Superconducting Phases in Lithium Decorated Graphene LiC6

A study of possible superconducting phases of graphene has been constructed in detail. A realistic tight binding model, fit to ab initio calculations, accounts for the Li-decoration of graphene with broken lattice symmetry, and includes s and d symmetry Bloch character that influences the gap symmetries that can arise. The resulting seven hybridized Li-C orbitals that support nine possible bond pairing amplitudes. The gap equation is solved for all possible gap symmetries. One band is weakly dispersive near the Fermi energy along Γ → M where its Bloch wave function has linear combination of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${d}_{{x}^{2}-{y}^{2}}$$\end{document}dx2−y2 and dxy character, and is responsible for \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${d}_{{x}^{2}-{y}^{2}}$$\end{document}dx2−y2 and dxy pairing with lowest pairing energy in our model. These symmetries almost preserve properties from a two band model of pristine graphene. Another part of this band, along K → Γ, is nearly degenerate with upper s band that favors extended s wave pairing which is not found in two band model. Upon electron doping to a critical chemical potential μ1 = 0.22 eV the pairing potential decreases, then increases until a second critical value μ2 = 1.3 eV at which a phase transition to a distorted s-wave occurs. The distortion of d- or s-wave phases are a consequence of decoration which is not appear in two band pristine model. In the pristine graphene these phases convert to usual d-wave or extended s-wave pairing.

Two dimensional superconducting phases have become of great interest since the discovery of the high temperature superconducting (HTS) cuprates and subsequent finding of Fe-pnictide and -chalcogenide HTSs. Interest was re-invigorated by the discovery of superconductivity onsets up to 75 K in single layer FeSe grown on SrTiO 3 and related substrates 1,2 . With the enormous research activity focused on graphene in recent years, it is not surprising that graphene-based superconductivity has become an active area of research. Very recently superconductivity up to 1.7 K has been reported 3 in magic angle bilayer graphene, which will buttress activity on two dimension superconductors and especially the related type that we discuss here.
Superconductivity has been known for some time in intercalated graphite compounds such as C 6 Ca and C 6 Yb 4 . With the many remarkable properties of graphene, it has been anticipated that doping by gating or by decorating with electro-positive elements, thereby moving the chemical potential away from the Dirac points, might induce superconductivity. However, graphene decorated with alkali metals has three valence bands with one weakly dispersive band near Fermi energy. Due to this flat band, there are additional available states around the Fermi level and the required pairing potential is reduced.
Discussion of superconductivity in doped graphene has been primarily within theoretical models, as we review below, but some encouraging data have been reported. Experimental evidence for a superconducting gap in Li-decorated monolayer graphene around 6 K has been reported by Ludbrook et al. based on angle-resolved photoemission spectroscopy 5 (ARPES). Scanning tunneling spectroscopy (STS) was applied by Palinkas et al. 6 to graphene suspended on tin nanoparticles, who concluded that superconductivity is induced in the graphene layer. Evidence of superconductivity in Li-decorated few layer graphene at 7.4 K has been reported by Tiwari and collaborators 7 . Low temperature mobility of K and Li atoms on graphene was observed by Woo et al., and suggest that mobility may persist at lower temperatures 8 , which would provide new challenges for theory.
Various mechanisms of pairing have been proposed. Uchoa and Castro-Neto modeled pristine and doped graphene with electron-phonon coupling or plasmon mediated in mind 9 . Repulsive electron-electron interactions were modeled by Nandkishore and collaborators 10,11 . Beginning from pristine graphene, varying the chemical potential leads to dominant chiral singlet + − d id x y xy 2 2 pairing for nearest neighbor interaction, according to Black-Schaffer et al. 12 a triplet f-wave state has been proposed to arise from next-nearest neighbor interaction with 1 Physics Department, Faculty of Science Razi University, Kermanshah, Iran. 2

Model Hamiltonian
Because the unit cell contains several atoms with important specific aspects, we provide many of the details of the expressions that can be obtained analytically. LiC 6 , as illustrated in Fig. 1, consists of a graphene layer decorated by a lithium layer in which Li atoms are located at the center of a carbon hexagon surrounded by six empty center hexagons. The height of Li above the carbon layer is calculated to be h z = 1.85, somewhat smaller than the value 1.93 Å obtained by Profeta et al. 15 . The nearest Li-C distances are h = 2.40. Since the Li 2s orbital energy is higher than the C 2p z orbital, charge transfer occurs. It is calculated that 0.685e from Li transfers to the six C atoms equally 19 . The positive Li ion and negative C ion provide a relative Coulomb (Madelung) shift in site potentials of the two atoms.
