-
Transport properties of fluorinated surfactants:viscosity and
diffusion of mixtures involving
fluorinated alcohols
João de Oliveira Mateus Afonso
Instituto Superior Técnico, Lisboa, PortugalUniversity of
Vanderbilt, Nashville, United States of America
November 2018
Abstract: This work will focus on studying thermodynamic
properties of fluorotelomeralcohols, which are linear highly
fluorinated molecules with a terminal OH group. Re-cently,
experimental densities and viscosities of (CF3(CF2)3(CH2)OH +
Butanol) and(CF3(CF2)5(CH2)OH + Hexanol) were reported and revealed
large positive excessvolumes and large negative excess
viscosities.This work consists on a computational methods namely,
Molecular Dynamics simula-tions focused on studying the significant
deviations to ideal behaviour of the transportproperties of this
mixtures and the possible existence of nano structures. It was
con-cluded that this behaviour results from the unfavourable
dispersion forces between thehydrogenated and fluorinated
chains.Molecular dynamics simulations were also used to calculate
densities and excess vol-umes for three different systems, namely,
(BuOH + HFB), (HexOH + UFH) and, (DOH+ PFO). Also, diffusion
coefficients were calculated for the (BuOH + HFB) mixture andthe
(HexOH + UFH) mixture.Experimental data for the viscosity of
2,2,3,3,4,4,5,5,6,6,7,7,7-tridecafluoroheptan-1-olas a function of
temperature between 283.15 K and 353.15 K was also
reported.Keywords: Viscosity, Molecular Dynamics, Fluorotelomer
Alcohols, Segregation
1. Introduction
Fluorinated surfactants are fluorocar-bon based surfactants that
match thetypical hydrophilic/hydrophobic surfactantbehaviour to
lyophobic behaviour to-wards hydrogenated organic medium.Inside
this class of surfactants, fluo-rotelomer alcohols, are linear
highly flu-
orinated molecules with a terminal OHgroup, described by the
general formula,CF3(CF2)n(CH2)mOH (n+1:m FTOH).
It is well known that mixtures of hy-drogenated and fluorinated
chains displaysignificant non-ideal behaviour, reflectedon a
tendency for phase separation, largepositive deviations from
Raoult’s law, and,
1
-
large positive excess properties, due toweak dispersion forces.
However, the ad-dition of a polar alcoholic group to boththese
mutually phobic segments providesthe existence of associative
interactionsbetween these molecules. As a result,this mixtures have
been reported to createan O · · ·HO network of hydrogen bondsand,
simultaneously, segregate into hy-drogenated and fluorinated
domains toachieve the best packing possible for thephobic
segments.[8][9]
Mixtures of alcohols and fluorotelomeralcohols have also been
shown to presentlarge non-ideal behaviour but the role pos-sible
molecular structural arrangementsand molecular interactions has on
the ther-modynamic properties of such mixtureshas not been
thoroughly analysed. In ad-dition, experimental data of this
mixturesis rather scarce.
Recently, experimental densitiesand viscosities of
(2,2,3,3,4,4,4-heptafluorobutan-1-ol + Butanol)
and(2,2,3,3,4,4,5,5,6,6,6-undecafluorohexan-1-ol + Hexanol) were
reported which re-vealed large positive excess volumes andlarge
negative excess viscosities[9][4]. Inthis work, molecular dynamics
simulationswere performed, mainly, to understand thereason beyond
the significant deviationsto ideal behaviour of the transport
proper-ties of this mixtures. Nonetheless, otherproperties such as
densities and excessvolumes were calculated and comparedwith
literature data for three different sys-tems, namely, (BuOH + HFB),
(HexOH +UFH) and, (DOH + PFO).
2. Computational Chemistry andMolecular Simulations
2.1. Molecular Dynamics Simulations
Molecular simulations provide a betterunderstanding experimental
observationsand allow the identification of underlyingmechanisms by
providing a insight of thesystem at a molecular level.
In Molecular Dynamics simulations, asimulation engine
numerically integratesNewtons laws of motion to move
particlesthrough time. This allows the observationof the system
evolution during time andthe calculation of its transport
properties.Therefore, MD simulations were the cho-sen method to
conduct this computationalstudy over Monte Carlo simulations.
