1. Introduction
The pairing mechanism of unconventional high-temperature superconductors (HTSCs) remains one of the greatest unsolved mysteries of physics. All unconventional superconductors, including cuprates [
1,
2] and iron-based HTSCs [
3,
4] but also heavy fermions [
5] and organic superconductors [
6], have in common that the superconducting phase occurs near a magnetic phase. Furthermore, their phase diagrams typically show at least one other form of electronic order, e.g., charge or orbital order [
7,
8], a pseudogap phase [
2], stripe order [
2] or nematic order [
9]. The proximity of the magnetic phases naturally suggests the involvement of magnetism [
10]. In most theoretical approaches, spin fluctuations play a leading role [
11,
12]. Alternative approaches consider, e.g., excitonic superconductivity [
13,
14], long-wavelength plasmonic charge fluctuations or orbital fluctuations [
15,
16,
17].
It is generally assumed that the Cooper pairing in these superconductors cannot be described within a standard phonon-mediated scenario. However, this assumption is based only on the consideration of electron–phonon coupling on the Fermi surface only. The
Tc calculation based on the McMillan
Tc formula typically uses an approximation valid for classical low-
Tc superconductors, where the superconducting electron concentration is only considered at the Fermi level. This approximation is no longer valid for high-temperature superconductors such as the iron-based superconductors, since high-energy phonons are excited at elevated temperatures, so that electron–phonon scattering influences the electron over a larger energy range around the Fermi energy. In the high-temperature limit, this energy range may be comparable to Debye energy. Experimental ARPES data actually show that in iron-based superconductors, electrons down to ~0.03–0.3 eV below the Fermi energy are influenced by the onset of superconductivity [
18,
19,
20]. In order to perform a comprehensive study of whether the electron–phonon coupling is related to the formation of Cooper pairs in iron-based superconductors or not, we decided to consider the true superconducting electron concentration in order to recalculate the electron–phonon coupling constant under an antiferromagnetic background. Several studies offered an alternative scenario for iron-based superconductors, suggesting that the role of electron–phonon coupling had previously been underestimated against antiferromagnetic (AF) backgrounds [
21,
22,
23]. An explicit Density Functional Theory (DFT) calculation by B. Li et al. [
22] showed that the phonon softening of AFeAs (A: Li or Na) under an AF background allows an increase in the electron–phonon coupling by a factor of ~2. While any orthogonal change in the phonon vector can be considered a phonon-softening phenomenon, the lattice dynamics studied by S. Deng et al. [
23] confirmed that out-of-plane lattice vibration amplifies electron–phonon scattering based on their first-principle linear response calculation. While the tetrahedral atom is better suited to attract electrons in terms of electronegativity, the vertical displacement of the lattice Fe transfers the charge of the electron to the tetrahedral regions to generate an additional
xy potential [
21]. S. Coh et al. [
21] calibrated the GGA + A functional, which made it possible to bring the simulation results much closer to the experiments [
21,
24]. The calibrated ab initio method explicitly demonstrates the occurrence of the induced
xy potential from the out-of-plane lattice dynamics in the AF background that increase the electron–phonon scattering matrix by this factor of ~2 (abbreviated as ratio
Rph). More importantly, they provide an analytical model [
21] to explain why the electron–phonon scattering computed by the ab initio method is always increased by a ratio of ~2 under the effect of the spin density wave (abbreviated as ratio R
SDW).