The attractive interaction between Li and C ions after charge transfer contracts the Li-C distance and reduces the C-C bond lengths in the Li-centered hexagon to a 1 = 1.425, while the bond length of nearest neighbor C atoms in different hexagons is slightly larger at a 2 = 1.426. For Ca instead of Li, this difference should be larger, hence we keep these lengths distinct. The hopping integral between short-bond carbons is t 1 , with that between stretched carbon sites is denoted ′ t 1 . We refer to this broken symmetry situation as "shrunk graphene". The difference in hopping amplitudes indicates that the new Li-C hopping parameter is the central new feature in LiC 6 compared to graphene. Symmetry breakdown leads to the opening of a small energy gap at the Γ point.
The lattice then becomes a two dimensional hexagonal Bravais lattice with seven atomic sites. These will be labeled as A 1 , A 2 , A 3 , B 1 , B 2 , B 3 and Li, as illustrated in Fig. 1. The Hamiltonian of this system is  Here H N and H P denote the non-interacting and interaction Hamiltonians respectively. In these expressions α and β run over A i , B i and Li. Here ασ † c i , ασ c i are creation and annihilation operators of an electron with spin σ on subsite α of ith lattice site, and = ασ ασ ασˆ † n c c i i i is the electron number operator. The noninteracting chemical potential is μ 0 and t iα,jβ is the hopping integral from the α site of ith cell to the β site of jth cell. We denote the on-site energy by ε α .
The interaction stated above corresponds to an extended (negative U) Hubbard model, which allows a variety of phenomenological values to be chosen and studied. It is largely for this reason that we provide substantial detail of the underlying, non-interacting C-Li lattice and electronic structure. The interactions that we study are introduced in Sec. IV.

Normal State of LiC 6
Many studies of graphene rely on tight binding parametrization of the band structure. The early parametrization of Wallace 20 already employed both first and second neighbors. Extensions in various ways have followed 21,22 , culminating in the application of Wannier functions by Jung and MacDonald 23 to provide simple but realistic five parameter model and a more accurate but more involved 15 parameter model. Our aim in this section is to construct a realistic seven band model for distorted LiC 6 , while also developing the formalism to allow exploration of superconducting phases once the interaction has been included.
The distortion of the graphene layer to shrunken graphene and the coupling to Li requires a considerable generalization of the underlying tight binding model Hamiltonian, and many of the details are relegated to appendices. The Hamiltonian of non-interacting LiC 6 is functions are defined in Supplementary Materials Eqs A.7, A.8, A.9 and A.10 respectively. For general → k vectors, it is challenging to obtain an exact analytical expression for the full Hamiltonian in Eq. 3 and it would not be transparent anyway. However, analytical expression for Eq. 3 can be achieved in two steps. Since hopping from Li atoms to nearest neighbor carbon sites t LiC 1 is small with respect to C-C nearest neighbor hopping t 1 , by first neglecting the lithium-carbon hopping → t 0 LiC 1 , first column and row of the Hamiltonian matrix in Eq. 3 are omitted, the remaining part given by Eq. B.1 is uncoupled shrunken graphene Hamiltonian which can be diagonalized exactly to obtain E sh,n . Finally, Li-C coupling is taken into account by perturbation theory to obtain eigenvalues E n , as presented in the appendices.
