ARTICLE IN PRESS

JID: PROCI [mNS; September 16, 2020;13:8 ]

Available online at www.sciencedirect.com

Proceedings of the Combustion Institute 000 (2020) 1–8 www.elsevier.com/locate/proci

Kinetics for the hydrolysis of Ti(OC 3

H 7

) 4

: A molecular

dynamics simulation study

Jili Wei a , b , Alireza Ostadhossein

a , Shuiqing Li b , Matthias Ihme

a , ∗

a Department of Mechanical Engineering, Stanford University, Stanford, CA 94305, United States b Key Laboratory for Thermal Science and Power Engineering of Ministry of Education, Department of Energy and Power

Engineering, Tsinghua University, Beijing 100084, China

Received 5 November 2019; accepted 28 June 2020 Available online xxx

Abstract

Using molecular dynamics (MD) simulations in conjunction with a reactive force field method, the chemical kinetics of the hydrolysis of titanium tetraisopropoxide (Ti(OC 3 H 7 ) 4 , TTIP) at high-temperature conditions is investigated. The MD simulations allow for presenting the complete dynamic process of the TTIP conversion at the atomic level. The rate constant of TTIP hydrolysis at 1 atm is estimated to be k = 1 . 23 × 10 14 × exp ( −11 , 323 /T ( K) ) mo l −1 c m

3 s −1 using a second-order reaction model. On the basis of Ti-containing intermediate species profiles, the evolutions of the main decomposition products during the TTIP hydrolysis are identified and key reaction pathways are elucidated. The results show that the clusters are formed before TiO 2 molecules are observed. During the decomposition, Ti-containing species with one or two C

–O bonds and carbon-free species with more than two Ti –O bonds are formed and undergo two

separate pathways. One is the combination via the formation of Ti –O

–Ti bridges, forming early clusters that serve as precursors for large TiO 2 nanoparticles. The other is the further decomposition to smaller molecules such as TiO 2 that participate in the subsequent cluster formation. Interactions between Ti and O atoms in

the cluster stabilize the large structure through the abstraction of water and –C x H y groups. © 2020 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Titanium tetraisopropoxide; Reaction mechanism; Molecular dynamics simulation; Reactive force field

1. Introduction

Titanium dioxide (TiO 2 ) nanoparticles are en-countered in a wide range of industrial applicationsthat include photocatalysis, pigments, semiconduc-tors and gas sensors [1] . The chemical and physicalproperties of nanomaterials are strongly influencedby the size, shape and crystalline phase. Therefore,

∗ Corresponding author. E-mail address: mihme@stanford.edu (M. Ihme).

https://doi.org/10.1016/j.proci.2020.06.345 1540-7489 © 2020 The Combustion Institute. Published by Elsev

Please cite this article as: J. Wei, A. Ostadhossein and S. Li et al.,dynamics simulation study, Proceedings of the Combustion Ins

understanding the reaction mechanism and nucle- ation behavior is essential toward enabling the con- trol of the particle growth during the synthesis pro- cess.

Gas-phase flame synthesis, one of the most scalable and economical technologies for produc- ing well-controlled nanomaterials, combines ad- vantages of a continuous single-step synthesis pro- cess with high throughput and fast processing time [2 , 3] . In this process, the high-temperature flame environment can be controlled to cover a wide range of operating conditions (e.g., temperature,

ier Inc. All rights reserved.

Kinetics for the hydrolysis of Ti(OC3H7)4: A molecular titute, https://doi.org/10.1016/j.proci.2020.06.345

2 J. Wei, A. Ostadhossein and S. Li et al. / Proceedings of the Combustion Institute xxx (xxxx) xxx

ARTICLE IN PRESS

JID: PROCI [mNS; September 16, 2020;13:8 ]

r

O

r

op

H

a

t

f

l

f

m

t

f

H

p

b

t

k

s

t

t

T

s

d

p

o

o

r

T

a

f

le

r

t

t

a

b

n

T

w

3

v

r

e

T

s

fle

p

m

t

s

eactant concentration, residence time, etc.) [4] .ver recent years, several studies have been car-

ied out to experimentally examine the synthesisf TiO 2 in flames focusing on the influence of ressure, temperature, and flame geometry [5 –7] .owever, due to the fact that chemical reactions

nd particle nucleation proceed on extremely shortime-scales in typical flame environments, only aew experimental studies reported the detailed evo-ution from precursors to small clusters [3] . There-ore, the gas-phase kinetics and precursor reaction

echanisms at the early stage of the particle forma-ion remain incompletely understood and requireurther investigations [8] .

