Coupled Colloidal Quantum Dot Molecules

Coupling of atoms is the basis of chemistry, yielding the beauty and richness of molecules. We utilize semiconductor nanocrystals as artificial atoms to form nanocrystal molecules that are structurally and electronically coupled. CdSe/CdS core/shell nanocrystals are linked to form dimers which are then fused via constrained oriented attachment. The possible nanocrystal facets in which such fusion takes place are analyzed. Coherent coupling is manifested by a red shift of the band gap, in agreement with quantum mechanical simulations. Single nanoparticle spectroscopy unravels the attributes of coupled nanocrystal dimers. This sets the stage for nanocrystals chemistry to yield a diverse selection of coupled nanocrystal molecules constructed from controlled core/shell nanocrystal building blocks, which are of direct relevance for numerous applications in optics, displays, sensing and biological tagging.


Introduction
Colloidal semiconductor Quantum Dots (CQDs) that contain hundreds to thousands of atoms have reached an exquisite level of control, side by side with gaining fundamental understanding of their size, composition and surface controlled properties leading to their implementation in technological applications 1 . The strongly quantum confined energetic levels of CQDs possess atomic like character, for example -s and p states, related to their spherical symmetry. This, alongside with the ability to manipulate CQDs into more elaborate structures, naturally led to their consideration as "artificial atoms". Inspired by molecular chemistry, in which functionality of molecules depends on how atoms couple, we apply analogous concepts to enrich CQDs based materials. If one considers CQDs as artificial atom building blocks 2,3 , how plentiful would be the selection of composition, properties and functionalities of the corresponding artificial molecules? Herein we introduce the utilization of CQDs as basic elements in "nanocrystal chemistry" for construction of coupled colloidal nanocrystals molecules focusing on homodimer quantum dots (QDs), in analogy to homonuclear diatomic molecules.
Coupled quantum dots were prepared by means of molecular beam epitaxy (MBE) [4][5][6] . However, MBEgrown double quantum dot structures exhibit some limitations. First, the size of MBE-grown QDs is larger than the colloidal ones, and the typically large distance between the QDs limits wave-function tunneling that yields coupling phenomenon. Correspondingly, such structures exhibit wave-function tunneling that typically yields coupling energies of a few meV confining their utility to low temperature operation in specialized cryogenic applications 7,8 . Furthermore, MBE grown structures are inherently buried within a host semiconductor 9 . In contrast, colloidal quantum dots are free in solution and accessible for wetchemical manipulations through their surface functionalization. Using such knobs, CQD molecules were constructed by connection with DNA strands providing geometrical control 10  CQDs showing dual emission peaks 11 . Other examples constitute synthesis of dot-in-rod structures and growing an additional quantum dot region on the rod apex thus yielding a coupled system 12 , and dumbbell architectures 13 . However, these progresses were either restricted by specific morphologies 14 , specific materials and relatively large coupling barrier distance and height [15][16][17] . Therefore, there is a lack of a general approach for producing coupled CQD molecules in which there is flexibility to tailor the potential energy landscape and to tune the coupling strength.
To this end we introduce a facile and powerful strategy for coupled CQD molecules with precise control over the composition and size of the barrier in between them to allow for tuning their electronic coupling characteristics and optical properties. This entails the use of core/shell CQDs as artificial atom building blocks. In terms of the band gap engineering, in first instance, tuning the core size is used to manipulate the wave-functions and energies of the electron and hole. On top of this, further control is afforded by the synthesis of shells on these cores.
While the chemical bond is the basis for combining atoms in molecules, connecting CQDs has to occur through adjoining of their crystal faces to form a continuous crystal. Thus, fusing two core/shell CQDs yields a homodimer with a tailored barrier dictated by the shell composition, thickness and fusion reaction conditions. With such control, using high resolution aberration corrected scanning transmission electron microscopy, we observed and analyzed the orientation relationships including homo-plane-attachment and hetero-plane-attachment in the fusion process. Moreover, the manifestations of quantum coupling were revealed by the broadening and red shift of the band gap transition observed in absorption and photoluminescence, in agreement with the quantum-mechanical calculations for the system. The scale of the hybridization energies and corresponding shifts are significantly lower as compared to diatomic molecules. This is expected considering the much larger dimensions of the CQD building blocks compared to atoms and the different potential energy landscape within the CQD molecule. The coupling also leads to broadening of the excited state transitions of the CQD dimers and the absorption spectrum for the high energy bands is modified as well. The emerging attributes of coupling are also revealed by single nanoparticle spectroscopy studies yielding modified electron-hole recombination rates and single photon statistics in CQD dimers in comparison to monomers.
