1.
Introduction
Propylene
glycol (PG) is a chemical commonly used in the food, cosmetics, and medical
industries.[1] The thermal treatment of PG-containing matrix may yield
decomposition reaction products.[2]
Several studies have investigated the thermal decomposition of PG and
highlighted the emission of toxic carbonyls.[3,4] In
2003, electronic cigarettes (ECIGS) were introduced to the market along with
corresponding liquids in which PG is the main constituent.[5,6] The
problem of thermal degradation of PG is particularly acute for ECIGs, which
produce an inhalable aerosol by heating and vaporizing the liquid.[7] The
inhalation exposure of ECIG users to the generated carbonyls has been a concern
since the introduction of ECIGs into the market.[8-10]
Predicting the amounts of toxicants yielded from a given process can help
designers minimize their production, and this can be useful in minimizing
carbonyl emissions from ECIGs.
The thermal degradation of PG
under oxygen rich conditions at 400-500 K was shown to yield formaldehyde (FA),
acetaldehyde (AA), and CO2 via oxidative C-C bond cleavage.[11] The
researchers suggested that PG follows a dehydration reaction to produce acetone
(Ace). Ayre et al. (1994) investigated the adsorption and reaction of PG
on oxygen activated Ag (110) surface, and they showed that PG adsorption
occurred at 215 K, while degradation peaked at 340K to form acetol and
lactaldehyde.[12] In 2012, Laino et al. reported the degradation mechanisms of PG
by performing a pyrolysis study at 800K with 30 seconds of residence time.[3]
Activation Gibbs free energy barriers and rate constants were calculated at
this temperature. Salaev et al. (2016) performed a theoretical study of
the gas phase oxidation of PG to methylglyoxal (MGA) over silver catalyst. It
consisted of optimizing geometry parameters of gas-phase molecules, as well as thermochemical data (heats of
formation) for the stages of PG oxidation.[13]
In ECIG research, in an
attempt to understand the mechanism of carbonyls generation, two ECIG
device-independent pyrolysis studies were reported.[14,15] Wang et
al. (2017) investigated the thermal degradation of five different ECIG
liquids, including prepared mixtures of PG and vegetable glycerin, and
commercially available liquids, in a stainless-steel tubular reactor.[14] They
confirmed that the decomposition of PG and vegetable glycerin at temperatures
above 450 K are the main sources of carbonyls in ECIGs: FA, AA, and acrolein
were detected at different concentrations for various ECIG liquids. Our group
reported a pyrolysis study of PG-saturated gas stream in a quartz tubular
reactor at a wide range of temperatures (~350-940 K).[15] The
influence of the atmosphere (air versus N2), the temperature and the
presence of coil material wires (nichrome, kanthal, stainless steel, and
nickel) on the thermal degradation of PG was assessed.
Even though PG is not
considered toxic for human health, its pyrolysis products, mainly FA
(Formaldehyde), AA (Acetaldehyde) and PA (Propanal or propionaldehyde), are
considered harmful.[3,16,17] In the 12th Report on Carcinogens
conducted by National Toxicology Program, FA was classified as a known human
carcinogen.[18] Acetaldehyde is classified as a Group B2 carcinogen by the United
States Environmental Protection Agency (EPA), i.e., it is a probable human
carcinogen, as more human data is needed.[17] Finally,
PA is linked to respiratory problems such as fluid build-up in lungs yet is not
considered carcinogenic.[19]
Laino et al. studied the
thermal decomposition mechanisms of PG by relying on simple decomposition based
on dehydration of PG into either propylene oxide or allyl alcohol.[3]
Propylene oxide can further decompose into acetone or propanal. They found
energies such as total system energy, enthalpy and Gibbs free energy,
theoretically from metadynamic simulations. The free energy surfaces found were
optimized at the Density Functional Theory (DFT) level and the mechanism
validated by GC-MS.
Herein, we present a kinetic
model that relies on the data produced from the PG pyrolysis experiment in a
quartz tube reactor under inert conditions.[15] By
doing that, the thermal effect is isolated, and accurate parameters can be
estimated for the thermal degradation process. The kinetic pathway mechanisms
are determined based on the experimental data, and a rate-based kinetic model
is developed. The activation energies for the reaction rate Arrhenius constants
are calculated, under various temperatures that mimicked ECIG conditions, by
employing sequential parameter estimation.[20] In
addition, results are validated via quantum chemistry simulations. ReaxFF
Molecular Dynamics (ReaxFF-MD), developed by van Duin et al., is employed in
this study, along with ReaxAMS, to predict reaction pathways, product
formation, transition states, and energies. The importance of such
atomistic-scale computational techniques has grown greatly over the years,
mainly due to their accessibility and accuracy.[21] This
study is critical to our efforts in putting forward a comprehensive
mathematical model of toxicants in ECIG emissions.[22] Our
works employs different perspectives: experiments, mathematical modeling,
molecular dynamics, and elementary scale computations; it thus manages to merge
these concepts into an unprecedented extensive study on PG pyrolysis.
2.
Materials and methods
2.1
Experimental setup
We
studied the homogeneous reaction of PG at various temperatures under inert
conditions in a quartz tube reactor (Fig. 1).
Controlled nitrogen gas flow (N2) was passed through an impinger
containing PG to generate a PG-saturated gas stream. A glass fiber filter was
placed downstream of the impinger to intercept any mechanically-generated
liquid particles. The 120 cm horizontal quartz tube reactor, with a radius of
0.33 cm, was placed in a two-zone rotary electric tube furnace (1200 °C max.).
The 45 cm heated zone was nominally set at 100, 300, 400, 500, 600, or 700 °C.
These temperatures corresponded to actual temperatures of 80, 256, 360, 460,
565, and 672 °C in the quartz tube reactor as measured using a K-type
thermocouple. For each temperature condition, the mass flow through the reactor
was adjusted to attain a fixed mean residence time of 1.6 s. Using the ideal
gas law, the corresponding mass flow rates for the above temperatures were 386,
271, 236, 209, and 187 standard mL/min, respectively. The carbonyls in the
outlet gas flow were trapped on DNPH-silica coated H10 cartridges and then
extracted with ethanol/acetonitrile (90/10) solution to be analyzed on HPLC.
The detailed analytical method of carbonyl quantification is detailed
elsewhere.[23] Six different carbonyls were detected and quantified:
propionaldehyde (PA), acetone (Ace), acetaldehyde (AA), formaldehyde (FA),
glyoxal (GA) and methylglyoxal (MGA).
Fig. 1: Scheme for the pyrolysis reactor used in this study.[15]
2.2
Computational methods
2.2.1
Reactive Force Field ReaxFF-MD
Quantum
mechanics (QM) is vastly used as a theoretical basis and a screening tool, as
it is limited by its computational cost while working for larger systems (more
than 100 atoms), and its inability to account for the dynamic progression of
the system in question. QM calculations essentially solve the famous
Schrodinger’s equation:
(1)
which
provides information on electron behavior and other events at the atomic scale.
