Formation of Colloidal Copper Indium Sulfide Nanosheets by Two-Dimensional Self-Organization.

Colloidal 2D semiconductor nanosheets (NSs) are an interesting new class of materials due to their unique properties. However, synthesis of these NSs is challenging, and synthesis procedures for materials other than the well-known Pb- and Cd-chalcogenides are still underdeveloped. In this paper, we present a new approach to make copper indium sulfide (CIS) NSs and study their structural and optical properties. The CIS NSs form via self-organization and oriented attachment of 2.5 nm chalcopyrite CuInS2 nanocrystals (NCs), yielding triangular- and hexagonal-shaped NSs with a thickness of ∼3 nm and lateral dimensions ranging from 20 to 1000 nm. The self-organization is induced by fast cation extraction, leading to attractive dipolar interactions between the NCs. Primary amines play a crucial role in the formation of the CIS NSs, both by forming in situ the cation extracting agent, and by preventing the attachment of NCs to the top and bottom facets of the NSs. Moreover, DFT calculations reveal that the amines are essential to stabilize the covellite crystal structure of the product CIS NSs. The NSs are indium-deficient and the off-stoichiometry gives rise to a plasmon resonance in the NIR spectral window.

Ultrathin 2-dimensional nanomaterials (nanosheets, NSs) are attracting increasing attention due to their unique physical, electronic, and structural properties.[1−11] Ultrathin colloidal semiconductor NSs with thickness (h) in the strong quantum confinement regime (i.e., h ≤ exciton Bohr radius) are of particular interest, since they combine the extraordinary properties of 2D nanomaterials with versatility in terms of composition, size, shape, and surface control, and the prospects of solution processability.[9] To date, colloidal NSs of a variety of binary semiconductors have been prepared: CdX (X = S, Se, Te),[1,6,10] PbS,[2] SnX (X = S, Se),[12,13] InSe,[14] In2S3,[15] Cu2–S,[5] Cu2–Se,[16] WS2,[17,18] and MoS2.[18] However, reports on NSs of multinary semiconductors are scarce, despite the interesting properties of this class of materials. CuInS2 (CIS) for instance, is a direct semiconductor with a bulk bandgap of 1.5 eV[19,20] and large absorption coefficients,[21] which yields nanocrystals (NCs) with photoluminescence tunable in the vis–NIR (600–1100 nm) spectral range.[22−25]

Size and shape control in a direct synthesis of colloidal multinary NCs is challenging, as the reactivity of three different precursors has to be precisely balanced to prevent formation of the binary system.[24] We circumvented this limitation earlier by a partial Cu+ for In3+ topotactic cation exchange reaction, through which template Cu2–S NCs are converted into CIS NCs with size and shape preservation.[25−27] A similar cation exchange protocol has been very recently followed by Mu et al.[28] to convert template Cu2–S NSs[5] into CIS NSs. The two-stage method developed in our work and described in this paper relies on a different mechanism: 2D self-organization of CIS NCs. The resulting 2D NCs have lateral dimensions ranging from 20 nm to 1 μm, thickness of ∼3 nm (similar to the height of the pyramidal-shaped parent NCs), and a Cu:In ratio of ∼3:1, which is different from the 1:1 stoichiometry of the parent CIS NCs. Our results indicate that the self-organization of the parent CIS NCs into 2D NCs and NSs is driven by a sudden change in their composition due to preferential extraction of In3+ by in situ generated reactive sulfur-containing species (e.g., H2S). The fast initial growth by oriented attachment of NC building blocks is followed by a slower phase in which NC building blocks are individually added to the edges of the NSs. The crystal structure changes during the reaction from the cubic chalcopyrite structure of the parent CIS NCs to hexagonal covellite. DFT calculations indicate that the ligands play a pivotal role in the formation of the 2D covellite structure and that the In atoms are preferentially located at the surfaces of the NSs. Interestingly, the Cu-rich stoichiometry of the product NSs gives rise to a plasmon resonance in the NIR, making the covellite CIS NSs synthesized in the present work promising materials for plasmonic applications requiring manipulation of NIR light,[29] such as chemical sensing,[29] IR spectroscopy,[29] photothermal therapy,[30]in vivo photoacoustic imaging,[31] smart windows,[32] and thermal doping.[33]

Experimental Section

Materials

Copper(I) iodide (CuI, Sigma-Aldrich, 98%), indium(III) acetate (In(Ac)3, Sigma-Aldrich, 99.99%), 1-dodecanethiol (DDT, Sigma-Aldrich, ≥98%), sulfur (Sigma-Aldrich, 99.98%), 1-octadecene (ODE, Sigma-Aldrich, technical grade 90%), oleylamine (OLAM, Sigma-Aldrich, technical grade 70%), octadecylamine (ODA, Sigma-Aldrich, technical grade 90%), indium(III) chloride (InCl3, Sigma-Aldrich, 99.999%), zinc(II) chloride (ZnCl2, Sigma-Aldrich, 99.99%), ammonium bromide (Fisher Scientific, +99%), ammonium chloride (Sigma-Aldrich, 99.99%), toluene (Sigma-Aldrich, 99.8%), methanol (Sigma-Aldrich, 99.8%), and butanol (Sigma-Aldrich, 99.8%) were obtained. ODE, OLAM, and ODA were degassed for 2 h at 150 °C before use. All other chemicals were used as received.

