Saturated Phase Densities of CO2 + H2O at Temperatures from (293 to

450) K and Pressures up to 64 MPa

Emmanuel C. Efika1, Rayane Hoballah1, Xuesong Li1, Eric F. May2,*, Manuela Nania1, Yolanda Sanchez-
Vicente1, and J. P. Martin Trusler1

1. Qatar Carbonates and Carbon Storage Research Centre, Department of Chemical

Engineering, Imperial College London, South Kensington Campus, London SW7 2AZ, U.K.

2. Centre for Energy, School of Mechanical & Chemical Engineering, The University of Western
Australia, Crawley WA 6009, Australia

17 June 2015

Abstract
An apparatus consisting of an equilibrium cell connected to two vibrating tube densimeters and two
syringe pumps was used to measure the saturated phase densities of CO2 + H2O at temperatures from
(293 to 450) K and pressures up to 64 MPa, with estimated average standard uncertainties of
1.5 kg·m-3 for the CO2-rich phase and 1.0 kg·m-3 for the aqueous phase. The densimeters were housed
in the same thermostat as the equilibrium cell and were calibrated in-situ using pure water, CO2 and
helium. Following mixing, samples of each saturated phase were displaced sequentially at constant
pressure from the equilibrium cell into the vibrating tube densimeters connected to the top (CO2-rich
phase) and bottom (aqueous phase) of the cell. The aqueous phase densities are predicted to within
3 kg·m-3 using empirical models for the phase compositions and partial molar volumes of each
component. However, a recently developed multi-parameter equation of state (EOS) for this binary
mixture [Gernert and Span, J. Chem. Thermodyn. (2015, submitted)] was found to under predict the
measured aqueous phase density by up to 13 kg·m-3. The density of the CO2-rich phase was always
within about 8 kg·m-3 of the density for pure CO2 at the same pressure and temperature; the
differences were most positive near the critical density, and became negative at temperatures above
about 373 K and pressures below about 10 MPa. For this phase, the multi-parameter EOS of Gernert
and Span describes the measured densities to within 5 kg·m-3, whereas the computationally-efficient
cubic EOS model of Spycher and Pruess [Transport in Porous Media 2010, 82, 173-196], commonly
used in simulations of subsurface CO2 sequestration, deviates from the experimental data by a
maximum of about 3 kg·m-3.

*
Corresponding author: [email protected]

1
Introduction
The amount of CO2 that can be stored in saline aquifers depends on the saturated phase densities of
mixtures of CO2 + brines at reservoir conditions. The estimation of solubility trapping capacities
requires knowledge of the aqueous phase density as a function of CO2 concentration. Convective
transport driven through the aquifer by the difference in density between the nascent brine and the
layer saturated with CO2 needs to be modelled accurately if total storage capacity is to be estimated
reliably [1]. Furthermore, estimates of residual (or capillary) trapping mechanisms require robust
models for interfacial tension (IFT), and standard methods of measuring IFT require knowledge of the
difference in the densities of both the CO2-rich and aqueous phases [2; 3].

Thermodynamic models used to estimate the saturated phase densities of the CO2 + brine mixtures
likely to be encountered in carbon sequestration scenarios involving saline aquifers can only be as
accurate as the quality of the experimental data used in their development. Haugan et al. [4] and Jones
et al. [5] stated that, ideally, the relative uncertainty of the measurements should approach 0.1 %.
This is a challenging specification given the extreme nature of the conditions under which the
measurements are needed. Achieving this measurement goal requires that the densimetry be both
sensitive and robust, and that the experimental system used be able to prepare and manipulate a
mixture without inadvertent composition changes occurring. Here we report an apparatus developed
for measuring densities (but not compositions) of the aqueous and CO2-rich phases of CO2 + brine
mixtures at saturation. In this work, we demonstrate the use of this apparatus through measurements
of the baseline CO2 + H2O system: while the aqueous phase of this mixture has been reasonably well-
studied [6; 7; 8; 9; 10; 11; 12] there are inconsistencies on the order of 1 % between some of these
data sets and most of the data are restricted to temperatures below 333 K and to pressures below 30
MPa. Furthermore, there are very few experimental density data available for the CO2-rich phase [9;
10], particularly at the requisite high pressure, high temperature conditions. A comprehensive
measurement of the CO2-rich phase density is particularly useful because the data produced will be
common to all CO2 + brine systems; one just needs to account for the change in the H2O content of
the CO2-rich phase caused by the presence of the salt in the aqueous phase.

In the following sections we discuss some existing models available for the prediction of the saturated
phase densities in CO2 + H2O and/or CO2 + brine systems. We then describe the experimental
apparatus used to measure the saturated densities of both phases over a wider range of conditions
and/or with lower uncertainty than achieved previously. Subsequently, the results of the
measurements are presented and compared with the densities of pure water and pure CO2 at the
same temperature and pressure. Finally the results are compared with data existing in the literature
for this binary system, as well as with predictions made using three models developed for predicting
the saturated phase densities of the aqueous and/or the CO2-rich phases of CO2 + H2O.

Existing Models for Predicting Saturated Phase Densities

To calculate the density of a saturated phase, it is first necessary to obtain the phase equilibria of the
mixture at the specified temperature and pressure. While there are several models in the literature
available for calculating the phase equilibrium for aqueous mixtures with CO2 [13] the correlation of
Duan et al. [14] for the solubility of CO2 in brines is one of the most widely used. This correlation was
developed using aqueous phase solubility data measured over the temperature range (274 to 523) K
at pressures to 200 MPa, and salt concentrations up to 4.5 mol⋅kg-1. In this model, the water content
of the CO2-rich phase is simply calculated by applying Raoult’s law to the water activity calculated for

2
the brine. Recently, however, Hou et al. [15] measured new vapour-liquid equilibrium (VLE) data for
the CO2 + H2O system over the temperature range (298 to 448) K at pressures to 18 MPa, and used
them to test predictions of the Duan et al. [14] correlation. Those data were represented well at
T ≤ 423 K but the correlation of Duan et al. [14] was found to over-predict the CO2 solubility at
T = 448 K by about 25 %, which is at least three times larger than its claimed uncertainty.

To facilitate sufficiently accurate calculations of phase partitioning for use in simulations of geological
CO2 sequestration, Spycher et al. [16; 17; 18] developed a simplified but computationally efficient
means of calculating the phase compositions of CO2 + H2O and CO2 + brine mixtures (with molalities
up to 6 mol⋅kg-1) at pressures up to 60 MPa. The impact of the non-ideality of the CO2-rich phase on
the phase equilibrium was accounted for by using the Redlich-Kwong [19] equation of state (EOS).
Several simplifying assumptions were implemented to avoid the need for iterative calculations and
any degradation in the performance of water-CO2 flow models for which the solubility calculations
were needed. At temperatures between (285 and 372) K [16; 17] the volume and the partial fugacity
coefficient of the CO2-rich phase were approximated to be that of pure CO2, and the water activity
was assumed to be equal to the mole fraction of water in the aqueous phase. At temperatures
between (382 and 573) K, these assumptions do not allow sufficient accuracy and a Margules activity
coefficient model was used to describe the aqueous phase and asymmetric binary interaction
parameters that were used in the mixing rule for the Redlich-Kwong EOS to describe the CO2-rich
phase [18]. At temperatures between (372 and 382) K, this extended model [18] gives a weighted
value of the properties calculated using the ideal (low temperature) and non-ideal (high-temperature)
sets of correlations.