Titanium tetraisopropoxide (TTIP, Ti(OC 3 7 ) 4 ) is commonly used as precursor for the

roduction of TiO 2 in flames with chlorine-freeyproducts and is safer in handling compared toitanium tetrachloride (TiCl 4 ). While the detailedinetics of TiCl 4 oxidation has been extensivelytudied both theoretically [9 , 10] and experimen-ally [11] , the detailed reaction process leading tohe TiO 2 formation from TTIP is less explored.wo main global chemical pathways have beenuggested for the TTIP conversion, namely thermalecomposition and hydrolysis. Thermal decom-osition takes place above 573 K and hydrolysisccurs at a temperature as low as 380 K [12] . Theverall reaction of the thermal decomposition waseported as [13] ,

TIP → TiO 2 + 4C 3 H 6 + 2H 2 O, (1)

nd a first-order rate constant was estimatedrom measuring the temperature-dependent propy-ene formation with a value of k = 3 . 96 × 10 5 ×xp ( −8479 . 7 /T ( K) ) s −1 . A reaction rate value de-ived from the measurement of the size distribu-ion of sub-2 nm TiO 2 particles formed throughhe thermal decomposition of TTIP showed goodgreement with these results [14] .

The gas-phase hydrolysis of TTIP was studiedy Seto et al. [15] by considering the global mecha-ism,

TIP + 2H 2 O → TiO 2 + 4C 3 H 7 OH, (2)

ith a first-order rate constant given by k = × 10 15 × exp ( −1013 . 9 /T ( K) ) s −1 without pro-iding further details on the evaluation of thisate constant in the original reference. Shmakovt al. [16] proposed a bimolecular reaction,TIP + H 2 O → Product, to describe the primary

tage of the TTIP decomposition in H 2 /O 2 /Arames with a rate constant of k = 2 × 10 12 ×xp ( −6140 /T ( K) ) mo l −1 c m

3 s −1 . Their results sup-ort that the hydrolysis reaction pathway is theain contributor to the conversion of TTIP at high

emperatures. Most modeling work has been done by con-

idering an overall one-step reaction without

Please cite this article as: J. Wei, A. Ostadhossein and S. Li et al.,dynamics simulation study, Proceedings of the Combustion Ins

intermediate species [8 , 17] . However, detailedchemical mechanisms and rate constants of the elementary reactions are crucial towardsthe improved modeling of processes leading to thenanoparticle formation. Several attempts have beenmade focusing on the kinetic mechanisms underly-ing the TTIP decomposition. The decompositionof TTIP was assumed to proceed by successiveβ-hydride elimination of propene and the conden-sation polymerization for waterless atomic layerdeposition [18] . Buerger et al. [19] derived a ther-modynamically consistent chemical mechanism forthe thermal decomposition of TTIP with Ti(OH) 4taken as the final stable product.

While progress has been made on examiningthe thermal decomposition of TTIP, the hydrol-ysis mechanisms and nucleation kinetics of nan-oclusters remain elusive. This is largely attributedto the absence of quantitative measurement tech-niques with high temporal resolution to resolve theultrafast reaction rates. Previous classical molec-ular dynamics (MD) studies have explored thehigh-temperature dynamics of TiO 2 nanoparticles,including particle-particle interaction [20] , colli-sion [21] , and coalescence [22] . By employing theReaxFF reaction force field method, molecular dy-namics simulations serve as a viable method tostudy the hydrolysis process prior to the nanoparti-cle formation at the atomic level.

The objective of this study is to obtain in-sight into the initial stage of the decompositionreaction of TTIP and the subsequent nanoparti-cle formation at high temperatures. For this pur-pose, ReaxFF MD simulations are performed for arange of temperatures and pressures to determinethe rate constant and identify elementary reactionpathways. By analyzing the evolution of the mainspecies and observing the cluster formation pro-cess, growth mechanisms are deduced. The remain-der of this manuscript has the following structure.The MD simulation method and computational de-tails are presented in Section 2 . Simulation resultsare discussed in Section 3 , and the paper finishedby drawing conclusions for the kinetics of TTIP hy-drolysis and cluster growth.

