Master Thesis Jeroen Brenkman Repository

High Pressure
Hydrogen Dehydration
Using Ionic Liquids
J.C.B. Brenkman
Technische Universiteit Delft
H IGH P RESSURE
H YDROGEN D EHYDRATION
U SING I ONIC L IQUIDS
by
J.C.B. Brenkman
in partial fulfillment of the requirements for the degree of
Master of Science
in Mechanical Engineering
Track: Energy & Process Technology
at the Delft University of Technology,

to be defended publicly on Wednesday June 26, 2019 at 9:00 AM.
Student number: 4169093

Supervisor: Rogier Schoon Royal Dutch Shell
Thesis committee: Prof. dr. ir. T.J.H. Vlugt Delft University of Technology
Prof. dr. ir. R.A.W.M Henkes Delft University of Technology
Dr. ir. M. Ramdin Delft University of Technology
Dr. ir. D.A. Vermaas Delft University of Technology
Ahmadreza Rahbari Delft University of Technology
An electronic version of this thesis is available at http://repository.tudelft.nl/.

ii
A CRONYMS
Acronym Definition
CAPEX Capital Expenditure
CMS Carbon Molecular Sieves
COSMO-RS Conductor like Screening Model for Real Solvents
EoS Equation of State
EHC Electrochemical Hydrogen Compressor
FCEV Fuel Cell Electric Vehicles
HEX Heat Exchanger
HP High Pressure
IL Ionic Liquid
ILM Immobilized Liquid Membranes
ISO International Organization for Standardization
KF Karl Fischer
NG Natural Gas
OPEX Operational Expenditure
PEM Proton Exchange Membrane
PR Peng-Robinson
PSA Pressure Swing Adsorption
PV Photovoltaic
SDS Safety Data Sheet
SMIRK Shell’s Modified Intensified Redlich Kwong
SRK Soave Redlich Kwong
TSA Temperature Swing Adsorption
UN United Nations
UNFCCC United Nations Framework Convention on Climate Change
VLE Vapor Liquid Equilibrium
C HEMICALS
Acronym Definition
DEG Diethylene glycol
TEG Triethylene glycol
[EMIM][Acetate] 1-Ethyl-3-methylimidazolium acetate
[BMIM][Chloride] 1-Butyl-3-methylimidazolium chloride
[EMIM][Ethyl sulfate] 1-Ethyl-3-methylimidazolium ethyl sulfate
[BMIM][Octyl sulfate] 1-Butyl-3-methylimidazolium octyl sulfate
A BSTRACT
Renewable energy projects, such as wind and solar (PV) parks, are rapidly being built to replace fossil fuel
based energy sources. The major drawback of electricity production from renewable sources is that they are
intermittent and therefore lead to a mismatch between the supply and the demand of energy. Energy storage
in the form of hydrogen production is a potential candidate to overcome this problem. However, due to its
low volumetric energy density, hydrogen must be compressed to use it for storage or transportation purposes.
HYET BV has developed an Electrochemical Hydrogen Compressor (EHC) that is capable of compressing hy-
drogen up to P = 1000 bar, and have the potential of bringing compression costs down to 3 kWh/kg. As the
compressed hydrogen is saturated with water, it must be dehydrated before it is used for refuelling Fuel Cell
Electric Vehicles (FCEV), as the ISO 14687-2:2012 standard has limited the water concentration to 5 µmol
water per mol gas mixture.
In this work, a landscaping study was performed to search for potential drying methods, including membrane
separation, adsorption, absorption and cooling. To select the best method for hydrogen dehydration, it is es-
sential to have a good understanding of the thermodynamics at the given pressure conditions, beginning with
the saturated water content in compressed hydrogen. To the best of our knowledge, the only experimental
data describing the water content in the H2 -H2 O mixture for pressures exceeding 300 bar are from 1927 and
are limited to T = 323.15K [J. Am. Chem. Soc., 1927, 49, pp 65-78]. In this thesis, it was concluded that con-
ventional Equations of State (EoS) failed to predict the equilibrium coexistence compositions of the liquid
and gas phase. Therefore, molecular simulations were used, which adequately predicted the water content
in compressed hydrogen.
The second part of this thesis was to select the best drying method for hydrogen dehydration. Previous work
showed that absorption using Ionic Liquids (IL) could be an interesting alternative to traditional drying meth-
ods. Therefore, absorption experiments were performed on the ILs 1-Ethyl-3-methylimidazolium acetate
[EMIM][Acetate], 1-Butyl-3-methylimidazolium chloride [BMIM][Chloride], and 1-Butyl-3-methylimidazolium
octyl sulfate [BMIM][Octyl Sulfate] to test the feasibility of ILs for hydrogen dehydration. The ILs were se-
lected on stability, safety, and on the activity coefficient of water in the given IL. The experiments showed that
ILs can be used as an absorbent to dry hydrogen gas. A part from the initial selection criteria, the experiments
showed that surface tension is an important criterion for future IL selection, as it is strongly related to foam
production. Finally, the very hygroscopic ILs tested turned out to be capable of absorbing large amounts
of water, but a lot of energy was required to regenerate the ILs, making them less attractive as a potential
absorbent.
iii
C ONTENTS
Abstract iii
1 Introduction 1
1.1 Research objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Drying Methods 5
2.1 Overview of possible drying methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.1 Membrane separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Adsorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.3 Cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.4 Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Method selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3 Thermodynamics 11
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Experimental Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3 The Gamma/Phi Formulation of VLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.4 Cubic Equations of State (EoS) for VLE calculations . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4.1 Density Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4.2 Fugacity Coefficient calculations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.5 Molecular Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.5.1 Molecular Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.5.2 Gas-Gas immiscibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4 Ionic Liquids 25
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2 Ionic Liquid Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.1 (Modified) UNIFAC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.2 COSMO-RS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.3 Molecular Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2.4 Literature and availibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5 Experiments 31
5.1 Assumptions & Simplifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.2 Materials & Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.3 Test Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.3.1 Absorption experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.3.2 Desorption experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.3.3 Karl Fischer titration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6 Results & Discussion 35

6.1 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.1.1 Bubble formation and Foaming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.1.2 Color change. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.1.3 Crystallization range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.2 Vapor Liquid Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.3 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6.3.1 EMIM Acetate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
v
vi C ONTENTS
7 Conclusions 43
7.1 Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
7.2 Ionic Liquids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
8 Recommendations 45
8.1 Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
8.2 Ionic Liquid Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
8.3 Ionic Liquid Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
8.4 Next phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Bibliography 47
A Experimental procedure 53
A.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A.2 apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
A.3 water content regulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
A.4 Ionic Liquid Scrubber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
B Solubility of Water in Hydrogen at High Pressure:
a Molecular Simulation Study 57
1
I NTRODUCTION
In 2015, The United Nations (UN) agreed on reducing the world-wide greenhouse gas emissions. As a result,
195 members of the United Nations Framework Convention on Climate Change (UNFCCC) signed the agree-
ment that aims to keep the global average temperature increase to well below 2°C, aiming for 1.5°C, above
pre-industrial levels [1]. On top of that, the UN has set 17 sustainable development goals for 2030 to end
poverty and hunger worldwide [2]. As access to energy is a key enabler to defeat poverty, the worldwide en-
ergy demand is expected to grow by 40% by 2040 [3]. To achieve the UN sustainable development goals, while
simultaneously reducing greenhouse gas emissions, new forms of energy sources are needed. Fossil fuels,
such as coal, oil, and natural gas, are currently accountable for 80% of the total primary energy supply [3] and
renewable energy projects, such as wind and solar (PV) parks, are rapidly being built to replace fossil fuels.
It is expected that 40% of the total generated electricity will be generated from renewable energy sources in
2040 [3].
However, cheap electricity production from renewable sources is only a part of the solution. The other part
is to integrate this electricity into the existing electricity grid. In the first place, electricity production from
renewable sources, like wind and solar (PV), are intermittent and therefore lead to a mismatch between sup-
ply and demand of energy [4]. To close this gap between supply and demand, energy storage is required
[5, 6]. In the second place, transporting electrons over large distances is expensive. When distances become
significant (>hundred kilometers), it becomes cheaper to transport molecules through a pipe rather than to
transport electrons through a power cable [7]. This holds even more where natural gas infrastructures are
already in place [7].
Hydrogen is a potential candidate to overcome both problems [7, 8], as hydrogen can be used to reproduce
electrons via fuel cells, for heating or as a raw material for the chemical industry [4, 9]. The major drawback
of hydrogen is the low volumetric energy density at room temperature and atmospheric pressure. To use hy-
drogen for storage or transportation, the density must be increased significantly [10]. One possible way to do
this is cryogenic cooling. However, hydrogen has a boiling temperature of 20.25K, which makes this process
very energy intensive. The other option, which is most common at the moment, is gaseous compression. For
instance, pressurized gaseous hydrogen is typically stored in Fuel Cell Electric Vehicles (FCEV) at P = 350 bar
or P = 700 bar [10]. Such a tank has a capacity of around 100 to 150 litres and stores 4 to 6 kg of hydrogen,
which provides a range of approximately 500 km [10]. High pressure storage tanks are installed at refueling
stations that temporary store the hydrogen up to 1000 bar. This high pressure is required to fill the hydrogen
tank within the target time of 3 to 5 minutes [10].
Currently, hydrogen gas is compressed at refueling stations, using mechanical compressors. These conven-
tional compressors, including piston, compressed air, diaphragm, or ionic compressors, require an average
of 6 kWh/kg of energy to compress hydrogen from 10 to 400 bar [11]. The required compressors at refueling
stations typically have a low throughput and a high pressure ratio. Under these constraints, the efficiency of
mechanical compressors drop significantly. Furthermore, mechanical compressors require a lot of mainte-
nance and produce a lot of noise, which is a problem on locations where acoustical emission is a constraint
[12].
An emerging alternative compressor is the Electrochemical Hydrogen Compressor (EHC). It has no moving
parts and therefore does not require a lot of maintenance and operates fully silent [12]. HYET BV has devel-
oped an EHC that has the potential of bringing compression costs down to 3 kWh/kg [11]. A pressure level
1
2 1. I NTRODUCTION
of 1000 bar was achieved at the compressor outlet. This shows that HYETT’s EHC is capable of compressing
to the desired pressure level of a refueling station storage tank. The working principle of an EHC operation
is similar to a Proton Exchange Membrane (PEM) fuel cell [13]. The compressor consists of a number of
stacks in parallel, where every single stack consists of a low pressure and a high pressure side, separated by a
membrane that is only permeable for hydrogen protons and not for molecules. The membrane is positioned
between two platinum catalysts containing electrodes. In Fig. 1.1, the working principle of the EHC is shown.
Just like a fuel cell, a potential difference over the electrodes splits a hydrogen molecule into two protons that
are transported through the membrane. However, unlike a fuel cell where hydrogen forms water with an oxy-
gen molecule, the protons are here reformed to hydrogen molecules at elevated pressures [12]. The reactions
are presented below in Eqs. (1.1) to (1.3) [12]:
Figure 1.1: Working principle of an Electrochemical Hydrogen Compressor (EHC). Source: HYET BV.
Anode HLP +
2 → 2H + 2e
−
(1.1)
Cathode 2H + 2e + −
→ HHP
2 (1.2)
Overall HLP
2 → HHP
2 (1.3)
The EHC has several advantages compared to traditional technologies. Apart from the higher efficiency,
especially at high compression ratios [12], low maintenance requirements, and fully silent operation. From
an electrochemical point, the compression ratio could be infinitely large. Only back diffusion of hydrogen
from high to low pressure, is limiting higher pressure ratios [12]. Furthermore, due to the highly selective
membrane, the compressor can also operate as a purifier, since other molecules than H2 are not able to pass
the membrane. Finally, the EHC is very well suited to scale up, due to its limited size [12]. A drawback of the
EHC, which is comparable to that of a fuel cell, is high material costs. Mainly the platinum catalyst, which
is required to resist the corrosive environments in the compressor, is very expensive [12]. Another disadvan-
tage is that the resulting hydrogen gas is saturated with water. This water comes from the membrane that is
always hydrated. This is because water enables the transport of protons through the membrane. This causes
a problem for FCEV applications, as the maximum water content in hydrogen used for PEM fuel cells in road
vehicles, is 5µmol water per mol hydrogen fuel mixture [14]. This maximum water content is described in
Table 1 of the ISO 14687-2:2012. The International Organization for Standardization (ISO) states that wa-
ter provides a transport mechanism for water-soluble contaminants such as K+ and Na+ when present as
an aerosol [14]. For conventional compressors, it is better to remove water prior to compression, as water
can damage the compressor and it accumulates as a liquid, due to the pressure increase. However, water
removal prior to compression is not applicable to EHC, as the output gas is saturated with water. Therefore,
the hydrogen gas needs to be dehydrated at high pressure.
1.1. R ESEARCH OBJECTIVES 3
1.1. R ESEARCH OBJECTIVES

A landscaping study, described in Chapter 2, was performed to search for potential drying methods, includ-
ing membrane separation, adsorption, absorption and cooling. The most common methods to dehydrate
gases are adsorption and absorption. This is elaborated on in Chapter 2. To select the best method for hydro-
gen dehydration, it is essential to have a good understanding of the thermodynamics at the given pressure
conditions, beginning with the saturated water content in compressed hydrogen. Unfortunately, there are
very little experimental data available on the Vapor Liquid Equilibrium (VLE) of the H2 -H2 O system at ele-
vated pressures. To the best of our knowledge, the only experimental data describing the water content in
the H2 -H2 O mixture for pressures exceeding 300 bar are from 1927 (limited to T = 323.15K) [15]. Therefore,
thermodynamic modeling was required to determine the water content in compressed hydrogen, resulting
in the first objective of this research:
1) Determine the water content in compressed hydrogen up to 1000 bar, in the temperature range of 0°C to 70°C.
In parallel, the different drying methods were evaluated on their feasibility to dry hydrogen gas at high pres-
sures. Previous studies showed that absorption using Ionic Liquids (IL) could be an interesting alternative to
traditional drying methods [16–19]. However, there has been no work produced on the use of ILs as a drying
agent for hydrogen dehydration. Therefore further research was required to test the feasibility of using ILs as
an absorbent. This resulted in the second objective:
2) Test the feasibility of Ionic Liquids for high pressure hydrogen dehydration.
1.2. A PPROACH
WATER CONTENT IN COMPRESSED HYDROGEN
Chapter 3 describes different thermodynamic models to calculate VLE of H2 -H2 O systems. In industrial ap-
plications, the most common method is to use cubic Equations of State (EoS), because of their simplicity
[20, 21]. First the Peng-Robinson (PR), Soave Redlich Kwong (SRK), and Shell’s Modified Intensified Redlich
Kwong (SMIRK) EoS were used to predict the water content at elevated pressures. The predicted composi-
tions of both the liquid and the vapor phase, calculated using all the different EoS, were not in good agreement
with the available experimental data. Since the conventional EoS failed to describe the VLE of the H2 -H2 O
system at elevated pressures, more physically based models were required. SAFT types EoS are more physi-
cally based, but they still require a temperature dependent binary interaction parameter k i j [22]. Due to the
lack of experimental data at high pressure, it is not possible to fit such a parameter. Therefore, force field
based molecular simulations were used to study the phase coexistence of the H2 -H2 O system. These molecu-
lar simulations showed excellent agreement with experimental data and resulted in the paper from [Rahbari,
Brenkman, and co-workers.] [23], attached in Appendix B.
I ONIC L IQUIDS AS DRYING AGENT

To test the feasibility of ILs as a drying agent, literature was consulted to see if ILs have been used for hydrogen
dehydration before. Although no work applied the concept on hydrogen, some work described the use of ILs
to dry natural gas [17–19], which is in many ways similar to hydrogen dehydration. Chapter 4 gives a brief
description of ILs in general and on what criteria they should be selected on for hydrogen dehydration. To
test the feasibility, preliminary experiments were executed, which are described in Chapter 5. Finally the
results are discussed in Chapter 6.
2
D RYING M ETHODS
There are several methods to remove impurities from a gas stream [24]. An overview of different drying meth-
ods is provided in this chapter, including their pros and cons. As elucidated in the introduction, the required
drying process is different from most conventional ones. Usually, gases are purified before they are com-
pressed. In this case, an Electrochemical Hydrogen Compressor (EHC) is used, which causes the hydrogen
gas to leave the compressor saturated with water. Therefore, it does not make sense to dry the hydrogen
prior to the compression stage, and the dehydration must take place at elevated pressures. Other specific
constraints for this project are the required purity and the small scale of the unit.
• Pressure
Present-day refueling stations use mechanical compressors, where water is removed from the hydro-
gen gas prior to the compression stage, as water could damage the compressor. Water is removed using
pressure, or temperature swing adsorption [25]. If hydrogen is compressed using an EHC, it leaves
the compressor saturated with water. Therefore one must dry the hydrogen at a pressure of 875 bar.
Dehydration units are usually operated at low pressure, so this constraint challenges the existing tech-
nologies.
• Purity
The desired purity of gases strongly depends on the application. If hydrogen is used for heating or in-
dustrial processes, it usually does not require a very low water content. For Fuel Cell Electric Vehicles
(FCEV) however, hydrogen specifications are very strict [14]. One of the key constraints is the maximum
water content in the pressurized hydrogen. The ISO standard requires a maximum water concentration
of 5 ppm, with the following motivation [14]:
"Water (H2 O) generally does not affect the function of a fuel cell, however; it provides a transport mech-
anism for water-soluble contaminants such as K+ and Na+ when present as an aerosol. Both K+ and
Na+ are recommended not to exceed 0.05 µmol/mol. In addition, water may pose a concern including
ice formation for onboard vehicle fuel and hydrogen dispensing systems under certain conditions. Water
should remain gaseous throughout the operating conditions of systems."
Since the temperature of the hydrogen gas increases significantly during refueling [26], it is not likely
that ice formation occurs in the dispensing system or the hydrogen tanks. In the dispensing system the
temperature of the hydrogen increases, because the Joule-Thomson coefficient at the pressure and the
temperature conditions of refueling is negative [26, 27]. Therefore, the hydrogen gas enters the cylinder
at an increased temperature. This temperature increases even further because of the compression of
the gas inside the cylinder, which is called heat of compression [26]. Various studies have modeled and
experimentally validated the temperature rise in hydrogen tanks during refueling [26, 28–30].
• Footprint/Scale
The final application specific criterion is that the whole unit must be installed at a refueling station,
instead of an industrial plant. Therefore the drying unit must have a small footprint and operate effi-
ciently at low throughput.
5
6 2. D RYING M ETHODS
Besides these project specific criteria, more general criteria are listed below:
• low energy use
• low hydrogen loss
• low Capital Expenditure (CAPEX) & Operational Expenditure (OPEX)
• low safety & environmental risks
2.1. O VERVIEW OF POSSIBLE DRYING METHODS

The considered methods are elaborated on in this section. Including membrane separation, adsorption, ab-
sorption, and cooling.
2.1.1. M EMBRANE SEPARATION

One of the possible methods for gas separation is membrane separation. The advantages over other methods
include low cost for low volumes, high mechanical reliability due to simplicity of the operation, low energy
costs and small footprint. Membrane separation is commercially available for dehydration of air and natural
gas. The biggest obstacle for applying this technology for hydrogen dehydration is the required purity of the
hydrogen gas. Polymeric membranes for example, which are commercially used for this application, achieved
dew points down to -40°C at atmospheric pressure, which is not dry enough [31].
I MMOBILIZED L IQUID M EMBRANES

Immobilized liquid membranes (ILM) can overcome this problem, as they can achieve very high selectivities.
ILMs are basically polymeric membranes, in which the pores are filled with an organic fluid. This increases
the selectivity significantly and it has been demonstrated to work for removing water and other impurities
from biogas. However, to reach this level of purity, multiple stages are required, resulting in a significant
pressure drop [32]. Unfortunately, there are some obstacles with this method that need to be overcome before
the technology can concluded to be feasible. Major problems are pressure drop, instability, flux decline and
liquid phase evaporation [16]. Although ILM may be a promising candidate in the future, the technology is
currently not ready.
C ARBON M OLECULAR S IEVE M EMBRANES

Carbon Molecular Sieves (CMS) are another promising alternative for polymer mebranes. CMS membranes
can be prepared by controlling the carbonization procedure during carbonization of a polymeric material.
They can be designed for very specific applications and therefore achieve very high selectivities [33]. The
main drawback of CMS membranes is that water can block the very small pores, blocking the transport of
hydrogen molecules.
2.1.2. A DSORPTION
Adsorption is the adhesion of a molecule to the surface area of the adsorbent and is used to remove impurities
in gases that have a low concentration of contaminants. The desired substance, in this case water, is adsorbed
by an adsorbent that eventually becomes saturated and must be regenerated or replaced. There are several
possibilities to regenerate the adsorbent [24]:
1. Temperature Swing Adsorption (TSA).

The absorbed water is evaporated out of the adsorbent by increasing the temperature and purging the
column with dry hydrogen or an inert gas.
2. Pressure Swing Adsorption (PSA).

Vaporization of the water by decreasing pressure, often by creating vacuum. Just like TSA, dry hydrogen
or an inert gas is used for purging.
3. Inert Purge Stripping (without changing T or P).

When the adsorbent is saturated with water, an inert gas is used to vaporize the water out of the adsor-
bent, without changing the temperature or pressure.
2.1. O VERVIEW OF POSSIBLE DRYING METHODS 7
4. Chemical desorption.
A fluid that adsorbs water even more than the used adsorbent is used to retrieve the water from the
adsorbent.
Inert purge stripping only works if the hydrogen input is relatively wet, resulting in a large difference in con-
centration between product gas and purge gas. That is clearly not the case here, so therefore inert purge
stripping is not an option. Chemical desorption is most likely not effective due to the required purity of the
hydrogen gas. The adsorbent must already be very hygroscopic, to achieve 5 ppm water content. Therefore, it
is not likely that the adsorbent easily releases the adsorbed water to a chemical absorbent. In similar applica-
tions, TSA and PSA, or a combination of both, are typically used. They also proved to be capable of achieving
the required level of purities. The major drawback of PSA for HP drying is the fact that it is not desirable to
depressurize the adsorbent column. Re-pressurizing the column back to 875 bar costs a lot of time and en-
ergy. Fig. 2.1 shows a typical scheme of an adsorption unit, in this case TSA [34]. The drying unit consists of a
minimum of two columns, often three. While one bed is drying the hydrogen, the other is regenerated. This
is done by purging hot gas over the column. This gas could be dry hydrogen, or an inert gas.
Figure 2.1: Schematic of a typical TSA unit. Solid lines indicate the formation where column 1 is used as adsorption column, and 2 is
being regenerated. The dashed lines indicate the process when it is the other way around. In the adsorption column, wet hydrogen
enters and leaves the column dehydrated. Heated purge gas enters the regeneration column and leaves it hydrated, to be cooled and
separated from water. Hydrogen loss can be prevented by recycling the purge gas (indicated by the dotdashed line).
Commercially available adsorbent agents that absorb water are activated alumina, silica gel, and synthetic
sodium or calcium aluminosilicate zeolites (molecular sieves). Literature shows that activated alumina and
zeolites have been proven to work at high pressure, respectively at 3000 atm [35] and 5000 bar [36]. No damage
of adsorbent or discontinuity in adsorption isotherms were reported. Silica gel, however, is affected by high
pressures. Each time silica gel is compressed it gets granulated, which leads to destruction of the skeleton
structure [25].
Adsorption, and in particular TSA, seems a very reasonable option for high pressure dehydration. It is capable
of obtaining very low ppm levels of water content and can operate at high pressure. The major disadvantage
is that the efficiency is strongly related to the hydrogen throughput. At refueling stations, the capacity is
relatively low, making this option less attractive.
2.1.3. C OOLING
C ONDENSING
The most straight forward method for reducing the water content in hydrogen gas is cooling the hydrogen
until all water condenses. However, there is little known about the dewpoints, i.e. at what temperature the
water condenses, of water at elevated pressures. Further research is required on the thermodynamics of high
pressure H2 -H2 O systems, to see if condensing is a serious option. Nonetheless, even if condensing does not
result in the desired level of purity, it could very well be a good pre-treatment method. The final bit of water
can then be removed by one of the other methods, described in this chapter.
F REEZING
It is also possible to cool the hydrogen gas even further, so that the water vapor will directly turn into ice,
called deposition. This method would require two units, just like adsorption. In this case, such a unit will be a
heat exchanger (HEX). Hydrogen gas is cooled well below zero °C, and the ice will build up in the HEX. When
the HEX reaches the maximum ice level, it goes in regeneration mode and the other unit is used. Regeneration
would simply mean heating the HEX, until all ice melts and the resulting water is removed. Disadvantages of
this method are very much comparable to those for adsorption.
2.1.4. A BSORPTION
The final drying method considered is absorption. In contrast to adsorption, where water is adsorbed by a
solid agent, absorption consists of liquids that absorb water from the vapor phase. Absorption is already used
for applications like natural gas dehydration, which is quite similar to hydrogen dehydration. In those cases,
Diethyleneglycol (DEG) or triethyleneglycol (TEG) are used as absorbents. Fig. 2.2 shows a typical absorption
unit [34]. The wet hydrogen is infused in the bottom of a tray column (left side of figure) and the dry absorbent
flows in from the top. During the contact, the absorbent is enriched with water where after the water is boiled
out in a heater (right side of figure). At elevated pressures, absorption can be favorable over adsorption, as
Figure 2.2: Schematic of a typical absorption unit. The dry absorbent enters the absorption column and absorbs the water from the wet
hydrogen gas. It leaves at the bottom where after it is heated and the water is stripped out the absorbent. Finally, the dry absorbent is
cooled back and it re-enters the absorption column.
the absorption column does not need to be depressurized and no purge gas is required. Due to the high
2.2. M ETHOD SELECTION 9
pressure, it is also likely to achieve the required purities. However, just like TSA, absorption also becomes
more efficient at higher flow rates. Another disadvantage is the contamination of the resulting hydrogen gas,
due to evaporation of the absorbent.
I ONIC L IQUIDS
The latter disadvantage can be overcome by using Ionic Liquids (ILs) instead of conventional absorbents like
TEG. Because ILs practically have no vapor pressure, contamination of the hydrogen gas by evaporation of
the absorbent is prevented. A few studies have shown that absorption using ILs could potentially work for
natural gas dehydration [17, 19]. To the best of our knowledge, no literature exists that describes the use of
ILs for hydrogen dehydration.
2.2. M ETHOD SELECTION

To select the best method for high pressure hydrogen dehydration, pros and cons of the different methods
were tested against the criteria stated in Chapter 2. To the best of our knowledge, no previous work described
the use of membranes that were capable of drying gases to a water content below 5 ppm, especially at el-
evated pressures. Therefore, this method is no longer considered for hydrogen gas dehydration. Note that
membranes can improve in the future, making this option interesting again.
The three other methods are, based on the three main criteria, expected to be capable of drying hydrogen gas
at the given pressures to the required purity.
• Adsorption is the most obvious method to go for, as it is most commonly used to remove contaminants
at low concentrations from gases. Previous work showed that adsorption systems are capable of drying
gases at elevated pressure [37]. PSA, which is the most efficient method based on mass transfer, requires
decompression of the adsorption column, which is not favorable. TSA is an alternative option that
does not require depressurization of the columns. However, the mass transfer to the purge gas is less
efficient with TSA and therefore this process requires much more purging gas. Even if this purging
gas is recycled, all this gas needs to be heated and cooled back again. Currently HyET uses a TSA in
a demo-unit that has a hydrogen gas loss of 20%. Although this can be reduced significantly, TSA will
unavoidably lead to a substantial hydrogen loss, or an efficiency reduction of the entire compressor
unit, due to recycling of the purge gas.
• Absorption has the major advantage that the absorbent can be pumped out the absorption column,
without decompressing the column. Therefore, only a single column is needed; this reduced number
of components is favorable at low volumes. The absorbent can be regenerated separately, by reducing
pressure and/or increasing temperature. This does not require a purge gas, resulting in hydrogen loss
or energy loss due to recycling. The major risk related to absorption is vaporization of the absorbent.
This is a concern, especially due to the strict specifications of the hydrogen gas [14]. Ionic Liquids
are a potential solution to this problem, as they have practically no vapor pressure. Table 2.1 shows
the vapor pressures of water, Triethylene Glycol (TEG), which is the most commonly used absorbent,
and six hygroscopic ILs [19]. To the best of our knowledge, no previous studies exist that used ILs to
dehydrate hydrogen gas. Previous studies also concluded that ILs are promising as absorbent, which
makes it an area that needs more research [25, 38]. Therefore, this work focuses on testing the feasibility
of ILs to dehydrate hydrogen gas.
Table 2.1: Vapor pressures of several liquids.