H is the Hamiltonian operator, which contains electron energies and other
system details, and Ψ is the wave function which depends on electron
position in space (r) and time
, and contains all physical
properties of the system. QM is commonly used as a structure and energy basis
for more advanced computations. Such calculations are referred to as empirical
force fields (EFF), yielding interatomic potentials that compute, based on atom
positions,
the energy of the system. These potentials include
harmonic potentials, van der Waals potentials and Coulombic interactions.[21] EFF
are proven to work for large systems. EFF methods are accurate in describing
non-reactive physical systems at or around equilibrium.[24]
Hence, reactive force fields (ReaxFF) were introduced to expand such potentials
to include a bond-order formalism, thus accounting for formation and breaking
of bonds i.e., chemical reactions. ReaxFF requires no pre-defined
reaction mechanisms, proving to be an innovative tool that predicts system
behavior based on environment-dependent reactive and non-reactive potentials,
rather than user reaction input.
We employed the Amsterdam
Modeling Suite’s ReaxFF developed originally by Adri van Duin.[25] This
package was provided by Software for Chemistry and Materials (SCM).[26]
Originally developed for hydrocarbons, ReaxFF was later expanded for different
systems. The general energy form in a ReaxFF simulation is computed as follows,
(2)
where
the system energy (Esystem) is defined as a sum of energy
terms.[27] Bond energy found from corrected bond order is defined as Ebond,
with the two valence theory correction terms Eover and Eunder
to account for atom over-coordination and under-coordination, respectively. In
addition, the valence angle energy is denoted by Eval in
addition to the penalty energy (Epen) term for systems
sharing one atom in a valence angle while having double bonds. Furthermore,
torsion angle energy (Etor) and conjugation effects (Econj),
are accounted for. Finally, for non-bonded interactions, Van der Waals energy
is accounted for in EVdW term, and Coulomb interactions in ECoul.
Note that bond order is always updated as atomic positions are updated.[28] In
this study, we investigated PG pyrolysis in ReaxFF, with the source code and
Graphical User Interface (GUI) provided by SCM via AMS.[25]
2.2.2
Simulation details
A box
of dimensions 23Å × 23Å × 23Å was filled with ten PG (1,2- Propanediol)
molecules as illustrated in Fig. 2, via the ‘Builder’ option
found in ReaxFF and provided by the Packmol package.[29] The
distance between molecules was 2.5Å. Such volume and molecule number yield a
density of 0.1038 g/cm3, which was chosen based on preliminary
trial-and-error simulations with other densities, where it provided better
particle collisions leading to enhanced carbonyl formation, replicating the
physical system of abundant PG input flow in e-cigarettes and in our
experiment.
After generating the reaction
system, the CHO.ff: (C/H/O) hydrocarbon oxidation force field was chosen mainly
since C, H and O are the only elements involved. This system was first
subjected to a Number-Volume-Energy (NVE) microcanonical ensemble at 5K for
10ps in order to minimize its energy. Then the system was equilibrated for 10
ps isothermally at 1000K to ensure that all initial energies are constant with
minimal fluctuations. It was set that no reaction takes place during these
first 10 ps (equilibration stage). In this equilibration step, the
Number-Volume-Temperature (NVT) canonical ensemble with a Berendsen thermostat
was used to statistically describe the system; such ideal state is the most
probable state for the real reaction system being simulated. The Berendsen
thermostat works by correcting system temperature during the simulation with
exponential decay.[30]
Thus, the optimal initial
configuration of the system is obtained and used for the following MD
simulations.
Fig. 2: PG Pyrolysis simulation box input.
2.2.2.1 Annealing simulations
In order to study overall PG pyrolysis and assess the products
formed, a temperature regime was employed to get the system at the desired
final temperature. Then, the temperature was increased linearly for 500 ps to
reach the desired maximum temperature (TMax). This was done to give
the system time to adjust to these temperature changes and to properly predict
reaction pathways (temperature increase stage). The system remained at this
temperature for another 500 ps (isothermal stage). Thus, a total of 10,100,000
iterations, or a simulation time of 1010 ps, were needed. A 10fs damping
constant and a time step of 0.1fs/iteration were applied throughout these
simulations. It is a common practice in ReaxFF to employ high temperatures (e.g.
up to 4600 K) in order to get results in real-time.[31] Initially, simulation
temperatures were set between 3000 K and 3750 K with an incremental temperature
ΔT of 250 K. This was changed to 100 K when further simulations were
conducted to generate more refined data. Three random simulations were conducted
for each temperature.
2.2.2.2 Isothermal simulations
To study reaction kinetics, isothermal simulations were conducted
to obtain reaction rate constants. The equilibrated system was subjected to
temperatures from 2000 K to 4000 K at 250 K intervals for 10 ps (for every
temperature) with a 100fs damping constant. Triplicates were also performed in
this stage. As stated in literature and based on our simulations, such high
temperatures do not affect kinetic rate constants since the relationship
between ln(K) and 1/T is generally linear, and data at high temperatures can be
interpolated to lower temperatures.
2.2.3 ReaxAMS computations
Small scale (at most 5 molecules) calculations were performed
using ReaxAMS, which is an SCM product that combines SCM’s ADF and BAND
functionalities with reactive molecular dynamics in one
interface.[26] The purpose of ReaxAMS was to study separately each elementary
reaction involved in PG pyrolysis. Unlike ReaxFF-MD, the focus here was not on
product concentrations and trends with respect to time. In fact, time is not a
factor in these small-scale computations. The path the reaction takes, from
reactant to product, was rather studied from an energy perspective. For every
reaction, geometry optimizations for initial and final systems were computed,
first using the universal force field (UFF) to improve bond lengths and angles,
and then CHO.ff to minimize the energy on the specified molecule. The Nudged
Elastic Band (NEB) method was then applied on the two systems to predict the
energy path. This optimizes the intermediate ‘images’ to find the lowest energy
possible from one image to the next. Spring forces along the band between
images were added, the potential perpendicular to the band found and projected
out, predicting the next image. From the energy path found, transition states
and associated energies (Gibbs free energy, bond energies and enthalpies) were
extracted.
3.
Results and discussion
3.1
Experimental results
3.1.1
Product concentrations
Table 1
summarizes the results of PG pyrolysis under an inert atmosphere.
Our results showed that thermal decomposition started at around 160 °C (433 K)
and increased with temperature producing mainly PA and Ace.
Table 1: PG pyrolysis products (g/m3)
under N2 at different temperatures and N2 flow rates
(mL/min).
|
T (° C)
|
Flow
Rate
|
PA
|
Ace
|
AA
|
FA
|
GA
|
MGA
|
|
160
|
386
|
0
|
0
|
0.00012
|
0
|
0.0008
|
0
|
|
360
|
271
|
0.00003
|
0.00065
|
0.00009
|
0
|
0.00181
|
0
|
|
460
|
236
|
0.01248
|
0.00356
|
0.00192
|
0
|
0.00241
|
0
|
|
565
|
209
|
0.11391
|
0.03911
|
0.00575
|
0.00164
|
0.00321
|
0.00384
|
|
670
|
187
|
0.14963
|
0.07464
|
0.01648
|
0.00844
|
0.0045
|
0.00085
|
3.1.2
Reaction mechanisms
The
degradation of PG under an inert atmosphere is expected to proceed mainly via
dehydration reactions as depicted in Fig. 3. PA
and Ace, the direct products of dehydration of PG, were detected at relatively
low temperatures, and they constituted the majority of the products formed (PA
and Ace constituted 61 % and 26 % of the total weight of carbonyls detected).