2. Computational method

The ReaxFF potential force-field [23] , imple-mented in the LAMMPS package [24] , is usedfor the MD simulations. The Ti/C/H/O ReaxFFforce field in the present study has been employedto investigate MXenes (i.e., pristine Ti 3 C 2 (OH) 2 )[25 , 26] , and the interaction parameters of theTi/O/H system were originally developed and vali-dated for modeling the reaction of water with TiO 2surfaces [27] . Here, the force field was validated bycalculating the enthalpy of formation, �H

◦f , 298 . 15K

,for TTIP by considering the unimolecular reac-tion ( 1 ), and all standard enthalpies of formation

Kinetics for the hydrolysis of Ti(OC3H7)4: A molecular titute, https://doi.org/10.1016/j.proci.2020.06.345

J. Wei, A. Ostadhossein and S. Li et al. / Proceedings of the Combustion Institute xxx (xxxx) xxx 3

ARTICLE IN PRESS

JID: PROCI [mNS; September 16, 2020;13:8 ]

of the reference species are taken from the NISTdatabase [28] . The computed value of �H

◦f , 298 . 15K

is −1497 kJ/mol, which is in good agreement com-pared to the literature value of −1509 ± 9.3 kJ/mol[19] .

In the present study, 410 water molecules and27 TTIP molecules in the gas phase with an initialwater/TTIP ratio of 15 are simulated inside three-dimensional cubic periodic systems to study thehydrolysis reaction. For the simulations discussedhere, three pressures with P = { 1 , 7 , 238 } atm areconsidered as the initial system pressure. The initialpressure of 238 atm is high enough to ensure thatthe complete process takes place within a nanosec-ond timescale. The size of the simulation box foreach temperature and pressure is calculated basedon the ideal gas law. To delineate hydrolysis fromthermal decomposition, separate simulations areconducted in which all water molecules are replacedby 27 TTIP molecules to maintain the total atomnumbers at 1 atm.

All MD simulations are performed under NVTconditions (constant number of atoms, volume,and temperature) to study the hydrolysis of TTIPwith fixed temperatures between 1500 K and3000 K for each pressure. The NVT ensemble isemployed to examine the influence of temperatureon the reaction rate and obtain rate constants of the primary reaction. Every simulation begins withan energy minimization using a conjugate gradientalgorithm to eliminate artifacts of the initializa-tion. For each case, the atomic configuration isequilibrated at target temperatures in the NVTensemble. The system temperature is controlled bythe Nosé–Hoover thermostat with a damping con-stant of 10 fs. For ReaxFF simulations, a smallertime step is preferred because the atom chargesand bond orders can change at every time step [29] .Here, a time step of 0.1 fs is adopted to ensure thatcollisions and reactions evolve smoothly. ReaxFFallows for bond formation and dissociation duringMD simulations by employing a bond-order/bond-energy relationship [30] . A 0.3 bond-order cutoff is employed to identify molecules and analyze thespecies formed during the simulation.

3. Results and discussions

3.1. Global kinetic parameters

Fixed-temperature NVT MD simulations be-tween 1500 K and 3000 K are performed for threedifferent pressures to study the reaction pathwaysof TTIP hydrolysis. Figure 1 a illustrates the initialconfiguration of the system inside an 8.55 nm pe-riodic cube from a simulation at 2500 K and 238atm. All molecules are randomly placed and energyminimized subsequently. Figure 1 b and c showssnapshots at the simulation time of 500 ps and1000 ps, respectively. At 500 ps, the TTIP molecules

Please cite this article as: J. Wei, A. Ostadhossein and S. Li et al.,dynamics simulation study, Proceedings of the Combustion Ins

are fully consumed and Ti-containing intermedi- ates are formed. The inset depicts the structure of a representative Ti-containing species from each

snapshot. A cluster containing 26 Ti atoms is ob- served in Fig. 1 c. We proceed with our discussion

by focusing on the formation process of clusters, including the detailed analysis of the chemical re- action pathways and subsequent nucleation.

Figure 2 represents the number of TTIP ( N T T IP ) and water ( N H 2 O

) molecules as a function of the simulation time obtained from NVT-MD simula- tions at a pressure of 238 atm and temperature ranging from 1500 K to 3000 K. The decomposi- tion of TTIP occurs first, followed by the absorp- tion of water with decomposition products. Here, these two processes are combined into one primary reaction path. The ratio of the consumed TTIP and

water molecules is approximately unity, which sup- ports an overall second-order reaction model dur- ing the primary stage. The kinetic parameters of the Arrhenius equation were calculated using the tem- perature dependence of the concentration profiles of TTIP and water following the second-order re- action model,

k =

1 t ( [ H 2 O ] 0 − [ TTIP ] 0 )

ln

([ TTIP ] 0 [ H 2 O ] [ H 2 O ] 0 [ TTIP ]

