Synthesis of Ti4Au3C3
Figure 1a shows a HRSTEM image of a sputter-deposited epitaxial Ti4SiC3 film on Al2O3(0001) substrates with a TiC(111) seed layer, taken along the [\(\:11\stackrel{-}{2}0\)] direction. The image reveals a characteristic laminated structure where Ti4C3 sheets are sandwiched between Si (c.f. silicene) layers. The film was sputter-coated with a 200 nm thick Au layer, and the Au-covered Ti4SiC3 sample was then annealed at 670°C for 30 h in a N2 atmosphere to obtain a high-quality Ti-Au-C phase. The HRSTEM image and correspoding EDX map of the annealed sample in Fig. 1b, and 1c show a characteristic nanolaminated structure corresponding to the M4A3X3 phase, where Ti4C3 sheets are now sandwiched between three-atom-thick Au layers. The c-lattice parameter increased from 22.8 Å for Ti4SiC3 to 32.5 Å for Ti4Au3C3, reflecting a 42.5% increase. Despite this significant lattice expansion, the coherent laminate structure is retained. The expansion is attributed to the insertion of Au, where each Au layer consists of atoms larger than those of the original Si in the A layer. The Ti and Au signals in the EDX map align well with the alternating laminated Ti and Au layers, confirming the insertion and location of Au. The corresponding EDX spectrum shows the presence of Ti, Au, and Si (Fig. 1d). The relative atomic ratio of Ti:(Au + Si) is about 4:3, consistent with the stoichiometry of a hybrid 433 MAX phase, as shown in the inset table of Fig. 1d. The low content of Si (at the1.5 at.% detection level) indicates that the Au intercalation was nearly complete. Figure 1e shows the XRD patterns of the as-grown Ti4SiC3 film and the annealed Au-covered Ti4SiC3 sample. The corresponding (000l) peaks shifted to lower angles after annealing, reflecting the lattice expansion along the c-axis. The c-parameter is calculated to be ~ 32.3 Å from the (0004) peak, which is in good agreement with the value obtained from direct observation in the STEM image (Fig. 1b).
The mechanism of the substitution intercalation reaction can be explained by the loosely bonded Si atoms that are provided with a reduced chemical-potential path to diffuse out into the Au capping layer, leaving behind vacancies that are subsequently backfilled with Au. This process is facilitated by the interdiffusion abilities of the two elements, in conjunction with their eutectic phase diagram25, 26. Similar transformations of Ti3AuxC2 (x = 1 and 2) from Ti3SiC2 and Tin+1Au2Cn from Tin+1AlCn (n = 1 and 2) have been reported25, 27. A novel aspect of this work is the demonstration that three atomic layers of Au can substitute each Si layer in Ti4SiC3.
In large-scale STEM images (Supplementary Section 1), peculiar examples of intercalation frontlines for the trilayer Au were observed, where the second and third gold layers closely follow the first layer, effectively pushing apart two adjacent Ti4C3 sheets. The deformation of the Ti4C3 sheets is conspicuous during the insertion of three atomic layers of Au, as indicated by the blue lines. Ti3C2Tx sheets have shown a comparable Young’s modulus, and exhibit high ductility, tensile strength as well as toughness22, 28. Similarly, high elasticity is expected for Ti4C3 sheets, which explains why the laminated structure was retained despite undergoing significant deformation and lattice expansion.
The intercalation of monolayer Au in Ti4SiC3 was also explored at lower annealing temperatures with shorter durations to synthesize Ti4AuC3 (Supplementary Section 2), however, this phase is not prevalent. To investigate the impact of inserting additional layers of Au into Ti4AuC3 from a theoretical perspective, multiple structures were considered. These are illustrated in Supplementary Section 3 and include different stackings of the Au layers as well as different stackings of the Ti4C3 subunits relative to the Au layers. Their energies have been calculed using DFT, with a focus on the configurations found to be lowest in energy. These are shown in Fig. 2a with calculated lattice parameters and space groups symmetries provided in Supplementary Table S1. It is important to note that for Ti4Au3C3, Ti4Au4C3, and Ti4Au5C3, there are additional structures with alternative stacking of Au that are close in energy to those illustrated in Fig. 2a. The common feature of all these low-energy structures is that the Au layers are stacked in an fcc, hcp, or mixed configuration. Using Eq. 1, we calculated the energy gain or cost associated with inserting additional layers into Ti4AuC3. Figure 3b shows that it is energetically favorable to insert additional layers of Au into Ti4AuC3, as indicated by the negative energy change of -0.285 eV when transitioning from Ti4AuC3 to Ti4Au2C3 and − 0.078 eV when transitioning from Ti4Au2C3 to Ti4Au3C3. At three layers of Au, i.e., Ti4Au3C3, a minimum energy of -0.363 eV per inserted Au layer is achieved relative to Ti4AuC3. The addition of a fourth or fifth Au layer, however, will cost energy by + 0.061 eV and + 0.032 eV, respectively, when transitioning from Ti4Au3C3 to Ti4Au4C3 and Ti4Au5C3. Additional phase stability data for Ti4Au1 + xC3 compared to its set of most competing phases can be found in Supplementary Table S2.
A reason why three layers of gold (Ti4Au3C3) are energetically most favorable, as shown in Fig. 2b, is revealed through the bonding analysis in Figs. 2c, d and Supplementary Fig. S11, combined with the evaluation of bond lengths shown in Fig. 2e. For Ti4AuC3, Au-Au interactions are only found in-plane. However, when additional layers of Au are inserted into the structure, out-of-plane Au-Au interactions appear. Beyond the trilayer Au configuration, for the Au layer in the middle, out-of-plane Au-Au interactions occur without direct bonding to Ti. This positively impacts bonding, as seen in Fig. 2c, where the integrated partial crystal orbital Hamiltonian population (IpCOHP) increases with additional Au layers. Analysis of the individual contributions in Fig. 2d shows that Ti-Au interactions also benefit from the presence of multiple Au layers. Figure 2d further demonstrates that three layers of Au in Ti4Au3C3 provide the largest total individual contribution. This is corroborated by the bond lengths shown in Fig. 2e, where Ti-Au and Au-Au bonds (adjacent to the Ti layer) are shortest for Ti4Au3C3, indicating improved bonding strengths compared to configurations with fewer – or more – than three Au layers. The experimentally calibrated in-plane Au-Au distances, which are slightly shorter than the calculated values for Ti4Au3C3 embedded in Fig. 2e, are found in Supplementary Section 4.
Another perspective on the relative stability of the tri-layer goldene is from the difference in electronegativity between Au (2.54), Ti (1.54), and C (2.55) compared to Si (1.90) or Al (1.61). As Au is more electronegative than Ti, there will be electronic charge transfer from Ti to Au. The stability of goldene stacks would then be a delicate balance of charge-transfer for the metal bonding between the Au atoms and a trilayer of Au happens to cause the “right” amount of charge transfer from Ti to Au for its highest stability, compared to less or more gold layers (see Supplementary Section 4, including test DFT calculations).
Atomic stackings in tri-layers Au in TiAuC
In this study, we found that trilayers of Au in Ti4Au3C3 contain both ABA and ABC stacking configurations (Fig. 1b and Supplementary Section 4), corresponding to 2H hcp and fcc structures, respectively. The coexistence of hcp and fcc stacking does not lead to a spread in the c-parameter. For the fcc phase of Au, the calculated a-parameter is 3.091 Å and the c- parameter is 32.529 Å. These values are very similar to those of the hcp phase of Au (Supplementary Table 1), which aligns with the experimental observations (Supplementary Section 4). The Ti4Au3C3 phase with ABA stacking of Au has an energy of -0.363 eV, whereas the ABC stacking has an energy of -0.354 eV (Supplementary Fig. S8). Both trilayer structures are therefore feasible as stable Au allotropes. From simulated ABA structures (Supplementary Fig. S8), Ti4C3 sheets are mirrored with Au layers. In contrast, for the ABC stacking, Ti4C3 sheets stack in a zig-zag pattern along [\(\:11\stackrel{-}{2}0\)] with respect to the Au layers. Since the Ti4C3 sheets in Ti4SiC3 and Ti4AuC3 are mirrored with A layers (Si or Au layers), as observations in STEM images and simulated structures (Fig. 1a, and Supplementary Figs. S2 and S6), the position of the Ti4C3 sheets did not shift laterally when the monolayer Au was introduced. This also applies to the Ti4Au3C3 with ABA-stacked Au, where the Ti4C3 sheets did not shift. To achieve the formation of ABC-stacked Au layers, displacement of two adjacent Ti4C3 sheets is required. Such a displacement was observed in STEM images (Fig. 1b), where Ti atoms move laterally by approximately 0.5 Å relative to their original mirrored positions. Similar phenomena have also been observed when inserting bi-layer Au into Ti3SiC2 to form Ti3Au2C225. However, gliding of Mn+1Xn slabs in the MAX family has not previously been reported. A possible mechanism for Ti4C3 gliding is discussed in Supplementary Section 5.
The co-existance of hcp and fcc quasi-goldene sheets can be understood by considering the reduced dimensionality of the metal, which increases the proportion of surface energy relative to the total system energy. Consequently, material properties can be tuned through crystal phase engineering. For instance, hcp metals often exhibit anisotropic properties due to the directional arrangement of atoms, while fcc metals typically have isotropic properties. These anisotropic materials are of interest across various disciplines, because of unique properties. Compared to their fcc counterparts, hcp Au nanoparticles (NPs) have shown plasmon and interband transitions29, and a 130-fold increase in in-plane resistivity and more pronounced plasmon absorption have been demonstrated in 4H Ag30. Zhang’s group synthesized ~ 2.4 nm-thick 2H hcp Au nanosquare sheets and 4H Au nanoribbons using wet-chemical methods31, 32. Ye et al. reported an hcp Au phase existing at the edge of two-atomic-thick fcc Au nanosheets9. Kondo et al. synthesized fcc Au nanowires encapsuled by an hcp Au outer shell33. Such non-fcc crystalline phase are known to stabilize ultrathin Au nanostructures34, 35. An hcp/fcc alternating structure was observed in a square-like Au plate due to the hcp to fcc transformation as the plate grows thicker36. To the best of our knowledge, the finding of large-scale, isolated, sub-nanometer-thick Au sheets with hcp structures is also original.
Preparation of trilayer goldene
To obtain isolated trilayer goldene, Ti4C3 sheets were selectively etched away using 0.5% Murakami’s reagent with 5 mM of CTAB for 168 h. A schematic illustration is shown in Fig. 3a. Ti4C3 sheets are stepwise oxidized by radical nascent oxygen [O] generated in an alkaline solution of potassium ferricyanide (K3[Fe(CN)6] in KOH)1, 37. CTAB surfactant was used to permeate the gaps once Ti4C3 slabs between freed and to impede the agglomeration of 2D gold layers into multilayers or nanoparticles, as applied for gold nanoparticles38, 39. We obtained an average in-plane Au-Au spacing of 2.86 Å in trilayer goldene (Supplementary Section 6), which is close to the equilibrium interatomic distance in fcc bulk Au (2.884 Å) and slightly larger than the DFT-calculated values (Supplementary Section 7). The Au-Au spacing in trilayer goldene is approximately 6.5% smaller than that in Ti4Au3C3 due to lattice contraction after exfoliation. Figure 3b shows an etching frontline observed at the edge of a Ti4Au3C3 film, where trilayer goldene is being exfoliated from the right side. The goldene sheets remain separate after etching, but begin to ripple from the edges once losing the support of the Ti4C3 sheets. The gap opening observed at the edges can be attributed to the fact that a CTAB chain, with a length of up to 20 Å, is larger than the distance between goldene sheets. CTAB molecules, which bind vertically to Au surfaces in a bilayer formation40, can expand the gaps between adjacent goldene sheets. The goldene sheets in the more deeply etched regions (the central area of the image) appear to maintain relatively better flatness, likely due to a slower infiltration of CTAB molecules parallel to them. Initial results indicate that an etched-free trilayer goldene is more stable than the corresponding monolayers1 and can maintain its structure with ripple features extending over a hundred nanometers. Meanwhile, blob formation of Au at the edges and their lateral diffusion through the sheets occurs (Supplementary Fig. S19). A magnified image reveals some four- and five-layer goldene sheets near the edges, as shown in Fig. 3c. The formation of blobs and thicker layers can be attributed to the rapid interactions between the released goldene sheets and excess spurious Au atoms during etching processes1. Additionally, this phenomenon is related to the coalescence of goldene sheets during ion milling in the TEM sample preparation processes (Supplementary Fig. S20). The black regions in Fig. 3c clearly indicate the complete removal of Ti4C3 sheets and confirm that the thickness of the three-atomic-layer Au is approximately 6.7 Å. The thickness of trilayer goldene inside of Ti4Au3C3 is measured to be 6.7 Å (see Fig. 3b). The corresponding EDX map of Au exhibits a distinct lamellar signal, while Ti is evenly dispersed (Figs. 3d, e), further confirming the complete etching. Solutions with higher concentration of Murakami’s reagent would be more aggressive towards Ti4AuC3, leading to faster etching of Ti4C3 sheets and the formation of Au particles through sheet clustering and curling up of sheets. Conversely, etching with a lower concentration of 0.2% requires a much longer time to achieve complete etching (Supplementary Fig. S21). In addition, AIMD simulations have shown that two adjacent goldene layers that contain Si impurities coalesce in few picoseconds if their interlayer spacing is within 7 Å1. The goldene-goldene interaction weakens significantly as the interlayer spacing increases from 10 Å to 12 Å and 14 Å. In this work, the distance between Au sheets is about 12 Å in Ti4Au3C3, which is larger than the ~ 9.3 Å spacing in Ti3AuC2. In contrast, the coalescence of two adjacent trilayer goldene after etching was less observed, as wider channels are provided for surfactants to stabilize the free-standing goldene sheets. Furthermore, the as-synthesized Ti4Au3C3 film is very close to ideal stoichiometry with negligible Si content, as confirmed by STEM-EDX and XRD (Fig. 1), implying that any disturbance of goldene layers by Si impurities is minimal.
Pure Au hcp square sheets have been found to become stable under ambient conditions when they are less than ~ 6 nm thick31, and ultrathin nanowires have been stabilized by hcp surface structures33. Therefore, it is not surprising that the present trilayer goldene contains hcp-stacked regions. These factors collectively suggest promising potential for the production of larger quasi-2D Au sheets.
DFT and AIMD investigations of free-standing trilayer goldene
To investigate the dynamic and energetic stability of fcc-like and hcp-like trilayer goldene sheets, DFT calcuations were carried out at 0 K and AIMD simulations at 300 K (Supplementary Section 7). DFT calculations show that ABA and ABC trilayer Au slabs have nearly equal energy. For both Perdew-Burke-Ernzerhof (PBE) and local density approximation (LDA) results, the energy difference between fcc and hcp slabs is less than 1 meV/atom. It has been experimentaly demonstrated that the hcp phase exists in Au nanostructures41, as the stacking fault energy in fcc metals is quite low.
On-the-fly machine learning (ML)-assisted AIMD simulations at 300 K were conducted to verify the dynamical stability of fcc- and hcp-stacked trilayer goldene. The dynamics of the trilayers were monitored for 0.18 ns (fcc) and 0.38 ns (hcp). The simulations indicate that both fcc and hcp stackings are dynamically stable, as shown by time-averaged Au positions in Supplementary Fig. S23. After a brief transient period, during which the Au layers shift to their equilibrium separation distance, the atoms continue vibrating around their respective fcc (or hcp) lattice positions for the entire simulation. The time-averaged potential energies suggest that, at room temperature, the hcp stacking is approximately 45 meV/atom more stable than the fcc stacking. An additional contribution arises from the vibrational free energy (Fvib), with the hcp-structured sheet being further stabilized by 5 meV/atom more than the fcc (Supplementary Fig. S22). Accordingly, the Helmholtz free energy (F) of the two allotropes, obtained by adding Fvib to the time-averaged potential energies, shows a difference of ∆Fhcp−fcc ≈ 50 meV/atom, indicating that hcp trilayer Au is substantially more stable than fcc trilayer Au at 300 K. Therefore, we predict that both structures would be retained after etching, with the hcp structure likely being more prevalent, as inferenced from the dominant ABA-Au in Ti4Au3C3 (Supplementary Section 4). Indeed, hcp-structured trilayer goldene is observed (Supplementary Fig. S18). The prevalence of trilayer goldene may thus be attributed to the relative stability of the hcp stacking.
Electronic properties of trilayer goldene
We performed XPS measurements on Ti4Au3C3 before and after etching with 0.5% Murakami’s reagent as well as on a reference sputter-etched Au foil. Figure 4 shows the corresponding Ti 2p, C 1s, and Au 4f core-level XPS spectra. The Au 4f7/2 peak of the pure reference Au film (Fig. 4, right panel) is located at 84.0 eV with a 4f spin-orbit splitting of 3.7 eV, consistent with reference values42. The Au 4f7/2 and 4f5/2 peaks of the unetched Ti4Au3C3 film are asymmetric revealing the presence of a second low-intensity doublet located at binding energies (Eb) of 84.9 eV and 88.6 eV, respectively, i.e., shifted by 0.9 eV to higher Eb with respect to the Eb of the reference Au metal (84.0 eV and 87.7 eV). The occurrence of the high Eb doublet is attributed to electronic charge transfer primarily from the Au 5d states to the Ti3C2 sheets in Ti3AuC2. Similar effects were observed in the Ti3AlC2 and Ti2AlC MAX phases1, 43, 44, where charge transfer takes place from Al to the Tin+1Cn (n = 1 or 2) sheets. The stronger 4f doublet in the spectrum from the unetched Ti4Au3C3 film appears at nearly the same Eb as for the reference Au. This can be attributed to the partially remaining capping Au layer on top of Ti4AuC3C after chemical-mechanical polishing and/or to Au-Au interactions in the trilayer Au. In contrast to Ti3AuC2, which has in-plane Au-Au interactions within each Au monolayer and Au-Ti interactions on both sides, in the case of Ti4Au3C3 Au-Au interactions are stronger relative to Ti-Au resulting in reduced charge transfer and a higher intensity of the main Au 4f doublet with respect to the high Eb pair.
The etched Ti3AuC2 film exhibits significant high Eb shoulders, shifted by ~ 0.9 eV with respect to the main Au 4f doublets. These shoulders are attributed to final state screening and charge transfer effects1. After etching, the Au 4f spectrum of the trilayer goldene, shown at the top of Fig. 4, reveals that the 4f doublet peaks are nearly at the same energy positions as those of the reference Au. In addition, a less intense high-energy tail is observed, that can be fitted with a second 4f7/2-4f5/2 doublet shifted to higher Eb values of 84.8 eV and 88.5 eV, respecitvely. Such Eb shifts in the high-energy tails of Au 4f spectra from monolayer goldene have been attributed to the final state effects45, 46. It has been demonstrated that both the Eb and the peak width of the Au 4f lines increase as the size of Au nano particles and clusters decreases, due to limited screening of the core hole left after photoionization47. In monolayer goldene produced from Ti3AuC2 (Fig. 4), the final state effects are enhanced due to the smaller coordination number of Au atoms (six or fewer) compared to bulk fcc Au and Ti3AuC21. In contrast, in trilayer goldene, the coordination number of the middle layer is 12, while the outerlayers have a coordination number of 9, likely decreasing further at the sheet edges. Therefore, the final state effects are weaker in trilayer goldene compared to the monolayer, resulting in a less intense tail and a slightly smaller Eb shift. The dominant Au 4f doublet in the spectrum from etched Ti4Au3C3 may originate from residuals of the capping Au layer or from sheets clustering and their curling-up. The tendency of goldene sheets to transform to 3D shapes can be expected from the surface morphology on the Ti4Au3C3 after etching (Supplementary Section 9).
The C 1s XPS spectrum of the etched sample (Fig. 4, middle panel), reveals that the intensity of the carbide peak at ~ 281.7 eV has drastically decreased as compared to the unetched sample. This indicates that almost all Ti4C3 has been removed during the etching of Ti4Au3C3. The corresponding Ti 2p spectra (Fig. 4, left panel) further confirm this conclusion: the Ti 2p3/2-Ti 2p1/2 peaks due to the carbide (at 454.8 and 460.8 eV, respectively) are very weak in the spectrum from the etched sample. The latter is dominated by a doublet peaks located at 458.6 eV and 464.5 eV, respectively, i.e., indicative of TiO2 formation on the Au layers. Minor impurity peaks from Fe 2p, Br 3d, and K 2p photoemissions48 imply that a small amount of iron and potassium residue from the etchant occurs at the surface after etching (Supplementary Section 10).
Prospects
We propose that, in addition to monolayer1 and trilayer goldene (present work), isolated bilayer goldene can also be obtained by etching Ti3Au2C2, a compound reported in 201725. Combined with the recently reported goldene1, 25, this work demonstrates that the thickness of isolated Au sheets can be tuned into one-, two-, and three-atomic layers using our deviced etching scheme, which includes the use of surfactants. This approach provides a foundation for future fundamental and applied studies on ultrathin Au with varying thicknesses. When using a MAX phase precursor with a larger interlayer distance, such as Ti4SiC3, there is a greater likelihood that surfactants will penetrate in-between the A layers and stabilize them during exfoliation. Furthermore, by selecting appropriate exchangeable elements, the A layers can be substituted with other noble metals such as silver, platinum or iridium, leading to the formation of both novel MAX phases and their corresponding metallenes through selective removal of the Tin+1Cn slabs.