At higher temperatures, minimal amounts of FA and AA were detected, in addition
to trace amounts of MGA and GA. To account for all these products reaction
pathways were suggested and their corresponding kinetic parameters were
estimated based on experimental results. In the first reaction step, PG loses
one water molecule to give propylene oxide (PO) as an intermediate. The
generation of propylene oxide via this pathway was validated by a computational
study.[13] Also, a recent report discussed the presence of these species in
e cigarette aerosols.[32] Epoxide can undergo ring opening to give two other intermediates,
namely propen-1-ol (P1) and/or propen-2-ol (P2) (Eqs. (2) and (3)).[33] These
intermediates are just the enol forms of their corresponding carbonyls;
propene-1-ol can tautomerize to give PA (Eq. (4)), and
propene-2-ol can tautomerize to give Ace (Eq. (5)). On
a less prominent route, the enols can undergo C=C bond cleavage to give smaller
molecules including FA, AA, and ethylene (Eqs (6) and (7)). Eq. (8)
illustrates a transformation known as glycol cleavage in which a vicinal diol
molecule undergoes C-C bond cleavage to give smaller carbonyls (FA and AA in
this case).[34] This cleavage is usually catalyzed by a metal-based catalyst,
which is not the case in our system, thus it is less noteworthy in our
mechanistic study. Another minor transformation is the direct oxidation of the
two alcohol functional groups to give MGA and a hydrogen molecule, via
dehydrogenation reaction (Eq. (9)).[8]
Oxidation can be accompanied with C-C bond cleavage leading to the formation of
GA (Eq. (10)). MGA and GA were detected in trace amounts. It should be noted
that the possible presence of small molecules like ethylene, methane and
hydrogen gases was not verified in our experiment.
This scheme is in accordance
to the one proposed by Laino et al.[3] for
the decomposition of PG where PA and Ace form from the dehydration reaction. In
fact, it adds more possible pathways such as FA and AA formation from
dehydration and formation of GA and MGA from dehydrogenation reactions.
Fig. 3: Reaction pathways for
the pyrolysis of PG under an inert atmosphere.
3.1.3
Kinetic model pathways
A
proposed kinetic model incorporating reactions 1-10 is illustrated in Fig. 4.
Products whose concentrations were not measured were ignored and assumed not to
affect reaction rate constants. The mole balance for PG is obtained by
combining reaction 1 with 8, giving:
(3)
Fig. 4: Proposed kinetic model
for PG thermal degradation.
Reactions
1 to 8 were assumed to follow the quasi-steady-state approximation. This has
been done since the intermediates—
namely
PO, P1 and P2—were not detected, as will be later on
shown, in the ReaxFF-MD simulations. In fact, experimentally, these
intermediates were not detected either. However, these compounds do form, they
are presented in other literature,[3] and P1
and P2 radicals appear in ReaxFF-MD simulations.
Adding reactions 2-5 and 7-8
yields the following expressions for PO, P1, and P2
concentrations:
[Reaction 2 + Reaction 3]: 
, 
(4)
[Reaction
4 + Reaction 6]: 
, 
(5)
[Reaction
5 + Reaction 7]: 
, 
(6)
Substituting the
expressions of PO, P1, and P2 in the mole balances of PA,
Ace, FA, and AA yields the following modified mole
balances:
(7)
(8)
(9)
(10)
The reaction rates were
defined as:
(11)
(12)
(13)
(14)
(15)
(16)
The number of reaction
rate constants that will be estimated in the model was reduced from ten (k1
to k10) to seven (K1 to K7), with K1 to K7 defined by,
(17)
(18)
(19)
(20)
And,
(21)
(22)
(23)
All reaction rate
constants, K (s-1), were assumed to follow the Arrhenius equation,
(24)
With A0:
pre-exponential term (1/s), T: temperature (K), Ea: Activation energy in
(J/mole), R: ideal gas constant (J/mol.K). The parameter estimation toolbox in Matlab was used to estimate
the Arrhenius constants that minimize the square of the errors between
experimental results and model predictions.
3.1.4
Model validation
The
estimated activation energies for the rate constants of equations (17) – (23)
are summarized in Table 2. PG degradation under N2 atmosphere mainly produced
propanal starting at 360 oC which continued to increase dramatically
as the temperature reached 565 oC. Similarly, Acetone was produced
but to a lesser extent. This 1 can be explained by PG degradation favoring the
formation of propylene-oxide intermediate at low temperatures, under which it
degrades to propene-1-ol and propene-2-ol along paths 1-2-4 and 1-3-5 as shown
in Fig. 4.
Table 2:
Estimated activation energies (J/mol)
|
Activation energy from rate constant
|
K1
|
K2
|
K3
|
K4
|
K5
|
K6
|
K7
|
|
Ea (J/mol)
|
1.14×105
|
1.4×105
|
2.41×105
|
1.86×105
|
1.64×105
|
1.17×105
|
1.14×105
|
As the temperature is
increased, PG and propene-ols underwent further cracking to produce new
carbonyls including FA and MGA from side reaction pathways 1-2-6 and 1-3-7
along with side reactions 8, 9 and 10. PA product formation slowed down as a
result of the new side-products being formed above 565 oC as shown in Fig. 5
and Fig. 6. This kinetic behavior was captured by the degradation model
which showed an excellent agreement with the experimental results of all
products.
Fig.5: Experimental
versus predicted concentration at different temperatures for PA and Ace.
Fig. 6:
Experimental versus predicted concentration of products at different
temperatures for AA, FA, GA, and MGA.
Note
that AA and FA carbonyls began to form at an accelerated rate as temperature
increased above 460 °C, whereas the production of GA and MGA was low and simultaneously taking place but at receding reaction
rates. At 565 °C PA represented 68
%(wt) of the products
and Ace was at 23 %(wt). This declined to 59 % (wt) for PA at 670 oC while Ace continued to increase
and reached 29 % (wt). On the other hand, AA, FA,
GA, and MGA were found at 3-6
%, 1-3%, 1.8-1.9%, and
2.3-0.3% respectively. This indicates that paths 1-2-4 and 1-3-5 in Fig. 4 are the dominant routes for PG
thermal degradation under an inert atmosphere.
3.2
ReaxFF-MD results
3.2.1
Product distribution
The
simulation results from both annealing and isothermal simulations were assessed
first with the Reaction Event Detection tool in ReaxFF’s ChemTraYzer. This tool
helps in viewing the radicals formed throughout the simulation time, the time
of formation of these radicals (via a timeline), the elementary reactions involved, and the
rates and fluxes of these reactions. One can also observe all the pathways
involved with a certain molecule, where it is involved as a reactant, or as a
product. PG undergoes pyrolysis as shown in Fig. 7. The
supplementary data (Fig. S1 through Fig. S4) further elaborates the mechanisms obtained from ChemTraYzer.