)(3)

where [TTIP] and [H 2 O] denote the concentra- tion of TTIP and H 2 O at time t and the sub- script 0 refers to initial conditions. For each

simulation point, three MD simulations are per- formed. From the Arrhenius plot in Fig. 3 , Ar- rhenius parameters are determined through the linear equation, ln k = ln A − ( E a /RT ) , where A

is the pre-exponential factor, E a is the activa- tion energy and R is the gas constant. The calculated activation energy is consistent at the three pressures with the standard deviation of 0.24 kJ/mol. At the ambient pressure of 1 atm, the rate constant has a value of k = 1 . 23 × 10 14 ×exp ( −11 , 323 /T ( K) ) mo l −1 c m

3 s −1 with a corre- sponding activation energy of E a = 94.1 kJ/mol. The result shows good agreement for the high- temperature regime with experimental results by Shmakov et al. [16] (black dashed line in Fig. 3 ) and

is also consistent with the rate parameters from the water-assisted surface deposition of TTIP vapor [31] . In comparison with thermal decomposition, the first-order rate constant for hydrolysis is also

calculated from k = −( 1 /t ) ln ( [ TTIP ] / [ TTIP ] 0 ) (see Fig. S3 in the Supplementary Material).

3.2. Time evolution of main species

Simulation results at 2500 K and 238 atm are considered to examine the kinetics of the TTIP de- composition. In addition, simulation of a large sys- tem containing 216 TTIP and 3280 water molecules at the same condition is performed to conduct statistical analysis for the species evolution. The species involved in the formation pathways of TTIP

Kinetics for the hydrolysis of Ti(OC3H7)4: A molecular titute, https://doi.org/10.1016/j.proci.2020.06.345

4 J. Wei, A. Ostadhossein and S. Li et al. / Proceedings of the Combustion Institute xxx (xxxx) xxx

ARTICLE IN PRESS

JID: PROCI [mNS; September 16, 2020;13:8 ]

Fig. 1. Snapshots from an MD simulation at 2500 K and 238 atm containing 27 TTIP molecules and 410 water molecules showing (a) initial energy-minimized configuration, (b) t = 500 ps, and (c) t = 1000 ps (for color figure, gray = titanium, red = oxygen from TTIP, blue = carbon, white = hydrogen from TTIP, yellow = oxygen from water, cyan = hydrogen from

water).

Fig. 2. Temporal evolution of TTIP and water molecules at a pressure of 238 atm and temperature ranging from

1500 K to 3000 K.

h

c

tT

i

b

o

a

m

c

D

r

i

b

v

s

Fig. 3. Arrhenius plot from MD simulations showing the temperature dependence of the reaction rate constant for hydrolysis (open symbols). Experimental results from

Shmakov et al. [16] are included for reference (solid sym- bols). Solid lines correspond to fitting formulas while the black dashed line fits the literature data above 1000 K.

ydrolysis are complicated and only major Ti-ontaining species are considered here. Based onhe structure of TTIP, the hydrolysis reaction of TIP can be divided into four stages, each involv-

ng the decomposition of one –O x C 3 H y ( x = {0, 1})ranch. The consumption of TTIP and the yieldsf major Ti-containing species during the reactionre shown in Fig. 4 .

Over the duration of the simulation, the pri-ary species is Ti(O

• )(OC 3 H 7 ) 3 due to the disso-iation of one C

–O bond at the first stage. PreviousFT calculations have shown that the energy bar-

ier for the reaction via the C

–O bond dissociations lower by 2.8 kcal/mol compared to that for Ti –Oond dissociation [16] , which supports our obser-ation. This initial hydrolysis step is followed by aecond stage that involves either the breaking of a

Please cite this article as: J. Wei, A. Ostadhossein and S. Li et al.,dynamics simulation study, Proceedings of the Combustion Ins

Ti –O bond forming TiO(OC 3 H 7 ) 2 or the breakingof a C

–O bond forming Ti(O

• )(OH)(OC 3 H 7 ) 2 . Thetransfer of primary or secondary hydrogen fromcarbon to oxygen can lead to the release of a propy-lene molecule instead of iso-propyl during the reac-tion, generating hydroxyl rather than O radicals. Atthe third stage, the dissociation of the C

–O bondis thermodynamically preferred for TiO(OC 3 H 7 ) 2leading to the formation of TiO(OH)(OC 3 H 7 ) andTiO(O

• )(OC 3 H 7 ). While for Ti(O

• )(OH)(OC 3 H 7 ) 2 ,the cleavage of Ti –O or C

–O leads to the formationof three species as shown in Fig. 4 c. The mole frac-tion of TiO(OH)(OC 3 H 7 ) increases slightly around100 ps due to the decomposition of species from thesecond stage and the transfer of one hydrogen atomfrom a nearby H 2 O to TiO(O

• )(OC 3 H 7 ).