2.2. Force Field - OPLS
A force field is a mathematical expres-sion chosen to describe
the intra- andinter-molecular potential energy of a col-lection of
atoms, and the correspondingparameters that will determine the
energyof a given ensemble.[2]
A force field requires the definition of in-termolecular
interactions, non-bonded pa-rameters, and intra-molecular
interactions,bonded parameters, existent in a system.
The intra-molecular or local contribu-tions to the total energy
include bondstretching, angle bending and dihedral tor-sions.
U =∑bonds
1
2kb(r − r0)2 +
∑angles
1
2ka(θ − θ0)2
+∑
torsions
V n
2[1 + cos(nΦ− δ)]
(1)
2
-
where r0 and kb are stretching and bend-ing constants,
respectively; ka is the equi-librium bond length and θ0 is the
equilib-rium angle. Torsional energy is usuallyrepresented by a
cosine function, whereΦ is the torsional angle, δ is the phase,
ndefines the number of minima or maximabetween 0 and 2π (also
called multiplicity)and, Vn determines the height of the po-tential
barrier.
Regarding the non-bonded terms, Vander Waals interactions arise
from the bal-ance between repulsive and attractiveforces. The 12-6
Lennard-Jones (LJ) po-tential is very often used to representthese
interactions:
V (r) = 4�
[(σ
r
)12−
(σ
r
)6](2)
in which the first term represents the re-pulsive part of the
potential while the sec-ond is the attractive term. The � is
depthof the potential well, σ is the finite dis-tance at which the
inter-particle potentialis zero and, r is the distance between
par-ticles. The final term of the non-bondedparameters serves to
describe the electro-static interactions. The electrostatic
inter-action arises due to the unequal distribu-tion of charge in a
molecule. This interac-tion between these point charges is
gen-erally modelled by a Coulomb potential:
V (r) =∑i
∑j 6=i
qiqj4π�0rij
(3)
where �0 is the permittivity of free space, qare atomic charges
and, rij is the distancebetween nuclei i and j.
3. Experimental Procedure
The 2,2,3,3,4,4,5,5,6,6,7,7,7-tridecafluoroheptan-1-ol (6:1
FTOH, CASnumber: 375-82-6) was purchased fromApollo Scientific, for
which was claimeda 97% purity. Prior to experimental mea-surements,
the 6:1 FTOH was dried withVWR Prolabo 4A molecular sieves toa
maximum water content of 200 ppm(analysed by Karl-Fischer
coulometry).
The viscosity of 6:1 FTOH was mea-sured in the temperature range
of 283.15K and 253.15 K at atmospheric pres-sure using an automated
SVM 3000 AntonPaar rotational Stabinger viscosimeter-densimeter.
The temperature uncertaintyis ±0.02 K from 288.15 to 378.15 K.
Theprecision of the dynamic viscosity mea-surements is ±0.5%.
4. Simulation procedure and details
Molecular dynamics simulations wereperformed by applying models
based onthe atomistic optimized potential for liq-uid simulations
all-atom (OPLS-AA) forcefield.[7] The necessary potential
parame-ters for Butanol were published in the origi-nal OPLS-AA
papers[7][13], while for Hex-anol and Decanol the parameters are
pub-lished in the L-OPLS-AA papers, namelythe extension for
alcohols.[11]
To describe the fluorinated alcohols, 3:1FTOH, 5:1 FTOH and 7:1
FTOH, themodel used in this work was built byusing the force-field
parameters of the(−CF2 − CH2 − OH) segment from themodel of
Trifluoroethanol developed byDuffy [3] and for the perfluoroalkyl
“tail”(−CF3 − CF2−) from the OPLS-AA workon perfluoroalkanes.[13]
The remaining di-
3
-
hedral torsion parameters were taken fromthe work of Padua[10],
and the partialcharge of the carbon atom in the first−CF2 group was
adjusted to give themolecule a net zero.
According to the OPLS parametrization,the non-bonded
Lennard-Jones interac-tions between different types of sites
werecalculated using the geometrical meanrule for both size and
energy. For thecross-interaction energy and the cross-interaction
diameter between alkyl hy-drogen atoms and perfluoroalkyl
fluorineatoms, in order to capture the weak unlikeinteractions
between hydrogenated andfluorinated chains, it it was introduced
cor-rections to the energy and size binary in-teraction parameters,
ξ = 0.80 and υ =1.035, for (BuOH + HFB) as suggested byMorgado for
the OPLS-AA force field [9]and ξ = 0.77 and υ = 1.035 for (HexOH
+UFH) in the L-OPLS-AA force field.