The pairing strength of iron-based superconductivity can be enhanced significantly with the help of nanostructuring [
22,
25,
26,
27]. The layer structure of FeSe makes it possible to grow monolayers of FeSe epitaxially on a substrate. In 2013, superconductivity was reported with a record
Tc of 70 K on monolayer FeSe on a SrTiO
3 substrate [
25], which was later increased to 100 K [
26]. Despite the complexity of the electronic phase diagram of iron-based superconductors, which suggests the presence of additional broken symmetries besides the broken U(1) gauge symmetry of the superconducting state and thus an unconventional pairing mechanism, recent works have suggested that electron–phonon coupling could play a certain role in the superconducting mechanism of iron-based superconductors [
22,
27,
28], although there is clear evidence that magnetic fluctuations must be taken into account. The high transition temperature of the monolayer FeSe on a SrTiO
3 substrate gives further indications of the importance of electron–phonon coupling. While growing FeSe films on graphene substrate suppresses
Tc [
29], the giant enhancement of
Tc is likely activated by the SrTiO
3 substrate, where the interfacial contribution cannot be ignored. Strong electron–phonon coupling at the interface of FeSe/SrTiO
3 has been identified in ARPES data [
19], with electrons located 0.1–0.3 eV below the Fermi level involved in superconductivity. Although the FeSe phonons do not depend on the thickness of the FeSe material, unusual phonons [
30,
31], such as the F-K phonon, across the interface may be responsible for the high
Tc [
31]. According to the experiment by S. Zhang et al. [
31], the F-K phonons of the FeSe/SrTiO
3 surface show new energy loss modes, and the line width is widened compared with bare SrTiO
3.
In this article, we revise the superconducting electron concentration and use an ab initio approach to examine if the Tc values of LiFeAs, NaFeAs and FeSe as a function of pressure can be calculated reasonably by taking into account the Rph and RSDW factors, etc. If successful, we use this model to test whether such an approach can be applied to the ~100 K superconductivity in the nanostructured FeSe/SrTiO3. Not all mechanisms of iron-based superconductivity have been encountered in this work, because the unified theory of iron-based superconductors remains an open question. We only apply mathematical techniques to convert the two models from the literature into Tc values, which may be important to find out the possible mechanism of iron-based superconductors.
2. Computational Methods
As a starting point, the electronic properties of all compounds investigated in this article are computed by the spin-unrestricted Generalized Gradient Approximation of the Perdew–Burke and Ernzerhof (GGA-PBE) functional (unless otherwise specified) [
31,
32,
33,
34,
35] in Wien2K. The SCF tolerance is 1 × 10
−5 eV, and the interval of the k-space is 0.025(1/Å). The maximum SCF cycle is 1000. The magnetism and phonon data are calculated by CASTEP. Finite displacement mode is chosen where the supercell defined by cutoff radius is 5 Å and the interval of the dispersion is 0.04(1/Å). Ultrasoft pseudopotential is assigned, and density mixing is chosen to be the electronic minimizer [
31,
32,
33,
34,
35]. The experimental lattice parameters are used [
36,
37]. In this article, only Fe and As atoms are imported for the 111-type compounds.
Instead of calibrating “A” in the GGA+A functional, which entails an enormous computational cost and time-consuming experimental effort [
21,
38,
39], we propose a two-channel model to more easily model the induced
xy potential, where the upper tetrahedral plane is called channel 1 and the lower tetrahedral plane is called channel 2, respectively. We apply the superposition principle to separately calculate the induced
xy potentials induced by channels 1 and 2. Our two-channel model has fulfilled an assumption that the probability of finding an Fe atom moving in the +
z and −
z directions is equal, but their vibrational amplitudes never cancel each other out. This assumption is justified by Coh et al., whose explicit calculation confirms that the iron-based system consists of an out-of-phase vertical displacement of iron atoms, with the first adjacent iron atoms moving in opposite directions [
21]. We define
. In the ARPES range,
represents the average electronic density of states for the structure that exclusively contains upper tetrahedral planes. Similarly,
indicates the average electronic density of states within the ARPES range for the structure that only contains lower tetrahedral planes. Meanwhile,
corresponds to the average electronic density of states within the ARPES range for the original structure that has coexisting upper and lower tetrahedral planes.
is the phonon density of states as a function of frequency
, and the integral
is taken over by the Fermi surface with the Fermi velocity
. The Eliashberg function is written as [
40]
The electron–phonon matrix elements are given by
, where
is abbreviated as
,
is the wavefunction of electron,
is the Planck constant divided by 2π and C is the material constant related to lattice [
40].
and
represent the displacement of the ion relative to its equilibrium position and the ionic potential.