The models available for estimating the density of the aqueous phase generally calculate the partial
molar volume V2 of the CO2 solute as a function of pressure, temperature, and brine molality; it is
assumed to be independent of mole fraction x of CO2. Treating the solvent as a pseudo-pure-
component, the molar volume Vm of the solution is then obtained as follows:

Vm =(1 − x )V1 + xV2 . (1)

Here, V1, the partial molar volume of the solvent, is also assumed to be independent of the CO2
concentration, and is calculated using an auxiliary model such as the equation of state of Wagner and
Pruss [20] for pure water, or the model of Al-Ghafri et al. [21] for brines. Finally, mass density ρ of the
aqueous solution containing CO2 is obtained as follows

(1 − x ) M 1 + xM 2
ρ= , (2)
(1 − x )V1 + xV2

where M1 is the molar mass of the solvent and M2 is the molar mass of the solute (CO2). Sedlbauer et
al. [22] presented an empirical expression for predicting the partial molar volume of various solutes
(including CO2) in pure water based on Fluctuation Solution Theory. In the case of nonelectrolyte
solutes, such as CO2, the model was tested over the temperature range (303 to 725) K at pressures to
35 MPa. Subsequently Duan et al. [23] and then Li et al. [24] presented alternative empirical
correlations for the partial molar volume of CO2 in both H2O + CO2 and NaCl(aq) + CO2 mixtures at
temperatures between (273 and 573) K and pressures to 100 MPa. Most recently, McBride-Wright et
al. [25; 26] reported measurements of the density of under-saturated aqueous CO2 + H2O mixtures

3
using a vibrating-tube apparatus at pressures up to 100 MPa, over the temperature range
(274 to 449) K, together with an empirical correlation for the partial molar volume of CO2 that was
able to represent his density data to within ±0.04 %. Partial-molar volumes predicted using the
correlations of Li et al. [24] and McBride-Wright et al. [25; 26] differed by up to 7 %, whereas partial
molar volumes calculated using the correlation of Sedlbauer et al. [22] had a maximum relative
deviation of only 2.5 % from the model of McBride-Wright et al. [25; 26].

The density of the CO2-rich phase in aqueous mixtures is often approximated by that of pure CO2 at
the same pressure and temperature [7]. This approximation is reasonable because the solubility of
water in the CO2 phase is much lower than that of CO2 in an aqueous phase, and was explicitly part of
the approach taken by Spycher et al. at temperatures below 373 K [16; 17]. At sufficiently low
densities, the ideal gas law implies that the mass density of the CO2 rich phase will be less than that of
pure CO2 because the molar mass of the mixture will always be less than the molar mass of pure CO2.
However, while this behaviour should be observable in the ideal gas limit, it is not clear under what
conditions the impact of molecular interactions will become significant.

Duan et al. [27; 28] attempted to account for those molecular interactions by modelling both pure CO2
and the CO2-rich phase of mixtures using a Benedict-Webb-Rubin [29] (BWR) type equations of state.
However, this work was done prior to the development of the reference EOS for CO2 by Span and
Wagner [30], which is considerably more accurate than the BWR models. Spycher et al. [16; 17; 18]
regressed their RK EOS to compressibility factors for pure CO2 calculated with the EOS of Span and
Wagner [30]. The binary interaction parameters within the RK EOS mixing rule used at temperatures
above 372 K were fit by Spycher et al. to phase equilibrium data for the CO2 + H2O system [18]. In the
process of calculating the fugacity coefficients using this tuned RK EOS, the Spycher et al. [16; 17; 18]
model also calculates the density of the CO2-rich phase.

However, cubic EOS are known to give poor representations of density, especially when they are
intended primarily for the description of phase equilibrium. Multi-parameter equations of state, such
as those of Span and Wagner [30] for CO2 and Wagner and Pruss [20] for H2O, are developed to
describe all thermodynamic properties for a system as accurately as possible and are increasingly
being applied successfully to the description of mixtures as in, for example, the GERG-2008 EOS for
natural gas [31]. Recently, Gernert and Span [32] have developed a multi-parameter EOS for the
Helmholtz energy optimized for the CO2 + H2O system as well as other humid or CO2-rich mixtures,
referred to here as the “EOS-CG”. To describe CO2 + H2O mixtures, the EOS-CG utilises the reference
EOS for water [20] and CO2 [30] together with a re-parameterised departure function with the same
functional form as that used in the GERG-2008 EOS for natural gas [31].

Apparatus and Method

A schematic and photographs of the apparatus used in this work are shown in Figure 1. A Hastelloy
pressure vessel (C) with an internal volume of 50 mL, containing a PTFE-coated magnetic stirrer bar
and sealed with a PTFE O-ring, was housed inside a fan-forced air bath that had a long-term stability
and spatial uniformity of ±0.2 K and an operating range of (293 to 473) K. To permit operation at
temperatures below about 308 K, a heat exchanger (HEX) was installed within the air bath through
which chilled water was circulated from an external circulating chiller (CC). A permanent magnet
mounted on a motor shaft below the pressure vessel was used to rotate the stirrer inside that cell at
900 r.p.m.; as discussed below, this degree of agitation was found to be sufficient for attaining

4
equilibrium within approximately 10 minutes. Hastelloy tubes, each of 6.35 mm o.d. and internal
volume < 1 mL, connected the top and bottom ports of the pressure vessel to one port of the
corresponding vibrating U-tube densimeter (VTD) (Anton Paar, DMA-1400), that was also constructed
from Hastelloy. These densimeters had an internal volume of less than 2 mL and a maximum operating
pressure and temperature of 140 MPa and 473 K, respectively.

(a)

(b) (c)

Figure 1. (a) Schematic of the apparatus: C, equilibrium vessel; VTD-A, VTD-B, vibrating-tube
densimeters; SP-A, SP-B, syringe pumps; F1, F2, filters; CV, Check valve; V2, V3, 5-port valves; V4,V5,
pressure relief valves; V1,V6,V7,V8,V9, isolation valves; BD; bursting disk safety devices; P1,P2,
pressure transducers; T1,T2,T3, temperature sensors; HEX, heat exchanger; CC, circulating chiller.
(b) Photograph showing the equilibrium cell and the two densimeters housed inside the air bath.
(c) Photograph showing the air bath, syringe pumps, densimeter readouts, and ancillary equipment.

The other port of each densimeter was connected via about 6 mL of coiled Hastelloy tubing to the
delivery valve (V8 or V9) of a syringe pump (SP) (Quizix model C-5000-10K) located outside the air
bath. The cylinder of each syringe pump was also made of Hastelloy and had a volume at maximum
extension of 10 mL. Tubing and fittings made from stainless steel or PTFE were used for all other
elements of the manifold used for transferring high- or low-pressure fluids, respectively, into the
pump cylinders. Syringe pump SP-A connected to VTD-A, the upper densimeter, could be filled from a
cylinder containing liquid CO2 at its saturation pressure of about 5.7 MPa via selector valve V2 and fill
valve V7. A circulating chiller (not shown in Figure 1) was used to regulate the temperature of SP-A at
T = 283 K to ensure any CO2 therein remained in the liquid phase. This allowed a CO2 sample loaded

5
into SP-A to be compressed up to 64 MPa when needed. The aqueous phase (de-ionised water in this
work) was stored in a bottle at ambient pressure. Immediately prior to use, degassing was carried out
by agitation under vacuum, the pressure in the sample bottle was restored to ambient, and the sample
was drawn by suction from below the liquid surface into syringe pump SP-B via selector valve V3 and
fill valve V6.

Periodically, leakage past the main PTFE O-ring seal was found to be a problem and this component
then had to be replaced. This problem was attributed to swelling and subsequent decompression
damage caused by dissolution of CO2 in the polymer. All experimental data reported in this work
pertain to leak-free operation of the equipment.