The approach introduced herein, serves as a basis for a wide selection of CQD molecules utilizing the rich collection of the artificial atom core/shell CQD building blocks. Such CQD molecules bear significant promise for their utilization in numerous applications, including in light-emitting devices, displays, photovoltaics and sensors. For example, the controlled formation of heterodimers consisting of CQD monomers with varying core sizes is of direct relevance for dual color emission. Similarly, forming a heterodimer with a staggered (type-II) band alignment between the two CQD building blocks is envisioned for electric field sensing. Additionally, there is high potential for CQD molecules to be used in emergent quantum technologies such as quantum computation 8 . MBE grown QD molecules were already demonstrated as quantum gates utilizing the interaction between the quantum dots through tunneling 5 .
CQD molecules offer enhanced coupling efficiency by their smaller sizes and by the small distances between the QDs resulting in quantum mechanical coupling an order of magnitude larger than prior MBE grown QD systems that is well resolved even at room temperature. This advance significantly widens the scope of quantum technologies applications of coupled quantum dot systems.

Formation of coupled CQDs molecules
Exemplary coupled homodimer molecules were generated from CdSe/CdS core/shell 18,19 CQDs via a procedure utilizing silica nanoparticles as a template for forming molecularly linked dimers 20 , which are then fused via a high temperature reaction (Fig. 1a, full scheme in Supplementary Fig.1). Three different CdSe/CdS core/shell CQDs were studied (1.9/4.0 nm, 1.4/2.1 nm, and 1.2/2.1 nm core-radius/shellthickness, see SI materials and methods for synthesis details and Supplementary Fig.2 for transmission electron microscopy (TEM) images and optical spectra). The TEM and high angle annular dark field (HAADF) scanning transmission electron microscopy (STEM) characterization manifests the wurtzite structure of the monomer CdSe/CdS QDs ( Supplementary Fig.3). These CdSe/CdS CQDs were bonded via thiol linking to the surface of a SiO2 nanoparticle template substrate ( Supplementary Fig.4-5). A second SiO2 layer was grown for masking the remaining SiO2 surface and to immobilize the bonded CQDs ( Supplementary Fig.6), followed by treatment with a tetra-thiol linker ( Supplementary Fig.7). Adding a second CQD leads to formation of a molecularly linked dimer structure (Fig. 1b, Supplementary Fig.8).
Next, the SiO2 template nanoparticles were selectively etched by HF treatment. Size-selective precipitation was used to separate out the monomers and obtain a highly dimer enriched sample (Fig. 1c). Calculated (red asterisk) and experimental (blue circles) band gap red shift of monomer-to-respectivehomodimer structures for CQD molecules with different core/shell dimensions.
The dimerization procedure yields a dimer structure with an organic insulating barrier. Hence, to achieve a coupled system, a last step of fusion is required. The fusion procedure was performed while adding Cd-oleate and heating to 180 °C for 20 h. Fig. 1d-e presents the fused dimer structure after a size selection procedure (Supplementary Fig.9-10). At this non-trivial important stage, the reaction parameters, including temperature, time and ligands type and concentration, have a significant influence on the coupled dimers formation. If the temperature was too high (above 240 °C), collapse of the dimer structures through linker bond cleavage may occur, as well as CQD ripening distorting the core/shell architectures. On the other hand, if the temperature was too low, the fusion rate would be too slow and inefficient. The dimer structure formation is also very sensitive to excess of ligands in the solution, which inhibits the fusion and leads to a decrease in the dimer yield. Therefore, careful tuning and choice of these reaction parameters is crucial for achieving high dimer yields and lower yields of dimer collapse and ripening, while achieving a continuous linking region of the shell materials forming the barrier between the two cores in the fused dimers. Further considerations of this important fusion step and the resultant interfacial structures are discussed later.

