Sunteți pe pagina 1din 23

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/he

New model of phase equilibria in metal e hydrogen


systems: Features and software

M.V. Lototskyy*
South African Institute for Advanced Materials Chemistry, University of the Western Cape, Private Bag X17,
Bellville, 7535, South Africa

article info abstract

Article history: A new semi-empirical model of phase equilibria of hydride forming metals and alloys with
Received 12 October 2015 hydrogen gas (PCT diagrams) has been developed. Within the model, pressure e compo-
Received in revised form sition isotherms (PCI’s) for hydrogen absorption and desorption are presented as pseudo-
6 December 2015 convolutions of one core function, or “ideal” PCI built according to Lacher model, or
Accepted 6 December 2015 model of Van der Waals lattice gas, with two modified asymmetric pseudo Voight distri-
Available online 8 January 2016 bution functions. The median of the absorption distribution is shifted from the one of the
desorption distribution towards higher pressures, by the value related to a minimum of a
Keywords: hysteresis free energy (assumed to be constant) and an excessive chemical potential of
Metal hydrides hydrogen in over-saturated a-solid solution.
Thermodynamics The model allows to simultaneously fit both H2 absorption and desorption datasets and
PCT diagram adequately simulates the features of experimental PCT diagrams in wide temperature
Modelling range including (i) presence of several plateau segments, (ii) hysteresis including its
disappearance when approaching to critical temperature, (iii) shape of transitions between
a-, (a þ b)- and b-regions, and (iv) flat or sloping plateaux including temperature and
concentration dependencies of the plateau slope.
The article also presents a demonstration software (macro-enabled Excel workbook
with an ActiveX calculation module) for the fitting of experimental PCT datasets verified on
the systems H2 e LaNi5exSnx, H2 e TiFe and H2 e bcc-V(Ti,M); M ¼ Zr,Ni.
Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights
reserved.

capacities, as well as pressure e temperature conditions


Introduction which correspond to the equilibrium of the reversible
reaction:
For gas phase applications of metal hydrides (MH) including
hydrogen storage, hydrogen compression and heat manage- absorption
MðsÞ þ x=2 H2 ðgÞ % MHx ðsÞ (1)
ment [1e5], the most important properties of hydride forming desorption

metals and alloys are their hydrogen absorption/desorption

* Corresponding author. Tel. þ27 21 959 9314; fax þ27 21 959 9312.
E-mail address: mlototskyy@uwc.ac.za.
http://dx.doi.org/10.1016/j.ijhydene.2015.12.055
0360-3199/Copyright © 2015, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.
2740 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1

Nomenclature Y generic fitting parameter


z ratio of critical and actual temperatures
a maximum H filling fraction in a-phase
a0 maximum H filling fraction in over-saturated a- Indexes
phase A absorption
a* attraction constant, Pa,m6 calc calculated
A asymmetry C critical
b minimum H filling fraction in b-phase D desorption
b' minimum H filling fraction in under-saturated b- exp experimental
phase g gas
b* repulsion constant, m3 G Gaussian
C hydrogen concentration in the solid, H/M H hydrogen atom
d generic distribution parameter h hysteresis
F concentration- and temperature-dependent term i number of plateau segment
of hydrogen equilibrium pressure I interstitial
f fugacity, atm L Lorentzian
G Gaussian distribution function M metal atom
G Gibbs free energy, J mol1 max maximum (asymptotic)
H enthalpy, J mol1 min minimum
h hysteresis energy loss, J mol1 o standard state
h1 coefficient of concentration dependence of s solid
enthalpy of hydrogen atoms in the solid, J mol1 SH entropyeenthalpy
k Boltzmann constant (J K1), total number of V pseudo-Voight
plateau segments
Greek symbols
L Lacher function
a, a0 H solid solution phase
m mixing factor
b hydride phase
M mixing coefficient
d delta function
N number of atoms, number of experimental points
d distance of experimental point from the
P pressure, Pa, atm
calculated curve
p ratio of actual and critical pressures
f fugacity coefficient
R Universal gas constant, J mol1 K1
h contribution of Lorentz profile
Rf goodness of fitting
m chemical potential, J mol1
S entropy, J mol1 K1
q filling fraction for ideal PCI
s slope factor
Q filling fraction for non-ideal PCI
SR total squared deviation
r Pearson correlation coefficient
T temperature, K
s standard deviation
t ratio of actual and critical temperatures
U specific matrix volume per a metal atom, m3
v ratio of actual and critical volumes
j distribution density
V molar volume, m3 mol1
J distribution integral
V Van der Waals lattice gas function
F Concentration- and temperature-dependent term
w width parameter
of hydrogen equilibrium pressure
W segment weight
X difference of logarithms of equilibrium and
plateau pressures

where M is the hydride forming metal (alloy, intermetallic rule, formal-empirical approach to the PCT modelling, by an
compound), the indexes s and g are related to the solid and gas arbitrary selection of a dependence which is more or less
phases, respectively. adequate to the experimentally observed behaviour (see, e.g.
Equilibrium of reaction (1) implies a unambiguous inter- Refs. [6e12]). This approach is convenient for the applications,
relation between gas pressure, P, hydrogen concentration in but unsatisfactory from the phenomenological point of view.
the solid, C, and temperature, T. Moreover, it often fails when the extrapolation out of the
available experimental PCT limits is necessary.
P ¼ PðC; TÞ; or C ¼ CðP; TÞ (2) PCT modelling based on statistical and / or thermodynamic
The corresponding analytical or numerical dependence prerequisites has a number of advantages including physical
(PCT diagram) has to be derived using this or that modelling interpretability of coefficients in the corresponding analytical
approach. expressions (model fitting parameters) that enables to analyse
Engineering calculations of MH reactors based on the the data for various materials, for example, with the purpose
modelling of heat-and-mass transfer in the MH beds use, as a of their optimal integration in the application. It enables the
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1 2741

extrapolation, frequently inaccurate but realistic, allowing to A simplified thermodynamic derivation of the PCT model
predict features of pressure e composition isotherms (PCI’s) similar to the Lacher one was suggested by Kierstead [28]. The
and to estimate hydrogen concentrations in the solid in wide model was extended for the description of the “multi-plateau”
pressure e temperature range. The model accuracy can be PCI presented as a superposition of several “single-plateau”
further improved by introducing semi-empirical corrections ones. Further refinement of the model by accounting the in-
which quantify additional features (e.g., sloping plateaux) for teractions between H atoms in the different types of the in-
the PCI’s in the real metal e hydrogen systems. terstitials was presented in Refs. [29,30].
Most of PCT models of this kind consider hydrogen in the The application of the thermodynamic approach for the
solid as a lattice gas. In doing so the H atoms are assumed to description of asymmetric PCI’s and considering H in metal as
be statistically distributed in interstitial sites between the an interacting lattice gas was suggested by Marinin, Lototsky,
metal atoms and to be able to randomly move towards vacant et al [31-33].
neighbouring interstitials. In the simplest case all the in- The deviations of PCT diagrams in real metal e hydrogen
terstitials are assumed to be energetically equivalent, and systems from the modelled “ideal” behaviour so far were
appearance of one plateau on a PCI at a temperature below mainly considered as regards to sloping plateaux. Larsen and
critical is associated with a decomposition of an interstitial Livesay [34] introduced the variation of the plateau pressure
solid solution onto two disordered phases characterised by with hydrogen concentration, so as the hydrogen equilibrium
different concentrations of the H atoms, similarly to gas e pressure was obtained as a convolution of an ideal flat-plateau
liquid transformation.1 isotherm with a function hðPÞ ¼ dCðPÞ=dP similar to a Gaussian
In 1936 Fowler [13] suggested general theory of adsorption distribution density. This approach was used for the solution
isotherms exhibiting plateau behaviour at temperatures of a particular subtask in the modelling of hydrogenation ki-
below critical. In 1937 Lacher [14] modified this theory for the netics in H2 e SmCo5 system and remained unnoticed for a
description of PCT diagram in H2 e Pd system. Both PCT long time as regards to the PCT modelling, maybe except for a
models were derived statistically using Bragg-Williams semi-empirical model by Bjurstro € m et al [35] where the
approximation of the attractive interaction between the sloping plateau PCI was built as a superposition of two
nearest neighbours of the adsorbate species. Further Langmuir-like isotherms.
improvement of the Lacher model taking into account volume Fujitani et al. [36] further elaborated the original approach
increase of the metal matrix in the course of H absorption and [34], by the combination of a regular interstitial hydrogen so-
explaining the PCI asymmetry was done by Harashima, lution model for the a- and b-phases of H2 e M system with the
Tanaka and Sakaoku in 1948 [15,16]. Accounting of energetic Gaussian distribution model for the hydrogen equilibrium
non-equivalency of interstitial sites in the metal matrix for pressures in the sloping plateau region. Two Gaussian distri-
the insertion of the H atoms thus describing the PCI’s with bution functions, one for a e (a þ b) and another for (a þ b) e b
several plateau segments was suggested by Rees in 1954 [17]. A transition were introduced to yield sloping plateau PCI as a
review of the PCT modelling in metal e hydrogen systems weighted superposition (pseudo-convolution) of the a- and b-
using statistical mechanics was presented by Speiser in segments. Later Lototsky et al. [33,37] modelled the sloping
Chapter 4 of well-known monograph edited by Mueller et al. plateau as a convolution of one Gaussian distribution of the
[18]. Later contributions in the application of statistical plateau pressure with “ideal” PCI built according to Lacher [37]
physics for the description of phase equilibria in metal e or Van der Waals lattice gas [33] models. The software for the
hydrogen systems were made by Hall and Futran [19], Bro- fitting of experimental PCT data utilising this approach is
dowsky and Yasuda [20], Beeri et al. [21,22], Senoh et al [23]. available on the author’s Research Gate webpage [38].
Some calculations based on the consideration of specific In Refs. [33,37] the origin of the deviation of the experi-
interstitial sites in the metal matrix available for H insertion mentally observed PCI’s from the “ideal” behaviour resulting
[20,22,23] yielded rather good correspondence of the calcu- in the sloping plateaux was assumed to be in inhomogeneities
lated PCI’s characterised by “multi-plateau” behaviour to the of the parent hydride forming alloys which, first of all, affect
experimentally observed ones. the plateau pressures [39]. The analysis of sloping plateaux in
Statistical model of phase equilibria in systems of hydrides of alloys or intermetallic compounds was elaborated
hydrogen gas with hydride-forming metals and alloys was by Park, Luo and Flanagan [40,41] who developed a quantita-
proposed by Ledovskikh et al in 2006 [24]. The model is based tive method for the determination of compositional distribu-
on the consideration of interstitial hydrogen as a lattice gas tion functions for hydride forming alloys / intermetallic
using the mean-field theory and Bragg-Williams approxima- compounds starting from the experimentally observed
tion. Further upgrade of the model [24] by the application of plateau sloping.
dynamic equilibrium approach based on first principles Zhou et al. [42] used the model of regular solid solution
chemical reaction kinetics [25,26] allowed very precise simu- with empirical correction for the plateau slope for the
lation of experimental PCI datasets for a number of systems modelling of PCT curves for H2 e AB2 systems. This approach
based on AB5-type intermetallides and characterised by one was further developed by Fang et al. [43,44] involving the
plateau segment [25,27]. correlation between atom size and electron properties of
metal atoms and enthalpy of the hydride formation, as well as
hysteresis factor.
1
We note that the models of this kind, including the one The PCT models involving an “ideal” PCT diagram built
developed in the present work, describe the formation of inter- according to the lattice gas model as a “core” combined with a
stitial hydrides only. distribution of plateau pressure allow to fit, with satisfactory
2742 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1