(a): calcuted using Wagner equation
Vapor Pressure
Liquid
/ (Pa)
Water 7.0 × 104(a)
TEG 2.4 × 101 [38]
[choline][glycolate] 2.5 × 10−5 [19]
[DMIM][DMPO4 ] 2.5 × 10−6 [19]
[EMIM][MeSO3 ] 3.9 × 10−8 [19]
[BMIM][BF4 ] 1.5 × 10−8 [19]
[EMIM][Et(EG)2 SO4 ] 7.5 × 10−9 [19]
[EMIM][EtSO4 ] 7.0 × 10−9 [19]
• The final method is cooling the hydrogen gas, until all water is condensed or frozen out. This is es-
pecially interesting for refueling applications, because the refueling protocol currently requires pre-
cooling. The reason for this is that the temperature of hydrogen increases as it is filled in a vehicle
tank [39] and the temperature in the tank may not exceed 85°C [40]. To determine if cooling is the
best method to dry the hydrogen gas, it is essential to know the saturated water content in compressed
hydrogen at different temperatures and pressures. Unfortunately, as mentioned in the introduction,
there are very little experimental data on the water content in high pressure hydrogen. Therefore, prior
to comparing the three different methods, the first part of this work is focused on calculating the satu-
rated water content in hydrogen gas.
3
T HERMODYNAMICS
3.1. I NTRODUCTION
As stated in Chapter 2, it is crucial to know the water content in compressed hydrogen at different tempera-
tures, to compare different drying methods. First, the water content of the hydrogen gas that leaves the EHC
is needed. In this specific application, hydrogen leaves the compressor at a pressure of 875 bar and temper-
atures ranging from 35°C to 40°C. Thereafter the water content at lower temperatures is required, to see if
cooling is a viable option. Will it be possible to remove the excess water by condensing it, or could this be
at least a part of the solution? The first major part of this thesis project was to determine the water content
in the compressed hydrogen, at different conditions. The saturated water content is the maximum amount
of water vapor that can be present in compressed hydrogen, before the water condenses. This chapter de-
scribes different methods to calculate the Vapor Liquid Equilibrium of hydrogen and water at high pressure
conditions. The calculated data are compared to experimental data. Due to the high pressure range, no VLE
experiments in this range were executed. At these conditions, the ideal gas law and the ideal mixing rules do
no longer apply [41].
3.2. E XPERIMENTAL D ATA

For the VLE, both the liquid phase and the vapor phase composition need to be studied. Therefore, experi-
mental data for both the solubility of hydrogen in water and the concentration of water vapor in compressed
hydrogen are relevant. Fig. 3.1 shows all available data for water vapor in compressed hydrogen [15, 42–45].
Unfortunately, very little data are available for the water content in hydrogen for pressures exceeding 300 bar.
The only paper is from Bartlett [15], who measured the water vapor in compressed hydrogen at 50°C. In that
paper there are only three data points (400/600/1000 atm.) higher than 300 bar. The most common method
for VLE calculations, is to fit an Equation of State (EoS) on available data. Since there are little data available
at high pressure conditions, the fitted EoS must be extrapolated, which is not very reliable.
3.3. T HE G AMMA /P HI F ORMULATION OF VLE

The simplest equation for VLE calculations is applying Raoult’s law. Application of Raoult’s law acquires two
major assumptions[46]:
• The vapor phase is considered an ideal gas (ideal gas law applies).
• The liquid phase is an ideal solution.
It is obvious that both assumptions do not apply for the given application. The ideal gas law does not hold, as
pressures are very high (950 bar). Fig. 3.2 shows the density of hydrogen, calculated using the ideal gas law,
and the density from RefProp®1 [27]. The increasing error shows that when the pressure increases, hydrogen
can no longer be treated as an ideal gas. At 100 bar and 50°C the compressibility factor Z already shows an
error of 6% compared to the ideal gas law (Z = 1) [27]. Since this error increases drastically for the output
1 Refprop® is a program, developed by the National Institute of Standards and Technology, that calculates the thermodynamic properties
of industrially important fluids and their mixtures [27].
11
12 3. T HERMODYNAMICS
Figure 3.1: Available experimental data of the water content in compressed hydrogen, at temperatures between 0°C and 250°C [15, 42–
45].
pressure of the EHC, the ideal gas law can not be used. Furthermore, water is a polar molecule and hydrogen
is not, which implies that they are not chemically similar and the liquid phase can therefore not be treated
as an ideal solution. Two common approaches for calculating Vapor Liquid Equilibrium data are the γ − φ
80
Ideal
70 Refprop
60
50
/[kg/m 3]
40
30
20
10
0
0 200 400 600 800 1000 1200
P/[bar]
Figure 3.2: Density of hydrogen as a function of pressure, at a temperature of 50°C. Predicted using ideal gas law(diamonds), and densities
from RefProp® [27] (triangles).
formulation and the φ−φ formulation. The γ−φ method is the traditional approach that requires an activity
coefficient of water in the liquid mixture γH2 O . This formulation is based on the fact that the fugacity in the
liquid phase is equal to the fugacity in the vapor phase, which is derived from the chemical potential [41]:
fˆVi = fˆLi , (3.1)
where the hat indicates that it concerns a property of a substance in a mixture. The fugacity can be in-
terpreted as the pressure, corrected for the non-idealities. Thus, for an ideal gas, the fugacity is equal to the
3.3. T HE G AMMA /P HI F ORMULATION OF VLE 13
pressure. The fugacity coefficient Φ is the ratio of the fugacity to the pressure:
f
Φ= (3.2)
P
The vapor phase fugacity of a component in a mixture is:
fˆVi = Φ̂i y i P. (3.3)
The fugacity coefficient of a component in a mixture Φ̂i should not be confused with the fugacity coefficient
of a pure component Φi . The latter varies only with temperature and pressure, while the former also varies
with the composition of the mixture. Section 3.4.2 describes different EoS to find Φ̂i . The liquid phase fugacity
of a component in a mixture is [46]:
Vi (P − P sat
" #
ˆL sat sat i )
f i = γi x i Φi P i exp , (3.4)
RT
where γi is the activity coefficient of i in the solution, Φsat

i is the fugacity coefficient of the pure component
at saturated pressure and Vi is the liquid molar volume of the pure component i. The exponential part of the
equation is called the Poynting factor. The Poynting factor corrects for the compressibility effects within the
liquid. This factor is usually neglected, as it is equal to 1 at low to moderate pressures. Since pressures are
very high in this work, it can not be neglected. Fig. 3.3 shows the Poynting Factor at 50°C and for different
pressures. At pressures close to 1000 bar, the Poynting factor is almost 2. Combining Eqs. (3.3) and (3.4) into
1.9
1.8
Poynting Factor/[-]
1.7
1.6
1.5
1.4
1.3
1.2
1.1
1
100 200 300 400 500 600 700 800 900 1000
P/[bar]
Figure 3.3: Poynting Factor, as a function of pressure, at 50°C.
Eq. (3.1) gives the γ − φ formulation:
Vi (P − P sat
" #
i )
Φ̂i y i P = γi x i Φsat sat
i P i exp (3.5)
RT
The detailed derivation of the γ − φ formulation is described in [Smith et al.]. P sat

i is the saturated vapor
pressure of water and is calculated with the Wagner equation [47]:
ln(P sat
r ) = (aτ + bτ
1.5
+ cτ3 + d τ6 )/T r 2 (3.6)
Where P sat
r =P
sat
/P c is the reduced saturated vapor pressure, T r = T /T c is the reduced temperature, and
τ = 1−T r . Since hydrogen can not be liquefied at temperatures above the critical temperature (= 32.98K ), the
saturated vapor pressure of hydrogen can not be calculated at the required temperatures.
2 The last two exponents are different for water than for other substances.
3.4. C UBIC E QUATIONS OF S TATE (E O S) FOR VLE CALCULATIONS

In industrial applications, the most commonly used method to describe VLE systems is using cubic Equations
of State (EoS). This section elaborates on different EoS to describe the VLE of the H2 -H2 O system. These EoS
can be used to describe pure components, but combined with mixing rules, they can also describe properties
of components in mixtures. The performance of each cubic EoS, used in this work, to describe the VLE of
the H2 -H2 O system will be evaluated. To give a better description of real gases, so where the compressibility
factor Z 6= 1 [46], extra parameters are required to describe the system using an EoS. These parameters are
component specific, based on the critical pressure and temperature of that component. These parameters
are typically a function of the reduced temperature T r = T /T c . An EoS that describes both gases and liquids
must be at least cubic in V. These cubic EoS are widely used and can be formulated in a general form [47]:
RT Θ(v − η)
P= − (3.7)
v − b (v − b)(v 2 + δv + ²)
Θ = a · α(T r ) (3.8)
where a, b, η, α(T r ), δ, ² differ for different cubic EoS. For all EoS applied in this work η = b.
3.4.1. D ENSITY C ALCULATIONS

P ENG -R OBINSON E O S
To make the parameters dependent on temperature, an alpha function is introduced. This alpha function
does not only vary with temperature but also introduces a third component specific parameter: the acentric
factor ω. This parameter corrects for the assumption that all fluids, when compared at the same reduced tem-
perature and reduced pressure, have the same compressibility factor [46]. The Peng-Robinson (PR) equation
is as follows (it is described in detail in [Lin and Daubert][48]):
RT a · α(T r )
P= − (3.9)
v − b v(v + b) + b(v − b)
0.45724(RT c )2
a= (3.10)
Pc
0.0778RT c
b= (3.11)
Pc
α(T r ) = (1 + (0.37464 + 1.54226ω − 0.26992ω2 )(1 − T r 1/2 ))2 (3.12)
Eq. (3.12) is the regular alpha function for PR. However, there are some small variations of this function opti-
mized for specific applications.
S OAVE ’ S M ODIFIED R EDLICH K WONG E O S

Soave’s modified Redlich Kwong (SRK) EoS is a variation on the more famous Redlich Kwong EoS. It is known
to perform well with respect to the vapor phase.
RT a · α(Tr )
P= − (3.13)
v − b v(v + b)
0.42724(RTc )2
a= (3.14)
Pc
0.08664RTc
b= (3.15)
Pc
α(Tr ) = (1 + (0.480 + 1.574ω − 0.176ω2 )(1 − Tr1/2 ))2 (3.16)

3.4. C UBIC E QUATIONS OF S TATE (E O S) FOR VLE CALCULATIONS 15
S HELL M ODIFIED I MPROVED R EDLICH -K WONG (SMIRK) E O S

The SMIRK-EoS is a 2-parameter Redlich-Kwong type EoS where not only a · α, but also b is made a function
of temperature in order to extend its applicability [49]. Both a and b have the following form [49]:
Y = y 1 + y 2 ln[1 + exp(y 3 (X − y 4 ))] − y 5 ln[1 + exp(−y 3 (X − y 4 ))] (3.17)
where: Y = a or b
y i = coefficient of the a- or b-function
X = Tc /T
The a- and b- parameters are obtained from regression of vapor pressures and liquid molar volumes along
the saturation curve (VLE) of pure components [49]. Since the details of the SMIRK-EoS are not within the
scope of this thesis, the reader is referred to the work of Drexhage, J.J. and Welsenes, for further details. Shell
developed their own software tool STFlash® for thermodynamic flash calculations including VLE, which is
based on Shell thermodynamic methods and parameter databased including the SMIRK EoS. SMIRK results
in this work are obtained by using this tool.
D ENSITY R ESULTS
Fig. 3.4 shows the densities of liquid water and hydrogen gas, which are predicted by the various EoS, com-
pared to data from Refprop® [27]. As expected, cubic EoS perform poorly in describing liquid densities using
conventional EoS. Only the SMIRK-EoS gives acceptable results, as Fig. 3.4a shows. However, it starts to de-
viate more as the pressure increases. Also note that the SMIRK-EoS is already fitted to available experimental
data. The vapor phase of hydrogen, however, is well described using this EoS.
1100 55
PR PR
SRK 50
SRK
SMIRK SMIRK
1050 Refprop Refprop
45
40
1000
35
950
/[kg/m 3]
/[kg/m 3]
30
25
900
20
850
15
10
800
750 0
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
P/[bar] P/[bar]
(a) water in the liquid phase (b) Hydrogen in the vapor phase
Figure 3.4: Densities of liquid water (a) and hydrogen gas (b) at 50°C, as a function of pressure. Predicted by EoS; PR (circles), SRK
(diamonds), SMIRK (squares), compared to experimental data from Refprop® (triangles) [27].
3.4.2. F UGACITY C OEFFICIENT CALCULATIONS

The fugacity and the fugacity coefficient depend on the temperature, pressure, and on the composition of the
mixture. The challenge in calculating the fugacity coefficients of components in mixtures is that the fugacity
depends on all mole fractions of all components in the mixture. All components have different interactions
with each other. To account for these interactions, mixing rules are introduced.
M IXING RULES
There are several ways to describe the dependence of different components with different mole fractions.
The most common mixing rule is the van der Waals mixing rule that uses a quadratic dependence between
components [50].
XX
am = xi x j ai j (3.18)
i j
X
bm = xi bi (3.19)
i
where a m and b m are the EoS parameters for the mixture, x is the mole fraction of each component in the
specific phase. Note that a i i and a j j are parameters corresponding to a pure component, while a i j and
a j i (i 6= j ) are called unlike-interaction parameters that describe the interaction between the components:
a i j = (1 − K i j )(a i a j )1/2 (3.20)
where
K 11 = K 22 = 0 & K 12 = K 21 = k i j (3.21)
k i j is called the binary interaction parameter, which is determined by fitting EoS results to experimental data.
As mentioned above, there are more complex mixing rules that also contain more interaction parameters. In
this work, the van der Waals mixing rules were applied for the PR and SRK EoS.
P ENG -R OBINSON E O S
The equation for the fugacity coefficient, predicted by the PR EoS, is [48]:
bi
ln(Φ̂i ) = (Zm − 1)−ln(Zm − B m )−
bm
 P
2 y i ai j

Ã p ! (3.22)
Am  j bi  Zm + (1 + 2)B m
p −  ln p
am bm

2 2B m Zm − (1 − 2)B m
where
am P
Am = (3.23)
R 2T 2
bm P
Bm = (3.24)
RT
P vg
Zm = (3.25)
RT
v g is the specific volume that is calculated by solving Eq. (3.9).
S OAVE ’ S R EDLICH K WONG E O S

The equation for the fugacity coefficient, predicted by the SRK EoS, is [48]:
 P
2 y i ai j

bi Am bi  Bm
µ ¶
 j
ln(Φ̂i ) = (Zm − 1) − ln(Zm − B m ) − −  ln 1 + (3.26)
bm Bm am bm Zm

SMIRK-E O S
The fugacity coefficients, predicted by the SMIRK-EoS, are again calculated using STFlash® . By default, the
SMIRK method uses the Second Michelsen-Huron-Vidal GExcess mixing rule [51]. This mixing rule uses 4
parameters that are fitted to experimental data. Since these parameters are fitted to a small set of data across
a relatively narrow range of pressure and temperature, it is not reliable to use these fitted parameters for
extrapolation.
F UGACITY C OEFFICIENT R ESULTS

Fig. 3.5a shows the water content in the H2 -H2 O mixture, calculated with the different EoS, compared to ex-
perimental data. This is calculated by substituting Eqs. (3.3) and (3.4) in Eq. (3.1), resulting in Eq. (3.27) for the
water fraction in the vapor mixture. Here the binary interaction parameter is not fitted to experimental data.
Referring to Section 3.4.2, this means that k i j = 0. For the SMIRK-EoS, this is slightly different, as it uses a 4-
parameter mixing rule. In this case all four parameters are set to zero. Fig. 3.5b shows the fugacity coefficients
Φ̂H2 O of water in the H2 -H2 O mixture, which follows directly from the fugacity coefficient equations.
V H2 O (P −P sat
· ¸
H2 O
)
Φsat H2 O P sat
H2 O exp RT
y H2 O = γH2 O x H2 O (3.27)
Φ̂H2 O P
Note that the amount of water dissolved in the liquid phase x H2 O if negligible for this equation and therefore
assumed to be equal to one. Accordingly, as x H2 O → 1, the activity coefficient of water γH2 O also goes to one.
Activity coefficient models have been used to calculate γH2 O and calculate x H2 O by iteration, but it did not
make any significant difference.
2000 1
Experimental data PR
1800 0.9
PR SRK
1600 SRK 0.8 SMIRK
SMIRK
1400 0.7
yH O /[PPM]
1200 0.6
/[-]
H O
1000 0.5
2
2
800 0.4
600 0.3
400 0.2
200 0.1
0 0
100 200 300 400 500 600 700 800 900 1000 100 200 300 400 500 600 700 800 900 1000
P/[atm] P/[atm]
(a) Water content in the H2 -H2 O mixture in the vapor phase. (b) Fugacity coefficients of water in the H2 -H2 O mixture in the vapor phase.
Figure 3.5: Water content (a) and fugacity coefficient of water (b), as a function of pressure, at a temperature of 50°C. Calculated using
PR-EoS (circles), SRK-EoS (diamonds) and SMIRK-EoS (x-mark), compared to experimental data (triangles) [15]. (k i j = 0)
Fig. 3.5 clearly shows that the experimental data are not accurately described by the different EoS. In these
calculations, the binary interaction parameter k i j is equal to zero. The next step is to fit k i j to experimental
data to see if better results are obtained. The parameter is gradually increased, until it results in an accurate
fit with the experimental data. For the SMIRK-EoS, the 4 interaction parameters, implemented in STFlash® ,
are used. Fig. 3.6 shows the fugacity coefficients and the water content fitted to the experimental data. In
this figure k i j is increased to 0.35, so that SRK fits with the experimental data. This value is very high for
a binary interaction parameter. k i j is used to make minor corrections in the a-function of the EoS. If this
value becomes very high (normal range is between -0.1 & 0.1), the a-function is not correct for this system at
these high pressure conditions. For the PR EoS, the binary interaction parameter needs to be increased even
further, in order to fit with the available data.
D IFFERENT T EMPERATURES
Although the fitted binary interaction parameter is very high and therefore not physical as explained in Sec-
tion 3.4.2, we also fitted this parameter to experimental data at other temperatures. Fig. 3.7 shows the water
content, predicted by the SRK-EoS and the SMIRK-EoS, compared to experimental data. For the SRK-EoS a
binary interaction parameter k i j = 0.35 is used and for SMIRK-EoS a more complex mixing rule, mentioned
in Section 3.4.2, is applied. Note that both SRK-EoS and SMIRK-EoS predict the water content accurately at
temperatures near 50°C, but as temperature increases, so do the errors. For SRK-EoS, this is not a surprise,
as the interaction parameter is fitted to experimental data at 50°C. The complex mixing rules used in SMIRK-
EoS, however, do not perform much better, as the predictions are very similar to SRK-EoS. It could well be
1
1500
Experimental data 0.9
PR
SRK 0.8
SMIRK
0.7
1000
yH O /[PPM]
0.6
/[-]
H O
0.5
2
2
0.4
500 0.3
0.2
PR
0.1
SRK
SMIRK
0
0 100 200 300 400 500 600 700 800 900 1000
100 200 300 400 500 600 700 800 900 1000
P/[atm]
P/[atm]
(b) Fugacity coefficients of water in the H2 -H2 O mixture in the vapor
(a) Water content in the H2 -H2 O mixture in the vapor phase. phase.
Figure 3.6: Water content (a) and fugacity coefficient of water (b) in the H2 -H2 O vapor mixture, as a function of pressure, at a temperature
of 50°C. Calculated using PR-EoS (circles), SRK-EoS (diamonds) and SMIRK-EoS (x-mark), compared to experimental data (triangles)
[15]. (k i j = 0.35)
possible that this EoS is also only fitted on the data at 50°C, but this is hard to find out. Even though a binary
interaction parameter of k i j = 0.35 is already very high, k i j is fitted to all experimental data at different tem-
peratures, with the purpose of finding k i j as a function of temperature. Fig. 3.8 shows the predicted water
content, where k i j is fitted to every single temperature. Fig. 3.9 shows the parameters at different tempera-
tures. It shows that, for T > 100°C , the binary interaction parameter seems very random and nowhere near
acceptable values. It is therefore not possible to find a temperature dependence of k i j , based on the available
data. Especially in the low temperature domain (T < 38°C ), which is of particular interest, no experimental
data are available and therefore it is not possible to predict the water content in this domain using a conven-
tional EoS. In the work of [Rahbari, Brenkman, and co-workers.], the fitted parameters were also tested for
the liquid phase [23]. These parameters predicted the solubility of hydrogen in the liquid phase even worse
than without a k i j .
350
150°C Experimental data 70°C

175°C
300 200°C SMIRK
SRK
250
200
P/[bar]
150 93°C 38°C
100
50
0
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4
- 10Log( yH O /[mol/mol])
2
Figure 3.7: Water content in the H2 -H2 O vapor mixture, as a function of pressure, at different temperatures. Calculated using SRK-EoS
(diamonds) and SMIRK-EoS (x-mark), compared to experimental data (triangles) [42–44]. (k i j = 0.35)
350
Experimental data
175°C 150°C SMIRK 70°C
300 200°C
SRK
250
200
P/[bar]
150 93°C 38°C
100
50
0
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4
- 10Log( yH O /[mol/mol])
2
Figure 3.8: Water content in the H2 -H2 O vapor mixture, as a function of pressure, at different temperatures. Calculated using SRK-EoS
(diamonds) and SMIRK-EoS (x-mark), compared to experimental data (triangles) [42–44]. (k i j fitted at every temperature (see Fig. 3.9))
0.4
0.2
-0.2
kij /[-]
-0.4
-0.6
-0.8
-1
-1.2
40 60 80 100 120 140 160 180 200
T/[°C]
Figure 3.9: Binary interaction parameter k i j , as a function of temperature, fitted to experimental data.
3.5. M OLECULAR S IMULATION

The results in Section 3.4, clearly show that the PR and SRK EoS are not capable of describing the H2 -H2 O
system at elevated pressures. The SMIRK-EoS only shows good results in the temperature range where it
is been fitted on experimental data. In this work, PR-EoS and SRK-EoS are used with the classical mixing
rules. An improvement of these conventional EoS, could be to use mixing rules that take the polarity of the
water molecule into account [21]. Another option is to use a more physically based equation of state, like a
SAFT-EoS [52]. However, Sun et al. showed that a temperature dependent k i j is still required [22]. Molecular
simulations are also physically based, but do not require a temperature dependent parameter. For this study,
molecular simulations are used to determine the water content in compressed hydrogen at different temper-
atures. These simulations resulted in the work of [Rahbari, Brenkman, and co-workers.], which accurately
describes the used method and results [23].
Figure 3.10: An overview of different modelling methods, based on their time and lenth scale. Source: T.J.H. Vlugt.
3.5. M OLECULAR S IMULATION 21
Fig. 3.10 shows the different scales of the modeling methods. Using cubic EoS, one can predict macro-
scopic properties of a given system using only a few parameters, at low computational cost. In many cases
cubic EoS perform really well. Therefore, they are most commonly used in industry [21]. Especially if abun-
dant experimental data is available, EoS are very reliable. The well known drawback is that the cubic EoS with
conventional mixing rules are not accurate for liquid density calculations or other thermodynamic properties
of complex fluids [21]. Therefore a k i j is always required, which is unfavorable due to the lack of experimen-
tal data. Furthermore mixing rules that for instance take polarity into account need to be modified [21].
Moving more to the bottom left of the figure, one finds modeling methods that are based on the physical
structure of single molecules, rather than entire systems. Molecular simulations take into account interac-
tions between the molecules in the system and can predict macroscopic properties of the entire system, using
statistical mechanics. To simulate the VLE of the H2 -H2 O system using molecular simulations, two boxes are
created that are filled with both molecule types. One box represents the liquid phase, the other box the vapor
phase. The interaction between molecules is based on the forcefields, selected for the specific molecules.
Typically, random configurations of the H2 -H2 O system are generated, proportional to Boltzmann distribu-
tion, and sampled using the Monte Carlo method. The Boltzmann distribution is a probability distribution
that indicates how likely it is that a given sample represents the real state of that system. For instance, if two
molecules overlap in the simulation, the internal energy becomes infinitely high. This results in a probability
of zero, meaning this sample is not used to calculate thermodynamic properties. Using this method, only
favorable configurations are used to predict properties like density and the chemical potential. Note that this
approach can not be used to calculate transport properties. Predicting transport properties requires molec-
ular dynamics, which is not included in this work. For every molecule, multiple forcefields are available. The
performance of the forcefield depends on the required properties and conditions of the given application. In
this study, the forcefields were selected on their prediction results of the pure components at high pressure.
The Marx [53] force field shows the best agreement with experimental data for hydrogen and the TIP3P [54]
predicts the chemical potential of water in good agreement with experimental data from Refprop® [27]. For
the comparison between the results with different forcefields, the reader is referred to the work of [Rahbari,
Brenkman, and co-workers.] [23], which is attached in Appendix B.
3.5.1. M OLECULAR S IMULATION R ESULTS

Fig. 3.11 shows the molecular simulation results. The results for the water content in the vapor phase are in
excellent agreement with the experimental data. The simulations are overpredicting the solubility of hydro-
gen in the liquid phase. Fortunately, these predictions are not essential for the given HP drying application.
The water content in the vapor phase is slightly overpredicted at low pressures, and marginally underpre-
dicted at high pressures. Fig. 3.12 clearly shows this overprediction at 100 atm and 50°C. The figure shows a
very good agreement at elevated pressure, which is the relevant pressure range for this work. The molecu-
lar simulations are a significant improvement for the EoS. However, further improvements could be made to
reduce the over- and under predictions. This can be done by including the polarity of the water molecule,
which is not included in the selected forcefield. At 10°C and 1000 bar, the water content is only 29 ppm. It
would be interesting to simulate even lower temperatures and higher pressures. Possibly a saturated water
content of 5 ppm can be achieved, meaning no further drying steps are required. It should be noted that the
model does not include ice formation, so it is questionable how reliable the results are at temperatures close
to 0°C.
100 0.03
0.025
10-1
0.02
yH O /[mol/mol]
xH /[mol/mol]
10-2
0.015
2
10-3
2
0.01
10-4
0.005
0
0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
P/[bar] P/[bar]
(a) vapor phase (b) liquid phase
Figure 3.11: Vapor Liquid Equilibrium of the H2 -H2 O system, plotted as the water content in the vapor phase as a function of pressure
at different temperatures(a), and the solubility of hydrogen in the liquid phase as a function of pressure at different temperatures (b).
Applied force fields are TIP3P[54] for water and Marx[53] for water. Both the GE(triangles) and the NPT(diamond, circle, square) are
applied. Simulated temperatures are: T = 283K (cyan), T = 310K (blue), T = 323K (red), T = 366K (green), T = 423K (magenta).
2500
Experimental data
PR
2000 SRK
SMIRK
Sim GE
Sim NPT
yH O /(PPM)
1500
2
1000
500
0
100 200 300 400 500 600 700 800 900 1000
P/(atm)
Figure 3.12: Water content in the H2 -H2 O vapor mixture, as a function of pressure, at a temperature of 50°C. Calculated using PR-
EoS (circles), SRK-EoS (diamonds), SMIRK-EoS (x-mark), Molecular simulations (Gibbs ensemble (stars) and NPT ensemble (squares)),
compared to experimental data (triangles) [15].
3.5.2. G AS -G AS IMMISCIBILITY
At very high pressures, something called gas-gas immiscibility, can arise [55]. Usually, gases become more
immiscible at increasing pressure and decreasing temperatures. The simulations clearly show this effect,
as the saturated water content decreases drastically with increasing pressure and decreasing temperature.
However, in some supercritical mixtures, the vapor phase can split in two separate phases [55]. As pressure
is increased or temperature decreased, the vapor phase can split in two phases with different compositions.
For the H2 -H2 O system, this can result in a higher saturated water content [56, 57]. Fig. 3.13 shows this effect.
Following the red arrow in Fig. 3.13, the miscibility initially decreases, but after the minimum in temperature
it increases again. The blue arrow follows the same process for pressure.
3.5. M OLECULAR S IMULATION 23
Figure 3.13: P − T diagram of the H2 -H2 O mixture, including four possible critical lines (dashed lines) of the H2 -H2 O mixture. The trend
from (a) to (d) indicates the increasing immiscibility [55]. The red arrow indicates a temperature increase at constant pressure, the blue
arrow a pressure increase at constant temperature. Both processes go through a peak in immiscibility, as explained above.
There are two kinds of gas-gas immiscibility, first and second kind, shown in Fig. 3.14. For the first kind
(Fig. 3.14a), gas-gas immiscibility occurs at the lowest temperature, while the second kind (Fig. 3.14b) has a
temperature minimum in the critical line. For further details, the reader is referred to the work of Rowlinson
et al.[55] and Seward and Franck[56]. Gas-gas immiscibility is not described by the performed molecular
simulations. Therefore, although the molecular simulations are a very good prediction tool, experiments are
needed to validate the model.
(a) First kind (b) Second kind
Figure 3.14: P − y diagrams of typical mixtures where gas-gas immiscibility of the first kind (a) and second (b) kind occurs. The temper-
ature increases form A to D. The critical locus is shown dashed [55]. At certain pressures and temperatures, the mixture shows gas-gas
immiscibility, meaning that the vapor phase splits in two different vapor phases with different compositions.
4
I ONIC L IQUIDS
4.1. I NTRODUCTION
As mentioned in Section 2.1.4, Ionic Liquids (ILs) are promising alternative absorbents for hydrogen dehy-
dration, mainly because of there negligible vapor pressure. This chapter gives a brief description of ILs in
general, the different classes of ILs and their properties. Although there is no strict definition for ILs, the gen-
eral way to define them is [58]:
"An Ionic Liquid is a salt with a melting point below 100°C".
Ionic liquids consist of an organic cation and (often) an inorganic anion. They are also referred to as room
temperature ILs or molten salts. The number of cations and anions that can be combined is high, result-
ing in an endless amount of possible combinations. ILs are classified into different groups, depending on
the cation. Fig. 4.1 shows five well-known classes: Imidazolium, Pyridinium, Quaternary Ammonium, Tetra
Alkylphosphonium and Pyrrolidinium [59].
(a) Imidazolium (b) Quaternary Ammonium (c) Pyrrolidinium (d) Pyridinium (e) Tetra Alkylphosphonium
Figure 4.1: An overview of the five most well-known IL classes. ILs are categorized based on their cation.
Ionic liquids are used for various applications, such as catalysis, synthesis, electrochemical devices, lu-
bricants, and absorbents [60]. The possible use of ILs as an absorbent for some molecules, like CO2 capture,
has been studied very widely [61–64]. To the best of our knowledge, no previous work describes the use of
ILs to dehydrate hydrogen gas. However [Yu et al., Han et al., Krannich et al., Heym et al.] describe systems
were ILs are used as an absorbent to dry natural gas, which is a physically similar process. Therefore some
data is available describing the solubility of water in different ILs. Water and hydrogen do not have a similar
chemical structure. The main difference is that water is polar and hydrogen not. Therefore, absorbents that
are very well suited for water absorption, show a very low solubility of H2 at similar conditions [66, 67].
25
26 4. I ONIC L IQUIDS
4.2. I ONIC L IQUID S ELECTION

