Bjørn Kvamme ,Na Wei ,Jinzhou Zhao ,Shouwei Zhou ,Liehui Zhang ,Wantong Sun ,Navid Saeidi
a State Key Laboratory of Oil and Gas Reservoir Geology and Exploitation,Southwest Petroleum University,Xindu Road No.8,Chengdu,610500,China
b State Key Laboratory of Natural Gas Hydrate,Sun Palace South Street No.6,Beijing,10027,China
c Hyzenenergy,26701 Quail Creek,Laguna Hills,CA,92656,USA
d Strategic Carbon LLC,20 Ladd St.,Suite 200,Portsmouth,NH,03801,USA
ABSTRACT Mineral surfaces adsorb water to extreme densities and corresponding low chemical potentials.This results in a dual effect in terms of hydrate.Water and slightly polar components adsorb directly on mineral surfaces and generate efficient conditions for hydrate nucleation.But due to the extremely low chemical potential of adsorbed water the hydrate nuclei formed towards mineral surfaces have to either detach from the vicinity of mineral surfaces,or be bridged by structured water in a dynamic attachment of hydrate cores some few nm outside mineral surfaces.During transport of gas(CH4,gas mixtures,CO2)the conventional water dew-point analysis will typically result in a substantially higher acceptable water concentration as compared to the concentration for adsorption of water from gas to rust surface.Direct formation of hydrate from water dissolved in gas is thermodynamically feasible,as discussed in open literature.In this work we demonstrate that it is also feasible in terms of mass transport.A new theory for enthalpy of hydrate dissociation has been extended to also direct hydrate formation from water dissolved in gas.The remaining question is whether direct hydrate formation from gas is also feasible in terms of transporting the hydrate formation heat away through a heat insulating medium.We propose further research strategies to enlighten this issue.Addition of glycols to critical points in processing of gas or transport is already in use by companies like for instance EQUINOR.There is,however,a need for more work on how efficient it is and if it can also be used for multiphase transport of hydrocarbons with significant water cut.Some research activities are in progress and briefly outlined here.
Keywords:Hydrate Thermodynamics Risk Inhibitors
Transport of light hydrocarbons in pipelines involves high pressures and frequently low temperatures.Offshore Norway there is a substantial network of pipelines transporting natural gas between Norway and Europe.Troll[1]is the largest gas field in Norway and also the supply to Europe that guarantees the contract with Germany.The gas from Troll gas is almost pure methane gas.Originally the separation plant was designed without drying unit to remove water from the gas.For this reason a more improvised solution involves injection of glycol at some critical points.Critical points in this context mean conditions of temperature and pressure that can lead to hydrate formation if liquid water is available as free phase,or water can drop out from gas.Pressures are normally not brought down to very low pressures in these plants since the work involved in compressing the gas to transport pressure is proportional to the ratio between outlet and inlet pressure of the compressor.Typical pressures at end of gas separation are higher than 70 bar.The Troll plant is constructed as a dew-point plant,which implies that the quality of the gas is controlled by the hydrocarbon dew-point in the final separator,which is set to-11°C at 80 bar.Final dew-point separator is therefore in the risk of hydrate formation from the gas in the separator as well as from the liquid phase in the separator.Cooling of the gas in turbines and still fairly high pressure leave the outlet conditions from turbines as hydrate risk points.
Total export of gas from Norway implies supply from many other sources as well.Troll gas is processed at the Kollsnes plant roughly 60 km west of Bergen in Norway.Gas from Kvitebjørn is another source that is processed in a separate plant at Kollsnes.Ormen Lange is the third largest Norwegian gas field.It is located in the Storegga region below hydrate filled sediments.
In a world with increasing focus on Carbon Separation and Storage(CCS),or CCSU(Carbon Separation,Storage and Utilization),also transport and processing of CO2under hydrate forming conditions is increasing.This might be CO2collected from various industrial point sources.Norwegian government is investing 25 Billion NOK(3 Billion US$)in a pilot project[2]on CO2separation from low pressure flue gas from cements industry in the eastern part of Norway,shipping to Kollsnes and then pipeline transport to injection into an aquifer formation west of Norway.
A higher focus on CO2is also motivated by the fact that many hydrocarbon reserves worldwide have a high content of CO2.If we use South-East Asia,as an example,gas from one the largest gas fields in the world,Natuna[3],contains roughly 70 per cent CO2.And in general for the whole region many of the rich hydrocarbon resources have been depleted.Hydrocarbon resources with a higher CO2content than what is generally considered as economically feasible are yet being produced.
It is very important to stress that this work is not in competition to other published works related to hydrate risk evaluation for transport and processing.And there are comprehensive reviews of this type of work in books like for instance the one by Sloan&Koh[4].We are therefore not referring to any of those papers.Many of the papers reporting research on hydrate formation risk also assume that the systems are in temperature pressure equilibrium and that temperature and pressure are the only factors to affect in terms of shifting hydrate stability limits.Despite the fact that these systems cannot reach thermodynamic equilibrium since there are too many constraints(conservation laws and conditions for equilibrium)as compared to independent variables[5].Yet another problem is that many hydrate phase transitions are neglected,like for instance hydrate dissociation in chemical potential gradients[6]towards surrounding phases.Hydrate will dissociate towards a hydrocarbon phase that is not saturated with water and hydrate will also dissociate are surrounding water contains less hydrate former than minimum needed to keep the hydrate stable.These are just a few of several issues that make some strategies and related papers uncertain as basis for hydrate problem prevention.One example is the strategy of cold flow.Converting the water that follows a gas into hydrate and/or ice is not a feasible strategy.Hydrate and ice will sublimate towards dry gas regardless of temperature and pressure.