The system pressure was measured using the two strain-gauge type transducers (P1 and P2) (Sensata
Technology, model 101HP2), one connected to each pump cylinder. By closing either of the delivery
valves V8 or V9 it was possible to isolate each transducer from the other and/or from the cell. Bursting
disks (BD), located in each pump and also on the equilibrium cell, were used to protect against over-
pressure in the high-pressure circuit, while proportional relief valves (V4 and V5) were used to protect
the filling lines to the syringe pumps. The two pressure transducers were calibrated at pressures up to
68 MPa against a reference quartz-crystal pressure gauge (Fluke, model PPCH-G-70M) with a relative
uncertainty of 0.02 % of full scale (i.e. 14 kPa). The strain-gauge transducers exhibited a hysteresis of
about ±30 kPa during the calibration but the mean of the pairs of readings taken with increasing and
decreasing pressure were linear to with about 5 kPa. During the measurement phase, the consistency
of the pressure transducer calibrations were checked by comparing the two transducers against each
other when the cell contained only a single phase fluid, and by monitoring their reading when under
vacuum. Periodically, the vacuum reading of the transducer in the lower cylinder (B) was observed to
shift by a small amount (up to 50 kPa). However, when a new offset correction was applied to this
transducer its reading remained consistent with that of the transducer in Cylinder A over the full
pressure range, indicating the sensitivities of the two gauges were stable. The overall standard
uncertainty of the measured pressures (including non-linearity, hysteresis, calibration uncertainty and
drift) was estimated to be 50 kPa.

Three 100 Ω platinum resistance thermometers (T1, T2, T3) (PRT) were used together with a data
acquisition unit/digital multimeter (Agilent, model 34970A) to measure the temperatures of the upper
and lower VTD and the equilibrium cell. These were calibrated on ITS-90 over the temperature range
(273 to 473) K using a triple-point-of-water cell and a standard 25 Ω PRT immersed in a stirred oil bath.
The measured resistances of the 100 Ω PRTs were regressed to the ITS-90 temperatures using the
Callander Van Dusen equation [33]. The resulting standard uncertainty of the thermometer calibration
was 0.02 K. However, the spatial variation of temperature within the stirred air bath meant that the
standard uncertainty in the temperature of the fluid phases contained within a given element was
estimated to be 0.2 K.

To ensure that liquid water did not condense in the upper densimeter, a resistive heating element
(not shown in Figure 1) was attached to the outside of the VTD housing (but inside the massive block
used for vibration isolation). Approximately 5 W of DC power was used to raise the temperature of
the VTD-A by around 2 K relative to the temperature of the equilibrium cell, as measured by the PRT
T1 mounted in the bore of the VTD block between the U tube’s inlet and outlet. This temperature shift
was consistent with the independent reading from the densimeter’s in-built temperature sensor,

6
which was monitored by the data acquisition unit supplied by the VTD manufacturer to measure the
resonance period of the vibrating tube. The temperatures of the equilibrium cell and the lower VTD,
as measured by the two 100 Ω PRTs, were consistent within ±0.2 K. The temperatures reported here
are those of the two VTDs: phase equilibrium was achieved at the temperature of the lower VTD and
this was the temperature at which the aqueous phase density was measured. However, it was
necessary to measure the density of the saturated CO2 phase at a slightly super-heated temperature.

To load a sample, the equilibrium cell, both densimeters and both syringe pumps were evacuated and
flushed with CO2 at pressures in the range (1 to 3) MPa several times, before V-9 was closed and SP-B
was evacuated. Approximately 10 mL of water, freshly degassed by agitation under vacuum, was
loaded into SP-B before the fill valve V6 was closed. With V6 and V9 both closed, the aqueous sample
in SP-B was compressed to a pressure greater than that in the equilibrium cell before the delivery
valve V9 was opened. The aqueous phase was then slowly injected through the VTD-B and into the
equilibrium cell. Once the pump reached its minimum volume, the delivery valve V9 was closed and
the cylinder was re-filled with a new 10 mL batch of aqueous phase. This procedure was repeated until
the volume of aqueous phase in the equilibrium cell was approximately 25 mL. Further increases or
decreases in system pressure were achieved by injecting or venting CO2 through the upper syringe
pump, SP-A. Following a change in pressure, it was found that approximately 10 min was required to
reach equilibrium under stirring as determined by monitoring the temperature and the pressure
(when the cell was isolated with V8 and V9 both closed) and/or the displacement of the piston in SP-
A with SP-A in constant-pressure mode, V8 open and V9 closed.

The control software for each of the syringe pumps allowed them to be operated in either constant
flow rate mode or constant pressure mode. By using one pump in constant flow mode (at a rate of 1
mL min-1) and the other in constant pressure mode it was possible to manipulate the locations of the
pistons in each cylinder prior to measuring the density of the saturated phases. To allow the density
measurement of both phases to occur at a given pressure, it was necessary to have the pump cylinder
volumes differ by at least 4 mL so that sufficient amounts of each phase could be displaced from the
equilibrium cell into the corresponding VTD to completely fill it. For example, prior to a density
measurement, the piston of SP-B was typically fully extended so that the volume of Cylinder B was
zero, while SP-A was set to have a cylinder volume of about 8 mL. A target pressure for the entire
system was chosen, and the SP-A was placed in constant pressure mode with this target as the set
point; SP-B remained in constant flow mode with a set rate of zero. Once the target pressure was
reached, the contents of the equilibrium cell were mixed for at least 10 minutes, the stirrer was
switched off and a further five minutes was allowed before SP-B was set to withdraw at a rate of
1 mL min-1. The SP-A’s control algorithm maintained the system pressure by extending the piston.
Usually, to maintain the pressure at the set-point, the flow rate at the upper pump was significantly
larger than 1 mL min-1, presumably because the compressibility of CO2-rich phase was greater than
that of the aqueous phase. After about 1.8 mL of aqueous phase sample had been withdrawn, a
breakthrough curve would be observed in the resonance period time-series data being logged from
the VTD-B, indicating the location and motion of the saturated aqueous phase composition front. After
about 3.5 mL had been withdrawn the shift in the densimeter’s resonance period was complete with
any further temporal variations corresponding to any small fluctuations in overall pressure or
temperature. The SP-B was stopped after between (4 and 5) mL had been withdrawn and, after a
further three minutes, the resonance period of the vibrating tube was recorded over a period of one

7
minute. During this time SP-A, which had often injected about (6 to 8) mL to compensate for the
amount withdrawn by the lower pump, remained in constant pressure mode. To be deemed stable,
the standard deviation of the VTD-B’s resonance period over the minute recording was required to be
less than 0.01 µs.

Figure 2. Breakthrough curve recorded for the upper VTD as water-saturated CO2 was displaced into
it from the equilibrium cell.

From this point, the procedure was reversed to enable the measurement of the density of the water-
saturated CO2–rich phase. The contents of the equilibrium cell were re-mixed for 10 minutes, the
stirrer was switched off and a further five minutes was allowed before SP-B was set to inject at a rate
of 1 mL min-1. Figure 2 shows a breakthrough curve recorded as a water-saturated sample of the CO2-
rich phase was displaced from the equilibrium cell through VTD-A. At least 5 mL of the CO2-rich phase
was withdrawn in this manner, although usually the lower piston was returned to an almost fully
extended position. Once the withdrawal stopped, a few minutes were allowed for equilibration before
the resonance period of the VTD-A was recorded.

