**3.2 Influence of atom displacements**

*Real Perspective of Fourier Transforms and Current Developments in Superconductivity*

*a c*

**Lattice parameters [Å] Enthalpy** 

**[eV]**

0.010 3.038657 3.487973 −1748.5037 8.4055 0.3424 0.018 3.038630 3.487921 −1748.5039 8.4126 0.3975 0.020 3.039107 3.486623 −1748.5046 8.3976 0.4038 0.040 3.042948 3.478367 −1748.5003 8.4704 0.3281 0.060 3.031477 3.514558 −1748.4685 8.2108 0.6365

0.005 3.582982 −1163.6112 11.0832 1.6164 0.010 3.582982 −1163.6114 11.0844 1.6152 0.015 3.582984 −1163.6121 11.0956 1.6039 0.017 3.582987 −1163.6120 11.0941 1.6056 0.020 3.583024 −1163.6102 11.0579 1.6410

**Fermi energy [eV]**

Δ**Ev [eV]**

differs by ~35 meV and the difference in energy to the vertex of the band at Γ (*i.e.* ΔEv)

We show a more detailed systematic comparison of calculated enthalpies as a function of the k-grid value for MgB2 in **Figure 4**. As noted above, all calculations are converged to the same ultra-fine criteria, or tolerance, where self-consistency is achieved. The variability of results shown in **Figure 4** is due to the discreteness of functions and values used to derive the full solution of the Schroedinger equation. The variability is not due to lack of convergence which in all cases is defined in

**Figure 4a** shows that the value of enthalpy for MgB2 oscillates around a consistent minimal value of −1,748.502 eV as the k-grid value is decreased to <0.015 Å−1. The context for this variation in enthalpy is shown in **Figure 4b** where the k-grid value is extended to 0.2 Å−1– a value that has been used in some DFT calculations as criterion for machine learning algorithms [41]. At these higher values for Δk, enthalpy calculations do not provide useful information on subtle structural

*Systematic comparison of calculated enthalpies for MgB2 (a) for fine values of k-grid (i.e. < 0.03 Å−1) and (b) for coarser grid values including those utilized for machine learning searches (arrowed) of materials databases [41]. Enthalpies shown in Figure 4(a) are reproduced in (b) for reference. The lightly shaded region* 

*in Figure 2(b) delineates the k-grid values used for EBS calculation in Figure 3.*

differs by up to ~35 meV depending on the Δk value.

*Parameters calculated for MgB2 and for B-doped diamond.*

Section 2 above.

**Compound k-grid** 

MgB2 Tc ~ 40 K

B-doped diamond Tc ~ 4–7.5 K

**Table 1.**

**value [Å−1]**

**118**

**Figure 4.**

In a dynamic system, other factors may also influence the position of key electronic bands with respect to the Fermi level. For example, atoms in all solids at temperatures above absolute zero vibrate [43] and in some cases, the resulting phonons may align with specific crystallographic real space features such as inter-atom bonds. This circumstance occurs for MgB2 in which one of the dominant E2g phonon modes – shown to be intimately involved in electron–phonon coupling at the onset of superconductivity [17, 34] – aligns with B–B bonds in the *ab* plane [36]. Using DFT, we can model the effect of bond deformation along specific planar orientations by displacing atoms from their structural equilibrium positions consistent with the direction of the E2g phonons [26, 28]. Under different extents of displacement, electron density distributions along the B–B bond and the corresponding EBS, can be determined [28, 36].

**Figure 5** shows the effect on the EBS for MgB2 of atom displacement along the B–B bond by ~0.6% (*i.e.* a shift of ~0.063 Å) from equilibrium [28]. **Figure 5** shows that the E2g phonon, which is degenerate at Γ with a peak parabola at 398 meV, splits into two separate non-degenerate bands above and below the equilibrium condition. The upper σ band - which we attribute to the heavy effective mass - has a calculated energy 813 meV above the Fermi level. Thus, parallel or nearly parallel FSs attributable to the superconducting condition [44], no longer exist with a 0.6% shift in atom position(s) [28]. A shift of atom position(s) is also reflected in the form and energy of key phonon modes in the corresponding PD for MgB2 [26]. An atom displacement of 0.6% along B–B for MgB2 is not unreasonable at temperatures >40 K [28].

### **Figure 5.**

*Enlarged view of EBS around* Γ *for MgB2 using LDA functional and* Δ*k = 0.018 Å−1 showing (a) degenerate*  σ *bands at equilibrium (blue lines) and (b) after atom displacement Dx = 0.063 Å (red lines) along the E2g mode direction; note the split of* σ *bands causing loss of degeneracy which coincides with loss of superconductivity [26].*

The σ bands for MgB2 consist of two bands degenerate at the Γ point, but degeneracy is lost when the vector k does not equal 0. The two bands thus have different effective masses (or curvatures) which appear to vary as a deformation potential is applied. In fact, the curvatures become very similar and/or identical along Γ–K at the point where a deformation shifts one band tangential to the Fermi level as shown in **Figure 5**. Along Γ–M, the effective masses are shown to cross over, where the top band has the curvature of the original light effective mass band, and for k-vectors away from the origin, the curvature remains the same as the original heavy effective mass band.

The response of the effective masses (*i.e.* the σ bands) to the deformation potential suggests that electronic behavior associated with superconductivity may be explained by directly analyzing the reciprocal space trajectories of these effective masses. Such an interpretation depends on the collective electronic response in reciprocal space which may, by inference, transform into localized information in real space. This outcome is, again, strongly dependent on the use of a fine k-grid value (< 0.015 Å−1) wherein a polynomial approximation to the bands can be calculated [45]. **Figure 6** shows this polynomial calculated as an average band (dotted line) of the two σ bands for the EBS of MgB2 at an equilibrium position and for the degenerate condition arising from a deformation potential shift of 0.063 Å along the B–B bond.

The calculated coefficients for the polynomials, fitted to σH, σL and the average trend line for these bands, σM, are dependent on the k-grid values used in DFT calculations as shown for the equilibrium condition (**Figure 6a**) in **Table 2**. The terms of these polynomial coefficients (*i.e.* for X0 in **Table 2**) for the σH and σL bands show an interesting characteristic with k-grid value. For example, if we assume that the term is in eV and describes the intersection of the polynomial trend curve with the vertical axis (*i.e.* along the y axis), then the difference between coefficients for the σH and σL bands not only varies with k-grid value but also approximates the SC gap energy for MgB2 at finer k-grid values [12, 34, 46]. As the k-grid value increases, the notional SC gap energy also changes and is zero for Δk = 0.03 Å−1. This trend supports the notion that a higher resolution DFT calculation (*i.e.* with Δk < 0.02 Å−1) may provide indicative energy gap values for an SC compound directly from an EBS calculation [45]. **Table 2** also shows coefficients for the polynomials at the displaced condition with boron atom positions along the E2g mode direction [28] calculated for different values of Δk. As expected, with σ<sup>H</sup>

### **Figure 6.**

*Sigma bands (heavy: Red, light: Blue and average: Black dotted) for the EBS of MgB2 along the* Γ*–K and* Γ*–M directions calculated with the LDA functional using* Δ*k = 0.01 Å−1 (a) for equilibrium boron atom positions and (b) for degenerate bands formed by displaced boron atom positions along E2g mode directions by 0.6% relative to the equilibrium position. The green dotted region delineates extent of polynomial trend lines matched to the EBS.*

**121**

*Insights from Systematic DFT Calculations on Superconductors*

**\*** σ**<sup>L</sup>**

**\*** σ**<sup>H</sup>**

and σL tracing parallel bands on either side of Γ, along Γ–K the linear terms show a large energy gap, almost constant as a function of Δk, of ~0.78 eV (**Figure 6b**).

**Coefficient X4 Coefficient X2 Coefficient X0 Coefficient** 

**\*** σ**<sup>H</sup>**

**\*** σ**<sup>L</sup>**

**\***

**\*** σ**<sup>L</sup>**

0.005 4.2955 18.572 8.161 14.749 0.3441 0.3266 0.0175 0.008 8.0342 36.301 11.159 20.344 0.3441 0.3287 0.0154 0.010 4.0135 17.807 7.960 14.371 0.3912 0.3745 0.0167 0.022 6.8879 36.014 10.381 19.477 0.3377 0.3289 0.0088 0.030 6.4614 42.492 10.330 20.213 0.3830 0.3830 0.0

0.005 382.94 36.37 46.15 25.80 0.753 −0.032 0.785 0.008 382.94 36.37 46.15 25.80 0.753 −0.032 0.785 0.010 321.72 19.84 44.98 26.12 0.751 −0.031 0.784

**X0** σ**H-** σ**<sup>L</sup> (eV)**

Conceptually, it is generally accepted that two approaches: (a) the free-electron theory and (b) the tight binding, or linear combination of atomic orbitals (LCAO), offer reasonably good approximations to the conduction and valence bands, respectively, in the electronic structure of materials [1, 2, 47–50]. With increased computational power, the distinction between these two models becomes negligible. In general, the actual EBS should be similar to an average of respective contributions

The detailed origin of particular bands and that of the zero of EF can be more readily appreciated when we examine an extended Brillouin zone scheme, instead of a reduced zone scheme [1, 2, 48, 51, 52]. For example, **Figures 7a**-**c** show periodically repeated reciprocal unit cells with reference to the extended Brillouin zone schemes for the electronic bands of MgB2 along the Γ-M and Γ-K directions, respectively. **Figure 7a** shows a 2D representation of multiple reciprocal unit cells viewed along c\* and identifies reciprocal directions for the calculated EBSs for extended

The origin of EF determined by the parabolic approximation is identified at (i) the cross-over of two parabolas "1" at M in **Figure 7b** and (ii) the inflection point of parabola 3 at K + M in **Figure 7c**. The calculated EF for MgB2 at zero pressure using the LDA functional in CASTEP (and Δk =0.01A−1) is 8.4055 eV. **Figure 7b** and **c** show the location for the origin of EF, at a K + M type reciprocal space position or the midpoint between two reciprocal space Γ vectors along Γ–K. This location is difficult to infer from a reduced BZ scheme, particularly for complex structures. Nevertheless, these two locations at M and ± K ± M nodal points, directly relate to

For MgB2, sections of the σ bands along the Γ–M and Γ–K directions are approximated by upward facing parabolas, even when inside the valence band region.