CuInS2 Parent Nanocrystal Synthesis

CuInS2 nanocrystals were synthesized following a protocol by Li et al.[34] First, 297 mg of In(Ac)3, 191 mg of CuI, and 5 mL of DDT were mixed and degassed for 1 h under vacuum at 100 °C. The reaction mixture was then heated under N2 pressure to 230 °C. After 5 min of reaction the heating mantle was removed, and the mixture was allowed to cool down to room temperature (RT). This yields trigonal-pyramidal-shaped CIS NCs with a height of 2.5 nm. The NCs were washed with a methanol/butanol 1/1 mixture, isolated by centrifugation and redispersed in 8 mL of ODE for further reactions or toluene for further analysis.

Nanosheet Synthesis

A 0.4 M stock solution of sulfur dissolved in ODE (S-ODE) was prepared by heating sulfur in ODE to 180 °C until a clear solution was obtained. A 1 mL portion of CuInS2 NCs in ODE (stock solution described above) was mixed with 0.25 mL of OLAM and 0.75 mL of S-ODE in a vial. The vial was then placed in a well in a preheated aluminum block on a hot plate at 200 °C. After 2 h the vial was removed from the aluminum block, and the product NCs were precipitated with a methanol/butanol 1/1 mixture, centrifuged, and redispersed in toluene. For separation of the nanosheets from small nanoparticle byproducts, the dispersion of the product NCs in toluene was left undisturbed for several days at room temperature under N2 to sediment. Subsequently, the supernatant was carefully removed with a pipet, and toluene was added to the sediment.

In experiments using ODA instead of OLAM, ODA was melted at 60 °C, and then, 0.75 mL was added to the reaction mixture.

In experiments designed to investigate the influence of excess cations in the reaction outcome, 0.0244 g (0.11 mmol) of InCl3 or 0.0252 g (0.11 mmol) of InCl3 and 0.0153 g (0.11 mmol) of ZnCl2 were added to the reaction mixture described above.

In experiments designed to investigate the influence of halide anions, 160.5 mg (3 mmol) of ammonium chloride or 294 mg (3 mmol) of ammonium bromide was added to the reaction mixture described above.

Absorption Spectroscopy

NC solutions in toluene were stored in sealed quartz cuvettes. Absorption spectra were measured on a double-beam PerkinElmer Lambda 950 UV/vis spectrophotometer.

Electron Microscopy

Transmission electron microscopy (TEM), cryogenic TEM, and electron diffraction measurements were performed on an FEI Tecnai-12 microscope. High-resolution transmission electron microscopy and energy dispersive X-ray spectroscopy (EDX) were performed on an FEI TalosF200X microscope. Samples for TEM imaging were prepared by drop-casting a toluene solution containing NCs or nanosheets on a carbon-coated copper grid. Samples for EDX were prepared by drop-casting on a carbon-coated aluminum grid. The sample for cryogenic TEM was prepared using an FEI Vitrobot instrument. The electron diffraction patterns were first radially integrated using the CrystTBox toolbox.[35] The pattern obtained was scaled with a scaling factor determined by fitting a measured gold reference electron diffraction pattern to a gold reference signal.

DFT Calculations

The binary Cu–S system has a very rich phase diagram, with a variety of equilibrium compositions and crystal structures, ranging from chalcocite Cu2S to covellite CuS.[24] Calculations were carried out for the CuS composition because the CIS NSs synthesized in the present work were observed to have adopted the covellite crystal structure. The phase diagram of the binary In–S system is also very rich, and therefore the 1:1 stoichiometry was chosen to allow a direct comparison with covellite CuS. The calculations of CIS bulk structures and bare CIS slabs were performed using the plane augmented wave (PAW) method[36] with the generalized gradient approximation (GGA) of Perdew–Burke–Ernzerhof (PBE)[37] as implemented in the Vienna ab initio simulation package (VASP) code.[38,39] The cutoff energies for the wave functions and augmentation functions were set to 550 and 770 eV, respectively, and sufficiently dense k-meshes were used to ascertain energy convergence to within 2 meV/atom. This resulted in k-meshes of 14 × 14 × 4 for the covellite structure, 6 × 6 × 4 for 2 × 2 × 1 supercells of the covellite structure, 14 × 14 × 8 for the chalcopyrite structure, and 16 × 8 × 18 for the orthorhombic InS structure.[40] Both the crystal lattice and the atomic coordinates were fully relaxed. The formation enthalpy is defined with respect to pure CuS covellite and pure InS covellite asFor ligand-covered configurations, supercells were constructed that were 58 Å in height, containing a vacuum slab at least 22 Å thick. Only for these configurations, the OptB88-vdW functional by Dion, as implemented in the VASP code by Klimeš et al.,[41] was used to take into account van der Waals interactions. After full relaxation (dimensions and atomic coordinates) the lateral cell dimensions were kept fixed for evaluation of the ligand coverage using OptB88-vdW while allowing full relaxation of atomic coordinates.