Prior to the mixture measurements, both densimeters were calibrated in-situ using vacuum and the
pure fluids helium, CO2 and water. The purities and other details of the sample materials used in this
work are listed in Table 1. The method and results of the calibration were described by May et al:[34]
the lower (VTD-B) and upper (VTD-A) densimeters used in this work correspond to VTD-2 and VTD-3
in ref. [34], respectively. The working equation used to calculate the density ρ from the measured
resonance period τ was

 2

ρ
( ρM S00 )  

τ 
 (1 + βτ p ) − 1 , (1)
( V V )  τ 00 (1 + ετ 1t + ετ 2 t ) 
1 + α t + β p   2
 
 

8
where t is the Celsius temperature, p is the pressure, ρM is the density of the tube material (taken to
be 8.89 g·cm-3) and S00, τ00, ετ1, ετ2, αV, βV, and β τ are apparatus parameters determined according to
the calibration procedure described by May et al. [34]. Values and statistical uncertainties of the VTD
parameters used in this work are listed in Table 2. Also listed are the root means square (r.m.s.)
deviations of the pure fluid densities calculated using eq. (1) from the densities calculated from the
reference equations of state for the pure fluids [20; 30; 35] using the measured temperature and
pressure. Figure 3 shows the results of the calibration. A subsequent calibration of the densimeters
using only vacuum and water and the constraint (β τ = -βV/3.87)[34] was conducted about 8 months
later and the resulting parameter values, which are also shown in Table 2, were in good agreement
with the original values indicating the stability of the VTDs over that extended period.

Table 1. Material purities and suppliers.

Material Supplier Purity Specification

Carbon dioxide BOC 0.9995 mole fraction

Water MilliPore Q Resistivity > 18 MΩ·cm at T = 298.15 K

Helium BOC 0.99999 mole fraction

Table 2. Parameter values in eq. (1) used to convert the measured resonance periods to densities.
The parameters were determined by calibration using reference EOS densities for helium, water and
CO2 as described in reference [34]. A check of that calibration was performed 8 months later using
only water and helium and the constraint (β τ = -βV/3.87) [34].

Lower VTD (VTD-2) Upper VTD (VTD-3)

Parameter
Ref. [34] This work Ref. [34] This work

S00 0.549715(75) 0.552041 0.551776(96) 0.550079

τ00 / µs 2578.331(95) 2577.997 2577.964(62) 2578.203

106ετ1 / K-1 128.21(80) 127.50 128.06(53) 128.55

108ετ2 / K-2 4.97(34) 4.99 4.78(23) 4.87

106αV /3 K-1 14.56(29) 14.25 13.88(36) 14.58

105βV / MPa-1 2.56(41) 1.51 1.87(52) 1.38

105β τ / MPa-1 -0.321(18) Constrained -0.371(23) Constrained

( ρVTD − ρEOS )r.m.s. 0.35 0.13 0.44

0.31
−3
kg ⋅ m
2 ( VTD
ρ − ρ EOS )r.m.s.
10 0.071 0.013 0.089 0.032
ρ EOS

9
The uncertainty of the mixture density measurements was larger than the value of 0.5 kg⋅m-3
suggested by the results of the pure fluid calibrations and the propagation of error calculation based
on the uncertainties in the measured temperature, pressure and resonance period. It was instead
limited by the repeatability of the mixture data: as discussed below several duplicate points were re-
measured over periods ranging from six weeks to ten months. The variation between these repeat
data points suggested there was an additional uncertainty contribution associated with the process of
sampling and transporting the saturated phases from the equilibrium cell to the respective VTDs.
Based on the consistency of the mixture density data obtained, we estimate that the uncertainty
contribution due to sampling for this binary system was 1 kg⋅m-3.

Figure 3. Lower panel: Water (blue), CO2 (red) and helium (green) densities ρEOS calculated from their
respective reference equations of state [14, 21, 24] as a function of the squared normalised resonance
period (τ/τ0)2 measured for lower VTD(-2), , and upper VTD(-3), +. Upper panel: Residuals between
the EOS density and the density, ρVTD, calculated using eq. (1).

Results
The densities measured for the saturated aqueous and CO2-rich phases of the CO2 + H2O system are
listed in Table 3 and Table 4, respectively. In Table 4 both the temperature at which the phase
equilibrium was established and the temperature at which the density of that phase was measured
are reported. For the aqueous phase measurements (Table 3), the two temperatures were consistent
within experimental uncertainty. The measured data are also presented as isothermal data series
within pressure-density plots in Figure 4. The aqueous phase densities shown in Figure 4(a) vary in a
near-linear fashion with pressure, although below about T = 373 K the isothermal densities exhibit a
change in slope which becomes more pronounced as the temperature decreases. The (p,ρ) data for

10
the CO2-rich phase shown in Figure 4(b) exhibit a more pronounced dependence on phase behaviour,
similar in nature to that of pure CO2.

Figure 4. Measured (p, ρ) isotherm data for the CO2 + H2O system. (a) Aqueous phase data, for which
the saturation temperature was the same as the density measurement. (b) CO2-rich phase data, for
which the saturation temperature (shown) was normally (1 to 2) K below the temperature of the
density measurement (as listed in Table 4). Symbols: , T = 293 K; , T = 298 K; , T = 308 K; , T
= 323 K; , T = 343 K; , T = 373 K; , T = 398 K; , T = 423 K; , T = 448 K. Lines are intended
only to guide the eye.

11
Table 3. Densities ρ of the saturated aqueous phase of CO2 + H2O measured using the lower VTD at
temperatures TVTD and pressures p. The temperature of the density measurement was equal to the
saturation temperature within the experimental uncertainty. Also listed is the combined standard
uncertainty of each density measurement uc(ρ), which combines in quadrature the propagated
uncertainties associated with T, p, sampling, and the measured resonance period. a

TVTD / K p / MPa ρ / kg m-3 uc(ρ) / kg m-3 TVTD / K p / MPa ρ / kg m-3 uc(ρ) / kg m-3
292.7 1.45 1004.9 1.0 322.8 5.91 998.8 1.0
292.7 1.87 1006.2 1.0 322.7 6.16 999.2 1.0
292.7 3.92 1012.8 1.0 322.8 7.28 1000.2 1.0
292.7 5.55 1016.2 1.0 322.7 8.44 1001.9 1.0
292.7 5.63 1016.6 1.0 322.8 8.55 1001.3 1.0
292.7 5.66 1016.9 1.0 322.7 8.57 1002.0 1.0
292.7 9.99 1019.4 1.0 322.8 9.28 1002.1 1.0
292.7 15.72 1022.2 1.0 322.8 10.25 1002.8 1.0
292.7 15.75 1022.3 1.0 322.7 10.33 1003.5 1.0
292.7 20.43 1024.7 1.0 322.8 11.05 1003.6 1.0
292.7 20.84 1025.0 1.0 322.8 12.05 1004.2 1.0
292.7 25.12 1027.2 1.0 322.8 13.34 1005.1 1.0
292.7 25.14 1027.2 1.0 322.8 17.05 1007.2 1.0
292.7 25.58 1027.4 1.0 322.8 21.53 1008.9 1.0
292.7 30.11 1029.7 1.0 322.7 21.73 1009.5 1.0
292.7 30.28 1029.8 1.0 322.7 26.83 1012.1 1.0
292.7 35.16 1032.2 1.0 322.8 31.08 1014.2 1.0
292.7 36.70 1033.0 1.0 322.8 38.03 1017.6 1.0
292.7 40.50 1034.9 1.0 322.7 47.33 1022.2 1.0
297.7 1.46 1002.7 1.0 322.8 56.51 1026.3 1.0
297.8 2.53 1006.1 1.0 322.7 63.44 1028.6 1.1
297.8 3.15 1008.0 1.0 342.8 2.15 981.1 1.0
297.8 4.25 1010.8 1.0 342.8 3.12 982.5 1.0
297.8 4.25 1010.8 1.0 342.8 4.24 984.0 1.0
297.7 5.13 1012.8 1.0 342.8 6.01 986.2 1.0
297.8 5.75 1013.9 1.0 342.8 10.65 990.3 1.0
297.8 6.31 1014.5 1.0 342.8 15.99 994.3 1.0
297.8 6.33 1014.6 1.0 342.8 20.37 996.7 1.0
297.8 6.34 1014.5 1.0 342.8 31.42 1001.7 1.0
297.8 6.65 1014.7 1.0 342.8 47.19 1010.1 1.0
297.8 7.09 1015.1 1.0 342.8 61.63 1016.9 1.0
297.8 9.45 1016.4 1.0 373.2 0.30 958.0 1.0
297.8 11.31 1017.4 1.0 373.1 0.70 959.0 1.0
297.8 15.27 1019.6 1.0 373.2 0.92 959.0 1.0
297.8 20.05 1022.2 1.0 373.2 1.82 960.0 1.0
297.8 24.65 1024.5 1.0 373.0 3.18 961.9 1.0
297.8 26.71 1025.6 1.0 373.0 4.35 963.1 1.0
297.8 30.71 1027.6 1.0 372.9 5.19 964.0 1.0
297.8 35.53 1030.0 1.0 373.0 6.13 964.9 1.0
297.8 40.49 1032.5 1.0 373.0 8.13 966.7 1.0