**3.3 Brillouin zone schemes and high order quartic approximations**

from these two types of approximations.

the real space B–B hexagonal plane in the MgB2 structure.

BZs in **Figure 7b** and **c**.

*DOI: http://dx.doi.org/10.5772/intechopen.96960*

σ**<sup>H</sup>**

At Equilibrium (Dx = 0.0)

Displaced along E2g (Dx = 0.006)

σ*H,* σ*L are the heavy and light sigma bands, respectively.*

*Calculated polynomial coefficients for sigma bands along Γ–K for MgB2.*

**Grid Value (Å−1)**

*\**

**Table 2.**


*Insights from Systematic DFT Calculations on Superconductors DOI: http://dx.doi.org/10.5772/intechopen.96960*

### **Table 2.**

*Real Perspective of Fourier Transforms and Current Developments in Superconductivity*

heavy effective mass band.

of these polynomial coefficients (*i.e.* for X0

the B–B bond.

The σ bands for MgB2 consist of two bands degenerate at the Γ point, but degeneracy is lost when the vector k does not equal 0. The two bands thus have different effective masses (or curvatures) which appear to vary as a deformation potential is applied. In fact, the curvatures become very similar and/or identical along Γ–K at the point where a deformation shifts one band tangential to the Fermi level as shown in **Figure 5**. Along Γ–M, the effective masses are shown to cross over, where the top band has the curvature of the original light effective mass band, and for k-vectors away from the origin, the curvature remains the same as the original

The response of the effective masses (*i.e.* the σ bands) to the deformation potential suggests that electronic behavior associated with superconductivity may be explained by directly analyzing the reciprocal space trajectories of these effective masses. Such an interpretation depends on the collective electronic response in reciprocal space which may, by inference, transform into localized information in real space. This outcome is, again, strongly dependent on the use of a fine k-grid value (< 0.015 Å−1) wherein a polynomial approximation to the bands can be calculated [45]. **Figure 6** shows this polynomial calculated as an average band (dotted line) of the two σ bands for the EBS of MgB2 at an equilibrium position and for the degenerate condition arising from a deformation potential shift of 0.063 Å along

The calculated coefficients for the polynomials, fitted to σH, σL and the average trend line for these bands, σM, are dependent on the k-grid values used in DFT calculations as shown for the equilibrium condition (**Figure 6a**) in **Table 2**. The terms

show an interesting characteristic with k-grid value. For example, if we assume that the term is in eV and describes the intersection of the polynomial trend curve with the vertical axis (*i.e.* along the y axis), then the difference between coefficients for the σH and σL bands not only varies with k-grid value but also approximates the SC gap energy for MgB2 at finer k-grid values [12, 34, 46]. As the k-grid value increases, the notional SC gap energy also changes and is zero for Δk = 0.03 Å−1. This trend supports the notion that a higher resolution DFT calculation (*i.e.* with Δk < 0.02 Å−1) may provide indicative energy gap values for an SC compound directly from an EBS calculation [45]. **Table 2** also shows coefficients for the polynomials at the displaced condition with boron atom positions along the E2g mode direction [28] calculated for different values of Δk. As expected, with σ<sup>H</sup>

*Sigma bands (heavy: Red, light: Blue and average: Black dotted) for the EBS of MgB2 along the* Γ*–K and* Γ*–M directions calculated with the LDA functional using* Δ*k = 0.01 Å−1 (a) for equilibrium boron atom positions and (b) for degenerate bands formed by displaced boron atom positions along E2g mode directions by 0.6% relative to the equilibrium position. The green dotted region delineates extent of polynomial trend lines matched* 

in **Table 2**) for the σH and σL bands

**120**

**Figure 6.**

*to the EBS.*

*Calculated polynomial coefficients for sigma bands along Γ–K for MgB2.*

and σL tracing parallel bands on either side of Γ, along Γ–K the linear terms show a large energy gap, almost constant as a function of Δk, of ~0.78 eV (**Figure 6b**).

### **3.3 Brillouin zone schemes and high order quartic approximations**

Conceptually, it is generally accepted that two approaches: (a) the free-electron theory and (b) the tight binding, or linear combination of atomic orbitals (LCAO), offer reasonably good approximations to the conduction and valence bands, respectively, in the electronic structure of materials [1, 2, 47–50]. With increased computational power, the distinction between these two models becomes negligible. In general, the actual EBS should be similar to an average of respective contributions from these two types of approximations.

The detailed origin of particular bands and that of the zero of EF can be more readily appreciated when we examine an extended Brillouin zone scheme, instead of a reduced zone scheme [1, 2, 48, 51, 52]. For example, **Figures 7a**-**c** show periodically repeated reciprocal unit cells with reference to the extended Brillouin zone schemes for the electronic bands of MgB2 along the Γ-M and Γ-K directions, respectively. **Figure 7a** shows a 2D representation of multiple reciprocal unit cells viewed along c\* and identifies reciprocal directions for the calculated EBSs for extended BZs in **Figure 7b** and **c**.

The origin of EF determined by the parabolic approximation is identified at (i) the cross-over of two parabolas "1" at M in **Figure 7b** and (ii) the inflection point of parabola 3 at K + M in **Figure 7c**. The calculated EF for MgB2 at zero pressure using the LDA functional in CASTEP (and Δk =0.01A−1) is 8.4055 eV. **Figure 7b** and **c** show the location for the origin of EF, at a K + M type reciprocal space position or the midpoint between two reciprocal space Γ vectors along Γ–K. This location is difficult to infer from a reduced BZ scheme, particularly for complex structures. Nevertheless, these two locations at M and ± K ± M nodal points, directly relate to the real space B–B hexagonal plane in the MgB2 structure.

For MgB2, sections of the σ bands along the Γ–M and Γ–K directions are approximated by upward facing parabolas, even when inside the valence band region.

### **Figure 7.**

*(a) Schematic of Brillouin zones for MgB2 viewed along c\* with nodal point nomenclature for primary reciprocal space orientations along* Γ*-M and* Γ*-K. This schematic shows the orientation of the real space asymmetric unit (Mg atoms are yellow; B atoms are gray) as well as nodal points and zones identified in Figure 7b and c. Extended Brillouin zone schemes for the EBS of MgB2 along: (b)* Γ*-M and (c)* Γ*-K. Representative values of energy at the zone boundary and at zone centres are indicated. Energy band sections are labeled as types 1 to 4. The red and blue lines in both EBS schemes refer to similar traces in Figure 1.*

Deviations from parabolas occur particularly at zone boundaries where the periodic crystal potential primarily influences free-electron like level crossings [49, 52, 53]. Along Γ–M, a parabola with convex inflection at Γ occurs at −12.55 eV (parabola 1, **Figure 7b**) and its translated homologs reproduce large sections of the light effective mass σ band distant from Γ.

Similarly, a parabola with convex inflection at M and at -M at −2.152 eV (parabola 3, **Figure 7b**) reproduces the heavy effective mass σ band. Along Γ–K, differentiation of the σ bands is less pronounced in the extended zones but is apparent at Γ (**Figure 7c**). In **Figure 7c** along Γ–K, an additional K–M section is shown because a hexagonal boundary edge, equivalent to K–M by symmetry, transects an adjacent reciprocal space point outside the first BZ (node M1 in **Figure 7a**). Both **Figure 7b** and **c** show the origin of EF for this structure. **Table 3** summarizes values of key parameters associated with these parabolic approximations to extended BZ schemes for the Γ–M and Γ–K directions, respectively.