Uncoupled C 6 Dispersion Relations. By first neglecting the lithium-carbon hopping, , the uncoupled shrunken graphene Hamiltonian given by Eq. B.1 can be diagonalized exactly. Even though Li-C hopping has been neglected but still remaining part of shrunken Hamiltonian in the most general case, include broken symmetries in the hopping integrals, bond lengths and on-site energies. The non trivial eigenvalues of uncoupled shrunken graphene Hamiltonian in general form are given by with details presented in Supplementary Materials Appendix B. However, the obtained equations are often complicated. To provide insight into the method, uncoupled shrunken graphene Hamiltonian can diagonalized in some particular cases. The Brillouin zone (BZ) of C 6 is one third of that of graphene, with the Dirac points folded back to the Γ point. In this mini-BZ, the two π bands of pristine graphene i.e. E ± = ±t 1 |η 0 | folds to six branches as illustrated in Fig. 2. These branches are solutions of Eq. B.1 in the limited case of pristine which in the nearest neighbor approximation they are given by, Exact analytical solutions for pristine graphene wherein next neighbor hopping integrals are taken into account are presented in Supplementary Materials Eqs B.7 and B.8. As shown in Fig. 2 one sees that → β ± E k ( ) is weakly dispersive near the van Hove singularity at the saddle points M at 3/8 or 5/8 filling (0.25 electron per carbon doping), this band plays a major role in the formation of superconductivity in graphene. Also, one can observe that the band structure is four-fold degenerate at the charge neutral Dirac points. Solution of the Schrödinger equation for pristine graphene in the mini-BZ has another advantage: the Bloch-wave symmetry character of each branch can be distinguished. The Bloch coefficients of the branch labeled by E γ are of s-wave character, as illustrated in Fig. 2 and demonstrated in more detail in Appendix B, Eqs B.4 and B.6. This becomes important when it is shown that different superconducting phases of graphene in a variety of doping regimes are due to electron pairing in each of these branches.
Decoration of graphene with metals reduces symmetries that lead to removal of bands degeneracy in some regions. While decoration causes expansion and contraction of bonds length in three inequivalent directions in the honeycomb lattice i.e. τ δ In shrunken graphene these probabilities are → k dependent and unequal in general. It will be seen that these small deviations influence the superconducting gap equation symmetries.

Coupled LiC 6 Dispersion Relations. Li-C hopping adds a perturbation term to the shrunken graphene
Hamiltonian. Obtaining exact dispersions from Eq. 3 is very challenging, so perturbation theory is applied to obtain approximate solutions, as presented in Appendix C. However, to get some insight into effects of the coupling, Eq. 3 can be solved exactly at the Γ point. At → k = 0 only the isolated Li band, E Li,0 (0) and the lowest valance band, E sh,6 (0), are mutually affected. The energies of these bands are, and other shrunk graphene bands given by (Supplementary) Eq. B.5 remain unchanged. Comparing the fit results from DFT to these equations suggests that − t Li C 1 is in the 0.3-0.5 eV range, and other next neighbor hopping from Li atoms to C sites are negligible.
There are two critical points in the pure graphene band structure which are affected by decoration and become important: the charge neutrality Dirac points folded at the Γ point, and the van Hove singularity at the M point. We define a hopping integral symmetry breaking index, = ≠ ′ w 1 t t t 1 1 indicates the degree of symmetry breaking. The difference in Li and C on-site energies can be considered to reflect the amount of doping. The Dirac points affected by w t open a small gap E g at Γ, which does not depend on t LiC , the three folded branches of pristine π * band structure in the mini-Brillouin zone of graphene C 6 and for E γ is f s |s〉 + f f | f〉. Here we use abbreviated notation www.nature.com/scientificreports/ was fit to the DFT band structure, with results illustrated in Fig. 3. In the graphene layer shown in Fig. 1(a,c), A 1 subsite chosen as central site labeled by 0 and B 1 subsite in adjacent hexagon considered as second neighbor while just slightly longer than the first neighbors atoms B 2 and B 3 in same hexagon, this neighbor labeled by n = 2 and so on the next neighbors are labeled. In Fig. 1(a), the big dashed hexagon included up to nine neighbors but for the pristine graphene it is surrounded by five neighbors. C-C hopping from 0-subsite to nth neighbor has been shown by t n CC 0 . In-plane Li-Li hopping, t m LiLi 0 obtained up to m = 4 neighbors. Li to C hopping integrals are very small with respect to those of C-C and Li-Li, so we keep only the near neighbor Li-C hopping amplitude.