Fig. 7: Main reaction pathways of PG pyrolysis from ChemTraYzer analysis
from ReaxFF-MD simulations. Captions (1) to (7)
indicate each possible pathway.
From the timeline and the reactions taking place in ChemTraYzer, a
simple mechanism was employed where PG
decomposes
to give each of the following products:
• CH2O
or FA
• C2H4O
or AA
• C3H7O
or PA/Ace radical
• C2H4O2/C2H5O2
or GA radicals
• C3H5O2/C3H4O2
or MGA radical
These products are in
accordance to the ones obtained in the previous section and other literature.[3] It is
important to note that the formation of radicals for a certain molecule varied
with temperature; noting each at every run was done to ensure the molecule in
question is properly recorded. After examining the reaction events in every
run, every molecule’s (or radical’s) number (N) versus iteration. xy file was
extracted. All simulations were repeated three times to ensure accuracy and
reliability of the results. Fig. 8 shows the averaged N versus
time of all the species at TMax of 3250 K. The number of molecules
initially remains at 10 indicating PG has not yet started decomposition.
Species start forming from PG pyrolysis at around 2500K with a sharp increase
in molecules at 3000K. Decomposition thus occurs prior to reaching the
isothermal temperature. Fig. 9a, b and c show the evolution of the different PG pyrolysis products for the
case of TMax = 3250 K. Simulations have shown that PG started
completely decomposing at simulation temperatures above 3000 K. It fully
decomposes to produce numerous carbonyls, as shown in Fig. 9. FA
did not form as soon as PG decomposed, but soon after, it gradually increased
to reach maximum values. Similarly, AA formed later during the simulation,
increasing somewhat linearly, then decomposing to reach almost zero value. It
possibly formed smaller carbonyl chains such as CH2O, CH3,
and lighter gases such as H2 and H2O. The PA/Ace radical
was formed in spikes, nonetheless, it appeared first relatively to other
aldehydes, suggesting that it might be forming directly from PG degradation. In
fact, isothermal simulations further showed that these two molecules,
especially PA, formed as soon as PG started degrading.
GA formed early on in the
simulation, around the time PG started decomposing, which was expected, as GA
is a longer chain carbonyl with more oxygen atoms than FA and AA. It reached a
low maximum N value and decreased back to zero, as it decomposed further into
smaller carbon chains. MGA formed in spikes and did not reach high N values.
This was expected since MGA is a relatively long chain hydrocarbon that will
decompose to give smaller carbonyls and shorter molecules.
In addition, H2O is
a direct product of PG dehydration and formed as soon as PG decomposed. It kept
increasing as it was formed from other carbonyls. H2 appeared later
in the simulation and increased as it was formed from carbonyl dehydrogenation.
It then decreased which could be attributed to the formation of carbonyls
possessing single bonds. The abundance of H2O over H2
indicates that dehydration is favored over dehydrogenation.
Fig. 8:
Averaged total number of molecules over time, along with the temperature
profile, at TMax = 3250K
Fig. 9: Number of molecules distribution over time for TMax = 3250K. (a)
time evolution of molecules for PG, AA, FA and PA (b) time evolution of
molecules for PG, MGA and GA (c) time evolution of molecules for PG, H2O,
H2 and C2H4.
3.2.2
ReaxFF-MD kinetics
The
rates of formation of the carbonyls were calculated from ReaxFF-MD isothermal
simulations over the temperature range of 2000 K-4000 K. In general, it was
found for a carbonyl A as follows,
(25)
Where
NA_formed is the number of A molecules formed at the time of
formation and NA0 is the initial number of A molecules. From Eq.11 to Eq.16, the
rate constants were obtained at each simulation temperature. The Arrhenius
equation 
(24) was then employed in
logarithmic form to find the Arrhenius parameters of each carbonyl:
(26)
Plotting ln(K) versus 
should give a line with slope
and an intercept of ln(A0).
The calculated Ea and A0 for the proposed pyrolysis mechanism
reactions are summarized in Table 3. For
the experimental model, rate constants K1 to K7 were found, however, in order
to clarify each formation reaction, the K-values are as follows based on Eq.11
through Eq.16:
l
K1: k value for PA formation
l K2: k
value for Ace formation
l K3 + K5
was summed up to give k value for AA formation
l K4 + K5
was summed up to give k value for FA formation
l K6: k
value for GA formation
l K7: k
value for MGA formation
Table 3: Activation energy
values (J/mol) of carbonyl formation from ReaxFF-MD and compared to
experimental results.
|
|
Activation
energy from rate constant k-formation of:
|
|
PA
|
Ace
|
AA
|
FA
|
GA
|
MGA
|
|
Experimental model Ea (J/mol)
|
1.14E+05
|
1.40E+05
|
1.64E+05
|
1.64E+05
|
1.17E+05
|
1.14E+05
|
|
ReaxFF-MD Ea (J/mol)
|
1.11E+05
|
8.52E+04
|
1.41E+05
|
1.73E+05
|
1.06E+05
|
1.33E+05
|
|
Error with respect
to experimental model
|
3%
|
39%
|
14%
|
5%
|
9%
|
17%
|
For the dehydration-related products, Ace formation has the lowest
activation energy barrier from ReaxFF-MD, however, it does not appear in
abundance or across the whole temperature range as further provided in the
supplementary data (Fig. S6). Thus, between PA, AA
and FA, the reaction constant for PA formation exhibited the lowest activation
energy, meaning it is the most favored reaction. This is in accordance with
experimental results where PA was the most detected aldehyde, with the highest
concentrations, and formed at low temperatures. It was formed most readily in
ReaxF-MD although at lesser concentrations when compared to FA. This could be
attributed to the fact that FA is formed from other side decomposition
reactions of longer chain carbonyls (AA, PA, Ace, GA and MGA). In addition,
previous literature eploying DFT calculations [3] has shown that the
formation of PA from PG dehydration is favored over that of Ace.
As for dehydrogenation mechanism, both rate constants for GA and
MGA formation exhibit close values, however, in both ReaxFF-MD and experiments,
GA was detected across the temperature range while MGA was not a favored
product.
Table 4: Pre-exponential
factor (1/s) values of carbonyl formation from ReaxFF-MD
|
Pre-exponential
factor from rate constant k-formation of:
|
|
PA
|
Ace
|
AA
|
FA
|
GA
|
MGA
|
|
A0 (1/s)
|
1.63E+12
|
3.99E+12
|
4.95E+12
|
1.41E+13
|
2.98E+12
|
5.65E+12
|
The high values of A0 obtained indicate a high
frequency of molecule collisions. In fact, FA formation rate constant exhibits
the largest value and indicates the abundance of FA formed.
3.2.3
PG overall thermal degradation
Another approach to assessing PG pyrolysis is studying the overall
decomposition of PG. Therefore, the rate of PG decomposition kPG decomposition
was found from ReaxFF-MD results. The total concentration profile of PG was
extracted from every isothermal simulation (including triplicates per one TMax)
and the results were averaged for every T. Then the integrated first-order rate
law was applied to determine kPG decomposition at each T:
(27)
Where N
is the number of PG molecules decomposing as the simulation progresses.