**123**

*Insights from Systematic DFT Calculations on Superconductors*

**(eV)**

Γ–M 1 −12.5501 −8.4219 1.2821

Γ–K 1 −12.5501 — −7.1550 1.3059

**Energy at M (eV)**

2 −2.9339 0.17329 1.3145 3 0.3979 −2.1520 1.4598 4 1.5331 4.0308 1.3236

2 −2.9339 — — 1.2900 3 — −8.4200 −7.1550 1.4311

**Energy at K (eV)**

**Effective Mass**

The value of EF for a particular DFT calculation is not only sensitive to the k-grid value as shown above but also to other extrinsic conditions such as compositional substitution in a solid-solution and/or changes in applied pressure. **Figure 8** displays the calculated Fermi energies for Al and Sc substitution in Mg1-xAlxB2 and Mg1-xScxB2 determined with the LDA and GGA functionals. For each calculated series, the value of k-grid is constant (i.e. Δk = 0.02 Å−1). **Figure 8** also shows the calculated Fermi energies for other end-member compositions with AlB2-type structure. Note that NbB2 and ZrB2 are reported superconductors at very low temperatures (Tc = 0.6 K and 5.5 K, respectively) albeit non-stoichiometric or substituted niobium diboride (*e.g.* NbB2.5 or Nb0.95Y0.05B2.5) is superconducting at

**Table 4** lists key parameters based on EBS calculations for MgB2 with applied external pressure. For each calculation, the LDA functional and k-grid value is constant and values are computed after geometry optimization. The value for the effective mass, *<sup>H</sup> meff* , is determined from the parabolic approximations described

**Table 4** shows that as pressure is applied, EF largely conforms to the textbook

2 <sup>0</sup>

by using the volume and heavy effective mass obtained at a particular external pressure (assuming that the electron density *n* does not change with pressure [55] as

> ( ) ( ) <sup>0</sup>

*m V* <sup>=</sup>

*eff P P*

*H eff*

2 <sup>3</sup> <sup>0</sup> <sup>0</sup>

*eff <sup>n</sup> <sup>E</sup> m V* <sup>=</sup>

*F*

( )

*P*

*F F H*

*<sup>m</sup> <sup>V</sup> E E*

In general, Eq. (2) provides estimates of EF at ~93% or more of the DFT calculated value without corrections for charge redistribution along bond directions

<sup>2</sup> <sup>2</sup> <sup>3</sup>

(1)

(2)

*DOI: http://dx.doi.org/10.5772/intechopen.96960*

**Orientation Band Type Energy at** Γ

*Calculated parameters at* Γ *and M, K points for MgB2.*

**3.4 Fermi energy values**

**Table 3.**

6.3 K and 9 K, respectively [54].

below in Section 3.4.

shown in Eq. (2):

equation:


### *Insights from Systematic DFT Calculations on Superconductors DOI: http://dx.doi.org/10.5772/intechopen.96960*

### **Table 3.**

*Real Perspective of Fourier Transforms and Current Developments in Superconductivity*

Deviations from parabolas occur particularly at zone boundaries where the periodic crystal potential primarily influences free-electron like level crossings [49, 52, 53]. Along Γ–M, a parabola with convex inflection at Γ occurs at −12.55 eV (parabola 1, **Figure 7b**) and its translated homologs reproduce large sections of the light effec-

*(a) Schematic of Brillouin zones for MgB2 viewed along c\* with nodal point nomenclature for primary reciprocal space orientations along* Γ*-M and* Γ*-K. This schematic shows the orientation of the real space asymmetric unit (Mg atoms are yellow; B atoms are gray) as well as nodal points and zones identified in Figure 7b and c. Extended Brillouin zone schemes for the EBS of MgB2 along: (b)* Γ*-M and (c)* Γ*-K. Representative values of energy at the zone boundary and at zone centres are indicated. Energy band sections are labeled as types 1 to 4. The red and blue lines in both EBS schemes refer to similar traces in* 

Similarly, a parabola with convex inflection at M and at -M at −2.152 eV (parabola 3, **Figure 7b**) reproduces the heavy effective mass σ band. Along Γ–K, differentiation of the σ bands is less pronounced in the extended zones but is apparent at Γ (**Figure 7c**). In **Figure 7c** along Γ–K, an additional K–M section is shown because a hexagonal boundary edge, equivalent to K–M by symmetry, transects an adjacent reciprocal space point outside the first BZ (node M1 in **Figure 7a**). Both **Figure 7b** and **c** show the origin of EF for this structure. **Table 3** summarizes values of key parameters associated with these parabolic approximations to extended BZ schemes

**122**

**Figure 7.**

*Figure 1.*

tive mass σ band distant from Γ.

for the Γ–M and Γ–K directions, respectively.

*Calculated parameters at* Γ *and M, K points for MgB2.*

### **3.4 Fermi energy values**

The value of EF for a particular DFT calculation is not only sensitive to the k-grid value as shown above but also to other extrinsic conditions such as compositional substitution in a solid-solution and/or changes in applied pressure. **Figure 8** displays the calculated Fermi energies for Al and Sc substitution in Mg1-xAlxB2 and Mg1-xScxB2 determined with the LDA and GGA functionals. For each calculated series, the value of k-grid is constant (i.e. Δk = 0.02 Å−1). **Figure 8** also shows the calculated Fermi energies for other end-member compositions with AlB2-type structure. Note that NbB2 and ZrB2 are reported superconductors at very low temperatures (Tc = 0.6 K and 5.5 K, respectively) albeit non-stoichiometric or substituted niobium diboride (*e.g.* NbB2.5 or Nb0.95Y0.05B2.5) is superconducting at 6.3 K and 9 K, respectively [54].

**Table 4** lists key parameters based on EBS calculations for MgB2 with applied external pressure. For each calculation, the LDA functional and k-grid value is constant and values are computed after geometry optimization. The value for the effective mass, *<sup>H</sup> meff* , is determined from the parabolic approximations described below in Section 3.4.

**Table 4** shows that as pressure is applied, EF largely conforms to the textbook equation:

$$E\_F = \frac{\hbar^2}{2m\_{\text{eff}}} \left(\frac{n}{V\_0}\right)^{\frac{1}{2}} \tag{1}$$

by using the volume and heavy effective mass obtained at a particular external pressure (assuming that the electron density *n* does not change with pressure [55] as shown in Eq. (2):

$$E\_{F\_{\mathcal{F}}} = E\_{\mathcal{E}\_0} \left( \frac{m\_{\mathcal{eff}(0)}^H}{m\_{\mathcal{eff}(P)}^H} \right) \left( \frac{\mathbf{V}\_0}{\mathbf{V}\_P} \right)^{\mathcal{Y}\_3} \tag{2}$$

In general, Eq. (2) provides estimates of EF at ~93% or more of the DFT calculated value without corrections for charge redistribution along bond directions

### **Figure 8.**

*Fermi energy (EF) as function of metal substitution in Mg1-xAlxB2 and Mg1-xScxB2 calculated with the LDA and GGA functionals using the CASTEP module of materials studio for* Δ*k = 0.02 Å−1. Calculated Fermi energies for end-member compositions of AlB2-type structures are also shown.*


### **Table 4.**

*List of calculated Fermi energy (EF) values for MgB2 at pressure using Eq. (2).*

which take place as pressure is applied [55]. Use of parabolic approximations in this manner may provide a useful benchmark (or "rule of thumb") for models of predicted new compounds.

### **3.5 Phonon dispersions – variation with k-grid**

As noted in earlier publications [17, 23, 25, 34], the k-grid value also influences the form and mode order of phonons in a DFT calculated PD. **Figure 6** demonstrates this influence on the MgB2 PD for the range 0.02 Å−1 < Δk < 0.06 Å−1. The more regularly shaped phonon anomaly becomes apparent with smaller k-grid value and is evident for Δk = 0.02 Å−1 (circled; **Figure 8a**). For values of Δk > 0.05 Å−1, the calculated PD for MgB2 implies that the phase is unstable yet we know from experimental evidence that this is not the case. For SC compounds with lower values of Tc and/or where Fermi surfaces closely intersect the Fermi

**125**

**Figure 9.**

*compound.*

*Insights from Systematic DFT Calculations on Superconductors*

level with minimal difference in reciprocal space, the sensitivity of the PD to k-grid value will be shifted towards smaller k-grids compared to the effect

Measurement of the parameter, δ, shown in **Figure 9a** provides a reliable estimate of Tc for MgB2 when the PD is calculated with Δk < 0.02 Å−1 [23]. This approach, which determines the thermal energy, Tδ, of the key E2g phonon mode using an empirical formula [23], precisely tracks the experimentally determined reduction of Tc for metal-substituted forms of MgB2 such as (Mg1-xAlx)B2 and (Mg1-xScx)B2 [23, 25]. When the value for δ is determined with two sequential DFT calculations using the LDA and the GGA functionals, error estimates (in terms of the amplitude of the spread of the DFT approximations) for the value of Tδ at each

We have used this approach to estimate the likely value(s) of Tc for other metal-substituted forms of MgB2 that have received limited attention or have not been identified previously in the literature. For example, we determined the PD for (Mg1-xBax)B2, and for (Mg1-xCdx)B2 where x = 0.33, 0.5 or 0.66 [23, 25]. **Figure 10a** shows the PD for (Mg0.5Ba0.5)B2 calculated using the LDA functional with Δk = 0.02 Å−1. Measurement of the four values for δ in **Figure 10a** (*i.e.* two non-degenerate E2g modes each in the Γ–K and Γ–M directions) and conversion to Tδ gives an average value of 58.1 ± 3.4 K for (Mg0.5Ba0.5)B2. **Figure 10b** is adapted from the work of Palnichenko *et al*. [56] in which Ba, Rb and Cs were substituted into MgB2 *via* solid state synthesis. In all cases, the Tc determined experimentally using magnetic susceptibility is higher than that for MgB2. Unfortunately, while the effects of substitution are evident, the explicit levels of substitution were not

*Phonon dispersions for MgB2 calculated using the LDA functional with different k-grids: (a)* Δ*k = 0.02 Å−1, (b)* Δ*k = 0.04 Å−1, (c)* Δ*k = 0.05 Å−1 and (d)* Δ*k = 0.06 Å−1. E2g (red) and B2g (blue) phonon modes are highlighted. The energy associated with the E2g phonon anomaly in (a) is ~16 meV [23]. Note the negative phonon frequencies for* Δ*k = 0.06 Å−1 which, due to insufficient k-grid resolution, implies an unstable* 

*DOI: http://dx.doi.org/10.5772/intechopen.96960*

level of metal substitution can be obtained.

with MgB2.

determined [45].

