe, G. Guiochon, Use of the
isopycnic plots in designing operations of supercritical fluid chromatography.
IV. Pressure and density drops along columns, J. Chromatogr. A 1238 (2012)
132–145.
[8] E.W. Lemmon, M. Huber, M.O. McLinden, D.G. Friend, National Institute of
Standards and Technology, Standard Reference Data Program, Gaithersburg
8.0, 2013.
[9] A. Tarafder, G. Guiochon, Use of isopycnic plots in designing operations of
supercritical fluid chromatography. I. The critical role of density in
determining the characteristics of the mobile phase in supercritical fluid
chromatography, J. Chromatogr. A 1218 (2011) 4569–4575.
[10] Dortmund Data Bank, 2017. www.ddbst.com.