accuracy, various experimental datasets in wide range of 1 1


temperatures and hydrogen pressures (see, e.g. Refs. mg ¼ ðHo  TSo þ RT lnf Þ ¼ fHo  TSo þ RT ln½fðP; TÞPg (4b)
2 2
[36,43e47]). At the same time, some problems still exist
where Ho and So are enthalpy and entropy of a standard
including:
state (To ¼ 0  C, Po z fo ¼ 1 atm), respectively; P (f) is a pressure
(fugacity), T is a temperature, f(P,T) is a fugacity coefficient;
 The fitting parameters related to hydrogenation enthalpy
and R is a universal gas constant.
and entropy for the core function exhibiting non Lacher
Hydrogen concentration, C, in the solid can be expressed
(asymmetric) plateau behaviour [33] are shifted from their
as:
observed (apparent) standard values calculated from Van’t
Hoff dependencies. The shifts are temperature-dependent NH
C¼ (5)
and the corresponding corrections must be done for every NM
temperature.
where NH and NM are the numbers of hydrogen and metal
 Poor accuracy of the calculated PCI’s at the beginning and
atoms, respectively. The total number of the interstitials, NI,
the end of plateau, especially when the plateau slope is
available for the accommodation of the hydrogen atoms de-
low. It is caused by the interdependence of the modelled
termines the maximum hydrogen concentration in the solid,
plateau slope and the curvature of its transitions to the
Cmax:
solid solution regions. As it was noted by Park et al. [40], the
origin of these transitions is in the stress effects accom- NI
Cmax ¼ (6)
panying the hydride phase transformations, and the NM
sloping of the P(C) dependence at the initial and final stages In doing so, the hydrogen concentration can be expressed
of hydride formation/decomposition takes place even for via a filling fraction, q (0  q  1):
pure metals characterised by flat plateaux in the midpoint.
 Asymmetric behaviour of the sloping plateau pressure for C NH
q¼ ¼ (7)
some metal e hydrogen systems especially pronounced for Cmax NI
the H absorption in H2 e V [48]. Assuming a single plateau on a pressure e composition
 Necessity to separately process H absorption and desorp- isotherm, the chemical potential, ms, can be expressed via the
tion PCT datasets while potentially they could be fitted by partial (per 1 g-atom H) enthalpy, H, and entropy, S, which are
the same sets of parameters including the one(s) respon- dependent on the hydrogen concentration, or the filling frac-
sible for the hysteresis. tion, q:

This article presents a new semi-empirical model of phase ms ¼ HðqÞ  T SðqÞ (8)
equilibria of hydride forming metals and alloys with hydrogen
where:
gas. The model is based on the approaches [33,37,40]
 
mentioned above with further updates allowing to address q
SðqÞ ¼ S0  R ln (9)
the specified problems. The software which realises the model 1q
for the fitting of experimental PCT datasets is presented in
Supplementary information. HðqÞ ¼ H0 þ H1 ðqÞ (10)

In so doing, the concentration dependence of the entropy


The model of H atoms in the solid is determined by configuration effects
(9), and the concentration dependence of the enthalpy in-
Core function: one segment PCI built according to a lattice cludes concentration-independent and concentration-
gas model dependent terms (10).
By introducing standard changes of the enthalpy, DHo, and
Following thermodynamic approach [28] in the derivation of the entropy, DSo, of reaction (1):
the PCT diagram for the reaction (1), the equilibrium between
DHo ¼ 2H0  H0 (11)
hydrogen molecules in the gas (g) and hydrogen atoms in the
solid (s) takes place if their chemical potentials (m) are equal:
DSo ¼ 2S0  S0 (12)
mg ðT; PÞ ¼ ms ðT; CÞ (3)
and taking into account Eqs. (4e10), we can re-write Eq. (3) as:
In doing so, the PCT diagram (2) can be obtained as a so-  
lution of Eq. (3) where mg is a function of temperature and q H1 ðqÞ
ln f ¼ ln f0 þ 2 ln þ2 (13)
hydrogen pressure, and mS is a function of temperature and 1q RT
hydrogen concentration in the solid.
The chemical potential, mg, of the gaseous H2 (per 1 g-atom DSo DHo
lnf0 ¼  þ (14)
R RT
H) is calculated as:
    Eqs. (13,14) represent the general form of a thermodynamic
1 f 1 fP model of PCT diagram with one segment corresponding to one
mg ¼ Ho  TSo þ RT ln o ¼ Ho  TSo þ RT ln o o (4a)
2 f 2 fP type of energetically equivalent interstitials available for the
or, when [P, f]¼ atm: insertion of H atoms. The specific form of the PCI at any
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1 2743

temperature, T, is dependent on the kind of the concentration value of ln f0 in this case, in addition to Eq. (14), will contain
dependence H1 ðqÞ of the partial molar enthalpy of the inter- additional term Fðq0 ; zÞ; which, generally, may not be equal to
action between H atoms accommodated in the metal matrix. zero. However, since the value of q0 is concentration-
In the simplest case (ideal solid solution) H1 ðqÞ ¼ 0, and Eq. (13) independent, this term can modify only the apparent values
takes form similar to Langmuir adsorption isotherm when the of DHo (11) and DSo (12), and the Van’t Hoff Eq. (14) will be still
shape of the PCI in coordinates ln f e q does not depend on the valid.
temperature, and the shift of the isotherm along ln f axis with Three examples of PCI’s described by Eq. (20) are shown in
changing the temperature is determined by Van’t Hoff Eq. (14). Fig. 1. At T  TC (z  1) a single feature of the X(q) function
At H1 ðqÞs0 the PCT behaviour becomes more complicated. (curves ABC) is an inflection point B at q ¼ qC and X ¼ 0; the
Of special interest here is a case of phase separation (i.e. for- derivative, vX=vq, is positive (or zero at z ¼ 1 and q ¼ qC). When
mation of a-solid solution and b-hydride2) accompanied by z > 1, the isotherm exhibits “spinodal” behaviour (curve AEFC)
the appearance of plateau region on a PCI. Formally, the phase with the change of sign of v2 X=vq2 from “e” to “þ” in between
separation phenomenon can be shown by the analysis of points E and F where vX=vq becomes negative. The segment EF
concentration dependence of ln f (13) which can be presented cannot be realised in practice. Instead, the system exhibits
as a function X(q,T) defined as: separation onto two phases: a (q ¼ a) and b (q ¼ b) corre-
  sponding to a plateau (straight segment DBG) at f ¼ f0 (or X ¼ 0)
Xðq; TÞ≡ lnf  ln f0 ¼ 2Fðq; TÞ (15) which crosses the calculated isotherm (curve AEFC) in the
where: point B’ (q ¼ q0). Commonly, the position of the plateau and,
accordingly, the value of q0, is determined using Maxwell’s
 
q H1 ðqÞ equal area rule, so as SDEB0 ¼ SB0 FG. Note that it is not always the
Fðq; TÞ ¼ ln þ (16)
1q RT case, since the Maxwell’s rule assumes independence of work
on the shape of path when passing from point D to point G
v 1 1 v h i
and, therefore, is valid in V e P coordinates. When the trans-
½Fðq; TÞ ¼ þ H1 ðqÞ (17)
vq qð1  qÞ RT vq formation of the volume, V, to the concentration (filling frac-
The first term of the right-hand side of Eq. (17) is always tion), q, is not linear, the Maxwell’s rule cannot be applied, and
positive (0 < q < 1), with the minimum value of 4, at q¼1/2. It the plateau should be re-constructed assuming the equality of
means that at the absence of H1 ðqÞ contribution the concen- both chemical potentials and lattice gas pressures in the
tration dependence of ln f (or X) is a steadily increasing points D and G [32], when the corresponding filling fractions
function, and the appearance of the phase separation features satisfy to the following conditions3:
takes place only when the whole right-hand side of Eq. (17) 
mH ða; TÞ ¼ mH ðq0 ; TÞ ¼ mH ðb; TÞ
changes its sign, i.e. when the following condition can be (21)
PH ða; TÞ ¼ PH ðb; TÞ
satisfied:
It has to be noted that the states corresponding to parts of
1 v h i 1 the PCI where vX > 0 can be physically realised, and the seg-
H1 ðqÞ ¼  (18) vq
RT vq qð1  qÞ ments DE (over-saturated a phase; a  q  a0 ) and GF’ (under-
Since the absolute value of the left-hand side in (18) de- saturated b phase; b0  q  b)4 were indeed observed in practice
creases with the temperature, this condition can be satisfied as, for example, was shown by Sieverts for H e Pd system (data
only if the temperature, T, is below a critical temperature, TC, presented by Lacher [14]). The aeb transformation can take
and. vqv ½H1 ðqÞ < 0: place at any X in between X1 and X2. Accordingly, the isotherm
Generally, at the critical temperature, TC, the derivatives can be represented by two segments for a (ADE) and b (F’GC)
v2 phases overlapping at X1  X  X2 as presented by the bold
vq s ðqÞ and vq2 ½ms ðqÞ are equal to zero at q ¼ qC [32]. The values
½m
v

of TC and qC are determined by a specific dependence of H1 ðqÞ solid lines in Fig. 1.


assumed in a PCT model. In doing so, Eq. (16) can be modified The function Xðq; TÞ ¼ ðlnf  ln f0 Þ defined in (15,20) in-
by the replacement of 2Fðq; TÞ with a function Fðq; zÞ where troduces a variable, X, directly related to hydrogen fugacity, f,
or pressure, P. Its specific forms considered below determine
TC concentration dependencies X(q) at various temperatures. In
z¼ (19)
T fact, the pressure (or X) is an independent variable which,
To separate concentration-dependent (shape of the PCI) together with the temperature, determines hydrogen con-
and concentration-independent (vertical shift of the PCI with centration in the solid according to the second form of the PCT
changing the temperature) parts of ln f (Eq. 13), Eq. (15) can be Eq. (2). Thus we introduce a core function, qðX; TÞ, as a function
presented as: which is inverse to Xðq; TÞbuilt according to this or that

Xðq; zÞ ¼ Fðq; zÞ  Fðq0 ; zÞ (20)

where z is determined by Eq. (19), Xðq0 ; zÞ ¼ 0, and the value of 3


Obviously (see Fig. 1), the chemical potentials are equal in the
q0 depends on z only and is equal to qC at z ¼ 1 (T ¼ TC). The
points D (q¼a), B’ (q¼q0) and G (q¼b). For the clarity, in Eq. (21) the
chemical potential is denoted as mH; it is equivalent to the nota-
2
Strictly speaking, the models under consideration deal with tion ms in Eqs. (3),(8).
4
disordered solid solution phase with higher H concentration Generally, only part GF’ of the segment GF corresponding to
which has to be denoted as a’. We denote here and below the positive values of the lattice gas pressure, PH, can appear (see
“hydride” phase as b for the convenience. below).
2744 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1