All MD simulations were carried outwith the open-source package
GROMACS(version 2018)[1][12], in cubic boxes withperiodic boundary
conditions imposed inall directions and a time step of 2 fs.All
bonds involving Hydrogen atoms wereconstrained to their equilibrium
lengths,using the LINCS algorithm.[6] Simulationswere conducted for
systems composed of300 molecules for (HexOH + UFH) and(DOH + PFO)
and, 500 molecules for(BuOH + HFB) with the system size be-ing
conserved as the molar ratio of alcoholto fluorinated alcohol was
changed. Thegeneral simulation scheme was the follow-ing:
Initially, the molecules were placedrandomly in the box and were
allowed toequilibrate in the NPT ensemble for 2.5 ns,during this
period, the density of the sys-
tem converged to mean values, then a 2.5ns long production run
is performed. After-wards, 2 ns long equilibration runs
wereperformed in the NVT emsemble at den-sities obtained from NPT
simulations atatmospheric pressure which are followedwith NVT
production runs from 5 to 50 nslong.
5. Densities and Excess Volumes
From the simulations performed, thesystems densities were
collected directlyfrom the average values of the box vol-ume in the
NPT production runs. Thedensities were calculated for (BuOH +HFB),
(HexOH + UFH) and (DOH + PFO)at 298.15 K. The calculated
densitiesshowed good agreement with experimen-tal data.
Figure 1: Densities of (BuOH + HFB) at 298.15 K
Figure 2: Densities of (HexOH + UFH)at 298.15 K
Figure 3: Densities of the mixture (PFO + Decanol) at
298.15K
4
-
Subsequently, the related excess vol-umes were calculated and
plotted in fig-ures 4 and 5. For the (BuOH + HFB)the values were
compared with the ex-perimental data published by Morgado[9]while
the (HexOH + UFH) system werecompared with experimental
measure-ments performed by research collaboratorMariana
Leitão.
V Em =x1M1 + x2M2
ρ12− x1M1
ρ1− x2M2
ρ2(4)
To model the unlike H − F interaction,the corrections on size
and energy param-eters proposed by Morgado[9], ξ = 0.8 andυ =
1.035, were introduced in the forcefieldfor the (BuOH + HFB)
mixture. The val-ues calculated for excess volume from
MDsimulations for this system demonstratea good agreement with the
experimentaldata.
Figure 4: Excess Volumes of (BuOH + HFB)
However, when the parameters mod-elled for (Hexane +
Perfluorohexane) inthe L-OPLS force field, η = 1.035 andυ = 0.77,
were applied to the mixture of(HexOH + UFH), the calculated
excessvolumes are significantly lower than theones experimentally
obtained. Accord-ingly, simulations were ran to parametrizethe best
correction for the (HexOH + UFH)which is η = 0.77 and υ =
1.039.
Figure 5: Excess Volumes of (HexOH + UFH)
This correction was kept for the (DOH+ PFO) system. For this
system, in-stead of comparing excess volumes, cal-culated and
experimental molar volumeswere compared, figure 6.
Figure 6: Molar Volume of the mixture (PFO + Decanol) at298.15
K
6. Viscosities and Excess Viscosity
6.1. Pure compounds
The experimental measurements of theviscosity of 6:1 FTOH as
function of tem-perature are plotted in the figure 7, along-side
other different fluorotelomer alco-hols. As expected, the
viscosities of flu-orotelomer alcohols increase with the in-crease
in chain length. As far as known,this are the first reported
viscosities for 6:1FTOH. Moreover, the viscosities of FTOHwere
compared with the viscosity of differ-ent types of compounds.
5
-
Figure 7: Experimental viscosities of fluorotelomer alcohols asa
function of temperature
Figure 8: Experimental viscosities at 293.15 K
The measurements executed for 6:1FTOH show good agreement with
pre-vious reported viscosities for other fluo-rotelomer alcohols.