is the electronic probability density in the nonmagnetic state. The resultant ionic interaction
on the
XY plane, due to the abnormal phonon, is calculated by multiplying the ionic potential by
, i.e.,
. Moreover, the antiferromagnetic interaction along the
XY plane modifies the electronic wavefunction
, and the probability density fulfills
. The spin density wave factor
can be considered as the amplification factor for electron–phonon scattering under an antiferromagnetic SDW state, relative to a nonmagnetic state [
21]. Rearranging the mathematical terms yields the electron–phonon matrix element as
To derive a superconducting transition temperature from the simulation parameters, we use the McMillan
Tc formula [
40]. Due to the high transition temperatures, the electron–phonon scattering matrix takes into account the full electronic DOS in a range from
to
and not only the value at Fermi level (i.e., increasing the effective electronic DOS). Here, we consider that
represents the upper limit of the phonon energies that can be transferred to electrons, and at the high transition temperatures of Fe-based superconductors, contributions from high-energy phonons become important in the electron–phonon scattering mechanism, as opposed to classical low-
Tc superconductors. Although this approach is a simple consequence of the conservation of energy, it is supported by experiments: a shift of the spectral weight between the normal and the superconducting state is clearly visible in the photoemission spectra below the superconducting energy gap of various iron-based compounds in an energy range of ~30–60 meV below the Fermi energy [
18,
19,
20]. This energy range is approximately on the order of Debye energy.
In Bardeen–Cooper–Schrieffer (BCS) superconductors, the electrons on the Fermi surface condense into the Bose–Einstein superconducting state, where the total number of electrons on the Fermi surface equals the total number of electrons on the superconducting state. Hence, the theoretical Tc of BCS superconductors remains the same if we substitute either the electronic DOS on the Fermi level or the electronic DOS of the condensed Bose–Einstein state. However, the situation is different in iron-based superconductivity, where the electrons located between and transfer energy to the electrons in the Bose–Einstein superconducting states. When this happens, we have to revise the resultant electron–phonon scattering matrix in the condensed Bose–Einstein state. The Bose–Einstein statistic favors more electrons occupying the superconducting state. The electrons within the ARPES range increases the effective electronic DOS in the condensed Bose–Einstein state indirectly. The electrons within the ARPES range cannot be excited to the Fermi surface due to electrostatic repulsion. However, these electrons have another route to follow the Bose–Einstein distribution, which can be argued as a reason why these electrons disappear below the Fermi level.
The computation of band structure produces discrete (E,k) points, where E and k are the energy and the wavevector of the electron, respectively. The ratio of the electron–phonon scattering matrix is , which is abbreviated as the ARPES factor. is 1 if . Similarly, if . Otherwise, . gives the total number of (E, k) points in the range . or is the percentage of electrons contributed to the term. To make a fair comparison, the intervals of k space in the numerator and denominator of are essentially the same. The term controls the proportion of electrons scattered below the Fermi level.
Due to the fact that the superconducting transition temperatures are low, we calculate the mean occupation number
in the Fermi–Dirac statistic at low temperatures (T < 100 K), where
and
are 0.5 and ~0.5005, respectively. If
,
and
, the tiny offset in the mean occupation number may allow the Eliashberg function to approximately obey the following form.
where
, and the velocity
is converted from the Debye energy.
is the sum of the surface integral
at different electron energies within the ARPES range. The form of the antiferromagnetically amplified electron–phonon coupling is expressed as
, where
. The
is the average square of the electron–phonon scattering matrix on the Fermi surface [
40]. In the case of strong coupling, the renormalized electron–phonon coupling is expressed as
[
41].