the two roots has to be found in the course of the numerical


solution.
Furthermore, the regions corresponding to the over-
saturated a- and under-saturated b-phase at z > 1 can be found
by the finding the filling fractions (a0 and b0 ) corresponding to
the maximum and minimum of the X(q) function (points E and
F in Fig. 1). In Lacher model they can be easily calculated by the
solution of the equation vq v
½Lðq; zÞz ¼ 0. But at z > 1.225 the
0
value of b calculated by this way will correspond to the
negative lattice gas pressure and should be increased until the
condition PH  0 will be satisfied (Fig. 2B).
The phase diagram of hydrogen e metal system calcu-
lated according to the Lacher model is presented in Fig. 3 A.
The boundary delimiting two-phase a þ b region (a < q < b) is
symmetric as respect to q0 ¼ 1/2 which is not changed with
Fig. 1 e Examples of X(q,z) plots. the temperature. The diagram also shows the areas within
the a þ b region where existence of the oversaturated a phase
(a(þ); a < q < a0 ) and under-saturated b phase (b(e); b0 < q < b) is
modelling procedure. The two examples, Lacher model and possible. The values of a’ and b’ calculated from the condi-
model of Van der Waals lattice gas, are presented below. v
tion vq ½Lðq; zÞz ¼ 0 show parabolic temperature dependence,
also symmetric as respect to q0 ¼ 1/2. However, at T/
Lacher model TC < 0.816 (z > 1.225) the appearance of negative pressures of
As it was shown by Fowler [13], the effect of phase separation lattice gas becomes possible (dashed line from the right).
on adsorption isotherms appears at H1 ðqÞ < 0 in (13) that can be Accordingly, when the temperature decreases, the b0 value
interpreted as an attraction between the adsorbate species. which, additionally, has to satisfy the condition PH  0,
Fowler assumed H1 ðqÞ ¼ h1 q (h1 > 0) that resulted in the quickly increases approaching the filling fraction, b, in the b
appearance of the phase separation at T < TC where: phase.
Fig. 3B shows temperature dependencies of
h1
TC ¼ (22) X ¼ ðlnf  ln f0 Þ ¼ Lðq; zÞ calculated at q¼b0 (X1) and q¼a0 (X2).
4R
As it can be seen, increase of z (or decrease of the temperature)
Lacher [14] presented H1 ðqÞ ¼ h1 ð1=2  qÞ that yielded the results in a significant monotonic increase of X2 determining
same value of the critical temperature. In our notation (20), the upper (for the coexistence of a- and b-phases) limit of the
within Lacher model the function Fðq; zÞ is equivalent to the deviation of H2 fugacity (pressure) from the value, f0, which
Lacher function Lðq; zÞ introduced by Kierstead [28]: corresponds to the midpoint (X ¼ 0) of Lacher’s PCI and is
    calculated from Van’t Hoff Eq. (14). In contrast, the lower limit,
q
Fðq; zÞ≡Lðq; zÞ ¼ 2 ln þ 8z 1 2  q
=
(23) X1, of this region, despite its more complicated temperature
1q
dependence, remains relatively small (close to zero) negative
For the Lacher model, independently on temperature (or z), value.
q0 ¼ 1/2, and L(q0,z) ¼ 0. Note that the higher limit, X2, can be achieved in practice
The pressure, PH, of the lattice gas in the Lacher model can in the course of H absorption while the lower one, X1, during
be calculated as [31] 5: H desorption. The limitations to satisfy the condition PH  0
  result in the fact that the value of X1 is closer to zero than X2
kTC Cmax lnð1  qÞ
PH ¼   2q2 (24) at the lower temperatures / higher z values (Fig. 3 B,D). It can
U z
explain the commonly-used statement that the desorption
where k is the Boltzmann constant, U is the specific matrix isotherms in two-phase region are closer to the real equi-
volume per a metal atom. Note that for the comparison pur- librium (X ¼ 0 according to the model) than the absorption
poses the constant and positive term kTC Cmax =U can be ones.
omitted. A feature of the Lacher model is a central symmetry of the
Fig. 2 (A, B) presents concentration dependencies of the X PCI as respect to the point q0 ¼ 1/2 and, in turn, the corre-
value (A) determining the shape of PCI (13,20), and the value of sponding symmetry of the H2 e M phase diagram as a whole.
lattice gas pressure (B; Eq. 24) within Lacher model (23). In all It is a direct consequence from the assumption of linear
temperature range the function X ¼ L(q) is odd (i.e., exhibits concentration dependence, H1 ðqÞ  q, of the HeH interac-
anti-symmetry) as respect to q0 ¼ 1/2. The values, a and b, of tion enthalpy. At the same time, in most cases such de-
filling fractions corresponding to a- and b-phases are equal to pendencies are non-linear. In doing so, hydrogen
the lower and higher roots of the equation Lðq; zÞ ¼ 0 solved as concentration in the critical point corresponds to the filling
respect to q at given z > 1. In doing so, a þ b ¼ 1, and only one of fractions below 1/2. Such a behaviour can be explained in the
framework of the general model (13, 14) if to assume that, in
5
Particular case of the interacting lattice gas model, when only addition to attraction, the HeH interaction also includes
attractive interaction between H atoms is taken into account, and repulsion whose contribution increases with the increase of
lattice expansion upon hydrogenation is neglected. the filling fraction.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1 2745

Fig. 2 e Plots of core functions, Xðq; zÞ (A, C), and lattice gas pressures, PH ðq; zÞ (B, D), for Lacher (A, B) and Van der Waals lattice
gas (C, D) models. The values of characteristic concentrations (a, a′, b, b′) are shown for the isotherms at z ¼ 1.4.

Model of Van der Waals lattice gas point of phase separation corresponding to the zeroing of both
The simplest approach accounting the repulsion of the H the first and the second derivatives of the concentration de-
atoms is a consideration of the lattice gas formed by them as a pendencies of ms(q) or X(q) is characterised by the following
Van der Waals one described by the following equation of values of temperature, TC, and hydrogen concentration, qC:
state: 8
>
> 8a* 4h1
! < TC ¼ ¼
a*   27Rb* 27R (27)
PH þ VH  b* ¼ RT (25) >
> 1
VH2 : qC ¼
3
where PH and VH are the pressure and the molar volume of the
Following the notation (20), within Van der Waals lattice
lattice gas, respectively, a* is an attraction constant, and b* is a
gas model the function Fðq; zÞ will be written as.
repulsion constant, or “own” molar volume of the H atoms
 
considered as rigid spheres. q q 27
Fðq; zÞ≡Vðq; zÞ ¼ 2 ln þ2  zq (28)
After transformation in (25) of the pressure, PH, to the 1q 1q 2
chemical potential, ms, and the volume, VH, to the filling frac-
For the determination of the lattice gas pressure, PH, in the
tion, q, we obtain the following solution of Eq. (3):
Van der Waals lattice gas model it is convenient to re-write
    (25) in the form:
q q 4a*
X ¼ ln f  ln f0 ¼ 2 ln þ  * q (26)
1q 1q b RT
  
3 1 8
In doing so, the attractive interaction term is the same as in pH þ 2 vH  ¼ t (29)
vH 3 3
Lacher model (h1 ¼ 2a*/b*), but the repulsion is accounted by
the additional configuration term, q=ð1  qÞ. where pH ; vH and t ¼ 1/z are the ratios of the lattice gas
Within the model of Van der Waals lattice gas [33] similar pressure, molar volume and the temperature to their values in
to Hill e de Boer equation of adsorption isotherms, the critical the critical point. Then:
2746 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1

Fig. 3 e Features of phase diagram of metal e hydrogen system built according to Lacher model (A, B) and model of Van der
Waals lattice gas (C, D). A, C e phase diagrams; the areas between solid lines correspond to the oversaturated a phase (a(þ))
and under-saturated b phase (b(e)). B, D e lower (X1) and higher (X2) limits of X ¼ ðlnf  ln f0 Þ corresponding to the
coexistence of a and b phases.

 Lowering the value of z (from 1.225 to 1.185) when the


8q appearance of negative lattice gas pressures becomes
pH ¼  27q2 (30)
zð1  qÞ possible.
 Faster increase, with the decrease of the temperature, of
The determination of q0 in the Van der Waals lattice gas
the higher limit, X2, of the range of X ¼ ðlnf  ln f0 Þ where a
model is reduced to the task of the reconstructing position of
and b phases co-exist.
the plateau, by the solution of (21) for chemical potentials and
lattice gas pressures to yield the values of a, b and q0.
The model of Van der Waals lattice gas applied by Lototsky
The features of the Van der Waals lattice gas model are
et al. [33] for phase equilibria in metal e hydrogen systems
illustrated by Fig. 2C,D showing concentration dependencies
allows more adequate (as compared to the Lacher one)
of X (C) and lattice gas pressures (D), as well as phase diagram
modelling of the features of the H2 systems with hydride
(Fig. 3C) and ranges of X where co-existence of a- and b-phases
forming metals and alloys, especially, at low hydrogen
can take place (Fig. 3D). The features are similar to the ones for
concentrations.
the Lacher model with the following distinctions:
Further refinement, by a general consideration of
hydrogen subsystem in the solid as an interacting lattice gas
 Slower approach of X(q) to the vertical asymptote (q ¼ 1).
was presented by Marinin et al. using a modified perturba-
 Deformation of the phase diagram, including metastable
tion theory [31,32]. The model, similar to the one by Hara-
a(þ) and b(e) regions, towards the lower filling fractions.
shima et al. [15,16] introduces a positive term proportional to
 As distinct to L(q,z) always equal to zero at q0 ¼ 1/2,
q2 into concentration dependence mS(q); it also takes into ac-
generally V(q0,z) takes non-zero values. However, accord-
count lattice expansion (dilatation) upon insertion of H
ing to (20), X(q0,z) ¼ 0, and the value q0 depends on z
atoms into the metal matrix. The model describes well the
covering the range from 1/3 at z ¼ 1 (T ¼ TC) to 1/2 at z ¼ ∞
phase equilibria at the temperatures close to the critical one
(T ¼ 0).
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1 2747

but yields unrealistic results at low temperatures (decrease Formally, the approach presented by (31,32) requires the
of hydrogen concentration, b, in the b-phase with the in- core function, q(X), to be two-valued, i.e. the segments qa(X)
crease of z at z > 3). Poor accuracy at the lower temperatures and qb(X) should be defined in the whole range of X (from e∞
of the models [15,16] with quadratic dependence mH (q) was to þ∞ ). In practice (see Fig. 1; bold solid lines), the segment
noticed in the review [18]. Due to this reason, the core qa(X) is defined at X ¼ e∞ … X2, and the segment qb(X) at
functions used in this work will be limited by the ones built X ¼ X1…þ∞. Again formally, we assume that for X > X2
according to the Lacher model and the model of Van der qa(X)¼a0 , and for X < X1 qb(X) ¼ b0 (Fig. 1) that can be inter-
Waals lattice gas described above. preted as the maximum hydrogen concentration in the a-
phase and minimum hydrogen concentration in the b-
Distribution function and its combination with the core one phase, respectively, which correspond to a given tempera-
ture (or z value).
Assuming variations of ln f0 due to compositional in- Fig. 4 illustrates the procedure of building of the desorp-
homogeneity of the hydride forming alloys, we can introduce tion (DES) and absorption (ABS) isotherms (solid lines). The
a probability density (distribution) function, j, of the value of plots shown by points represent the segments qa(X) and qb(X)
ln f0, or X.6 Further combination of the distribution function of the core function; the dashed line plots show probability
with the modelled core function qðXÞ inverse to XðqÞ intro- density functions, jD(X) and jA(X) for hydrogen desorption
duced in (15,20), to yield the calculated non-ideal PCI, is and absorption, respectively. Note that within the model
possible by the application of two procedures. presented in this work hydrogen desorption is assumed to
The first one is a convolution of the distribution function, present the true equilibrium, and the mode of distribution
j(X), with a core function qðXÞ characterised by a discontinuity function, jD(X), for the desorption is equal to zero. The mode
at X ¼ 0 when the abrupt change of the filling fraction from a to of the absorption distribution function, jA(X), is shifted
b takes place. This approach was used in Refs. [33,37] taking from the one of the desorption distribution towards higher
the core function as the one inverse to Lacher [37] or Van der pressures, by the value, Xh, related to hysteresis energy loss,
Waals lattice gas [33] models, with building the plateau at h, as
X ¼ 0 and omitting the over- and under-saturated regions.
h
Another procedure implies a weighted superposition of the Xh ¼ (33)
RT
integrated distribution functions and hydrogen concentra-
tions in a- and b-phases calculated according to the “ideal” flat The model features related to the hysteresis will be dis-
plateau PCI. Fujitani et al. [36] calculated hydrogen concen- cussed in the next Section.
tration in the sloping plateau region using two integral dis- Similarly to the most of previously published PCT models
  describing sloping plateaux [33,36,37,40], both jD(X) and jA(X)
tribution functions of Y ¼ ln PP0 scaled by (bea). Park et al. are presented in"Fig. 4 as Gaussian distribution densities.
#
[40] started from the composition distribution function com- 1 ðX  X0 Þ2
GðXÞ ¼ pffiffiffiffiffiffi exp  (34)
bined with hydrogen concentrations in a- and b-phases in the 2ps 2s2
vicinities of the phase boundaries to obtain the correlation
between the concentration and plateau pressure. Both ap-
proaches represent a simplified convolution procedure, or
pseudo-convolution, which can be easier (than the true
convolution) realised numerically, especially for the cases
when the probability density distribution functions are not
Gaussian.
Within the model presented in this work we use the second
approach when the resulting non-ideal PCI, Q(X), is calculated
starting from the core function, q(X), and the integral distri-
Z X
bution functions, JðXÞ ¼ jðXÞdX, as:
∞