Since Li is small with respect to alkaline earths such as Ca, the pristine band structure is less affected by decoration by lithium than by calcium, as can be seen in Fig. 2 of ref. 15 . The fitted hopping amplitudes and on-site energies are presented in Table 1. Note that by comparing band structure of LiC 6 with pristine graphene in ref. 23 , it is observed that Li decoration only slightly changes the pristine graphene band structure. These changes are due to electron transfer from Li to graphene, which changes the pristine on site in which s = 1 is for particles and s = −1 for holes, and E m are the normal state eigenvalues.
3 are shown in Fig. 4(a), and where  E ( ) i j are Bloch wave coefficients of the j-th band. Possible order parameter symmetries in Eq. 8 are related to symmetries of Bloch wave functions, through Ω → k ( ) ij functions in Eq. 9. In the limiting case of (folded) six band pristine graphene, the symmetry character of different conduction bands along high symmetry lines were provided in Fig. 2. Bloch symmetry character of non-interacting bands specifies the symmetry of the band order parameter. Superconducting States. The linearized gap equation, obtained by minimizing the quasiparticle free energy with respect to nearest neighbor order parameters, is  3 3  3 3  3 3   3 3  3 3  3 3   3 3  3 3  3 3 1 1 ; the subscripts 〈ij〉 has been dropped for brevity. The A 3×3 , B 3×3 , C 3×3 , and D 3×3 matrices, given by Eq. 12, have identical structures, hence they share eigenvectors: , where the latter two are degenerate. Their eigenvalues, in obvious notation, are For folded six band pure graphene g 0 = g 1 , the Bloch wave coefficients appearing in Eq. 9 can be replaced from Eq. B.7 to show that Ω and similarly relations for other elements, hence C 3×3 = A 3×3 and D 3×3 = B 3×3 . Eq. 11 takes the more symmetric form  12 . These three solutions preserve symmetry of the two band unit cell as illustrated in Fig. 4(b,c). In addition to these three states, there are six more non-orthogonal solutions Φ 0n = (V sy 0 −V sy ) and Φ 1n = (V sy −V sy 0) that break symmetries of pristine graphene two band model. Inserting these solutions into Eq. 14 leads to a new two band gap equation , which is unphysical because of an unreachably high energy pairing potential g 0 . In the following section the superconducting gap equation has been solved for LiC 6 and it is demonstrated how Li-C coupling influences superconducting phases.
Nine Superconducting Phases. Self-consistent solutions of the gap equation Eq. 11 can be obtained analytically. There are three superconducting states with island character (discussed in more detail below) that can be expressed in compact form as where V sy refers to one of the V s , V d xy or − V d x y 2 2 -wave symmetries. Pairing in these phases cannot propagate, as may be pictured in Fig. 5. The other six superconducting states of Eq. 11 have the explicit form In these expressions ± J d s ( ) and α ± By comparing the gap equations introduced in Eqs 11 and 14 the gap equation symmetry reduction of decorated graphene with respect to folded but pristine graphene becomes clear. This symmetry reduction results in an α sy coefficient appearing in the pairing amplitudes of stretched bonds as shown in Eq. 16 and Fig. 4. We refer to these symmetry reduction phases as "distorted phases".