*Insights from Systematic DFT Calculations on Superconductors DOI: http://dx.doi.org/10.5772/intechopen.96960*

*Real Perspective of Fourier Transforms and Current Developments in Superconductivity*

which take place as pressure is applied [55]. Use of parabolic approximations in this manner may provide a useful benchmark (or "rule of thumb") for models of

*Fermi energy (EF) as function of metal substitution in Mg1-xAlxB2 and Mg1-xScxB2 calculated with the LDA and GGA functionals using the CASTEP module of materials studio for* Δ*k = 0.02 Å−1. Calculated Fermi* 

 27.8912 0.5070 8.4055 8.4055 100.0 27.5344 0.5059 8.5368 8.4964 99.5 26.5752 0.5035 8.9059 8.7411 98.1 25.7498 0.5018 9.2430 8.9572 96.9 25.0255 0.5003 9.5549 9.1565 95.8 24.3834 0.4990 9.8450 9.3409 94.9 23.8050 0.4983 10.1183 9.5049 94.0 23.2793 0.4981 10.3776 9.6513 93.0

**Fermi energy [eV]**

**EF(P)**

**[eV]** *F formula* (**1**)

*E E* − − **[%]**

*F DFT calculated*

*<sup>H</sup> meff* **[me] \***

*energies for end-member compositions of AlB2-type structures are also shown.*

the form and mode order of phonons in a DFT calculated PD. **Figure 6** demonstrates this influence on the MgB2 PD for the range 0.02 Å−1 < Δk < 0.06 Å−1. The more regularly shaped phonon anomaly becomes apparent with smaller k-grid value and is evident for Δk = 0.02 Å−1 (circled; **Figure 8a**). For values of Δk > 0.05 Å−1, the calculated PD for MgB2 implies that the phase is unstable yet we know from experimental evidence that this is not the case. For SC compounds with lower values of Tc and/or where Fermi surfaces closely intersect the Fermi

As noted in earlier publications [17, 23, 25, 34], the k-grid value also influences

**124**

*\**

**Table 4.**

predicted new compounds.

*me is the electron mass.*

**Figure 8.**

**Pressure [GPa]**

**Unit cell volume [A3 ]**

**3.5 Phonon dispersions – variation with k-grid**

*List of calculated Fermi energy (EF) values for MgB2 at pressure using Eq. (2).*

level with minimal difference in reciprocal space, the sensitivity of the PD to k-grid value will be shifted towards smaller k-grids compared to the effect with MgB2.

Measurement of the parameter, δ, shown in **Figure 9a** provides a reliable estimate of Tc for MgB2 when the PD is calculated with Δk < 0.02 Å−1 [23]. This approach, which determines the thermal energy, Tδ, of the key E2g phonon mode using an empirical formula [23], precisely tracks the experimentally determined reduction of Tc for metal-substituted forms of MgB2 such as (Mg1-xAlx)B2 and (Mg1-xScx)B2 [23, 25]. When the value for δ is determined with two sequential DFT calculations using the LDA and the GGA functionals, error estimates (in terms of the amplitude of the spread of the DFT approximations) for the value of Tδ at each level of metal substitution can be obtained.

We have used this approach to estimate the likely value(s) of Tc for other metal-substituted forms of MgB2 that have received limited attention or have not been identified previously in the literature. For example, we determined the PD for (Mg1-xBax)B2, and for (Mg1-xCdx)B2 where x = 0.33, 0.5 or 0.66 [23, 25]. **Figure 10a** shows the PD for (Mg0.5Ba0.5)B2 calculated using the LDA functional with Δk = 0.02 Å−1. Measurement of the four values for δ in **Figure 10a** (*i.e.* two non-degenerate E2g modes each in the Γ–K and Γ–M directions) and conversion to Tδ gives an average value of 58.1 ± 3.4 K for (Mg0.5Ba0.5)B2. **Figure 10b** is adapted from the work of Palnichenko *et al*. [56] in which Ba, Rb and Cs were substituted into MgB2 *via* solid state synthesis. In all cases, the Tc determined experimentally using magnetic susceptibility is higher than that for MgB2. Unfortunately, while the effects of substitution are evident, the explicit levels of substitution were not determined [45].

### **Figure 9.**

*Phonon dispersions for MgB2 calculated using the LDA functional with different k-grids: (a)* Δ*k = 0.02 Å−1, (b)* Δ*k = 0.04 Å−1, (c)* Δ*k = 0.05 Å−1 and (d)* Δ*k = 0.06 Å−1. E2g (red) and B2g (blue) phonon modes are highlighted. The energy associated with the E2g phonon anomaly in (a) is ~16 meV [23]. Note the negative phonon frequencies for* Δ*k = 0.06 Å−1 which, due to insufficient k-grid resolution, implies an unstable compound.*

### **Figure 10.**

*(a) Phonon dispersion for (Mg0.5Ba0.5)B2 calculated using the LDA functional for* Δ*k = 0.02 Å−1, showing the extent,* δ*, of the E2g phonon anomaly (in red) along the* Γ*–K and* Γ*–M directions; (b) magnetic susceptibility for MgB2 and metal substituted forms of MgB2 showing experimentally determined Tc values (arrowed); adapted from Figure 1 of Palnichenko et al. [56]. Substitution of Ba, Rb and Cs shows a higher Tc than for MgB2.*

### **4. Discussion**

The calculated EF and Fermi level allow systematic comparison of EBSs from a structural family or group of materials with varying properties. Moreover, the FS is generally of a well-defined orbital character and topology determined by the value of EF as a result of bands that cross the Fermi level. The Fermi level determined from DFT calculations is defined as at zero energy while the calculated value of EF obtained after an accurate DFT calculation is seldom described in the published literature. By definition, the Fermi level is determined in the ground state by the filling of lower energy electronic states by all the (nearly) free electrons up to a highest possible value of the energy, which in practice should correspond to EF [49, 50, 55, 57, 58].

### **4.1 Fermi energy**

The superconducting gap in many compounds (*e.g.* diborides, disilicides, A15 compounds, B-doped diamond) is in the meV range of energy [59]. For many SC materials, the gap is directly linked to the separation of parallel, or nearly parallel, FSs that may not be identifiable if the k-grid value is at an insufficient resolution [45]. **Table 1** also shows that for MgB2 differences in EF and enthalpy of a few tens of meV are associated with exceptionally small differences in lattice parameter, of the order ~10−5 Å. This attribute highlights the robustness and sensitivity of DFT calculations, particularly when represented in reciprocal space. More importantly, these differences in EF, attributed to differences in k-grid value, are substantially greater than the superconducting gap for MgB2 [60]. Hence, detection of a gap – which in an EBS for MgB2 is related to the separation of σ bands crossing the Fermi level – may not be achieved with low resolution DFT calculations.

We demonstrate this issue using the EBS for MgB2 as shown in **Figure 11**. In this figure, we have reproduced the EBS for MgB2 as calculated using the LDA functional for Δk = 0.018 Å−1. The Fermi level is set at 0 eV and a notional "Fermi level 2" is also shown as a red dotted line at −250 meV. As noted in **Table 1**, a change in calculated EF ~ 200 meV may occur with choice of Δk > 0.04 Å−1. The intersection of σ bands with the calculated Fermi level are separated by a distance λ1, which also defines the separation between Fermi surfaces for MgB2.

However, if the value of Δk, or the calculated value for EF, results in a shift of the Fermi level by ~250 meV, the separation of Fermi surfaces, illustrated by λ2,

**127**

~0.2 eV) for EF.

determination of Tc.