QD ðXÞ ¼ ð1  JD ðXÞÞ qa ðXÞ þ JD ðXÞqb ðXÞ (31)

QA ðXÞ ¼ ð1  JA ðXÞÞqa ðXÞ þ JA ðXÞqb ðXÞ (32)

where subscripts a and b represent hydrogen concentrations


in the a- and b-phases, respectively, calculated as the corre-
sponding segments of the common core function, q(X). Both Fig. 4 e Model dependencies Q(X) for H concentrations in a-
JD(X) and JA(X) are the integrated distribution functions, (qa) and b / a’ (qb) phases and pressure e composition
jD(X) and jA(X), representing the probabilities of the appear- isotherms for H absorption (ABS) and desorption (DES)
ance of the b-phase in the course of desorption (D) and ab- calculated assuming Gaussian probability density
sorption (A), respectively. functions, j(X), as jA and jD, respectively. Core function:
Van der Waals lattice gas, TC ¼ 500 K, T ¼ 416.67 K
6
The distribution functions j(X) and j[ln(f0)] will differ only by (z ¼ 1.2); hysteresis loss h ¼ 2.8 kJ/mol H (Cmax ¼ 1H/M);
the values of their modes. standard deviations sA ¼ 0.2, sD ¼ 0.02.
2748 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1

where X0 and s are the mode and the standard deviation,


respectively. This commonly used approach is quite conve-
nient and easy in the numerical implementation. At the same
time, for the improvement of the modelling accuracy in the
transient concentration regions corresponding to the begin-
ning and the end of the plateau, the distribution function
needs to be further modified.
The task of modelling the non-ideal PCI characterised by
plateau slope and (generally, independent on it) smooth
transitions between a-, (a þ b)- and b-regions resembles the
modelling of peak shape in spectroscopy [49,50], but in the
integral form. Herein, several factors influencing on the peak
shape can be accounted by a combination of two distribution
functions, viz. modified Gaussian (35,36) and Lorentzian
(37,38):
rffiffiffiffiffiffiffiffi "  2 #
1 ln2 X  X0
jG ðXÞ ¼ exp  ln2 (35)
w p w

( "pffiffiffiffiffiffiffiffiffiffi #!)
1 2ln2ðX  X0 Þ
JG ðXÞ ¼ 1 þ erf (36)
2 w

, "
 2 #
jL ðXÞ ¼ 1 pw 1 þ X  X0 (37)
w

1 1
JL ðXÞ ¼ þ   (38)
2
p arctan XX
w
0

where 2w is the full width at the peak half maximum (FWHM).


Note that jG(X) ¼ G(X) when s ¼ pffiffiffiffiffiffi
w ffi
2ln2
.
A frequently used method of the combination of the
Gaussian and Lorentzian distributions is their weighted su-
perposition (pseudo-Voight distribution) [49,50]:

jV ðXÞ ¼ hjL ðXÞ þ ð1  hÞjG ðXÞ (39a)

JV ðXÞ ¼ hJL ðXÞ þ ð1  hÞJG ðXÞ (39b)

where 0  h  1 is a contribution of the Lorentz profile.


The plots of pseudo-Voight distribution functions (39) with
various Lorentz contributions, h, are shown in Fig. 5A. The
differential and integral curves exhibit symmetric and anti-
symmetric shapes, respectively, as respect to X ¼ X0 with
the slope of the integral curve approaching to constant in the Fig. 5 e Influence of the distribution parameters on the
centre. At the same time, the increase of the Lorentz contri- features of integral (solid lines) and differential (dot lines)
bution results in the smoother transition from the “horizon- modified pseudo-Voight distribution function (39-41). A e
tal” parts of the integral curve at X / ±∞ to the sloping one at Lorentz contribution, h, at A ¼ 0, s ¼ 1; B e asymmetry, A,
X ¼ X0. at h ¼ 0.5, s ¼ 1; C e slope factor, s, at A ¼ 0, h¼0.5. For all
The pseudo-Voight distribution (39) was used for the fitting the plots X0 ¼ 0, w0 ¼ 0.3; the values of the varied
of PCI’s in real metal e hydrogen systems according to the parameters are shown as curve labels.
procedures (31,32). The fitting was characterised by better
accuracy and shorter calculation time than the one in the may include (i) existence of at least two “non-ideal” hydride
previously used approach [33,37] of the convolution of a core phases with close values of the mean plateau pressures, (ii)
function with the Gaussian distribution (34). However, the non-linear concentration dependence of hysteresis originated
modelling of some isotherms was still inaccurate. It was from the features of stress effects during the growth of b-hy-
manifested as non-linear behaviour of the experimentally dride in the matrix of a-phase, etc.
observed plateau slope in two-phase (a þ b) region that can be To further improve accuracy of the modelling of the iso-
interpreted as asymmetry of the distribution density as therms characterised by the non-linear asymmetric plateau
respect to the plateau midpoint. Probable origins of this effect slope (first of all, for the absorption in the systems H2 e BCCeV
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1 2749

alloy), an asymmetry parameter was introduced in the dis- which require quantification within the present model have to
tribution (39). The procedure was taken from Ref. [49] intro- be noted:
ducing the variation of the width parameter, w, according to
the formula:  Hysteresis is observed in only two-phase (a þ b)-regions of
pressure-composition isotherms and is absent in the re-
2w0
w¼ (40) gions of a- and b-solid solutions.
1 þ exp½AðX  X0 Þ
 At temperatures far below the critical one, the observed
where w0 is a width parameter at X ¼ X0, and A is asymmetry value of PA/PD (fA/fD) ratio corresponds to the constant
factor. change of free energy associated with the hysteresis en-
Fig. 5B shows the influence of the asymmetry, A, on the ergy losses:
behaviour of the distribution functions. At A ¼ 0 the curves
are the same as the ones shown in Fig. 5A. The non-zero A fA
DGh ¼ RT ln ¼ RT$Xh (42)
values result in the deformation of the distribution density fD
function with insignificant shift of the mode from X ¼ X0.
However, the value JV ðX0 Þ ¼ 0:5 remains unchanged, so as  The hysteresis decreases when increasing the temperature
X0 is still the median, and introduction of additional term and vanishes when it approaches the critical one.
responsible for keeping the peak maximum constant [50] is
not necessary in our case. At the same time, the asymmetry As it was mentioned above, our model assumes equilib-
results in the different curvatures of the integral distribu- rium at X ¼ 0 that corresponds to the mode (or median when
tion function when approaching to X0 from the lower the asymmetry is taken into account) of ln f0 distribution for
and higher X values. At A < 0 slower growth of JV ðXÞ is the desorption isotherm (f0 ¼ fD D D D D
0 ; parameters w0 , h , A , s ).
A
observed at X>X0, and at A>0 the slower growth takes place Assuming absorption plateau pressure (f0 ) to also have a
at X < X0. random nature with own distribution parameters (wA A A
0, h , A ,
A
The application of the distribution (39,40) allowed us to s ), we similarly build the absorption branch of the PCI start-
obtain very good fit of the PCI’s characterised by sloping ing from the same core function, q(X), and the absorption
plateaux but not close to the flat ones. Further improvement distribution, JA(X); X ¼ ln f e ln fA 0 . We assume that the me-

of the modelling was achieved by the modification of the dian of the absorption distribution function, JA(X), is shifted
asymmetric pseudo-Voight distribution assuming that the from X ¼ 0, by the value, Xh, related to hysteresis energy loss,
width parameter is different for the Lorentz and Gauss com- h, as shown in (33). The h value is assumed to be an average
ponents. For the clarity, we ascribe the width parameter, w0, hysteresis energy loss also having the random nature with the
to the Lorentz component, wL0, and the Gauss one, wG 0 , is corresponding distribution parameters.
assumed to be related to it via slope factor, s: To account the disappearance of the hysteresis when
approaching to critical temperature, we assume the h value to
wL0 ¼ w0 be equal to DGh until the latter will be lower than the
(41)
wG0 ¼ s$w0 ; 0  s  1
maximum excessive chemical potential of H atoms, RT,X(q,z),
At s ¼ 1 we have a commonly used asymmetric pseudo- which can be calculated from the core model (Section Core
Voight distribution where the slope in the centre and the function: one segment PCI built according to a lattice gas
curvatures of the integral distribution function correlate with model). It happens when the value X2 ¼ X(a0 ,z) (point E in
each other depending on the width parameter, w0, and the Fig. 1) corresponds to the local maximum of the model
Lorentz contribution, h. When s < 1, the Gaussian part results isotherm. It is temperature-dependent monotonically
in a steeper growth of the integral distribution function in the increasing with the decrease of the temperature, or increase
proximity to X0 approaching to the vertical at s / 0 of z (see Fig. 3B,D). As it can be seen from the example shown
(G(X) / d(X)) without significant change of the curvature in Fig. 6, the graph Xh(z) plotted according to (42) crosses the
determined by the Lorentzian part (Fig. 5C). graph X2(z) at a certain value z ¼ z0 which depends on the
critical temperature (19) and DGh (42). The resulting mean
hysteresis, Xh, will be equal to the value corresponding to the
Hysteresis top of the hatched area, or a minimum of a hysteresis free
energy, DGh, (assumed to be constant; 1500 J/mol H in Fig. 6)
Hysteresis is a general feature of metal e hydrogen systems and an excessive chemical potential of hydrogen in over-
resulting in higher plateau pressures (fugacities), PA (fA), saturated a-solid solution, RT,X2(z).
observed for H absorption than those, PD (fD), for H desorption.
Hysteresis was studied in a number of experimental and Temperature dependencies of the model parameters
theoretical works [48,51e54, etc.]; its main origin was associ-
ated with the stresses which appear in the course of growth of Temperature dependence of the plateau fugacity, f0, corre-
hydride nuclei inside the matrix of the metal or alloy having sponds to the Van’t Hoff Eq. (14) while the shape of the core
lower molar volume. Hysteresis results in energy losses in the function, q(X), depends on the temperature via the value of
operation of MH-based heat engines and has to be taken into z ¼ TC/T according to Lacher (23) or Van der Waals lattice gas
account in their modelling [4]. (28) models.
Discussion about mechanisms of the hysteresis is out of Some systems, e.g. H2 e PrNi5 [55], exhibit temperature
scope of this work. However the following general features dependent plateau slope when at a certain temperature, T0,
2750 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1