The main objective of this paper is to shed light on some aspects that are normally not considered in hydrate risk evaluations.The usual strategies for hydrate risk evaluation are based on evaluation of water dew-point.If the water dew-point is exceeded then free liquid water is assumed to be present and hydrate formation conditions must be evaluated.We extended these evaluations to take into account water drop-out at rusty pipelines[7-15]and direct hydrate formation from water dissolved in gas[8-15].We will return to some aspects of these alternative routes to hydrate formations.
A secondary objective is to investigate the impact of injected glycol on the water drop out conditions.Adsorption of glycol on rusty surfaces like pipelines is thermodynamically facilitated and is even reducing further corrosion.The question is whether glycol layers on rusty pipeline walls promote or prevent water drop-out as compared to adsorption on rust.
The primary scientific methods in this work are classical thermodynamics and statistical mechanics.The statistical mechanics is limited to application of earlier published models and derivatives of these.But these models are also derived in order to achieve models for other thermodynamic properties.The fundamental details behind the original derivation from statistical mechanics are also important.Mass transport is utilized in combination with classical nucleation theory from physics.Results from molecular dynamics is also utilized for properties of water in different phases as well as concentration profiles of hydrate forming molecules across interfaces between liquid water and hydrate.These results are based on earlier publications and not directly methods used in this work.Similar applies to models for diffusivity coefficients of hydrate forming molecules across the interface between liquid water and hydrate.
Injection of glycols at certain points in a gas processing plant which do not have a gas drying unit may be a way to prevent hydrate from creating flow blockage.As a first step we need to know how the glycol distributes inside a pipeline.
In this study we limit ourselves to some few relevant systems.As example for hydrocarbon we use pure CH4gas containing water.There are plenty of other examples in open literature[1,3,8-14].The other examples are CO2and mixtures of CO2and N2.Before using calculations in a thermodynamic analysis of water drop-out we need to verify the thermodynamic models.Note that our model utilizes residual thermodynamics for all components in all phases.If full thermodynamic equilibrium can be achieved the pressure temperature limits for hydrate formation are equilibrium curves.If thermodynamics equilibrium is not possible,as in real cases,then these curves are stability limits in specific sub-sets of thermodynamic independent variables.
Chemical potential of water in aqueous solution in excess thermodynamics is trivially given by Refs.[18-22]:
lim(γH2O)=1.0 whenxH2Oapproaches unity.
γH2Ois the activity coefficient of the liquid water as function of dissolved hydrate formers,as well as additives like methanol and salt.In the ice region then the latter term in(1)is zero.Molecular Dynamics(MD)simulations of ice[18]provided a temperature dependent chemical potential for water as ice at 1 bar pressure.Using experimental data for enthalpy of ice dissociation at 273.15 K,and specific heat capacity of liquid water then also resulted in chemical potential of liquid water above 273.15 K[18].Since liquid water is almost incompressible it is quite trivial to correct pure water chemical potential from 1 bar to higher pressures:
Chemical potential for water in hydrate is given by Ref.[18]:
number of cavities per water,with subscripts k for large and small cavities respectively.For structure I,which is the main focus here,νøarge=3/24 and νsmall=1/24.The second term on the right hand side of(3)is a simple Poynting correction from 1 bar to actual pressure,similar to the one in equation(2)for liquid water.Chemical potential for water in empty clathrate is also calculated from MD[18]and therefore also by definition residual thermodynamic values,with ideal gas as reference state for water.Equation(3)is identical with the original van der Waal &Platteeuw[23]derivation.The fundamental difference between the version derived by Kvamme &Tanaka[18]and van der Waal &Platteuw[23]is the option for flexible lattice and water movements[18].Specifically the canonical cavity partition function can be expressed as[18]:
β is the inverse of Boltzmann constant times temperature in molecular units and the inverse of the universal gas constant times temperature in molar units.The rigid water lattice version of this derivation is:
miis the molecular mass of guest molecule i,ħ is the Plank-Dirac constant.wki(Vk)is the interaction energy between guest molecule i in the cavity of type k and all surrounding molecules.Frequently all surrounding water molecules are limited to surrounding water molecules.Guest interactions with guest molecules in neighbouring cavities may also have significant impact on the integral[24].Slightly polar molecules in which the average partial charge(when rotating molecule around center of mass inside cavity)is positive outwards to cavity walls will have extra coulombic force stabilization effects[25].A typical example is H2S,which forms hydrates at low pressures.Opposite effect for molecules in which the average charge is pointing outwards,like for instance CO2.Equations(5)and(6)is the best choice for small molecules relative to available space in the cavity,like for instance N2.CH4is typically a turning point in size relative to available space since the results from(5)and(6)are almost the same as results derived from harmonic oscillator approach,equation(7)below.
Large molecules relative to available space for movements inside the cavity will significantly interact with the librational movements of waters,relative to minimum energy positions.Samplings from harmonic oscillator simulations provide detailed sampling of frequency ranges of guest movements which interfere with water librational frequencies.The result is specific values for the destabilization effects due to size.The harmonic oscillator formulation is:
A typical example for the need for(7)is CO2.The interactions between CO2and water movements results in a destabilization effect of 1 kJ/mol which is not accounted for when equations(5)and(6)are used instead of equation(7).The last term in equation(7)is the free energy of inclusion and is listed in Table 1 below in terms of a fitting equation and corresponding parameters.Note that the results for N2were derived from(5)and(6)and then fitted to a corresponding harmonic oscillator expression for use in equation(7).