**Figure 11.**

*EBS for MgB2.*

**4.2 Computational resolution**

*Insights from Systematic DFT Calculations on Superconductors*

is different. We estimate that this order of Fermi level or EF shift may result in discrepancies between 20% and 35% of the value(s) for λ. As shown in **Figure 11**, the shape of the σ bands around Γ are asymmetric. The difference in value(s) for λ with variation in EF, will accordingly be dependent on the form, or shape, of these parabolas. Projections of the density of states at the Fermi level will also be affected by this shift of EF as will the outcomes of Eliashberg or McMillan equations for the

*Representative EBS showing the effect of a change in Fermi level of 250 meV (red dotted line). The values for*  λ*1 and* λ*2, which define the distance between parallel Fermi surfaces, and the values for d1 and d2 (the energy above the Fermi level at the vertex of the* σ *band parabola), are not equivalent. This example is based on the* 

The apparent discrepancy in determination of the value for EF, noted in the Introduction, may be elucidated by examination of **Figure 11**. For example, we suggest that some researchers define a value for EF as equivalent to the energy shown as d1 in **Figure 11** (*i.e.* the distance to the vertex of the parabolic band [61]). If this definition for EF is used, then a shift in the Fermi level as described above would lead to an estimate equivalent to d2 in **Figure 11**. The literature contains calculations of EBS where the specific approach to determine EF is not identified; for superconductors, this practice provides an unfortunate level of uncertainty. An uncertain position for the Fermi level will also result in varying cross-sectional areas of the FS at the Fermi level (see **Figure 2**), which directly determines the period of sensitive quantum oscillation measurements [62]. Discrepancies between DFT predicted values and experimental quantum oscillations may be reconciled by revisiting the choice of k-grid.

**Table 5** provides a summary of reports on previous DFT calculations for MgB2 and of systematic calculations from this study. This table highlights the diversity of computational methods used to date as well as wide variations in parameters such as k-grid value and the cut-off energy. Systematic evaluation of these two parameters shows that the value for EF may differ by several hundred meV for the same cut-off energy with change in Δk value. For our systematic calculations of these parameters shown in **Table 5**, the LDA functional is used for consistency. Calculations with the GGA functional show similar trends albeit at different absolute values (by

**Table 5** shows that a low value for cut-off energy (*i.e.* < 500 eV) results in a value for EF > 1 eV different to that with cut-off energy >500 eV for calculations using

*DOI: http://dx.doi.org/10.5772/intechopen.96960*

*Insights from Systematic DFT Calculations on Superconductors DOI: http://dx.doi.org/10.5772/intechopen.96960*

### **Figure 11.**

*Real Perspective of Fourier Transforms and Current Developments in Superconductivity*

The calculated EF and Fermi level allow systematic comparison of EBSs from a structural family or group of materials with varying properties. Moreover, the FS is generally of a well-defined orbital character and topology determined by the value of EF as a result of bands that cross the Fermi level. The Fermi level determined from DFT calculations is defined as at zero energy while the calculated value of EF obtained after an accurate DFT calculation is seldom described in the published literature. By definition, the Fermi level is determined in the ground state by the filling of lower energy electronic states by all the (nearly) free electrons up to a highest possible value of the energy, which in practice should

*(a) Phonon dispersion for (Mg0.5Ba0.5)B2 calculated using the LDA functional for* Δ*k = 0.02 Å−1, showing the extent,* δ*, of the E2g phonon anomaly (in red) along the* Γ*–K and* Γ*–M directions; (b) magnetic susceptibility for MgB2 and metal substituted forms of MgB2 showing experimentally determined Tc values (arrowed); adapted from Figure 1 of Palnichenko et al. [56]. Substitution of Ba, Rb and Cs shows a higher Tc than* 

The superconducting gap in many compounds (*e.g.* diborides, disilicides, A15 compounds, B-doped diamond) is in the meV range of energy [59]. For many SC materials, the gap is directly linked to the separation of parallel, or nearly parallel, FSs that may not be identifiable if the k-grid value is at an insufficient resolution [45]. **Table 1** also shows that for MgB2 differences in EF and enthalpy of a few tens of meV are associated with exceptionally small differences in lattice parameter, of the order ~10−5 Å. This attribute highlights the robustness and sensitivity of DFT calculations, particularly when represented in reciprocal space. More importantly, these differences in EF, attributed to differences in k-grid value, are substantially greater than the superconducting gap for MgB2 [60]. Hence, detection of a gap – which in an EBS for MgB2 is related to the separation of σ bands crossing the Fermi

We demonstrate this issue using the EBS for MgB2 as shown in **Figure 11**. In this figure, we have reproduced the EBS for MgB2 as calculated using the LDA functional for Δk = 0.018 Å−1. The Fermi level is set at 0 eV and a notional "Fermi level 2" is also shown as a red dotted line at −250 meV. As noted in **Table 1**, a change in calculated EF ~ 200 meV may occur with choice of Δk > 0.04 Å−1. The intersection of σ bands with the calculated Fermi level are separated by a distance λ1, which also

However, if the value of Δk, or the calculated value for EF, results in a shift of the Fermi level by ~250 meV, the separation of Fermi surfaces, illustrated by λ2,

level – may not be achieved with low resolution DFT calculations.

defines the separation between Fermi surfaces for MgB2.

**126**

**4. Discussion**

**Figure 10.**

*for MgB2.*

**4.1 Fermi energy**

correspond to EF [49, 50, 55, 57, 58].

*Representative EBS showing the effect of a change in Fermi level of 250 meV (red dotted line). The values for*  λ*1 and* λ*2, which define the distance between parallel Fermi surfaces, and the values for d1 and d2 (the energy above the Fermi level at the vertex of the* σ *band parabola), are not equivalent. This example is based on the EBS for MgB2.*

is different. We estimate that this order of Fermi level or EF shift may result in discrepancies between 20% and 35% of the value(s) for λ. As shown in **Figure 11**, the shape of the σ bands around Γ are asymmetric. The difference in value(s) for λ with variation in EF, will accordingly be dependent on the form, or shape, of these parabolas. Projections of the density of states at the Fermi level will also be affected by this shift of EF as will the outcomes of Eliashberg or McMillan equations for the determination of Tc.

The apparent discrepancy in determination of the value for EF, noted in the Introduction, may be elucidated by examination of **Figure 11**. For example, we suggest that some researchers define a value for EF as equivalent to the energy shown as d1 in **Figure 11** (*i.e.* the distance to the vertex of the parabolic band [61]). If this definition for EF is used, then a shift in the Fermi level as described above would lead to an estimate equivalent to d2 in **Figure 11**. The literature contains calculations of EBS where the specific approach to determine EF is not identified; for superconductors, this practice provides an unfortunate level of uncertainty. An uncertain position for the Fermi level will also result in varying cross-sectional areas of the FS at the Fermi level (see **Figure 2**), which directly determines the period of sensitive quantum oscillation measurements [62]. Discrepancies between DFT predicted values and experimental quantum oscillations may be reconciled by revisiting the choice of k-grid.

### **4.2 Computational resolution**

**Table 5** provides a summary of reports on previous DFT calculations for MgB2 and of systematic calculations from this study. This table highlights the diversity of computational methods used to date as well as wide variations in parameters such as k-grid value and the cut-off energy. Systematic evaluation of these two parameters shows that the value for EF may differ by several hundred meV for the same cut-off energy with change in Δk value. For our systematic calculations of these parameters shown in **Table 5**, the LDA functional is used for consistency. Calculations with the GGA functional show similar trends albeit at different absolute values (by ~0.2 eV) for EF.

**Table 5** shows that a low value for cut-off energy (*i.e.* < 500 eV) results in a value for EF > 1 eV different to that with cut-off energy >500 eV for calculations using

the same Δk value (*e.g.* compare Δk = 0.03 Å−1 calculated for MgB2 in **Table 5**). The importance of such parameters has been noted in the literature primarily in relation to PD calculations [17, 63]. However, the specific impact of both computational parameters on EF and the effect on band structures has not previously been enumerated for MgB2 nor for other SCs.

**Table 5** also lists the variation in energy, ΔEv (in eV), between the Fermi level and the vertex of the parabola at Γ for different values of Δk and for two cut-off energies using the LDA functional for the EBS of MgB2 (in **Figure 11**, this energy is represented as d1). As we have noted for EF, there are substantial variations (*i.e.* > 100 meV) in ΔEv with choice of Δk and cut-off energy. Calculated outcomes in our systematic study of MgB2 parameters over a wide range of input parameters as listed in **Table 5**, show that for MgB2, Δk < 0.008 Å−1 and a cut-off energy >900 eV, provides reliable determination of meV phenomena in this structure and in substitutional analogues of MgB2. We note that these attributes apply to plane wave calculations. We are yet to undertake a systematic evaluation of augmented plane wave calculations using similar strategies.

The calculations by de la Pena-Seaman [71] on the transformation of Fermi surfaces with substitution of Al and C into MgB2 and recent work by Pesic *et al*. [72] are notable exceptions on the previous studies shown in **Table 5** albeit each with a low cut-off energy. Note that a cut-off energy of 500 eV in **Table 5** results in Fermi energies similar to those obtained for molecular fragments obtained by the ADF software (data not shown). This suggests that calculations with smaller cut-off energy do not adequately capture periodic crystal behavior, but instead, model a set of values that are molecule-like. Some DFT studies reveal inherent inconsistencies in EBS and PD calculations for known superconductor materials due to insufficient computational resolution. This aspect of DFT models also appears to confuse the peer review process for some journal papers.