Plotting ln (Nt) versus time t (in
s) provides a linearly-fitted plot with a slope of kPG decomposition. After
finding the kinetic rate constants for the whole temperature range, they were
linearly fitted using ln(kPG decomposition) versus 1/T. Arrhenius
parameters were then extracted from the linear fitting and presented in Table 5.
Table 5: Activation
energy values (J/mol) and pre-exponential factor (1/s) values of PG
decomposition from ReaxFF-MD.
|
|
PG Decomposition
|
|
Ea
|
1.86E+05
|
|
A0
|
6.84E+13
|
The results show that the A0
values for PG decomposition are at least one order of magnitude larger than all
k-formation terms, except for PA formation where PG decomposition is the same
order of magnitude. In addition, the energy barrier to break PG bonds is higher
than the ones for product formation. As will be later seen in ReaxAMS
small-scale computations, for the PG molecule to crack, energy is needed to be
inputted to the system (endergonic reaction), thus explaining the presence of a
high energy barrier.
3.3
ReaxFF sensitivity analysis
3.3.1
Effect of temperature
Temperature
played an important role during ReaxFF simulations and PG starts generating
decomposition carbonyls at 2000K, with increasing carbonyl concentrations at
2750K. PA, GA and MGA radicals form over the whole temperature range of 2000 to
4000K. FA and AA begin forming from 2500 up to 4000K. Finally, Ace, which is
generally not detected in abundance in both experiments and MD, forms only
between the 2000 to 3000K range. Such high temperatures were chosen as it is a
simulation condition where reactions in ReaxFF-MD are generally expected to
occur, as seen in literature.[31]
3.2.2
Effect of density
Density
is an important input variable that needs to be optimized before precise
simulations are run. By varying the volume of the box to 503 Å3,
233 Å3, and 183 Å3, densities of
0.01, 0.1 and 0.22 g/cm3 were obtained respectively. Simulations
were run for TMax values of 3000 K to 4000 K with 250 K intervals. N profiles
were parsed, with the number of PG molecules detected compared at TMax
= 3250 K in Fig. 10 Number of PG molecules distribution for the three densities over
time for TMax = 3250 K.
It can be seen that a density
of 0.01 g/cm3 does not lead to PG decomposing (except for a few
points in time which can be considered as noise). This is attributed to the low
collision of molecules. On the other hand, a density of 0.22 g/cm3
shows a sharp decrease in PG molecules, with the reappearance of numerous
spikes after its supposed decomposition. Here, side reactions are occurring,
due to the high density and thus high collisions between all molecules
involved. Nonetheless, the physical system, in the experiments specifically and
the e cigarette generally, do not produce propylene glycol gas during
pyrolysis. At the 0.1 g/cm3 density, PG decomposes less sharply than
at 0.22 g/cm3, nevertheless, it decomposes fully and does not
reappear after its consumption.
In addition, pyrolysis
products form in both densities 0.1 and 0.22 g/cm3 TMax =
3250 K, and in all densities as temperature increases. Therefore, to further
study the difference between densities, k-formation values and Arrhenius
fitting parameters were obtained. This has been done prior to the triplicates
method as this was the preliminary test. The results are provided in the
supplementary data in Table S1. All in all, increasing system density will increase particle
collisions, initiate decomposition sooner, and speed up the decomposition of
reactants, as noted in literature, such as the pyrolysis of n-dodecane.[27] As
density was increased, activation energy and pre-exponential factors of decomposition increased too, with the
latter increasing at a faster rate. Temperature-depended simulations where then
based on the system density (ρ=0.08g/cm3) of which results
showed closer Arrhenius values to those of experiments. Nonetheless, it has
been concluded that the effect of temperature on the system and on the results
is higher than the effect of density.
It can be seen that a density
of 0.01 g/cm3 does not lead to PG decomposing (except for a few
points in time which can be considered as noise). This is attributed to the low
collision of molecules. On the other hand, a density of 0.22 g/cm3
shows a sharp decrease in PG molecules, with the reappearance of numerous
spikes after its supposed decomposition. Here, side reactions are occurring,
due to the high density and thus high collisions between all molecules
involved. Nonetheless, the physical system, in the experiments
specifically and the e-cigarette generally, do not produce propylene glycol gas
during pyrolysis. At the 0.1 g/cm3 density, PG decomposes less
sharply than at 0.22 g/cm3, nevertheless, it decomposes fully and
does not reappear after its consumption.
In addition, pyrolysis products form in both densities 0.1 and
0.22 g/cm3 TMax = 3250K, and in all densities as
temperature increases. Therefore, to further study the difference between
densities, k-formation values and Arrhenius fitting parameters were obtained.
This has been done prior to the triplicates method as this was the preliminary
test. The results are provided in the supplementary data. All in all,
increasing system density will increase particle collisions, initiate
decomposition sooner, and speed up the decomposition of reactants, as noted in
literature, such as the pyrolysis of n-dodecane [27]. As density was
increased, activation energy and pre-exponential factors of decomposition
increased too, with the latter increasing at a faster rate.
Temperature-depended simulations where then based on the system density
(ρ=0.08g/cm3) of which results showed closer Arrhenius values
to those of experiments. Nonetheless, it has been concluded that the effect of
temperature on the system and on the results is higher than the effect of
density.
Fig. 10:
Number of PG molecules distribution for the three densities over time for TMax
= 3250K.
3.4
ReaxAMS results
3.4.1
Transition states and bond energies
The
elementary reactions presented in Fig. 3 were
simulated each on ReaxAMS using the NEB method. The energy pathways were
computed, and transition states identified as illustrated in Fig. 11.
ReaxAMS calculated the bond energy of every system involved. The
change in bond energy from reactants to products was determined, in addition to
the change between transition states and reactants, and transition states and
products. The values can be found in the supplementary data (Table S1).
Reaction information can be derived from bond energies and
interpreted using the TS diagrams in Fig. 11. For instance, for
reaction 1, energy was required to be added into the system to break the PG
bonds. PG undergoes a C-O bond cleavage, and one O-H bond formation, with the
oxygen still attached to the carbon in the molecule, to reach the TS. Then, the
product resulted from the TS by an O-H cleavage (to obtain H2O), and
a C-O bond formation.
In reaction 2, taking the reactant to the TS involved a change in
bond angles and lengths, with no bond cleavage. However, the TS underwent a C-O
bond cleavage in the middle (C2), a hydrogen abstraction on the outer carbon
atom (C1), resulting in the formation of an O-H bond and a double C=C bond
between C1 and C2, resulting to a higher energy on the TS-product side.
For reaction 3, the outer carbon atom (not bonded with oxygen,
i.e. C3) broke from C2. In addition, oxygen broke from C1, and bonds with a
hydrogen atom abstracted from C2. A double bond formed between C1 and C2, and a
single bond between C2 and C3.
