The site-total-charge approach

As the sequential use of the dominant-valency and the dominant-constituent rules are recommended by the IMA−CNMNC, it is worthwhile to analyse their basic aspects, here identified as follows: (1) the principle of the formula electroneutrality; and (2) the definition of an end-member formula (Hawthorne, Reference Hawthorne2002) that includes at most one site with double occupancy (i.e. heterovalent pair of ions or ion–vacancy pair). In addition, the concepts of total charge and charge constraint at the crystallographic sites should be considered. The charge constraint can be defined as an integer number close (or next) to the observed site total charge (STC). It provides information on possible root-charge and atomic arrangements; if such arrangements satisfy all the criteria of an end-member formula, the mineral may be identified.

Again, considering the composition ^[8](Ca_0.4Na_0.35K_0.25)^[4](Al_1.4Si_2.6)O₈, the key aspect is to explore the STC at the [8]- and [4]-fold coordinated sites: ^[8]STC = +1.4 and ^[4]STC = +14.6. As the former is close to +1 and the latter is close to +15, the charge constraints (+1 and +15) indicate that the mineral species may be consistent with the end-member root-charge arrangement (R ⁺)^Σ1+(R ³⁺R ⁴⁺₃)^Σ15+O₈, rather than (R ²⁺)^Σ2+(R ³⁺₂R ⁴⁺₂)^Σ14+O₈. In terms of atoms per formula unit (apfu), the atomic arrangement 0.60[(Na,K)(AlSi₃)] (= 3.00 apfu, limited by Na + K content) is larger than 0.40[(Ca)(Al₂Si₂)] (= 2.00 apfu, limited by Ca content). Thus, in accord with the dominant-constituent rule, Na_0.35 > K_0.25, the species corresponds to albite.

Note that the site total charge and charge constraint are another important way to manifest the dominant-valency rule in which the STC is used to identify the dominant root-charge arrangement consistent with an end-member formula.

Extension of the dominant-valency rule

For complex mineral compositions, it may appear that the simple application of the dominant-valency rule does not lead to a charge-balanced end-member formula. This is due to the occurrence of heterovalent substitution mechanisms at more than one crystallographic site (or group of sites). Starting from the albite composition ^[8](Na_0.60Ca_0.40)^[4](Al_1.40Si_2.60)O₈, we could assume a hypothetical heterovalent substitution mechanism affecting the larger cation site: 0.12 ^[8](Pb²⁺ + □) → 0.12 ^[8](Na⁺ + Na⁺). This substitution will lead to the composition ^[8](Na_0.36Pb²⁺_0.12□_0.12Ca²⁺_0.40)^[4](Al_1.40Si_2.60)O₈, which has 0.52 divalent (R ²⁺) apfu; the simple application of the dominant-valency rule thus leads to the unbalanced charge formula (CaAlSi₃O₈)^Σ1+. Application of the site-total-charge approach to this composition indicates +1.4 charges at the [8]-fold coordinated site, close to +1, thus confirming that the root-charge arrangements with total charge +1 are dominant at that site: in fact, [0.36(R ⁺) + 0.24(R ²⁺_0.5□_0.5)^Σ1+] = 0.60 apfu is larger than the largest amount of charge arrangement entirely composed by R ²⁺ cations = 0.52 apfu. In other words, the aggregate-charge arrangement [(R ¹⁺) + (R ²⁺_0.5□_0.5)^Σ1+] is more abundant than the largest amount of (R ²⁺) we can calculate from the mineral composition. As the arrangement (R ⁺_0.36) (limited by the Na content) is larger than (R ²⁺_0.12□_0.12)_Σ0.24 (limited by the number of vacancies), the relative dominant cation arrangement is given by ^[8](Na⁺), which leads to the end-member formula NaAlSi₃O₈. Figure 2 shows the boundaries between generalised mineral species in such a complex system involving a heterovalent substitution at two sites (Fig. 2a), followed by a heterovalent substitution at a single site (Fig. 2b).

Fig. 2. Schematic representations of coupled heterovalent substitutions at two sites (a), followed by a heterovalent substitution at a single site (b) leading to the site-total-charge approach.