### **4.3 Phonon dispersions and k-grid**

We have examined the changes in PD form and mode order for the substitutional series Mg1-xAlxB2 [23] and Mg1-xScxB2 [25] where 0 < x < 1. For PDs, the value of k-grid in a DFT calculation may obscure phenomena that imply superconductivity such as the presence or absence of a phonon anomaly [5, 34]. We have also demonstrated for MgB2 that the change in the E2g phonon anomaly varies with applied pressure and correlates with the experimentally determined change in Tc [27]. For these cases, we show that a temperature, calculated from the extent of the anomaly, Tδ, is a reliable *ab initio* indicator of Tc determined by experiment [23, 24, 27]. A fine k-grid (or a k-grid value smaller than ~0.025 Å−1 depending on the structure) is important for PD plots of SCs with AlB2-type structures and for estimations of Tc for BCS-type compounds that display a phonon anomaly [5, 23].

In an earlier publication [25], we compare for MgB2 the calculation of Tc (i) using the McMillan formalism of the Eliashberg model [12] and (ii) using the E2g phonon anomaly energy, Tδ, as noted above [23, 27, 28]. In both cases, with suitable assumptions for the McMillan formalism, the "predictive" fidelity of either method adequately matches experimental data. However, the Eliashberg model requires an estimate for two key parameters, λ and μ\*, based on average values of electron–phonon behavior summed over all orientations. In practice, determination of λ and/or μ\* by *a priori* methods is non-trivial for compounds with indeterminate physical properties [67]. In this regard, we applaud the recent advances in mathematical formalism and computational implementation of the Eliashberg model by Sanna *et al*. [19].

**129**

**DFT Code**

[64] Eliashberg Eliashberg+ Eliashberg+

SCDFT

VASP WIEN2k\*

WIEN2k

SIESTA QExpr'o

MBPP\* QExpr'o+ CASTEP CASTEP CASTEP CASTEP CASTEP CASTEP CASTEP CASTEP

0.020 *~0.022* *~0.018* *~0.050* *~0.015* *~0.019* *~0.025* *~0.021* *~0.018* *~0.015* *~0.010* *~0.007*

0.100 0.060 0.050 0.040 0.030 0.020 0.015 0.008

15x15x11 18x18x12 18x18x18

8x8x8 24x24x22 19x19x15 14x14x14 18x18x14 20x20x20 24x24x24 36x36x36 48x48x48

4x4x4 6x6x6 8x8x6 9x9x8 12x12x10 19x19x14 25x25x20 47x47x36

3744

990

8.4038

0.343

LDA

650

990

8.3991

0.352

LDA

280

990

8.3976

0.348

LDA

95

990

8.3773

0.383

LDA

48

990

8.4706

0.328

LDA

[5]

[5]

[23]

[5]

\*\*

30

990

8.5190

0.419

LDA

21

990

8.2110

0.636

LDA

8

990

8.8103

−0.097

LDA

na

408

na

na

GGA

na

218

na

na

Both

na

612

na

na

GGA

na

na

na

na

GGA

na

82

na

na

LDA

144

122

na

na

GGA

na

503

na

na

GGA

na

340

na

na

GGA

na

245

na

na

na

na

218

na

na

LDA

na

816

na

na

LDA

na

500

na

na

GGA

[38]

[65]

[17]

[66]

[20]

[67]

[44]

[68]

[69]

[70]

[71]

[72]

\*\*

\*\*

\*\*

Δ**k value (Å−1)**

**Grid**

**No. of** 

**Energy Cut-off** 

**Fermi Energy**

Δ**Ev**

**# (eV)**

**LDA** 

**Ref.**

**or GGA**

**(eV)**

**(eV)**

**k-points**

*Insights from Systematic DFT Calculations on Superconductors*

*DOI: http://dx.doi.org/10.5772/intechopen.96960*


### *Insights from Systematic DFT Calculations on Superconductors DOI: http://dx.doi.org/10.5772/intechopen.96960*

*Real Perspective of Fourier Transforms and Current Developments in Superconductivity*

merated for MgB2 nor for other SCs.

wave calculations using similar strategies.

peer review process for some journal papers.

**4.3 Phonon dispersions and k-grid**

phonon anomaly [5, 23].

model by Sanna *et al*. [19].

the same Δk value (*e.g.* compare Δk = 0.03 Å−1 calculated for MgB2 in **Table 5**). The importance of such parameters has been noted in the literature primarily in relation to PD calculations [17, 63]. However, the specific impact of both computational parameters on EF and the effect on band structures has not previously been enu-

**Table 5** also lists the variation in energy, ΔEv (in eV), between the Fermi level and the vertex of the parabola at Γ for different values of Δk and for two cut-off energies using the LDA functional for the EBS of MgB2 (in **Figure 11**, this energy is represented as d1). As we have noted for EF, there are substantial variations (*i.e.* > 100 meV) in ΔEv with choice of Δk and cut-off energy. Calculated outcomes in our systematic study of MgB2 parameters over a wide range of input parameters as listed in **Table 5**, show that for MgB2, Δk < 0.008 Å−1 and a cut-off energy >900 eV, provides reliable determination of meV phenomena in this structure and in substitutional analogues of MgB2. We note that these attributes apply to plane wave calculations. We are yet to undertake a systematic evaluation of augmented plane

The calculations by de la Pena-Seaman [71] on the transformation of Fermi surfaces with substitution of Al and C into MgB2 and recent work by Pesic *et al*. [72] are notable exceptions on the previous studies shown in **Table 5** albeit each with a low cut-off energy. Note that a cut-off energy of 500 eV in **Table 5** results in Fermi energies similar to those obtained for molecular fragments obtained by the ADF software (data not shown). This suggests that calculations with smaller cut-off energy do not adequately capture periodic crystal behavior, but instead, model a set of values that are molecule-like. Some DFT studies reveal inherent inconsistencies in EBS and PD calculations for known superconductor materials due to insufficient computational resolution. This aspect of DFT models also appears to confuse the

We have examined the changes in PD form and mode order for the substitutional series Mg1-xAlxB2 [23] and Mg1-xScxB2 [25] where 0 < x < 1. For PDs, the value of k-grid in a DFT calculation may obscure phenomena that imply superconductivity such as the presence or absence of a phonon anomaly [5, 34]. We have also demonstrated for MgB2 that the change in the E2g phonon anomaly varies with applied pressure and correlates with the experimentally determined change in Tc [27]. For these cases, we show that a temperature, calculated from the extent of the anomaly, Tδ, is a reliable *ab initio* indicator of Tc determined by experiment [23, 24, 27]. A fine k-grid (or a k-grid value smaller than ~0.025 Å−1 depending on the structure) is important for PD plots of SCs with AlB2-type structures and for estimations of Tc for BCS-type compounds that display a

In an earlier publication [25], we compare for MgB2 the calculation of Tc (i) using the McMillan formalism of the Eliashberg model [12] and (ii) using the E2g phonon anomaly energy, Tδ, as noted above [23, 27, 28]. In both cases, with suitable assumptions for the McMillan formalism, the "predictive" fidelity of either method adequately matches experimental data. However, the Eliashberg model requires an estimate for two key parameters, λ and μ\*, based on average values of electron–phonon behavior summed over all orientations. In practice, determination of λ and/or μ\* by *a priori* methods is non-trivial for compounds with indeterminate physical properties [67]. In this regard, we applaud the recent advances in mathematical formalism and computational implementation of the Eliashberg

**128**


**Table 5.**

**131**

*Insights from Systematic DFT Calculations on Superconductors*

An and Pickett [36] estimate that the influence of the E2g mode is at least a factor of 25 times greater than all other phonon modes in MgB2. The E2g mode is predominantly associated with movement within the boron planes of MgB2; that is, along specific orientations [73]. Nevertheless, use of an average value for phonon frequencies integrated over all directions in reciprocal space is a feature of the McMillan formalism that provides a reasonable "*post facto*" estimate of Tc presumably because the E2g mode is so dominant. Such coincidence does not enable, nor guarantee, *ab initio* predictive capacity for *a priori* models, particularly if evaluating structures for which experimental data are limited or unavailable. Thus, we advocate an alternative approach for superconductivity prediction that complements the McMillan formalism. In this alternative approach, appropriate values for Δk and the cut-off energy enable *ab initio* DFT calculations to estimate values for Tδ that correlate with experimentally determined values for the Tc of MgB2 [23, 27], for compounds of the form (Mg1-xMx)B2 (where M = Al, Sc, Ti) [23–25], and for disilicides [23] and metal

The predictive value of the approaches we advocate to estimate Tc that utilizes calculation of a value for Tδ using a phonon anomaly [23–25, 74] is evident for Ba-substitution into MgB2 [56]. Our estimates for (Mg1-xBax)B2 at three levels of Ba substitution (x = 0.33, 0.5 and 0.66) and using both LDA and GGA approximations suggest that 62.1 K < Tδ < 64.4 K with an error of ±4.9 K. These estimates are higher by ~15 K than the experimentally determined value of ~45 K by Palnichenko *et al*. [56]. However, the extent of Ba substitution in MgB2 was not determined in this experimental work; albeit 11B NMR analysis shows that the final product has the same site symmetry as MgB2 [56]. Substitution of Ba in MgB2 at levels less than 33%