When all the terms in the pairing strength at any pressure are entirely calculated by the spin-unrestricted GGA-PBE functional [
33], this approach is defined as a “traditional combination of DFT functional”, in which
may be neglected, as the effect of SDW should be included in
,
,
and
automatically in the spin-unrestricted mode. On the other hand, we propose an “empirical combination of DFT functional” which imposes the antiferromagnetic effect on the pairing strength separately. In this case, the antiferromagnetically amplified pairing strength is separately calculated by multiplying the nonmagnetic pairing strength with the antiferromagnetic factors.
and
are computed by spin-restricted mode, but
always needs an operation of the spin-unrestricted mode in order to add the effect of SDW. As the two-channel model has already mimicked the contributions of the abnormal phonon under antiferromagnetism manually, it is recommended to apply the spin-restricted mode to calculate
. Otherwise, the effect of antiferromagnetism on the abnormal phonon may be overestimated.
For the “empirical combination of DFT functional”, the pairing strength is further corrected by the AF Ising Hamiltonian in the presence of pressure. To include the magnetic effect, this AF Ising Hamiltonian is acquired by the spin-unrestricted GGA-PW91 functional. The pairing strength formulas of LiFeAs (111-type), NaFeAs (111-type) and FeSe (11-type) under pressure are given as
, where
. The ratio
monitors the pressure dependence of the AF energy at each external pressure
P, and
is the exchange–correlation energy. We use
to correct the antiferromagnetism under pressure instead of recalculating the
. The Debye temperature of the FeSe/SrTiO
3 is replaced by the vibrational energy of the F-K phonon across the interface [
31]. The pairing strength is substituted into the McMillian
Tc formula [
27], which includes the enhanced electron–phonon scattering matrix elements:
.
3. Results
The atomic spring constants between the FeFe bond
kFeFe and FeSe bond
kFeSe in the iron-based superconductors are compared. Our DFT calculation shows that
kFeSe/
kFeFe is ~0.25, while the
kFeAs is almost 2 times stronger than
kFeSe. As the atomic spring constants of the tetrahedral bonds are comparable to the FeFe bond, the orthogonal phonon appearing is feasible. Our two-channel model demonstrates that the induced
xy potential is good enough to be emerged at the “GGA-PBE” level. We calculated that the electron–phonon scattering matrix of FeSe under the induced
xy potential amplified by
Rph = 2.8. While the accuracy of our two-channel model is comparable to the
Rph = 2.2 obtained from the calibrated GGA + A functional [
21], we determine the
Rph of NaFeAs and LiFeAs to be 1.97 and 1.8, respectively. The pressure dependence on
Rph is less than ~5% due to
c >>
a.
A critical parameter in any ab initio approach is the value of the renormalized Coulomb pseudopotential.
Figure 1 estimates the error of the theoretical
Tc by tuning
μ*. Despite that the calculation of
μ* as a function of Debye temperature and Fermi level [
41] may not be very accurate in such a strongly correlated electron system [
42], it has been argued that for the most Fe-based superconductors,
μ* should be 0.15–0.2 [
12]. In this paper, we choose the value (
µ* = 0.15) of the Coulomb pseudopotential to calculate the
Tc of LiFeAs, NaFeAs and FeSe to make a fair comparison. Our calculated
µ* value of the uncompressed NaFeAs is 0.13. The error of our
Tc calculation due to the uncertainty of
µ* between
µ* = 0.15 and
µ* = 0.13 is within ~15%.
Figure 2a shows that our approach can generate the theoretical
Tc values in an appropriate range. The ARPES data confirm that LiFeAs and FeSe require the use of the
Rg term, while the NaFeAs does not [
18,
20,
43]. The theoretical
Tc of NaFeAs at 0 GPa and 2 GPa are 11 K and 12.5 K, respectively [
44]. The antiferromagnetically enhanced electron–phonon interaction on the Fermi surface and the AF exchange Hamiltonian compete in the compressed NaFeAs, as illustrated in
Figure 2b. We observe that the antiferromagnetism is slightly weaker at finite pressure, but the antiferromagnetically assisted electron–phonon coupling on the Fermi layer is increased almost linearly at low pressure. We show the steps to estimate the
Tc of NaFeAs at 0 GPa as an example. After activating the spin-unrestricted mode, the
is 1.625. The antiferromagnetically assisted electron–phonon coupling on the Fermi surface is
Figure 2.