Fig. 6 e Higher (X2) limit of X ¼ ðlnf  ln f0 Þ corresponding to


Fig. 7 e Temperature dependencies of the slope factor, s
the coexistence of a and b phases within Van der Waals
(44) at various Pearson correlation coefficients, rSH (the
lattice gas model (Fig. 3D). The straight lines represent
values presented as curve labels).
hysteresis, Xh ¼ DGh/RT, calculated at TC ¼ 500 K starting
from the values DGh [J/ mol H] shown by the line labels. Top
of the hashed area corresponds to the mean hysteresis
value assumed within the present model. Within the model this effect is taken into account by the
assumption that the distribution parameters (d ¼ w0, h, or A)
for the absorption and desorption are related with each other
via mixing factor, m, as:
the plateau becomes close to flat, with the increase of the
slope as the temperature decreases or increases. This effect dD ¼ mdD0 þ ð1  mÞdA0
(45)
was explained by the variations of DSo and DHo caused by the dA ¼ mdA0 þ ð1  mÞdD0
alloy inhomogeneities [39]. In Ref. [33] we quantified this
phenomenon by the assumption that both DSo and DHo are 1
m ¼ ½2  expðMðb  aÞÞ (46)
characterised by the Gaussian distributions with standard 2
deviations sS and sH, respectively. Then the standard devia- where indexes D and A relate to the distribution parameters
tion, s, of ln P0 (or ln f0), according to Van’t Hoff Eq. (14), will be for desorption and absorption, respectively, M  0 is the
calculated as: mixing coefficient, a and b are hydrogen filling fractions in the
a and b phases, respectively (assumed to be equal to qC at
s2S s2 2r sS sH T  TC). In doing so, the mixing factor, m, varies between ~1 at
s2 ¼ þ 2 H 2  SH2 (43)
R 2 R T R T T / 0 (a / 0, b / 1) and ½ when T / TC. In the latter case
where rSH is the Pearson correlation coefficient between DSo (M > 0) the corresponding distribution parameters for the ab-
and DHo. sorption and desorption will approach to each other while in
In the present model we apply (43) to the slope factor, s, the former (when M is close to zero) they will take their values
since the changes in the plateau slope with the temperature close to the initial independent ones, dA D
0 and d0 .

seem to be not related to the shape of PCI in the transient Within this model, the temperature dependencies (44e46)
concentration region (mainly determined by the Lorentz are applied to both absorption and desorption distribution
contribution). For the clarity, we present the temperature parameters, with common parameters T0, rSH and M.
dependence, s(T), in the form:
Modelling of PCI with several plateau segments
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
T2 T0 We use a simplified Kierstead approach [28] for the modelling
sðTÞ ¼ s0 1 þ 02  2rSH (44)
T T of the “multiplateau” PCT diagrams, by representing the
Fig. 7 presents the dependencies of s(T)/s0 calculated ac- resulting hydrogen concentration, C, as.
cording to (44) when the correlation coefficient, rSH, varies
X
k
between 0.05 and 0.99. When rSH is close to one, the slope C ¼ Cmax Wi Qi ðT; PÞ
changes linearly with 1/T approaching to zero at T ¼ T0. At i¼1 (47)
P
k
rSH / 0 the plateau slope monotonically increases when the Wi ¼ 1
i¼1
temperature decreases; the intermediate values of the corre-
lation coefficient result in the appearance of minimum non- where Cmax is the overall maximum hydrogen concentration,
zero plateau slope at T > T0. i ¼ 1…k is the number of the plateau segment with the weight
Finally, the shapes of absorption and desorption PCI’s must Wi, Qi(T,P) is a filling fraction for the i-th segment calculated
coincide when approaching to the critical temperature. according to the model as described above, taking
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1 2751

temperature, T, and hydrogen pressure, P, (recalculated to the core function (A) exhibit faster approach to the maximum H
fugacity, f) as independent variables. concentration (Cmax ¼ 1 in both cases) than for the Van der
Waals lattice gas one (B).
Summary The modelled isotherms show three segments charac-
terised by the total lengths proportional to Wi,Cmax and by
Tables 1 and 2 present the lists of the fitting parameters for the pressure plateaux (desorption) at P z f0. The plateau slopes
modelling of single pressure e composition isotherm (PCI) and correlate with the values of w0(i), h(i) and s(i); and the curvature
a set of the isotherms for different temperatures (polytherm; of the transition to / from the i-th plateau depends on w0(i) and
PCT), respectively. The tables refer to the main equations h(i). Variation of the values of the width parameter, w0, slope
which include the listed parameters within this model, as well factor, s, and asymmetry, A, allows to build the asymmetric
as present basic constrains in the parameters variation and flat or sloping plateaus as, for example, was observed for the
their number for the fitting of PCI (PCT) with several plateau systems H2 e Pd0.85Ni0.15 [41] (similar to first absorption
segments, i ¼ 1…k. segment in Fig. 8), or H2 e VH [48] (similar to the second ab-
The number of the fitting parameters in the present model sorption segment).
is quite big, 12 k for the PCI and 16 k for the PCT. For the Hysteresis is observed in the plateau region, with PA/PD in
comparison, the number of fitting parameters in Kierstead’s the plateau midpoint proportional to the value of the hyster-
PCT model is from 4 k [28] to 6 k [29], the model by Lototsky esis energy loss parameter, DGh, as well as the deviation of the
et al. (2003) accounting the temperature dependent plateau temperature, T, from the critical one, TC.
slope has 7 k fitting parameters [33], and the model by As it can be seen from the second desorption and the third
Ledovskikh et al. requires from 8 [24] to 18 [25] parameters for absorption segments, the model can describe even flat
the simulation of single PCI. At the same time, the present plateaux with smooth transitions to single-phase regions
model allows the simultaneous fitting of both absorption and assuming low values of the slope factor, s.
desorption datasets while in the previous models they have to The modelled PCT behaviour is illustrated by Fig. 9 (the
be processed separately that implies, for example, the varia- parameters are listed in Table 4), for three plateau segments
tion of 14 k fitting parameters using the model [33] (see, e.g. described by Lacher (A) and Van der Waals lattice gas (B) core
Ref. [46]). For the modelling of only desorption (or absorption) functions.
PCT without taking hysteresis and mixing into account, this In addition to the features of the single isotherm discussed
model requires 11 fitting parameters per a plateau segment above (Fig. 8), the model PCT (Fig. 9) exhibits (a) temperature
(see examples for LaNi4.78Sn0.22Hx, Section Validation). dependence of the desorption plateau pressure according to
Moreover, some fitting parameters in the present model Van’t Hoff Eq. (14) and parameters DSo and DHo for the cor-
are necessary only in special cases. For example, the slope responding i-th segment; (b) faster increase of hydrogen
factor, s, should be varied only when a PCI segment can equilibrium pressure with the temperature at high H con-
exhibit close to flat plateau; in other cases it could be fixed as centrations for the core function described by Van der Waals
s ¼ 1. lattice gas model; (c) decrease of PA/PD in the plateau midpoint
Fig. 8 shows the model three-segment PCI’s (T ¼ 300 K) built with the increase of the temperature and disappearance
assuming values of the parameters listed in Table 3. It can be of the hysteresis when approaching to critical temperature;
seen that different core functions result in the similar (d) coinciding the absorption and desorption isotherms at
isotherm features, but the isotherms built assuming Lacher T / TC.

Table 1 e Fitting parameters (PCI modelling).


Group Notation Description [units] Equations General Number (for k
constrains plateau segments)
General Cmax Maximum hydrogen concentration [a] 6, 7, 47 Cmax > 0 1
P
Wi Segment weight [-] 47 Wi  0; ki¼1 Wi ¼ 1 ke1b
Core function TC(i) Critical temperature [K] 19 TC > 0 k
f0(i) Plateau fugacity [atm] 14 f0 > 0 k
Distribution DGh(A,i) Hysteresis energy loss [J / mol H] 42 DGh  0 k
(Absorption) w0(A,i) Width parameter, 1/2 FWHM at 35e38, 40 w0  0 k
X ¼ 0 (f ¼ f0) [-]
h(A,i) Contribution of Lorentz profile [-] 39 0h1 k
A(A,i) Asymmetry [-] 40 e k
s(A,i) Slope factor [-] 41 0s1 k
Distribution w0(D,i) Width parameter (1/2 FWHM at 35e38, 40 w0  0 k
(Desorption) X ¼ 0 (f ¼ f0) [-]
h(D,i) Contribution of Lorentz profile [-] 39 0h1 k
A(D,i) Asymmetry [-] 40 e k
s(D,i) Slope factor [-] 41 0s1 k
Total 12 k
a
Any units proportional to H/M atomic ratio (5,6).
b
For k ¼ 1 W1 ¼ 1 (not varied).
Table 2 e Fitting parameters (PCT modelling).
2752

Group Notation Description [units] Equations General constrains Number (for k plateau segments)
General Cmax Maximum hydrogen concentration [a] 6, 7, 47 Cmax > 0 1
P
Wi Segment weight [-] 47 W i 0; ki¼1 Wi ¼ 1 ke1b
Core function TC Critical temperature [K] 19 TC>0 k
DSoi Hydrogenation entropy [J/(mol H2 K)] 14 DSo<0 k
DHoi Hydrogenation enthalpy [J / mol H2] 14 DHo < 0 k
Distribution (absorption) DGh(A,i) Hysteresis energy loss [J / mol H] 42 DGh  0 k
w0(A,i) Width parameter, 1/2 FWHM at X ¼ 0 (f ¼ f0) [-] 35e38, 40 w0  0 k
h(A,i) Contribution of Lorentz profile [-] 39 0h1 k
A(A,i) Asymmetry [-] 40 e k
s(A,i) Slope factor [-] 41 0s1 k
Distribution (desorption) w0(D,i) Width parameter (1/2 FWHM at X ¼ 0 (f ¼ f0) [-] 35e38, 40 w0  0 k
h(D,i) Contribution of Lorentz profile [-] 39 0h1 k
A(D,i) Asymmetry [-] 40 e k
s(D,i) Slope factor [-] 41 0s1 k
Distributions temperature dependence T0(i) Characteristic temperature related to 44 T0 > 0 k
the minimum plateau slope [K]
rSH(i) Pearson correlation coefficient between DSo and DHo [-] 44 0  rSH  1 k
Mi Mixing coefficient [-] 45,46 M0 k
Total 16 k
a
Any units proportional to H/M atomic ratio (5,6).
b
For k ¼ 1 W1 ¼ 1 (not varied).

Wi

f0(i)
TC(i)

s(D,i)
s(A,i)

h(D,i)
h(A,i)
Cmax

A(D,i)
A(A,i)

w0(D,i)
w0(A,i)
DGh(A,i)
Notation

1500
500

1
0
1
0
1

0.3
0.3
0.7
0.1
0.1
0.4
parameters are listed in Table 3.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1

1500
450

1
1
-1
1
2

0.1
0.5
0.1
0.5
0.5
0.3
Value

Segment number (i)


Waals lattice gas (B) models. The common model
Fig. 8 e Model isotherms at T ¼ 300 K for hydrogen

Table 3 e Parameters of the model isotherms (Fig. 8).

1000

0
0
10
350
3

0.1
0.7
0.2
0.3
0.3
0.3

0.15
using the core function according to Lacher (A) and Van der
absorption (solid lines) and desorption (dashed lines) built
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1 2753

By the variation of the model parameters, it is possible to


Table 4 e Parameters of the model PCT (Fig. 9).
change the features of the model isotherms in very wide
range. A reader can try it using the file HyPCT501.xlsm Notation Value
(worksheets ModelsPCI, ModelPCTLacher and ModelPCTVdW) Cmax 1
presented in the Supplementary information. Segment number (i)
1 2 3
Wi 0.4 0.3 0.3
TC 500 450 400
DSoi -110 -105 -100
Software DHoi 35,000 30,000 25,000
DGh(A,i) 1500 1500 1500
The model described above was implemented for PC (32 bit w0(A,i) 0.2 0.5 0.7
h(A,i) 0.7 0.6 0.5
Windows operating system) as a stand-alone package entitled
A(A,i) 0 -1 1
“HyPCT 5.0” and written in Microsoft Visual Basic, version 6.0.
s(A,i) 0.1 0.5 1
The demo version presented in the Supplementary Informa- w0(D,i) 0.02 0.03 0.05
tion includes: h(D,i) 0.6 0.5 0.4
A(D,i) 0 1 -1
 Main calculation procedures in ActiveX calculation module s(D,i) 0.05 0.1 1
HyPCT501.dll; Mi 1 0.7 0.3
T0(i) 600 450 400
rSH(i) 0.25 0.5 0.95

 User interface in Macro-enabled Excel workbook


HyPCT501.xlsm;
 User manual, file HyPCT501.pdf.