Hatert and Burke (Reference Hatert and Burke2008) have in fact addressed the cases of (1) coupled heterovalent substitutions at a single site; (2) heterovalent substitutions at two sites; and (3) heterovalent substitutions at two sites plus homovalent substitution at a single site, but have not considered the case (4) of coupled heterovalent substitutions at two sites along with the heterovalent substitution at a single site. This more complex case is now clarified (Fig. 2).

The jervisite example

The application of the site-total-charge approach should be used when the simple application of the dominant-valency rule leads to an unbalanced end-member formula. Let us consider the empirical formula of the jervisite pyroxene of Mellini et al. (Reference Mellini, Merlino, Orlandi and Rinaldi1982):

$$ {^{M {\left( 2 \right)}} \left({\rm N}{\rm a}^ + _{0.43} {\rm C}{\rm a}^{2 + }_{0.31} {\rm F}{\rm e}^{2 + }_{0.14} \squ_{0.12}\right) _{\Sigma 1.00}} {^{M{(1)}}\left({{\rm S}{\rm c}^{3 + }_{0.66} {\rm M}{\rm g}^{2 + }_{0.19} {\rm F}{\rm e}^{2 + }_{0.15} }\right)}_{\Sigma 1.00} {\rm Si}_2{\rm O}_6, $$

in which divalent cations seem to predominate at the M(2) site with Ca²⁺ as dominant constituent, and trivalent cation predominate at the M(1) site with Sc³⁺. As a result, the simple application of the dominant-valency rule yields (^M(2)Ca^M(1)ScSi₂O₆)^Σ1+, which is not charge-balanced. However, the STC at M(2) (= +1.33) and M(1) (= +2.66) indicate the following charge constraints: (i) M(2)^Σ1+ compatible with both ^M(2)(R ⁺) and ^M(2)(R ²⁺_0.5□_0.5)^Σ1+; (ii) M(1)^Σ3+ compatible with R ³⁺ (Fig. 3). As a result, the root-charge arrangements consistent with an end-member are ^M(2)(R ⁺)^M(1)(R ³⁺)Si₂O₆ and ^M(2)(R ²⁺_0.5□_0.5)^M(1)(R ³⁺)Si₂O₆, which are related by the heterovalent substitution ^M(2)(Na⁺) ↔ ^M(2)(R ²⁺_0.5 + □_0.5). Regarding the dominant charge arrangements, ^M(2)[(R ⁺) + (R ²⁺_0.5□_0.5)^Σ1+] prevails over the divalent ^M(2)(R ²⁺) in terms of apfu: the sum of the charge arrangements with total charge +1 at the M(2) site, 0.43 apfu of ^M(2)(R ⁺) (limited by Na contents) and 0.24 apfu of ^M(2)(R ²⁺_0.5□_0.5)^Σ1+ (limited by the number of vacancies), yields an aggregate-charge arrangement ^M(2)[0.43(R ⁺) + 0.24(R ²⁺_0.5□_0.5)^Σ1+] equal to 0.67 apfu. The latter is even larger than the largest amounts of charge arrangement characterised by the sole occurrence of ^M(2)(R ²⁺), 0.45 apfu. Note that ^M(2)(R ²⁺_0.45) is a hypothetical value that we can calculate to compare with the dominant-charge arrangements: in fact, the occurrence of 0.24(R ²⁺_0.5□_0.5) will actually reduce ^M(2)(R ²⁺) to 0.33 apfu. Between the two root-charge arrangements characterised by a total charge +1 at M(2), 0.43[^M(2)(R ⁺)^M(1)(R ³⁺)R ⁴⁺₂O₆] is more abundant than 0.24[^M(2)(R ²⁺_0.5□_0.5)^M(1)(R ³⁺)R ⁴⁺₂O₆] and as the atomic arrangement 0.43[^M(2)(Na⁺)^M(1)(Sc³⁺)Si₂O₆] > 0.24[^M(2)(Ca²⁺_0.5□_0.5)^M(1)(Sc³⁺)Si₂O₆], the end-member formula is NaSc³⁺Si₂O₆.

Fig. 3. Schematic representations of site total charge and atomic charge arrangements at the M(2) (a) and M(1) (b) sites of jervisite (see text).