Results and Discussion

Briefly, the nanosheets (NSs) are synthesized by suspending dodecanethiol-capped chalcopyrite CIS NCs in a solution of elemental sulfur in octadecene and a primary amine (octadecylamine or oleylamine, OLAM), and subsequently heating the mixture to 200 °C (see the Experimental Section for details). This induces the formation of triangular and hexagonal NSs at the expenses of the parent CIS NCs (Figure ). Interestingly, the NSs only form when both sulfur and a primary amine (either octadecylamine or OLAM) are present (Supporting Information, Figure S1). Elemental sulfur has been previously shown to react with alkylamines forming alkylammonium polysulfides at room temperature, and alkylthioamides, dialkylamidines, alkyl-thioketoamidines, and H2S at elevated temperatures (<100 °C).[42] Under appropriate conditions the released H2S was shown to readily combine with metal precursors present in the reaction medium to form metal sulfide NCs.[42] In a control experiment (Supporting Information, Figure S2) we show that H2S is also formed under the conditions used in our NS synthesis method. This implies that H2S (or another product of the reaction between OLAM/ODA and sulfur) is crucial to trigger the conversion of the parent CIS NCs into the product NSs. A mechanism for this transformation will be proposed later in this paper. Figure A shows that after 1 min at 200 °C the NSs have already grown to 25 ± 4 nm in lateral size, which is 10 times larger than the parent CIS NCs (trigonal-pyramidal-shaped with a base of 2.5 nm and height of 2.4 nm).[23] Nanosheets also form at lower temperatures (e.g., 80 °C, Figure S3; or even at room temperature, Figure S4), but the growth kinetics are much slower. The NS thickness is determined to be 3–4 nm by analyzing TEM images of NS stacks (Figure S5).

Figure 1

Formation of nanosheets from 2.5 nm parent CIS NCs at 200 °C followed over time with ex situ TEM: (A) 1 min, (B) 2 min, (C) 5 min, (D) 1 h, and (E) 5 h of reaction time. (F) Cryo-TEM image of a sample after 2 h at 200 °C. Scale bars correspond to 100 nm.

The growth of the NSs at 200 °C was followed ex situ with TEM. Over time, both the lateral dimensions and the polydispersity of the NSs increase, with sizes ranging from tens to several hundreds of nanometers after 1 h of reaction (Figure A–D). Interestingly, most of the growth occurs in the first 5 min. If the reaction is allowed to continue for several hours, the NSs start to shrink and break down (Figure E). Small irregularly shaped nanoparticles, with sizes ranging from 4 to 11 nm, are observed to coexist with the nanosheets at all times (Figure and Figure S6). Cryo-TEM (Figure F) shows that these small nanoparticles are present both as free-standing particles in solution and as adsorbates attached to the nanosheet surface. Most importantly, the cryo-TEM images also clearly demonstrate that the NSs are already present in the reaction medium, thereby excluding the possibility that they are formed on the TEM grid by drying effects or electron-beam-induced aggregation.

Elemental Composition of the NSs

The formation of nanosheets at 200 °C was also followed over time with ex situ absorption spectroscopy (Figure ). The parent CIS NCs show a rather featureless absorption spectrum extending to ∼650 nm with a shoulder at ∼550 nm, which is ascribed to the band-edge (1S–1S) transition.[23] This shoulder broadens and shifts to longer wavelengths as the reaction proceeds, indicating a reduction in the bandgap of the NCs, which is consistent with a decrease in the quantum confinement as a result of the growth of the nanosheets. Furthermore, an additional broad absorption band appears in the near-infrared (NIR) region, which further broadens and shifts to longer wavelengths over time (see Figure ; peak is at 1700, 2000, and 2200 nm for the 1 min, 1 h, and 5 h samples, respectively). A (strong) absorption band in the NIR is often observed in Cu-chalcogenide NCs and is ascribed to a localized surface plasmon resonance (LSPR) due to excess charge carriers.[24,29] These excess charge carriers are present because of stoichiometry deviations, typically Cu-vacancies, which lead to excess holes in the valence band (i.e., p-doping).[24,29] The observation of a strong NIR absorption band in the spectra of the product CIS NSs thus implies the presence of excess carriers, suggesting that the NSs no longer have the 1:1 Cu:In stoichiometry of the parent CIS NCs.