The demonstration software shows features of the


modelled PCI and PCT plots built starting from the sets of
model parameters (Tables 1,2) which can be introduced by a
user for up to five plateau segments. It also allows the fitting of
experimental PCI or PCT datasets containing both H absorp-
tion and desorption data. The user can change the fitting pa-
rameters manually, or carry out automated refinement
procedure. In both cases the software shows the PCI / PCT
plots corresponding to the actual values of the fitting param-
eters, as well as the total squared deviations of the experi-
mental points from the calculated curves.
Since the slope of a model isotherm, ln P(C), can vary from
0 (in plateau region) to þ∞ (when C approaches 0 or Cmax), the
refinement procedure uses minimisation of sum, SR, of the
squared shortest distances, dk, of the experimental points from
the calculated PCI curves built in coordinates C/Cmax e ln P
(see Fig. 10):

X
N
SR ¼ d2k ¼ min (48)
k¼1

where N is the total number of the experimental points. As it


can be seen from Fig. 10, the distance, d, can be considered as a
vector sum of the relative deviations of the pressure,
D(lnP) z DP/P, and hydrogen concentration, DC/Cmax, corre-
sponding to the experimental point and the closest point on
the calculated PCI. In doing so, the goodness of the fit is
expressed via the factor, Rf, which can be interpreted as a
relative mean-square fitting error:
Fig. 9 e Model isotherm sets (the temperatures [K] are
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
specified as curve labels) for hydrogen absorption (solid
SR
2
2
lines) and desorption (dashed lines) built using the core Rf ¼ z DP=P þ DC=Cmax (49)
N ðN  1Þ
function according to Lacher (A) and Van der Waals lattice
gas (B) models. The common model parameters are listed The refinement algorithm is based on the gradient mini-
in Table 4. misation method with a changeable step. Each fitting
2754 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1

Further details about the model software are presented in


the Supplementary information, file HyPCT501.pdf.

Validation

For the validation of the model the author used experimental


datasets for the systems H2 e LaNi4.8Sn0.2 (P ¼ 0.01e150 atm,
T ¼ 0e240  C, one plateau segment [57]), “fresh” and thermally
aged LaNi4.78Sn0.22 (P ¼ 0.01e150 atm, T ¼ 19e203  C, one or
two plateau segments first of which is hidden due to too low
plateau pressures [27]), H2 e TiFe (P ¼ 0.3e100 atm,
T ¼ 0e70  C, two plateau segments [58]), and H2 e bcc-V alloy
(P ¼ 0.01e25 atm, T ¼ 30e90  C, two plateau segments first of
which is hidden [59]). The data [27,57,58] were taken from the
Fig. 10 e Schematic representation of distances, d, of figures while for the vanadium alloy (V0.75Ti0.1Zr0.075Ni0.075)
experimental points (triangles) from the calculated the author used the original data partially presented in
isotherm (the closest points are shown by stars). Ref. [59].
The results presented below (Table 5, Figs. 11e15) were
calculated assuming the core function for Van der Waals lat-
parameter, Y, varies between the minimum, Ymin, and the tice gas model (28). The comparison of main experimental and
maximum, Ymax, which have to be set by a user. To avoid exit calculated PCT properties (reversible hydrogen storage ca-
of the parameter, Y, out of the range, (Ymin…Ymax), the varia- pacities, critical temperatures, dehydrogenation entropies
tion is applied to the transformed value, Y, e related to the
and enthalpies, and hysteresis) is presented in Table 6.7.
original one as:
 
e ¼ Log Y  Ymin H2 e LaNi5exSnx
Y
Ymax  Ymin
  (50) Example of the refinement of a separate absorption e
Ymin þ Ymax exp Ye
Y¼   desorption isotherm (T ¼ 373.15 K) for the H2 e LaNi4.8Sn0.2
1 þ exp Ye
system is shown in Fig. 11. The refinement was found to be
Numerical variables in the calculation procedures were set sensitive to the variation of all the fitting parameters except
as Double type; the calculation accuracy (machine Epsilon) for asymmetries, A, which were fixed to zero for both ab-
was assumed to be 5,10e16, and the minimal variations of the sorption and desorption.
fitting parameters during refinement session were set as As it can be seen, the fitting is very good. The deviations of
5,10e15. the experimental points from the calculated curves (see right
Errors of the refined fitting parameters were assumed to be graph in Fig. 11) are maximal in the plateau regions, but the
equal to their increments or decrements resulting in ~1% in- absolute values of the corresponding distances, d, do not
crease of the total squared deviation, SR. In the case of the exceed 0.02. The goodness of the fit corresponds to Rf ¼ 0.0017,
different values of the corresponding increments and decre- and the calculated average concentration error corresponded
ments, the error was ascribed to the biggest one. to DC/Cmax ¼ 0.007 where DC is the mean-squared absolute
The input (experimental) and output (calculated) PCT data value of the difference between experimental, Cexp, and
present hydrogen pressures (atm), concentrations (any units calculated, Ccalc(Pexp) concentrations taken at the same pres-
proportional to H/M ratio) and temperatures (K). In all model sure, Pexp. Despite the fitting was done only for one absorption
calculations the pressures, P, are converted to fugacities, f, / desorption isotherm, it yielded quite accurate value of the
and vice versa, using the SoaveeRedlicheKwong (SRK) equa- critical temperature, TC; the extrapolation is based on the
tion of state [56]. observed hysteresis.
The starting experimental PCT data should not be Still good fitting (Fig. 12; Rf ¼ 0.0019, jdjmax z 0.11, DC/
compulsory presented as isotherms, but for the clear visual- Cmax ¼ 0.0407) was achieved for the polythermal experimental
isation of the fitting process it is desirable. PCT dataset, though some systematic deviations of the
Despite some inconvenience of working with this demon- experimental and calculated data at T > 423 K were observed.
stration software which requires, for example, manual prep- The fitted value of the critical temperature, TC ¼ 487 K, simi-
aration of the PCI or PCT data in the Excel workbook for their larly to our previous modelling work [33] (TC ¼ 492 K), was
correct plotting, it is fully functional and can be used for fitting found to be lower than the one estimated experimentally
of the experimental PCT data for new MH materials for
7
hydrogen storage, compression and other applications. Once The reversible hydrogen storage capacities are presented in
refined, the model parameters can be used for the building the pressure e temperature ranges typical for the applications: H2
calculated PCT diagram which can be further incorporated in compression for LaNi5exSnx and H2 storage for other materials.
The experimental values of dehydrogenation entropies and en-
a software for precise modelling of heat-and-mass transfer in
thalpies and hysteresis (experimental and calculated) are related
MH beds. to the midpoint of corresponding plateau segments.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1
Table 5 e PCT fitting parameters (Figs. 12e15).
Parameter System
H2eLaNi4.8Sn0.2 H2eLaNi4.78Sn0.22 H2eLaNi4.78Sn0.22 aged H2eTiFe H2eV75Ti10Zr7.5Ni7.5
Segment 1 Segment 2 Segment 1 Segment 2 Segment 1 Segment 2
a
Cmax [ ] 7.60(5) 7.28(5) 6.52(2) 1.12 (2) 1.809(2)
W 1(eb) 1(e) 0.139(2) 0.861(e) 0.5862(e) 0.4138(5) 0.47525(e) 0.52475(2)
TC [K] 487(3) 633(12) 692(254) 566(10) 778(14) 377(19) 381(7) 509(1)
DSo [J mol He1 e1
2 K ] e106.90(3) e109.17(2) e150(18) e99.62(5) e107.65(6) e128.6(5) e138(1) e176.553(8)
DHo [J mol He1
2 ] e33,913(10) e35,097(6) e100,090(12,850) e34,270(40) e28,660(16) e32,889(16) e69,710(440) e54,008(2)
DGh 501(9) e e e 2146(39) 900(800) 0(e) 3691(68)
wA0 0.109(1) e e e 0.14(2) 0.5(2) 0.1(1) 0.80(1)
hA 0.43(3) e e e 0.58(6) 0.9(1) 0.5(5) 0.7(3)
AA 0(e) e e e e3(2) 3.0(9) 0(e) e2.1(2)
sA 0.8(3) e e e 0(e) 0.5(2) 1(e) 0.99(5)
wD0 0.132(2) 0.189(9) 0.5(e) 1.15(2) 0.169(8) 0.11(4) 0.1(1) 0.10(4)
hD 0.43(8) 0.57(6) 0.5(e) 0.82(2) 0.398(3) 0.5(2) 0.5(5) 0.7(1)
AD 0(e) e0.8(2) 0(e) 0.01(6) e0.1(2) e0.5(5) 0(e) 1.0(2)
sD 0.62(9) 0.50(2) 1(e) 0.86(4) 0(e) 0.5(5) 1(e) 0.20(3)
M 2(2) e e e 1(e) 1(e) 0.5(e) 0.73(4)
T0 156(57) 60(5) 0(e) 48(18) 0(e) 0(e) 0(e) 349(17)
rSH 1(e) 0.90(4) 0(e) 0.98(2) 0(e) 0(e) 0(e) 0.30(4)
Rf 0.0019 0.0015 0.0016 0.0034 0.0017
DC/Cmax 0.0407 0.0212 0.0107 0.0112 0.0042
a
H/AB5 for H2eLaNi5exSnx, H/M for other systems.
b
Not refined.

2755
2756 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1

Fig. 11 e Experimental points and calculated PCI curves for


H2 absorption (filled symbols, solid line) and desorption
(open symbols, dashed line) in LaNi4.8Sn0.2. The graph from
the right shows the corresponding shortest distances, d, of
the experimental points to the calculated curves in DC/
Cmax e ln P coordinates. Fitting parameters: Cmax ¼ 7.49(2)
H/AB5, TC ¼ 513.1(5) K, f0 ¼ 7.106(2) atm, DGh ¼ 502(9) J mol
H¡1, wD D D
0 ¼ 0.1383(9), h ¼ 0.431(4), s ¼ 0.52(3),
A A A
w0 ¼ 0.103(5), h ¼ 0.425(8), s ¼ 0.48(2). The asymmetries,
AD ¼ AA ¼ 0 were not refined.

Fig. 13 e Experimental points and calculated PCI curves


(the temperatures in K are shown as labels) for H2
desorption from hydrogenated LaNi4.78Sn0.22 before (A) and
after (B) thermal aging under H2 pressure. The graph from
the right shows the corresponding shortest distances, d, of
the experimental points to the calculated curves in DC/
Cmax e ln P coordinates. Fitting parameters are listed in
Table 5.

Further validation of the model on Sn-substituted LaNi5


was done by the processing of the experimental H2 desorption
Fig. 12 e Experimental points and calculated PCI curves PCT data for LaNi4.78Sn0.22Hx before and after thermal aging at
(the temperatures in K are shown as labels) for H2 T > 460 K and H2 pressure P > 150 bar for 1000 h [27]. The pro-
absorption (filled symbols, solid lines) and desorption cessing of both datasets (Fig. 13) yielded very good fit
(open symbols, dashed lines) in LaNi4.8Sn0.2. The graph (Rf  0.0016, jdjmax z 0.05). The comparison of the data for
from the right shows the corresponding shortest distances, LaNi4.8Sn0.2 (Fig. 12; Table 5, first column) and non-aged
d, of the experimental points to the calculated curves in DC/ LaNi4.78Sn0.22 (Fig. 13A; Table 5, second column) shows that
Cmax e ln P coordinates. Fitting parameters are listed in slight (10%) increase of tin content results in the lowering of
Table 5. the maximum hydrogen concentration, Cmax, and the decrease
(513 K [57]). We note that the lower values of TC calculated in of hydrogen equilibrium pressures. The strongest substitution
Ref. [33] and this work are in a good correspondence with the effect was observed for the critical temperature that increased
results obtained by Laurencelle et al. (TC ¼ 490 ± 25 K [60]) by almost 150 K when increasing x in LaNi5exSnx from 0.2 to
using another approach. 0.22. The increase of Sn substitution also results in the
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1 2757

The PCT of aged LaNi4.78Sn0.22 exhibits two plateau seg-


ments first of which can be ascribed to decomposition of LaHx
formed in the course of degradation of the AB5 intermetallide
during the aging. The fitted values of DSo and DHo for the first
segment (Table 5) correspond to plateau pressure of 0.67 Pa at
T ¼ 470 K, that is close to the value (0.1 Pa) estimated in
Ref. [27] for the same temperature.
The calculated Cmax ¼ 6.52 for the aged alloy is lower than
for the non-aged one (7.28H/AB5) to be 6.52/7.28 ¼ 89.6% of the
latter. The lowering fits in the range (83.1%e90.6%) calculated
in Ref. [27] using the model based on the approach [24,25].
From the other hand, the fraction, W, of the high-pressure
segment in the aged sample is of 0.861 corresponding to the
maximum hydrogen concentration in the AB5Hx-phase
Cmax(AB5Hx) ¼ Cmax/W ¼ 6.52/0.861 ¼ 7.51H/AB5. This value fits
in the range calculated for non-aged LaNi5exSnx (7.28e7.60H/
AB5, see Table 5), that allows to assume that the degradation is
not accompanied by the formation of individual Ni resulting in
Fig. 14 e Experimental points and calculated PCI curves
further reduction of hydrogen storage capacity, and the
(the temperatures in K are shown as labels) for H2
degradation mechanism is related to the substitution of
absorption (filled symbols, solid lines) and desorption
withdrawn La atoms with Ni (Sn) ones as shown by Eqs. (1)e(2)
(open symbols, dashed lines) in TiFe. The graph from the
in Ref. [27].
right shows the corresponding shortest distances, d, of the
The high pressure segment of the aged LaNi4.78Sn0.22 is
experimental points to the calculated curves in DC / Cmax e
characterised by the lowering critical temperature by 74 K as
ln P coordinates. Fitting parameters are listed in Table 5.
compared to the non-aged material. Despite quite big de-
viations of the fitted DSo and DHo for the second segment from
the values reported in Ref. [27] (Table 6), the calculated H2
equilibrium pressures remain very close to the experimental
data (Fig. 13B) exhibiting lower PD at C<4H/AB5. At the same
time, the plateau segment exhibits significant sloping when
the distribution width parameter and slope factors are higher
in 6 and 1.7 times, respectively, as compared to the non-aged
hydride. It testifies about significant inhomogeneity of the
intermetallic phase in the aged material. Interesting that the
aging results in almost symmetric plateau segment (A ~ 0) that
allows us to suppose that the composition inhomogeneities
have a random nature with a regular unimodal distribution.
Finally, the aged sample is characterised by a significant (in
1.44 times) increase of the contribution of the Lorentz profile
that allows us to suppose that the aging also results in higher
contribution of stress effects in the PCT behaviour.

H2 e TiFe

Fig. 15 e Experimental points and calculated PCI curves Despite of low accuracy of the experimental data taken by the
(the temperatures in K are shown as labels) for H2 digitizing poor quality images in Ref. [58] and only one ab-
absorption (filled symbols, solid lines) and desorption sorption isotherm available, a satisfactory fitting (Fig. 14) has
(open symbols, dashed lines) in V75Ti10Zr7.5Ni7.5 b c.c. alloy been achieved. Assuming slope factors, sA and sD, for the first
The graph from the right shows the corresponding segment equal to zeros, we could model the PCI’s features
shortest distances, d, of the experimental points to the characterised by a smooth sloping at the ends of the flat first
calculated curves in DC/Cmax e ln P coordinates. Fitting plateau. The calculated enthalpies and entropies of the hy-
parameters are listed in Table 5. drogenation for both plateau segments are close to the ones
presented in the literature (see Table 6), but the fitted critical
temperatures are higher than it can be expected from the
smoother transitions to / from plateau region (increase of the phase diagram in Ref. [58].
width parameter and Lorentz contribution) while plateau Because of availability of only one experimental absorption
slopes in the midpoints remain similar due to a lower value of isotherm at the temperature close to the calculated critical
the slope factor for LaNi4.78Sn0.22. The smoothening is asym- one for the second segment, the refinement accuracy of the
metric; the fitted value A<0 corresponds to the lower PCI cur- hysteresis energy loss, DGh, for the second segment was poor.
vature when passing from (a þ b)- to b-region. Actually, the calculated hysteresis corresponded to the break
2758
Table 6 e Comparison of main PCT properties (experimental vs calculated) using the present model.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1
Parameter System
H2eLaNi4.8Sn0.2 H2eLaNi4.78Sn0.22 H2eLaNi4.78Sn0.22 aged H2eTiFe H2eV75Ti10Zr7.5Ni7.5
Segment 1 Segment 2 Segment 1 Segment 2 Segment 1 Segment 2
Reversible H sorption capacity [wt.%] Experimental 1.20 1.25 0.48 1.66 1.20
Calculated 1.12 1.28 0.51 1.67 1.47
P [atm] 2e100 2e30 2.5e30 64e1.7 25e0.1
T [oC] 0e240 21e203 21e203 40 30e100
Critical Experimental 513 e e e ~400 333 443d e
temperature [K] Calculated 487 633 669 559 778 377 381 509
DSo [J mol He1 e1
2 K ] Experimental e104.2 e107.0 PD ¼ 0.1 Paa e108.0 e106.0 e130.14c e118.6d e174.6e
Calculated e106.9 e109.2 e99.6 e107.7 e128.6 e138.0 e176.6
(168.2)f
DHo [kJ mol He1
2 ] Experimental e33.2 e34.53 PD ¼ 0.67 Pab e36.74 e28.11 e33.41c e83.0d e52.69e
Calculated e33.9 e35.10 e34.27 e28.66 e32.89 e69.7 e54.01
(46.04)f
PA/PD Experimental 1.15 e e e 2.06 1.81 e 3.19g
Calculated 1.24 e e e 2.23 1.44 e 3.98g
T [oC] 100 e e e 40 40 e 30
References [57] [27] [47,58] [59,61,62]
a
Estimated for LaHx at T ¼ 470 K [27].
b
Calculated from the values DSo and DHo (Table 5) at T ¼ 470 K using Van’t Hoff equation.
c
For the composition TiFeH1.4.
d
For H2 e V [61,62].
e
Calculated from Van’t Hoff plots of desorption data at H/M ¼ 1.11.
f
Calculated in Ref. [59] using model [33].
g
Taken from experimental data [59] and calculated (this work) at H/M ¼ 1.11.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1 2759

of the polyline (at z0 ¼ 1.13 according to the fitted TC value) of (i) the isotherms with several plateau segments, as well as
showing temperature dependence of the hysteresis shift, Xh, (ii) changing the isotherms with the temperature.
assumed in the present model (Fig. 6). The poor accuracy was The model was implemented as a demonstration software
also observed for other fitting parameters of the second able to show the features of the modelled PCT diagrams, as
segment, especially, the ones responsible for its absorption well as to process experimental PCT data with the refinement
part. Due to this reason, the fitting parameters responsible for of the model fitting parameters.
the temperature dependence of the plateau slope were not The developed model allows to fit, with high accuracy,
refined in this case. experimental PCT data in wide range of temperatures and
Nevertheless, the presented example shows good possi- hydrogen pressures, including the simulation of quite
bilities of the proposed model in the fitting of quite compli- complicated features of real hydrogen e metal systems viz. (i)
cated PCT diagrams using even inaccurate and incomplete presence of several plateau segments, (ii) hysteresis including
experimental data. its disappearance when approaching to critical temperature,
(iii) shape of transitions between a-, (aþb)- and b-regions, and
H2 e bcc-V alloy (iv) flat or sloping plateaux including concentration and tem-
perature dependencies of the plateau slope. For the first time,
Fig. 15 shows results of the refinement of experimental PCT we succeeded in the simultaneous modelling of both
data in the system H2 e b.c.c.-V75Ti10Zr7.5Ni7.5. Although the hydrogen absorption and desorption experimental PCT data
first plateau could not be observed at the experimental con- using the same set of the fitting parameters.
ditions, the refinement was found to be sensitive to some Even inaccurate and incomplete experimental PCT data
parameters of the first plateau segment and allowed to fit can be processed with the model to yield their reasonable
reasonable values of the critical temperature (although ~60 K approximation.
lower than it is observed in H e V phase diagram [61]) and Despite of big number of the fitting parameters and asso-
hydrogenation enthalpy and entropy [62]. ciated difficulties in their refinement, the model can become a
The refinement of all parameters for the second segment useful tool for the systematisation of experimental PCT data
allowed us to achieve quite good accuracy of the modelling in various metal e hydrogen systems, as well as for the in-
(Rf ¼ 0.0017, jdj < 0.09) including such feature as change of the crease of precision in the simulation of heat-and-mass
plateau slope with the concentration on the absorption iso- transfer for gas phase applications of metal hydrides.
therms, first of all, affected by the asymmetry, AA < 0. We also
note that the fitted values of DSo and DHo for the second
segment are closer to the corresponding values obtained from
the experimental points by building Van’t Hoff plots, than the Acknowledgements
apparent hydrogenation entropies and entropies calculated in
Ref. [59] using the previous model [33] (see Table 6). The author is grateful to South African National Research
In general, PCT parameters calculated using the present Foundation and the Department of Trade and Industry (NRF/
model are in a reasonable correspondence with the data DTI; NRF incentive funding grant 76735, THRIP project
published in the literature (Table 6). Typical deviations were TP1207254249) whose support significantly promoted this
observed when the original data were inaccurate due to errors work. Financial contribution from the South African Depart-
in digitizing the figures (TiFe) or incomplete (first segments for ment of Science and Technology (DST) via projects KP3-S02
aged LaNi4.78Sn0.22 and V75Ti10Zr7.5Ni7.5). Significant but not and KP8-S05 within Hydrogen South Africa (HySA) program,
critical deviations of DSo and DHo also took place when as well as HENERGY project within ERAfrica international
refining desorption segments characterised by sloping collaboration program is acknowledged as well.
plateaux (second segment for aged LaNi4.78Sn0.22). The origin The author also thanks Prof Andrei Kolesnikov (Tshwane
can be in different H concentrations taken for the determi- University of Technology, Pretoria, South Africa) for his valu-
nation of the plateau properties. able comments regarding this manuscript and the code of the
supplementary software.

Conclusions
Appendix A. Supplementary data
Within the present work, the phase equilibria in real metal e
hydrogen systems were modelled by statistical processing of Supplementary data related to this article can be found at
an “ideal” pressure e composition isotherm (core function) http://dx.doi.org/10.1016/j.ijhydene.2015.12.055.
built assuming hydrogen in the solid as an interacting lattice
gas. The processing assumed random nature of plateau pres-
sures for H desorption and absorption, and involved pseudo-
references
convolutions of the core function with the corresponding
special distribution functions with the parameters responsible
for plateau slope (including its change with the concentration),
[1] Dantzer P. Metal-hydride technology: a critical review. In:
curvatures of the isotherms in the course of transitions to / Wipf H, editor. Hydrogen in metals iii. Properties and
from the plateau region, and absorption / desorption hyster- applications. Berlin e Heidelberg: Springer-Verlag; 1997.
esis. Furthermore, the model was extended for the simulation p.279e340.
2760 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1

[2] Sandrock G. A panoramic overview of hydrogen storage RNi5 (R ¼ La, Pr, Nd and Sm) systems based on statistical
alloys from a gas reaction point of view. J Alloys Compds mechanics. J Alloys Compds 2009;470:360e4.
1999;293e295:877e88. [24] Ledovskikh A, Danilov D, Rey WJJ, Notten PHL. Modeling of
[3] Bowman RC, Fultz B. Metallic hydrides I: hydrogen storage hydrogen storage in hydride-forming materials: statistical
and other gas-phase applications. MRS Bull thermodynamics. Phys Rev 2006;B73:014106. http://
2002;27(9):688e93. dx.doi.org/10.1103/PhysRevB.73.014106.
[4] Lototskyy MV, Yartys VA, Pollet BG, Bowman Jr RC. Metal [25] Ledovskikh A, Danilov D, Notten PHL. Modeling of hydrogen
hydride hydrogen compressors: a review. Int J Hydrogen storage in hydride-forming materials: equilibrium gas-phase
Energy 2014;39:5818e51. kinetics. Phys Rev 2007;B76:064106. http://dx.doi.org/10.1103/
[5] Lototskyy M, Satya Sekhar B, Muthukumar P, Linkov V, PhysRevB.76.064106.
Pollet BG. Niche applications of metal hydrides and related [26] Ledovskikh A, Danilov D, Notten PHL. Equilibrium kinetics of
thermal management issues. J Alloys Compds chemisorption processes. ChemPhysChem 2008;9:1040e5.
2015;645:S117e22. [27] Bowman Jr RC, Payzant EA, Wilson PR, Pearson DP,
[6] Nishizaki T, Miyamoto K, Yoshida K. Coefficients of Ledovskikh A, Danilov D, et al. Characterization and
performance of hydride heat pumps. J Less-Common Met analyses of degradation and recovery of LaNi4.78Sn0.22
1983;89:559e66. hydrides following thermal aging. J Alloys Compds
[7] Isselhorst A. Heat and mass transfer in coupled hydride 2013;580:S207e10.
reaction beds. J Alloys Compds 1995;231:871e9. [28] Kierstead HA. A theory of multiplateau hydrogen absorption
[8] Jemni A, Ben Nasrallah S. Study of two-dimensional heat and isotherms. J Less-Common Met 1980;71:303e9.
mass transfer during absorption in a metal e hydrogen [29] Kierstead HA. A generalized theory of multiplateau hydrogen
reactor. Int J Hydrogen Energy 1995;20:43e52. absorption isotherms. J Less-Common Met 1982;84:253e61.
[9] Lee SG, Lee KY, Kim TG, Lee ZH, Lee JY. Mathematical model [30] Kierstead HA. A theory of hydrogen absorption with
for the dynamic P-C-T curves of the MmNi4.6Al0.2Fe0.2V0.03 interactions. J Less-Common Met 1984;96:141e52.
alloy in a tubular reactor. Int. J. Hydrogen Energy [31] Marinin VS, Umerenkova KR, Shmalko YF, Lobko MP,
1996;21:733e40. Lototsky MV. Interacting lattice gas model for hydrogen
[10] Gambini M, Manno M, Vellini M. Numerical analysis and subsystem of metal hydrides. Funct Mater 2002;9:395e401.
performance assessment of metal hydride-based hydrogen [32] Marinin VS, Shmalko YF, Umerenkova KR, Lototsky MV,
storage systems. Int J Hydrogen Energy 2008;33:6178e87. Lobko MP. Critical separation point of disordered metal
[11] Paya a J, Linder M, Laurien E, Corbera n JM. Mathematical hydride phases in the model of interacting lattice gas. Funct
models for the PeCeT characterization of hydrogen Mater 2002;9:609e16.
absorbing alloys. J Alloys Compds 2009;484:190e5. [33] Lototsky MV, Yartys VA, Marinin VS, Lototsky NM. Modelling
[12] Ye J, Jiang L, Li Z, Liu X, Wang S, Li X. Numerical analysis of of phase equilibria in metalehydrogen systems. J Alloys
heat and mass transfer during absorption of hydrogen in Compds 2003;356e357:27e31.
metal hydride based hydrogen storage tanks. Int J Hydrogen [34] Larsen JW, Livesay BR. Hydriding kinetics of SmCo5. J Less-
Energy 2010;35:8216e24. Common Met 1980;73:79e88.
[13] Fowler RH. Adsorption isotherms. Critical conditions. Proc [35] € m H, Suda S, Lewis D. A numerical expression for the
Bjurstro
Camb Philos Soc 1936;32:144e51. P-C-T properties of metal hydrides. J Less-Common Met
[14] Lacher JR. A theoretical formula for the solubility of 1987;130:365e70.
hydrogen in palladium. Proc Roy Soc Lond 1937;A161:525e45. [36] Fujitani S, Nakamura H, Furukawa A, Nasako K, Satoh K,
[15] Harashima A, Tanaka T, Sakaoku K. Co-operative Imoto T, et al. A method for numerical expressions of p-c
phenomena in the palladium-hydrogen system. I. J Phys Soc isotherms of hydrogen-absorbing alloys. Z Phys Chem
Jap 1948;3:208e13. 1993;179:27e33.
[16] Tanaka T, Sakaoku K, Harashima A. Co-operative [37] Lototsky MV. A modification of the Lacher - Kierstead theory
phenomena in the palladium-hydrogen system. Ii. J Phys Soc for simulation of PCT-diagrams of real “hydrogen - hydride-
Jap 1948;3:213e8. forming material” systems. Kharkov Univ Bull
[17] Rees ALG. Statistical mechanics of two-component 2000;477:45e53. Chem Series 5(28).
interstitial solid solutions. Trans Faraday Soc [38] https://www.researchgate.net/publication/251572144_
1954;50:335e42. HyPCT-v41-July2013.
[18] Speiser R. Statistical mechanics of metal e hydrogen [39] Shilov AE, Efremenko NE. Effect of sloping plateaux in two-
systems. In: Mueller WM, Blackledge JP, Libowitz GG, editors. phase regions of hydride systems. Russ J Phys Chem
Metal hydrides. London e New York: Academic Press; 1968. 1986;60:3024e8.
p. 90e118. [40] Park C-N, Luo S, Flanagan TB. Analysis of sloping plateaux in
[19] Hall CK, Futran M. Statistical theory of phase change alloys and intermetallic hydrides. I. Diagnostic features. J
behavior of metal e hydrogen systems. J Less-Common Met Alloys Compds 2004;384:203e7.
1980;74:237e42. [41] Luo S, Park C-N, Flanagan TB. Analysis of sloping plateaux in
[20] Brodowsky H, Yasuda K. From partition function to phase alloys and intermetallic hydrides. II. Real systems. J Alloys
diagram e statistical thermodynamics of the LaNi5 e H Compds 2004;384:208e16.
system. Z Phys Chem 1993;179:45e55. [42] Zhou Z, Zhang J, Ge J, Feng F, Dai Z. Mathematical modeling
[21] Beeri O, Cohen D, Gavra Z, Johnson JR, Mintz MH. High- of the PCT curve of hydrogen storage alloys. Int J Hydrogen
pressure studies of Laves phase intermetallic hydrides e Energy 1994;19:269e73.
Adaptation of statistical thermodynamic models. J Alloys [43] Fang S, Zhou Z, Zhang J, Yao M, Feng F, Northwood DO. Two
Compds 1999;293e295:14e8. mathematical models for the hydrogen storage properties of
[22] Beeri O, Cohen D, Gavra Z, Mintz MH. Sites occupation and AB2 type alloys. J Alloys Compds 1999;293e295:10e3.
thermodynamic properties of the TiCr2exMnxeH2 (0x1) [44] Fang S, Zhou Z, Zhang J, Yao M, Feng F, Northwood DO. The
system: statistical thermodynamics analysis. J Alloys application of mathematical models to the calculation of
Compds 2003;352:111e22. selected hydrogen storage properties (formation enthalpy
[23] Senoh H, Takeichi N, Yasuda K, Kiyobayashi T. A theoretical and hysteresis) of AB2-type alloys. Int J Hydrogen Energy
interpretation of the pressureecomposition isotherms of 2000;25:143e9.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 1 ( 2 0 1 6 ) 2 7 3 9 e2 7 6 1 2761

[45] Singh RK, Gupta BK, Lototsky MV, Srivastava ON. On the [54] Esayed AY, Northwood DO. Hysteresis in metallic solid
synthesis and hydrogenation behaviour of MmNi5exFex solution and intermetallic compound e hydrogen systems.
alloys and computer simulation of their PeCeT curves. J Int J Hydrogen Energy 1997;22:77e82.
Alloys Compds 2004;373:208e13. [55] Shilov AL. Kinetics and thermodynamics of formation and
[46] Wu Y, Lototskyy MV, Solberg JK, Yartys VA. Effect of decomposition of PrNi5 hydride. Russ J Inorg Chem
microstructure on the phase composition and hydrogen 1983;57:1305e8.
absorption-desorption behaviour of melt-spun Mg-20Ni- [56] Zhou L, Zhou Y. Determination of compressibility factor and
8Mm alloys. Hydrogen Energy 2012;37:1495e508. fugacity coefficient of hydrogen in studies of adsorptive
[47] Davids MW, Lototskyy M. Influence of oxygen introduced in storage. Int J Hydrogen Energy 2001;26:597e601.
TiFe-based hydride forming alloy on its morphology, [57] Luo S, Luo W, Clewley JD, Flanagan TB, Bowman Jr RC.
structural and hydrogen sorption properties. Int J Hydrogen Thermodynamic and degradation studies of LaNi4.8Sn0.2-H
Energy 2012;37:18155e62. using isotherms and calorimetry. J Alloys Compds
[48] Flanagan TB, Noh H, Clewley JD, Bowman Jr RC. The effect of 1995;231:473e8.
cold-work on hysteresis in the V-H(D) system. Scr Met Mater [58] Reilly JJ, Wiswall RH. Formation and properties of iron
1993;28:355e9. titanium hydride. Inorg Chem 1974;13(1):218e22.
[49] Stancik AL, Brauns EB. A simple asymmetric lineshape for [59] Lototsky MV, Yartys VA, Zavaliy IY. Vanadium-based BCC
fitting infrared absorption spectra. Vibr Spectr 2008;47:66e9. alloys: phase-structural characteristics and hydrogen
[50] Schmid M, Steinrücka HP, Gottfried JM. A new asymmetric sorption properties. J Alloys Compds 2005;404e406:421e6.
Pseudo-Voigt function for more efficient fitting of XPS lines. [60] Laurencelle F, Dehouche Z, Goyette J. Hydrogen sorption
Surf Interface Anal 2014;46:505e11. cycling performance of LaNi4.8Sn0.2. J Alloys Compds
[51] Bowerman BS, Wulff CA, Biehl GE, Flanagan TB. Calorimetry 2006;424:266e71.
within hysteresis loops: application to LaNi5 e H. J Less- [61] Dolan MD, McLennan KG, Chandra D, Kochanek MA, Song G.
Common Met 1980;73:1e13. Suppression of the critical temperature in binary vanadium
[52] Flanagan TB, Clewley JD. Hysteresis in metal hydrides. J Less- hydrides. J Alloys Compds 2014;58:385e91.
Common Met 1982;83:127e41. [62] Lamb J, Chandra D, Coleman M, Sharma A, Cathey WN,
[53] Flanagan TB, Park CN, Oates WA. Hysteresis in solid state Paglieri SN, et al. Low and high-pressure hydriding of
reactions. Prog Solid St Chem 1995;23:291e363. Ve0.5at.%C. J Nucl Mater 2010;399:55e61.

S-ar putea să vă placă și