The reason why the dominant-valency rule apparently fails may be shown by the substitution mechanisms involving jervisite. In this regard, it is instructive to arrive at the jervisite empirical formula from the diopside composition CaMgSi₂O₆:

1. (1) the heterovalent substitution at two sites

   $$\eqalign{&\hskip-15pt ^{M {\left( 2 \right)}}\left( {{\rm N}{\rm a}^ + } \right)_{0.66} \,+\, ^{M{(1)}} \left( {{\rm S}{\rm c}^{3 + }} \right)_{0.66}\;\to\; ^{M{(2)}}\!\! \left( {{\rm C}{\rm a}^{2 + }} \right)_{0.66} \,+\, ^{M{(1)}} \left( {{\rm M}{\rm g}^{2 + }} \right)_{0.66} \cr & {\rm yield}{\rm s}\;^{M{(2)}} \left( {{\rm N}{\rm a}^ + _{0.66} {\rm C}{\rm a}^{2 + }_{0.34} } \right)\;^{M{(1)}} \left( {{\rm S}{\rm c}^{3 + }_{0.66} {\rm M}{\rm g}^{2 + }_{0.34} } \right){\rm S}{\rm i}_2{\rm O}_6;} $$

2. (2) the homovalent substitution at a single site

   $$\eqalign{& ^{M {\left( 1 \right)}}\left( {{\rm F}{\rm e}^{2 + }} \right)_{0.15}\; \to\; ^{M{(1)}}\!\! \left( {{\rm M}{\rm g}^{2 + }} \right)_{0.15} \cr & {\rm yield} {\rm s}\;^{M{(2)}} \left( {{\rm N}{\rm a}^ + _{0.66} {\rm C}{\rm a}^{2 + }_{0.34} } \right)\;^{M{(1)}} \left( {{\rm S}{\rm c}^{3 + }_{0.66} {\rm M}{\rm g}^{2 + }_{0.19} {\rm F}{\rm e}^{2 + }_{0.15} } \right){\rm S}{\rm i}_2{\rm O}_6;}$$

3. (3) the heterovalent substitution at a single site

   $$\eqalign{& ^{M {\left( 2 \right)}} \left( {{\rm F}{\rm e}^{2 + } + {\squ}} \right)_{0.12} \,\to\; ^{M{(2)}}\!\! \left( {2{\rm N}{\rm a}^ + } \right)_{0.12} \cr & \hskip-15pt {\rm yield}{\rm s}\;^{M{(2)}} \left( {{\rm N}{\rm a}^ {+} _{0.42} {\rm C}{\rm a}^{2 + }_{0.34} {\rm F}{\rm e}^{2 + }_{0.12} {\squ}_{0.12} } \right)^{M{(1)}} \left( {{\rm S}{\rm c}^{3 + }_{0.66} {\rm M}{\rm g}^{2 + }_{0.19} {\rm F}{\rm e}^{2 + }_{0.15} } \right){\rm S}{\rm i}_2{\rm O}_6,} $$
   which substantially corresponds to the empirical composition of jervisite.

The alluaudite example

The alluaudite supergroup contains phosphates and arsenates with the structural formula A(2)’A(1)M(1)M(2)₂(TO₄)₃; the cations are distributed mainly among four distinct crystallographic sites: the large cavities coordinated by the A(2)’ and A(1) sites in the channels of the structure, and the octahedrally-coordinated M(1) and M(2) sites. A wet chemical analysis of an alluaudite from the Buranga pegmatite, Rwanda, was published by Héreng (Reference Héreng1989) and gives the following cation distribution:

$\eqalign { {^{A(2)^{\prime}}(\squ)^{A(1)} ({\rm Mn}^{2 + }_{0.34} {\rm Ca}^{2 + }_{0.17} {\rm Na}_{0.29} {\squ}_{0.21})_{\Sigma 1.00}} {^{M(1)}( {\rm Mn}^{2 + })} \cr {^{M(2)} ( {\rm Fe}^{3 + }_{1.70} {\rm Fe}^{2 + }_{0.15} {\rm Mn}^{2 + }_{0.10} {\rm Mg}_{0.04}{\rm Zn}_{0.01} )_{\Sigma 2.00}} {^{T} ( {\rm PO}_4 )_3}.}$