Figure 2

Absorption spectra of the parent CIS NCs (black) and the product CIS NSs at different stages of the reaction. The dashed line at ∼550 nm marks the position of the band-edge absorption transition of the parent NCs, while the dashed arrow highlights the shift of the broad NIR absorption band to lower energies over time. It is noted that this band is absent in the spectrum of the parent NCs. The sharp peak at 1400 nm is due to residual methanol, which was used to wash the samples.

Indeed, energy-dispersive X-ray spectroscopy (EDX) on wide areas, encompassing a large ensemble of NSs, reveals that the average elemental composition of the nanosheets is Cu/In/S 0.6(±0.01):0.3(±0.004):1 (see the Supporting Information, Figure S7, for a representative example), which is clearly different from the 0.5:0.5:1 stoichiometry of the parent CIS NCs, suggesting that, to keep charge neutrality, (part of) the sulfur ions formed covalent bonds, similarly to the CuS covellite phase.[43] This should result in excess holes in the valence band, giving rise to an LSPR, consistent with the NIR absorption band observed in the spectra of the NSs (Figure ). Elemental analysis of individual NSs (Figure and Figure S8) reveals small variations within a sample: The average Cu:In ratio of 11 measured nanosheets is 1:0.3 ± 0.06. The highest Cu:In ratio is 1:0.39, while the lowest is 1:0.21. Additionally, the small NCs seem to be In-rich CIS, as they appear brighter in the In map than in the Cu map (see Figure and Supporting Information, Figure S8). The elemental maps provide no evidence for the formation of InS NCs, since all regions examined contained both Cu and In, albeit in different ratios (i.e., NSs are Cu-rich, NPs are In-rich). The difference between the average Cu:In ratio determined for the NS ensemble (viz., 2:1) and for individual NSs (viz., 3:1) can be ascribed to the fact that the contribution of the In-rich NCs is much larger in the ensemble measurements. The average composition of the NSs at different stages of the reaction was also analyzed with EDX and observed to be essentially constant. This shows that the cation extraction is fast and is, in fact, already completed within the first minute of the reaction. This observation also indicates that the shift of the NIR LSPR absorption band during the reaction (Figure ) is due to the lateral growth of the nanosheets, rather than a change in their composition. The consequences of these observations for the proposed formation mechanism will be discussed below. It should be noted that the LSPR band originates from the In-poor NSs, rather than from the In-rich CIS NCs. This becomes evident when comparing the absorption spectrum of a sample of NSs obtained by sedimentation with that of the original ensemble prior to sedimentation, when a high concentration of small In-rich NCs was still present (Figure S9). The LSPR band is very pronounced in the absorption spectrum of the sample obtained by sedimentation, but is hardly observable in the spectrum of the sample prior to sedimentation, which is dominated by the absorption of the small In-rich NCs.

Figure 3

(B, C, D) STEM-EDX elemental maps of the nanosheet shown in the STEM-HAADF image in part A. The yellow frame highlights a region where clusters of small nanoparticles are observed at the edge of the nanosheet. The diagonal yellow line marks the edge of the nanosheet. The nanoparticles appear brighter in the In map than in the Cu map, suggesting that they are In-rich CIS. This is clearer visible in the close-up of this image in the Supporting Information, Figure S8.

Shape of the NSs

As shown in Figure (and also in Figures S1D, S6, and S8) the shape of the nanosheets ranges from triangles to hexagons, through truncated triangles and irregular hexagons. Remarkably, the NS edges are often kinked and irregular (see Figure A,B and also Figures S6 and S8A). Figure B also clearly shows small particles adsorbed on the nanosheets’ surface. Moreover, the edges of the NS in Figure B are noticeably brighter than its interior, suggesting that the edges are richer in indium. This observation is consistent with the structural model derived from DFT calculations and will be discussed in more detail below. The high-resolution TEM (HR-TEM) image in Figure C shows lattice fringes of small particles on the nanosheet surface and the nanosheet itself. It is clear that the fringes have different directions, indicating that the small particles have different orientations on the surface and are not atomically aligned with the underlying nanosheet. The wobbly edge of the NS in Figure C is deformation caused by the intense electron beam. This deformation happens in situ while imaging the NS and worsens over time. The implications of the irregularities and kinks observed in the edges of the NSs in the low-resolution TEM images will be discussed in the mechanism section below.