In each of reactions 4 and 5, one O-H bond broke, and so did one
bond of the C=C double bonds. One C-H bond was formed, in addition to a double
C=O bond.
In reaction 6, one double C=C bond and one O-H bond broke. One C=O
bond and one C-H bond formed FA. Finally, one C=C double bond and one C-H bond
formed ethene.
In reaction 7, Propene-2-ol formed Ace as in reaction 5, then,
each Ace molecule broke at the C-C bond between C1 and C2. Each of the two AA
contained one C=O and one C-H bond on C2 and three C-H bonds on C3. Ethene
formed from two CH2 radicals bonding via a C=C bond.
For reaction 8, C-C bond cleavage between C1 and C2 resulted in
the TS that then undergoes two O-H abstractions and a H-H bond formation–one H2
molecule.
Reaction 9 involved −OH cleavage from C2, only to lose one
hydrogen atom, and bonding again to C2 with a double bond. The other Oxygen
atom also formed a double bond with C1 by losing a hydrogen atom. The resulting
two hydrogen atoms bonded to form H2. C1 and C2 both lose one
hydrogen atom, and thus, another H2 molecule is formed. MGA and two
H2 molecules resulted from reaction 9.
Reaction 10 contained two steps. In 10 (1), C3 broke off from C2
forming a −CH3 radical. In step 10 (2), each oxygen lost one
attached hydrogen and formed a double bond with carbon (two C=O bonds form).
This was possible since carbon lost one attached hydrogen (three C-H bonds
breaking). One C-H bond formed transforming CH3 into CH4.
Finally, the two hydrogen atoms left formed one H-H bond, or H2.
Fig. 11:
Reaction steps with transition states from ReaxAMS
3.4.2
Enthalpy and Gibbs free energy
The enthalpy plots for reactions (1), (2) and (3) are shown in Fig. 12,
while information for the other reactions can be found in the supplementary
data. The temperature presented is 1000 K, which is close to the experimental
condition where the temperature was highest and all products appeared, as seen
in Table 1. It can be noted from the enthalpy changes between reactants and
products, that only reactions 2 and 3 (and the second step of reaction 10) were
exothermic, while the other reactions were endothermic. It can be seen from the
free energy profiles in Fig. 13 that reactions 2, 3, 6, and 7(2) are exergonic reactions, as they
released free energy (Gfinal < Ginitial), and thus occurred without the need
of energy being added to the system (spontaneous reactions). The other
reactions are endergonic, notably reaction 1, and therefore it need energy to
proceed. Such energy initially originated from thermal heating and could also
be attributed to the high-energy products formed in previous endergonic reactions.
Fig. 12: Reactions (1), (2) and (3) enthalpies with respect to reaction
coordinate at 1000K.
Fig. 13: Reactions (2), (3),
(6) and (72) Gibbs free energies with respect to reaction coordinate
at 1000K.
4.
ReaxAMS kinetics
To
study favored pathways, small-scale computations were run on the individual
reactions presented in Fig. 3 Reaction pathways for the pyrolysis of PG under an inert
atmosphere.. For each reaction, the Gibbs free energy of the reactant and of
the transition state were found. The change in these Gibbs free energies
(∆G‡) was used to find the activation of each elementary reaction. Energy
calculations were found over a temperature range similar to experimental
temperatures (400K to 1000K at 100K increments). For each reaction, in order to
obtain an accurate Ea value that applies over this temperature range, and
rather than simply averaging ∆G‡ obtained at each T, the following
equation is applied:
(28)
Where kB is
the Boltzmann constant and h is the Plank constant.
The logarithmic form of the
Arrhenius equation was plotted for each k value, and the Arrhenius
parameters are found. The activation energies of the elementary reactions are
presented in Table 6. Note
that this section covers the activation energy of the reaction steps and not
the overall formation of certain products. The aim of this section is to cover
the preferred pathways rather than focus on rate of overall aldehyde formation.
Table 6: Ea values (J/mol) of
pathway reactions of PG Pyrolysis from ReaxAMS.
|
Mechanism
|
1
|
2
|
3
|
4
|
5
|
|
|
Ea (J/mol)
|
2.04E+05
|
8.71E+04
|
1.58E+05
|
2.82E+05
|
3.64E+05
|
|
|
Mechanism
|
6
|
7 (2)
|
8
|
9
|
10 (1)
|
10 (2)
|
|
Ea (J/mol)
|
7.50E+05
|
6.00E+05
|
2.91E+05
|
1.07E+06
|
3.23E+05
|
7.80E+05
|
As shown in Fig. 14, PG
can undergo cracking from either pathways (1), (8), (9), or (10). From the
activation energy barriers, pathway (1)—the dehydration of PG—has the lowest Ea
value and is thus the preferred reaction. It shows similar barrier heights to
the one computed by Laino et al. for the same reaction (∆G‡=66.5kcal/mol
or 2.78E+05 J/mol) [3] which was also considered the preferred pathway. This further
justifies why more dehydration products (PA, Ace, FA, and AA) were detected
over dehydrogenation products (GA and MGA) in experiments and ReaxFF-MD.
Pathway (2), prop-1-ol
formation from PO, had the lowest activation energy and was thus the favored
reaction in PG pyrolysis. The next lowest was pathway (3), the formation of
prop-2-ol from PO. After prop-1-ol formation, two pathways can occur: the formation
of PA—pathway (4)—and the formation of FA—pathway (6). From elementary reaction
activation energies, one can deduce that pathway (4) is the preferred reaction
from prop-1-ol, thus, PA formation is the dominant pathway. While Laino et
al. model PA forming directly from PO, our work proposes the intermediate
step of prop-1-ol formation. Nonetheless, similar barrier heights are observed
for this PA formation step: 2.45E+05 J/mol at 800K [3]
2.82E+05 J/mol (our work).
The same reasoning is applied
for reactions of prop-2-ol (pathways 5 and 7), where pathway (5)—formation of
Ace—is the preferred pathway over AA formation. Comparing Ace formation barrier
height from literature, Lano et al. DFT methods computed 2.80E+05 J/mol at 800K [3] and
our work computed 3.64E+05 J/mol, over the temperature range.
Fig. 14: Dominant and least preferred
pathways for PG pyrolysis.
5.
Conclusion
In
this study, experimental data was used to develop a kinetic model to estimate
the activation energies of PG degradation
over a temperature range of 160-670 °C (333-943 K), and under an
inert N2 atmosphere where the dominant degradation pathway was found to be the
dehydration of PG to give PA and Ace. We employed high-temperature MD
simulations to investigate further these findings and the main products
observed were propanal, acetone, formaldehyde, acetaldehyde, glyoxal and methyl
glyoxal. FA was a dominant product that could form from several cracking
reactions. PA formation reaction had the lowest activation energy and proved to
be the dominant pathway. Elementary reactions were studied in ReaxAMS to
determine energies and transition states. Rate-limiting steps and favored
pathways were identified in addition to bond formation and breakage mechanisms.
PG pyrolysis favored the pathway of propylene oxide to prop-1-ol and prop-2-ol.