Table 1 Coefficients dm,ki in the polynomial for free energies of inclusion;in kJ/moles for cavity type k and guests i = CO2,CH4 and N2. TCi is critical temperature for component i..

Table 1 Coefficients dm,ki in the polynomial for free energies of inclusion;in kJ/moles for cavity type k and guests i = CO2,CH4 and N2. TCi is critical temperature for component i..
M CO2CH4N2 LargeLargeSmallLargeSmallLargeSmall T ≤283.14T ≤283.14 1 41.52168-17.87093.19928817.97150-42.476838.85531-143.99019 2-41.96874-17.89249-28.28735-23.44013119.2412410.87598-95.44723 3-70.7269117.38136-11.94528-161.81535-183.1956525.065454.68052 4-11.81084-29.68940-2.6625045.20561128.3925254.1707859.13995 5 16.73045-19.903213.8565336.67261-54.9878420.64268109.87325 6 21.9162125.221123.21040138.00217-78.55671-133.44618113.79746 7 0 0 0 0 0-130.99357149.13992 8 0 0 0 0 0-85.72083139.08208 9 0 0 0 0 0-106.74770153.82431 1000000-78.83698141.94965
The solution to hydrate equilibrium curves are then the following applied to equilibrium between liquid water and hydrate water:
Assuming equilibrium between guest molecules in small and large cavities for molecules that can enter both cavity types:
For this work the guest molecules i are CH4,CO2and N2.For equilibrium between gas and hydrate we then have:
We utilize Soave Redlich Kwong(SRK)[26]equation of state to calculate the fugacity coefficientsfor chosen temperatures between 273.16 K and 285 K and the iterative solution for pressure,at each temperature,that satisfies equation(8)when values from(4),(7)and(10)has been inserted into(8)for the cavity partition functions.
In this work we are only interested in injection gas with maximum 30 mole per cent N2in CO2/N2mixture,and rather more focusing on 20 mole per cent N2in CO2/N2mixture as most interesting.For this purpose there are some experimental data available as plotted in Figs.4 and 5 together with calculated values.There are some uncertainties about the experimental data from Kanget al.[27]since they could almost seem to be extrapolated/smoothened past the CO2phase transition point.The only reason for that indication is the final experimental point for 77.80 mol per cent CO2which fall into a smooth curve across the CO2phase transition point.Based on the three last experimental points the final point should be expected to be above the calculated curve.This CO2phase transition point will also be reflected in mixtures with CO2and is fairly clear in Fig.6.
In Fig.6 we verify the rest of the model systems in terms of hydrates from pure CO2and pure CH4respectively.
Note that figures(4)-(6)verify the thermodynamic model for Gibbs free energy calculation since Gibbs free energy for the hydrate is:
The hydrate composition is trivially available from the hydrate filling fractions and given by:
And at equilibrium the following condition is fulfilled:
θkiis the filling fraction of componentiin cavity typek.Also:
where ν is the fraction of cavity per water for the actual cavity type,as indicated by subscripts.The corresponding mole-fraction water is then given by:
Two very important misunderstandings are related to the use of Fig.6 in many publications.Temperature and pressure are independent thermodynamic variables and do not tell anything about stability of the hydrates.Fig.6 is only a two-dimensional projection of conditions for hydrate possible hydrate formation.The dependent thermodynamic function for phase stability is Gibbs free energy.In contrast to this many publications state that methane hydrate is more stable after the temperature for CO2phase transition.Fig.7 tells that CO2and all the experimental mixtures investigated are thermodynamically more stable than the pure CH4hydrate for the whole range of conditions in figures(4)-(6).The second misunderstanding that frequently appears in publications and software for hydrate equilibrium curve is the missing phase transition effect for CO2.Many curves like Fig.6 do not have these steep changes but rather a smooth curve for the CO2hydrate stability limit curve.
Enthalpies of hydrate formation for CH4and some mixtures of CO2and N2are plotted in Fig.8.
Fig.7 also confirms that injection of CO2/N2mixtures with up to 22.20 mol per cent is able to create a new hydrate with free pore water which is more stable than the in situ CH4hydrate.Hydrate production is not the focus here but rather the transport aspects.Nevertheless we can still examine if also the available enthalpy of hydrate formation for the CO2and CO2/N2mixtures are also able to dissociate the in situ hydrate.The equations for enthalpy of hydrate formation will be presented and discussed later,as connected to discussion on direct hydrate formation from dissolved water in gas during pipeline transport.
For pure CO2and all the investigated experimental compositions of CO2/N2the enthalpy of hydrate formation is sufficient to dissociate in situ CH4hydrate.
The model equations for two ways water can drop out tp a separate phase are given in Fig.9.The dew-point calculation,as maximum water mole-fraction in gas before water drops out as liquid,is the more common way to evaluate hydrate today in industry.This mole-fraction is found by solving for the chemical potential of water in gas being equal to pure liquid water for a defined set of temperature in a pipeline or equipment.If the condition of temperature and pressure for the actual gas is inside pressure temperature hydrate stability limits,like for instance those in figures(4)-(6)then hydrate is formed.A mass balance between initial water in gas and the remaining water after drop-out can then tell something about amount of hydrate formed.