In this formula, vacancies are dominant at A(2)’, divalent cations with Mn²⁺ > Ca at A(1), Mn²⁺ at M(1) and Fe³⁺ at M(2), thus leading to the unbalanced end-member formula [□Mn²⁺Mn²⁺Fe³⁺₂(PO₄)₃]^Σ1+. The dominant-valency rule cannot be simply applied to the A(1) site population, due to the occurrence of the heterovalent substitution ^A(1)(R ²⁺ + □) → ^A(1)(2Na⁺) which produces a decrease of the Na content and an increase of the R ²⁺-cations and vacancies at the A(1) site. However, this substitution mechanism preserves the STC, whose value (+1.31) is close to +1. In fact, the alluaudite structure is dominated by the aggregate-charge arrangement ^A(1)[(R ²⁺_0.5□_0.5)^Σ1+ + (R ⁺)]: the sum of ^A(1)(R ²⁺_0.21 + □_0.21) (Σ = 0.42 apfu, limited by the number of vacancies) plus ^A(1)(R ⁺_0.29) (= 0.29 apfu, limited by Na contents) is equal to 0.71 apfu; the latter is larger than the largest amount of charge arrangement calculated using only R ²⁺-cations at A(1): 0.51 apfu. Looking at the dominant aggregate-charge arrangement with total charge +1, ^A(1)(R ²⁺_0.21□_0.21)_Σ0.42 > ^A(1)(R ⁺_0.29). In accord with the dominant-constituent rule, ^A(1)Mn²⁺ > ^A(1)Ca²⁺, the relative dominant atomic arrangement at the A(1) site is hence (Mn²⁺_0.5□_0.5)^Σ1+ and the end-member formula is □(Mn²⁺_0.5□_0.5)Mn²⁺Fe³⁺₂(PO₄)₃. It should be noted that the heterovalent-pair ^A(1)(R ²⁺_0.5□_0.5)^Σ1+ and the homovalent ^A(1)(R ⁺) play the same role in the charge balance (overall electroneutrality).

The garnet example

The approved garnet nomenclature is based on the dominant-valency rule, but this rule was unsuccessful in discriminating schorlomite from morimotoite/andradite. Let us consider the schorlomite composition from the type locality Magnet Cove, Arkansas (Grew et al., Reference Grew, Locock, Mills, Galuskina, Galuskin and Hålenius2013): ^X[(Ca²⁺_2.907Fe²⁺_0.043Mn²⁺_0.035)Na⁺_0.015]_Σ3.000^Y[(Ti⁴⁺_1.069Zr⁴⁺_0.055)Fe³⁺_0.517(Fe²⁺_0.204Mg²⁺_0.155)]_Σ2.000^Z[Si⁴⁺_2.250(Fe³⁺_0.588Al³⁺_0.162)]_Σ3.000O₁₂ in which the dominant valence at the X site is +2, and at the Y and Z sites is +4. The simple application of the dominant-constituent rule fails as the dominant constituents ^XCa²⁺, ^YTi⁴⁺ and ^TSi⁴⁺ lead to unbalanced end-member formula (Ca²⁺₃Ti⁴⁺₂Si⁴⁺₃O^2–₁₂)^Σ2+. This issue was overcome by Grew et al. (Reference Grew, Locock, Mills, Galuskina, Galuskin and Hålenius2013) by using the most abundant end-member composition: schorlomite Ca₃Ti₂(SiFe³⁺₂)O₁₂ (37.5%) > morimotoite Ca₃(TiFe²⁺)Si₃O₁₂ (35.9%) > andradite Ca₃(Fe³⁺₂)Si₃O₁₂ (25.9%). This is equivalent to stating the schorlomite-group minerals (which include the mineral species hutcheonite, irinarassite, kerimasite, kimzeyite, schorlomite and toturite) are identified by the dominant-end-member approach.

With regard to the substitution types involved in the empirical formula of this garnet, starting from andradite Ca₃(Fe³⁺₂)Si₃O₁₂, we have a heterovalent substitution at two sites (schorlomite → andradite substitution type):

$$^Y {\rm T}{\rm i}^{4 + }_{0.75} \,+\; ^Z\!\!{\rm F}{\rm e}^{3 + }_{0.75} \to ^Y\!\!{\rm F}{\rm e}^{3 + }_{0.75} \,+\; _{}^Z\!\! {\rm S}{\rm i}^{4 + }_{0.75} $$

yielding the formula Ca₃^Y(Ti⁴⁺_0.75Fe³⁺_1.25)_Σ2.00^Z(Si_2.25Fe³⁺_0.75)_Σ3.00O₁₂, plus a heterovalent substitution at a single site (morimotoite → andradite substitution type):