Also, FTOH have anhigher viscosity than the remaining typesof
compounds due to the existence of hy-drogen bonds combined with the
highermolecular weight of Fluorine.[5]
6.2. Mixtures
Due to its simplicity, the Green-Kubo re-lation based
equilibrium molecular dynam-ics (MD) simulations is perhaps the
mostwidely used method to calculate viscosi-ties. In this approach,
the shear viscosityis calculated from the integral over time ofthe
pressure tensor autocorrelation func-tion
η =V
kBT
∫ ∞0
〈Pαβ(t) · Pαβ(0)〉dt (5)
where V is the system volume, kB is theBoltzmann constant, T is
temperature, Pαβdenotes the off-diagonal elements of the
pressure tensor, and the angle bracket in-dicates the average
ensemble. To obtainthe viscosities from the MD simulations,the
methodology proposed by Zhang wasused.[14]
The calculated viscosities for the (BuOH+ HFB) mixture , at
293.15 K were com-pared with viscosities experimentally mea-sured
by Costa. [4].
Figure 9: Experimental and simulated viscosities of (BuOH +HFB)
and simulated densities at 293.15 K
Figure 10: Experimental and simulated viscosities of (HexOH+
UFH) and simulated densities at 293.15 K
The simulations are capable of predict-ing the singular excess
behaviour of vis-cosity in this mixtures, showing resultsclose to
the experimental and revealinglarge negative deviations.
Figure 11: Experimental and simulated excess viscosities
for(BuOH + HFB)
6
-
Figure 12: Experimental and simulated excess viscosities
for(HexOH + UFH)
7. Diffusion Coefficients and Hydrody-namic Radii
The diffusion coefficients of differentmixtures were calculated
from the linearpart of the mean square displacement ofthe center of
mass of the solute moleculesaccording to the Einstein equation:
D =1
6Nlimt→∞
d
dt
N∑i=1
< [ri(t)−ri(0)]2 > (6)
where [ri(t)− ri(0)]2 is the mean squaredisplacement of the
solute and the brack-ets stand for average over time. The
sum-mation extends to all solute molecules inthe simulation. The
final value of diffusioncoefficient was obtained from the averageof
5 values obtained independently.
Figure 13: Diffusion coefficients at 293.15 K for (BuOH +
HFB)and (UFH + Hexanol)
To further understand the molecules mo-tion in this mixtures the
effective radiuswas calculated. By cancelling the ef-
fect different viscosities have on moleculesmovement, the
effective radius allows abetter evaluation of the molecules
motion.Therefore, a higher effective radius meansthe motion of the
molecule in the mixtureis more difficult.
The translational motion of a solute in afluid solution at
infinite dilution can be de-scribed by the Stokes-Einstein
relation.
D =kBT
Cπηr(7)
where D is the diffusion coefficient, kB isthe Boltzmann
constant, T is the abso-lute temperature, and C is equal to 6
forthe case of “stick” boundary conditions.Furthermore, because the
calculated vis-cosities from MD simulations have an highof
uncertainty, these hydrodynamic radiiwere calculated using
experimental vis-cosities reported by Costa.[4]
Simulations of HexOH and UFH at infi-nite dilution in UFH and
HexOH, respec-tively, were also performed and the corre-sponding
diffusion coefficients and effec-tive radii calculated. These are
included infigure 14. Figure ?? consisted of a singlemolecule of
hexane and a single moleculeof perfluorohexane inserted in pure
hex-anol, pure UFH and an equimolar mixture.This allows
understanding the influenceof the weak interactions between
hydro-genated and fluorinated chains have onthe flow in the absence
of hydrogen bond-ing. This study was conducted at 343.15K to speed
up the molecular dynamics ofthe (HexOH + UFH) system, making
thesamplings faster.
7
-
Figure 14: Hydrodynamic radius of HexOH and UFH at 343.15K with
experimental viscosities
Figure 14 suggests that when diluted inHexanol, molecules of UFH
tend to have ahigher motion than when pure as shownby the clear
tendency of decreasing itshydrodynamic radii when the
concentra-tion on Hexanol increases. RegardingHexanol, the change
of effective radius ismuch smaller. Nevertheless, the resultsseem
to indicate that there is an initialslight decrease of the
effective radius ofHexanol in the mixtures, increasing
after-wards.