The use of a simple equation of state for water in gas in justified because of the low water mole-fractions in the gas that reduces the water-water contributions in the attractive parameter to practically zero.If we for example look at just water dissolved in methane then the Lorentz-Berthelot mixing rule for the attractive parameter a reads:
With 1 for methane and 2 for water and solubility of water in methane in the order of 10-3in mole-fraction then the latter term practically vanishes and the remaining terms end up as non-polar terms(methane-methane and methane-water).Even the mixed term is small.Strictly speaking the SRK equation“does not see”water polarity since equation of state parameters depends on critical properties and pure component saturation pressures.The alternative use of virial equation makes the water polarity directly visible but the use of virial equation with up to third coefficient is still too limited compared to the range of pressures needed in the calculations.
In Fig.10 we plot the solutions to the equations in Fig.9 for pure CH4containing dissolved water.Similar calculations for water in pure CO2are plotted in Fig.11.And since the experimental condition of 77.80 mol per cent CO2in CO2/N2mixture is representative for a relevant mixture we use that as example in Fig.12.
The absolute values for the three systems plotted in Fig.10-12 are different.An analysis based on dew-point water mole-fraction Fig.10.a)Water dew-point mole-fraction water from solution in CH4.B)Water mole-fraction in CH4before water adsorb on Hematite.Upper curves are for 50 bars,then 100 bars,150 bars,200 bars,250 bars and lowest curve is for 300 bars.as acceptable limit for RISK evaluation involves acceptance of 18-19 times higher mole-fraction than if adsorption of water on rust is used as criteria.Ordinary oxidation rust is a mixture of iron oxide(FeO),hematite(Fe2O3)and magnetite(Fe3O4).For pipelines that corrode in open air at atmospheric conditions magnetite typically dominates in early stages due to access of much oxygen from air.Hematite is the thermodynamically most stable and rapidly dominates.All three rust forms structure water to extreme densities(and correspondingly low chemical potentials for adsorbed water).Even if the absolute values for water chemical potentials adsorbed on iron oxide or magnetite might be slightly different the dominance of hematite makes it a fair model for rust.
Direct formation of hydrate from dissolved water is thermodynamically given by the conditions that satisfy same chemical potential for water in the gas and in the hydrate that is formed.Thermodynamically then the following equation is solved:
For defined temperature and pressure equation(18)is iteratively solved for mole-fraction water in gas while keeping the composition of the carrying gas constant.Thermodynamically direct hydrate formation is feasible and the mole-fraction water in gas before direct hydrate formation is slightly more than 90 percent of the water dew-point mole-fractions.Specific mole-fractions for pressures from 50 to 300 bars are plotted in Fig.13.
3.2.1.Mass transport limitations
The asymptotic solution to a Fick's type of molecular diffusion transport process is available in any College level book on fluid mechanics,like for instance the web book from Rice University[33].For the purpose of this work this is sufficient for providing rough estimates of mass transport needed to supply hydrate growth with enough water from a dilute solution in CH4.
As given by the definitions of all molecular models used in this work,the interatomic or pair interaction energies U are approximated by a sum of Lennard-Jones and Coulomb interaction potentials,given by
whererijis the distance between a pair of interacting particlesi,j.εijand σijare the well-depth and radii of the Lennard-Jones potential for the interacting pair,andqi,qjare the partial charges.ε0is permittivity in vacuum.Sets of parameters for the used models were taken from the referred papers.Whenever the Lennard-Jones parameters ε and σ were only provided for pairs of identical atom types,i.e.εiiand σii,the parameters for unequal pairs were calculated using Lorentz-Berthelot mixing rules,given byandAs mentioned above OPLS CH4is simply a spherical monoatomic model in which all hydrogens are merged together with carbon to a neutral atom model with σCH4CH4=3.73 Å and εCH4CH4= 1,2309 kJ/mol.As a consequence all interactions between OPLS CH4and water will be non-polar and the polar nature of water is invisible.Practically this means that the higher mass of water and relatively weak attraction compared to CH4will imply that water diffuses through CH4faster than self diffusion of CH4due to the smaller“friction”.TIP4P[34]model parameters and intramolecular coordinates are given in Table 2.Note that the oxygen partial charge is located slightly outside of oxygen mass centre in this model and the denoted as PC(Point Charge).Parameters for the 5-site CH4model[36]is given in Table 3 below.Due to the Coulumbic interactions between partial charges in TIP4P[34]and the 5-site CH4[36]

Table 2 Intramolecular atom coordinates and interaction model parameters for equation(20)for the TIP4P model[34].

Table 3 Intramolecular atom coordinates and interaction model parameters for equation(20)for the 5-site CH4 model[36].
Diffusivity coefficients are calculated from auto-velocity correlation function integrals.Within the need for only approximate calculations we do not adjust for temperature dependency and reduce number of simulations to the six pressures in Fig.13 but only for 278.16 K.Based on the relationship between kinetic energy and velocity the diffusivity coefficients scales as:
which,on this rough level of estimations,may also be used as an approximation.In Table 4 we list calculated diffusivities for the 6 different pressures and with densities calculated from SRK[26]to make them consistent with other equations of state calculations in this work.For all the simulations an initial stabilization period of 500000 time steps with a time step of 0.5×10-16s.Minimum 3.5 Million time steps,and more if changes in sampled average Diffusivity exceeds more than 2 per cent as compared to samplings over the previous 0.5 Million time steps.Thermostat masses for the TIP4P water[34]were chosen[38-40]as 44 fs for translation and 21 fs for the rotation.Rotational thermostat is not actually relevant for infinite dilution water in CH4since TIP4P hydrogen only has partial charge and no short range contribution.Barostat mass was set to 5000 fs.