Due to the numerous possible combinations of cations and anions, it is challenging to select the IL that is
the best for this specific application. The selection can be narrowed down, as they should at least satisfy the
following constraints:
1. Hygroscopic
2. Negligible vapor pressure
3. Water/Hydrogen/thermally stable
Unfortunately, little experimental data are available on ILs and there is a countless number of them. The
alternative way to select ILs, in the absence of experimental data, is thermodynamic modeling. There are
a few possible methods available for predicting IL properties. Several studies have shown accurate density
and viscosity predictions for various ILs [68, 69], validated with experiments. However, these volumetric
properties are not decisive for selecting the best IL for this application. Referring back to the γ−φ formulation,
described in Section 3.3, efficient drying is achieved if the water content in the vapor phase y H2 O is low, while
the water content in the liquid phase x H2 O is high. The only parameter that varies with a different IL is the
activity coefficient of water γH2 O in that particular Ionic Liquid. A low activity coefficient means that the IL
can absorb a larger amount of water at a constant water content in the vapor phase. However, as the drying
process is a repeating cycle, a low γH2 O is only desirable if the IL is also capable of releasing the water again at
other conditions. This results in a more detailed description of the desired hygroscopic capability of the IL:
1. The activity coefficient should be low enough to dry the hydrogen gas to a water content of 5 ppm.
2. The difference in the activity coefficient, between absorption mode and regeneration mode, is prefer-
ably as large as possible.
The latter determines the amount of water that the IL actually can absorb; this does not take into account
the water in the IL that is not removed in the regeneration mode and will permanently be present. While the
activity coefficient is of large importance for the selection, previous studies show that it is hard to model prop-
erties without having any experimental data to fit on [17]. Possible methods to predict the activity coefficients
are COSMO-RS, (Modified) UNIFAC and Molecular Simulation.
4.2.1. (M ODIFIED ) UNIFAC

UNIFAC is the most commonly used group contribution method for the prediction of activity coefficients in
non-ideal mixtures [70]. Group contribution methods subdivide a molecule into functional groups. While
the thermodynamics of the system were originally based on interactions between the different molecules, it
is now reduced to interactions between these functional groups, which are assumed to be independent of
each other [71]. The thermodynamic properties of mixtures can be calculated by adding up all interactions,
depending on the temperature, pressure and composition. The full method is explained in detail by Fre-
denslund et al. [70]. Unfortunately, the original UNIFAC method has a few disadvantages. It is not possible
to accurately describe VLE data and excess enthalpies simultaneously. Therefore, the temperature depen-
dency of the activity coefficient can not be described correctly [72]. The newer modified UNIFAC (Dortmund)
method solves this problem by implementing the temperature dependent parameters. These parameters are
fitted simultaneously to a large (Dortmund) database, covering the temperature dependency of the activity
coefficients [72]. However, the modified UNIFAC does not eliminate the largest disadvantage of these group
contribution methods. The described methods always need at least a few experimental data to fit the re-
quired group interaction parameters [71]. Therefore, the selection of ILs, using this method, is limited to ILs
that have experimental data available. However, due to the countless possibilities of ILs, a method is required
that does not need this kind of data. COSMO-RS gives a good solution, as it does not rely on fitted parameters.
4.2.2. COSMO-RS
Conductor like Screening Model for Real Solvents (COSMO-RS) is a method, based on quantum mechanics,
that can predict chemical potentials in liquids [73]. Properties, such as the activity coefficient, can be derived
from the chemical potential. Analogously to the group contribution methods, properties are predicted by
modeling the interactions between molecules. However, instead of dividing the molecules into functional
groups, the molecule structure is simplified to its outer surface [71]. The molecular surface is subdivided into
4.2. I ONIC L IQUID S ELECTION 27
surface segments, with equal surface areas, that interact with each other. The surface is simulated by a sigma
(σ) profile. A sigma profile is the probability distribution of such a segment having a specific charge density
[74]. The main advantage of COSMO-RS is that, if one has a correct sigma profile of the pure component, no
fitted interaction parameters are necessary. This method is therefore very well suited for the selection of a
wide variety of different potential ILs, which have no experimental data available. Although COSMO-RS does
not require fitted parameters that describe interactions between different molecules, a few element-based
parameters are still required [71]. Therefore, experimental data are still needed to get the surface charge
density profile of a pure element right. Furthermore, COSMO-RS captures only approximately ion-dipole
interactions, which are existing in this system due to the water-ion interactions. During the course of this
project, we approached a company that provides a tool that includes COSMO-RS calculations. Initial predic-
tions using COSMO-RS were executed and compared to the available experimental data [75]. Fig. 4.2 shows
these predictions, using three different sets of parameters, and it also includes the experimental data [75].
Based on Fig. 4.2, the following conclusions were drawn:
1. The predicted activity coefficients vary strongly with the used element specific parameters. Without
comparing it to experimental data, this already indicates that the used parameters are of much impor-
tance.
2. The COSMO-RS predictions show a large deviation with respect to the experimental data. This devi-
ation increases at low water fraction. While COSMO-RS clearly shows a minimum at infinite dilution,
the experimental data show that this minimum is somewhere between 0.7-0.8 mole fraction.
3. The temperature dependency of the activity coefficient, predicted by COSMO-RS, is opposite to the
experimental data. The predictions suggest that the activity coefficient increases with temperature,
while the experiments show that it is the other way around.
Unfortunately, the COSMO-RS predictions show that COSMO-RS also needs some fitting to experimental
data, to get reliable results. Previous studies mainly use COSMO-RS on a qualitative basis instead of on a
quantitative basis [17, 18, 76–78]. Although COSMO-RS could be used to get a relative understanding of how
ILs behave compared to other ILs, it can not be used to validate activity coefficients that follow from experi-
ments. For this reason, we chose not to go ahead with COSMO-RS.
4.2.3. M OLECULAR S IMULATION

The third method to predict IL properties is using molecular simulations, like we used for the VLE of H2 and
H2 O. The working principle of molecular simulation methods is described in Chapter 3. The major drawback
of this method is that it requires a lot of time and manual work for every molecule that needs to be simulated.
Therefore it is not a very efficient method for screening a large amount of ILs. When an IL is selected though,
molecular simulations can be very useful to precisely predict properties like activity coefficients. Molecular
simulations for IL systems have not been used in this work, but they could be a useful tool for future work.
4.2.4. L ITERATURE AND AVAILIBILITY

The objective of this study is to test the feasibility of hydrogen gas drying, using Ionic Liquids. In this phase
of the project, it is not required, nor expected to find the optimal IL. Based on the challenges associated with
property prediction, a more pragmatic approach is followed for the IL selection. We first created a long-list of
ILs, based on the following:
1. Section 4.1 already mentioned the previous work on testing ILs for natural gas dehydration [17–19, 65].
The ILs used in these works are also interesting for this work, as the process is physically quite similar.
2. As mentioned in Section 4.2.1, Modified UNIFAC is the most commonly used method to predict activity
coefficients. It only works though, if interaction parameters are available. The Dortmund Data Bank
(DDBST [79]), that provides the worlds largest factual data bank, was used to check which interaction
parameters between water and different ILs were included in the data bank. All the Ionic Liquids that
belonged to one of the five classes shown in Fig. 4.1 were checked on available parameters in DDBST
[79].
3. Finally, the ILs mentioned above were checked for availability, price, and safety. At this stage of the
research, little is known about which IL will show the best results, if the concept works at all. There-
fore, we decided not to synthesize the ILs ourselves. The most common suppliers of Ionic Liquids are
28 4. I ONIC L IQUIDS
(a) Parameters: ADFcombi2005 (b) Parameters: ADFcombi2005(no H-bond restrictions)
0 0
-2 -2
-4 -4
293K 293K
Ln( H O )
Ln( H O )
-6 -6 313K
313K
2
2
333K 333K
-8 353K -8 353K
373K 373K
-10 -10
-12 -12
-14 -14
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
mol fraction H2O (mol/mol) mol fraction H2O (mol/mol)

(c) Parameters: Slightly fitted IL parameters (d) Experimental data
0 0
-2 -2
-4 -4
293K 293K
Ln( H O )
Ln( H O )
-6 313K -6 313K
2
333K 333K
-8 353K -8 353K
373K 373K
-10 -10
-12 -12
-14 -14
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
mol fraction H2O (mol/mol) mol fraction H2O (mol/mol)
Figure 4.2: Activity coefficient predictions by COSMO-RS (a-c), using three different parameters (ADFcombi2005, ADFcombi2005 (no
H-bond restrictions), and slightly fitted parameters). The activity coefficients are plotted as a function of the H2 O mole fraction at five
different temperatures (293K, 313K, 333K, 353K, 373K). Experimental data at the same conditions are shown for comparison (d). [75]
Sigma Aldrich® and Iolitec® . The Safety Data Sheets (SDS) were checked to see if the ILs were safe to
do experiments on. Due to the limited experience in experimental work with ILs, it was decided to
only select ILs that have a GHS07 classification, or less harmful. This classification indicates a warning
instead of a real danger [80]. Finally, the price was checked and considered in the decision making. It
should be noted that, in a further phase of this research, it is possible to select ILs that do not fulfill
these requirements, while they still will deliver significantly better results.
P REVIOUS GAS DEHYDRATION STUDIES

Table 4.1 shows ILs that were used in previous studies describing gas dehydration [17, 65, 81–83]. It is clear
that, except for the last one, all cations are imidazolium based.
D ORTMUND D ATA B ANK (DDBST )

The availability of ILs with Modified UNIFAC parameters in DDBST was checked [79]. As expected, imida-
zolium based Ionic Liquids have the most parameters available. Table 4.2 gives an overview of the most occur-
ring anions and cations. The values represent the activity coefficients at infinite dilution and 25°C, in increas-
ingly order from left to right, bottom to top. In line with previous work [17, 18], this table clearly shows that the
activity coefficient mainly depends on the anion instead of the cation. The ILs 1-Ethyl-3-methylimidazolium
4.2. I ONIC L IQUID S ELECTION 29
Table 4.1: An overview of ILs that were used for gas dehydration in previous work.
Abbreviation Full Name Ref

[BEIM][EtSO4 ] 1-Butyl-3-ethylimidazolium ethyl sulfate [65]
[BMIM][Ac] 1-Butyl-3-methylimidazolium acetate [81]
[EMIM][BF4 ] 1-Ethyl-3-methylimidazolium tetrafluoroborate [82]
[EMIM][Cl] 1-Ethyl-3-methylimidazolium chloride [83]
[EMIM][dca] 1-Ethyl-3-methylimidazolium dicyanamide [82]
[EMIM][EtSO4 ] 1-Ethyl-3-methylimidazolium ethyl sulfate [65]
[EMIM][MeSO3 ] 1-Ethyl-3-methylimidazolium methanesulfonate [83]
[EMIM][OTF] 1-Ethyl-3-methylimidazolium trifluoromethane sulfonate [83]
[EMIM][TF2 N] 1-Ethyl-3-methylimidazolium bis(trifluoromethanesulfonyl)amide [17, 82]
[EMIM][TFA] 1-Ethyl-3-methylimidazolium trifluoroacetate [83]
[N(4)111][Tf2 N] Trimethyl(butyl)ammonium bis(trifluoromethanesulfonyl)amide [82]
ethyl sulfate ([EMIM][EtSO4 ]), 1-Ethyl-3-methylimidazolium tetrafluoroborate ([EMIM][BF4 ]), and 1-Ethyl-
3-methylimidazolium bis(trifluoromethanesulfonyl)amide ([EMIM][TF2 N]) have been studied for gas dehy-
dration before in previous work (Table 4.1) and have modified UNIFAC parameters. Therefore, these ILs are
of particular interest.
Table 4.2: Activity coefficients of water in the given ILs at infinite dilution and 25°C, calculated with modified UNIFAC if parameters were
available. For the calculations, the ’Activity Coefficient Calculation’ tool from DDBST was used [79]. The ILs are ordered on the activity
coefficient, increasing from left to right and from bottom to top.
Anion(down) and Cation(right) [EMIM] [PMIM] [BMIM] [MMIM] [HMIM] [OMIM] [HOEMIM] Average
bis(trifluoromethylsulfonyl)imide 3.120 3.268 3.426 2.987 3.761 4.105 2.177 3.263
trifluoromethylsulfonate 0.857 N/A N/A N/A 1.117 1.266 N/A 1.080
tetrafluoroborate 1.016 0.928 0.885 1.193 0.867 0.895 1.063 0.978
trifluoromethanesulfonate N/A N/A 0.977 0.809 N/A N/A N/A 0.893
octyl sulfate 0.304 N/A 0.341 0.287 N/A N/A N/A 0.311
hexylsulfate 0.271 N/A N/A N/A N/A N/A N/A 0.271
ethylsulfate 0.223 0.231 0.242 N/A 0.271 0.304 N/A 0.254
butylsulfate 0.242 0.255 N/A N/A N/A N/A N/A 0.249
propylsulfate 0.231 0.242 0.255 N/A N/A N/A N/A 0.243
methylsulfate 0.220 0.223 0.231 0.226 0.255 0.287 N/A 0.241
hydrogensulfate 0.143 0.152 0.164 0.138 0.192 0.224 N/A 0.169
Average 0.663 0.757 0.815 0.940 1.077 1.180 1.620
AVAILIBILITY AND S AFETY

The ILs used for experiments, are initially selected on the outcomes of previous work, existing modified UNI-
FAC parameters, and stability. Based on these final criteria, the tetrafluoroborate ([BF4 ]) cation was no longer
considered, as Freire et al. showed that BF4 -based ILs are not water stable [84]. As safety is of the most impor-
tance, the Safety Data Sheets (SDS) of the ILs were checked on the safety criteria mentioned in Section 4.2.4.
Finally, the price of the ILs was checked. Both safety and price strongly depend on the way the ILs are synthe-
sized, so if certain ILs look promising, it is interesting to examine what method is the best to synthesize them.
Based on all the above mentioned criteria, the ILs, displayed in Table 4.3, were ordered for the experiments.
Table 4.3: The ordered ILs that are used for the experiments. Properties from left to right are Molar weight, melting temperature, density,
viscosity, and surface tension all at specified temperature. Properties are from literature, or found on the web via:
a) Iolitec® , b) Sigma-Aldrich® , c) Chemicalbook®
Full name M /(g/mol) Tmelt /(°C) ρ/(g/cm3 ) η/(cP) γ/(mN/m)

1-Ethyl-3-methylimidazolium acetate 170.21a -45a 1.10(25°C)a 162(25°C)[85] 42.9(25°C)[86]
1-Butyl-3-methylimidazolium chloride 174.67a 65a 1.08(25°C)[87] 142(80°C)[85] 48.2(25°C)[88]
1-Ethyl-3-methylimidazolium ethyl sulfate 236.29a <25a 1.24(25°C)a 94(25°C)a 46.9(25°C)[86]
1-Butyl-3-methylimidazolium octyl sulfate 348.50b 37c 1.07(25°C)c N/A 35.7(30°C)[89]
5
E XPERIMENTS
This chapter describes the experiments that were executed to test the feasibility of high pressure drying of
hydrogen gas using Ionic Liquids. The main target of the experiments was to examine if ILs did actually
function as an absorbent, when contacted with wet hydrogen gas. Therefore, first experiments were executed
that were focused on showing the proof of concept. If the selected ILs did indeed absorb water from the
hydrated gas, as expected, the research was extended with an additional experiment, to test if it was also
possible to regenerate the IL.
• The main experiment was meant to bubble wet hydrogen gas through an IL and to measure if the re-
sulting hydrogen gas contained less water. The mass transfer from the gas to the liquid was qualitatively
determined by measuring if Vapor Liquid Equilibrium was achieved. During these experiments, close
attention was paid to properties like viscosity and foam production. This experiment is referred to as
the absorption experiment.
• The other experiment was to test if the IL was also capable of releasing the water. To test this, the IL was
transfered to a round bottom flask, which was heated and put under vacuum, to shift the VLE of water
from the liquid phase to the vapor phase. This experiment is referred to as the desorption experiment.
5.1. A SSUMPTIONS & S IMPLIFICATIONS

The following assumptions and simplifications were made for the experiments in relation to the actual final
application, which is HP drying of hydrogen:
1. For safety and simplicity reasons, the absorption experiment is executed at atmospheric pressure, while
the final application operates in a pressure range of 800 - 1000 bar. Fortunately, the activity coefficient
mostly depends on the temperature and the composition, instead of the pressure. Although these ex-
periments can not be translated one to one to high pressures, these experiments can provide a good
impression on how the system would behave at elevated pressures.
2. It is assumed that hydrogen does not dissolve in the Ionic Liquid. This assumption is reasonable, as
hygroscopic ILs are selected that are suited for water absorption. Since water is a polar molecule and
hydrogen is not, one can assume that the hydrogen dissolved in the IL is negligible [66, 67].
3. Previous studies have shown that the vapor pressure of ILs are so low that they can be neglected [19, 65,
82]. The evaporation of the IL, for the most volatile IL from Table 2.1, was calculated using Raoult’s law
[46]. This resulted in an IL content in the vapor phase of 2.9 × 10−7 ppm, which is more than a billion
times less than water. Therefore, vaporization of ILs is not taken into account in this work.
For another project, a unit was already in place to calibrate and test sensors that can measure contaminants
in hydrogen gas up to ppb (parts per billion) level. This unit consists of a number of cylinders that can store
hydrogen, contaminated with H2 O, H2 S, CO, and H2 S. This unit has been slightly modified to use it for the
absorption experiment.
31
32 5. E XPERIMENTS
5.2. M ATERIALS & A PPARATUS

The tested ILs were 1-Ethyl-3-methylimidazolium acetate [EMIM][Acetate], 1-Butyl-3-methylimidazolium
chloride [BMIM][Chloride], 1-Ethyl-3-methylimidazolium ethyl sulfate [EMIM][Ethyl Sulfate], and 1-Butyl-
3-methylimidazolium octyl sulfate [BMIM][Octyl Sulfate]. The rational for the selection can be found in
Section 4.2. The suppliers and purities are specified in Appendix A, along with the specifications of other
materials used.
Fig. 5.1 gives an overview of the experimental apparatus. The modified part of the unit is indicated with the
dashed rectangle. The left part of the flow diagram is used to regulate the water content in the hydrogen gas.
The inlet hydrogen (1) was first purified (2) to remove contaminants using a palladium catalyst. The purified
hydrogen was then used to create wet hydrogen (3 and 4) or directly used as a product. The exact procedure
for creating wet hydrogen can be found in Appendix A. Using the computer controlled valves (5a and 5b), the
wet hydrogen was accurately mixed with dry hydrogen to regulate the desired water content. The hydrogen
could then either be directed through the scrubber (7), whose temperature was controlled using a water bath,
or bypassed (9). Samples of the resulting hydrogen were then measured by the sensor (10) and finally the hy-
drogen was directed to the incinerator, where it was burned. Appendix A gives a detailed overview of the used
materials and equipment, including a technical drawing of the designed scrubber.
Figure 5.1: Schematic diagram of experimental apparatus for hydrogen dehydration, using ILs. The unit consists of the following com-
ponents:
1, Hydrogen gas storage; 2, purifier; 3, water scrubber; 4, buffer vessel; 5(a and b), computer valves; 6, mixing vessel; 7, IL scrubber; 8,
thermocouple; 9(a and b), three-way hand valves; 10, Sensor; 11, Incinerator. The text above the figure explains how the apparatus works
exactly.
5.3. T EST P ROCEDURE

The absorption and desorption experiments were executed as followed:
5.3.1. A BSORPTION EXPERIMENTS

In this work, the ’dry’ IL is defined as the state of the IL, right before it was used in the absorption experiment.
After this experiment, the IL is referred to as the ’wet’ IL. Note that this classification is not directly related to
the absolute water content in the IL. In a typical experiment, a certain amount (about 40 gram) of the ’dry’
IL was put into the scrubber (7). Then scrubber 7 was installed in the unit (Fig. 5.1), and the temperature
of the scrubber was raised to the desired temperature. When the desired IL temperature, measured by the
thermocouple 8, was reached, it was tested if the liquid phase was in equilibrium with the vapor phase. This
started by bubbling dry hydrogen gas through the IL in the scrubber. The output water content was measured
by sensor 10. After this measurement, the hydrogen flow was bypassed (via valves 9a and 9b) and the water
content was increased to 1755 ppm. If the water content was constant (this always took about 5 minutes,
due to the memory effect of pipes and valves), the hydrogen gas was again directed over the scrubber. The
5.3. T EST P ROCEDURE 33
VLE was checked by comparing the output water content with the initial output, where dry hydrogen gas was
bubbled through the IL. The VLE was achieved, once these measurements were equal. This process is visual-
ized in Section 6.2. If VLE was reached, the measurements could continue. Because the IL is absorbing water
from the wet hydrogen gas, the water fraction in the IL gradually increases. The absorption experiment was
continued until the water content in the hydrogen gas reached 500 ppm, due to the limitations of the sensor,
specified in Appendix A. When this water content was reached, it was tested again if a VLE has been estab-
lished. This procedure was the same as at the start of the experiment, only now starting with wet hydrogen
gas, instead of dry hydrogen. Finally the IL was drained from the scrubber into a round bottom flask, to start
the desorption experiment. At the same time, a sample of the ’wet’ IL was taken, for the KF titration.
5.3.2. D ESORPTION EXPERIMENT

For the desorption experiment, the round bottom flask containing the ’wet’ IL, was placed in a heating mantle
and was via a vacuum adapter connected to a vacuum pump at 15 mbar. The flask was heated up to 150°C for
[BMIM][Chloride] and[BMIM][Octyl sulfate]. The temperature of [EMIM][Acetate] was limited to a maximum
of 100°C, as the work of Cao and Mu showed that acetate based ILs are less stable and can start to decompose
at temperatures lower than 150°C [90].
5.3.3. K ARL F ISCHER TITRATION

The water content in the dry and wet IL was measured using Karl Fischer titration. Because of the high vis-
cosity of the liquids, samples were diluted with methanol. The specifications of the KF coulomat and the
methanol are described in Appendix A. Since the ILs are very hygroscopic, contact with the surrounding air
should be avoided. Therefore, transfer of IL was always performed in a glove box, or through syringe needles
that were flushed with nitrogen prior to use.
6
R ESULTS & D ISCUSSION
The experimental results are divided into two sections. Section 6.1 outlines the observations of the different
Ionic Liquids for the absorption and desorption experiments. These observations include bubble formations,
foaming, color change, and crystallization. Section 6.3 reports the more quantitative experimental results.
The data from the sensor, that measures the water content in the dehydrated hydrogen gas is used to calculate
the VLE of the H2 O - IL mixture.
Table 6.1: The ordered ILs that are used for the experiments. Properties from left to right are Molar weight, melting temperature, density,
viscosity, and surface tension all at specified temperature. Properties are from literature, or found on the web via:
a) Iolitec® , b) Sigma-Aldrich® , c) Chemicalbook®
Full name M /(g/mol) Tmelt /(°C) ρ/(g/cm3 ) η/(cP) γ/(mN/m)

1-Ethyl-3-methylimidazolium acetate 170.21a -45a 1.10(25°C)a 162(25°C)[85] 42.9(25°C)[86]
1-Butyl-3-methylimidazolium chloride 174.67a 65a 1.08(25°C)[87] 142(80°C)[85] 48.2(25°C)[88]
1-Ethyl-3-methylimidazolium ethyl sulfate 236.29a <25a 1.24(25°C)a 94(25°C)a 46.9(25°C)[86]
1-Butyl-3-methylimidazolium octyl sulfate 348.50b 37c 1.07(25°C)c N/A 35.7(30°C)[89]
6.1. O BSERVATIONS
The experiments are outlined in Chapter 5 and in Appendix A.
[EMIM][E T SO4 ]
Unfortunately, during the first experiment with [EMIM][EtSO4 ], we found that the Safety Data Sheet (SDS)
was modified by Sigma-Aldrich® compared to the date at which we ordered this IL. The SDS was updated with
safety classifications GHS05, GHS06, and GHS08, which we concluded to form a too high risk, as discussed in
Section 4.2.4. Therefore, we decided to stop the experiments with this particular IL.
6.1.1. B UBBLE FORMATION AND F OAMING

Fig. 6.1 shows the three ILs during the dehydration experiment. The [BMIM][Chloride] clearly shows a differ-
ent bubble formation than the other two ILs. The hydrogen gas that enters the scrubber forms very large bub-
bles that slowly rise as the bubbles grow. There is no layer of foam on top of the IL. For the ILs [EMIM][Acetate]
and [BMIM][Octyl Sulfate] on the other hand, a clear separation between the fluid and foam is observed.
Foam producing can cause two significant problems:
1. If the foam layer grows to a certain height, the IL-foam leaves the scrubber and enters the rest of the
system. This causes pollution of the sensor and a loss of IL. Since the VLE calculations assume a con-
stant amount of IL, as explained in Chapter 5, a loss of IL means that the results are not representative.
To prevent this overflow, the volume of the IL used for the experiments, which was originally set at 40
ml, needed to be reduced to 25 ml for [BMIM][Octyl Sulfate]. The amount of IL could not be reduced
further, otherwise the hydrogen gas inlet was not submerged in the IL.
2. The second problem that occurs due to foaming is low diffusion of water in the IL. Hydrogen gas does
not enter the scrubber from the bottom, so there is a layer of IL where hydrogen bubbles are not forced
35
36 6. R ESULTS & D ISCUSSION
(a) [BMIM][Chloride] (b) [EMIM][Acetate] (c) [BMIM][Octyl Sulfate]
Figure 6.1: Snapshots of the experiments with three different ILs at 70°C and atmospheric pressure. The input water content in the
hydrogen gas is 1755ppm and the flow rate is 40 Nl/h
through. The risk is that the foam reaches an equilibrium very quickly, while the liquid at the bottom
still has enough capacity to absorb water. Section 5.3.1 describes how it was tested if this was indeed
occuring.
It is well known that surface tension, which acts to reduce the surface of a liquid to a minimum, has a
large impact on foam production [91]. To prevent problems related to foaming, ILs with a relatively high
surface tension must be selected. As mentioned in Chapter 4, it is hard to predict IL properties, such as
surface tension. However, many anions of ILs consist of an alkyl group that lowers the surface tension of a
molecule [92]. The longer the alkyl chain, the lower the surface tension [92]. It is therefore not surprising that
[BMIM][Octyl Sulfate] produces a lot of foam. The IL properties in Table 6.1 confirm the relation between
alkyl chain length and surface tension. Note that the properties in Table 6.1 are mostly at at a temperature of
25°C, while the experiments shown in Fig. 6.1 were executed at a temperature of 70 °C. A higher temperature
results in a lower viscosity, which also increases foam formation [93]. Furthermore, as temperature rises,
cohesive forces between molecules decrease, resulting in a lower surface tension [91]. Fig. 6.2 clearly shows
the temperature effect on foam production. To prevent foaming, it is therefore recommended to absorb at
low temperatures.
6.1.2. C OLOR CHANGE

During the regeneration process, described in Chapter 5, the ILs [EMIM][Acetate] and [BMIM][Octyl Sulfate]
changed color. This process is pictured in Fig. 6.3. When heated, the colors transformed from light yellow to
very dark brown. Previous work also observed this color change [94]. The color change is possibly due to the
impurities in the IL [94].
6.1.3. C RYSTALLIZATION RANGE

A final observation that was made was the large crystallization range of the ILs that have a melting point
above room temperature([BMIM][Octyl Sulfate] and [BMIM][Chloride]). Both ILs were sometimes a liquid at
room temperature, and sometimes a solid.
6.1. O BSERVATIONS 37
(a) [EMIM][Acetate] at 70°C (b) [EMIM][Acetate] at 30°C
Figure 6.2: Foam production in [EMIM][Acetate] during dehydration experiment. Both experiments are at atmospheric pressure, con-
stant water content and constant flow rate of 40 Nl/h. The only difference is the temperature (70°C in a and 30°C in b)
(b) [BMIM][Octyl Sulfate]

(a) [EMIM][Acetate]
Figure 6.3: Observed color change during the desorption experiment, for the ILs [EMIM][Acetate] (a) and [BMIM][Octyl Sulfate] (b).
6.2. VAPOR L IQUID E QUILIBRIUM

To compare the ILs on a more quantitative basis, the water content in the output stream is measured and
subtracted from the input water content. Appendix A describes how the input water content is regulated. A
Vapor Liquid Equilibrium between the liquid phase and the vapor phase is essential to draw conclusions on
absorption capacity and the activity coefficients of the ILs. Therefore, during every experiment it is tested
if, at the specified conditions, equilibrium is achieved. This procedure is described in Section 5.3.1. Fig. 6.4
presents the results of such an experiment for all three ILs. A VLE is achieved if the water content in the re-
sulting hydrogen gas remains constant when the input water content is instantly changed from wet hydrogen
to dry hydrogen. Fig. 6.4 shows that [BMIM][Chloride] and [EMIM][Acetate] are in equilibrium with the va-
por phase, at the specified conditions. [BMIM][Octyl Sulfate], however, does not achieve VLE. This has two
possible reasons. Firstly, the absorption rate of water from the vapor phase to the liquid phase could be too
low. Secondly, the diffusion in the IL could be the problem. The latter implies that a certain amount of the
IL is in equilibrium, but the water does not diffuse through the rest of the IL quickly enough. The significant
foam production, which is visualized in Fig. 6.1c, is the most plausible reason for preventing a fast VLE. The
foam is in equilibrium with the hydrogen gas, but the water absorbed by the foam is not (quickly enough)
transported to the liquid layer below the foam.
Figure 6.4: Experimental results of the VLE tests. The left figure shows the start of the experiment, where the IL is still dry. The right figure
shows the measured water content at the end of the experiment. VLE is achieved, if the output water content remains constant when
the input stream is instantly changed from dry to wet hydrogen, or the other way around. The small difference can be explained by a
memory effect in the pipes.
6.3. M EASUREMENTS
The VLE experiments showed that the ILs [BMIM][Chloride] and [EMIM][Acetate] achieved VLE during the
experiment and are in equilibrium with the hydrogen gas. However, as [EMIM][Acetate] absorbed more
water, the foam production increased. Additional experiments at lower temperatures are required to test
[EMIM][Acetate]. Therefore, the main focus of this chapter is on the results obtained with [BMIM][Chloride].
Therefore, the γ − φ approach, described in Section 3.3, can be applied to calculate the properties of the ILs.
Since the experiments are executed at atmospheric conditions, the Poynting Effect is negligible. Furthermore,
the ideal gas law is applied, resulting in fugacity coefficients equal to unity. Therefore the γ − φ equation re-
duces to Eq. (6.1), which is also referred to as modified Raoult’s law [41].
sat
y H2 O P = γH2 O x H2 O P H 2O
(6.1)
6.3. M EASUREMENTS 39
The output water content in the vapor phase y H2 O is measured by the sensor. Fig. 6.5a shows the water
content in the hydrogen gas, after it was bubbled through the IL [BMIM][Chloride]. The input water content
is constant over time and the output water content is increasing, as the IL absorbs water. The difference
between the water content in the input and the output is the amount of water that is absorbed by the IL.
Therefore, we know the increase of the water fraction in the IL over time. The results in Fig. 6.5a show that
the experiments are in good agreement with one another. Especially experiments 2 and 3, which both started
with a regenerated IL, instead of a new one. It is remarkable that the regenerated ILs start with a lower water
content than the new one. The next step is to calculate the VLE from these results. The only parameter that is
missing to calculate the activity coefficient is the initial water fraction in the IL. The water content in the IL is
measured with the Karl Fischer (KF) titration method. At the end of the experiment, the water content in the
IL is measured again. This is to check if the water content is now equal to the initial water content plus the
absorbed water. Unfortunately, Fig. 6.5b shows that this is not the case. The KF-results should correspond to
the end values of the plotted curves. It is clear that the KF measurements are far off, which may be explained
as follows:
1. The water absorbed by the IL is not measured accurately. This is unlikely, because the input water
content is constant and the output water content is measured by a sensor that is well calibrated and
capable of measuring contaminants like water at ppb level.
2. Another explanation could be the accuracy of the KF titration. At every KF measure, three samples were
tested. The measurements had a deviation of around 5%. If the error exceeded this error percentage,
additional samples were taken. The dashed lines in Fig. 6.5b represent this 5% error. It is obvious that
the difference between the KF samples and the sensor measurements is much larger.
3. A more likely explanation is that the samples absorbed water from surrounding air before they were
injected in the KF device. Surrounding air easily contains 10-100 times more water than the hydro-
gen gas in the experiment. Therefore, the ILs strongly attract water molecules, before and after the
experiment. Although air contact was avoided at all times using glove-boxes and needles, it is hard to
completely eliminate any form of air contact. Furthermore, due to the occasionally unavailability of the
KF-equipment, it was not always possible to directly take the KF samples. Therefore, the time between
the experiments and the KF titration differed quite a lot. Although the samples were sealed, some water
penetration in the sample is still possible, especially with this long waiting times.
600 12000
500 10000
400 8000
xH O /[PPMw]
yH O /[PPMv]
300 6000
2
200 4000
100 BMIM Cl (exp 1) 2000 BMIM Cl (exp 1)