Figure 15: Hydrodynamic radius at 343.15 K with
experimentalviscosities
The Perfluorohexane molecule de-creases slightly its effective
radius wheninfinitely diluted in Hexanol and evenmore significantly
when in the equimolarmixture. This overall decrease could indi-cate
that the weak dispersion interactionsbetween hydrogenated and
fluorinatedchains can ”help” the movement of thefluorinated
chain.
8. Liquid Structure
Intermolecular radial distribution func-tions (rdfs) provide a
measure of the localstructure in liquids. They were calculatedfrom
the simulation trajectories for the thefluorinated chain of UFH and
PFO and, thehydrogenated chain of Hexanol and De-canol excluding
the Hydrogens bonded toCarbon-1 and, the Oxygen and Hydroxylgroup,
not only for pure like interactionsbut also for
cross-interactions.
-Figure 16: Intermolecular rdfs between the Hydrogen and
Flu-orine atoms (H or F) for (HexOH + UFH ) mixtures at
differentcompositions, from molecular dynamics simulations.
-Figure 17: Intermolecular rdfs between the Hydrogen and
Flu-orine atoms (H or F) for (DOH + PFO) mixtures at different
com-positions, from molecular dynamics simulations.
Radial distribution functions generatedfor H − H and F − F
group, increasenotably with the increase in concentra-tion of the
second component. This astrong indication of segregation
betweenhydrogenated and fluorinated segmentsand supports the
conclusion of the pres-ence of different clusters of fluorinated
and
8
-
hydrogenated chains in the mixture.
(a) HexOH + UFH (b) DOH + PFOFigure 18: Snapshot of a MD
simulation box consisting of anequimolar mixture. Alcohols (HexOH
and DOH) are the bluechains and, fluorinated alcohols (UFH and PFO)
are the greenchains
Intermolecular rdfs between the oxy-gen and hydroxyl hydrogen
atoms for the(HexOH + UFH) system show that theaddition of the
different alcohol also in-creases the variety of H-bonds formedupon
mixing. Even so, the new es-tablished cross-interaction are
strongerthan the intersections between identicalmolecules with the
curve correspondingto hydrogen bonding between the Oxygenatom in
HexOH, highest negative partialcharge, and the Hydrogen atom,
highestpositive partial charge, in UFH being muchmore intense than
the others, which fol-lows the difference of the partial
chargesassigned to the interacting atoms.
Figure 19: Intermolecular rdfs between the oxygen and hy-droxyl
hydrogen atoms for the (HexOH + UFH) at equimolarmixture.
The analysis of the rdfs indicate thatsimilar to what was
previously seen byMorgado[9] for the (BuOH + HFB) sys-tem, mixtures
of (HexOH + UFH) can be
regarded as nanostructured. This nanos-tructure consists of a O
· · ·H network ofhydrogen bonds formed between the hy-droxyl
headgroups, surrounded by the car-bon chain tails, which, in turn,
segre-gate into hydrogenated and fluorinated do-mains. Figure 20,
is a molecular dy-namics simulation snapshot obtained foran
equimolar mixture of (HexOH + UFH)which illustrates such
structure.
Figure 20: Snapshot of the O · · ·H network in a MD
simulationbox consisting of an equimolar mixture of HexOH (blue
chains)and UFH (green chains); coloured red are the Oxygen’s of
bothmolecules and white are the Hydrogen’s
9. Conclusions and Future Work
The molecular dynamics simulationsperformed indicate that the
large negativeexcess viscosities of mixtures of (BuOH +HFB) and
(HexOH + UFH) results from anenhanced motion of fluorinated
segmentswhen in contact with hydrogenated alco-hols due to the
unfavourable dispersionforces.
The existence of mutual segregationbetween fluorinated and
hydrogenatedchains was identified using radial distribu-tion
functions. Also, the existence of anano structure composed by a O ·
· ·HOnetwork was identified for the (HexOH +UFH) mixture.
The comparison of the reported vis-cosities for
2,2,3,3,4,4,5,5,6,6,7,7,7-
9
-
tridecafluoroheptan-1-ol with other fluo-rotelomer alcohols,
present in literature,shows a good agreement with
viscositiesreported for other fluorotelomer alcohols.