Table 4 Estimated Diffusivity coefficients for water in methane calculated from Molecular Dynamics simulation using TIP4P[34]for water and two model for CH4;single site spherical OPLS for CH4[35]and the 5-site CH4 model due to Sun et al.[36].Gas density calculated from SRK[26].Temperature is 278 K yHH2Ois water mole-fraction for direct hydrate formation from water dissolved in gas.ρNH2Ois the molar density of water at direct hydrate formation mole-fraction from water dissolved in gas.
In Fig.14 we plot experimental data from open literature,as well as some results from very recentab initiosimulations.Accuracy in experimental measurements of methane self-diffusivity is generally low.Dawson[41]indicates roughly 10% accuracy in density times Diffusivity coefficient as function of density,and less accuracy for Diffusivity coefficient as function of density.Similar accuracy applies to the data from Oosting&Trappeniers[42].Higgoda et al.[43]claims that their data fromab initiosimulations can be used as reference values in view of the experimental complexities and corresponding uncertainties.
If we consider the highest mole-fraction water for direct hydrate formation for 50 bar pressure in Fig.13 it varies around 0.003.Similarly the lowest mole-fraction water at 300 bar is roughly 0.0005.Using CH4,Fig.13 a),as an example and utilizing the average densities and molecular weights for the two conditions a water molecule will on the average find another water molecule in a radial distance of roughly 68.6 Å for 300 bar and 81.8 Å for 50 bar.These are as expected of course on nano-scale and have to be compared to the phase transition mass transport related to the hydrate formation.At this level we do not consider whether the hydrodynamic level of mass transport is able to maintain this concentration during a hydrate growth.
The diffusivity coefficients for methane are reasonable as compared to experimental data and Quantum mechanics simulated data plotted in Fig.14,as seen from Fig.15.To our knowledge there are no available experimental data for water diffusivity coefficients dissolved in water.It should be stressed that these simulations are fairly rough since a larger simulation ensemble would be desirable for more accurate results.But relative to other uncertainties and the need for fair estimates rather than rigorous results we consider these values as good enough for the purpose of calculations in this section.
In Classical Nucleation Theory(CNT)the total free energy involved in a phase transition consists of the free energy change related to the phase transition itself and a penalty term which is the work needed to push away the old phases.For the heterogeneousformation of hydrate on the liquid water/gas interface,this involves a push work towards liquid water on one side and gas phase on the other side.During the nucleation period the clusters are fairly small,typically less than 3 nm[9,22],and it is fair to assume that the hydrate cluster is facing water on all sides due to capillary forces.
For the simplest possible geometry of a crystal,which is a sphere,with radiusRwe then get:
Asis contact surface area between the phases.For non-polar gas/liquid interfaces the difference between σ and γ is limited for low to moderate pressures.For solid/liquid the difference is generally significant.The interface free energy between hydrate and liquid water is not known and as approximation we use the experimental value for ice/liquid water,equal to 29.1 × 10-3J/m2[44].
The nucleation stage is a period of unstable growth and decay due to the competition between the free energy gain and the penalty of pushing away the old phases.Mathematically this appears as a maximum in total free energy in Equation(30).The cluster size for maximum is called critical nuclei size.See for instance Kvamme et al.[16].Within the simple CNT this is the transition over to stable growth.Limitations in supply of mass,and/or limitations related to transport of released heat,require a different approach,such as,for instance,PFT[45-51].For a simple sphere the critical radius is then given by:
in which the *indicates critical radius.The critical radius is also a turning point for Gibbs free energy and opens up for the possibility of an additional shape factor to describe a non-spherical shape.See for instance Kvamme and Kuznetsova[52].
In Fig.16 we plot critical radius for hydrate forming directly from water dissolved in gas for water supersaturations between 10 and 40%above water hydrate formation mole-fraction limits in Table 4.Corresponding number of molecules in critical size as function of pressure are plotted in Fig.17.
The kinetics of a phase transition is implicit functions of thermodynamic control and related mass and heat transport.Heat transport through condensed(ice,liquid water,hydrate)water phases is normally two-to-three orders of magnitudes faster than mass transport[53].There are certainly exceptions to that.Hydrate formation from water dissolved in gas,as discussed here,is thermodynamically feasible.This is illustrated above for methane,and illustrated elsewhere[54]for CO2.The kinetics related to Classical Nucleation Theory(CNT)can be expressed by:
whereJ0is the mass transport flux supplying building blocks for the hydrate growth.For the heterogeneous hydrate phase formation it will be the supply of methane to the interface growth.
There are two stages of mass transport.The first is the mass transport to the growing hydrate core and the second is the transport of hydrate former across the interface between hydrate and the surrounding phase.For hydrate formation at the interface between liquid water and a separate hydrate former phase the hydrate core is covered by water and the interface transport is typically the mass transport limiting term.See for instance Kvamme et al.[5],Kvamme and Clarke[19],Kvamme[21]and Kvamme et al.[22].In the case of surrounding gas containing low concentration of water there will not be such a thick(relative)interface of structured water and mass transport of water through gas will be the mass transport flux defining the mass transport term in equation(25).