The six bands of pristine graphene support nine pairing amplitudes while in the two band model there are three possible pairing amplitudes along three different bonds. These two notions can be mapped onto each other only if α sy = 1 as illustrated in Fig. 4(b,c). Therefor the three island superconducting phases given by Eq. 15 in the special case of pristine cannot be mapped onto the two band model. These three phases are unphysical even in the case of decorated graphene because the Cooper pairs in these phases require a large pairing potential. In the special case of pristine graphene in which κ = 1, a sy = c cy and b sy = d cy from Eq. 17, and it follows that if b sy > 0 then α =  All states are orthogonal except those with same subscript, viz. Φ − s and Φ + s . Such solutions are orthogonal if κ = 1, i.e. g 1 = g 0 . Only for this case the matrix gap equation becomes Hermitian, then band order parameters takes the following form in terms of the band Green function and g 0 , annihilates an electron with spin σ in the ith band with energy Although it is assumed that g 1 = g 0 but deviation from pristine leads to distortion of Green's functions 〈 〉

Discussion and Relation To Previous Work
The possibility of a superconductivity state in metal decorated graphene has been suggested theoretically by a few groups 9,12,15 . Some have suggested phonon-mediated superconductivity in single layer graphene. Most prominently, Profeta et al. 15 calculated on the basis of density functional theory for superconductors that decoration by electron donating atoms such as Ca and Li will make single layer graphene superconducting, up to 8 K for the case of Li. The ab initio anisotropic Migdal-Eliashberg formalism was used by Zheng and Margine 24 , who predicted a single anisotropic superconducting gap with critical temperature T c = 5.1-7.6 K, in surprisingly good agreement with experimental reported superconductivity around 6 K in LiC 6 5 . Using a phenomenological microscopic Hamiltonian in a nearest-neighbor tight-binding approximation, possible superconducting phases of pristine graphene have been discussed by Uchoa and Castro-Neto 9 and also by Black-Schaffer and Doniach 12 . The possibility of a singlet p + ip phase pairing near the Dirac points between nearest neighbors subsites were suggested by Uchoa and Castro-Neto 9 . They worked in terms of a plasmon mediated mechanism for metal coated graphene, and discussed the conditions under which attractive electron-electron interaction can be mediated by plasmons.
Singlet superconducting gap phases of pristine graphene have been proposed and discussed by Black-Schaffer and Doniach 12 . For the nearest neighbors pairing amplitudes Δ =Δ where δ → j are the vectors that connects the iA site to its three nearest neighbors, it was observed that there are three states that minimize the free energy in various regimes of the parameters, which here have been denoted by V s = (1, 1, 1 preserves the graphene band symmetry. Depending on the position of the Fermi energy with respect to Dirac points, d + id or s states tend to dominate. Their numerical calculation showed that d-wave solutions will always be favored for electron or hole doping in the regime < < . c , it was suggested that chiral d + id superconductivity, which breaks time-reversal symmetry, can be stabilized. In this regime d-wave superconductivity may arise from repulsive electron-electron interaction 11 .
Although doping by a gate voltage is normally considered to change only the chemical potential but not the band structure, gating cannot be expected to push the Fermi energy to the van Hove singularity without altering the band dispersion. The most likely way to do this is by decoration with electropositive atoms, which has been our focus. We note that doping is essential, when graphene decorated, in addition to the expected charge migration from the decorating atoms to the graphene sheet, it is then necessary the interlayer state is partially occupied to induce superconductivity as happens in GICs. Hybridization of interlayer s-band and graphene π bands changes the graphene band structure. The s orbitals of Ca have more overlap with C orbitals than Li and lead to stronger and longer range interactions as well as increasing the doping level, effects that become detrimental to superconductivity. For this reason our emphasis here is on the Li decorated graphene.
We review some of our main points. When graphene is decorated by Li, electron transfer from Li atoms to C contracts the Li-C distance and reduces the C-C bond lengths in the Li-centered hexagon. In this kekulé -type structure, hopping amplitude symmetries of all C-C neighbors are broken (our "shrunken graphene"). This model allows study of multiband effects on the superconducting phase diagram. To gain insight into our model, solutions of superconducting gap equation in both cases of folded bands otherwise pristine C 6 and the usual two band model of C 2 were compared. These two viewpoints coincide if the same pairing paradigms are considered. For pristine graphene with its two site cell, in real space picture electrons can pair with near neighbors in three inequivalent directions, 3 which must respect honeycomb symmetries. The V sy quantities are the three vectors that belong to the irreducible representation of crystal point group D 6h i.e. V sy T = (1, 1, 1), (−1, 1, 0) and (2, −1, −1) for which the sy subscript stands for symmetries s, d xy and − d x y 3 . The gap equation is a 9 × 9 matrix equation given by Eq. 14. The folded bands supercell include three vertices numbered 5, 6, 7, and nine bonds as shown in Fig. 4(a). There are nine orthogonal solutions that preserve symmetries of this supercell. One of these configurations has s-wave symmetry (1, 1, 1, 1, 1, 1, 1, 1, 1) the other eight solutions are constructed by all possible permutations of (−1, 1, 0) along these bonds that preserve our supercell symmetry. There are only three solutions which can preserve symmetry of both two and six atoms cells simultaneously which they are of the form Φ = sy sy sy sy T as illustrated in Fig. 4. For these solutions, the folded 9 × 9 gap equation reduces to 3 × 3 gap equations of ordinary pristine graphene. The Cooper pair formation energy for these three modes are significantly less than the other six phases which are not reducible to the two band model.