BMIM Cl (exp 2) BMIM Cl (exp 2)
0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
time / [hour] time / [hour]
(a) water content in the vapor phase, as a function of time. (b) mass fraction of water in the IL, as a function of time.
Figure 6.5: Water content in the hydrogen gas, which is bubbled over the IL [BMIM][Chloride] (a). As the IL absorbs water from the
hydrogen gas, the mass fraction of water in the IL increases (b). All experiments are at atmospheric pressure and a constant flow rate of
40 Nl/h and a water content input of 1755ppm.
Due to the fact that the absolute water content in the IL is unknown, it is not possible to calculate the
activity coefficient very accurately. The lines plotted in Fig. 6.6, showing the VLE of H2 O and the calculated
activity coefficients, should coincide. It is clear from Fig. 6.6a, that the different experiments show similar
slopes. If the individual slopes would only be translated in the x-direction, they would coincide very well.
This strengthens the theory that the bad results are caused by the absence of the initial water content. Based
on the results shown in Fig. 6.6b, it is not possible to determine the activity coefficient as a function of the
water fraction in the IL. However, it is fair to say that in this low water fraction region, the activity coefficient
is low. Even if it would be off by a factor two, it would still be well below 0.1, which is still very low.
0.04
600
0.035
500
0.03
400
yH O /[PPMv]
/ [-]
0.025
H O
300
2
0.02
2
200
0.015

100 0.01
0 0.005
0 1000 2000 3000 4000 5000 6000 0 1000 2000 3000 4000 5000 6000
xH O /[PPMw] xH O /[PPMw]
2 2
(a) Calculated Vapor Liquid Equilibrium of the IL-H2 O system. (b) Activity Coefficients, calculated with the Gamma/Phi approach, as a func-
tion of the mass fraction of the water in the IL.
Figure 6.6: The measured water content in the vapor phase is plotted against the calculated mass fraction of water in the IL (a). This is
calculated by adding the absorbed water to the initial mass fraction of water. The activity coefficient easily follows from modified Raoult’s
law and is plotted against the mass fraction of water in the IL (b).
6.3.1. EMIM A CETATE

Fig. 6.7a shows the same experiment for [EMIM][Acetate]. The first experiment looks very promising, as it
takes two times longer to reach the 500 ppm water content in the vapor phase, compared to the experiment
with [BMIM][Chloride]. However, it is much harder to regenerate the IL. The green line in Fig. 6.7a shows the
second experiment, so after the IL has been regenerated. It is clear that the IL can not be dried to a water
content near the new IL. One reason for this is that the [EMIM][Acetate] is regenerated at 100°C instead of
150°C, due to the fact that it is less stable, as explained in Section 5.3.2. The other reason is that Acetate is very
hydrophilic, even more hydrophilic than the Chloride anion [95]. Although this has the advantage that it can
absorb a lot of water, a clear downside is that it costs a lot of energy to get the water out again. A third exper-
iment with this IL resulted in a water content at the start of the experiment of 500 ppm. This is the same as
the end value of the previous experiment, meaning that no water was removed during regeneration. Another
downside of hydrophilic ILs is the effect of the water content on the resulting properties [96]. Increasing the
alkyl chain length of the anion creates a more stable and hydrophobic IL, which makes it easier to regenerate
the IL [96]. However, as mentioned in Section 6.1.1, a longer alkyl chain reduces the surface tension which
results in foam production.
L OWER T EMPERATURES
Fig. 6.7b shows the absorption experiment fro [EMIM][Acetate], executed at 30°C, instead of 70°C. It was not
possible to reduce the temperature for [BMIM][Chloride], due to the melting temperature of 65°C. It is clear
that a lower temperature results in an enormous increase of the water absorption capacity of the IL. This
is due to the fact that the vapor pressure of water drastically decreases with temperature. The experiment at
30°C took almost 30 hours. For safety reasons, the experiments did not continue overnight. Therefore, a single
experiment took almost a week, resulting in only a single run. More experiments at multiple temperatures
are required to test the temperature dependency on the IL properties, like the activity coefficient.
6.3. M EASUREMENTS 41
600 600
500 500
400 400
yH O /[PPMv]
yH O /[PPMv]
300 300
2
2
200 200
100 100
EMIM Ac (exp 1)
EMIM Ac (exp 2) EMIM Ac (exp 1)
0 0
0 1 2 3 4 5 6 7 0 5 10 15 20 25 30
time / [hour] time / [hour]

(a) 70°C (b) 30°C
Figure 6.7: Water content in the hydrogen gas, which is bubbled over the IL [EMIM][Acetate], at a temperature of 70°C (a), and 30°C (b).
All experiments are at atmospheric pressure and a constant flow rate of 40 Nl/h and a water content input of 1755ppm.
7
C ONCLUSIONS
The goal of this project was twofold. The first objective was to obtain a better understanding of the H2 -H2 O
system, by calculating the saturated water content in compressed hydrogen. The second objective was to test
the feasibility of using Ionic Liquids as a drying agent for HP hydrogen dehydration.
7.1. T HERMODYNAMICS
1. Chapter 3 showed that conventional EoS, with classical mixing rules, are not capable of describing the
H2 -H2 O system at elevated pressures. The EoS that were tested included Peng-Robinson (PR), Soave’s
Redlich Kwong (SRK), and Shell’s in-house developed (SMIRK) EoS.
2. Molecular simulation results showed excellent agreement with experimental data, especially for the
vapor phase. The TIP3P [54] forcefield gave the best results for water, and the Marx [53] forcefield for
hydrogen. As expected, the simulation with the highest pressure and lowest temperature conditions,
resulted in the lowest water content. Namely 29 µmol per mol H2 -H2 O mixture, at 10 °C and 1000 bar.
7.2. I ONIC L IQUIDS

The following conclusions can be drawn from the experimental results.
1. First of all, Ionic Liquids can be used as an absorbent to dehydrate hydrogen gas. As expected, the
experiments showed that bubbling wet hydrogen gas through an IL resulted in a dehydrated hydrogen
output.
2. Foam production is more relevant than initially expected. Foam production drastically decreases the
mass transfer of water through the IL. Due to the lack of diffusion, equilibrium is not achieved, reducing
the capacity of the IL for water absorption drastically. Since surface tension is directly related to foam
production, this should be an important property to check in the selection process.
3. ILs that are capable of absorbing large amounts of water can dry the hydrogen gas to a very low water
content. However, this does not necessarily mean that they are suited for this application. Due to their
high attraction to water, it costs a lot of energy to regenerate very hydrophilic anions, like acetate. Non-
polar molecules with long alkyl chains are more hydrophobic, but they have the drawback that they
produce more foam.
4. The temperature of the absorption process is preferably as low as possible, even though lower tem-
peratures result in higher viscosities. Firstly, surface tension increases with decreasing temperatures,
resulting in less foam production (see Section 6.1.1). Secondly and most importantly, the saturated va-
por pressure of water decreases drastically with temperature. Referring to the γ−φ equation (Eq. (7.1)),
a low K is desired, as it results in a higher water content in the liquid phase. Even if low temperatures
result in higher activity coefficients, which is observed by previous studies [75], the saturated vapor
pressure of water P sat
H2 O decreases drastically with temperature. For example, the activity coefficient of
water in [EMIM][Acetate] γH2 O decreases by a factor 4 as temperature rises from 10°C to 70°C, while at
43
44 7. C ONCLUSIONS
the same time, the saturated vapor pressure P sat

H2 O increases by a factor 25 [47]. Therefore, for future
selection, ILs must have a melting temperature below 0°C.
γH2 O Φsat sat

V H2 O (P − P sat
" #
y H2 O H O PH O
2 2 H O)
2
K H2 O = = exp (7.1)
x H2 O Φ̂H2 O P RT
5. Although preliminary experiments show low activity coefficients for water in the tested ILs, no firm
conclusions can be drawn on this matter. Additional experiments are required at different temperatures
to determine the temperature dependency of the activity coefficient. Furthermore, due to the limited
range of the sensor, it was not possible to measure the composition dependency of the ILs, as all ILs
were still relatively dry. Due to this low range, it was also not possible to measure the maximum amount
of water per unit of weight that the ILs can absorb. Finally, no reliable absolute values were obtained,
based on the KF titrations.
Based on the molecular simulation results and the results obtained from the Ionic Liquid experiments, it
is too premature to select the best method for high pressure hydrogen dehydration. However, based on the
thermodynamics, it can be concluded that most of the water can be removed by condensing the water out.
Experiments are required to see if it would be possible to get the water content down to 5 ppm, by reducing
the temperature and/or increasing the pressure even further.
If this is not possible, an additional drying unit is still required. It is not possible to point out the best dehydra-
tion method, based on this work. There are two drying methods favorable over the others. Firstly Temperature
Swing Adsorption (TSA), but only if the purging gas is recycled, to prevent hydrogen losses. The other option
is absorption, in which case ILs look like a promising alternative for conventional absorbents, mainly for their
negligible vapor pressure.
8
R ECOMMENDATIONS
8.1. T HERMODYNAMICS
1. The molecular simulation predictions of the vapor phase were in good agreement with experimental
data. However, to further improve the molecular simulations, the polarizability of the water molecules
must be included.
2. It is highly recommended to do VLE experiments on the H2 -H2 O system at high pressures, for several
reasons.
(a) Even though the molecular simulations are in excellent agreement with the available experimental
data, the simulated data are still predictions. Since there is only one paper available that describes
the system at pressures exceeding 300 bar, which is also from 1927, more experimental data are
required to validate the molecular simulations. For the experiments of Bartlett, compressed hy-
drogen was bubbled through water and then expanded to atmospheric pressure. The resulting
gas was dehydrated by means of phosphorus pentoxide. As a result, the phosphorus pentoxide
gained weight, which was equal to the water that was originally in the vapor phase. Although
these experiments can be repeated relatively easy, state-of-the-art sensors like the one used in
this work, can measure water content at low pressures more accurate. An additional advantage of
the experiments would be that the experimental data can be used to fit EoS with more complex
physically based mixing rules. These fitted EoS can than easily be integrated within existing tools,
which significantly reduces calculating time compared to molecular simulations.
(b) Since the absorption will take place at temperatures near the melting point of water, ice forma-
tion may occur, which is not included in the model. Experiments are required to measure ice
formation at temperatures near and below 0°C.
(c) Finally, as mentioned in Section 3.5.2, gas-gas immiscibility may occur. This can result in the
behaviour that if temperature is gradually reduced, the saturated water content will initially de-
crease, but at a certain moment this can potentially increase again. This phenomenon is also not
included in the molecular simulations and therefore experiments are needed.
8.2. I ONIC L IQUID S ELECTION

1. To select the optimal IL, a software tool is required that is capable of predicting IL properties more
accurately. To do this, more experimental data are required that can validate this tool.
2. On top of the criteria stated in Section 4.2, new criteria are added for future selection.
• The melting point of the IL Tmelt must be below zero.

• The surface tension is preferably high, to avoid foam production.
• The IL must be hygroscopic, but does not necessarily have to be hydrophilic. A balance should
be found between the capability of the IL for absorption and for desorption, to avoid high energy
requirements for regeneration of the IL.
45
46 8. R ECOMMENDATIONS
8.3. I ONIC L IQUID E XPERIMENTS

1. For safety reasons, it is recommended to frequently check the SDS of the chemicals one is working with,
as we have experienced that these sheets can change over time.
2. An additional sensor would be helpful to continuously measure the input water content. Although
the flowrate through the water was unchanged, the water content in the hydrogen input showed some
instabilities. If the input and the output would always be measured simultaneously, the results would
be more reliable.
3. Unfortunately, it was not possible to obtain valid results for the activity coefficient, as a function of
temperature and composition. A few measures are recommended to improve the experiments.
(a) Firstly, the absolute water content was not accurately measured. It is recommended to install
the entire KF equipment in the glove box. During the filling and draining of the scrubber, which
should also happen in the glove box, the KF titrations must be executed. This way additional mea-
sure are taken to avoid surrounding air contact, and all samples are taken at the same time and
conditions. It is also advised to explore other methods to measure the water content in substances.
(b) To determine the temperature dependency of the activity coefficients, additional experiments at
different temperatures are required.
(c) Finally, a sensor is required that has a higher range, so it is capable of detecting higher levels of
water content in the hydrogen gas. This is needed to determine the maximum water content in
the IL, as well as to calculate the composition dependency of the activity coefficient.
4. In a further phase of the research, it is recommended to do the absorption experiments at elevated

pressure, to check if the activity coefficient does indeed not depend significantly on pressure.
8.4. N EXT PHASE

Based on this work, the following actions are recommended, in order to find the best method for HP drying
of hydrogen gas:
1. Even though it is unlikely to get it adjusted, the rational for the 5 ppm restriction needs to be investi-
gated. The risk of ice formation during the refueling process is rejected. This is not possible, as multiple
studies showed that the temperature will only rise during this process. Water can provide a transport
mechanism for other contaminants, but it is questionable if the water content must be this low to avoid
that.
2. The first step is to validate the molecular simulation results with accurate HP VLE experiments. There-
after, the temperature can be reduced further to see if a water content of 5ppm can be achieved with
condensing only.
3. If an additional dehydration step is still required, both TSA and absorption should be investigated more
extensively. For absorption, it is recommended to calculate and experimentally validate the vapor pres-
sure of conventional absorbents, like TEG. Even though these absorbents have the disadvantage of con-
taminating the hydrogen gas, the high pressure and low temperature conditions can result in a very low
vapor pressure. At these conditions, the vapor pressure can become so low, that the advantage of ILs
almost disappears.
4. If more extensive comparison between TSA and absorption shows that absorption is favorable, but TEG
is not suitable due to evaporation, further research in ILs is recommended.
B IBLIOGRAPHY
[1] Adoption of the Paris Agreement, Tech. Rep. FCCC/CP/2015/L.9/Rev.1 (United Nations, Framework Con-
vention on Climate Change, Paris, 2015).
[2] Transforming our world: the 2030 Agenda for Sustainable Development, Tech. Rep. A/RES/70/1 (United
Nations, General Assembly, New York, 2015).
[3] World Energy Outlook 2018, Tech. Rep. (International Energy Agency, 2018).
[4] A. Gallo, J. Simões-Moreira, H. Costa, M. Santos, and E. Moutinho dos Santos, Energy storage in the en-
ergy transition context: A technology review, Renewable and Sustainable Energy Reviews 65, 800 (2016).
[5] B. Zakeri and S. Syri, Electrical energy storage systems: A comparative life cycle cost analysis, Renewable
and Sustainable Energy Reviews 42, 569 (2015).
[6] A. Evans, V. Strezov, and T. J. Evans, Assessment of utility energy storage options for increased renewable
energy penetration, Renewable and Sustainable Energy Reviews 16, 4141 (2012).
[7] A. van Wijk and C. Hellinga, Hydrogen the key to the energy transition, Tech. Rep. (Delft University of
Technology and TVVL, 2018).
[8] T. Mahlia, T. Saktisahdan, A. Jannifar, M. Hasan, and H. Matseelar, A review of available methods and
development on energy storage; technology update, Renewable and Sustainable Energy Reviews 33, 532
(2014).
[9] B. Johnston, M. C. Mayo, and A. Khare, Hydrogen: the energy source for the 21st century, Technovation
25, 569 (2005).
[10] J. Adolf, C. Balzer, and J. Louis, Energy of the Future? Sustainable Mobility through Fuel Cells and H2 ,
Tech. Rep. (Shell Deutschland Oil GmbH, Wuppertal Institut, Hamburg, 2017).
[11] P. Bouwman, Electrochemical Hydrogen Compression (EHC) solutions for hydrogen infrastructure, Fuel
Cells Bulletin 2014, 12 (2014).
[12] M. Nordio, F. Rizzi, G. Manzolini, M. Mulder, L. Raymakers, M. Van Sint Annaland, and F. Gallucci, Exper-
imental and modelling study of an electrochemical hydrogen compressor, Chemical Engineering Journal
369, 432 (2019).
[13] M. Bampaou, K. D. Panopoulos, A. I. Papadopoulos, P. Seferlis, and S. Voutetakis, An Electrochemical

Hydrogen Compression Model, Chemical Engineering Transactions 70 (2018), 10.3303/CET1870203.
[14] Hydrogen fuel - Product specification - Part 2: Proton exchange membrane (PEM) fuel cell applications for
road vehicles, Tech. Rep. ISO 14687-2:2012(E) (International Organization for Standardization, Switzer-
land, 2012).
[15] E. P. Bartlett, The concentration of water vapor in compressed hydrogen, Nitrogen and a mixture of these
gases in the presence of condensed water, Journal of the American Chemical Society 49, 65 (1927).
[16] J. Hallet and W.-C. Tu, Possible use of ionic liquids to dry hydrogen gas, Tech. Rep. (Imperial College, 2016).
[17] G. Yu, C. Dai, L. Wu, and Z. Lei, Natural Gas Dehydration with Ionic Liquids, Energy and Fuels 31, 1429
(2017).
[18] J. Han, C. Dai, Z. Lei, and B. Chen, Gas drying with ionic liquids, AIChE Journal 64, 606 (2018).
47
48 B IBLIOGRAPHY
[19] M. Krannich, F. Heym, and A. Jess, Characterization of Six Hygroscopic Ionic Liquids with Regard to
Their Suitability for Gas Dehydration: Density, Viscosity, Thermal and Oxidative Stability, Vapor Pressure,
Diffusion Coefficient, and Activity Coefficient of Water, Journal of chemical & engineerging data 61, 1162
(2016).
[20] E. Hendriks, G. M. Kontogeorgis, R. Dohrn, J.-C. De Hemptinne, I. G. Economou, L. Fele, Z. . Ilnik, and
V. Vesovic, Industrial Requirements for Thermodynamics and Transport Properties, Ind. Eng. Chem. Res
49, 11131 (2010).
[21] J. S. Lopez-Echeverry, S. Reif-Acherman, and E. Araujo-Lopez, Peng-Robinson equation of state: 40 years

through cubics, Fluid Phase Equilibria 447, 39 (2017).
[22] R. Sun, S. Lai, and J. Dubessy, Calculations of vapor-liquid equilibria of the H2 O − N2 and H2 O − H2
systems with improved SAFT-LJ EOS, Fluid Phase Equilibria 390, 23 (2015).
[23] A. Rahbari, J. Brenkman, R. Hens, M. Ramdin, J. P. V. D. Broeke, R. Schoon, R. Henkes, O. A. Moultos, and
T. J. H. Vlugt, Solubility of Water in Hydrogen at High Pressure : a Molecular Simulation Study, .
[24] J. D. Seader, Separation Process Principles, 3rd Edition (2006).
[25] T. C. Yan and K. Sciamanna, High Pressure Hydrogen Design, MIT Practice School Project Report (2017).
[26] C. J. B. Dicken and W. Mérida, Measured effects of filling time and initial mass on the temperature distri-
bution within a hydrogen cylinder during refuelling, Journal of Power Sources 165, 324 (2007).
[27] E. W. Lemmon, I. Bell, M. L. Huber, and M. O. McLinden, NIST Standard Reference Database 23: Reference
Fluid Thermodynamic and Transport Properties-REFPROP, Version 10.0, National Institute of Standards
and Technology, (2018).
[28] M. Monde, P. Woodfield, T. Takano, and M. Kosaka, Estimation of temperature change in practical hy-
drogen pressure tanks being filled at high pressures of 35 and 70 MPa, International Journal of Hydrogen
Energy 37, 5723 (2012).
[29] N. De Miguel, R. Ortiz Cebolla, B. Acosta, P. Moretto, F. Harskamp, and C. Bonato, Compressed hydrogen
tanks for on-board application: Thermal behaviour during cycling, International Journal of Hydrogen
Energy 40, 6449 (2015).
[30] I. Simonovski, D. Baraldi, D. Melideo, and B. Acosta-Iborra, Thermal simulations of a hydrogen storage
tank during fast filling, International Journal of Hydrogen Energy 40, 12560 (2015).
[31] L. M. Robeson, Polymeric Membranes for Gas Separation, (2016), 10.1016/B978-0-12-803581-8.03297-5.
[32] M. Harasimowicz, P. Orluk, G. Zakrzewska-Trznadel, and A. G. Chmielewski, Application of polyimide

membranes for biogas purification and enrichment, Journal of Hazardous Materials 144, 698 (2007).
[33] M.-B. Hägg and X. He, Chapter 15. Carbon Molecular Sieve Membranes for Gas Separation, 2, 162 (2011).
[34] M. Netusil and P. Ditl, Comparison of three methods for natural gas dehydration, Journal of Natural Gas
Chemistry 20, 471 (2011).
[35] C. Heald, H. W. Thompson, P. Roy Soc, and P. G. Menon, Adsorption of Carbon Monoxide on Alumina at
High Pressures, Tech. Rep. 11 (1962).
[36] J. Vermesse, D. Vidal, and P. Malbrunot, Gas Adsorption on Zeolites at High Pressure, Tech. Rep. (1996).
[37] Y. Zhou and L. Zhou, Fundamentals of High Pressure Adsorption †, Langmuir 25, 13461 (2009).
[38] Triethylene Glycol, Tech. Rep. XXX-0207X CRCG (The Dow Chemical Company, 2007).
[39] R. Ortiz Cebolla, B. Acosta, N. De Miguel, and P. Moretto, Effect of precooled inlet gas temperature and
mass flow rate on final state of charge during hydrogen vehicle refueling, International Journal of Hydro-
gen Energy 40, 4698 (2015).
[40] SURFACE VEHICLE TECHNICAL INFORMATION REPORT , Tech. Rep. (SAE Technical Standards, 2014).
B IBLIOGRAPHY 49
[41] R. Hołyst and A. Poniewierski, Thermodynamics for Chemists, Physicists and Engineers (Springer, 2012)
pp. 1–343.
[42] W. G. Gillespie P.C., Vapor-liquid equilibrium data on water-substitute gas components: N2 − H2 O,

H2 − H2 O, CO − H2 O, H2 − CO − H2 O, and H2 S − H2 O, GPA Research Rep. Rep.No. RR-41, 1 (1980).
[43] U. V.V., Equilibrium compositions of vapor-gas mixtures over solutions, Russ.J.Phys.Chem. 70(7), 1240
(1996).
[44] S. L.-T. D. Maslennikova V.Y., Goryunova N.P., The solubility of water in compressed hydrogen.
Russ.J.Phys.Chem. 50(2), 240 (1976).
[45] K. P.-L. E. B. DeVaney W., Berryman J.M., High temperature v-l-e measurements for substitute gas compo-
nents, GPA Research Rep. Rep.No. RR-30, 1 (1978).
[46] J. Smith, H. Van Ness, and M. Abbott, Introduction to Chemical Engineering Thermodynamics. Sixth
Edition in SI Units (2001).
[47] O. J. Poling B.E., Prausnitz J.M., The Properties of Gases and Liquids, fifth edition.
[48] C.-t. Lin and T. E. Daubert, Estimation of Partial Molar Volume and Fugacity Coefficient of Components
in Mixtures from the Soave and Peng-Robinson Equations of State, 1979, 51 (1980).
[49] A. Drexhage, J.J. and Welsenes, Physical properties of pure compounds. Parameters for the SMIRK equa-
tion of state, Shell Internal Report (1990).
[50] T. Y. Kwak E H Benmekki and C. A. Ransoori, VAN DER WAALS MIXING RULES FOR CUBIC EQUATIONS
OF STATE (Applications for Supercritical Fluid Extraction Nodeling and Phase Equilibrium Calculations),
Tech. Rep.
[51] M. L. Michelsen, Fluid Phase Equilibria, Tech. Rep. (1990).
[52] J. Gross and G. Sadowski, Perturbed-Chain SAFT: An Equation of State Based on a Perturbation Theory
for Chain Molecules, Ind. Eng. Chem. Res 40, 1244 (2001).
[53] D. Marx and P. Nielaba, Path-integral monte carlo techniques for rotational motion in two dimensions:
Quenched, annealed, and no-spin quantum-statistical averages, Phys. Rev. A 45, 8968 (1992).
[54] W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, and M. L. Klein, Comparison of simple

potential functions for simulating liquid water, The Journal of Chemical Physics 79, 926 (1983).
[55] J. Rowlinson, J. Baldwin, and A. Buckingham, Liquids and Liquid Mixture, 3rd ed. (1982) p. 328.
[56] T. Seward and E. Franck, The System Hydrogen - Water up to 440°C and 2500 bar Pressure, Ber. Bunsenges.
Phys. Chem. 85, 2 (1981).
[57] W. B. Streett, A. L. Erickson, and J. L. E. Hill, PHASE EQUILIBRIA IN FLUID MIXTURES AT HIGH PRES-
SURES : THE He − CH4 SYSTEM, Phys. Earth Planet. Interiors 6, 69 (1972).
[58] S. T. Handy, Ionic Liquids - Classes and Properties (InTech, 2011).
[59] J. F. Brennecke and E. J. Maginn, Ionic liquids: Innovative fluids for chemical processing, AIChE Journal
47, 2384 (2001).
[60] Y. Y. Abe, T. Takekiyo, Y. Imai, and Hiroshi, High Pressure Phase Behavior of Two Imidazolium-Based Ionic
Liquids, [bmim][BF4] and [bmim][PF6], Ionic Liquids Classes and Properties (2011), 10.5772/32009,
arXiv:9809069v1 [arXiv:gr-qc] .
[61] M. Ramdin, T. W. De Loos, and T. J. H. Vlugt, State-of-the-Art of CO2 Capture with Ionic Liquids, Ind. Eng.
Chem. Res 51, 8149 (2012).
[62] M. Hasib-Ur-Rahman, M. Siaj, and F. Larachi, Ionic liquids for CO2 capture-Development and progress,
Chemical Engineering and Processing 49, 313 (2010).
50 B IBLIOGRAPHY
[63] J. Huang and T. Rther, Why are ionic liquids attractive for CO2 absorption? An overview, Australian Journal
of Chemistry 62, 298 (2009).
[64] E. Torralba-Calleja, J. Skinner, and D. Gutiérrez-Tauste, CO2 Capture in Ionic Liquids: A Review of Solu-
bilities and Experimental Methods, Journal of Chemistry 2013, 1 (2013).
[65] F. Heym, J. Haber, W. Korth, B. J. Etzold, and A. Jess, Vapor Pressure of water in mixtures with hydrophilic
ionic liquids - A contribution to the design of processes for drying of gases by absorption in ionic liquids,
Chemical Engineering and Technology 33, 1625 (2010).
[66] J. Kumelan, A. P.-S. K. Kamps, D. Tuma, and G. Maurer, Solubility of H2 in the Ionic Liquid [BMIM][PF6 ],
J. Chem. Eng. Data 41, 11 (2006).
[67] A. Kordi and F. Sabzi, Thermodynamic modeling of hydrogen solubility in a series of ionic liquids, Inter-
national Journal of Hydrogen Energy , 1 (2018).
[68] V. V. Chaban and O. V. Prezhdo, Computationally Efficient Prediction of Ionic Liquid Properties, J. Phys.
Chem. Lett 5, 1973 (2014).
[69] J. Jacquemin, P. Nancarrow, D. W. Rooney, M. F. C. Gomes, P. Husson, V. Majer, A. A. H. Pádua, and