Using equation(19)we can calculate the time for supply of mass to hydrate core growth between two core sizes and corresponding times.
In equation it is approximated that water density is constant.Strictly speaking there will be a higher water density close to the hydrate core.And also the water Diffusivity coefficient will be affected in the vicinity of the hydrate core.Nevertheless - for our purpose equation(26)is considered as good enough.
In Fig.18 we plot mass transport rates in terms of molecules of water transported as function of time together with the thermodynamically determined critical numbers of water for generating a stable hydrate core that is able to continue growth.This first plot is for 50 bar pressure and 278 K.Even for only 10 mole per cent water super saturation nucleation is feasible from a mass transport point of view.We already know that t is feasible in terms of Gibbs free energy changes as discussed above.Whether it is feasible from a heat transport point of view will be discussed more in the next section.
Similar plots like Fig.18 for 100 bar,150 bar,200 bar,250 bar and 300 bar are plotted in Figs.19-23.All calculations are for 278 K.
As expected the times for hydrate nucleation based on mass transport increases with pressure since the molecular diffusion rate decreases with gas pressure.All of these calculations were based on the non-polar one-site model for methane.The trend would be the same with the 5-site methane model but the nucleation would be slightly slower.The expected difference is,however,not large enough to deserve space in this section of the manuscript.
Comparison with homogeneous hydrate formation form dissolved methane in liquid water is also relevant.This type of hydrate formation is feasible in between liquid water solubility of methane and minimum concentration of methane in water for keeping the hydrate stable.This concentration limit is the reason that in situ methane hydrate dissociates when seawater enters hydrate filled sediments through fracture systems and gives rise to methane leakage fluxes.In Fig.25 we plot solubility of methane in water as function of temperature and pressure and in Fig.26 we plot the corresponding lowest mole-fraction of methane in water needed to keep hydrate stable.The pressure dependencies of these molefractions are very limited since this is a phase transition between condensed phases.The Poynting corrections for water in liquid on one side and the Poynting correction for water on the hydrate side almost cancel.The same can also be seen in Fig.27 which illustrate the window of water methane mole-fractions for which methane hydrate can form from solution.
Homogeneous hydrate nucleation is plotted in Fig.28 below.
3.2.2.Heat transport limitations
A consistent model for enthalpy of hydrate formation from the free energy model is given by the standard thermodynamic relationship[21,55-58]and formulated as follows for hydrate formation from a separate hydrate former phase and water as ice or liquid water:
For comparison of nucleation times we can compare with nucleation times for heterogeneous hydrate nucleation on the interface between methane and liquid water.These are plotted in Fig.24 below for the same range of pressures and 278 K.
Differentiation of equation(3)gives the following results for partial molar enthalpy of water in hydrate:
The temperature derivatives of the free energy of inclusions in equation(28)is plotted in Fig.29 for the components in this study.
From equation(8)using the same basic thermodynamic relationship as equation(27):
and the enthalpy of hydrate formation is then:
The last term is the partial molar enthalpy of the guest molecules in the gas(or liquid)hydrate former phase.This is trivially given by the equation of state using the same type of relationship as equation(27).The analytical expression for the SRK version of this term is available form engineering handbooks and does not need space here.The partial molar enthalpy of each guest molecule in the hydrate,needs some more attention.
The ideal gas enthalpy depends only on temperature and will be the same for the guest molecule inside the cavities and in the gas mixture.As such we only need the residual contributions for these two terms.While entropy related properties like chemical potentials and free energies are sensitive to possible interactions between water movements and guest movements the average interaction energy is less sensitive to these effects.We have therefore used a Monte Carlo approach[59,60]to evaluate average interaction energies between guest and water for the various cavity types in structure I.We also sample the efficient volume for the guest in the cavity in order to calculate a compressibility factor for the guest molecule inside the two types of cavities.
whereVklis the sampled molar volume of guest molecule i in cavity type k.
Presently the theory is also developed for hydrate formation from dissolved hydrate formers in water[55].It is not difficult to extend the theory to hydrate formation from water dissolved in gas.The derivation is fairly trivial and simply involves replacing equation(29)with a similar expression for water partial molar enthalpy in gas.The free energy expression is already in used and is the basis for the water solubility calculations in the previous section,as well as the super saturation calculations in the same section.But the programming still needs to be done and results verified.As such this will be the topic of a separate follow-up paper.
But a model for enthalpy of hydrate formation is not all that is needed.Even for extremely small hydrate nuclei,in principle from the first clusters that struggle between further growth or decay due to the competition between benefits(Gibbs free energy gain by the phase transition)and the penalty of pushing the surroundings to make space for the new phase).See equations(21)and(22).This can be accomplished by either Molecular Dynamics(MD)simulations in the classical regime,or by Quantum mechanical modelling.MD simulations are more efficient in terms of possibilities for many molecules and long simulations,which is needed.In addition there are modern tools for tracking local entropy changes and consequences for chemical potentials of individual types of molecules.The adaptive biasing force(ABF)method[59,60]is actually based on entropy perturbations and particularly suitable.