In fact reduction of symmetry leads to increasing of the system free energy. After the orthogonalization procedure, one obtains three sy sy sy T 0 . These phases have been designated as island phases, as illustrated in Fig. 5(b) for Φ f , within which a pairing amplitude is localized within island hexagons and cannot propagate. For these island phases, numerical calculation of the electron pair potential energy g 0 shows that g 0 is large. This kind of solutions is a consequence of the six atom basis and does not appear for the two atom basis. Also, there are three solutions of the form sy sy sy sy which also break symmetry of two atom cell. For these reasons, in association with the normal state band structure of graphene, we concentrate on superconductivity in the three Φ + sy symmetry phases. For pristine graphene C 2 , two normal bands are E ± = ±t 1 |η 0 | which fold to six branches in mini-BZ of C 6 i.e. To understand how superconducting phases of graphene can be affected by decoration by Li, one can compare the LiC 6 gap solutions with those of folded bands C 6 at the same doping. Numerical results for pristine graphene gap equation performed in the nearest neighbor approximation in ref. 12 have been extended by applying a more accurate tight binding model fit to the DFT band structure of pristine graphene 23 . Although a quantum critical point for zero doping reported by Black-Schaffer and Doniach 12 at dimensionless coupling = .
1 91 g t 0 which dand s-wave solutions are degenerate. In the more realistic tight binding model we applied, this degeneracy is not observed at the Γ point, and the d-wave solution is dominant. This difference may be consequence of particle-hole symmetry breaking of valence and conduction bands. Also the van Hove singularity at the M point is moved from 0.25 doping for nearest neighbor hopping to 0.16 doping in the accurate model. The phase transition from d-wave to s-wave is shifted to 0.35 doping instead of the 0.4 doping reported for nearest neighbor hopping 12 . Numerical calculations for this more detailed model are illustrated in Fig. 7.
When graphene is decorated by Li, around 0.68 electron per lithium atom transfers to neighboring C sites, viz. = . n 0 11 c , and the Dirac points folded to Γ move to −1.52 eV. Symmetry breaking of the hopping partially removes degeneracies of band structure of pristine graphene, which leads to creation of the small gap at Γ, with energy = | − | = .
. Also two of four-fold degeneracies between valence and conduction bands at the Dirac points are removed. Compression between band structure of decorated graphene and folded pristine graphene at the same doping shows that hybridization of the Li s band and C π band is small. This means nearest neighbor Li-C hopping is in the range t LiC 1 ~ 0.3-0.5, and further hoppings are negligible.
Li decoration of graphene changes not only the band structure but also the Bloch wave coefficients from those of pristine graphene. While pristine graphene Bloch wave coefficients have pure s-or d-wave character and their magnitudes are → k -independent. In the case of LiC 6 they become mixed and vary with → k , hence gap equation symmetry is reduced. Because of this symmetry reduction, for the longer C-C bonds, a new coefficient α sy appears in the pairing amplitudes. In terms of this coefficient we have classified superconducting phase symmetries into three groups. Eqs 18,19, and 20 present all nine possible pairing phases of LiC 6 . There are three categories of solutions which have not appeared in complete form in the literature. The total of nine phases arise from spatial, and therefore hopping parameter, symmetry breaking.