Figure 4

(A) TEM image showing kinked and irregular edges of NSs. (B) STEM-HAADF image showing kinked edges and small particles adsorbed on the surface and at the edges of a NS. (C) HR-TEM image showing lattice fringes of both the nanosheet and small particles adsorbed on the nanosheet surface in different orientations.

Crystal Structure of the NSs

Electron microscopy was also used to study the crystal structure of the nanosheets, by using selected area electron diffraction (SAED) and comparing the integrated pattern to bulk reference patterns. Figure A shows the SAED pattern of a batch of nanosheets on the TEM grid. Azimuthal integration of the SAED pattern yields the signal shown in Figure B, which is in good agreement with a covellite bulk reference pattern. The SAED pattern of a single NS (Figure C,D) shows diffraction spots consistent with a single crystal. The hexagonal pattern can be indexed to covellite lattice planes (Figure S11). The crystallographic properties of the product CIS NSs are further studied with DFT calculations below.

Figure 5

(A) Selected area showing a batch of deposited nanosheets. The electron diffraction pattern of this batch is shown in the inset. A larger version of this panel is shown in Figure S10. (B) Azimuthally integrated SAED signal is compared to a covellite bulk reference pattern (JCPDS PDF card 00-036-0380). SAED of (C) a single nanosheet shows (D) a hexagonal diffraction pattern.

DFT Modeling of Crystal Structures

During the reaction, the 2.5 nm parent CIS NCs, with chalcopyrite crystal structure, transform to NSs with much larger (∼100-fold) lateral dimensions and covellite crystal structure. For insight into the energetics of the crystallographic transformation from chalcopyrite to covellite, plane-wave density functional theory (DFT) calculations were performed using the plane augmented wave (PAW) method with the generalized gradient approximation (GGA) as implemented in the VASP code.[38,39] Computational details are given in the Experimental Section. The calculations, over the entire cation range but limited to compositions with a cation/anion ratio of one, were performed both for the CuS-type covellite structure and for the CuInS2-type chalcopyrite structure displayed in Figure A,B. Also the known most stable orthorhombic phase of InS shown in Figure C was considered. The formation enthalpies, defined with respect to CuS covellite and InS covellite (see the Experimental Section for details), are shown in Figure . Tangent lines are drawn connecting the most stable covellite phases, and the overall most stable phases. The total-energy calculations confirm that covellite CuS, chalcopyrite CuInS2, and orthorhombic InS are overall the energetically most stable phases. In the compositional range of Cu1–InS with x between 0.5 and 1.0, phase separation into chalcopyrite CuInS2 and orthorhombic InS is predicted. For the experiments, only the compositional range with In concentrations between 0.0 and 0.5 is relevant. The DFT calculations show that the chalcopyrite structure is more stable over this compositional range, except for In concentrations below x ∼ 0.05 where the CuS covellite structure is more stable. At an In concentration of x ∼ 0.25 (corresponding to the experimentally determined composition of the nanosheets), the tangent line connecting the most stable covellite phases and the tangent line connecting the most stable chalcopyrite phases are very close, with a formation enthalpy difference of only 104 meV per formula unit (fu). The fact that the covellite structure is found experimentally shows that the ligands, which are not included in these bulk calculations, are required to stabilize the covellite phases with a nonzero In content, thereby bridging this difference in formation enthalpy.

Figure 6

Schematic bulk crystal structures of (A) covellite, (B) chalcopyrite, and (C) orthorhombic InS. Cu atoms are displayed in blue, In atoms in magenta, and S atoms in yellow. The simulation cells are periodic in all dimensions and are drawn with black lines. (D–F) Proposed transformation sequence from chalcopyrite to covellite, with projections corresponding to side views of the NSs. Gray arrows drawn at the right-hand side of panels A, D, E, and F indicate the directionality of bilayers, pointing from anion-filled atomic layers to cation-filled atomic layers. Gray dots indicate atomic layers filled with both anions and cations. The {112} atomic planes of the (D) chalcopyrite structure consist of hexagonal (Cu, In)S2 bilayers which we draw aligned with the {001} planes of the (E, F) covellite superstructure. Upon extraction of In atoms, the Cu atoms occupy the vacant sites, and reordering causes the cation–anion separated bilayers to become (E) mixed atomic layers, thereby providing opportunities for S–S bond formation. Further atomic rearrangements result in structure F, with oppositely directed bilayers that are typical of the covellite structure. This last structure is predicted to be the lowest-energy covellite Cu3In1S4 phase.

Figure 7

Formation enthalpies of Cu1–InS bulk phases, relative to CuS covellite and InS covellite, calculated using GGA–PBE. The covellite phases are indicated by black circles, chalcopyrite phases by red triangles, and the orthorhombic InS phase by a green square. The top black line is the common tangent connecting the most stable covellite phases; the bottom black line is the common tangent line connecting the overall most stable phases. The blue arrow indicates the change induced by the NS formation reaction.