Kinetics for the hydrolysis of Ti(OC3H7)4: A molecular titute, https://doi.org/10.1016/j.proci.2020.06.345

J. Wei, A. Ostadhossein and S. Li et al. / Proceedings of the Combustion Institute xxx (xxxx) xxx 5

ARTICLE IN PRESS

JID: PROCI [mNS; September 16, 2020;13:8 ]

Fig. 4. TTIP consumption and yields of Ti-containing in- termediate species during NVT simulation of TTIP/water mixture at a temperature of 2500 K and pressure of 238 atm. Species containing (a) more than 6 carbon atoms; (b) 6 carbon atoms; (c) 3 carbon atoms; and (d) no carbon atoms. (e) Clusters with more than one Ti atom.

Previous studies on the thermal decompositionof TTIP indicated that although Ti(OH) 4 is of-ten identified as the thermodynamically most stablebyproduct at temperatures below 1250 K, speciessmaller than Ti(OH) 4 that contain Ti –OH groupsand Ti = O bonds are more important at high tem-peratures [19 , 32] . Our simulations show that themole fraction of Ti(OH) 4 is significantly lower thanthat of other smaller species at 2500 K. The ma-jor Ti-containing species at the fourth stage in-

Please cite this article as: J. Wei, A. Ostadhossein and S. Li et al.,dynamics simulation study, Proceedings of the Combustion Ins

clude TiO(OH) 2 , TiO(O

• )(OH), Ti •• (OH) 2 , TiO 2 , Ti • (OH) 3 and Ti • O(OH), as shown in Fig. 4 d.

It is worth noting that clusters, containing more than one Ti atom, appear before the first TiO 2 molecule is formed, as depicted in Fig. 4 e. These nascent clusters contain propyl groups that are re- leased during the subsequent nucleation. In order to examine the cluster formation and growth pro- cess completely, we consider the smaller system

with 27 TTIP molecules in the next section in more detail.

3.3. Formation of nanoclusters

The conversion from precursors to TiO 2 nanoparticles is usually separated into two

processes, namely the chemical reaction from

precursors to TiO 2 molecule and the subsequent nucleation of single TiO 2 molecules to (TiO 2 ) n clusters. A semi-empirical reaction scheme was proposed earlier to describe the decomposition of TTIP, leading to the formation of Ti(OH) 4 as a stable intermediate and ultimately to a TiO 2 molecule [16] . Extensive computational studies on (TiO 2 ) n clusters have been performed, starting with TiO 2 monomers [33] . By measuring charged

clusters with an atmospheric pressure interface time-of-flight (APi-TOF) during flame synthesis using TTIP as the precursor, Fang et al. [1] ob- served the formation of intermediate organic species in the form of Ti n C x O y H z prior to com- plete conversion to TiO 2 . Here, by performing ReaxFF simulations, we are able to consider the conversion process as a whole and gain insight into

the underlying mechanisms. The initial steps of TiO 2 formation after TiCl 4

hydrolysis have been studied [9 , 34] . Partially oxy- genated Ti-containing molecules (Ti x O y Cl z (OH) w ) are formed that serve as precursors to TiO 2 nanoparticles. Energy analysis of hydrated (TiO 2 ) n clusters has shown that the dehydration reac- tion of two spatially available Ti(OH) groups is thermodynamically favored and enables further aggregation of clusters, which can be expressed

as 2(–Ti(OH)) → –TiOTi– + H 2 O [35] . A similar reaction pathway is observed from our MD simu- lation for TTIP hydrolysis. Figure 5 illustrates one example of the formation of a Ti 2 O 4 cluster with

Ti • O(OH) and TiO(OC 3 H 7 ) 2 as initial species. The reaction is initiated by forming the Ti –O

–Ti bridge to combine the two molecules. Subsequently, Ti and

O atoms tend to interact with nearby O or Ti atoms, resulting in the distortion and reconstruction of the whole molecule [36] . After releasing two –C 3 H y

groups and one water molecule, Ti 2 O 4 is generated

having a thermodynamically stable structure. The detailed formation process of the large

cluster, shown in Fig. 1 c, is illustrated through

the atomistic snapshots in Fig. 6 . Each row starts with an early cluster containing two Ti atoms and

only the stable structure of the cluster is shown

Kinetics for the hydrolysis of Ti(OC3H7)4: A molecular titute, https://doi.org/10.1016/j.proci.2020.06.345

6 J. Wei, A. Ostadhossein and S. Li et al. / Proceedings of the Combustion Institute xxx (xxxx) xxx

ARTICLE IN PRESS

JID: PROCI [mNS; September 16, 2020;13:8 ]

Fig. 5. Ti 2 O 4 formation pathway observed from the ReaxFF NVT-MD simulation at 2500 K and 238 atm.