(
a) The theoretical and experimental [
44]
Tc values of NaFeAs. (
b) The antiferromagnetically assisted electron–phonon coupling on the Fermi surface and the AF energy as a function of pressure. The DFT parameter can be found in
Table 1.
Figure 2.
(
a) The theoretical and experimental [
44]
Tc values of NaFeAs. (
b) The antiferromagnetically assisted electron–phonon coupling on the Fermi surface and the AF energy as a function of pressure. The DFT parameter can be found in
Table 1.
According to the McMillian
Tc Formula, the
Tc becomes
We compare our theoretical
Tc by substituting the raw data of other groups [
15,
21]; their calculated
is 0.39 [
15], and the induced
xy potential by the out-of-plane phonon reinforces the electron–phonon coupling matrix by 2.2 [
21].
After renormalization, these two couplings are softened to , and the renormalized Coulomb pseudopotential .
Based on the data in other groups [
15,
21], the theoretical
Tc becomes
Our calculated value of the electron–phonon coupling on the Fermi surface of the uncompressed LiFeAs is ~0.1 [
45], but the magnetic amplification factors increase the pairing strength to 0.82, remarkably. The Debye temperature
of LiFeAs remains at ~385 K below 8 GPa [
46], as shown in
Table 2. A reduction in the theoretical
Tc is also observed in the compressed LiFeAs, and the weakening effect of
and
under pressure is identified, as shown in
Figure 3b. In compressed FeSe [
24], however, a gain in
is observed that triggers the increase in
Tc under pressure (
Figure 4). It should be noted that our approach is a mean field approach, and we treat the spin fluctuations as being proportional to the mean field Hamiltonian. The vanishing of the macroscopic AF order observed in real samples is due to the strong fluctuation effects in these layered compounds. The magnetism considered here in the nonmagnetic regimes of the phase diagrams is of a fluctuating microscopic nature. The optimized pairing strength of LiFeAs and FeSe is achieved at a pressure of 0 GPa and 0.7 GPa, respectively. The differences between
DOS(EF–EDebye) and
DOS(
EF) in LiFeAs and FeSe are less than 4%. The
Rg term in LiFeAs is reduced with pressure, but the
Rg term of FeSe is optimized at medium pressure (see
Table 2 and
Table 3).
Figure 3.
(
a) The theoretical and experimental [
47]
Tc values of LiFeAs are consistent. (
b) The antiferromagnetically assisted electron–phonon coupling and the AF exchange Hamilton under pressure.
equals 1.75 at 0 GPa.
Figure 3.
(
a) The theoretical and experimental [
47]
Tc values of LiFeAs are consistent. (
b) The antiferromagnetically assisted electron–phonon coupling and the AF exchange Hamilton under pressure.
equals 1.75 at 0 GPa.
Figure 4.
(
a) Both theoretical and experimental [
24]
Tc values increase with pressure. (
b) The pressure dependence of the antiferromagnetically assisted electron–phonon coupling and the AF interaction.
at 0 GPa is 1.59.
Figure 4.
(
a) Both theoretical and experimental [
24]
Tc values increase with pressure. (
b) The pressure dependence of the antiferromagnetically assisted electron–phonon coupling and the AF interaction.
at 0 GPa is 1.59.
Table 1.
The DFT parameter of NaFeAs [
46]. The
Rg term is computed by the “empirical combination of DFT functional”.
Table 1.
The DFT parameter of NaFeAs [
46]. The
Rg term is computed by the “empirical combination of DFT functional”.
P/GPa | a (Å) | c (Å) | FeAs Length (Å) | Rg | TDebye (K) |
---|
0 | 3.929 | 6.890 | 2.400 | 1.00 | 385.0 |
1 | 3.914 | 6.833 | 2.388 | 1.00 | 385.5 |
2.0 | 3.900 | 6.777 | 2.376 | 1.00 | 386.0 |
Table 2.
The DFT parameter of LiFeAs [
46]. The
Rg term is compiled by the “empirical combination of DFT functional”.
Table 2.
The DFT parameter of LiFeAs [
46]. The
Rg term is compiled by the “empirical combination of DFT functional”.