$$^Y \left( {{\rm T}{\rm i}^{4 + }_{0.38} + {\rm F}{\rm e}^{2 + }_{0.38} } \right) \to 2^Y\left( {{\rm F}{\rm e}^{3 + }} \right)_{0.38}$$

yielding the formula Ca₃^Y(Ti⁴⁺_1.13Fe³⁺_0.49Fe²⁺_0.38)_Σ2.00^Z(Si_2.25Fe³⁺_0.75)_Σ3.00O₁₂, which reasonably corresponds to the empirical one, with Ca = ^X(Ca²⁺+ Fe²⁺+ Mn²⁺), ^YTi⁴⁺ = ^Y(Ti⁴⁺+ Zr⁴⁺), ^YFe²⁺ = ^Y(Fe²⁺+ Mg²⁺) and ^ZFe³⁺ = ^Z(Fe³⁺+ Al³⁺). Similarly to the jervisite example above, the garnet from Magnet Cove can also be identified by the dominant-valency rule using the concepts of STC and charge constraint to select the possible root-charge and atomic arrangements among the end-members ^X(Ca₃)^Σ6+ Y(Ti₂)^Σ8+ Z(SiFe³⁺₂)^Σ10+O₁₂ (schorlomite), ^X(Ca₃)^Σ6+ Y(Fe³⁺₂)^Σ6+ Z(Si₃)^Σ12+O₁₂ (andradite) and ^X(Ca₃)^Σ6+ Y(TiFe²⁺)^Σ6+ Z(Si₃)^Σ12+O₁₂ (morimotoite). In detail: the total charge at X (= +5.99) is very close to +6, at Y (= +6.77) is closer to +6 than +8, and at Z (= +11.25) is closer to +12 than +10; note that the charge constraint Y ^Σ7+ and Z ^Σ11+ are ruled out as leading to a formula, (Ca₃)^Y(TiFe³⁺)^Z(Si₂Fe³⁺)O₁₂, inconsistent with the end-member definition (Hawthorne, Reference Hawthorne2002); the charge constraints (X ^Σ6+, Y ^Σ6+ and Z ^Σ12+) are only compatible with the root-charge arrangements ^X(R ²⁺₃)^{Σ6+ Y}(R ⁴⁺R ²⁺)^{Σ6+ Z}(R ⁴⁺₃)^Σ12+O₁₂ (morimotoite type) and ^X(R ²⁺₃)^{Σ6+ Y}(R ³⁺₂)^{Σ6+ Z}(R ⁴⁺₃)^Σ12+O₁₂ (andradite type), but not with the arrangement ^X(R ²⁺₃)^Σ6+ Y(R ⁴⁺₂)^Σ8+ Z(R ⁴⁺R ³⁺₂)^Σ10+O₁₂ (schorlomite type) that can be considered as a minor component; in this regard, note that amount of the aggregate-charge arrangement at Y with total charge +6, ^Y[0.718(R ⁴⁺_0.5R ²⁺_0.5) + 0.517(R ³⁺)] = 1.235 apfu, is larger than the largest amount of charge arrangement characterised solely by the ^YR ⁴⁺-cations (1.124 apfu); as ^Y(R ⁴⁺_0.5R ²⁺_0.5)_0.718 > ^Y(R ³⁺)_0.517, the relative dominant charge arrangement is ^Y(R ⁴⁺_0.5R ²⁺_0.5); in terms of constituents, it corresponds to ^Y[Ti⁴⁺_0.5(Fe²⁺,Mg²⁺)_0.5], which in accord with the dominant-constituent rule, ^Y(Fe²⁺_0.204) > ^Y(Mg²⁺_0.155), leads to ^Y(Ti⁴⁺_0.5Fe²⁺_0.5); therefore, the garnet species from Magnet Cove corresponds to morimotoite, even though the most abundant end-member component is schorlomite (Grew et al., Reference Grew, Locock, Mills, Galuskina, Galuskin and Hålenius2013).