Fig. 6. Snapshots of cluster formation over 1000 ps.

w

t

g

n

a

h

g

[

a

w

a

s

t

t

a

t

c

t

t

ith the corresponding formula at each step. Athe early stage of the cluster formation, the clusterrows slowly and subsequently combines withearby small Ti-containing molecules one afternother. The intermediates produced from TTIPydrolysis aggregate into large clusters, showingood agreement with previous experimental results1] . It is noted that small species containing Ti = Ond Ti –OH play an important role. For speciesith Ti = O, e.g., TiO 2 , the double bond breaksnd forms Ti –O

–Ti to combine two molecules. Forpecies involving a Ti –OH group, e.g., Ti •• (OH) 2 ,he molecule has one or more radical sites andends to form stable bonds through recombination.

To stabilize the structure, the unsaturated Tind O atoms on the outer surface of the clus-er either continue to aggregate with other Ti-ontaining species from the surrounding or restruc-ure the surface. In the meantime, the abundant wa-er molecules in the surrounding environment will

Please cite this article as: J. Wei, A. Ostadhossein and S. Li et al.,dynamics simulation study, Proceedings of the Combustion Ins

adsorb on the clusters molecularly or dissociatively.For molecular adsorption, the oxygen atom of thewater molecule binds to an under-coordinated Tiatom, as seen from the structure of Ti 26 O 31 H 6 inFig. 6 . While in dissociative adsorption, the watermolecule forms a hydrogen bond with a surface Oatom forming a hydroxyl group, which can be ob-served for the formation of Ti 3 O 6 H 3 in the secondrow of Fig. 6 .

For bulk TiO 2 , the coordination number of Tiatoms in anatase, rutile, and brookite is 6 and thatof O atoms is 3. While for amorphous titaniumdioxide of small size, the averaged Ti –O coordina-tion number reduces to around 5 mainly becausemost Ti atoms are located in the outer shell andthe lattice is contracted in the core [37] . Here, wetracked the averaged Ti –O bond number for all theTi and O atoms that have contributed to the for-mation of the large cluster (Ti 26 O 31 H 6 ), as shownin Fig. 7 . From the beginning to around 100 ps, the

Kinetics for the hydrolysis of Ti(OC3H7)4: A molecular titute, https://doi.org/10.1016/j.proci.2020.06.345

J. Wei, A. Ostadhossein and S. Li et al. / Proceedings of the Combustion Institute xxx (xxxx) xxx 7

ARTICLE IN PRESS

JID: PROCI [mNS; September 16, 2020;13:8 ]

Fig. 7. Temporal evolution of averaged Ti –O bond num- bers of Ti and O atoms that contribute to the cluster Ti 26 O 31 H 6.

main Ti –O bond number drops from 4 to approx-imately 3 for Ti atoms due to the decompositionof TTIP, while the coordination number remainsalmost constant for O atoms that originate fromTTIP and water. The coordination number of Ti –Obonds continually increases and reaches a value of 3 for O atoms due to cluster formation, which isconsistent with the theoretical value. Since most Tiatoms are still unsaturated, the value of the averagenumber of Ti –O bonds increases to 3.8 for Ti withan increasing trend at 1000 ps. This result indicatesthat generally only one Ti –O bond breaks duringthe whole conversion process for TTIP molecules.Compared with the breaking of two Ti –O bonds toobtain TiO 2 and the formation of one Ti –O bondto build clusters, this is an energetically more favor-able pathway. High resolution and ultrafast mea-surements are necessary to experimentally validatethese findings.

4. Conclusions

In this work, the kinetics of the TTIP hydroly-sis reaction is investigated by performing ReaxFFmolecular dynamics simulations at temperaturesranging from 1500 K to 3000 K. The simulation re-sults are used to identify chemical reactions, im-portant intermediate species, and cluster forma-tion pathways during the conversion process fromthe gas-phase precursor TTIP to small clusters.Knowledge of the nascent particle formation andgrowth can bridge the gap between chemical reac-tions and further particle growth at the nanome-ter scale, providing guidance for nanomaterialproduction.