12
 297.8 45.26 1034.8 1.0 373.0 8.17 966.7 1.0
307.8 0.22 994.6 1.0 373.0 8.42 966.9 1.0
307.8 0.80 996.5 1.0 373.0 10.08 968.3 1.0
307.7 2.01 999.9 1.0 373.0 15.87 972.6 1.0
307.8 2.03 999.9 1.0 373.0 17.60 973.5 1.0
307.8 2.95 1002.3 1.0 373.0 27.82 979.4 1.0
307.7 2.96 1002.2 1.0 373.0 33.24 982.3 1.0
307.7 3.94 1004.4 1.1 373.0 33.37 982.5 1.0
307.8 4.02 1004.4 1.1 373.1 40.29 986.0 1.0
307.7 5.20 1006.6 1.1 373.0 46.45 989.4 1.0
307.8 5.25 1006.7 1.1 373.0 50.01 990.3 1.0
307.7 7.27 1010.0 1.1 373.0 62.08 997.0 1.0
307.8 7.69 1010.0 1.1 398.5 19.46 954.0 1.0
307.7 8.07 1010.4 1.1 398.4 24.31 957.1 1.0
307.7 8.27 1010.6 1.1 398.4 27.05 958.7 1.0
307.8 8.62 1010.9 1.1 398.4 32.85 962.1 1.0
307.7 8.84 1010.9 1.1 398.4 41.52 966.9 1.0
307.8 9.58 1011.5 1.1 398.3 47.22 970.1 1.0
307.7 11.10 1012.4 1.1 423.3 2.43 918.5 1.0
307.8 11.48 1012.5 1.1 423.2 3.12 919.1 1.0
307.8 15.06 1014.5 1.0 423.3 4.17 919.8 1.0
307.7 17.73 1016.0 1.0 423.3 5.34 920.8 1.0
307.8 22.32 1018.3 1.0 423.2 17.73 929.8 1.0
307.7 31.36 1022.8 1.0 423.2 25.01 934.6 1.0
307.7 38.08 1025.9 1.0 423.3 46.22 947.7 1.0
307.7 46.54 1030.2 1.0 423.3 63.89 957.7 1.0
322.7 2.19 992.7 1.0 448.5 21.77 905.9 1.0
322.8 3.15 994.4 1.0 448.5 27.76 910.0 1.0
322.7 3.26 994.7 1.0 448.5 33.25 913.7 1.0
322.8 4.14 996.1 1.0 448.5 37.75 916.7 1.0
322.7 4.35 996.5 1.0 448.5 43.58 920.5 1.0
322.8 5.00 997.5 1.0 448.5 50.00 924.5 1.0
322.8 5.09 997.8 1.0
a
Standard uncertainties in temperature and pressure are u(T) = 0.2 K and u(p) = 0.05 MPa.

13
 Table 4. Densities ρ for the near-saturated CO2-rich phase of CO2 + H2O measured using the upper VTD
at temperature TVTD and pressure p when the phase was saturated at the same pressure but at
temperature Tsat. Also listed is the combined standard uncertainty of each density measurement uc(ρ),
which combines in quadrature the propagated uncertainties associated with T, p, sampling, and the
measured resonance period. a

Tsat / K TVTD / K p / MPa ρ / kg m-3 uc(ρ) / kg m-3 Tsat / K TVTD / K p / MPa ρ / kg m-3 uc(ρ) / kg m-3
292.7 292.7 1.87 38.5 1.1 322.8 323.9 9.12 291.8 2.4
292.7 292.7 3.91 97.4 1.2 322.8 323.9 9.28 305.3 2.6
292.7 292.7 5.61 187.0 2.5 322.8 323.9 10.25 406.0 4.2
292.7 293.1 5.63 187.8 2.4 322.8 323.9 11.05 499.1 4.6
292.7 292.7 5.65 781.6 2.6 322.8 323.8 12.05 581.6 3.6
292.7 292.7 5.65 781.1 2.6 322.8 323.9 13.34 645.2 2.7
292.7 292.8 6.04 796.2 2.9 322.8 323.7 17.05 740.2 1.8
292.7 292.5 10.40 867.0 1.7 322.8 323.9 17.05 739.2 1.8
292.7 292.7 15.78 913.3 1.4 322.8 323.8 21.49 800.5 1.5
292.7 292.7 15.81 913.4 1.4 322.8 324.1 21.52 799.9 1.5
292.7 292.7 20.85 944.9 1.3 322.8 324.1 26.81 847.0 1.3
292.7 292.7 25.60 968.2 1.3 322.8 323.8 31.10 876.3 1.3
292.7 292.8 35.55 1008.1 1.2 322.8 323.8 38.05 913.1 1.2
292.7 292.8 40.18 1022.5 1.2 322.8 324.1 47.31 951.2 1.2
297.8 297.8 1.42 27.7 1.1 322.8 323.8 56.48 982.4 1.1
297.8 297.8 4.24 103.8 1.2 322.8 324.1 63.01 1000.7 1.1
297.8 297.8 4.24 103.8 1.2 342.8 344.2 4.16 76.2 1.1
297.8 297.8 5.08 138.5 1.4 342.8 344.2 5.88 115.2 1.1
297.8 297.8 5.68 172.0 1.8 342.8 344.2 15.90 542.5 2.1
297.8 297.8 6.21 217.8 3.0 342.8 344.2 20.28 664.2 1.7
297.8 297.8 6.79 747.5 3.9 342.8 344.2 31.37 799.0 1.3
297.8 297.8 8.33 794.5 2.4 342.8 344.2 47.11 891.3 1.2
297.8 297.8 15.45 884.0 1.5 373.0 374.1 0.30 4.6 1.0
297.8 297.8 17.97 904.9 1.4 373.0 374.1 0.93 13.8 1.1
297.8 297.8 20.25 918.8 1.4 373.0 374.4 2.15 32.7 1.1
297.8 297.8 24.73 944.1 1.3 373.0 374.4 3.17 49.1 1.1
297.8 297.8 26.52 954.6 1.3 373.0 374.4 4.44 71.0 1.1
297.8 297.8 30.75 973.7 1.2 373.0 374.3 5.14 83.7 1.1
297.8 297.8 36.61 995.4 1.2 373.0 374.4 6.09 101.7 1.1
297.8 297.8 40.49 1008.7 1.2 373.0 374.4 8.71 157.6 1.1
297.8 297.8 45.34 1023.0 1.2 373.0 374.4 9.78 183.1 1.1
297.8 297.8 45.59 1023.4 1.2 373.0 374.3 15.76 356.7 1.4
307.7 308.7 0.25 4.2 1.1 373.0 374.1 17.59 413.2 1.4
307.7 308.9 0.79 14.8 1.1 373.0 374.1 20.20 487.0 1.5
307.7 308.9 0.80 14.9 1.1 373.0 374.1 27.82 634.6 1.4
307.7 309.1 2.00 39.2 1.1 373.0 374.1 32.04 686.7 1.3
307.7 308.9 2.01 38.7 1.1 373.0 374.1 33.24 699.4 1.3
307.7 308.8 2.92 59.1 1.1 373.0 374.1 38.04 742.5 1.2
307.7 309.0 2.94 60.1 1.1 373.0 374.1 40.30 759.5 1.2