As can be seen in Figure , there are multiple data points for the formation enthalpies of the covellite phases (black circles) at one particular composition. This is because multiple orderings of the Cu/In atoms inside the (Cu, In)S covellite structure were considered, varying from concentrating In atoms in particular layers to a more homogeneous distribution in the structure. From all of these calculations, it can be concluded that it is remarkable that the covellite structure is found experimentally, as—at least for bulk phases—the covellite structure cannot accommodate large concentrations of In atoms without severe distortion. The covellite crystal structure is characterized by triplets of CuS atomic planes (Figure A) that are bonded by S–S bonds oriented perpendicular to the layered structure. The central CuS atomic layer in the triplets is atomically flat. However, upon introduction of In atoms in the structure, in general this central layer becomes buckled as shown in Figure E. It is noted that, in the chalcopyrite structure, the directionality of the anion–cation bilayers is uniform in the [221]ch direction, as indicated with the gray arrows in Figure D. In the covellite structure of Figure A, though, the bilayers show opposing directionality in the corresponding [001]cov direction (also indicated with gray arrows). Figure E is therefore an intermediate structure, where the bilayers do not consist of an anion-filled atomic plane and a cation-filled atomic plane, but of cation–anion mixed layers. The energetically most favorable orderings in the covellite structures were those whereby the In atoms occupy the atomic planes adjacent to the S–S bonds, as shown in Figure F. However, this also leads to strong weakening of the S–S bonds at particular planes, with the interatomic S–S distance increasing substantially from 1.7 Å for pure CuS (Figure A) to ∼3.0 Å for the S atomic layers shown at the bottom of Figure F. In addition, these In-coordinated S layers adopt a more conventional stacking. The large interatomic distances suggest that, along these planes, the structure could be more easily cleaved or layers exfoliated; i.e., surfaces will be formed more easily at these lattice planes.

Although the dynamics of the transformation sequence (with cation extraction, ligand replacement, and oriented attachment all taking place over a very short time window) is much too complex to be resolved by means of total-energy DFT calculations, the calculations point to a plausible transformation pathway which is shown in Figure D–F. The {112} atomic layers of CuInS2 chalcopyrite show structural resemblance to the (001) layers of (Cu, In)S covellite. Starting from the chalcopyrite structure (Figure D), In extraction provides flexibility in the atomic environments, allowing the Cu(I)-sublattice to quickly reorganize itself by occupation of the vacant sites. This results in mixed anion–cation bilayers in the {112} chalcopyrite planes, which then transform into {100} covellite planes (Figure E). This process is likely favored by the very high solid-state mobility of Cu(I) cations in chalcogenide lattices.[25−28,44] Such a transformation can also be much enhanced when chalcopyrite NCs come in a direct solid–solid contact with a larger covellite nanosheet by means of directed attachment. As a result, layered structures are formed that are terminated with S atoms at both sides so that S–S bonds can be formed. Next, atomic rearrangements take place whereby In atoms become concentrated near S–S bonded planes, which enables the formation of the nearly flat CuS atomic planes at the center of the triplet layers that are so typical of the covellite structure (Figure A). Because of the strong S–S bonds in the case of a Cu-rich atomic environment and the weak S–S bonds in the case of an In-rich environment (both shown in Figure F), we hypothesize that the interior of the covellite NSs are Cu-rich, while the surfaces of the NSs are In-rich, and that these InS-terminated surfaces preferably bind to ligands.

Calculations were also performed on a 6-bilayer slab of the chalcopyrite structure, and a 6-bilayer slab of the covellite structure (both at In concentration x = 1/3). Upon comparison of the formation enthalpies of bare {001} covellite CIS NSs with bare {112} chalcopyrite CIS NSs, it was found that the chalcopyrite NSs are energetically slightly more favorable by 57 meV/fu. Therefore, the DFT calculations predict that the 2D covellite structure cannot be stabilized by surface energies alone, and that the ligands likely play a pivotal role in stabilizing the colloidal 2D CIS covellite phase. Additional simulations were performed in which 6-bilayer covellite slabs were covered with amine or thiol ligands. To take into account van der Waals interactions, these calculations were performed with the computationally more expensive OptB88-vdW functional instead of the GGA–PBE functional (details in the Experimental Section). Figure S12 shows the configurations found for bare, crystal-bound thiol-covered, surface-bound thiol-covered, and amine-covered 2D CIS covellite slabs. We note here that not all possible terminations of the CIS covellite structure (Cu, In orderings) were simulated, neither were different ligand areal densities, as this configurational space is too large to be explored with quantum mechanical DFT calculations. As the end groups mainly determine the interaction with the 2D slabs, for computational efficiency only 6-carbon-atom-chain variants of the ligands (i.e., the C6H13 hexyl radical, the C6H13S thiyl radical, and C6H15N hexylamine) were considered. Not surprisingly, all ligands were found to have a favorable binding energy with the CIS nanosheets when taking the energies of the bare nanosheets and of the ligand 2D layers as a reference. However, in the thiol-covered configurations of Figure S12B,C, the remaining forces on the atoms are very high (up to 0.7 and 1.0 eV/Å for the crystal-bound and surface-bound configurations, respectively) rendering thiol coverage unlikely, while the amine-covered configuration displayed in Figure S12D relaxed well, with forces on the atoms on average 0.03 eV/Å and at most 0.09 eV/Å. Therefore, the DFT calculations indicate that OLAM coverage is preferred over DDT coverage of the covellite CIS nanosheets. It is thus likely that the formation of covellite (rather than chalcopyrite) CIS nanosheets was driven by the presence of large concentrations of OLAM (or other alkylamines) in the reaction medium.