C. Hardacre, Prediction of Ionic Liquid Properties. II. Volumetric Properties as a Function of Temperature
and Pressure, Journal of chemical & engineerging data 53, 2133 (2008).
[70] A. Fredenslund, R. L. Jones, and J. M. Prausnitz, Group-Contribution Estimation of Activity Coefficients

in Nonideal Liquid Mixtures, AIChE Journal 21, 1086 (1975).
[71] H. Grensemann and J. Gmehling, Performance of a Conductor-Like Screening Model for Real Solvents
Model in Comparison to Classical Group Contribution Methods, Ind. Eng. Chem. Res 44, 1610 (2005).
[72] J. Lohmann, R. Joh, and J. Gmehling, From UNIFAC to Modified UNIFAC (Dortmund), Ind. Eng. Chem.
Res 40, 957 (2001).
[73] A. Klamt, Conductor-like Screening Model for Real Solvents: A New Approach to the Quantitative Calcula-
tion of Solvation Phenomena Starting from the question of why dielectric continuum models give a fairly
good description of molecules, J. Phys. Chem 99, 2224 (1995).
[74] E. Mullins, R. Oldland, Y. A. Liu, S. Wang, S. I. Sandler, C.-C. Chen, M. Zwolak, and K. C. Seavey, Sigma-
Profile Database for Using COSMO-Based Thermodynamic Methods, Ind. Eng. Chem. Res 45, 4389 (2006).
[75] K. Guo, Y. Bi, L. Sun, H. Su, and L. Hungpu, Experiment and Correlation of Vapor-Liquid Equilibrium
of Aqueous Solutions of Hydrophilic Ionic Liquids: 1-Ethyl-3-methylimidazolium Acetate and 1-Hexyl-3-
methylimidazolium Chloride, Journal of chemical & engineering data 57, 2243 (2012).
[76] G. Gonfa, M. A. Bustam, A. M. Sharif, N. Mohamad, and S. Ullah, Tuning ionic liquids for natural gas
dehydration using COSMO-RS methodology, Journal of Natural Gas Science and Engineering 27, 1141
(2015).
[77] M. Mohammadi, M. Asadollahzadeh, and S. Shirazian, Molecular-level understanding of supported ionic

liquid membranes for gas separation, Journal of Molecular Liquids 262, 230 (2018).
[78] C. Jork, C. Kristen, D. Pieraccini, A. Stark, C. Chiappe, Y. A. Beste, and W. Arlt, Tailor-made ionic liquids,
J. Chem. Thermodynamics 37, 537 (2005).
[79] Ddbst - dortmund data bank software & separation technology gmbh, activity coefficient calculation,
(2017 (2.9.0.24)).
[80] Sigma-Aldrich, The Global Harmonised System (GHS), Tech. Rep.
[81] L. Wu, W. Geng, L. Gao, J. Chen, and H. Zhao, Study of Gas Dehydration Process by Ionic Liquid Method
in a Rotating Packed Bed, Energy and Fuels 31, 13400 (2017).
[82] P. Scovazzo, Testing and evaluation of room temperature ionic liquid (RTIL) membranes for gas dehumid-
ification, Journal of Membrane Science 355, 7 (2010).
B IBLIOGRAPHY 51
[83] F. Radakovitsch, F. Heym, and A. Jess, Gas Drying Using Supported Ionic Liquids, CHEMICAL ENGINEER-
ING TRANSACTIONS 69 (2018).
[84] M. G. Freire, C. M. S. S. Neves, I. M. Marrucho, J. A. P. Coutinho, and A. M. Fernandes, Hydrolysis of

Tetrafluoroborate and Hexafluorophosphate Counter Ions in Imidazolium-Based Ionic Liquids †, J. Phys.
Chem. A 114, 3744 (2010).
[85] S. Fendt, S. Padmanabhan, H. W. Blanch, and J. M. Prausnitz, Viscosities of Acetate or Chloride-Based

Ionic Liquids and Some of Their Mixtures with Water or Other Common Solvents, J. Chem. Eng. Data 56,
31 (2011).
[86] E. Quijada-Maldonado, S. Van Der Boogaart, J. H. Lijbers, G. W. Meindersma, and A. B. De Haan,

Experimental densities, dynamic viscosities and surface tensions of the ionic liquids series 1-ethyl-3-
methylimidazolium acetate and dicyanamide and their binary and ternary mixtures with water and
ethanol at T = (298.15 to 343.15 K), J. Chem. Thermodynamics 51, 51 (2012).
[87] H. Saba, X. Zhu, Y. Chen, and Y. Zhang, Determination of physical properties for the mixtures of [BMIM]Cl
with different organic solvents, Chinese Journal of Chemical Engineering 23, 804 (2015).
[88] M. H. Ghatee and A. R. Zolghadr, Surface tension measurements of imidazolium-based ionic liquids at
liquid-vapor equilibrium, Fluid Phase Equilibria 263, 168 (2008).
[89] M. P. Singh, S. K. Mandal, Y. L. Verma, A. K. Gupta, R. K. Singh, and S. Chandra, Viscoelastic, Surface,
and Volumetric Properties of Ionic Liquids [BMIM][OcSO4 ], [BMIM][PF6 ], and [EMIM][MeSO3 ], Journal
of chemical & engineering data 59 (2014), 10.1021/je5000617.
[90] Y. Cao and T. Mu, Comprehensive Investigation on the Thermal Stability of 66 Ionic Liquids by Thermo-
gravimetric Analysis, Industrial and Engineering Chemistry Research 53, 8651 (2014).
[91] I. Cantat, S. Cohen-Addad, F. Elias, F. Graner, R. Hohler, O. Pitois, F. Rouyer, and A. Saint-Jalmes, Foams
structure and Dynamics, edited by S. Cox, September (Oxford, 2010).
[92] M. Moosavi, F. Khashei, A. Sharifi, and M. Mirzaei, The effects of temperature and alkyl chain length on
the density and surface tension of the imidazolium-based geminal dicationic ionic liquids, The Journal of
Chemical Thermodynamics 107, 1 (2017).
[93] T. F. HOLDEN, N. C. Aceto, and E. F. Schoppet, EFFECTS OF VISCOSITY AND TEMPERATURE ON THE
FOAMING CHARACTERISTICS OF CONCENTRATED WHOLE MILK, journal of dairy science , 359 (1964).
[94] Z. Zhang, R. Yang, W. Gao, and X. Yao, Investigation of [Emim][OAc] as a mild pretreatment solvent for
enhancing the sulfonation efficiency of alkali lignin, RSC Advances 7, 31009 (2017).
[95] Y.-S. Ye, J. Rick, and B.-J. Hwang, Ionic liquid polymer electrolytes, J. Mater. Chem. A 1, 2719 (2013).
[96] J. G. Huddleston, A. E. Visser, W. M. Reichert, H. D. Willauer, G. A. Broker, and R. D. Rogers, Characteri-

zation and comparison of hydrophilic and hydrophobic room temperature ionic liquids incorporating the
imidazolium cation, Green Chemistry 3, 156 (2001).
A
E XPERIMENTAL PROCEDURE
This appendix includes the used materials and apparatus.
A.1. M ATERIALS
I ONIC L IQUIDS
Ionic Liquid Purity Supplier

[BMIM][Chloride] Purity >98% Sigma Aldrich
[EMIM][Acetate] Purity >97% Alfa Aesar
[BMIM][Octyl Sulfate] Purity >99% Alfa Aesar
[EMIM][Ethyl Sulfate] Purity >95% Sigma Aldrich
OTHER C HEMICALS
Substance Purity Supplier

Hydrogen hydrogen 3.0 dry, Purity >99.9%, H2 O < 50 ppmw
Nitrogen nitrogen 5.0, Purity >99.999%
Methanol H2 O < 20 ppmw, VWR Chemicals
KF coulomat CombiCoulomat fritless Merck Milipore
A.2. APPARATUS
Equipment Specifications Range

Hydrogen purifier HEA EP40 max 21 bar and 300°C
Hydrogen sensor ap2e ProCeas 0-550ppmv - H2 O/CO/CH4 /NH3 /H2 S/CO2
Water bath Julabo F32 Mw-basis 10°C - 210°C
Karl Fischer apparatus coulometric, fritless
Mass flow controller Brooks 0-200 Nl/h
53
54 A. E XPERIMENTAL PROCEDURE
A.3. WATER CONTENT REGULATION
Figure A.1: Schematic diagram of experimental apparatus. 1, Hydrogen gas storage; 2, purifier; 3, water scrubber; 4, buffer vessel; 5(a&b),
computer valves; 6, mixing vessel; 7, IL scrubber; 8, thermocouple; 9(a&b), three-way hand valves; 10, Sensor; 11, Incinerator.
The hydrogen gas was hydrated by flowing dry hydrogen gas through water (3) (see Fig. A.1). Thereafter, the
hydrogen flowed through cylinder 4, to make sure no liquid water was left in the hydrogen gas. The drawback
of this unit was that the sensor could not measure water content >550 ppmv. Higher water content was
required though as input, to make sure the ILs could absorb water. This problem was solved by increasing
the flow rate of valve 5a significantly. The resulting water content was measured at a flow rate of 150 NL/h for
5a, and 10.67 Nl/h for 5b. This way, it was calculated that the water content of the saturated hydrogen was
6578 ppmv. For the absorption experiment, the flow rate through 5b was alway kept constant, as a change in
flow rate could have an impact on the hydration process. The saturated hydrogen was diluted with hydrogen,
which resulted in a constant flow rate of 40 Nl/h with a water content of 1755 ppmv.
Registration Number (s) D6618
PED Category Art. 4, Par. 3 X Cat. I Cat. II Cat. III Cat. IV
ACC.:
DIRECTIVE 2014/68/EU Mod. Mod. Mod. Mod. Mod.
DESIGN CONDITIONS
Max working Design Max working Design Test Pressure Barg Vessel
Vessel Pressure Pressure Temp. Temp. Volume
Tag no. Inner Vessel
Barg Barg New Inspection
C C 100ml
Outer Vessel
A A V-410 4 60 5.7
60ml
A.4. I ONIC L IQUID S CRUBBER
2
SECTION A-A
A.4. I ONIC L IQUID S CRUBBER
D6618
45 inbranden
3
20
6
20
4
15
2.8
5
36 8 3 O-ring speedflens 12 10.82 x 1.78 Kalrez 6375 AS 013
7 1 Septum GL14 dop Siliconen rubber
6 2 O-ring 6.02x2.62 Kalrez 6375 AS 108
65 5 3 Assembly Speed Clamp 12 Aluminium
7 4 3 Speedflens 12-3_8 tubing AISI 316 Siliceren
150
1 3.5 3 2 GL14->1/4" Tubing AISI 316L Siliceren
2 3 GL14 open dop PPS Schott
1 1 Assembly Scrubber V-410 Borosilicaatglas
8 ITEM QTY DESCRIPTION DIMENSIONS MATERIAL REMARKS
15 GENERAL TOLERANCES ACC. ISO 2768-mK PROJECTION METHOD
25 SURFACE ROUGHNESS ACC. ISO 1302/4287
PREFERD RANGE
3,2 1,6 0,8 0,4
ROUGHNESS Ra in m (0,001 mm)
vingerfilter p1
8
15
20/12/18
Ref. Ref.
8 6.5 ISSUE DATE Designer Designer DESCRIPTION
APPROVED BY
This document is confidential. Neither the whole nor any part of this document may be disclosed to any third party without the prior written consent of Shell Global
Solutions International B.V. and Shell International Exploration and Production B.V., The Hague, The Netherlands. The copyright of this document is vested in these
Companies All rights reserved. Neither the whole nor any part of this document may be reproduced, stored in any retrieval system or transmitted in any form or by
any means (electronic, mechanical, reprographic, recording or otherwise) without the prior written consent of the copyright owners.
H2 Contamination Unit (2016)

Scrubber V-410
SCALE:
STANDARD DRAWING UNITS: mm Sheet No. 001 of Size: A3
SHELL GLOBAL SOLUTIONS INTERNATIONAL B.V.
SHELL INTERNATIONAL EXPLORATION AND PRODUCTION B.V. ID 6667 DRG No. 6667 - 18 - 3100 Rev. 0
FILENAME: 6667-18-3100n001.dwg
55
B
S OLUBILITY OF WATER IN H YDROGEN AT
H IGH P RESSURE :
A M OLECULAR S IMULATION S TUDY
The following paper is submitted to the Journal of Chemical and Engineering Data.
https://pubs.acs.org/journal/jceaax
57
Solubility of Water in Hydrogen at High Pressure:
a Molecular Simulation Study
Ahmadreza Rahbari,† Jeroen Brenkman,‡ Remco Hens,† Mahinder Ramdin,† Leo
J. P. van den Broeke,† Rogier Schoon,‡ Ruud Henkes,‡ Othonas A. Moultos,† and
Thijs J. H. Vlugt∗,†
†Engineering Thermodynamics, Process & Energy Department, Faculty of Mechanical,

Maritime and Materials Engineering, Delft University of Technology, Leeghwaterstraat 39,
2628CB, Delft, The Netherlands
‡Shell Global Solutions International, PO Box 38000, 1030BN, Amsterdam, the Netherlands
E-mail: t.j.h.vlugt@tudelft.nl
Abstract
Hydrogen is one of the most popular alternatives for energy storage. Due to its low
volumetric energy density, hydrogen should be compressed for practical storage and
transportation purposes. Recently, Electrochemical Hydrogen Compressors (EHC) have
been developed that are capable of compressing hydrogen up to P = 1000 bar, and have
the potential of reducing compression costs to 3 kWh/kg. As EHC compressed hydrogen
is saturated with water, the maximum water content in gaseous hydrogen should meet
the fuel requirements issued by the International Organization for Standardization (ISO)
when refuelling Fuel Cell Electric Vehicles (FCEV). The ISO 14687-2:2012 standard has
limited the water concentration in hydrogen gas to 5 µmol water per mol hydrogen fuel
mixture. Knowledge on the vapor liquid equilibrium of H2 O − H2 mixtures is crucial for
designing a method to remove H2 O from compressed H2 . To the best of our knowledge,
1
the only experimental high pressure data (P > 300 bar) for H2 O − H2 phase coexistence
is from 1927 [J. Am. Chem. Soc., 1927, 49, pp 65-78]. In this paper, we have used
molecular simulation and thermodynamic modelling to study the phase coexistence of
the H2 O − H2 system for temperatures between T = 283 K to T = 423 K and pressures
between P = 10 bar and P = 1000 bar. It is shown that the PR-EoS and SRK-EoS
with van der Waals mixing rules fail to accurately predict the equilibrium coexistence
compositions of the liquid and gas phase, with or without fitted binary interaction
parameters. We have shown that the solubility of water in compressed hydrogen is
adequately predicted using force field based molecular simulations. The modelling of
phase coexistence of H2 O − H2 mixtures will be improved by using polarizable models
for water. In the Supporting Information, we present a detailed overview of available
experimental VLE and solubility data for the H2 O − H2 system at high pressures.
1 Introduction
The world population is expected to grow rapidly, from 7.6 billion currently, to about 9.8
billion in 2050. 1 Due to increasing prosperity, the worldwide consumption of energy per
individual will also increase. Even in the current modern world, several billion people still do
not have access to basic needs, such as clean water, sanitation, nutrition, health care, and
education. 2 These are all examples of the Sustainable Development Goals, adopted by all
United Nations Member States in 2015. 2 Access to energy is a key enabler to reach these
basic needs. The worldwide energy demand is therefore expected to increase by 40% by
2040. 3 At the same time, CO2 emissions need to be reduced to reach the goals of the Paris
agreement. 4 80% of the total primary energy supply is currently produced by fossil fuels,
such as coal, oil, and natural gas. 3 To reach the goals of the Paris agreement, attempts are
made to replace fossil fuels with renewable alternatives such as wind and solar (PV) energy.
Current expectations are that by 2040, 40% of the total generated electricity will be from
renewable energy sources. 3
2
Unlike fossil fuels, energy production from intermittent renewable sources, including wind
power and solar energy, critically depend on the availability of these sources leading to
an uncontrollable energy output. 5 For direct integration to the power grid, uncontrollable
availability of intermittent renewable energy sources within 10% of the installed capacity is
acceptable without major technical problems. 5 However, large scale integration of intermittent
energy sources above this limit is expected to cause frequent mismatches between the supply
and demand of energy. To avoid this, integration of energy storage technologies is proposed
as one of the promising solutions for stable and flexible supply of electricity. 6,7 Different
types of technologies have been developed for electrical energy storage including: hydro
storage, flywheels, batteries, and hydrogen produced by electrolysis etc. 5,6,8,9 One of the most
popular alternatives for energy storage is hydrogen. 8 Hydrogen has the advantage that it
can be stored for long periods and converted to electricity without pollution. 10 Hydrogen
has a broad span of applications, like fuel cells, fuel for heating, transportation, or even
as a raw material for the chemical industry. 5,10,11 Since hydrogen has a very low density
at standard conditions, it has a very low volumetric energy density. For practical storage
and transportation purposes, the density of hydrogen must be increased significantly. 12 The
density of hydrogen can be increased by compression, cooling, or a combination of both,
depending on the scale and application. 12 One of the emerging applications for hydrogen
is found in sustainable transportation. 10 In Fuel Cell Electric Vehicles (FCEV), hydrogen
is stored in compressed form in pressurized cylinders at P = 350 bar or P = 700 bar. 12 In
practice, a passenger car needs a tank capacity of around 100 to 150 litres to store 4 to 6
kg of hydrogen, which provides a range of approximately 500 km. 12 High pressure storage
tanks with pressures of at least P = 875 bar 12,13 are installed at refuelling stations, to fuel a
vehicle within the target time of three to five minutes. 12 Conventional compressor types that
are currently used are piston, compressed air, diaphragm, or ionic compressors, depending
mainly on the capacity of the refuelling station. 12 The conventional compressor requires on
average 6 kWh/kg of energy to compress hydrogen from 10 to 400 bar. 13
3
An alternative compressor is the Electrochemical Hydrogen Compressor (EHC). HyET
has developed an EHC that works with pressures up to 1000 bar, and has the potential of
bringing compression costs down to 3 kWh/kg. 13 The working principle of an EHC operation
is similar to a Proton Exchange Membrane (PEM) fuel cell. 14 A single EHC stack consists of
a low pressure and a high pressure side, separated by a membrane that is only permeable for
hydrogen protons, and not for molecules. The membrane is positioned between two platinum
catalysts containing electrodes. Once a potential difference is applied over the electrodes, a
hydrogen molecule splits into two protons. The protons then travel through the membrane
where conversion to hydrogen molecules takes place at elevated pressure. 14 In the EHC, the
proton transfer through the membrane is enabled by water. The EHC has several advantages
compared to traditional technologies: 13,15–17 (1) the EHC has a higher efficiency, especially
at high compression ratios. 18 In theory, the compression ratio using EHC can go to infinity,
from an electrochemical perspective. The mechanical strength and back diffusion losses are
the main limitations for higher pressure ratios for the EHC; 18 (2) due to the highly selective
membrane that only allows the permeation of protons, contaminants are prevented from
passing the membrane. 18 This means that the EHC performs both as a compressor and
a purifier of hydrogen gas; 18 (3) the compressor has no moving parts, resulting in lower
maintenance costs and making lubricants, which may contaminate the compressed hydrogen,
redundant; (4) the EHC operates silently, since it has no rotating parts. This makes the EHC
suitable for locations such as refuelling stations, where acoustical emission is a constraint; (5)
the EHC is a compact device that is well suited to scale up. 18 Disadvantages of the EHC are
similar to those of fuel cells, mainly high material costs. For instance, the platinum catalyst
which is required to resist the corrosive environments in the compressor, is very expensive. 18
Another disadvantage is related to the proton transport through the membrane. Water enables
the proton transport through the membrane and therefore the membrane always needs to
be hydrated. 19 Therefore, the resulting hydrogen gas is saturated with water which can be
an issue depending on the application. The International Organization for Standardization
4
(ISO) stated that water provides a transport mechanism for water-soluble contaminants such
as K+ and Na+ when present as an aerosol. 20 Both K+ and Na+ can affect the fuel cell and
are not recommended to exceed 0.05 µmol K+ or Na+ per mol hydrogen fuel mixture. 20 To
avoid potential issues, the ISO has directed the maximum allowed concentration of impurities
for gaseous hydrogen, including water in Table 1 of the ISO 14687-2:2012. 20 The maximum
concentration of water in the gaseous hydrogen, used for PEM fuel cells in road vehicles is
limited to 5 µmol water per mol hydrogen fuel mixture. 20 This poses two important questions:
(1) what is the solubility of water in hydrogen at high pressures? (2) if this solubility is too
large, what is the best method to reduce the water content? To answer these questions, an
accurate description of Vapor Liquid Equilibrium (VLE) of the H2 O − H2 system at high
pressures is required. Published experimental data that describe these systems is scarce.
To the best of our knowledge, the only experimental data describing phase coexistencce of
H2 O − H2 for pressures exceeding 300 bar are from 1927 (limited to T = 323 K 21 ). Wiebe
and Gaddy studied also the solubility of hydrogen gas in liquid water at high pressures up
to P = 1013.25 bar. 22 Therefore, molecular simulation and thermodynamic modelling are
needed to determine the water content in the compressed hydrogen. In industrial applications,
cubic type Equations of State (EoS) are one the most commonly used methods to study VLE,
because of their simplicity. 23–27 In this work, the Peng-Robinson (PR) EoS and the Soave
Redlich-Kwong (SRK) EoS with van der Waals mixing rules are used to predict the phase
coexitence of H2 O − H2 at elevated pressures. However, molar volumes of the liquid phase
and fugacity coefficients at high pressures obtained from PR-EoS and SRK-EoS modelling
(with conventional mixing rules) are known to deviate significantly from experiments. 28–31
In this work, it is shown that both the PR-EoS and SRK-EoS fail to describe the liquid
phase and the gas phase compositions, with or without fitted binary interaction parameters
(kij ’s). Since water is a highly polar molecule, either modifications of the conventional mixing
rules are required, 24 or more physically based models (i.e. SAFT types EoS 32 or molecular
simulations 33 ) should be used to describe the phase behavior of the H2 O − H2 system. 32 It
5
was found that a temperature dependent parameter kij is still required for SAFT type EoS
modelling. 34 Therefore, force field based molecular simulation could be considered as a natural
tool to study the phase coexistence of the H2 O − H2 system. In this work, different molecular
force fields for water and hydrogen are considered for describing the phase coexistence
compositions of the liquid and gas phase of the H2 O − H2 system, especially at high pressures.
To evaluate the accuracy of the results from molecular simulations, we have performed an
extensive literature survey on the VLE of H2 O − H2 mixtures, at high pressures. 21,22,35–42 In
this work, it is shown that the best predictions of the VLE of the H2 O − H2 system at high
pressures (in both phases) are obtained using molecular simulations. No adjustable kij ’s were
used for molecular simulations in this study.
This paper is organized as follows. In Section 2, the molecular simulation techniques used
in this study are explained and simulation details (molecular simulations and EoS modelling)
and force field details for water and hydrogen are provided. Our results obtained from
molecular simulations and EoS modelling are presented and compared with experimental data
in Section 3. Our conclusions are summarized in Section 4. In the Supporting Information,
we present a detailed overview of available experimental VLE and solubility data for the
H2 O − H2 system at high pressures.
2 Modelling and Methodology
2.1 Simulation Techniques
The natural choice for VLE phase equilibrium calculations is the Gibbs Ensemble (GE)
method introduced by Panagiotopoulos, 43–45 which is used extensively in molecular simulation
studies. 33 In the GE, the vapor and liquid phase are simulated in two simulation boxes, which
can exchange molecules, volume and energy. At coexistence, the pressures, temperatures and
chemical potentials of each component are equal in both boxes. 33 The GE is reliable, and the
finite size effects are small unless conditions close to the critical point are considered. 46,47 To
6
accurately predict coexistence densities, simulations in the GE rely on sufficient molecule
exchanges between the two phases. 33,44 The well-known drawback of the conventional GE is
that at high densities, particle insertions/deletions have a low acceptance probability, also
leading to poor estimates of chemical potentials in both phases. Although chemical potentials
of different component types are not strictly needed for calculating the coexistence densities,
the equality of chemical potentials is an important condition for phase equilibrium. Chemical
potential calculations in the GE follow from a modification of the Widom’s Test Particle
Insertion (WTPI), 33,48 taking fluctuations in density into account. It is well known that
WTPI method often performs poorly for dense liquids. 49,50
Based on the work of Shi and Maginn, 51,52 Vlugt and co-workers expanded the conventional
GE with so-called fractional molecules to improve the efficiency of molecule exchanges between
the simulation boxes. 53,54 In contrast to the normal or "whole" molecules, the interactions of
fractional molecules are scaled between zero and one with a coupling parameter λi . λi = 0
means that the fractional molecule of type i has no interactions with the surrounding molecules
and acts as an ideal gas molecule. λi = 1 means that the fractional molecule of type i has
fully scaled interactions and interacts as a whole molecule. The fractional molecule of
each component type can be in either one of the phases. 53 In addition to the conventional
thermalization trial moves (translation, rotation and volume changes), three additional trial
moves are associated with the fractional molecule of each component: (1) changes in λi while
keeping the positions and orientations of all molecules including the fractional molecule(s)
fixed; (2) Reinsertion of the fractional molecule to a randomly selected position in the other
simulation box (phase) while keeping the value of λi , positions and orientations of all other
molecules fixed; (3) Changing the identity of the fractional molecule with a randomly selected
molecule of the same type in the other box, while keeping the value of λi , positions and
orientations of all molecules fixed. The use of fractional molecules significantly improves the
efficiency of the VLE calculations and the calculations of chemical potentials at coexistence.
For details, the reader is referred to Refs. 49,53,55
7
Since molecule exchanges in the Continuous Fractional Component GE (CFCGE) are
performed using fractional molecules with scaled interactions, molecule transfers between
coexisting phases are facilitated leading to a more efficient sampling of coexistence densities.
The chemical potential of component type i in phase j (gas or liquid) is obtained from: 53,54
!
hρij i p(λij = 1)
µij = kB T ln − kB T ln (1)
ρ0 p(λij = 0)
where kB is the Boltzmann constant, ρij is the number density of component i in phase j and
p(λij ) is the probability distribution of λij in phase j. The term ρ0 is an arbitrary reference
density to make the argument of the logarithm dimensionless. The first term on the right
hand side of Eq. 1 is the ideal gas contribution of the chemical potential (µid
ij ). The second
term on the right hand side of Eq. 1 is the excess chemical potential (µex
ij ). The brackets
h· · · i denote an ensemble average. The fugacity coefficient of component type i in phase j