On a macroscopic level we can compare thermal conductivity of methane.It varies with density of course.At 1 bar and 298 K it is 0.034 Watt/m·K while at 100 bar the values at 273.15 K and 298 K are almost the same;0.044 Watt/m·K.The values for liquid water are more than an order of magnitude higher.At 273.15 K and 100 bar pressure thermal conductivity for water is around 0.560 Watt/m·K.It is possible to make some simplified modelling that includes conductivity and convection of released hydrate formation heat with similar level of sophistication as the mass transport analysis in the previous section.But it makes more sense to do this as part of an analysis that also involves the challenges of transporting away heat from sub-critical size hydrate cores that try to grow.The nucleation challenge might be in that step of the hydrate formation process.
In summary there is a need for a separate study that can address the heat transport challenge on all stages of hydrate formation,from nucleation to visible hydrate cores.And it is important to keep in mind that direct hydrate formation from water dissolved in a non-polar gas always will compete with other possible routes towards hydrate formation.As discussed earlier water can condense to liquid water and then create hydrate but even more important water can drop out by adsorption on rust surfaces in pipelines and equipment.And this also opens up a discussion on presence of alcohols.This is briefly mentioned in the next section but an ongoing effort is expected to result in a separate paper on that within a few months.
Finally we just plot(Fig.30 below)an example of heterogeneous hydrate formed from water and methane with ethane as basis for some indications of expected range of enthalpies of hydrate formation from water dissolved in gas.
Note that experimental measurements of hydrate dissociation enthalpies are complex since it is hard to measure all the necessary properties within the same experiment.Obviously hydrate needs to be formed outside of equilibrium(lower T and/or higher P)and also has to be dissociated at higher T than equilibrium.Hydrate coordination number is frequently missing or has been estimated.These are just some of the challenges in hydrate formation(or dissociation)enthalpy experiments.See Kvamme et al.[56]for a more detailed discussion on experimental challenges and Aromada et al.[57]for variations in experimental data.
As discussed elsewhere[55]the enthalpies of hydrate formation from dissolved methane in water are lower in absolute value because the change in partial molar enthalpy for guest molecules is smaller in that case as compared to bringing gas molecules into a smaller space inside a cavity.For direct hydrate formation from gas there will not be any such difference.Unlike chemical potentials of water at solubility,which will be equal to liquid water chemical potential the difference in chemical between liquid water partial molar enthalpy and partial molar volume of water in gas will be very different and sensitive to pressure.Partial molar enthalpy of water in gas is given by:
In which y with an arrow indicates mole-fractions in the gas.In Fig.31 below we plot water enthalpies in the gas for 278 K and the same pressure as discussed above.
If water actually goes directly from a state of dissolved in gas the numbers in Fig.31 will result in fairly extreme enthalpies of hydrate formation.The partial molar enthalpies for dissolved water in gas are extremely much higher than liquid water enthalpies,as seen from Fig.30.Direct transfer of water molecules from solvation in gas to hydrate structure therefore results in the large enthalpies illustrated in Fig.32.Corresponding coordination numbers,which are number of water molecules in hydrate per gas molecule are plotted in Fig.33.
It is quite unrealistic that water goes directly from a state of low concentration in a gas over to a solid hydrate structure.It is more realistic that the water molecules adsorb on a hydrate like nuclei structure and restructure with methane over to cavity structure.A detailed mechanism for the process is not known but at least the overall enthalpy change illustrated in Fig.32 will obviously then split up in enthalpy changes related to the various steps of a transformation mechanism.Nevertheless - the total sum is obviously the same given the very low enthalpy of hydrate water compared to the high numbers of partial molar enthalpy of dissolved water in methane in Fig.30.
Methanol is frequently used to prevent hydrate formation also during the transport of hydrocarbons from a well to local processing facilities or in multiphase flow to more remote processing facilities offshore or onshore.Some countries with access to low cost ethanol like for instance Brasil also use ethanol for these purposes.These small alcohols have a dual effect relative to hydrate.In fairly high concentrations they are,as is well known,efficient thermodynamic hydrate inhibitors.When a water phase containing these types of alcohols is exposed to a non-polar phase the concentration of these alcohols will be higher at the water/nonpolar interface than in“bulk”water solution.If the alcohol concentration is small then these surfactant effects will dominate and the alcohol presence in the water side of the interface will serve as hydrate formation catalyst.See Kvamme et al.[11]for a more detailed discussion on the various mechanisms related to that effect.The paper by Selvåg et al.[62]also provides useful insight.
The hydrate formation acceleration effect generated by small amounts of alcohols in water is not new knowledge.There are several laboratory studies that confirm this although this is not a main focus in this paper and we only mention a large scale pilot plant study on the use of methanol in a real pipeline flow offshore Norway[63].Methanol and Ethanol also distribute in gas along with water and need to be accounted for in all analysis of water drop out concentrations or direct hydrate formation from gas.
Glycols are also frequently used because they also have corrosion inhibitor effect,which is logical because they adsorb very well on rust[16].In the context of this work it is important how glycol films on rusty pipeline walls affect water drop out from gas and how these hydrate films prevent hydrate problems since they reduce water possibilities for structuring towards rust and makes itthermodynamically feasible to“glue”hydrate towards rust and eventually block the pipeline.More work is needed along this pathway.Some of this work will be a continuation of earlier work[16]but some will include classical thermodynamic analysis.For this purpose some fairly rough thermodynamic models are provided in Table 5.For methanol and ethanol the model are verified towards experimental data and proven very accurate[65],while there are few data to compare model calculated results to experiments for glycols.Nevertheless - the model and parameters are available for anyone to use at their own discretion.They should be used with caution and compared to experimental data for quality control.