14
307.7 309.0 3.92 85.7 1.2 373.0 374.4 46.46 798.9 1.2
307.7 308.8 3.98 86.3 1.2 373.0 374.1 50.04 819.1 1.2
307.7 309.1 5.17 125.6 1.3 373.0 374.4 62.00 872.4 1.1
307.7 308.8 5.24 126.5 1.3 398.4 399.4 19.44 373.7 1.2
307.7 308.8 6.02 159.0 1.4 398.4 399.3 24.29 476.7 1.3
307.7 308.8 6.70 196.1 1.8 398.4 399.3 27.04 525.3 1.3
307.7 308.8 7.68 290.3 4.8 398.4 399.4 32.84 607.1 1.2
307.7 308.9 9.58 686.0 3.7 398.4 399.3 41.54 692.6 1.2
307.7 309.0 11.04 740.4 2.5 398.4 399.2 47.24 734.7 1.2
307.7 308.9 11.45 750.5 2.3 423.3 424.8 2.42 28.4 1.0
307.7 308.9 15.06 812.5 1.7 423.3 424.8 3.11 37.8 1.0
307.7 309.0 17.61 842.0 1.5 423.3 424.9 4.17 52.4 1.0
307.7 308.9 22.32 881.2 1.4 423.3 424.8 5.32 69.0 1.0
307.7 309.1 31.40 934.0 1.3 423.3 424.8 8.73 121.3 1.1
307.7 309.1 46.68 993.5 1.2 423.3 424.8 17.59 279.8 1.1
322.8 324.2 2.16 39.5 1.1 423.3 424.8 24.98 415.6 1.2
322.8 323.9 3.15 61.2 1.1 423.3 424.9 46.18 661.4 1.1
322.8 324.1 3.20 61.3 1.1 423.3 424.9 63.77 765.7 1.1
322.8 323.9 4.15 83.7 1.1 448.5 449.5 4.18 47.1 1.0
322.8 324.1 4.28 86.9 1.1 448.5 449.5 21.74 312.4 1.1
322.8 323.8 5.00 105.8 1.2 448.5 449.5 27.74 403.8 1.1
322.8 324.1 5.07 107.9 1.2 448.5 449.5 33.24 476.9 1.1
322.8 324.1 6.05 137.7 1.2 448.5 449.5 37.74 527.9 1.1
322.8 324.1 8.45 244.8 1.8 448.5 449.5 43.58 583.6 1.1
322.8 323.9 8.55 250.2 1.9 448.5 449.6 50.01 634.0 1.1
a
Standard uncertainties in temperature and pressure are u(T) = 0.2 K and u(p) = 0.05 MPa.

15
Discussion
Since no composition measurements were made in this work, comparisons are made below with
models (or combinations of models) that permit the saturated-phase densities to be calculated at
specified temperatures and pressures.

Aqueous phase of CO2 + H2O

The absolute densities plotted in Figure 4 reflect primarily the dependence of the parent phase density
on pressure as a function of temperature. The contribution of the minor component on the saturated
mixture density is better assessed by considering the differences in density between the mixture and
that of either pure water (for the aqueous phase) or pure CO2 (for the CO2-rich phase). Figure 5 shows
the differences between the measured CO2-saturated aqueous phase densities from those of pure
water at the same temperature and pressure, calculated using the reference EOS for water of Wagner
and Pruss [20]. On this scale, some scatter in the measured data is evident but this is consistent with
the estimated densities uncertainties listed for the aqueous phase in Table 3, which have an average
of 1.0 kg·m-3. The change of slope at pressures near 10 MPa for isotherms below the critical
temperature of CO2 is very clear in Figure 5, and represents the transition from vapour-liquid
equilibrium at low pressures, where the CO2-solubility approximately follows Henry’s law, to liquid-
liquid equilibrium at high pressures.

Figure 5. Differences between the measured aqueous phase densities ρaq for CO2 + H2O from the
density ρw of pure water at the same pressure and temperature as calculated using the EOS of Wagner
and Pruss [20]. Symbols: , T = 293 K; , T = 298 K; , T = 308 K; , T = 323 K; , T = 343 K; , T
= 373 K; , T = 398 K; , T = 423 K; , T = 448 K. Lines are intended only to guide the eye.

16
Figure 6. Deviations (ρaq - ρcalc) of the saturated aqueous-phase densities ρaq for the CO2 + H2O system
from values ρcalc calculated at the same temperature and pressure using the models of Duan et al. [14]
(for saturated phase composition), Sedlbauer et al. [22] (for CO2 partial molar volume) and Wagner
and Pruss [20] (for solvent molar volume). (a) Symbols: this work , T = 293 K; , T = 298 K; , T =
308 K; , T = 323 K; , T = 343 K; , T = 373 K; , T = 398 K; , T = 423 K; , T = 448 K. Curves:
ρaq predicted using the EOS of Spycher et al. [16],[18] (for saturated phase composition) in place of the
Duan et al. [14] at the experimental conditions studied in this work. (b) Symbols: , this work; ,
Chiquet et al. [9]; , Yan et al. [36]; , Hebach et al. [7]; ,Tabasinejad et al. [10]; , Yaginuma et

17
al. [11]. Curves: ρaq predicted (entirely) with the EOS-CG of Span and co-workers [32] at the
experimental conditions studied in this work.

Figure 6 shows the deviations of the measured CO2-saturated aqueous-phase densities, ρaq, from
those calculated, ρcalc, using the empirical models of Duan et al. [14] (for saturated phase composition),
Sedlbauer et al. [22] (for CO2 partial molar volume) and Wagner and Pruss [20] (for solvent molar
volume). The deviations, which are all within the experimental uncertainty, indicate that this
combination of models is capable of describing the saturated aqueous-phase densities of this well -
studied system as accurately as they can currently be measured. The largest deviations occur at low
temperatures and high pressures, where the difference between the mixture and pure water densities
is largest (but still only 20 kg·m-3). At the highest temperature measured here (448 K) the deviations
from the model are of a similar magnitude but the opposite sign. Figure 6(a) also shows the effect of
replacing the model of Duan et al. [14] with the EOS of Spycher et al. [16; 18] for prediction of the
saturated phase composition in the calculation of ρcalc. The differences are almost negligible, ranging
from (-0.6 to 0.4) kg·m-3.

Figure 6(b) shows how data available in the literature for the aqueous phase densities of the CO2 + H2O
binary compare with the new data, as well as densities calculated using the EOS-CG of Gernert and
Span [32]. The new data extend, and are consistent with, the data sets reported by Hebach et al. [7],
Yan et al. [25] and Yaginuma et al. [11], which have a similar degree of scatter to the current work. In
contrast, the data of Chiquet et al. [9] have relative deviations of as much as -1.2 % from the calculated
densities, with an average of -0.3 % and a standard deviation of 0.3 %, which is about three times
larger than the standard deviation of the data reported here. The densities reported by Tabasinejad
et al. [10] at T = (382, 423 and 468) K have low scatter and are consistent with the other data and the
model predictions at their lowest pressure around 3.5 MPa. However, the data exhibit a systematic
negative deviation with increasing pressure that exceeds -1 % at all three temperatures. Additionally,
the data of Nighswander et al. [12], which are not plotted in Figure 6(b), deviate by between -1 % and
-3 %.