These are exergonic and exothermic, and their products lead to the direct formation
of PA and Ace, corroborating experimental
and MD findings. Similar to ReaxFF-MD and experimental results, the formation
of PA and Ace respectively were the dominant pathways. In addition, the
formations of FA and AA respectively, were found to be exergonic while all
elementary reactions that have PG as reactant were endothermic and endergonic.
Future work will concentrate on the oxidative degradation of PG to account for
the formation of oxygenated carbonyls observed in ECIG emissions. Other e-liquids
and flavors could be studied using the same methodology.
Acronyms
A0
pre-exponential factor
Ea
activation energy
TMax
maximum temperature
AA
acetaldehyde
Ace
acetone
DFT
Density Functional Theory
EFF
empirical force fields
FA
formaldehyde
GA
glyoxal
GUI
Graphical User Interface
MD
molecular dynamics
MGA
methylglyoxal
NEB
The Nudged Elastic Band NVT Number-Volume-Temperature
P1
propen-1-ol
P2
propen-2-ol
PA propanal
PG Propylene Glycol
PO Propylene Oxide
QM quantum mechanics
R ideal gas constant
ReaxAMS
Amsterdam Modeling Suite
ReaxFF-MD
Molecular Dynamics Reactive Force Field
SCM
Software for Chemistry and Materials
TS
transition state
UFF
universal force field
Acknowledgements
This
work was supported by the National Institute on Drug Abuse of the National
Institutes of Health (NIH grant number P50DA036105) and the Center for Tobacco
Products of the U.S. Food and Drug Administration (FDA). The content is solely
the responsibility of the authors and does not necessarily represent the
official views of NIH or FDA.
Conflict
of Interest
There
is no conflict of interest.
Supporting
Information
Applicable.
References
[1] M. A.
McGowan, A. Scheman, S. E. Jacob, Propylene glycol in contact dermatitis: a
systematic review, Dermatitis, 2018, 29, 6-12, doi: 10.1097/DER.0000000000000307.
[2] W. J.
RossiterJr, M. Godette, P. W. Brown, K. G. Galuk, An investigation of the
degradation of aqueous ethylene glycol and propylene glycol solutions using ion
chromatography, Solar Energy Materials, 1985, 11, 455-467, doi: 10.1016/0165-1633(85)90016-4.
[3] T.
Laino, C. Tuma, P. Moor, E. Martin, S. Stolz, A. Curioni, Mechanisms of
propylene glycol and triacetin pyrolysis, The Journal of Physical Chemistry A, 2012, 116,
4602-4609, doi: 10.1021/jp300997d.
[4] Y.
Zhao, L. Ding, J. Liu, H. Yang, X. Yin, S. Tuo, Q. Zhang, Pyrolysis
characteristics of different humectant at high temperature, Asian Journal of
Chemistry, 2014, 26, 4893-4896, doi: 10.14233/ajchem.2014.16580.
[5] S. Han,
H. Chen, X. Zhang, T. Liu, Y. Fu, Levels of selected groups of compounds in
refill solutions for electronic cigarettes, Nicotine & Tobacco Research, 2016, 18,
708-714, doi: 10.1093/ntr/ntv189.
[6] A.
Breland, E. Soule, A. Lopez, C. Ramôa, A. El-Hellani, T. Eissenberg, Electronic
cigarettes: what are they and what do they do?, Annals of the New
York Academy of Sciences, 2017, 1394, 5-30, doi: 10.1111/nyas.12977.
[7] A.
EL-Hellani, R. Salman, R. El-Hage, S. Talih, N. Malek, R. Baalbaki, N.
Karaoghlanian, R. Nakkash, A. Shihadeh, N. A. Saliba, Nicotine and carbonyl
emissions from popular electronic cigarette products: correlation to liquid
composition and design characteristics, Nicotine & Tobacco Research, 2018, 20,
215-223, doi: 10.1093/ntr/ntw280.
[8] K.
Bekki, S. Uchiyama, K. Ohta, Y. Inaba, H. Nakagome, N. Kunugita, Carbonyl
compounds generated from electronic cigarettes, International Journal
of Environmental Research and Public Health, 2014, 11, 11192-11200, doi:
10.3390/ijerph111111192.
[9] N. L.
Benowitz, J. B. Fraiman, Cardiovascular effects of electronic cigarettes, Nat. Rev. Cardiol., 2017, 14,
447-456, doi: 10.1038/nrcardio.2017.36.
[10] G.
Zhang, Z. Wang, K. Zhang, R. Hou, C. Xing, Q. Yu, E. Liu, Safety assessment of
electronic cigarettes and their relationship with cardiovascular disease, International Journal
of Environmental Research and Public Health, 2018, 15, 75, doi:
10.3390/ijerph15010075.
[11] E.
Díaz, M. E. Sad, E. Iglesia, Homogeneous oxidation reactions of propanediols at
low temperatures, ChemSusChem, 2010, 3, 1063-1070, doi: 10.1002/cssc.201000142.
[12] C.
R. Ayre, R. J. Madix, The adsorption and reaction of 1, 2-propanediol on
Ag(110) under oxygen lean conditions, Surface
Science, 1994, 303, 279-296, doi:
10.1016/0039-6028(94)90776-5.
[13] M. A.
Salaev, O. K. Poleshchuk, O. V. Vodyankina, Propylene glycol oxidation over
silver catalysts: a theoretical study, Journal of Molecular Catalysis A: Chemical, 2016, 417,
36-42, doi: 10.1016/j.molcata.2016.03.011.
[14] P.
Wang, W. Chen, J. Liao, T. Matsuo, K. Ito, J. Fowles, D. Shusterman, M.
Mendell, K. Kumagai, A device-independent evaluation of carbonyl emissions from
heated electronic cigarette solvents, PLoS One, 2017, 12, e0169811, doi: 10.1371/journal.pone.0169811.
[15] N. A.
Saliba, A. El Hellani, E. Honein, R. Salman, S. Talih, J. Zeaiter, A. Shihadeh,
Surface chemistry of electronic cigarette electrical heating coils: effects of
metal type on propylene glycol thermal decomposition, Journal of Analytical
and Applied Pyrolysis, 2018, 134, 520-525, doi: 10.1016/j.jaap.2018.07.019.
[16]
“Formaldehyde and Cancer Risk,” Natl. Cancer Inst. (2011).
[17]
“Acetaldehyde: Hazard Summary,” Environ. Prot. Agency (2016).
[18]
“National Toxicology Program, 12th Report on Carcinogens,” U.S. National
Library of Medicine (2011).
[19]
“Toxicological Review of Propionaldehyde,” U.S. Environmental Protection Agency
(2008).
[20] B.
Urych, Determination of kinetic parameters of coal pyrolysis to simulate the
process of underground coal gasification (UCG), Journal of
Sustainable Mining, 2014, 13, 3-9, doi: 10.7424/jsm140102.
[21] T. P.
Senftle, S. Hong, M. M. Islam, S. B. Kylasa, Y. Zheng, Y. K. Shin, C.
Junkermeier, R. Engel-Herbert, M. J. Janik, H. M. Aktulga, T. Verstraelen, A.