In the first category Φ f , Φ p x and Φ p y , there is α sy = 0 identical to that of folded pristine C 6 . For the second category, α sy (denoted by α − ) is negative, in the case of pristine α − = −2 as discussed. These three phases break the two site cell symmetry, and numerical calculation shows that the pairing potential g 0 must be large to realize these phases. For the last category α + is positive. Three phases which correspond to α + > 0 include Φ + − d x y 2 2 , Φ + d xy , and Φ + s , and these have the lowest pairing potentials with respect to the other six phases.
In the limiting case of folded six band pristine graphene α + − d x y 2 2 , α + d xy , and α + s are all equal to unity, which maps the results to the two-band symmetries as it should. But when Li decorated, depending on doping strength viz. w t and t LiC 1 these coefficients α + sy no longer remain unity. The pairing amplitude distortion along longer C-C bonds α + , for s-wave phase is significant due to its spatial isotropic symmetry. In spite of the pristine nature this phase no longer preserves two band model symmetry. On the other hand, d-wave phases are hardly affected by doping and their superconductivity is more persistent against perturbation. The chirality or non-chirality of Cooper pairs in these phases is undetermined, however. As shown in Fig. 6 is approximately equal to unity and varies little with temperature. At a given critical temperature T c and chemical potential μ 0 , for each of nine possible superconducting phases, Eqs 10, 13 and 17 were evaluated numerically over the BZ of LiC 6 to find the corresponding pairing potential = g J 0 1 sy and α sy coefficient. Smaller g 0 means less Cooper pair formation energy is required. Figure 6(a) provides the phase boundaries for T c in terms of the pairing potential g 0 for LiC 6 in which μ 0 = 0. For a given transition temperature T c , by changing the chemical potential μ 0 of LiC 6 via gating, one can engineer the pairing potential g 0 . Figure 8 gives a g 0 -μ 0 phase boundary diagram at T c = 0.1 K. As illustrated in this figure, similarly to pristine graphene, decoration with Li atoms makes it is possible to change the dominant pairing and to have a symmetry-change phase transition from d to "distorted s-wave." Changing μ o up to μ o−v ≈ 0.22 eV so that the distance between the Fermi energy and the saddle points decreases, leads to a decrease in g 0 . Continuously increasing μ o up to 0.5 eV causes g 0 to increase for both d-wave and "distorted s-wave" pairing, and after that a smooth decrease proceeds. For both symmetries at critical μ o−c = 1.3 eV mixed state exist.
Up to μ o−c = 1.3 eV, the flat band plays a primary role in formation of Cooper pairs with lowest energy. The Bloch wave function of this band consists of d and p character, therefore Γ 12 , Γ 15 , Γ 45 and Γ 48 in Eq. 13 carry minus signs. This makes it evident from Eq. 17 that d-wave pairing is dominant. Beyond that, the uneven part of the "flat band" and also upper bands assume a major role. These bands consist of d, p, s, and f character Bloch wave functions (as defined in earlier sections) with a significantly low density of states. In this case Γ 12 , Γ 15 , Γ 45 and Γ 48 change their sign, hence s-wave pairing is favored.
Numerically we have demonstrated that electron pairing g 0 in the limit of pristine graphene is minimal for all dopings. Our calculations indicate that any perturbation of the flat band reduces T c . The flat band can be perturbed through electron hopping from decorating atoms to carbon sites (t LiC 1 ) or by hopping symmetry breaking index w t . For fixed doping at n = 0.11 electron per carbon site and for fixed w t = 0.94 as obtained for lithium decorated, in a variety of Li-C hopping between 0.3-0.4 eV, numerical calculation doesn't show significant To summarize, our calculations indicate that d-wave phases exist and are dominant symmetry of pairing in both pristine and Li decorated graphene. Pure s-wave phase does not appear in LiC 6 , and s-wave superconductivity in metal decorated graphene is disfavored because of spatially increased overlap for s-wave symmetry. These results show that while degree of doping plays a major role in the graphene superconductivity, perturbation effects of decorating atoms finally determine the phase diagram. Our work also provides a new type of classification of superconducting phases in LiC 6 -like nanostructures, and certain aspects of the formalism may be useful in modeling the recently observed superconductivity in magic angle bilayer graphene 3 .