The value of the new data reported here for the aqueous phase densities is highlighted by the
comparison in Figure 6(b) of the recently developed EOS-CG of Gernert and Span [32] with the
literature density data. The new experimental results presented here help resolve the ambiguity
associated with the inconsistency in the literature data and demonstrate that the data sets of Chiquet
et al. [9], Tabasinejad et al. [10] and Nighswander et al. [12] should not be weighted highly in the
development of future models.

CO2-rich phase
The density of the CO2-rich phase is even closer to that of pure CO2 than the aqueous phase density is
to pure water. Figure 7(a) shows the deviations of the (near)-saturated CO2-rich phase densities from
those of pure CO2 at the same temperature and pressure, calculated using the reference EOS for
carbon dioxide of Span and Wagner [30].

18
Figure 7. (a) Differences between the measured densities ρnon-aq of the CO2-rich phase from the density
ρCO2 of pure CO2 at TVTD as calculated from the EOS of Span and Wagner [30]. (b) Deviations (ρnon-aq -
ρEOS-CG), where ρEOS-CG is the density of the CO2-rich phase calculated using the EOS-GC of Gernert and
Span [32]. The ρEOS-CG values used Tsat in Table 4 to predict the mixture composition and then used TVTD
(≈ Tsat + 1 K) to calculate the density of that mixture at the measurement temperature. Symbols: ,
TVTD = 293 K; , TVTD = 298 K; , TVTD = 309 K; , TVTD = 324 K; , TVTD = 344 K; , TVTD = 374 K; ,
TVTD = 399 K; , TVTD = 425 K; , TVTD = 450 K.

19
In the limits of both high and low densities, the deviations of the mixture mass densities from those
of pure CO2 tend to zero, except at the two higher temperatures, 425 K and 450 K, where the
deviations become negative as expected from ideal-gas mixing (because the molar mass of H2O is
lower than that of CO2). At lower temperatures, the vapour pressure of water and, hence, the mole
fraction of water in the CO2-rich phase is small, and thus it might be expected that any reduction of
the mixture’s density relative to pure CO2 due to ideal-gas mixing would be smaller than the resolution
of the experiment. However, at these lower temperatures the mixture becomes denser than pure CO2
with the effect being clearly larger than the experimental uncertainty at densities above about
100 kg⋅m-3. This indicates that the small decrease in the mixture’s molar mass at these conditions is
offset by a slightly larger increase in the mixture’s molar density. Above about 800 kg·m-3, where the
CO2-rich mixture has reached molar densities characteristic of a liquid phase, the decrease in the
mixture’s molar mass at all temperatures appears to be nearly balanced by the molar density increase.
Between these two limits, the mixture densities deviate positively from those of pure CO2, with a
maximum of about 8 kg·m-3 across all measurements. The location and value of the maximum appears
to vary with temperature: at the lower temperatures it is in the vicinity of the critical density for CO2
(468 kg·m-3) [30], whereas at the highest temperature, it has shifted to around 600 kg·m-3. It is
apparent from this maximum that at near-critical densities the mixture can support a molar density
that more than offsets the associated decrease in molar mass.

Figure 7(b) compares the new data for the CO2-rich phase with mixture densities, ρEOS-CG, calculated
using the EOS-CG of Gernert and Span [32]. To calculate ρEOS-CG, the EOS-CG was first used to calculate
the saturated phase composition at Tsat in Table 4 and then again to calculate the density of this
mixture at TVTD in Table 4 (where it is slightly under-saturated). While the difference between Tsat and
TVTD in Table 4 is small (≈ 1 K to 2 K), the sensitivity of density to temperature in the vicinity of carbon
dioxide’s critical point means that assuming TVTD = Tsat for data around 500 kg⋅m-3 in the range (308 to
344) K causes ρEOS-CG to shift by (10 to 30) kg⋅m-3. The maximum and r.m.s. deviations of multi-
parameter EOS-GC from the data are 5 kg⋅m-3 and 2 kg⋅m-3, respectively; the data have a standard
deviation about the model predictions of 1.5 kg⋅m-3, which is close to the average experimental
combined standard uncertainty for the CO2-rich phase. In contrast, the GERG-2008 multi-parameter
EOS [31] which was developed for natural gas mixtures, has maximum and r.m.s. deviations from the
data of 23 kg⋅m-3 and 6 kg⋅m-3, respectively, while the EOS for pure CO2 by Span and Wagner [30] has
maximum and r.m.s. deviations of 8 kg⋅m-3 and 3 kg⋅m-3, respectively.

Figure 8 shows a comparison of the results from this work with the other two sets of density data
available in the literature for the CO2-rich phase. The data of Chiquet et al. [9], which cover the
temperature range (307 to 384) K again exhibit the most scatter, with a range of 49 kg·m-3 and a
standard deviation of 8 kg·m-3. Three of the four isotherms reported by Tabasinejad et al. [10] have a
similar scatter to the data reported here, and also have a maximum positive difference from the
density of pure CO2 in the vicinity of its critical density, although this difference is about twice as large
in magnitude as observed in this work. However, data on one of the intermediate isotherms measured
by Tabasinejad et al. [10] appear to be inconsistent with their other data and those measured in this
work, with an apparent minimum of similar magnitude also occurring near the critical density. At
higher densities, the deviations increase linearly, reaching nearly 60 kg·m-3. It is likely this subset of
the measurements by Tabasinejad et al. [10] suffer from a systematic error and should be excluded
from further consideration.

20
Figure 8. Differences between the densities ρnon-aq of the CO2-rich phase from the density ρCO2 of pure
CO2 as calculated from the EOS of Span and Wagner [30]. Symbols: , Tabasinejad et al. [10] at T =
(423, 462 and 478) K; , Tabasinejad et al. [10] at T = 446 K; , Chiquet et al. [9]; , this work (with
ρCO2 evaluated at TVTD). Curves show calculations made using the EOS of Spycher and Pruess [18] at:
, T = 373 K; , T = 398 K; , T = 423 K; , T = 448 K.

Figure 8 also shows a comparison of CO2-rich phase densities calculated using the RK-EOS within the
model of Spycher and Pruess [18]. At temperatures below 372 K, this model explicitly treats the CO2-
rich phase as pure CO2, so only comparisons for the isotherms measured in this work at higher
temperatures are shown. Between (372 and 382) K, densities calculated using the model of Spycher
and Pruess [18] are a weighted average of the RK-EOS predictions for pure CO2 and the water-
saturated mixture. At each of the four temperatures above 372 K considered, the CO2-rich phase
densities calculated using the model of Spycher and Pruess [18] are in remarkably good agreement
with the experimental results obtained in this work. At 373 K, the cubic RK-EOS predicts a maximum
density difference from pure CO2 of +6 kg⋅m-3 at a density of around 540 kg⋅m-3, which is consistent
with the measured differences within the experimental uncertainty. The location of the predicted
maximum does not change significantly with temperature, although the value of the predicted
maximum difference does increase systematically with temperature up to +11 kg⋅m-3 at 448 K.
Furthermore, the model correctly predicts the density differences observed in the limit of both low-
and high-density limits, including the negative density differences observed at lower pressures and
higher temperatures.

Conclusions
A new apparatus capable of precise measurements of the saturated densities of both the aqueous and
CO2-rich phases over a wide range of temperature and pressure has been described and
demonstrated. Measurements made for both the saturated aqueous phase and CO2-rich phase of the
CO2 + H2O system over the temperature range (293 to 450) K at pressures to 64 MPa helped identify

21
erroneous literature data and extend the range of the available accurate literature data. For the
saturated aqueous phase of the CO2 + water system, the new data were consistent with two existing
data sets [7; 25] that exhibit comparable uncertainties and with the predictions of an empirical model
[14; 20; 22] to within 0.3 % of the density. Using the computationally efficient EOS of Spycher et al.
[16; 18] to calculate the aqueous phase composition in the empirical model rather than the method
of Duan et al. [14] produced negligibly different results. However, the EOS-CG model developed
recently by Gernert and Span [32] reproduces one of the erroneous data sets [10] and its predictions
for the aqueous phase density of the CO2 + H2O system deviated systematically with increasing
pressure at all temperatures, becoming increasingly negative to a relative value of about -1.3 % at T =
448 K.

Both of the literature data sets available for the CO2-rich phase density of the CO2 + H2O system were
found to contain problems, either in the form of significantly larger data scatter [9] (±20 kg⋅m-3) than
the measurements in this work (±1.5 kg⋅m-3), or in the form of systematic inconsistencies [10]. The
new data reported here demonstrate that, for the conditions studied, the mass density of the
saturated CO2-rich phase is always within 8 kg⋅m-3 of the density of pure CO2 at the same pressure and
temperature. A maximum difference of about +8 kg⋅m-3 was observed to occur in the vicinity of the
critical density of CO2 at all temperatures, indicating that the increase in the molar density at these
conditions arising from the presence of water made a larger contribution to the mass density than the
associated decrease in the molar mass. Only at temperatures higher than about 423 K and at pressures
below about 10 MPa was the mass density of the CO2-rich mixture found to be unambiguously lower
than that of pure CO2 as might be expected from a consideration of ideal gas mixing. At densities above
800 kg⋅m-3, the mixture density was less than 3 kg⋅m-3 above that of pure CO2. For the CO2-rich phase,
the EOS-CG model [32] was able to describe the measured densities within ±5 kg⋅m-3, with the
deviations having a scatter comparable with the experimental uncertainty and exhibiting only slight
systematic trends at higher temperatures. At temperatures of 373 K and above, the cubic EOS used in
the model of Spycher and Pruess [18] was in suprisingly good agreement with the experimental data:
its predictions of the maximum differences for the CO2-rich phase densities from pure CO2 ranged
from (+6 to +11) kg⋅m-3 from (373 to 448) K, and it also correctly represented the observed behaviour
in the limit of both high and low densities.

Acknowledgements
This work was carried out as part of the activities of the Qatar Carbonates & Carbon Storage Research
Centre (QCCSRC). We gratefully acknowledge the funding of QCCSRC provided jointly by Qatar
Petroleum, Shell, and the Qatar Science and Technology Park, and their permission to publish this
research. The authors thank Roland Span, Monika Thol and Stefan Herrig at Ruhr-Universität Bochum
for assisting with calculations made using the EOS-CG. EFM thanks the Department of Chemical
Engineering and QCCSRC at Imperial College London for hosting him on his sabbatical where this work
was done, and is grateful to everyone there who made that stay stimulating and enjoyable.

22
References
[1] K. Pruess, J. Garcia, Environmental Geology 42 (2002) 282-295.
[2] X. Li, E. Boek, G.C. Maitland, J.P.M. Trusler, Journal of Chemical & Engineering Data 57 (2012)
1078-1088.
[3] X. Li, E.S. Boek, G.C. Maitland, J.P.M. Trusler, Journal of Chemical & Engineering Data 57
(2012) 1369-1375.
[4] P.M. Haugan, H. Drange, Nature 357 (1992) 318-320.
[5] P.D. Jones, T.M.L. Wigley, P.B. Wright, Nature 322 (1986) 430-434.
[6] H. Teng, A. Yamasaki, M. Chun, H. Lee, Journal of Chemical Thermodynamics 29 (1997) 1301-
1310.
[7] A. Hebach, A. Oberhof, N. Dahmen, Journal of Chemical & Engineering Data 49 (2004) 950-
953.
[8] Z. Li, M. Dong, S. Li, L. Dai, Journal of Chemical & Engineering Data 49 (2004) 1026-1031.
[9] P. Chiquet, J. Daridon, D. Broseta, S. Thibeau, Energy Conversion and Management 48 (2007)
736-744.
[10] F.B. Tabasinejad, Yalda; Moore, Robert Gordon; Mehta, Sudarshan A.; Van Fraassen, Kees
Cornelius; Rushing, Jay; Newsham, Kent Edward, Water/CO2 System At High Pressure And
Temperature Conditions: Measurement And Modeling Of Density In Equilibrium Liquid And Vapor
Phases., SPE EUROPEC/EAGE Annual Conference and Exhibition, SPE, Barcelona, Spain, 2010, pp.
SPE-131636-MS.
[11] R. Yaginuma, Y. Sato, D. Kodama, H. Tanaka, M. Kato, Nihon Enerugi Gakkaishi 79 (2000)
144-146.
[12] J.A. Nighswander, N. Kalogerakis, A.K. Mehrotra, Journal of Chemical & Engineering Data 34
(1989) 355-360.
[13] Y.B. Chang, B.K. Coats, J.S. Nolen, SPE Reservoir Evaluation & Engineering (1998) 155-160.
[14] Z. Duan, R. Sun, C. Zhu, I. Chou, Marine Chemistry 98 (2006) 131-139.
[15] S.-X. Hou, G.C. Maitland, J.P.M. Trusler, Journal of Supercritical Fluids 73 (2013) 87-96.
[16] N. Spycher, K. Pruess, J. Ennis-King, Geochimica et Cosmochimica Acta 67 (2003) 3015-3031.
[17] N. Spycher, K. Pruess, Geochimica et Cosmochimica Acta 69 (2005) 3309-3320.
[18] N. Spycher, K. Pruess, Transport in Porous Media 82 (2010) 173-196.
[19] O. Redlich, J.N.S. Kwong, Chemical Reviews 44 (1949) 233-244.
[20] W. Wagner, A. Pruss, J Phys Chem Ref Data 31 (2002) 387-535.
[21] S. Al-Ghafri, G.C. Maitland, J.P.M. Trusler, Journal of Chemical & Engineering Data 58 (2013)
402-412.
[22] J. Sedlbauer, J.P. O'Connell, R.H. Wood, Chemical Geology 163 (2000) 43-63.
[23] Z. Duan, J. Hu, D. Li, S. Mao, Energy & Fuels 22 (2008) 1666-1674.
[24] D. Li, B.J. Graupner, S. Bauer, Energy Procedia 4 (2011) 3817-3824.
[25] M. McBride-Wright, G.C. Maitland, J.P.M. Trusler, Journal of Chemical & Engineering Data
ASAP article (2015).
[26] M. McBride-Wright, Viscosity and Density of Aqueous Fluids with Dissolved CO2, Chemical
Engineering, Imperial College London, 2013.
[27] Z. Duan, N. Moller, J.H. Weare, Geochimica et Cosmochimica Acta 56 (1992) 2605-2617.
[28] Z. Duan, N. Moller, J.H. Weare, Geochimica et Cosmochimica Acta 56 (1992) 2617-2631.
[29] M. Benedict, G.B. Webb, L.C. Rubin, Journal of Chemical Physics 8 (1940) 334-345.
[30] R. Span, W. Wagner, J Phys Chem Ref Data 25 (1996) 1509-1596.
[31] O. Kunz, W. Wagner, Journal of Chemical & Engineering Data 57 (2012) 3032-3091.
[32] G.J. Gernert, R. Span, Journal of Chemical Thermodynamics (Submitted) (2015).
[33] H. Preston-Thomas, Metrologia 77 (1990) 3-10.
[34] E.F. May, W.J. Tay, M. Narnia, A. Aleji, S. Al-Ghafri, J.P.M. Trusler, Review of Scientific
Instruments 85 (2014) 095111.

23
[35] D.O. Ortiz-Vega, K.R. Hall, J.C. Holste, V.D. Arp, E.W. Lemmon, Helmholtz equation of state
for helium of Ortiz-Vega et al. (2013), 2013.
[36] W. Yan, S. Huang, E.H. Stenby, International Journal of Greenhouse Gas Control 5 (2011)
1460-1477.

24