Cation Extraction

As shown above, the nanosheets are indium-poor, and the final stoichiometry is already reached after 1 min of reaction. The DFT calculations discussed above imply that the extraction of In from the parent CIS NCs is crucial to initiate the transformation from the chalcopyrite to the covellite structure, which, in the presence of OLAM ligands, can form stable In-poor CIS nanosheets. For verification of whether the extraction of indium is indeed essential for the formation of the CIS nanosheets and whether the final composition of the product NSs can be controlled, cation precursor salts were added to the reaction mixture. Both InCl3 and ZnCl2 were added, and in both cases nanosheets were not formed (Figure S13A,B). In contrast, addition of ammonium halides (chloride and bromide) to the reaction mixtures had no impact on the formation of the nanosheets (Figure S13C,D). This confirms that extraction of In(III) cations from the parent CIS NCs (which is prevented when excess In(III) and Zn(II) cations are present) is essential for the formation of nanosheets.

Formation Mechanism of the CIS Nanosheets

Several mechanisms have been proposed in the literature for the formation of two-dimensional (2D) colloidal semiconductor NCs.[1,2,5,9,10,13−18,45−48] These mechanisms can be roughly divided into three different categories:[9] (i) 2D anisotropic growth by monomer addition to specific crystallographic facets (CdS,[1] CdSe,[1,6,47,48] WS2,[17,18] MoS2[18]); (ii) self-organization by 2D-oriented attachment of NC building blocks (PbS,[2] SnSe,[13] In2S3,[15] Cu2–Se,[16] CdTe[49−51]); and (iii) 2D-constrained growth within soft lamellar templates (Cu2–S,[5] CdS,[10] CdSe,[10] InSe,[14] PbS[46]). These mechanisms have been recently discussed in detail,[9] so we will here only highlight their essential aspects. In the first mechanism, monomers (i.e., molecular [MX] units) formed upon reaction between M- and X-precursors attach to specific crystallographic facets of a growing NC (or a magic-size cluster) resulting in 2D anisotropic growth. The 2D constraints in this mechanism are attributed either to selective capping of the top and bottom facets with suitable ligands,[6,47] or to higher activation energies for monomer addition on the large top and bottom facets.[48] In the self-organization mechanism the 2D geometry of the resulting NC superlattice is ascribed to oriented attachment due to both directional (dipolar) interactions between specific high-free-energy facets[2,49−51] and dense ligand layers on the facets perpendicular to the attachment plane.[2,16] In the soft-template mechanism, the nanosheets form within pre-existing 2D lamellar templates consisting of dense self-assembled monolayers of linear alkyl chain ligands, metal cations, and small anions (such as halides),[10,14,45,46] which impose 2D constraints on growth by both monomer addition to NC nuclei[14,45,46] and self-organization of NC (or magic-size cluster) building blocks.[10]