The following conclusions are drawn from thisstudy:

(1) Kinetic parameters for TTIP hydrolysisare estimated from simulations over a

Please cite this article as: J. Wei, A. Ostadhossein and S. Li et al.,dynamics simulation study, Proceedings of the Combustion Ins

wide range of temperature and pressure conditions. Using a second-order reaction

model, the determined rate constant at a pressure of 1atm is k = 1 . 23 × 10 14 ×exp ( −11 , 323 /T (K )) mo l −1 c m

3 s −1 . (2) The species analysis indicates that the main

reaction scheme of TTIP hydrolysis at high

temperatures proceeds through the follow- ing stages: dissociation of a C

–O bond to

form Ti(O

• )(OC 3 H 7 ) 3 ; breakage of a Ti –O

bond or a C

–O bond to form TiO(OC 3 H 7 ) 2 or Ti(O

• )(OH)(OC 3 H 7 ) 2 ; and C

–O cleav- age preferred for TiO(OC 3 H 7 ) 2 leading to

the formation of TiO(OH)(OC 3 H 7 ) and

TiO(O

• )(OC 3 H 7 ) at the third stage. Then the species directly combine to form clusters or continue to release the remaining –O x C 3 H y

groups to create small intermediates, such as TiO(OH) 2 , Ti • O(OH), Ti •• (OH) 2 .

(3) Early clusters are observed before the first TiO 2 molecule appears in the simulation. The cluster formation is initiated by forming a Ti –O

–Ti bridge to combine two molecules, which is followed by the interactions between

Ti and O atoms, resulting in the restructuring and release of water or –C x H y groups. Fur- ther work is necessary to experimentally val- idate the mechanism proposed from the sim- ulations.

Declaration of Competing Interest

The authors declare that they have no known

competing financial interests or personal relation- ships that could have appeared to influence the work reported in this paper.

Acknowledgments

Acknowledgment is made to the donors of the American Chemical Society Petroleum Research

Fund for partial support of this research. We ac- knowledge support through the DFG Mercator Fellowship SPP1980 “SpraySyn: Nanopartikelsyn- these in Sprayflammen”. This work is partially funded by the National Natural Science Founda- tion of China (No. 51725601 and NSFC-NSF Joint Program No. 51761125012 ) for co-author S.Q.L. J.W acknowledges the support from the China Scholarship Council (CSC). We would like to thank

Shujia Liang for helpful discussions on the reaction

pathways.

Supplementary materials

Supplementary material associated with this ar- ticle can be found, in the online version, at doi: 10. 1016/j.proci.2020.06.345 .

Kinetics for the hydrolysis of Ti(OC3H7)4: A molecular titute, https://doi.org/10.1016/j.proci.2020.06.345

8 J. Wei, A. Ostadhossein and S. Li et al. / Proceedings of the Combustion Institute xxx (xxxx) xxx

ARTICLE IN PRESS

JID: PROCI [mNS; September 16, 2020;13:8 ]

R

[

[

[

[

[

[

[

[

[

[

eferences

[1] J. Fang , Y. Wang , J. Kangasluoma , M. Attoui ,H. Junninen , M. Kulmala , T. Petäjä, P. Biswas ,Aerosol Sci. Technol. 51 (9) (2017) 1071–1081 .

[2] S. Li , Y. Ren , P. Biswas , S.D. Tse , Prog. Energy Com-bust. Sci. 55 (2016) 1–59 .

[3] S.E. Pratsinis , Prog. Energy Combust. Sci. 24 (3)(1998) 197–219 .

[4] M.S. Wooldridge , Prog. Energy Combust. Sci. 24 (1)(1998) 63–87 .

[5] H. Zhao , X. Liu , S.D. Tse , J. Aerosol Sci. 40 (11)(2009) 919–937 .

[6] S. Memarzadeh , E.D. Tolmachoff, D.J. Phares ,H. Wang , Proc. Combust. Inst. 33 (2) (2011)1917–1924 .

[7] J. Wei , Y. Ren , Y. Zhang , B. Shi , S. Li , J. Aerosol Sci.133 (2019) 72–82 .

[8] S. Tsantilis , H.K. Kammler , S.E. Pratsinis , Chem.Eng. Sci. 57 (12) (2002) 2139–2156 .

[9] T.H. Wang , A.M. Navarrete-López , S. Li ,D.A. Dixon , J.L. Gole , J. Phys. Chem. A 114(28) (2010) 7561–7570 .

10] R.H. West , R.A. Shirley , M. Kraft , C.F. Gold-smith , W.H. Green , Combust. Flame 156 (9) (2009)1764–1770 .

11] S.E. Pratsinis , H. Bai , P. Biswas , M. Frenklach ,S.V.R. Mastrangelo , J. Am. Ceram. Soc. 73 (7) (1990)2158–2162 .