Table 5 Model and parameters for water activity coefficients for equation(1).
All mineral surfaces are water wetting and result in water structuring to extreme densities.Densities of water in the first adsorbed layer can typically be in the order of three times liquid water density.Slightly polar components like CO2and H2S can adsorb directly on mineral surface and supply mass for hydrate growth slightly outside the mineral surfaces.Hydrate formed towards mineral surfaces cannot attach directly on mineral surfaces since adsorbed water has significantly lower chemical potential than hydrate water.Another argument is that partial charges of hydrogen and water,on the hydrate surface,are not compatible with average partial charge distribution in liquid water.An interface in the order of 1.2 nm thick between hydrate and“bulk”water will therefore establish itself by molecular physics.Non-polar molecules like for instance methane will not be able to compete with water on direct adsorption on mineral surface but can be trapped in structured water slightly outside of the mineral surfaces[5].
In sediments typical minerals can be Quartz,Calcite and a large variety of mixed minerals.Most offshore hydrates are in a dynamic stationary balance in which new hydrate is formed from gas that enters hydrate conditions from below trough fracture system,and hydrate dissociation towards incoming seawater from above.This seawater contains little or almost no hydrocarbons and hydrate dissociates because almost infinite dilution chemical potential of these components are lower than chemical potential of the same components in hydrate.
In pipelines and processing equipment the most typical mineral surfaces are rust.Oxitated rust consists of Hematite(Fe2O3),Magnetite(Fe3O4)and Iron oxide(FeO).Magnetite typically dominates the first rust formation due to easy access of oxygen from air,while Hematite is thermodynamically more stable and dominate at later stages.Dew-point concentrations of water are typically 18 to 20 times higher than concentration of water that leads to adsorption of water from gas to Hematite.A higher focus in hydrate prevention should therefore be on effects of mineral surfaces.In particular there is need for more fundamental work on how glycols adsorb on rust surfaces and the subsequent hydrate inhibition effects.
Direct formation of hydrate is thermodynamically feasible and in this work we have demonstrated that it is also feasible in terms of mass transport dynamics.A challenge that needs more work is the impact of heat transport dynamics.On a molecular level the entropy changes during small hydrate cores(smaller than critical size for stable growth)that are in competition between growth and decay are crucial.Are these small hydrate cores able to get rid of heat generated by entropy changes from gas solution over to lower entropy in hydrate? The answer to this needs fundamental modelling studies.Methane gas is a heat insulator.On a higher level we have extended our enthalpy model to hydrate formation from water dissolved in gas.As expected the enthalpies of this direct hydrate formation are very large in absolute numbers compare to enthalpies of heterogeneous hydrate formation on water/gas interface and enthalpies of hydrate formation from dissolved hydrate formers in liquid water.For heterogeneous hydrate formation the enthalpy difference for liquid water going into hydrate is in the order of -6 kJ/mol.The difference in water enthalpy from water dissolved in methane over to hydrate is in the order of-30 kJ/mol.More research on heat transport from growing hydrate,through heat insulating gas,is needed in order to decide whether it is feasible all for hydrate to form directly from water dissolved in hydrate forming gas.
Mineral surfaces act as hydrate accelerators because they adsorb water as well as slightly polar hydrate formers(CO2and H2S are examples)and therefore result in convenient symbiosis of structured water and close availability of hydrate formers.Non-polar hydrate formers,like for instance methane,can be trapped in structured water and create hydrate from what we might call secondary adsorption state.We have demonstrated that these routes to hydrate formation are feasible.And some relevant thermodynamic data are also available from open literature to facilitate hydrate nucleation studies.It is,however,important to keep in mind that hydrate nucleated towards mineral surfaces will not be able to attach to mineral surfaces due to very low chemical potential of adsorbed water.In industrial setting adsorption of water on rust surfaces dominated thermodynamically compared to water dropping out as liquid(dew-point).
We have earlier demonstrated that hydrate formation from water dissolved in hydrate forming gas like for instance methane or carbon dioxide is feasible.In this work we have demonstrated that it is also feasible from a mass transport perspective.The critical question is whether it is also feasible from a heat transport perspective.On a molecular level the entropy changes from water in gas to water in a hydrate core lattice result in heat that needs to be transported away through a heat insulating medium.On a higher level we have extended our enthalpy model to also hydrate formation from gas.In absolute values these enthalpies are very high.There is a need for extending a model for enthalpy of hydrate formation to also include hydrate formation from water dissolves in gas.And there is a need for heat transport modelling that can shed light on whether the heat can be transported away fast enough,or the released heat will result in local temperature increase and redissociation of formed hydrate cores.Work is in progress in these directions.
Glycols have very favourable adsorption characteristics on rust and there is a need for more insight into how this feature can prevent possible hydrate core attachments on glycol covered rust surfaces.And there is also a need to understand the dynamics of water adsorption from gas to glycol covered rust surfaces.Work is in progress on these issues.
It is hereby declared that the submission involves no conflict of interests with other people or institutions.
The authors are grateful for financial support through 111 Project(No:D21025),National Key Research and Development Program(No:2019YFC0312300),National Natural Science Foundation Item of China(No:U20B6005-05,51874252 and 5177041544),Open Fund Project of State Key Laboratory of Oil and Gas Reservoir Geology and Exploitation(No:PLN2021-02 and PLN2021-03).