The mechanism we propose here for the formation of In-poor covellite CIS nanosheets from parent chalcopyrite CuInS2 NCs is based on 2D self-organization of in situ produced NC building blocks, followed by fusion and recrystallization (Figure ). The first step, completed within the first minute of the reaction, consists of a fast extraction of indium by reactive sulfur-containing species formed in situ upon reaction between elemental sulfur and OLAM or ODA (probably H2S, which is the most reactive of the in situ generated species).[42] This sudden change in composition most likely leads to drastic modifications in the free energies and charge distributions of the different crystallographic facets of the parent pyramidal chalcopyrite CIS NCs, thereby inducing dipolar interactions between the NCs, similarly to previous observations on tetrahedral zinc-blende CdTe NCs that had been subjected to reactions changing the surface composition of the NCs.[49−51] We propose that these dipolar interactions between the In-poor CIS NCs lead to their fast self-organization into 2D NSs by oriented attachment driven by minimization of the imbalanced surface free energies (Figure A). On the basis of the DFT calculations discussed above, it is likely that the oriented attachment process is accompanied by a fast ligand-assisted structural reorganization through which the In-poor CIS NC superlattice fuses and transforms into single-crystalline OLAM-capped In-poor covellite CIS nanosheets. The OLAM ligands not only are crucial to stabilize the In-poor covellite CIS phase (see DFT calculations above), but also play an important adjuvant role in directing the 2D self-organization, both by stabilizing the top and bottom facets of the nanosheets and by preventing attachment of NC building blocks on them, thereby constraining the growth to the lateral directions only, as has been observed in other systems.[2] This is clearly illustrated by the observation of randomly oriented small nanoparticles adsorbed on the NS surface, without fusing to it (see, e.g., Figures F, 3, and 4). It should be noted that the reaction between elemental sulfur and alkylamines produces a number of species that can also function as ligands (viz., alkylthioamides, alkyl-thioketoamidines, dialkylamidines).[42] However, these molecules either are bulkier than OLAM (or ODA) or contain a thiol donor group, which, according to the DFT calculations discussed above, should lead to less effective stabilization of the NS surfaces as compared to OLAM.

Figure 8

Schematic representation of the proposed formation mechanism for the In-poor CIS NSs. (A) The first step is the extraction of In from the parent chalcopyrite CIS NCs, which induces a fast initial growth by oriented attachment of In-poor CIS NC building blocks, followed by fusion and recrystallization into single-crystalline covellite NSs. (B) Subsequent growth occurs by addition of CIS NC units to the side facets of the NSs. (C) The NS growth stops when the parent (In-poor) CIS NCs (i.e., “monomers”) in the reaction medium have been almost completely depleted.

The initial fast growth by self-organization is a multibody event in which several NCs must collide to form a 2D superstructure. This process can only be sustained if the concentration of NC building blocks is sufficiently high. At low concentrations of parent NCs, growth can only proceed by sequential addition of individual In-poor CIS NCs to the edges of the NSs, in a process that is equivalent to NC growth by monomer addition[52] (Figure B, the NCs are the “monomers”). This results in the kinked and ragged edges observed in many of the NSs (e.g., Figures F and 4, and Figure S6). Again, the ligands prevent the incorporation of NCs through addition to the bottom and top facets of the nanosheets. The concentration of available In-poor NC building blocks (i.e., free in the reaction medium) eventually decreases below a critical limit, causing the growth of the NSs to stop (Figure C). We note that, as discussed above, the NCs observed simultaneously with the NSs (either attached to their edges and surfaces, or dispersed throughout the TEM grid) are In-rich, and are thus unlikely to be leftover building blocks of the In-poor NSs. We propose that these In-rich NCs are byproducts of the reaction that induces the formation of the NSs, being formed by incorporation of the In(III) extracted from other CIS NCs. This process establishes a partition of the initial ensemble of parent 2.5 nm CIS NCs into 2.5 nm In-poor CIS NCs (which become the nanosheet building blocks) and larger than 2.5 nm (due to incorporation of “InS” units) In-rich NCs that remain as isolated NCs. We note that CIS NCs are very tolerant to stoichiometry deviations and can be readily made with Cu:In ratios ranging from 0.3 to 2.9.[24] Interestingly, the oxidation state of Cu and In in CuInS is the same in all compositions (i.e., +1 and +3, respectively) because the neutrality of the NCs is ensured by the change in the formal oxidation state of the sulfur atoms, leading to excess carriers in either the valence or the conduction bands. As discussed above, this is consistent with the observation of LSPR bands in the absorption spectra of the covellite CIS NSs.

Conclusions

Colloidal indium-deficient covellite CIS nanosheets (NSs) with a thickness of ∼3 nm and lateral dimensions up to 1 μm were obtained via a two-stage synthetic procedure, in which pyramidal chalcopyrite CuInS2 NCs (height, 2.5 nm) were used as parent NCs. Fast indium extraction by in situ generated reactive sulfur species (e.g., H2S) triggers a rapid self-organization and oriented attachment process that is accompanied by fusion and recrystallization of the In-poor CIS NC building blocks, yielding single-crystalline covellite CIS NSs. Further growth occurs by addition of individual In-poor CIS NCs to the side facets of the NSs, until the NC building blocks available for growth are depleted. Alkylamine ligands (OLAM or ODA) are essential for the formation of the NSs both because they are required for the in situ generation of the cation extraction agent from elemental sulfur and because they play a pivotal role in stabilizing the In-poor covellite structure and in directing the 2D anisotropic growth by preventing addition of NC building blocks to the top and bottom facets of the NSs. The Cu-rich stoichiometry of the product CIS NSs gives rise to a plasmon resonance in the NIR, making them promising materials for plasmonic applications.[29−33]