12] F. Kirkbir , H. Komiyama , Adv. Ceram. Mater. 3 (5)(1988) 511–515 .

13] K. Okuyama , R. Ushio , Y. Kousaka , R.C. Flagan ,J.H. Seinfeld , AIChE J. 36 (3) (1990) 409–419 .

14] Y. Wang , P. Liu , J. Fang , W.N. Wang , P. Biswas , J.Nanopart. Res. 17 (3) (2015) 147 .

15] T. Seto , M. Shimada , K. Okuyama , Aerosol Sci.Technol. 23 (2) (1995) 183–200 .

16] A.G. Shmakov , O.P. Korobeinichev , D.A. Knyazkov ,et al. , Proc. Combust. Inst. 34 (1) (2013) 1143–1149 .

17] N. Baguer , E. Neyts , S.V. Gils , A. Bogaerts , Chem.Vap. Depos. 14 (11–12) (2008) 339–346 .

18] K.Y. Park , M. Ullmann , Y.J. Suh , S.K. Friedlander ,J. Nanopart. Res. 3 (4) (2001) 309–319 .

19] P. Buerger , D. Nurkowski , J. Akroyd , S. Mos-bach , M. Kraft , J. Phys. Chem. A 119 (30) (2015)8376–8387 .

Please cite this article as: J. Wei, A. Ostadhossein and S. Li et al.,dynamics simulation study, Proceedings of the Combustion Ins

[20] W. Yan , S. Li , Y. Zhang , Q. Yao , S.D. Tse , J. Phys.Chem. C 114 (24) (2010) 10755–10760 .

[21] Y. Zhang , S. Li , W. Yan , Q. Yao , S.D. Tse , J. Chem.Phys. 134 (8) (2011) 084501 .

[22] E. Goudeli , S.E. Pratsinis , Part. Sci. Technol. 34 (4)(2016) 483–493 .

[23] A.C.T. van Duin , A. Strachan , S. Stewman ,Q. Zhang , X. Xu , W.A. Goddard , J. Phys. Chem. A107 (19) (2003) 3803–3811 .

[24] H.M. Aktulga , J.C. Fogarty , S.A. Pandit ,A.Y. Grama , Parallel Comput. 38 (4–5) (2012)245–259 .

[25] D. Zhang , M. Ashton , A. Ostadhossein , A.C.T. vanDuin , R.G. Hennig , S.B. Sinnott , ACS Appl. Mater.Interfaces 9 (39) (2017) 34467–34479 .

[26] A. Ostadhossein , J. Guo , F. Simeski , M. Ihme , Com-mun. Chem. 2 (1) (2019) 1–11 .

[27] S.Y. Kim , N. Kumar , P. Persson , J. Sofo , A.C.T. vanDuin , J.D. Kubicki , Langmuir 29 (25) (2013)7838–7846 .

[28] P.J. Linstrom, W.G. Mallard (Eds.), NIST Chem-istry webBook, NIST Standard Reference DatabaseNumber 69, National Institute of Standards andTechnology, Gaithersburg MD, United States, 2005 .

[29] M.R. Weismiller , M.F. Russo , A.C.T. van Duin ,R.A. Yetter , Proc. Combust. Inst. 34 (2) (2013)3489–3497 .

[30] M. Raju , S.Y. Kim , A.C.T. van Duin ,K.A. Fichthorn , J. Phys. Chem. C 117 (20) (2013)10558–10572 .

[31] M. Reinke , Y. Kuzminykh , P. Hoffmann , J. Phys.Chem. C 119 (50) (2015) 27965–27971 .

[32] P. Buerger , D. Nurkowski , J. Akroyd , M. Kraft , Proc.Combust. Inst. 36 (1) (2017) 1019–1027 .

[33] D. Çakır , O. Gülseren , J. Phys. Condens. Matter 24(30) (2012) 305301 .

[34] A. Tarasov , V. Goertz , E. Goodilin , H. Nirschl , J.Phys. Chem. C 117 (24) (2013) 12800–12805 .

[35] M. Chen , T.P. Straatsma , D.A. Dixon , J. Phys. Chem.A 119 (46) (2015) 11406–11421 .

[36] Y. Xu , M. Wang , N. Hu , J. Bell , C. Yan , RSC Adv. 6(33) (2016) 28121–28129 .

[37] H. Zhang , J.F. Banfield , Chem. Rev. 114 (19) (2014)9613–9644 .

Kinetics for the hydrolysis of Ti(OC3H7)4: A molecular titute, https://doi.org/10.1016/j.proci.2020.06.345