Grama, A. C. T. van Duin, The ReaxFF reactive force-field: development,
applications and future directions, Npj Computational Materials, 2016, 2, 15011, doi:
10.1038/npjcompumats.2015.11.
[22] S.
Talih, Z. Balhas, R. Salman, R. El-Hage, N. Karaoghlanian, A. El-Hellani, M.
Baassiri, E. Jaroudi, T. Eissenberg, N. Saliba, A. Shihadeh, Transport
phenomena governing nicotine emissions from electronic cigarettes: model
formulation and experimental investigation, Aerosol Science and Technology, 2017, 51,
1-11, doi: 10.1080/02786826.2016.1257853.
[23] M. Al
Rashidi, A. Shihadeh, N. A. Saliba, Volatile aldehydes in the mainstream smoke
of the narghile waterpipe, Food and Chemical Toxicology, 2008, 46, 3546-3549, doi:
10.1016/j.fct.2008.09.007.
[24]
"Reaxff User Manual," Software for Chemistry and Materials (2002).
[25] Reaxff: reactive
md with graphical interface and analysis tools,
https://www.scm.com/product/reaxff/.
[26]
"Included force fields," Software for Chemistry and Materials (2021).
https://www.scm.com/doc/ReaxFF/Included_
Forcefields.html.
[27]
Q.-D. Wang, J.-B. Wang, J.-Q. Li, N.-X. Tan, X.-Y. Li, Reactive molecular
dynamics simulation and chemical kinetic modeling of pyrolysis and combustion
of n-dodecane, Combustion and Flame, 2011,
158, 217-226, doi: 10.1016/j.combustflame.2010.08.010.
[28] T.
Zhao, T. Li, Z. Xin, L. Zou, L. Zhang, A ReaxFF-based molecular dynamics
simulation of the pyrolysis mechanism for polycarbonate, Energy & Fuels, 2018, 32,
2156-2162, doi: 10.1021/acs.energyfuels.7b03332.
[29] L.
Martínez, R. Andrade, E. G. Birgin, J. M. Martínez, PACKMOL: A package for
building initial configurations for molecular dynamics simulations, Journal of
Computational Chemistry, 2009, 30, 2157-2164, doi: 10.1002/jcc.21224.
[30] T.
Morishita, Fluctuation formulas in molecular-dynamics simulations with the weak
coupling heat bath, The Journal of Chemical Physics, 2000, 113,
2976-2982, doi: 10.1063/1.1287333.
[31] S.
Arvelos, O. AbrahãoJr, C. Eponina Hori, ReaxFF molecular dynamics study on the
pyrolysis process of cyclohexanone, Journal of Analytical and Applied Pyrolysis, 2019, 141,
104620, doi: 10.1016/j.jaap.2019.05.009.
[32] P. G.
Shields, M. Berman, T. M. Brasky, J. L. Freudenheim, E. Mathe, J. P. McElroy,
M.-A. Song, M. D. Wewers, A review of pulmonary toxicity of electronic
cigarettes in the context of smoking: a focus on inflammation, Cancer Epidemiology, Biomarkers &
Prevention, 2017, 26, 1175-1191, doi: 10.1158/1055-9965.epi-17-0358.
[33] R. P.
Jensen, R. M. Strongin, D. H. Peyton, Solvent chemistry in the electronic
cigarette reaction vessel, Scientific Reports, 2017, 7, 42549, doi: 10.1038/srep42549.
[34] A. S. Perlin,
Glycol-cleavage oxidation, Advances in Carbohydrate Chemistry and Biochemistry, 2006, 60,
183-250, doi: 10.1016/S0065-2318(06)60005-X.
Author
Information
Christina
AlGemayel is a research assistant at the Aerosol Research Laboratory at the
American University of Beirut (AUB) where she obtained her B.E. and M.E. in
chemical engineering. Her research focuses on molecular dynamics, kinetic
modeling, and analytical chemistry.
Dr.
Ahmad El Hellani Assistant Professor in the Environmental
Health Sciences, College of Public Health, The Ohio State University. He is an
Organic/Analytical chemist that went from fundamental organometallic chemistry
research to be interested in the assessment of human exposure to toxicants from
various sources, especially tobacco products. He studies tobacco emissions in a
controlled analytical lab setting using smoking machines, in a clinical lab
setting using real-time sampling devices, and in a chamber lab setting for the
assessment of secondhand emissions. His work is interdisciplinary by nature,
and he is currently conducting clinical studies to assess tobacco product abuse
liability
and toxicity.
Rola
Salman obtained her BS in Biochemistry from Lebanese University. She
joined the Aerosol Research Lab in 2008 as a research scientist. Her research
focuses on measuring toxicant yields from tobacco and non-tobacco products in
waterpipe, conventional cigarette, and different designs of electronic
cigarettes.
Najat A. Saliba is currently a member
of the Parliament of the Lebanese Government. She is also a Professor in
Chemistry at the American University of Beirut. Saliba leads several locally
relevant and globally important projects related to understanding the chemistry
of inhalable tobacco and non-tobacco smoke and atmospheric aerosols. She is the
co-founder and executive director of Khaddit Beirut and the Director and
founder of the Environment Academy (EA). In 2021, she was invited to serve as a
Member of the Scientific Board of the International Basic Science Programme
(IBSP) at UNESCO and as a Chair of a Technical Advisory Group on the Global Air
Quality Platform and Health (GAPH-TAG) at WHO. She was also nominated by
Apolitical’s the 12 most influential people in climate justice and 100 most
influential people in Gender Policy in 2021. She received the 2019
L’Oreal-UNESCO International Award for Women in Science, the National Order of
the Cedar from the President of the Lebanese Republic, the Honorary Cedar
Shield from the Speaker of the Parliament of Lebanon and the Paul Harris Fellow
Pin from the Rotary Club Beirut Cedars. Also, in 2019 she was voted among the
top 100 most influential women by BBC. In 2016 she received the Lebanese
National Council for Scientific Research Award in the Environmental Category.
Alan Shihadeh is Dean of the Maroun Semaan
Faculty of Engineering and Architecture at the American University of Beirut
(AUB). Shihadeh earned his undergraduate degree from the University of Texas at
Austin and his doctorate in mechanical engineering at the Massachusetts
Institute of Technology (MIT) in the field of combustion in 1998. He taught at
Birzeit University as a Fulbright fellow before joining AUB as an assistant
professor of mechanical engineering in 2000. Shihadeh is also Project Director
at the Virginia Commonwealth University Center for the Study of Tobacco
Products, founding director of the Aerosol Research Laboratory, and co-director
of the Collaborative for the Study of Inhaled Atmospheric Aerosols (CARS).
Dr. Joseph Zeaiter received his PhD in Chemical
Engineering (Process Systems Engineering) from the University of Sydney and
joined the process industry, as an advanced process control /senior consultant
working on projects in oil and gas, petrochemicals, power plants and mining. He
is currently an associate professor in Chemical Engineering at the American
University of Beirut (AUB) where his research interests are focused on
catalytic pyro-gasification, hydrogen from natural gas conversion, and thermal
solar energy for renewable fuel production.