The presence of multiple phases in the Rb- and Cs- substituted forms of MgB2 synthesized by Palnichenko *et al*. [56] is difficult to verify from the data presented due to limited microstructural and compositional characterization. However, we note that PD calculations on a nominal 50:50 ratio for Rb:Mg and Cs:Mg for substituted MgB2 results in asymmetric and multi-level anomalies (data not shown) similar to that shown in **Figure 11a**. By measuring the extent of the anomaly in each of these cases, the values for Tδ are similar to the onsets of transitions shown for these compositions in **Figure 11b**. While circumstantial, this combination of modeling and experiment suggests that these substituted MgB2 compositions may be homogeneous single phase. Further analyses of this compositional suite, and that of (Mg1-xCdx)B2 may reveal additional SC compounds in the AlB2-type structural

Fully converged PDs are a useful indicator of phase stability [26, 74]. The sensitivity of PDs to changes in stoichiometry, composition or Δk is significantly higher than typically encountered in an EBS [26]. The PD calculated at a deliberately large k-grid value 0.06 Å−1 in **Figure 9d** may be interpreted as a dynamic instability. MgB2 is a well-studied case and we know that this is not correct; however, for unknown or other materials with closer FSs in reciprocal space, we would expect similar phenomena to be manifest at smaller k-grids. Thus, sometimes conclusions about phase transitions may be artifacts of the DFT calculation if k-grids of insufficient resolution are used for materials with approximately parallel FSs in close

Electronic bands and FSs of constant energy possess all point symmetries of a crystal as a function of position in reciprocal space [48, 51]. The intersections of the σ bands with the Fermi level, as shown in **Figure 1**, determine points that, by

group with significantly enhanced superconducting properties to MgB2.

*DOI: http://dx.doi.org/10.5772/intechopen.96960*

hexaborides [74].

may result in a lower value for Tc.

reciprocal space proximity [45].

**4.4 Fermi surfaces and superconductivity**

*#Equivalent energy to d1 in* **Figure 11***.*

*Comparison of computational settings for MgB2.*

*Real Perspective of Fourier Transforms and Current Developments in Superconductivity*

**130**

**DFT Code**

CASTEP CASTEP CASTEP CASTEP CASTEP

0.005 0.005 0.008 0.020 0.030 *"na" – data not available in publication.\*\*This work; italicized values are estimates by this study.*

*\*Also describes (Mg,Al)B2; grid size of EBS calculation is 18x18x18 for [71].*

*+Uses a 12x12x12 grid for PD calculations [72].*

*#Equivalent energy to d1 in* **Figure 11***.*

**Table 5.**

*Comparison of computational settings for MgB2.*

75x75x58 75x75x58 47x47x36 19x19x14 12x12x10

95

500

7.2449

0.475

LDA

280

500

7.3218

0.370

LDA

3744

500

7.3285

0.372

LDA

14703

500

7.3295

0.371

LDA

14703

990

8.4040

0.343

LDA

\*\*

\*\*

\*\*

\*\*

\*\*

Δ**k value (Å−1)**

**Grid**

**No. of** 

**Energy Cut-off** 

**Fermi Energy**

Δ**Ev**

**# (eV)**

**LDA** 

**Ref.**

**or GGA**

**(eV)**

**(eV)**

**k-points**

An and Pickett [36] estimate that the influence of the E2g mode is at least a factor of 25 times greater than all other phonon modes in MgB2. The E2g mode is predominantly associated with movement within the boron planes of MgB2; that is, along specific orientations [73]. Nevertheless, use of an average value for phonon frequencies integrated over all directions in reciprocal space is a feature of the McMillan formalism that provides a reasonable "*post facto*" estimate of Tc presumably because the E2g mode is so dominant. Such coincidence does not enable, nor guarantee, *ab initio* predictive capacity for *a priori* models, particularly if evaluating structures for which experimental data are limited or unavailable. Thus, we advocate an alternative approach for superconductivity prediction that complements the McMillan formalism. In this alternative approach, appropriate values for Δk and the cut-off energy enable *ab initio* DFT calculations to estimate values for Tδ that correlate with experimentally determined values for the Tc of MgB2 [23, 27], for compounds of the form (Mg1-xMx)B2 (where M = Al, Sc, Ti) [23–25], and for disilicides [23] and metal hexaborides [74].

The predictive value of the approaches we advocate to estimate Tc that utilizes calculation of a value for Tδ using a phonon anomaly [23–25, 74] is evident for Ba-substitution into MgB2 [56]. Our estimates for (Mg1-xBax)B2 at three levels of Ba substitution (x = 0.33, 0.5 and 0.66) and using both LDA and GGA approximations suggest that 62.1 K < Tδ < 64.4 K with an error of ±4.9 K. These estimates are higher by ~15 K than the experimentally determined value of ~45 K by Palnichenko *et al*. [56]. However, the extent of Ba substitution in MgB2 was not determined in this experimental work; albeit 11B NMR analysis shows that the final product has the same site symmetry as MgB2 [56]. Substitution of Ba in MgB2 at levels less than 33% may result in a lower value for Tc.

The presence of multiple phases in the Rb- and Cs- substituted forms of MgB2 synthesized by Palnichenko *et al*. [56] is difficult to verify from the data presented due to limited microstructural and compositional characterization. However, we note that PD calculations on a nominal 50:50 ratio for Rb:Mg and Cs:Mg for substituted MgB2 results in asymmetric and multi-level anomalies (data not shown) similar to that shown in **Figure 11a**. By measuring the extent of the anomaly in each of these cases, the values for Tδ are similar to the onsets of transitions shown for these compositions in **Figure 11b**. While circumstantial, this combination of modeling and experiment suggests that these substituted MgB2 compositions may be homogeneous single phase. Further analyses of this compositional suite, and that of (Mg1-xCdx)B2 may reveal additional SC compounds in the AlB2-type structural group with significantly enhanced superconducting properties to MgB2.

Fully converged PDs are a useful indicator of phase stability [26, 74]. The sensitivity of PDs to changes in stoichiometry, composition or Δk is significantly higher than typically encountered in an EBS [26]. The PD calculated at a deliberately large k-grid value 0.06 Å−1 in **Figure 9d** may be interpreted as a dynamic instability. MgB2 is a well-studied case and we know that this is not correct; however, for unknown or other materials with closer FSs in reciprocal space, we would expect similar phenomena to be manifest at smaller k-grids. Thus, sometimes conclusions about phase transitions may be artifacts of the DFT calculation if k-grids of insufficient resolution are used for materials with approximately parallel FSs in close reciprocal space proximity [45].

### **4.4 Fermi surfaces and superconductivity**

Electronic bands and FSs of constant energy possess all point symmetries of a crystal as a function of position in reciprocal space [48, 51]. The intersections of the σ bands with the Fermi level, as shown in **Figure 1**, determine points that, by

construction, belong to the FSs. The FS corresponding to σ bands in the reduced BZ become two approximately parallel tubes [28], as schematically represented in **Figure 12** below. As shown for MgB2 in earlier work [23, 24, 27, 28] and by others [65, 75, 76], these σ band FSs are not strictly cylindrical, but form as warped tubes with a narrowing in all directions towards Γ (sketched more accurately in **Figure 2**).

Since the FS tubes represent hole carrier sections, their interior will be empty in the ground state, while their exterior will be occupied. In a reduced zone schema, this construct creates ambiguous electron/hole character for the inter-tubular region. Ambiguity arises because this inter-tubular region should be, in the ground state, empty (*i.e.* without electrons) relative to the outer heavy effective mass σ band, but, in the ground state, filled with electrons relative to the inner light effective mass σ band.

This notion creates an apparent dilemma, although according to Ziman [55], "*There can be points in the zone where one cannot assign the label 'hole' or 'electron' uniquely to the states*". Further, *"the excitations of the superconducting state are peculiar quasi-particles which change from being 'electrons' to being 'holes' as they pass through the Fermi level"* [53]*.* Alternatively, we may reconcile this dilemma by considering that the reduced BZ scheme merges two different diameter tubes from points in reciprocal space within an extended BZ scheme [27, 28].

Given the indeterminate nature of the origin in reciprocal space, specific diameter tubes may be selected interchangeably by the DFT calculation; thus, implying a potential resonating behavior [28]. Analysis of electron–phonon behavior determined by DFT calculations suggests that this inter-tubular region of FSs (or other regions enclosed by parallel surfaces of different topology) is a region in reciprocal space that reveals the extent of superconductivity in typical BCS-type materials [5, 23, 24, 26–28]. Our calculations for both MgB2 and B-doped diamond show that this inter-tubular region is of meV energy scale from the Fermi energy.

Parallel FSs are common features of superconducting compounds albeit their identification is dependent on crystal symmetry and the choice of k-grid value for DFT calculations [5, 23, 25]. The "resolution" of reciprocal space calculations using DFT (*i.e.* the value of k-grid) is a critical factor for identification of phenomena that may be influenced by changes of a few meV. For example, we show above that the value of EF for MgB2 may change by several hundred meV with a difference of ~0.02 Å−1 for Δk (noting that Tc ~ 40 K). Such changes in EF may shift the apparent Fermi level to a position where parallel FSs are not shown in an EBS. For compounds with a higher EF value, closer to the parabola vertex (likely associated with lower Tc) and with larger difference in effective masses (*i.e.* the light mass displays a steeper EBS variation with k), the impact of this sensitivity to Δk increases.