P/GPa | a (Å) | c (Å) | FeAs Length (Å) | Rg | TDebye (K) |
---|
0 | 3.769 | 6.306 | 2.44 | 2.66 | 385.00 |
2.4 | 3.745 | 6.134 | 2.42 | 2.38 | 385.25 |
4.5 | 3.723 | 5.985 | 2.35 | 1.67 | 385.5 |
6.3 | 3.702 | 5.918 | 2.33 | 1.56 | 385.75 |
Table 3.
The DFT parameter of FeSe [
48,
49]. The
Rg term is simulated by the “empirical combination of DFT functional”.
Table 3.
The DFT parameter of FeSe [
48,
49]. The
Rg term is simulated by the “empirical combination of DFT functional”.
P/GPa | a (Å) | c (Å) | FeSe Length (Å) | Rg | TDebye (K) |
---|
0 | 3.767 | 5.485 | 2.390 | 3.04 | 240 |
0.7 | 3.746 | 5.269 | 2.388 | 2.05 | 256 |
2.0 | 3.715 | 5.171 | 2.384 | 4.92 | 274 |
3.1 | 3.698 | 5.114 | 2.382 | 2.50 | 290 |
Based on the successful
Tc calculation of the bulk FeSe, LiFeAs and NaFeAs, we start our journey to acquire the theoretical
Tc of the FeSe monolayer on a SrTiO
3 substrate step by step using the model of an antiferromagnetically enhanced electron–phonon coupling. The flowchart is shown in
Figure 5. After the geometric relaxation of FeSe/SrTiO
3, the angles of the unit cell are 89.81°, 90.88° and 89.05°, with a tiny internal shear force being captured. The relaxed tetrahedral angle of Fe–Se–Fe is 108 degrees. The antiferromagnetic energy of FeSe can be amplified by low dimensionality when it is deposited in the form of a monolayer on SrTiO
3 [
26]. Compared with a FeSe monolayer without substrate, the FeSe film on SrTiO
3 shows an increased exchange correlation energy of ~16% on FeSe. Apart from this, the local Fe moment in the isolated FeSe monolayer is only ~0.5 µ
B. However, contact with SrTiO
3 amplifies the local Fe moment of the FeSe film up to ~1.3 µ
B. Our calculated electron–phonon coupling on the Fermi surface without any amplification factor is
. Based on our simulation, the antiferromagnetism of FeSe/SrTiO
3 is still as strong as of the FeSe monolayer without substrate. Hence, the simultaneous occurrence of antiferromagnetism and tetrahedral atoms makes the Coh factor unavoidable. The analytical result of C
AF = 2 is used [
21], and our calculated C
Ph in FeSe/SrTiO
3 is 2.9. After amplification of the Coh factor, the theoretical
Tc is only 14 K. However, a massive enhancement of the pairing strength can be observed when the interfacial F-K phonon is involved [
31]. The F-K phonon actuated via the interface contributes the vibrational energy of ~100 meV (~1159 K) [
31]. With this enormous Debye temperature, the theoretical
Tc is increased to 69 K, although the electron–phonon interaction is limited to the Fermi energy. In the ARPES data, it is evident that a shift in spectral weight occurs in the superconducting state 0.1~0.3 eV below the Fermi level [
19], which means that electrons in this energy range are affected by electron–phonon scattering as a result of the high phonon frequencies. This means that electrons in this energy range contribute to superconductivity, since the high phonon frequencies can scatter them up to the Fermi energy and need to be considered in the McMillan formula, and not only those at the Fermi energy, as in the usual approximation applied to classical low-
Tc superconductors. The superconducting electron concentration is thus corrected, and the average electron–phonon scattering matrix in these multienergy layers is 1.96 times higher than the matrix considering only the Fermi level. This is the last factor with which our theoretical
Tc can reach 91 K, which corresponds quite well to the experimental
Tc of 100 K.
The pairing strength is renormalized as
The pseudopotential is diluted as
We substitute all parameters into the McMillian
Tc formula: