Supplementary Figures Supplementary Figure 1 | the Evolution of the Band Structure of Franckeite With

3 increased thickness for a n = 1. b, n = 2. c, n = 3 and d, bulk.

Sb in the Q layer and 50% Sn and 50% Sb in the H layer. c, Calculated band structure of 27 franckeite with 50% Sn and 50% Sb in the Q layer and 100% Sb in the H layer.        The temperature gradient across the junction was set to ΔT ≈ 36 K. Note that the       i.e., 50% (or 100%) of the Sn atoms have been replaced by Sb atoms in the H layer. In  Figure 6a) is placed on a Au substrate which serves as electrode (indium is used to ensured 396 electrical contact between the two of them), and a mechanically cut Au wire is used as a tip.
Silver is typical in cylindrite group of minerals like incaite and potosiite, and it is present in 438 this sample in a small amount too: Ag 3d 5/2 core level (Supplementary Figure 10b) appears 439 at 367.8 eV binding energy that could be compatible with Ag inclusions in an oxide state.

440
As contaminants we find carbon and oxygen: C 1s peak at 285.0 eV; oxygen core level O1s 441 is not directly observable due to its coincidence with Sb 3d that it is one of the most intense As in the single flake, the best fit with the minor number of meaningful components results 460 in two doublets for each of the measured element: we name "A-doublet" to the doublet with 461 lower binding energy and "B-doublet" to the one with higher binding energy.  , S18b-f and S19a show optical images and the corresponding 515 extinction spectra of these various IPA redispersions. From these results, it appears that 516 exfoliation is occurring in IPA/water 4/1 and water, although in the latter case scattering is 517 dominating the extinction spectrum and the quantity of produced material is very low.

518
Exfoliated material can also be recovered from the other samples, in similar proportions to 519 that obtained from IPA/water 4/1. To be noticed is the particularly high amount of material 520 recovered from the IPA/water 1/4 sediment; cumulated with the material initially dispersed 521 in this mixture, the total amount of exfoliated material compares with that achievable in the 522 best dispersing media (i.e. IPA and IPA/water 1/1; Supplementary Figure 18b   Colloidal stability of the suspensions depends on many parameters, but mainly on the 562 dispersing abilities of the corresponding solvent, the weight (related to size and thickness) 563 of the particles, and the temperature. All of these parameters are involved in our exfoliation 564 process and are to be considered to explain the results obtained and ensure reproducibility.

565
All samples prepared in NMP show long-term colloidal stability due to the strong 566 coordination abilities of this solvent. Such a property also ensures highly reproducible 567 exfoliation experiments because it can compensate for small variations in the size of 568 exfoliated nanosheets during bath sonication or in the temperature during centrifugation.

569
On the contrary some experiments, in IPA/water 1/4 for example, were found more difficult 570 to replicate at the beginning, sometimes leading to particle-free supernatants after   The photoresponse of the device is analyzed upon illumination with light wavelengths of 717 940 nm and 885 nm (Fig. 5d in the main text). Here, a photovoltaic behavior is observed 718 due to the presence of a short-circuit current (I SC ) and an open-circuit voltage (V OC ), which 719 increase with increasing photon energy due to a higher absorption of the franckeite flake.

720
The electrical power harvested in the device, calculated as P el = │I ds │·V ds , is represented in