Acronyms
BuOH - ButanolHexOH - HexanolDOH - DecanolHFB -
2,2,3,3,4,4,4-heptafluorobutan-1-olUFH -
2,2,3,3,4,4,5,5,6,6,6-undecafluorohexan-
1-olPFO - 2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-
pentadecafluorooctan-1-olFTOH - Fluorotelomer alcoholMD -
Molecular DynamicsOPLS - Optimized Potential for Liquid
SimulationAA - All AtomH-Bonds - Hydrogen Bondsrdf - Radial
Distribution Function
References[1] M. Abraham, E. Lindahl, B. Hess, and GRO-
MACS Development Team. GROMACS UserManual version 2018,
2018.
[2] M. P. Allen and D. J. Tildesley. Computer Sim-ulation of
Liquids. Oxford Science Press, sec-ond edi edition, 2017.
[3] R. Chitra and P. E. Smith. A comparison of theproperties of
2,2,2-trifluoroethanol and 2,2,2-trifluoroethanol/water mixtures
using differentforce fields. Journal of Chemical
Physics,115(12):5521–5530, 2001.
[4] M. A. Costa. Property Data and Phase Equi-libria for the
Design of Chemical Processesinvolving Carbon Dioxide, 2017.
[5] M. G. Freire, A. G. Ferreira, I. M. Fonseca,I. M. Marrucho,
and J. A. Coutinho. Viscosi-ties of liquid fluorocompounds. Journal
ofChemical and Engineering Data, 53(2):538–542, 2008.
[6] B. Hess, H. Bekker, H. J. Berendsen, and J. G.Fraaije.
LINCS: A Linear Constraint Solver formolecular simulations. Journal
of Computa-tional Chemistry, 18(12):1463–1472, 1997.
[7] W. L. Jorgensen, D. S. Maxwell, and J. Tirado-Rives.
Development and testing of the OPLSall-atom force field on
conformational en-ergetics and properties of organic
liquids.Journal of the American Chemical
Society,118(45):11225–11236, 1996.
[8] P. Morgado, J. Black, J. B. Lewis, C. R. Ia-covella, C.
McCabe, L. F. Martins, and E. J.Filipe. Viscosity of liquid systems
involv-ing hydrogenated and fluorinated substances:Liquid mixtures
of (hexane+perfluorohexane).Fluid Phase Equilibria, 358:161–165,
2013.
[9] P. Morgado, A. R. Garcia, L. M. Ilharco, J. Mar-cos, M.
Anastácio, L. F. G. Martins, andE. J. M. Filipe. Liquid Mixtures
Involving Hy-drogenated and Fluorinated Alcohols: Ther-modynamics,
Spectroscopy, and Simulation.The Journal of Physical Chemistry B,
pageacs.jpcb.6b04297, 2016.
[10] A. A. Pádua. Torsion energy profiles and forcefields
derived from Ab initio calculations forsimulations of
hydrocarbon-fluorocarbon di-blocks and perfluoroalkylbromides.
Journal ofPhysical Chemistry A, 106(43):10116–10123,2002.
[11] K. Pluhackova, H. Morhenn, L. Lautner,W. Lohstroh, K. S.
Nemkovski, T. Unruh,and R. A. Böckmann. Extension of theLOPLS-AA
Force Field for Alcohols, Esters,and Monoolein Bilayers and its
Validation byNeutron Scattering Experiments. Journal ofPhysical
Chemistry B, 119(49):15287–15299,2015.
[12] D. V. Spoel, E. Lindahl, B. Hess, andG. Groenhof. GROMACS:
Fast, Flexible, andFree. Journal of Computational
Chemistry,26(16):1701–18, 2005.
[13] E. K. Watkins and W. L. Jorgensen. Per-fluoroalkanes:
Conformational Analysis andLiquid-State Properties from ab Initio
andMonte Carlo Calculations. Journal of Physi-cal Chemistry A,
105(16):4118–4125, 2001.
[14] Y. Zhang, A. Otani, and E. J. Maginn. ReliableViscosity
Calculation from Equilibrium Molec-ular Dynamics Simulations: A
Time Decom-position Method. Journal of Chemical Theoryand
Computation, 11(8):3537–3546, 2015.
10