It is worth noting that this garnet example showed that the STC at Y (= +6.77) and Z (= +11.25) were very close to +7 and +11, but we selected the integer numbers +6 and +12, respectively, as the choice of charge constraints Y ^Σ7+ and Z ^Σ11+ would lead to the formula ^X(Ca₃)^Y(TiFe³⁺)^Z(Si₂Fe³⁺)O₁₂ with double occupancy of two sites. As a rule, if the integer number closest to the STC is not consistent with an end-member, another integer number in line with the end-member definition must be selected. The incorrect integer number can be recognised as it results in root-charge and atomic arrangements leading to a formula with double occupancy of two sites or charge imbalance.

The detailed procedure reported above for garnet can be summarised as follows:

1. (1) The STC at X (+5.99), Y (+6.77) and Z (+11.25) indicate the charge constraints X ^Σ6+, Y ^Σ6+ and Z ^Σ12+.

2. (2) These constraints lead to the possible root-charge arrangements ^X(R ²⁺₃)^{Σ6+ Y}(R ⁴⁺R ²⁺)^{Σ6+ Z}(R ⁴⁺₃)^Σ12+O₁₂ and ^X(R ²⁺₃)^{Σ6+ Y}(R ³⁺₂)^{Σ6+ Z}(R ⁴⁺₃)^Σ12+O₁₂, which are end-members.

3. (3) As the condition of the charge balance in end-members is satisfied, we can calculate the amounts of root-charge and atomic arrangements in terms of apfu: ^XR ²⁺ ≅ Ca, ^ZR ⁴⁺ = Si and as ^Y(R ⁴⁺_0.5R ²⁺_0.5)_0.718 > ^Y(R ³⁺)_0.517 leads to ^Y[Ti⁴⁺_0.5(Fe²⁺,Mg²⁺)_0.5] in which ^Y(Fe²⁺)_0.204 > ^Y(Mg²⁺)_0.155 leads to the end-member Ca₃(TiFe²⁺)Si₃O₁₂.

The dominant-valency rule is able to identify this garnet if the concepts of site total charges and charge constraints, dictated by the mineral composition and the electroneutrality principle, are taken into account. These concepts emphasise the role of the root-charge arrangements compatible with the site populations and the end-member definition. The dominant-constituent rule can hence be applied to the relative dominant root-charge arrangement.

The aforementioned concepts are also reflected in the ternary composition diagrams. Morimotoite/andradite and schorlomite are related by coupled heterovalent substitutions at the Y and Z sites, whereas morimotoite and andradite are related by a coupled heterovalent substitutions at the Y site. Plotting the compositions of the Y site in a triangular diagram with (R ⁴⁺)^Σ4+, (R ⁴⁺_0.5+R ²⁺_0.5)^Σ3+ and (R ³⁺)^Σ3+ placed at each corner, the boundaries crossing at the centre of the diagram are displaced as follows: 25% for the component (R ⁴⁺)^Σ4+, 37.5% for both (R ⁴⁺_0.5+R ²⁺_0.5)^Σ3+ and (R ³⁺)^Σ3+ (Fig. 4). This displacement is dictated by the different charge or cation arrangements at the Z site between morimotoite/andradite and schorlomite: ^Z(Si⁴⁺₃)^Σ12+ and ^Z(Si⁴⁺R ³⁺₂)^Σ10+, respectively. As a result, Fig. 4 may be used to discriminate the schorlomite-group minerals from the ‘Ca-garnet-group minerals’ in line with the dominant-valency rule (cf. Fig. 4 of this study with fig. 7 of Grew et al., Reference Grew, Locock, Mills, Galuskina, Galuskin and Hålenius2013) as well as with the current garnet nomenclature (the first criteria for distinguishing the group in garnets are the total charge at the Z site and symmetry). A similar argumentation should be applied to the ternary diagram andradite–morimotoite–menzerite-(Y), where the boundaries between the species are redefined in Fig. 5.

Fig. 4. Diagram for discriminating the species in the schorlomite−andradite−morimotoite system: solid blue lines are consistent with the dominant-valency rule, whereas dashed red lines are consistent with the dominant-end-member approach (cf. with fig. 7 of Grew et al., Reference Grew, Locock, Mills, Galuskina, Galuskin and Hålenius2013).

Fig. 5. Diagram for discriminating menzerite-(Y) from andradite and morimotoite: solid blue lines are consistent with the dominant-valency rule, whereas dashed red lines are consistent with the dominant-end-member approach (cf. with fig. 10 of Grew et al., Reference Grew, Locock, Mills, Galuskina, Galuskin and Hålenius2013).