follows from:
1 p (λij = 0)
φij = × (2)
Zmix p (λij = 1)
where Zmix is the compressibility factor of the mixture. Eq. 2 is derived in the Supporting
Information. Based on the limited experimental solubility data available in literature at
T = 323 K and pressures above P = 300 bar, 21 we know that the solubility of water in the
gas phase at high pressures (P = 100 bar to P = 1000 bar) is about a couple of hundred
PPMs (molar), or less. At lower temperatures, due to the low the solubility of water in
hydrogen, a very large number of hydrogen molecules (up to a million) in the gas phase
would be required in the simulations to have on average a single water molecule in the gas
phase. The solubility of hydrogen in the liquid phase is also very low, e.g. mole fractions
ranging from between 0.003 to 0.115 at T = 323 K and pressures between of P = 25 bar and
P = 1000 bar. This makes most simulations of the H2 O − H2 system in the CFCGE at low
temperatures and high pressures impractical, as a very large system is needed to have at least
8
a single component of each type in each box. One could in principle simulate the VLE of
H2 O − H2 in the CFCGE using a smaller system size. This would lead to poor statistics for
the average number of H2 molecules in the liquid phase, and H2 O molecules in the gas phase.
To circumvent these issues, both the gas and liquid phases (almost pure hydrogen gas and
pure liquid water, respectively) are simulated independently in the Continuous Fractional
Component NPT (CFCNPT ) ensemble. 55 By varying the mixture composition in the gas
and liquid phases around the equilibrium state, the coexistence compositions are obtained by
imposing equal chemical potentials for both phases. Vlugt and co-workers considered the
conventional NPT ensemble expanded with a fractional molecule, 55 similar to earlier work
with the GE. 53 Similar to the CFCGE, trial moves for the fractional molecule are performed
in addition to the usual thermalization moves. The only difference is that in the simulations
in the CFCNPT ensemble, the trial moves related to the fractional molecule are performed
in the same simulation box. By applying the CFCMC method to the NPT ensemble, one
can calculate the chemical potential of each species (similar to Eq. 1). For details the reader
is referred to Refs. 55,56
At high pressures we know that the solute is almost pure in both phases, i.e. hydrogen
in the gas phase and water in the liquid phase. For a solution close to infinite dilution, one
can express the variation of the excess chemical potential of the solute, i.e. hydrogen in the
liquid phase and water in the gas phase as a function of the number density of the solute:
µex 2
ij (ρij ) = Aij + Bij ρij + Cij ρij + · · · (3)
To obtain the terms Aij , Bij , · · · , multiple simulations are performed at constant temperature
and pressure, for different concentrations of the solute. In the region of interest (very dilute
solutions) µex
ij (ρij ) depends linearly on the number density. As the solvent in both phases is
almost a pure component, one can assume that the excess chemical potential of the solvent is
independent of the number of few solute molecules in that phase. The coexistence densities
9
are then obtained by imposing equal chemical potentials of each component using Eq. 3.
Note that at conditions where both methods are applicable to obtain phase coexistence (i.e.
simulations in the GE and the CFCNPT ensemble), we have verified that both methods (i.e.
GE and imposing equal chemical potentials) yield the same results.
2.2 Simulation Details
Depending on the temperature and pressure, molecular simulations are performed in the
CFCGE or in the CFCNPT ensemble. All simulations were performed using our in-house code.
It was verified that our results are identical to those from the RASPA software package. 57,58
In all simulations, periodic boundary conditions were used. All molecules are rigid and
the interactions between the molecules only consist of LJ and Coulombic interactions. LJ
potentials were truncated but not shifted. Analytic tail corrections and the Lorentz-Berthelot
mixing rules were applied. 33,59 To treat the electrostatic interactions, the Ewald summation
was used with a relative precision of 1 × 10−6 . In CFCGE simulations of H2 O − H2 mixtures,
fractional molecules of water and hydrogen are present which are used to facilitate molecule
exchanges between the phases. To protect the charges from overlapping, the (repulsive) LJ
interactions of the fractional molecules are switched on before the electrostatics. 60–65 For
details about scaling the LJ and Coulombic interactions of the fractional molecule, the reader
is referred to Refs. 54,56,66 Details about the force field parameters for different water and
hydrogen models and cutoff radii for LJ interactions are provided in tables S1 and S2 of the
Supporting Information.
Simulations in the CFCGE ensemble were started with 730 molecules of water and 600
molecules of hydrogen. For all temperatures and pressures, 105 equilibration cycles were
carried out followed by 4 · 106 production cycles. Each MC cycle consists of NMC Monte Carlo
trial moves, where NMC equals the total number of molecules, with a minimum of 20. Trial
moves in the CFCGE simulations were selected with the following probabilities: 1% volume
changes, 35% translations, 30% rotations, 17% λ changes, 8.5% reinsertions of fractional
10
molecules at randomly selected positions in the other box, and 8.5% identity changes of
fractional molecules between the boxes. Independent CFCNPT simulations of the liquid
phase, close to infinite dilution of hydrogen, were performed with 730 water molecules with
NH2 ∈ h0, 10i hydrogen molecules. Similarly, independent CFCNPT simulations of the gas
phase, close to infinite dilution of water, were performed with 600 hydrogen molecules with
NH2 O ∈ h0, 7i water molecules. Trial moves in the CFCNPT simulations were selected with
the following probabilities: 1% volume changes, 35% translations, 30% rotations, 17% λ
changes, 8.5% reinsertions of fractional molecules at randomly selected positions, and 8.5%
identity changes of fractional molecules.
2.3 Force Fields
To model the VLE of H2 O − H2 mixtures, molecular force fields are considered to predict
the density and composition of the gas and liquid phases. As the most commonly used
force fields are developed based on single-phase coexistence data, 67,68 we have screened these
force fields using single-phase hydrogen (gas phase) and single-phase water (liquid phase)
simulations. Force fields for water and hydrogen are selected based on predicting bulk
properties of pure phases such as densities, chemical potentials and fugacity coefficients. The
densities and fugacity coefficients of molecular hydrogen in the gas phase are computed at
different pressures using several force fields from the literature. The results are compared
with REFPROP. 69,70 Common force fields for molecular hydrogen in literature include single
site, 71–73 two-site 74 and multi-site potentials with (permanent) charge interactions. 75–77 Single-
site hydrogen models are capable of predicting bulk thermodynamic properties of hydrogen
accurately. The single-site hydrogen model by Buch 71 reproduces the bulk properties of
hydrogen accurately up to high pressures. Multi-site hydrogen potentials that consider charge-
quadrupoles and polarizability are more relevant for modeling hydrogen sorption in highly
heterogeneous systems. 74,75,77–80 The densities and the excess chemical potentials predicted
by different force fields of water in the liquid phase are computed as a function of pressure.
11
The results are compared to those obtained from REFPROP. 69,81 Even though water is
a flexible and polarizable molecule, to date most molecular simulations studies consider
rigid molecular potentials of water with constant point-charges. 68,82–85 It is computationally
advantageous to use these simplified water potentials, which can predict thermodynamic
and transport properties of water in good agreement with experiments. To obtain a more
physical description of water, polarizable force fields have been developed to account for
polarization effects. 68,86–95 Compared to the fixed-charge water potentials, thermodynamic
properties of polarizable force fields are not fully known. 87 Commonly used fixed-charge
force fields for water are three-site potentials: TIP3P, 96 SPC 97,98 and SPC/E, 99 four-site
potentials: TIP4P/2005, 83 TIP4P/Ew, 100 OPC, 67 and a five-site potential: TIP5P/Ew. 101
In our previous studies, 49,56 we have shown that the computed excess chemical potentials
of water for the three-site potentials TIP3P and SPC are in good agreement with values
obtained from an empirical Helmholtz equation of state 81 based on experimental data. 70 It is
well-known that the TIP4P/2005 water outperforms the three-site models for predicting bulk
properties of water such as the density. 83 In our previous studies, we have shown that the
computed excess chemical potentials of water obtained from four-site and five-site potentials
show larger deviations from experimental data compared to three-site potentials 49 .
2.4 Equation of State Modelling
The PR-EoS 102 and SRK-EoS 103 with the conventional van der Waals mixing rules are used
to predict the H2 O − H2 VLE. These equations of state are the most widely used in industry
and perform best for describing the VLE of non-polar mixtures. 104 It is well-known that the
molar volume of the liquid phase predicted by cubic equations of state is inaccurate. 105,106
Since the solubility of small gas nonpolar molecules in the liquid phase are dominated by
entropic effects (i.e., molar volume), the solubility of H2 in H2 O is predicted poorly. We have
used both zero kij ’s and kij ’s fitted on high pressure experimental data. Details on the EoS
modelling are provided in the Supporting Information.
12
3 Results and Discussion
3.1 Molecular Simulations
The densities and the fugacity coefficients of pure hydrogen between P = 100 bar and
P = 1000 bar obtained from CFCNPT simulations and EoS modelling are compared to
those obtained from REFPROP, 107 see Fig. 1. Since the differences between the results
obtained for P < 400 bar is very small, only the results between P = 400 bar and P = 1000
bar are shown, and the raw data are provided in table S3 of the Supporting Information.
Hydrogen models used for this study include single-site models: i.e. Hirschfelder, 72 Vrabec, 73
Buch, 71 two-site model: i.e. Cracknell 74 and the multi-site model of Marx. 77 It is clear that
the densities obtained using the Buch 71 and Marx 77 force fields are in excellent agreement
with experimental data up to P = 1000 bar. The results obtained from the PR-EoS and
SRK-EoS deviate from experimental data for P > 400 bar. The calculated fugacity coefficients
of pure hydrogen in the gas phase are best predicted using the Buch 71 and Marx 77 force
fields. The calculated fugacity coefficients from the SRK-EoS are in excellent agreement
with experiments. The simulation results show that both the Buch and Marx force fields
outperform the other molecular models in predicting bulk densities and fugacity coefficients of
hydrogen at high pressures. This means that considering a quadrupole moment for hydrogen
does not strictly improve the bulk properties of hydrogen in the gas phase. Including the
quadrupole moment may improve the prediction of phase coexistence in the liquid phase,
as observed by Sun et al. 34 Therefore, the Marx force field is considered further for VLE
simulations of H2 O − H2 mixtures.
The densities and chemical potentials of TIP3P, 96 SPC, 98 SPC/E, 99 TIP4P/2005, 83
TIP4P/Ew, 108 OPC 67 and TIP5P/Ew 101 force fields between P = 100 bar and P = 1000
bar obtained from CFCNPT simulations are compared to the IAPWS empirical EoS, 69,81
see Fig. 2. Raw data are provided in table S4 in the Supporting Information. It is shown
13
in Fig. 2(a) that the force fields TIP5P/Ew and TIP4P/2005 clearly outperform the TIP3P
and SPC force fields in predicting the density of liquid water (on average around 2%) over
the whole pressure range. The TIP4P/2005 water is parameterized based on temperature of
maximum density of liquid water, the stability of several ice polymorphs etc. 83 The TIP5P/Ew
model is obtained from reparametrization of the TIP5P model 109 which is also a very accurate
model capable of predicting maximum density of liquid water around 4◦ C. 101 Note that the
deviations of the densities obtained from the TIP3P and SPC models decrease with increasing
pressure. As shown in Fig. 2, the chemical potential of water is best predicted using the
TIP3P and SPC force fields over the whole temperature range. This observation is also in
agreement with previous works. 49,56 The performance of the TIP3P and the SPC force fields
are very similar in calculating the densities and chemical potentials of water. The TIP3P
force field has been parameterized to the vaporization energy and density of liquid water. 96
This is consistent with the fact that the computed chemical potential of TIP3P water is
in better agreement with IAPWS empirical EoS, compared to TIP4P/2005 or TIP5P/Ew
models. Raw data for Fig. 2 are provided in table S4 of the Supporting Information. As
shown in Fig. 2, the average deviation of the chemical potential of TIP3P force field from the
IAPWS empirical EoS 69,81 is about +50 K (in units of energy/kB ) for the whole pressure
range. The average deviations of the chemical potentials for the TIP4P/2005 and TIP5P/Ew
force fields from IAPWS empirical EoS are ca. −500 K and +250 K, respectively. The
performance of the SPC/E force field is very similar to the TIP4P/2005 force field for
predicting the densities and chemical potentials of water. For the 4-site water force fields,
the densities and chemical potentials of the TIP4P/2005 force field show the best agreement
with the experiments. Due to overall difference between the predicted densities and chemical
potentials of these water models, it is not a priori clear which water model is best fitted for
predicting the VLE of H2 O − H2 mixtures. Therefore, three water models are considered
(TIP3P,TIP4P/2005,TIP5P/Ew) in combination with the Marx force field (for hydrogen) for
phase coexistence calculations of H2 O − H2 mixtures, using molecular simulations.
14
The water content in the gas phase and the solubility of hydrogen in the liquid phase
for the mixture defined by the TIP3P-Marx force fields are obtained from phase coexistence
equilibrium calculations, see Fig. 3. To check the consistency between the results with both
methods, phase coexistence calculations at T = 323 K and P > 100 bar are performed for both
(i.e. CFCGE and CFCNPT ). It is shown that both methods yield the same results within
the error bars. At T = 283 K, all simulations are performed only in the CFCNPT ensemble
for the whole pressure range. At T = 310 K and P > 100 bar, phase coexistence calculations
are also performed using simulations in the CFCNPT ensemble. At T = 366 K and T = 423
K, phase coexistence calculations are performed using simulations in the CFCGE. Raw data
from experimental results are provided in tables S5 and S6 of the Supporting Information,
and the simulations results are provided in table S7 in the Supporting Information. Based on
available experimental data at pressures above P = 300 bar, 21 it is clear that the predicted
solubility of TIP3P water in the gas phase is in good agreement with experimental data.
At T = 283 K, no experimental solubilities have been found, and therefore only the results
obtained from molecular simulations are shown. For all isotherms of water vapor in the gas
phase, it can be observed that the water content is slightly overpredicted at low pressures. At
high pressures, the solubility of water in the gas phase is marginally underpredicted. From
the condition of chemical equilibrium, we know that the chemical potential of water in the gas
phase is equal to the chemical potential of water in the liquid phase. Therefore, it seems that
good performance of the TIP3P force field to predict the isotherms of water in the gas phase
is most likely related to how accurate it can predict µH2 O in the liquid phase. Based on the
results shown in Figs. 2 and 3 it can be concluded that parametrization of the TIP3P force
field based on the evaporation energy as one of the target quantities is essential for predicting
the VLE of H2 O − H2 mixtures. For all temperatures in this study (between T = 283 K and
T = 423 K), it is observed that the solubility of water in the gas phase at coexistence is
significantly higher than 5 µmol water per mol hydrogen (as allowed by the ISO standard 20 ).
Therefore, an additional step for removing water is needed.
15
The calculated isotherms for hydrogen in the liquid phase (TIP3P-Marx) are clearly
overpredicted compared to experimental data as shown in Fig. 3(b). To the best of our
knowledge, experimental solubility data for hydrogen isotherms in the liquid phase at pressures
above ca. P = 140 bar are not available in the literature, except at T = 323 K. 21 The
deviation from experimental solubilities of hydrogen at T = 323 K ranges from about 36%
to 18% between P = 50 bar and P = 1000 bar, respectively. At T = 366 K, the deviation
from experimental data is about 50% between P = 50 bar and P = 100 bar. At T = 423 K
the deviation from experimental data is about 110% between P = 50 bar and P = 80 bar.
Therefore, it can be concluded that the deviation of simulation results from experimental
data increases with increasing temperature. Based on these results, it can also be concluded
that the deviation from experimental solubilities decreases with increasing pressure. Similarly,
better agreement is observed between experimental densities of water and those obtained
based on TIP3P water at high pressures as also shown in Fig. 2. This suggests that predicting
the density of the liquid phase (almost pure water) accurately may result in predicting the
mixture compositions in better agreement with experiments.
The solubilities obtained from phase coexistence at equilibrium for the H2 O − H2 mixture
defined by the TIP4P/2005-Marx force fields are shown in Fig. 4. Raw data are provided in
table S8 of the Supporting Information. For this mixture, all simulations are performed in the
CFCGE, at T = 323 K, T = 366 K and T = 423 K. It is clear from Fig. 4 that the solubilities
of water in the gas phase are significantly underestimated for the whole pressure range. This
is mainly due to the fact that the chemical potential of TIP4P/2005 water is significantly
underpredicted, as shown in Fig. 2. Since the predicted water solubilities in the gas phase are
systematically lower for the TIP4P/2005-Marx mixture (see Figs. 3a and 4a), the statistics
for water solubilities obtained from CFCGE simulations are worse. This sampling issue is
explained in Section 2.1. Similarly, the computed isotherms of hydrogen in the liquid phase
are slightly underpredicted. For the mixture defined by TIP4P2005-Marx force fields, better
agreement with experiments is observed for solubilities in the liquid phase for all temperatures.
16
At T = 366 K, the deviation from experimental data is about 14% between P = 50 bar and
P = 100 bar. At T = 423 K the deviation from experimental data is about 5% between
P = 50 bar and P = 80 bar.
The solubilities obtained from phase coexistence at equilibrium for H2 O − H2 mixture
defined by TIP5P/Ew-Marx force fields are shown in Fig. 5. Raw data are provided in table S9
of the Supporting Information. For this mixture, all simulations are performed in the CFCGE,
at T = 323 K, T = 366 K and T = 423 K. In sharp contrast to the TIP4P/2005-Marx
mixture, both calculated solubilities in the liquid and gas using the TIP5P/Ew-Marx mixture
are overpredicted. The solubilities of hydrogen in the liquid phase are very similar to those
obtained form the TIP3P-Marx force fields. To explain the results in a coherent way, it is
important to consider the predicted water isotherms in the gas phase in Figs. 3 to 5 and the
calculated chemical potentials of pure water in Fig. 2(b) simultaneously. From these figures,
it can be concluded that underpredicting the solubilties in the gas phase is directly related to
underpredcting the chemical potential of water (TIP4P/2005). Similarly, overpredicting the
solubilties of water in the gas phase is directly related to overpredcting the chemical potential
of water (TIP5P/Ew).
3.2 Equation of State Modelling
The water content in the gas phase and the solubility of hydrogen in the liquid phase are
also calculated using the PR-EoS and SRK-EoS. High pressure experimental solubilites at
T = 323 K were used to obtain the Binary Interaction Parameters (kij ’s) for the PR-EoS
and SRK-EoS. For T = 323 K, the isotherms of water and hydrogen in the gas and liquid
phase are shown in Fig. 6 using both zero kij ’s and non-zero kij ’s. In Fig. 6, it is shown that
the predicted solubilities in the liquid phase are significantly lower compared to experiments,
using zero kij ’s. The solubility of (nonpolar) gases is dominated by entropic effects which are
related to the molar volume. 110 It is well-known that the predicted volumes of the liquid phase
from PR-EoS or SRK-EoS, using conventional mixing rules, have significant errors. 24,25,106
17
Since 2017, more than 220 modifications of mixing rules for pure components and extensions
to mixtures with the PR-EoS have been reported in literature. 24 This clearly indicates
the need for more physically based models for thermodynamic modelling. In addition, the
H2 O − H2 system is highly polar in the liquid phase, and the performance of the conventional
mixing rules for PR-EoS and SRK-EoS for polar mixture are known to be poor. 24 Therefore,
it is expected that PR-EoS or SRK-EoS are not able to predict solubilities of hydrogen in
liquid water accurately. With the fitted kij ’s, the obtained solubilities of hydrogen in the
liquid phase are in excellent agreement with experimental data for p < 400 bar. However, the
solubilities in the gas phase deviate significantly using the fitted kij ’s. Therefore, calculations
of VLE of H2 O − H2 mixtures using PR-EoS and SRK-EoS do not yield satisfactory results
for both phases simultaneously, with or without adjusted kij ’s.
4 Conclusions
Molecular simulations are used to model the VLE behavior of H2 O − H2 mixtures for pressures
between P = 10 bar and P = 1000 bar. In tables S5 and S6, a detailed overview of available
experimental data has been provided for this system. It is shown that commonly used
cubic EoS, with conventional mixing rules fail to predict the composition of the gas and the
liquid phases accurately. For the different molecular models for hydrogen, the Buch force
field 71 (single-site model) and the Marx force field (including quadrupole moment) predict
the density and fugacity coefficient of hydrogen in good agreement with experiments up to
P = 1000 bar. In this study, no force field for rigid water with fixed point charges could
accurately predict both the chemical potential and the density of water. The computed
chemical potentials of TIP3P water 96 have the best agreement with experimental data from
REFPROP 69 with a deviation of about +50 K (µ/kB ) for pressures between P = 100 bar and
P = 1000 bar. This may be partly due to the fact that one of the target fitting parameters
for the TIP3P force field is the heat of vaporization, unlike the TIP4P/2005 and TIP5P/Ew
18
force fields. The computed chemical potentials (µ/kB ) of the TIP4P/2005 and TIP5P/Ew
deviate on average by −500 K and +250 K from experimental data in this pressure range,
respectively. Both the TIP4P/2005 and TIP5P/Ew force fields can predict the density of
liquid water in good agreement with the experiments for the whole pressure range. From the
simulations result, it is observed that solubilities of water in the gas phase are systematically
underpredicted when using the TIP4P/2005 force field. This force field also underpredicts
the chemical potential of liquid water compared to experiments. The highest solubilities
in the gas phase are predicted using the TIP5P/Ew force field with the largest values for
the calculated chemical potential of water. The best agreement between the predicted
gas phase compositions and experiments for the whole pressure range are observed for the
TIP3P fore field. This suggests that a suitable water force field for studying the VLE of
H2 O − H2 mixtures can be screened based on the chemical potential of the water model in
the liquid phase. Based on the screening of seven water force fields in this study, it turns
out that the TIP3P and SPC force fields (with very similar values for chemical potential of
liquid water) can best predict the equilibrium vapor phase coexistence composition of the
H2 O − H2 system. For all temperatures in this study, we observed that the solubility of water
in the gas phase at coexistence is significantly higher than 5 µmol water per mol hydrogen
(as allowed by the ISO standard). Therefore, an additional step for removing extra water
from the gas phase is required. Despite the fact that the molecular simulations significantly
outperform cubic EoS modelling for the VLE of H2 O − H2 mixtures, the predicted liquid
phase compositions need further improvements. The solubilities of hydrogen in the liquid
phase are overpredicted using the TIP3P-Marx and TIP5P/Ew-Marx force fields. The best
agreement between the calculated liquid phase composition and experiments is observed for
the TIP4P/2005-Marx system (although the predicted solubilities are slightly lower). Further
improvements in simulations of H2 O − H2 systems may be realized taking polarizability of
water molecules into account. Therefore, further molecular simulations of the H2 O − H2 are
recommended using polarizable force fields for water, especially to improve the predictions
19
for the liquid phase composition.
Acknowledgement
This work was sponsored by NWO Exacte Wetenschappen (Physical Sciences) for the use
of supercomputer facilities, with financial support from the Nederlandse Organisatie voor
Wetenschappelijk Onderzoek (Netherlands Organization for Scientific Research, NWO). TJHV
acknowledges NWO-CW for a VICI grant.
Supporting Information Available
Derivation of Eq. 2; force field parameters for water and hydrogen, calculated densities and
fugacity coefficients for pure hydrogen in the gas phase (different force fields); calculated
densities and chemical potentials for pure liquid water (different force fields); experimental
solubilities of hydrogen in the H2 O − H2 mixture (liquid phase) at coexistence; experimental
solubilities of water vapor in the H2 O − H2 mixture (gas phase) at coexistence; raw data
for the computed coexistence compositions of H2 O − H2 mixtures at equilibrium using MC
simulations; equation of state parameters for the PR-EoS and SRK-EoS
20
(a)
50
45
40
35
30
25
20
400 500 600 700 800 900 1000
(b)
2
1.9
1.8
1.7
1.6
1.5
1.4
1.3
1.2
1.1
400 500 600 700 800 900 1000
Figure 1: Comparison of different models to predict (a): the density and (b): the fugacity
coefficient of pure hydrogen in the gas phase at T = 323 K and pressures ranging between
P = 10 and P = 1000 bar. PR-EoS (left-pointing triangle), SRK-EoS (asterisk), experimental
data from REFPROP 69,70 (lines), molecular force fileds: Hirschfelder 72 (squares), Vrabec 73
(Plus signs), Buch 71 (upward-pointing triangles), Cracknell 74 (downward-pointing triangles)
and Marx 77 (right-pointing triangles). Parameters for the EoS are provided in table S10 of
the Supporting Information. Raw simulation data are provided in table S3 of the Supporting
Information.
21
(a)
1040
1030
1020
1010
1000
990
980
970
960
0 200 400 600 800 1000
(b)
-3600
-3800
-4000
-4200
-4400
-4600
-4800
-5000
0 200 400 600 800 1000
Figure 2: Comparison of different force fields of water to predict (a): the density and (b):
the chemical potential in the liquid phase at T = 323 K and pressures ranging between
P = 10 and P = 1000 bar: TIP3P 96 (diamonds), SPC 98 (circles), SPC/E 99 (right-pointing
triangles), TIP4P/2005 83 (squares), TIP4P/Ew 108 (downward-pointing triangles), OPC 67
(upward-pointing triangles), TIP5P/Ew 101 (left-pointing triangles). In both subfigures, the
lines are obtained from REFPROP. 69,70 Raw data are provided in table S4 of the Supporting
Information.
22
(a)
100
10-1
10-2
10-3
10-4
10-5
0 200 400 600 800 1000
(b)
0.03
0.025
0.02
0.015
0.01
0.005
0
0 200 400 600 800 1000
Figure 3: Vapor-Liquid equilibrium of H2 O − H2 (TIP3P 96 -Marx 77 ) at pressures ranging

between P = 10 and P = 1000 bar. (a): yH2 O in the gas phase and (b): xH2 in the liquid
phase. T = 423 K (upward-pointing triangles), T = 366 K (downward-pointing triangles),
T = 323 K (squares), T = 310 K (circles), T = 283 (right-pointing triangles). Experimental
data for T = [423, 366, 323, 310] K are shown with dashed lines, dash-dot lines, solid lines
and dotted lines, respectively. Published high pressure data are only available for T = 323
K. 21 Raw data are provided in tables S5, S6 and S7 of the Supporting Information.
23
(a)
100
10-1
10-2
10-3
10-4
10-5
0 200 400 600 800 1000
(b)
0.03
0.025
0.02
0.015
0.01
0.005
0
0 200 400 600 800 1000
Figure 4: Vapor-Liquid equilibrium of H2 O − H2 (TIP4P/2005 83 -Marx 77 ) at pressures ranging

between P = 10 and P = 1000 bar. (a): yH2 O in the gas phase and (b): xH2 in the liquid
phase. T = 423 K (upward-pointing triangles), T = 366 K (downward-pointing triangles),
T = 323 K (squares). Experiential data for T = [423, 366, 323, 310] K are shown with dashed
lines, dash-dot lines, solid lines and dotted lines, respectively. Published high pressure data
are only available for T = 323 K. 21 Raw data are provided in tables S5, S6 and S8 of the
Supporting Information..
24
(a)
100
10-1
10-2
10-3
10-4
10-5
0 200 400 600 800 1000
(b)
0.03
0.025
0.02
0.015
0.01
0.005
0
0 200 400 600 800 1000
Figure 5: Vapor-Liquid equilibrium of H2 O − H2 (TIP5P/Ew 101 -Marx 77 ) at pressures ranging

between P = 10 and P = 1000 bar. (a): yH2 O in the gas phase and (b): xH2 in the liquid phase.
T = 423 K (upward-pointing triangles), T = 366 K (downward-pointing triangles), T = 323
K (squares). Experiential data for T = [423, 366, 323, 310] K are shown with dashed lines,
dash-dot lines, solid lines and dotted lines, respectively. Published high pressure data are
only available for T = 323 K. 21 Raw data are provided in table S5, S6, S9 of the Supporting
Information.
25
(a)
10-2
10-3
10-4
0 200 400 600 800 1000
(b)
0.012
0.01
0.008
0.006
0.004
0.002
0
0 200 400 600 800 1000
Figure 6: VLE of H2 O − H2 at T = 323 K and pressures ranging between P = 100 and

P = 1000 bar, obtained from EoS modelling. (a): mole fraction of water in the gas phase,
(b): mole fraction of hydrogen in the liquid phase. Experimental solubilities are shown with
circles. In both subfigures, the results are shown for kij = 0: PR-EoS 102 (lines), SRK-EoS 103
(dashed lines). The results from the γ-φ method are shown with open symbols: PR-EoS
(upward-pointing triangles) and the SRK-EoS (downward-pointing triangles). The results for
the fitted BIP for the PR-EoS (kij = −0.89) are shown with dash-dot lines. The results for
the fitted BIP for the SRK-EoS (kij = −1.51) are shown with dotted lines.
26
References
(1) World Population Prospects: The 2017 Revision, Key Findings and Advance Tables,
ESA/P/WP/248.; 2017; https://population.un.org/wpp/Publications/Files/
WPP2017_KeyFindings.pdf.
(2) United Nations. (2015) General Assembly Resolution A/RES/70/1. Transforming our
world: the 2030 Agenda for Sustainable Development; 2015; http://www.un.org/ga/
search/view_doc.asp?symbol=A/RES/70/1&Lang=E.
(3) World Energy Outlook 2018 ; 2018; www.iea.org/weo.
(4) Adoption of the Paris Agreement, FCCC/CP/2015/L.9/Rev.1; 2015; https://unfccc.

int/resource/docs/2015/cop21/eng/l09r01.pdf.
(5) Gallo, A.; Simões-Moreira, J.; Costa, H.; Santos, M.; Moutinho dos Santos, E. Energy
storage in the energy transition context: A technology review. Renew. Sust. Energy
Rev. 2016, 65, 800–822.
(6) Zakeri, B.; Syri, S. Electrical energy storage systems: A comparative life cycle cost
analysis. Renew. Sust. Energy Rev. 2015, 42, 569 – 596.
(7) Evans, A.; Strezov, V.; Evans, T. J. Assessment of utility energy storage options for
increased renewable energy penetration. Renew. Sust. Energy Rev. 2012, 16, 4141 –
4147.
(8) Mahlia, T.; Saktisahdan, T.; Jannifar, A.; Hasan, M.; Matseelar, H. A review of
available methods and development on energy storage; technology update. Renew. Sust.
Energy Rev. 2014, 33, 532 – 545.
(9) Chen, H.; Cong, T. N.; Yang, W.; Tan, C.; Li, Y.; Ding, Y. Progress in electrical energy
storage system: A critical review. Progress in Natural Science 2009, 19, 291 – 312.
27
(10) Johnston, B.; Mayo, M. C.; Khare, A. Hydrogen: the energy source for the 21st century.
Technovation 2005, 25, 569 – 585.
(11) Rahbari, A.; Ramdin, M.; van den Broeke, L. J. P.; Vlugt, T. J. H. Combined steam
reforming of methane and formic acid To produce syngas with an adjustable H2 :CO
ratio. Ind. Eng. Chem. Res. 2018, 57, 10663–10674.
(12) Adolf, J.; Balzer, C.; Louis, J. Energy of the Future? Sustainable Mobility through Fuel
Cells and H2 ; 2017.
(13) Bouwman, P. Electrochemical Hydrogen Compression (EHC) solutions for hydrogen

infrastructure. Fuel Cells Bulletin 2014, 2014, 12–16.
(14) Bampaou, M.; Panopoulos, K. D.; Papadopoulos, A. I.; Seferlis, P.; Voutetakis, S. An
electrochemical hydrogen compression model. Chem. Eng. Trans. 2018, 70 .
(15) Ströbel, R.; Oszcipok, M.; Fasil, M.; Rohland, B.; Jörissen, L.; Garche, J. The
compression of hydrogen in an electrochemical cell based on a PE fuel cell design. J.
Power Sources 2002, 105, 208 – 215, 7th Ulmer Elektrochemische Tage.
(16) Suermann, M.; Kiupel, T.; Schmidt, T. J.; Büchi, F. N. Electrochemical hydrogen
compression: efficient pressurization concept derived from an energetic evaluation. J.
Electrochem. Soc. 2017, 164, F1187–F1195.
(17) Rohland, B.; Eberle, K.; Ströbel, R.; Scholta, J.; Garche, J. Electrochemical hydrogen
compressor. Electrochim. Acta. 1998, 43, 3841 – 3846.
(18) Nordio, M.; Rizzi, F.; Manzolini, G.; Mulder, M.; Raymakers, L.; Van Sint Anna-
land, M.; Gallucci, F. Experimental and modelling study of an electrochemical hydrogen
compressor. Chem. Eng. J. 2019, 369, 432–442.
(19) Casati, C.; Longhi, P.; Zanderighi, L.; Bianchi, F. Some fundamental aspects in
28
electrochemical hydrogen purification/compression. J. Power Sources 2008, 180, 103–
113.
(20) Hydrogen fuel - Product specification - Part 2: Proton exchange membrane (PEM) fuel
cell applications for road vehicles; International Standard, 2012; Vol. 2012; www.iso.
org/committee/54560/x/catalogue/.
(21) Bartlett, E. P. The concentration of water vapor in compressed hydrogen, Nitrogen and
a mixture of these gases in the presence of condensed water. Journal of the American
Chemical Society 1927, 49, 65–78.
(22) Wiebe, R.; Gaddy, V. L. The solubility of hydrogen in water at 0, 50, 75 and 100◦
from 25 to 1000 atmospheres. 2005, 56, 76–79.
(23) Hendriks, E.; Kontogeorgis, G. M.; Dohrn, R.; de Hemptinne, J.-C.; Economou, I. G.;
Z̆ilnik, L. F.; Vesovic, V. Industrial requirements for thermodynamics and transport
properties. Ind. Eng. Chem. Res. 2010, 49, 11131–11141.
(24) Lopez-Echeverry, J. S.; Reif-Acherman, S.; Araujo-Lopez, E. Peng-Robinson equation

of state: 40 years through cubics. Fluid Phase Equilib. 2017, 447, 39 – 71.
(25) Iwai, Y.; Margerum, M. R.; Lu, B. C.-Y. A new three-parameter cubic equation of
state for polar fluids and fluid mixtures. Fluid Phase Equilib. 1988, 42, 21 – 41.
(26) Diamantonis, N. I.; Boulougouris, G. C.; Mansoor, E.; Tsangaris, D. M.; Economou, I. G.
Evaluation of cubic, SAFT, and PC-SAFT equations of state for the vapor-liquid
equilibrium modeling of CO2 mixtures with other gases. Ind. Eng. Chem. Res. 2013,
52, 3933–3942.
(27) Kwak, T.; Mansoori, G. A. Van der Waals mixing rules for cubic equations of state.
Applications for supercritical fluid extraction modelling. Chem. Eng. Sci. 1986, 41,
1303–1309.
29
(28) Harstad, K. G.; Miller, R. S.; Bellan, J. Efficient high-pressure state equations. AIChE
J. 1997, 43, 1605–1610.
(29) Jhaveri, B. S.; Youngren, G. K. Three-parameter modification of the Peng-Robinson

equation of state to improve volumetric predictions. SPE reservoir engineering 1988,
3, 1033–1040.
(30) Poling, B. E.; Prausnitz, J. M.; O’Connell, J. P. The properties of gases and liquids,
5th ed.; McGraw-Hill New York: New York, USA, 2001.
(31) Valderrama, J. O. The State of the Cubic Equations of State. Ind. Eng. Chem. Res.
2003, 42, 1603–1618.
(32) Gross, J.; Sadowski, G. Perturbed-Chain SAFT: An equation of state based on a

perturbation theory for chain molecules. Ind. Eng. Chem. Res. 2001, 40, 1244–1260.
(33) Frenkel, D.; Smit, B. Understanding molecular simulation: from algorithms to applica-
tions, 2nd ed.; Academic Press: San Diego, California, 2002.
(34) Sun, R.; Lai, S.; Dubessy, J. Calculations of vapor-liquid equilibria of the H2 O-N2 and
H2 O-H2 systems with improved SAFT-LJ EOS. Fluid Phase Equilib. 2015, 390, 23 –
33.
(35) Meyer, M.; Tebbe, U.; Piiper, J. Solubility of inert gases in dog blood and skeletal
muscle. Pflügers Archiv European Journal of Physiology 1980, 384, 131–134.
(36) Gillespie, P. C.; Wilson, G. M. Vapor-Liquid Equilibrium Data on Water-Substitue Gas

Components; 1980; pp RR_41, 1–34.
(37) Kling, G.; Maurer, G. The solubility of hydrogen in water and in 2-aminoethanol
at temperatures between 323 K and 423 K and pressures up to 16 M Pa. Chem.
Thermodynamics 1991, 23, 531–541.
30
(38) Devaney, W.; Berryman, J. M.; Kao, P.-L.; Eakin, B. High Temperature V-L-E
Measurements for Substitute Gas Components; 1978; pp 1–27.
(39) Jung, J. Loslichkeit von Kohlenmonoxid und Wasserstoff in Wasser zwisen 0 C und
300 C. Ph.D. thesis, RWTH Aachen, 1962.
(40) Ipatev, V.; Teodorovich, V. Equilibrium compositions of vapor-gas mixtures over

solutions. Zh.Obshch.Khim. 1934, 4, 395–399.
(41) Ugrozov, V. V. Equilibrium compositions of vapor-gas mixtures over solutions. 1996,

70, 1240–1241.
(42) Maslennikova, V. Y.; Goryunova, N.; Subbotina, L.; Tsiklis, D. The solubility of water
in compressed hydrogen. Russian Journal of Physical Chemistry 1976, 50, 240–243.
(43) Panagiotopoulos, A. Z. Molecular simulation of phase equilibria: simple, ionic and

polymeric fluids. Fluid Phase Equilib. 1992, 76, 97–112.
(44) Panagiotopoulos, A. Z. Direct determination of fluid-Phase equilibria by simulation in

the Gibbs ensemble - a Review. Mol. Simul. 1992, 9, 1–23.
(45) Panagiotopoulos, A. Z. Direct determination of phase coexistence properties of fluids

by Monte Carlo simulation in a new ensemble. Mol. Phys. 1987, 61, 813–826.
(46) Recht, J.; Panagiotopoulos, A. Z. Finite-size effects and approach to criticality in Gibbs
ensemble simulations. Mol. Phys. 1993, 80, 843–852.
(47) Siepmann, J. I.; McDonald, I. R.; Frenkel, D. Finite-size corrections to the chemical
potential. J. Phys.: Condens. Matt. 1992, 4, 679.
(48) Smit, B.; Frenkel, D. Calculation of the chemical potential in the Gibbs ensemble. Mol.
Phys. 1989, 68, 951–958.
31
(49) Rahbari, A.; Poursaeidesfahani, A.; Torres-Knoop, A.; Dubbeldam, D.; Vlugt, T. J. H.
Chemical potentials of water, methanol, carbon dioxide and hydrogen sulphide at low
temperatures using continuous fractional component Gibbs ensemble Monte Carlo.
Mol. Simul. 2018, 44, 405–414.
(50) Coskuner, O.; Deiters, U. K. Hydrophobic interactions by Monte Carlo simulations.