### **Figure 12.**

*Schematic of the FSs for MgB2 viewed along: (a) the c-axis and (b) perpendicular to the c-axis. In this schematic, these FSs are simplified by neglecting warping in the DFT calculated model for MgB2 [28]. Hatched section represents the inter-tubular region.*

**133**

*Insights from Systematic DFT Calculations on Superconductors*

superconductivity may emerge and valid "hits" may be missed.

Thus, the value of k-grid used for DFT calculations is paramount. For PDs, this computational requirement has previously been well documented [14, 17, 20, 24] and, we suggest, is equally requisite for the use of EBS to predict, or design, new

The recent development of an ML-based scheme to efficiently assimilate the function of the Kohn-Sham equation, and to directly and rapidly, predict the electronic structure of a material or a molecule, given its atomic configuration [41] is of salient interest with regard to k-grid value. This ML approach maps the atomic environment around a grid-point to the electron density and local density of states at that grid-point. The method clearly demonstrates more than two orders of magnitude improvement in computational time over conventional DFT calculations to generate accurate electronic structure details [41]. Utilization of this methodology at a k-point spacing <0.2 Å−1 to initialise ML-training for charge density [41] may enable very rapid determination of potential SC materials with many hundreds of atoms in the base structure. Nevertheless, as we have shown in this article, caution in the use of such values for Δk using ML is suggested because "false positives" for

Thermal effects on electronic properties are generally included in DFT calculations as a smearing of electron behavior. However, high structural symmetry, or the lack of it, may impose significant anisotropy and/or preferred directionality of ionic movement that remains active even as temperature is increased. For reference, thermal excitation of the free-electron gas is kBT or about 26 meV at ambient temperatures [51, 57]. As noted above, variations in EF for superconducting phases may be in the meV range depending on the structure. We also note the importance of the smearing parameter in DFT calculations. We suggest that for particular superconducting cases where the Tc and/or phonon energy is low (*i.e.* Tc < 10 K) default values (~ 0.1–0.2 eV) in software packages for the smearing parameter may

Calculated Fermi energies and Fermi levels are essential attributes for determination of materials properties in a range of other applications, such as for the energy band alignment of components in solar cell materials [78, 79], with solidelectrolyte interfaces [80], as well as for interface induced phenomena such as the substantial increase in Tc of monolayer FeSe on SrTiO3 substrates [81]. Improved interpretation and understanding of electronic behavior in SCs and SC systems can be achieved with reliable calculated output values determined by *ab initio* DFT [82]. Indeed, Kohn posits that to achieve high accuracy with comprehensible representations of multi-particle systems, it is necessary to focus on real, three-dimensional coordinate space, *via* electron density distributions calculable using DFT [29].

The EBS encapsulates a wealth of information for superconductivity that may be misinterpreted due to the quality, or resolution, of DFT computations. A tendency to be satisfied with poor or limited computational resolution is evident in superconductivity literature unlike other fields that compute electronic properties using DFT. Translation of reciprocal space detail to real space periodicity for DFT-based design of new materials in an EBS with appropriate k-grid resolution can provide evidence for structures that may be viable SCs. As we have shown above, the EF value is explicitly determined in DFT computations and, with consistent use of k-grid resolution, can provide comparable estimates of SC properties for proposed structures of a compositional suite. We encourage inclusion of these DFT calculated

*DOI: http://dx.doi.org/10.5772/intechopen.96960*

superconducting materials.

be misleading [77].

**5. Conclusions**

parameters in reports of SC materials.

*Real Perspective of Fourier Transforms and Current Developments in Superconductivity*

construction, belong to the FSs. The FS corresponding to σ bands in the reduced BZ become two approximately parallel tubes [28], as schematically represented in **Figure 12** below. As shown for MgB2 in earlier work [23, 24, 27, 28] and by others [65, 75, 76], these σ band FSs are not strictly cylindrical, but form as warped tubes with a narrowing in all directions towards Γ (sketched more accurately in **Figure 2**). Since the FS tubes represent hole carrier sections, their interior will be empty in the ground state, while their exterior will be occupied. In a reduced zone schema, this construct creates ambiguous electron/hole character for the inter-tubular region. Ambiguity arises because this inter-tubular region should be, in the ground state, empty (*i.e.* without electrons) relative to the outer heavy effective mass σ band, but, in the ground state, filled with electrons relative to the inner light effec-

This notion creates an apparent dilemma, although according to Ziman [55], "*There can be points in the zone where one cannot assign the label 'hole' or 'electron' uniquely to the states*". Further, *"the excitations of the superconducting state are peculiar quasi-particles which change from being 'electrons' to being 'holes' as they pass through the Fermi level"* [53]*.* Alternatively, we may reconcile this dilemma by considering that the reduced BZ scheme merges two different diameter tubes from points in

Given the indeterminate nature of the origin in reciprocal space, specific diameter tubes may be selected interchangeably by the DFT calculation; thus, implying a potential resonating behavior [28]. Analysis of electron–phonon behavior determined by DFT calculations suggests that this inter-tubular region of FSs (or other regions enclosed by parallel surfaces of different topology) is a region in reciprocal space that reveals the extent of superconductivity in typical BCS-type materials [5, 23, 24, 26–28]. Our calculations for both MgB2 and B-doped diamond show that

Parallel FSs are common features of superconducting compounds albeit their identification is dependent on crystal symmetry and the choice of k-grid value for DFT calculations [5, 23, 25]. The "resolution" of reciprocal space calculations using DFT (*i.e.* the value of k-grid) is a critical factor for identification of phenomena that may be influenced by changes of a few meV. For example, we show above that the value of EF for MgB2 may change by several hundred meV with a difference of ~0.02 Å−1 for Δk (noting that Tc ~ 40 K). Such changes in EF may shift the apparent Fermi level to a position where parallel FSs are not shown in an EBS. For compounds with a higher EF value, closer to the parabola vertex (likely associated with lower Tc) and with larger difference in effective masses (*i.e.* the light mass displays a steeper

this inter-tubular region is of meV energy scale from the Fermi energy.

EBS variation with k), the impact of this sensitivity to Δk increases.

*Schematic of the FSs for MgB2 viewed along: (a) the c-axis and (b) perpendicular to the c-axis. In this schematic, these FSs are simplified by neglecting warping in the DFT calculated model for MgB2 [28]. Hatched* 

reciprocal space within an extended BZ scheme [27, 28].

**132**

**Figure 12.**

*section represents the inter-tubular region.*

tive mass σ band.

Thus, the value of k-grid used for DFT calculations is paramount. For PDs, this computational requirement has previously been well documented [14, 17, 20, 24] and, we suggest, is equally requisite for the use of EBS to predict, or design, new superconducting materials.

The recent development of an ML-based scheme to efficiently assimilate the function of the Kohn-Sham equation, and to directly and rapidly, predict the electronic structure of a material or a molecule, given its atomic configuration [41] is of salient interest with regard to k-grid value. This ML approach maps the atomic environment around a grid-point to the electron density and local density of states at that grid-point. The method clearly demonstrates more than two orders of magnitude improvement in computational time over conventional DFT calculations to generate accurate electronic structure details [41]. Utilization of this methodology at a k-point spacing <0.2 Å−1 to initialise ML-training for charge density [41] may enable very rapid determination of potential SC materials with many hundreds of atoms in the base structure. Nevertheless, as we have shown in this article, caution in the use of such values for Δk using ML is suggested because "false positives" for superconductivity may emerge and valid "hits" may be missed.

Thermal effects on electronic properties are generally included in DFT calculations as a smearing of electron behavior. However, high structural symmetry, or the lack of it, may impose significant anisotropy and/or preferred directionality of ionic movement that remains active even as temperature is increased. For reference, thermal excitation of the free-electron gas is kBT or about 26 meV at ambient temperatures [51, 57]. As noted above, variations in EF for superconducting phases may be in the meV range depending on the structure. We also note the importance of the smearing parameter in DFT calculations. We suggest that for particular superconducting cases where the Tc and/or phonon energy is low (*i.e.* Tc < 10 K) default values (~ 0.1–0.2 eV) in software packages for the smearing parameter may be misleading [77].

Calculated Fermi energies and Fermi levels are essential attributes for determination of materials properties in a range of other applications, such as for the energy band alignment of components in solar cell materials [78, 79], with solidelectrolyte interfaces [80], as well as for interface induced phenomena such as the substantial increase in Tc of monolayer FeSe on SrTiO3 substrates [81]. Improved interpretation and understanding of electronic behavior in SCs and SC systems can be achieved with reliable calculated output values determined by *ab initio* DFT [82]. Indeed, Kohn posits that to achieve high accuracy with comprehensible representations of multi-particle systems, it is necessary to focus on real, three-dimensional coordinate space, *via* electron density distributions calculable using DFT [29].