Zeitschrift für Phys. Chem. 2006, 220, 349–369.
(51) Shi, W.; Maginn, E. J. Continuous Fractional Component Monte Carlo: an adaptive
biasing method for open system atomistic simulations. J. Chem. Theory Comput. 2007,
3, 1451–1463.
(52) Shi, W.; Maginn, E. J. Improvement in molecule exchange efficiency in Gibbs ensemble
Monte Carlo: development and implementation of the continuous fractional component
move. J. Comput. Chem. 2008, 29, 2520–2530.
(53) Poursaeidesfahani, A.; Torres-Knoop, A.; Dubbeldam, D.; Vlugt, T. J. H. Direct free
energy calculation in the Continuous Fractional Component Gibbs ensemble. J. Chem.
Theory Comput. 2016, 12, 1481–1490.
(54) Poursaeidesfahani, A.; Hens, R.; Rahbari, A.; Ramdin, M.; Dubbeldam, D.; Vlugt, T.
J. H. Efficient application of Continuous Fractional Component Monte Carlo in the
reaction ensemble. J. Chem. Theory Comput. 2017, 13, 4452–4466.
(55) Rahbari, A.; Hens, R.; Nikolaidis, I. K.; Poursaeidesfahani, A.; Ramdin, M.;
Economou, I. G.; Moultos, O. A.; Dubbeldam, D.; Vlugt, T. J. H. Computation
of partial molar properties using continuous fractional component Monte Carlo. Mol.
Phys. 2018, 116, 3331–3344.
(56) Rahbari, A.; Hens, R.; Jamali, S. H.; Ramdin, M.; Dubbeldam, D.; Vlugt, T. J. H.
Effect of truncating electrostatic interactions on predicting thermodynamic properties
of water-methanol systems. Mol. Simul. 2019, 45, 336–350.
32
(57) Dubbeldam, D.; Calero, S.; Ellis, D. E.; Snurr, R. Q. RASPA: molecular simulation
software for adsorption and diffusion in flexible nanoporous materials. Mol. Simul.
2015, 42, 81–101.
(58) Dubbeldam, D.; Torres-Knoop, A.; Walton, K. S. On the inner workings of Monte
Carlo codes. Mol. Simul. 2013, 39, 1253–1292.
(59) Allen, M. P.; Tildesley, D. J. Computer simulation of liquids, 2nd ed.; Oxford university
press: Oxford, United Kingdom, 2017.
(60) Klimovich, P. V.; Shirts, M. R.; Mobley, D. L. Guidelines for the analysis of free energy
calculations. J Comput. Aided. Mol. Des. 2015, 29, 397–411.
(61) Naden, L. N.; Pham, T. T.; Shirts, M. R. Linear basis function approach to efficient
alchemical free energy calculations. 1. Removal of uncharged atomic sites. J. Chem.
Theory Comput. 2014, 10, 1128–1149.
(62) Naden, L. N.; Shirts, M. R. Linear basis function approach to efficient alchemical free
energy calculations. 2. Inserting and deleting particles with Coulombic interactions. J.
Chem. Theory Comput. 2015, 11, 2536–2549.
(63) Shirts, M. R.; Pande, V. S. Solvation free energies of amino acid side chain analogs for
common molecular mechanics water models. J. Chem. Phys. 2005, 122, 134508.
(64) Shirts, M. R.; Pitera, J. W.; Swope, W. C.; Pande, V. S. Extremely precise free
energy calculations of amino acid side chain analogs: Comparison of common molecular
mechanics force fields for proteins. J. Chem. Phys. 2003, 119, 5740–5761.
(65) Shirts, M. R.; Mobley, D. L.; Chodera, J. D. In Annual Reports in Computational

Chemistry; Spellmeyer, D., Wheeler, R., Eds.; Elsevier: United States, 2007; pp 41–59.
(66) Rahbari, A.; Hens, R.; Dubbeldam, D.; Vlugt, T. J. H. Improving the accuracy of
33
computing chemical potentials in CFCMC simulations. Mol. phys. in press (uploaded
as SI for review only).
(67) Izadi, S.; Anandakrishnan, R.; Onufriev, A. V. Building Water Models: a different
approach. J. Phys. Chem. Lett. 2014, 5, 3863–3871.
(68) Vega, C.; Abascal, J. L. F.; Conde, M. M.; Aragones, J. L. What ice can teach us
about water interactions: a critical comparison of the performance of different water
models. Faraday Discuss. 2009, 141, 251–276.
(69) Lemmon, E. W.; Span, R. Short fundamental equations of state for 20 industrial fluids.
J. Chem. Eng. Data 2006, 51, 785–850.
(70) Lemmon, E. W.; Huber, M. L.; McLinden, M. O. NIST reference fluid thermodynamic
and transport properties–REFPROP. NIST standard reference database 2002, 23, v7.
(71) Buch, V. Path integral simulations of mixed para-D2 and ortho-D2 clusters: The
orientational effects. J. Chem. Phys. 1994, 100, 7610–7629.
(72) Hirschfelder, C.; Curtiss, F.; Bird, R. B. Molecular Theory of Gases and Liquids; Wiley:
New York, 1954.
(73) Köster, A.; Thol, M.; Vrabec, J. Molecular Models for the Hydrogen Age: Hydrogen,
Nitrogen, Oxygen, Argon, and Water. J. Chem. Eng. Data 2018, 63, 305–320.
(74) Cracknell, R. F. Molecular simulation of hydrogen adsorption in graphitic nanofibres.

Phys. Chem. Chem. Phys. 2001, 3, 2091–2097.
(75) Belof, J. L.; Stern, A. C.; Space, B. An accurate and transferable intermolecular
diatomic hydrogen potential for condensed phase simulation. J. Chem. Theory Comput.
2008, 4, 1332–1337.
(76) Forrest, K. A.; Pham, T.; McLaughlin, K.; Belof, J. L.; Stern, A. C.; Zaworotko, M. J.;
Space, B. Simulation of the mechanism of gas sorption in a metal-organic framework
34
with open metal sites: molecular hydrogen in PCN-61. J. Phys. Chem. C 2012, 116,
15538–15549.
(77) Marx, D.; Nielaba, P. Path-integral Monte Carlo techniques for rotational motion in
two dimensions: Quenched, annealed, and no-spin quantum-statistical averages. Phys.
Rev. A 1992, 45, 8968–8971.
(78) Camp, J.; Stavila, V.; Allendorf, M. D.; Prendergast, D.; Haranczyk, M. Critical factors
in computational characterization of hydrogen storage in metal-organic frameworks. J.
Phys. Chem. C 2018, 122, 18957–18967.
(79) Yang, Q.; Zhong, C. Molecular simulation of carbon dioxide/methane/hydrogen mixture

adsorption in metal-organic frameworks. J. Phys. Chem. B 2006, 110, 17776–17783.
(80) Darkrim, F.; Levesque, D. Monte Carlo simulations of hydrogen adsorption in single-
walled carbon nanotubes. J. Chem. Phys. 1998, 109, 4981–4984.
(81) Wagner, W.; Pruß, A. The IAPWS formulation 1995 for the thermodynamic properties
of ordinary water substance for general and scientific use. J. Phys. Chem. Ref. Data
2002, 31, 387–535.
(82) Vega, C.; Abascal, J. L. F. Simulating water with rigid non-polarizable models: a
general perspective. Phys. Chem. Chem. Phys. 2011, 13, 19663–19688.
(83) Abascal, J. L. F.; Vega, C. A general purpose model for the condensed phases of water:
TIP4P/2005. J. Chem. Phys. 2005, 123, 234505.
(84) Tsimpanogiannis, I. N.; Moultos, O. A.; Franco, L. F. M.; de M. Spera, M. B.; Erdős, M.;
Economou, I. G. Self-diffusion coefficient of bulk and confined water: a critical review
of classical molecular simulation studies. Mol. Simul. 2019, 45, 425–453.
(85) Vega, C. Water: one molecule, two surfaces, one mistake. Mol. Phys. 2015, 113,
1145–1163.
35
(86) Bauer, B. A.; Patel, S. Properties of water along the liquid-vapor coexistence curve via
molecular dynamics simulations using the polarizable TIP4P-QDP-LJ water model. J.
Comput. Phys. 2009, 131, 084709.
(87) Jiang, H.; Moultos, O. A.; Economou, I. G.; Panagiotopoulos, A. Z. Hydrogen-bonding

polarizable intermolecular potential model for water. J. Phys. Chem. B 2016, 120,
12358–12370.
(88) Chen, B.; Xing, J.; Siepmann, J. I. Development of polarizable water force fields for
phase equilibrium calculations. J. Phys. Chem. B 2000, 104, 2391–2401.
(89) Yesylevskyy, S. O.; Schäfer, L. V.; Sengupta, D.; Marrink, S. J. Polarizable Water
Model for the Coarse-Grained MARTINI Force Field. PLoS Comput. Biol. 2010, 1–17.
(90) Gladich, I.; Roeselová, M. Comparison of selected polarizable and nonpolarizable water
models in molecular dynamics simulations of ice Ih . Phys. Chem. Chem. Phys. 2012,
14, 11371–11385.
(91) Kunz, A.-P. E.; van Gunsteren, W. F. Development of a nonlinear classical polarization
model for liquid water and aqueous solutions: COS/D. J. Phys. Chem. A 2009, 113,
11570–11579.
(92) Lamoureux, G.; MacKerell, A. D.; Roux, B. A simple polarizable model of water based
on classical Drude oscillators. J. Chem. Phys. 2003, 119, 5185–5197.
(93) Lamoureux, G.; Harder, E.; Vorobyov, I. V.; Roux, B.; MacKerell, A. D. A polarizable
model of water for molecular dynamics simulations of biomolecules. Chem. Phys. Lett.
2006, 418, 245 – 249.
(94) Ren, P.; Ponder, J. W. Polarizable atomic multipole water model for molecular me-
chanics simulation. J. Phys. Chem. B 2003, 107, 5933–5947.
36
(95) Laury, M. L.; Wang, L.-P.; Pande, V. S.; Head-Gordon, T.; Ponder, J. W. Revised
parameters for the AMOEBA polarizable atomic multipole water model. J. Phys.
Chem. B 2015, 119, 9423–9437.
(96) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L.
Comparison of simple potential functions for simulating liquid water. J. Chem. Phys.
1983, 79, 926–935.
Comparison of simple potential functions for simulating liquid water. J. Comput. Phys.
1983, 79, 926–935.
(98) Mark, P.; Nilsson, L. Structure and dynamics of the TIP3P, SPC, and SPC/E water
models at 298 K. J. Phys. Chem. A 2001, 105, 9954–9960.
(99) Berendsen, H. J. C.; Grigera, J. R.; Straatsma, T. P. The missing term in effective
pair potentials. J. Phys. Chem. 1987, 91, 6269–6271.
(100) Horn, H. W.; Swope, W. C.; Pitera, J. W.; Madura, J. D.; Dick, T. J.; Hura, G. L.;
Head-Gordon, T. Development of an improved four-site water model for biomolecular
simulations: TIP4P-Ew. J. Comput. Phys. 2004, 120, 9665–9678.
(101) Rick, S. W. A reoptimization of the five-site water potential (TIP5P) for use with
Ewald sums. J. Comput. Phys. 2004, 120, 6085–6093.
(102) Peng, D.-Y.; Robinson, D. B. A new two-constant equation of state. Ind. Eng. Chem.
Fundam. 1976, 15, 59–64.
(103) Soave, G. Equilibrium constants from a modified Redlich-Kwong equation of state.

Chem. Eng. Sci. 1972, 27, 1197 – 1203.
(104) Twu, C. H.; Coon, J. E.; Bluck, D. Comparison of the Peng-Robinson and Soave-
Redlich-Kwong equations of state using a new zero-pressure-based mixing Rule for the
37
prediction of high-pressure and high-temperature phase equilibria. Ind. Eng. Chem.
Res. 1998, 37, 1580–1585.
(105) Péneloux, A.; Rauzy, E.; Fréze, R. A consistent correction for Redlich-Kwong-Soave
volumes. Fluid Phase Equilib. 1982, 8, 7 – 23.
(106) Lin, C.-T.; Daubert, T. E. Estimation of partial molar volume and fugacity coefficient
of components in mixtures from the soave and Peng-Robinson equations of state. Ind.
Eng. Chem. Process Des. Dev. 1980, 19, 51–59.
(107) Leachman, J. W.; Jacobsen, R. T.; Penoncello, S.; Lemmon, E. W. Fundamental

equations of state for parahydrogen, normal hydrogen, and orthohydrogen. J. Phys.:
Chem. Ref. Data. 2009, 38, 721–748.
simulations: TIP4P-EW. J. Chem. Phys. 2004, 120, 9665–9678.
(109) Mahoney, M. W.; Jorgensen, W. L. A five-site model for liquid water and the repro-
duction of the density anomaly by rigid, nonpolarizable potential functions. J. Chem.
Phys. 2000, 112, 8910–8922.
(110) Wilhelm, E.; Waghorne, E.; Hefter, G.; Hummel, W.; Maurer, G.; Rebelo, L. P. N.;
da Ponte, M. N.; Battino, R.; Clever, L.; van Hook, A.; Domanska-Zelazna, U.;
Tomkins, R. P. T.; Richon, D.; de Stafani, V.; Coquelet, C.; Costa Gomes, M.;
Siepmann, J. I.; Anderson, K. E.; Eckert, F.; Grolier, J.-P.; Boyer, S.; Salminen, J.;
Dohrn, R.; Leiberich, R.; Fele Zilnik, L.; Fages, J.; Macedo, M. E.; Gmehling, J.;
Brennecke, J.; Cordes, W.; Prausnitz, J.; Goodwin, A. R.; Marsh, K.; Peters, C. J.;
Voigt, W.; Koenigsberger, E.; May, P.; Kamps, A. P.-S.; Shariati, A.; Raeissi, S.;
Padua, A. A. H.; Sauceau, M.; Pinho, S. P. P.; Kaskiala, T.; Kobylin, P. In Developments
38
and Applications in Solubility; Letcher, T. M., Ed.; The Royal Society of Chemistry,
2007.
39
-1 T = 323 K
10
-2
10
-3
10
-4
10
0 200 400 600 800 1000
for Table of Contents use only
40
Supporting Information for: Solubility of Water
in Hydrogen at High Pressure: a Molecular
Simulation Study
Ahmadreza Rahbari,† Jeroen Brenkman,‡ Remco Hens,† Mahinder Ramdin,† Leo
J. P. van den Broeke,† Rogier Schoon,‡ Ruud Henkes,‡ Othonas A. Moultos,† and
Thijs J. H. Vlugt∗,†
†Engineering Thermodynamics, Process & Energy Department, Faculty of Mechanical,

Maritime and Materials Engineering, Delft University of Technology, Leeghwaterstraat 39,
2628CB, Delft, The Netherlands
‡Shell Global Solutions International, PO Box 38000, 1030BN, Amsterdam, the Netherlands
E-mail: t.j.h.vlugt@tudelft.nl
Fugacity Coefficient
Consider a multicomponent system that is simulated in an ensemble that is either open

(e.g. grand-canonical ensemble) or closed (NPT ensemble). In this system, we would like to
calculate the fugacity coefficient φi of component i. We assume that a fractional molecule of
component i is present. The chemical potential of component i equals 1,2
hρi i hρi i p(λi = 1)

µi = µ0i + RT ln + µiex = µ0i + RT ln − RT ln (S1)
ρ0 ρ0 p(λi = 0)
S1
in which µ0i is the reference state of the chemical potential, hρi i is the average number density
of i, µiex is the excess chemical potential of i, ρ0 is an arbitrary reference density (to make
the argument of the logarithm dimensionless), and p(λi ) is the probability distribution of the
coupling parameter of the fractional molecule of i. In classical thermodynamics, the chemical
potential of i is usually expressed as 3,4
!
y i P φi
µi = µ?i + RT ln (S2)
P0
in which µ?i is a reference chemical potential (which is different from µ0i ), yi is the mole
fraction of i, P is the pressure, and P0 is a reference pressure (usually 1 bar). The reference
chemical potentials µ0i and µ?i only depend on the temperature and not on the pressure or
composition of the system. To find an expression for the fugacity coefficient φi , consider a
system in which the pressure P is approaching zero while the composition of the mixture is
constant. In this limit, φi = 1 and µiex = 0. We have
hρi i yi P

µ0i + RT ln = µ?i + RT ln (S3)
ρ0 P0
In this limit, the ideal gas law can also be used to calculate the average number density of i,
yi P
hρi i = (S4)
RT
This leads to
ρ0 RT

µ0i − µ?i = RT ln (S5)
P0
This equation can be used to eliminate the reference state in Eq. S2 leading to
!
y i P φi
µiex = RT ln (S6)
RT hρi i
S2
so
RT hρi i h i
φi = exp µiex /(RT ) (S7)
yi P
If the system consists of Nt molecules in total (including component i, and not counting
fractional molecules), we have hρi i ≈ Ni / hV i (hV i being the average volume) and yi = Ni /Nt .
We finally have
Nt RT h i exp [µiex /(RT )]
φi = exp µiex /(RT ) = (S8)
P hV i Zm
P hV i
in which Zm = Nt RT
is the compressibility of the mixture. The fugacity coefficient φi thus
depends on both the excess chemical potential of i and the overall deviation from ideal gas
behavior of the mixture.
S3
Table S1: Force field parameters for the water models used in this study. L is the dummy
site for four-site models. For the TIP5P/Ew 5 model, L and M are the dummy sites. θ is
the angle between atoms OHO, in degrees. For four-site water models ε = 0.5θ. ϕ is the
angle between the point charges of the dummy sites for the TIP5P/Ew. A cutoff radius of
12 Å was used for all LJ interactions. Analytic tail corrections and the Lorentz-Berthelot
mixing rules were applied. 6,7
Force Field Parameters TIP3P 8 SPC 9 SPC/E 10 OPC 11 TIP4P/2005 12 TIP4P/Ew 13 TIP5P/Ew 5
OO /kB /[K] 76.500 78.203 78.175 107.086 93.196 81.899 89.57888
σOO /[Å] 3.151 3.1656 3.166 3.1666 3.1589 3.1644 3.097
qO /[e] -0.834 -0.82 -0.8476 -1.3582 - -
qH /[e] 0.417 0.41 0.4238 0.6791 0.5564 0.52422 -
qL /[e] - - - -1.3582 -1.1128 -1.04844 -0.241
qM /[e] - - - - - - 0.241
rOH /[Å] 0.957 1 0.8724 0.9572 0.9572
rOL /[Å] - - - 0.1594 0.1546 0.125
θ 104.52 109.47 109.47 103.6 104.52 104.52 104.52
ϕ - - - 51.8 52.26 52.26 109.47
S4
Table S2: Force field parameters for the hydrogen models used in this study. Dummy site L
is the geometric center of mass for the Marx model. The H-H bond length of the two-site
and three-site force fields is 0.74 Å. A cutoff radius of 12 Å was used for all LJ interactions.
Analytic tail corrections and the Lorentz-Berthelot mixing rules were applied. 6,7
Force Field Parameters Buch 14 Hirschfelder 15 Vrabec 16 Cracknell 17 Marx 18

HH /kB /[K] 34.2 38 25.84 12.5 -
σHH /[Å] 2.96 2.915 3.0366 2.59 -
qH /[e] - - - - 0.468
qL /[e] - - - - -0.936
LL /kB /[K] - - - - 36.7
σLL /[Å] - - - - 2.958
S5
Table S3: Calculated densities and fugacity coefficients of different hydrogen models obtained
from CFCNPT simulations at T = 323 K and pressures between P = 100 and P = 1000 bar.
The results from equation of state modelling are obtained using parameters from table S10.
σx is the uncertainty of x. Results from cubic EoS and REFPROP 19,20 are also included.
P /[bar] hρi/[kg/m3 ] σhρi µ/kB /[K] σφ

Hirschfelder 15
100 7.121 0.001 1.05 0.02
200 13.556 0.002 1.10 0.02
400 24.650 0.006 1.23 0.02
600 33.84 0.01 1.37 0.03
800 41.58 0.01 1.53 0.03
1000 48.22 0.02 1.71 0.03
16
Vrabec
100 7.027 0.001 1.06 0.02
200 13.239 0.002 1.14 0.02
400 23.733 0.005 1.30 0.02
600 32.30 0.01 1.48 0.02
800 39.49 0.01 1.67 0.03
1000 45.65 0.01 1.92 0.03
14
Buch
100 7.088 0.001 1.05 0.02
200 13.443 0.003 1.12 0.02
400 24.314 0.006 1.25 0.02
600 33.26 0.01 1.40 0.02
800 40.78 0.01 1.59 0.02
1000 47.22 0.01 1.79 0.03
17
Cracknell
100 7.133 0.001 1.05 0.02
200 13.611 0.003 1.10 0.02
400 24.891 0.005 1.21 0.02
600 34.37 0.01 1.35 0.02
800 42.45 0.01 1.50 0.03
1000 49.45 0.01 1.66 0.03
18
Marx
100 7.100 0.002 1.05 0.02
200 13.482 0.004 1.11 0.02
400 24.416 0.010 1.25 0.02
600 33.41 0.02 1.40 0.03
800 40.97 0.02 1.57 0.03
1000 47.42 0.02 1.76 0.04
Continued on next page
S6
P /[bar] hρi/[kg/m3 ] σhρi µ/kB /[K] σφ
PR-EoS 21
100 7.35 - 1.03 -
200 14.05 - 1.07 -
400 25.60 - 1.17 -
600 35.00 - 1.28 -
800 42.73 - 1.42 -
1000 49.17 - 1.57 -
SRK-EoS 22
100 7.19 - 1.05 -
200 13.58 - 1.11 -
400 24.36 - 1.25 -
600 33.00 - 1.41 -
800 40.04 - 1.60 -
1000 45.87 - 1.82 -
REFPROP 19,20
100 7.10 - 1.06 -
200 13.46 - 1.12 -
400 24.29 - 1.26 -
600 33.17 - 1.42 -
800 40.63 - 1.60 -
1000 47.01 - 1.80 -
S7
Table S4: Calculated densities and chemical potentials of different water models obtained
from CFCNPT simulations at T = 323 K and pressures between P = 100 and P = 1000 bar.
σx is the uncertainty of x. Results from cubic EoS and REFPROP 19,20 are also included.
P /[bar] hρi/[kg/m3 ] σhρi µ/kB /[K] σµ

9
TIP3P
100 965 2 -4057 31
200 970 1 -4033 25
400 982 2 -3989 32
600 992 1 -3948 29
800 1002 1 -3905 30
1000 1011 1 -3861 25
SPC 8
100 962 2 -4075 24
300 973 2 -4021 32
500 982 1 -3987 26
800 996 2 -3912 29
1000 1005 1 -3873 28
10
SPC/E
100 987 2 -4482 36
300 997 1 -4424 33
500 1004 2 -4385 31
800 1017 2 -4311 42
1000 1025 2 -4271 42
TIP4P/2005 12
100 990 2 -4606 47
200 995 2 -4584 44
400 1003 2 -4554 49
600 1012 2 -4494 48
800 1020 2 -4467 55
1000 1027 2 -4408 45
TIP4P/Ew 13
100 987 2 -4433 35
300 996 2 -4388 25
500 1006 1 -4355 38
800 1018 2 -4291 41
1000 1025 2 -4242 38
11
OPC
100 991 2 -4946 38
300 998 2 -4902 52
500 1005 2 -4861 43
S8
P /[bar] hρi/[kg/m3 ] σhρi µ/kB /[K] σµ
800 1017 2 -4808 59
1000 1025 2 -4730 61
TIP5P/Ew 5
100 990 2 -3845 24
300 1000 1 -3821 24
500 1011 2 -3769 27
800 1025 2 -3699 30
1000 1034 2 -3651 29
REFPROP 19,23
100 992.31 - -4112.58 -
200 996.53 - -4089.42 -
400 1004.7 - -4043.47 -
600 1012.6 - -3997.99 -
800 1020.1 - -3952.96 -
1000 1027.4 - -3908.32 -
S9
Table S5: Experimental solubilities of hydrogen in H2 O − H2 mixtures (liquid phase) at
coexistence. Experimental data are converted to mole fractions for different temperatures
and pressures. For the original units for each data set, see the indicated references below.
For conversion to mole fractions, standard conditions at T = 273.15 K and P = 1.01325 atm
are considered, unless otherwise mentioned in the reference.
T /[K] P /[bar] x H2 Ref

−4 24
273.15 25 4.31 × 10
−4 24
273.15 51 8.63 × 10
273.15 101 1.71 × 10−3 24
273.15 203 3.35 × 10−3 24

−3 24
273.15 405 6.40 × 10
273.15 608 9.25 × 10−3 24
−2 24
273.15 811 1.19 × 10
273.15 1013 1.42 × 10−2 24
298.15 25 3.50 × 10−4 24

−4 24
298.15 51 6.94 × 10
298.15 101 1.39 × 10−3 24
−3 24
298.15 203 2.71 × 10
298.15 405 5.24 × 10−3 24
298.15 608 7.64 × 10−3 24

−3 24
298.15 811 9.90 × 10
298.15 1013 1.20 × 10−2 24
−5 25
310.15 1.013 1.33 × 10
310.93 3.4 4.50 × 10−5 26
310.93 13.8 1.81 × 10−4 26

−4 26
310.93 31.0 4.10 × 10
310.93 65.5 8.62 × 10−4 26
−3 26
310.93 103.4 1.33 × 10
310.93 137.9 1.76 × 10−3 26
323.15 25 3.26 × 10−4 24

−4 27
323.15 31.8 4.02 × 10
323.15 51 6.49 × 10−4 24
−4 27
323.15 60.3 7.63 × 10
323.15 101 1.29 × 10−3 24
323.15 119.3 1.51 × 10−3 27

−3 24
323.15 203 2.54 × 10
323.15 405 4.92 × 10−3 24
−3 24
323.15 608 7.18 × 10
323.15 811 9.34 × 10−3 24
323.15 1013 1.14 × 10−2 24

−4 24
348.15 25 3.32 × 10
348.15 51 6.63 × 10−4 24
−3 24
348.15 101 1.32 × 10
S10
−3 24
348.15 203 2.60 × 10
348.15 405 5.04 × 10−3 24
−3 24
348.15 608 7.35 × 10
348.15 811 9.54 × 10−3 24
348.15 1013 1.17 × 10−2 24

−5 26
366.48 3.4 3.74 × 10
366.48 13.8 2.0 × 10−4 28
366.48 13.8 1.80 × 10−4 26

−4 28
366.48 27.6 3.7 × 10
366.48 31.0 4.26 × 10−4 26
−4 28
366.48 55.2 7.5 × 10
366.48 65.5 8.93 × 10−4 26
−3 28
366.48 110.3 1.50 × 10
366.48 137.9 1.840 × 10−3 26
373.15 21 3.01 × 10−4 29

−4 24
373.15 25 3.70 × 10
373.15 31 4.48 × 10−4 29
−4 29
373.15 42 5.94 × 10
373.15 42 5.17 × 10−4 30
373.15 51 7.30 × 10−4 24

−4 27
373.15 57.0 8.20 × 10
373.15 62 8.77 × 10−4 29
−4 30
373.15 62 7.61 × 10
373.15 82 1.15 × 10−3 29
373.15 82 1.02 × 10−3 30

−3 24
373.15 101 1.45 × 10
373.15 102 1.32 × 10−3 30
−3 27
373.15 120.9 1.75 × 10
373.15 153.7 2.23 × 10−3 27
373.15 203 2.85 × 10−3 24

−3 24
373.15 405 5.46 × 10
373.15 608 7.96 × 10−3 24
−2 24
373.15 811 1.03 × 10
373.15 1013 1.25 × 10−2 24
398.15 23 3.44 × 10−4 29

−4 29
398.15 33 5.04 × 10
398.15 43 6.54 × 10−4 29
−4 30
398.15 52 7.18 × 10
398.15 63 9.32 × 10−4 29
−3 30
398.15 82 1.13 × 10
398.15 83 1.21 × 10−3 29
398.15 87 1.21 × 10−3 30

−3 30
398.15 102 1.43 × 10
S11
−4 26
422.04 31.0 5.00 × 10
422.04 65.5 1.16 × 10−3 26
−3 26
422.04 103.4 1.88 × 10
423.15 21 3.26 × 10−4 29
423.15 30 4.95 × 10−4 29

−4 30
423.15 31 4.370 × 10
423.15 40 6.55 × 10−4 29
423.15 42 5.98 × 10−4 30

−4 30
423.15 52 7.79 × 10
423.15 51.8 9.00 × 10−4 27
−4 27
423.15 54.3 9.79 × 10
423.15 55 8.75 × 10−4 29
−4 30
423.15 62 9.43 × 10
423.15 76 1.15 × 10−3 29
423.15 75.9 1.35 × 10−3 27

−3 30
423.15 82 1.34 × 10
423.15 87.1 1.60 × 10−3 27
−4 29
448.15 22 3.28 × 10
448.15 29 4.83 × 10−4 29
448.15 39 6.59 × 10−4 29

−4 30
448.15 52 7.46 × 10
448.15 60 9.56 × 10−4 29
−4 30
448.15 62 9.43 × 10
448.15 80 1.24 × 10−3 29
473.15 26 3.30 × 10−4 29

−3 29
473.15 31 1.00 × 10
473.15 36 5.74 × 10−4 29
−4 30
473.15 42 5.52 × 10
473.15 46 8.44 × 10−4 29
473.15 52 7.51 × 10−4 30

−3 29
473.15 76 1.22 × 10
473.15 82 1.24 × 10−3 30
−3 30
473.15 102 1.40 × 10
473.15 118 1.62 × 10−3 30
477.59 27.6 3.5 × 10−4 28

−4 26
477.59 31.0 4.29 × 10
477.59 55.2 1.05 × 10−3 28
−3 26
477.59 65.5 1.46 × 10
477.59 103.4 2.57 × 10−3 26
477.59 110.3 2.71 × 10−3 28

−4 29
498.15 33 3.22 × 10
498.15 36 4.47 × 10−4 29
−4 29
498.15 41 5.91 × 10
S12
498.15 51 8.10 × 10−4 29
498.15 62 9.93 × 10−4 30
498.15 71 1.21 × 10−3 29
498.15 92 1.45 × 10−3 30
523.15 44 2.70 × 10−4 29
523.15 47 4.45 × 10−4 29
523.15 52 6.50 × 10−4 29
523.15 60 8.93 × 10−4 29
523.15 70 1.18 × 10−3 29
548.15 63 3.01 × 10−4 29
548.15 65 4.70 × 10−4 29
548.15 68 6.54 × 10−4 29
548.15 72 8.48 × 10−4 29
548.15 80 1.19 × 10−3 29
573.15 89 2.86 × 10−4 29
573.15 91 5.42 × 10−4 29
573.15 93 7.24 × 10−4 29
573.15 96 8.80 × 10−4 29
573.15 100 1.25 × 10−3 29
574.81 110.3 1.41 × 10−3 28
588.70 110.3 2.26 × 10−3 28
588.71 137.9 2.98 × 10−3 26
S13
Table S6: Experimental solubilities of water in H2 O − H2 mixtures (gas phase) at coexistence.
Experimental data are converted to mole fractions for different temperatures and pressures.
For the original units for each data set, see references below. For conversion to mole fractions,
standard conditions at T = 273.15 K and P = 1.01325 atm are considered, unless otherwise
mentioned in the reference.

−2 26
310.93 3.4 1.96 × 10
−3 26
310.93 13.8 4.88 × 10
310.93 31.0 2.22 × 10−3 26
310.93 65.5 1.16 × 10−3 26

−4 26
310.93 103.4 7.6 × 10
310.93 137.9 6.0 × 10−4 26
−3 31
310.95 13.8 4.88 × 10
310.95 31.0 2.22 × 10−3 31
310.95 65.5 1.16 × 10−3 31

−4 31
310.95 103.4 7.60 × 10
323.15 50 2.66 × 10−3 32
−3 32
323.15 100 1.48 × 10
323.15 101.3 1.38 × 10−3 33
323.15 150 1.08 × 10−3 32

−4 32
323.15 200 8.8 × 10
323.15 202.7 7.84 × 10−4 33
−4 32
323.15 250 7.6 × 10
323.15 300 6.8 × 10−4 32
323.15 405.3 4.64 × 10−4 33

−4 33
323.15 608.0 3.61 × 10
323.15 1013.3 2.80 × 10−4 33
−3 32
343.15 50 6.60 × 10
343.15 100 3.58 × 10−3 32
343.15 150 2.57 × 10−3 32

−3 32
343.15 200 2.07 × 10
343.15 250 1.77 × 10−3 32
−3 32
343.15 300 1.57 × 10
366.45 13.8 3.86 × 10−3 31
366.45 31.0 2.64 × 10−3 31

−3 31
366.45 65.5 1.32 × 10
366.45 137.9 6.68 × 10−3 31
366.48 3.4 2.47 × 10−1 26

−2 28
366.48 13.8 5.54 × 10
366.48 13.8 5.86 × 10−2 26
−2 28
366.48 27.6 3.08 × 10
366.48 31.0 2.64 × 10−2 26
−2 28
366.48 55.2 1.63 × 10
S14
T /[K] P /[bar] yH2 O Ref
−2 26
366.48 65.5 1.32 × 10
366.48 110.3 6.51 × 10−3 28
−3 26
366.48 137.9 6.68 × 10
422.04 31.0 1.570 × 10−1 26
422.04 65.5 7.53 × 10−2 26

−2 26
422.04 103.4 5.13 × 10
423.15 50 1.06 × 10−1 32
423.15 100 6.00 × 10−2 32

−2 32
423.15 150 4.40 × 10
423.15 200 3.60 × 10−2 32
−2 32
423.15 250 3.12 × 10
423.15 300 2.83 × 10−2 32
448.15 50 2.06 × 10−1 32

−1 32
448.15 100 1.21 × 10
448.15 150 9.27 × 10−2 32
−2 32
448.15 200 7.80 × 10
448.15 250 6.88 × 10−2 32
448.15 300 6.23 × 10−2 32

−1 32
473.15 50 2.66 × 10
473.15 100 1.48 × 10−1 32
−1 32
473.15 150 1.08 × 10
473.15 200 8.8 × 10−2 32
473.15 250 7.6 × 10−2 32

−2 32
473.15 300 6.9 × 10
477.59 27.6 6.34 × 10−1 28
−1 26
477.59 31.0 5.55 × 10
477.59 55.2 3.26 × 10−1 28
477.59 65.5 2.85 × 10−1 26

−1 26
477.59 103.4 1.88 × 10
477.59 110.3 1.72 × 10−1 28
−1 32
498.15 50 5.54 × 10
498.15 100 3.22 × 10−1 32
498.15 150 2.44 × 10−1 32

−1 32
498.15 200 2.08 × 10
498.15 250 1.83 × 10−1 32
−1 32
498.15 300 1.67 × 10
523.15 50 8.26 × 10−1 32
523.15 100 4.96 × 10−1 32

−1 32
523.15 150 3.83 × 10
523.15 200 3.27 × 10−1 32
−1 32
523.15 250 2.94 × 10
523.15 300 2.71 × 10−1 32
548.15 100 6.86 × 10−1 32
S15
T /[K] P /[bar] yH2 O Ref
548.15 150 5.36 × 10−1 32
548.15 200 4.61 × 10−1 32
548.15 250 4.16 × 10−1 32
548.15 300 3.86 × 10−1 32
573.15 100 9.00 × 10−1 32
573.15 150 7.22 × 10−1 32
573.15 200 6.33 × 10−1 32
573.15 250 5.82 × 10−1 32
573.15 300 5.51 × 10−1 32
574.81 110.3 8.45 × 10−1 28
588.7 110.3 9.67 × 10−1 28
588.71 137.9 8.10 × 10−1 26
S16
Table S7: Computed compositions of H2 O − H2 mixtures at coexistence using MC simulations.
xH2 is the mole fraction of hydrogen in the liquid phase, yH2 O is the mole fraction of water in
the gas phase. The H2 O − H2 mixture is defined by the TIP3P 9 -Marx 18 force fields. The
simulation techniques are described in the main text. σx is the uncertainty of x.
T /[K] P/[bar] xH2 /[−] σxH2 yH2 O /[−] σ y H2 O Sim. Tech.

−4 −7 −3 −4
283 10 1.510 × 10 9 × 10 1.8 × 10 1 × 10 CFCNPT
283 50 7.48 × 10−4 6 × 10−6 3.7 × 10−4 3 × 10−5 CFCNPT
283 80 1.174 × 10−3 8 × 10−6 2.4 × 10−4 2 × 10−5 CFCNPT
283 100 1.47 × 10−3 1 × 10−5 2.0 × 10−4 1 × 10−5 CFCNPT
283 300 4.28 × 10−3 4 × 10−5 7.3 × 10−5 5 × 10−6 CFCNPT
283 500 6.79 × 10−3 7 × 10−5 4.8 × 10−5 4 × 10−6 CFCNPT
283 800 9.51 × 10−3 9 × 10−5 3.5 × 10−5 2 × 10−6 CFCNPT
283 1000 1.162 × 10−2 9 × 10−5 2.9 × 10−5 4 × 10−6 CFCNPT
310 10 1.6 × 10−4 2 × 10−5 9.3 × 10−3 8 × 10−4 GE
310 50 8.1 × 10−4 8 × 10−5 2.3 × 10−3 4 × 10−4 GE
310 80 1.29 × 10−3 9 × 10−5 1.3 × 10−3 5 × 10−4 GE
310 100 1.6 × 10−3 1 × 10−4 1.2 × 10−3 3 × 10−4 GE
310 100 1.64 × 10−3 2 × 10−5 9.8 × 10−4 7 × 10−5 CFCNPT
310 300 4.59 × 10−3 9 × 10−5 3.5 × 10−4 4 × 10−5 CFCNPT
310 500 7.3 × 10−3 1 × 10−4 2.4 × 10−4 5 × 10−5 CFCNPT
310 800 1.08 × 10−2 2 × 10−4 1.7 × 10−4 4 × 10−5 CFCNPT
310 1000 1.31 × 10−2 2 × 10−4 1.4 × 10−4 5 × 10−5 CFCNPT
323 10 1.7 × 10−4 1 × 10−5 1.74 × 10−2 1 × 10−3 GE
323 50 8.8 × 10−4 7 × 10−5 3.57 × 10−3 7 × 10−4 GE
323 80 1.4 × 10−3 1 × 10−4 2.35 × 10−3 6 × 10−4 GE
323 100 1.7 × 10−3 1 × 10−4 2.00 × 10−3 4 × 10−4 GE
323 300 4.8 × 10−3 4 × 10−4 6.89 × 10−4 2 × 10−4 GE
323 500 7.6 × 10−3 6 × 10−4 4.45 × 10−4 2 × 10−4 GE
323 800 1.14 × 10−2 8 × 10−4 3.12 × 10−4 1 × 10−4 GE
323 1000 1.4 × 10−2 1 × 10−3 3.04 × 10−4 1 × 10−4 GE
323 100 1.72 × 10−3 2 × 10−5 1.8 × 10−3 1 × 10−4 CFCNPT
323 300 4.90 × 10−3 5 × 10−5 6.4 × 10−4 8 × 10−5 CFCNPT
323 500 7.7 × 10−3 1 × 10−4 4.2 × 10−4 7 × 10−5 CFCNPT
323 800 1.15 × 10−2 3 × 10−4 3.1 × 10−4 6 × 10−5 CFCNPT
323 1000 1.35 × 10−2 2 × 10−4 2.6 × 10−4 6 × 10−5 CFCNPT
366 10 2.1 × 10−4 1 × 10−5 1.08 × 10−1 7 × 10−3 GE
366 50 1.16 × 10−3 5 × 10−5 2.2 × 10−2 2 × 10−3 GE
366 80 1.86 × 10−3 8 × 10−5 1.41 × 10−1 9 × 10−4 GE
366 100 2.28 × 10−3 8 × 10−5 1.14 × 10−1 6 × 10−4 GE
366 300 6.4 × 10−3 3 × 10−4 4.2 × 10−1 5 × 10−4 GE
366 500 1.03 × 10−2 6 × 10−4 2.7 × 10−1 4 × 10−4 GE
366 800 1.49 × 10−2 7 × 10−4 1.8 × 10−1 4 × 10−4 GE
S17
366 1000 1.80 × 10−2 9 × 10−4 1.4 × 10−1 4 × 10−4 GE
423 50 1.70 × 10−3 5 × 10−5 1.38 × 10−1 6 × 10−3 GE
423 80 2.85 × 10−3 9 × 10−5 8.6 × 10−2 4 × 10−3 GE
423 100 3.5 × 10−3 1 × 10−4 7.2 × 10−2 3 × 10−3 GE
423 300 1.03 × 10−2 4 × 10−4 2.5 × 10−2 1 × 10−3 GE
423 500 1.62 × 10−2 6 × 10−4 1.58 × 10−2 8 × 10−4 GE
423 800 2.35 × 10−2 9 × 10−4 1.02 × 10−2 5 × 10−4 GE
423 1000 2.8 × 10−2 1 × 10−3 8.2 × 10−3 6 × 10−4 GE
S18
the gas phase. The H2 O − H2 mixture is defined by the TIP4P/2005 12 -Marx 18 force fields.
σx is the uncertainty of x.

323 10 1.2 × 10−4 1 × 10−5 3.6 × 10−3 6 × 10−4 GE
323 50 5.9 × 10−4 9 × 10−5 9 × 10−4 5 × 10−4 GE
323 80 9.1 × 10−4 9 × 10−5 6 × 10−4 4 × 10−4 GE
323 100 1.2 × 10−3 1 × 10−4 5 × 10−4 4 × 10−4 GE
323 300 3.3 × 10−3 4 × 10−4 2 × 10−4 1 × 10−4 GE
323 500 5.3 × 10−3 5 × 10−4 1 × 10−4 1 × 10−4 GE
323 800 8.0 × 10−3 1 × 10−3 1 × 10−4 1 × 10−4 GE
323 1000 1.0 × 10−4 1 × 10−3 8 × 10−5 1 × 10−4 GE
366 10 1.3 × 10−4 1 × 10−5 2.8 × 10−2 2 × 10−3 GE
366 50 6.6 × 10−4 4 × 10−5 5.8 × 10−3 5 × 10−4 GE
366 80 1.0 × 10−3 6 × 10−5 3.8 × 10−3 5 × 10−4 GE
366 100 1.27 × 10−3 8 × 10−5 3.3 × 10−3 5 × 10−4 GE
366 300 3.6 × 10−3 3 × 10−4 1.2 × 10−3 5 × 10−4 GE
366 500 5.9 × 10−3 6 × 10−4 9 × 10−4 4 × 10−4 GE
366 800 8.7 × 10−3 8 × 10−4 5 × 10−4 3 × 10−4 GE
366 1000 1.02 × 10−2 8 × 10−4 4 × 10−4 2 × 10−4 GE
423 50 8.4 × 10−4 4 × 10−5 4.5 × 10−2 2 × 10−3 GE
423 80 1.37 × 10−3 6 × 10−5 2.8 × 10−2 2 × 10−3 GE
423 100 1.73 × 10−3 8 × 10−5 2.3 × 10−2 1 × 10−3 GE
423 300 4.9 × 10−3 2 × 10−4 8.0 × 10−3 6 × 10−4 GE
423 500 7.8 × 10−3 4 × 10−4 5.0 × 10−3 5 × 10−4 GE
423 800 1.18 × 10−2 5 × 10−4 3.3 × 10−3 4 × 10−4 GE
423 1000 1.41 × 10−2 8 × 10−4 2.8 × 10−3 4 × 10−4 GE
S19
the gas phase. The H2 O − H2 mixture is defined by the TIP5P/Ew 5 -Marx 18 force fields. σx
is the uncertainty of x.
T /[K] P/[bar] xH2 /[−] σxH2 yH2 O /[−] σyH2 O Sim. Tech.
323 10 1.67 × 10−4 8 × 10−6 2.9 × 10−2 2 × 10−3 GE
323 50 8.5 × 10−4 6 × 10−5 6 × 10−3 1 × 10−3 GE
323 100 1.7 × 10−3 1 × 10−4 3.0 × 10−3 6 × 10−4 GE
323 300 4.9 × 10−3 4 × 10−4 1.4 × 10−3 5 × 10−4 GE
323 500 7.9 × 10−3 8 × 10−4 8 × 10−4 4 × 10−4 GE
323 800 1.2 × 10−2 1 × 10−3 6 × 10−4 3 × 10−4 GE
366 10 1.9 × 10−4 1 × 10−5 1.75 × 10−1 7 × 10−3 GE
366 50 2.34 × 10−3 7 × 10−5 2.27 × 10−2 9 × 10−4 GE
366 100 2.12 × 10−3 9 × 10−5 1.9 × 10−2 1 × 10−3 GE
366 300 6.1 × 10−3 3 × 10−4 6.8 × 10−3 5 × 10−4 GE
366 500 9.7 × 10−3 5 × 10−4 4.4 × 10−3 4 × 10−4 GE
366 800 1.42 × 10−2 5 × 10−4 2.9 × 10−3 4 × 10−4 GE
366 1000 1.73 × 10−2 9 × 10−4 2.4 × 10−3 4 × 10−4 GE
423 50 1.64 × 10−3 6 × 10−5 2.2 × 10−1 1 × 10−2 GE
423 100 3.6 × 10−3 1 × 10−4 1.14 × 10−1 5 × 10−3 GE
423 300 1.05 × 10−2 3 × 10−4 4.1 × 10−2 2 × 10−3 GE
423 500 1.65 × 10−2 5 × 10−4 2.6 × 10−2 1 × 10−3 GE
423 800 2.40 × 10−2 8 × 10−4 1.7 × 10−2 1 × 10−3 GE
423 1000 2.8 × 10−2 1 × 10−3 1.41 × 10−2 7 × 10−4 GE
S20
Table S10: Parameters used for equation of state modeling: critical temperatures (Tc ),
pressures (Pc ), and acentric factors (ω) the components at the standard reference state (1
bar). 34–36
Component Tc /[K] Pc /[Pa] ω

H2 33.14 1296400 -0.219
H2 O 647.1 22064000 0.3443
S21
References
(1) Poursaeidesfahani, A.; Hens, R.; Rahbari, A.; Ramdin, M.; Dubbeldam, D.; Vlugt, T.
J. H. Efficient application of Continuous Fractional Component Monte Carlo in the
Reaction Ensemble. J. Chem. Theory Comput. 2017, 13, 4452–4466.
(2) Poursaeidesfahani, A.; Torres-Knoop, A.; Dubbeldam, D.; Vlugt, T. J. H. Direct Free
Energy Calculation in the Continuous Fractional Component Gibbs Ensemble. J. Chem.
Theory Comput. 2016, 12, 1481–1490.
(3) McQuarrie, D. A.; Simon, J. D. Physical Chemistry: A Molecular Approach, 1st ed.;
University Science Books: Sausalito, California, 1997.
(4) Walas, S. M. Phase Equilibria in Chemical Engineering, 1st ed.; Butterworth-Heinemann:

USA, 1985.
(5) Rick, S. W. A reoptimization of the five-site water potential (TIP5P) for use with Ewald
sums. J. Comput. Phys. 2004, 120, 6085–6093.
(6) Allen, M. P.; Tildesley, D. J. Computer simulation of liquids, 2nd ed.; Oxford university
press: Oxford, United Kingdom, 2017.
(7) Frenkel, D.; Smit, B. Understanding molecular simulation: from algorithms to applica-
tions, 2nd ed.; Academic Press: San Diego, California, 2002.
(8) Mark, P.; Nilsson, L. Structure and dynamics of the TIP3P, SPC, and SPC/E water
models at 298 K. J. Phys. Chem. A 2001, 105, 9954–9960.
Comparison of simple potential functions for simulating liquid water. J. Chem. Phys.
1983, 79, 926–935.
(10) Berendsen, H. J. C.; Grigera, J. R.; Straatsma, T. P. The missing term in effective pair
potentials. J. Phys. Chem. 1987, 91, 6269–6271.
S22
(11) Izadi, S.; Anandakrishnan, R.; Onufriev, A. V. Building Water Models: a different
approach. J. Phys. Chem. Lett. 2014, 5, 3863–3871.
(12) Abascal, J. L. F.; Vega, C. A general purpose model for the condensed phases of water:
TIP4P/2005. J. Chem. Phys. 2005, 123, 234505.
simulations: TIP4P-EW. J. Chem. Phys. 2004, 120, 9665–9678.
(14) Buch, V. Path integral simulations of mixed para-D2 and ortho-D2 clusters: The
orientational effects. J. Chem. Phys. 1994, 100, 7610–7629.
(15) Hirschfelder, C.; Curtiss, F.; Bird, R. B. Molecular Theory of Gases and Liquids; Wiley:
New York, 1954.
(16) Köster, A.; Thol, M.; Vrabec, J. Molecular models for the hydrogen age: hydrogen,
nitrogen, oxygen, argon, and water. J. Chem. Eng. Data 2018, 63, 305–320.
(17) Cracknell, R. F. Molecular simulation of hydrogen adsorption in graphitic nanofibres.

Phys. Chem. Chem. Phys. 2001, 3, 2091–2097.
(18) Marx, D.; Nielaba, P. Path-integral Monte Carlo techniques for rotational motion in
two dimensions: Quenched, annealed, and no-spin quantum-statistical averages. Phys.
Rev. A 1992, 45, 8968–8971.
(19) Lemmon, E. W.; Span, R. Short fundamental equations of state for 20 industrial fluids.
J. Chem. Eng. Data 2006, 51, 785–850.
(20) Lemmon, E. W.; Huber, M. L.; McLinden, M. O. NIST reference fluid

thermodynamic and transport properties–REFPROP. NIST standard ref-
erence database 2002, 23, v7, https://www.nist.gov/publications/
nist-standard-reference-database-23-reference-fluid-thermodynamic-and-transport.
S23
(21) Peng, D.-Y.; Robinson, D. B. A new two-constant equation of state. Ind. Eng. Chem.
Fundam. 1976, 15, 59–64.
(22) Soave, G. Equilibrium constants from a modified Redlich-Kwong equation of state.

Chem. Eng. Sci. 1972, 27, 1197 – 1203.
(23) Wagner, W.; Pruß, A. The IAPWS formulation 1995 for the thermodynamic properties
of ordinary water substance for general and scientific use. J. Phys. Chem. Ref. Data
2002, 31, 387–535.
(24) Wiebe, R.; Gaddy, V. L. The solubility of hydrogen in water at 0, 50, 75 and 100Âř
from 25 to 1000 atmospheres. 2005, 56, 76–79.
(25) Meyer, M.; Tebbe, U.; Piiper, J. Solubility of inert gases in dog blood and skeletal
muscle. Pflügers Archiv European Journal of Physiology 1980, 384, 131–134.
(26) Gillespie, P. C.; Wilson, G. M. Vapor-Liquid Equilibrium Data on Water-Substitue Gas

Components; 1980; pp RR_41, 1–34.
(27) Kling, G.; Maurer, G. The solubility of hydrogen in water and in 2-aminoethanol
at temperatures between 323 K and 423 K and pressures up to 16 M Pa. Chem.
Thermodynamics 1991, 23, 531–541.
(28) Devaney, W.; Berryman, J. M.; Kao, P.-L.; Eakin, B. High Temperature V-L-E Mea-
surements for Substitute Gas Components; 1978; pp 1–27.
(29) Jung, J. Loslichkeit von Kohlenmonoxid und Wasserstoff in Wasser zwisen 0 C und 300
C. Ph.D. thesis, RWTH Aachen, 1962.
(30) Ipatev, V.; Teodorovich, V. Equilibrium compositions of vapor-gas mixtures over solu-
tions. Zh.Obshch.Khim. 1934, 4, 395–399.
(31) Ugrozov, V. V. Equilibrium compositions of vapor-gas mixtures over solutions. 1996,

70, 1240–1241.
S24
(32) Maslennikova, V. Y.; Goryunova, N.; Subbotina, L.; Tsiklis, D. The Solubility of Water
in Compressed Hydrogen. Russian Journal of Physical Chemistry 1976, 50, 240–243.
(33) Bartlett, E. P. The concentration of water vapor in compressed hydrogen, Nitrogen and
a mixture of these gases in the presence of condensed water. Journal of the American
Chemical Society 1927, 49, 65–78.
(34) Yaws, C. L. Thermophysical properties of chemicals and hydrocarbons, 2nd ed.; Gulf
Professional Publishing: Oxford, UK, 2014.
(35) Chase, M. W.; Curnutt, J.; Prophet, H.; McDonald, R.; Syverud, A. JANAF thermo-
chemical tables, 1975 supplement. J. Phys. Chem. Ref. Data. 1975, 4, 1–176.
(36) Chase, M. W. NIST-JANAF Themochemical Tables, Fourth Edition. J. Phys. Chem.

Ref. Data 1998, 4, 1–1951.
S25

Master Thesis Jeroen Brenkman Repository

Uploaded by

Copyright:

Available Formats

Master Thesis Jeroen Brenkman Repository

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Master Thesis Jeroen Brenkman Repository

Uploaded by

Copyright:

Available Formats

High Pressure

in partial fulfillment of the requirements for the degree of

at the Delft University of Technology,

Student number: 4169093

An electronic version of this thesis is available at http://repository.tudelft.nl/.

6 Results & Discussion 35

1.1. R ESEARCH OBJECTIVES

I ONIC L IQUIDS AS DRYING AGENT

• low energy use

• low hydrogen loss

• low Capital Expenditure (CAPEX) & Operational Expenditure (OPEX)

• low safety & environmental risks

2.1. O VERVIEW OF POSSIBLE DRYING METHODS

2.1.1. M EMBRANE SEPARATION

I MMOBILIZED L IQUID M EMBRANES

C ARBON M OLECULAR S IEVE M EMBRANES

1. Temperature Swing Adsorption (TSA).

2. Pressure Swing Adsorption (PSA).

3. Inert Purge Stripping (without changing T or P).

2.2. M ETHOD SELECTION

Table 2.1: Vapor pressures of several liquids.

3.2. E XPERIMENTAL D ATA

3.3. T HE G AMMA /P HI F ORMULATION OF VLE

• The liquid phase is an ideal solution.

fˆVi = fˆLi , (3.1)

fˆVi = Φ̂i y i P. (3.3)

where γi is the activity coefficient of i in the solution, Φsat

Figure 3.3: Poynting Factor, as a function of pressure, at 50°C.

Eq. (3.1) gives the γ − φ formulation:

The detailed derivation of the γ − φ formulation is described in [Smith et al.]. P sat

3.4. C UBIC E QUATIONS OF S TATE (E O S) FOR VLE CALCULATIONS

3.4.1. D ENSITY C ALCULATIONS

α(T r ) = (1 + (0.37464 + 1.54226ω − 0.26992ω2 )(1 − T r 1/2 ))2 (3.12)

S OAVE ’ S M ODIFIED R EDLICH K WONG E O S

α(Tr ) = (1 + (0.480 + 1.574ω − 0.176ω2 )(1 − Tr1/2 ))2 (3.16)

S HELL M ODIFIED I MPROVED R EDLICH -K WONG (SMIRK) E O S

Y = y 1 + y 2 ln[1 + exp(y 3 (X − y 4 ))] − y 5 ln[1 + exp(−y 3 (X − y 4 ))] (3.17)

3.4.2. F UGACITY C OEFFICIENT CALCULATIONS

a i j = (1 − K i j )(a i a j )1/2 (3.20)

S OAVE ’ S R EDLICH K WONG E O S

F UGACITY C OEFFICIENT R ESULTS

150°C Experimental data 70°C

150 93°C 38°C

150 93°C 38°C

3.5. M OLECULAR S IMULATION

3.5.1. M OLECULAR S IMULATION R ESULTS

(a) First kind (b) Second kind

"An Ionic Liquid is a salt with a melting point below 100°C".

4.2. I ONIC L IQUID S ELECTION

2. Negligible vapor pressure

4.2.1. (M ODIFIED ) UNIFAC

4.2.3. M OLECULAR S IMULATION

4.2.4. L ITERATURE AND AVAILIBILITY

(a) Parameters: ADFcombi2005 (b) Parameters: ADFcombi2005(no H-bond restrictions)

mol fraction H2O (mol/mol) mol fraction H2O (mol/mol)

mol fraction H2O (mol/mol) mol fraction H2O (mol/mol)

P REVIOUS GAS DEHYDRATION STUDIES

D ORTMUND D ATA B ANK (DDBST )

Abbreviation Full Name Ref

AVAILIBILITY AND S AFETY