Optical signatures for coupling and wavefunction hybridization
After the fusion step, the resultant CQD dimer leaves an interesting optical signature of a red-shift in the absorption and photoluminescence spectra along with broadening of the band gap and excited state spectral features ( Fig. 1 and Supplementary Fig. 11). Generally, there are several factors which can lead to a red-shift: the formation of alloying shell 21 , alteration of the dielectric environment 22 (surface ligands) or interfacial strain 23 . To address these different possibilities, we also studied the spectral properties of the monomers, which underwent the fusion reaction under similar conditions (Supplementary Fig. 9-10), and found them to be identical to the original monomer particles ( Supplementary Fig. 12). Hence the possibility of observing a red shift in the band gap transition due to formation of an alloy shell or altered dielectric environment can be ruled out. Furthermore, strain effects and change in the dielectric properties during the fusion procedure can be considered negligible as we did not grow an additional shell, but rather the fused shell material is the same (CdS), and the surface ligands are also the same for the CQDs monomers and dimers. Moreover, no shift was observed after the fusion of the large 1.9/4.0nm CQDs ( Fig.1), where dielectric and strain effects, if significant, would be expected to contribute as well. In fact, the red shift in the band gap transitions was found to depend systematically on the alteration of the core size and shell thickness of the monomer counterparts, increasing for small core and shell dimensions (Fig  1j). This is consistent with the difference in the delocalization of the wavefunctions in the various CQDs that lead to different degree of coupling of the corresponding wavefunctions in the CQD molecules.
To this end, we have employed quantum mechanical calculations to visualize the wavefunction hybridization and to calculate the expected red-shift in the different CQD dimers. The changed potential energy landscape upon fusion leads to hybridization of the monomer QD wave-functions in the dimers ( Fig. 1f), in analogy to homonuclear diatomic molecules. We utilized finite element software (COMSOL) to calculate the energy levels and wave-functions of the fused CdSe/CdS dimer and monomers within an effective-mass based approximation (Supplementary Discussion and Supplementary Table 1 for details).
The conduction band in this system is demonstrating the fundamental textbook example of hybridization. According to this model, when the distance between two atoms is decreased, their wave-functions will hybridize to form a symmetric bonding state and anti-symmetric anti-bonding state with energy difference of twice the hopping energy. The bonding and anti-bonding electron wave-functions, which, respectively, are in-phase and anti-phase superpositions of the monomer wave-functions, are presented in Fig.1g for the case of 1.4 nm core radius and a potential energy barrier between the dots of 4.2 nm (0.1eV band-offset), corresponding to the CQD molecules formed from 1.4/2.1 nm core/shell CQDs. Because of the quasi-type II nature of the CdSe/CdS interface, the monomers electron wave-functions are easily hybridized and leading to 23 meV energy spacing between the bonding and anti-bonding electron states. For the hole however, the valence band potential manifests a relatively high band offset of 0.64 eV, and this, combined with the heavier hole effective mass, yields minimal hole hybridization.
Considering the case of one exciton residing in the dimer and taking into account the Coulombic interaction between the electron-hole pair, since the hole wave-function is essentially not hybridized, the hole is in one of the dots and consequently the electron does not see a symmetric double QDs' potential anymore. The calculated two lowest energy levels wave-functions of the electron including the Coulombic interaction are presented in Fig.1h. The Coulombic interaction for the first electron level, localized around the hole, is greater than the second electronic state in the opposite dot, increasing the energy spacing between the bonding and anti-bonding states to 60 meV. One can see that the electron is localized in the dot which contains the hole as well. However, there is still significant tunneling-coupling observed for the electron wavefunction and a red shift is predicted. This is indeed confirmed experimentally in the emission and absorption spectrum, where only in the case of the fused dimer a red shift is observed compared to the monomer (603-607 nm in case of 1.4/2.1 nm core/shell QD), whereas, for the unfused organically linked dimer no red shift is seen ( Fig. 1i and Supplementary Table 2). The control in the magnitude of the red-shift, for the monomer to fused dimer transition, for three different types of CQDs is depicted in Fig. 1j. In the the case of 1.2/2.1 nm core/shell CQDs the red shift is increased (13/14 meV calculated/experimental) due to the greater spill out of the electron wave-function to the shell because of 8 the smaller core size. This is in contrast to the case of 1.9/4 nm core/shell CQD where the red shift is negligible (0.5/0 meV calculated/experimental) because of the localization of the electron wave-function in the larger core ( Supplementary Fig.13, and Supplementary Table 2).
An additional signature for the coupling in fused dimers is observed in the absorption spectra at higher energies. Figure 1i shows broadening only upon fusion consistent with coupling forming multiple states in dimers. Furthermore, the spectra normalized at the band gap manifest a significantly stronger relative absorbance in high energies for the fused dimers compared with monomers and unfused dimers (see also Supplementary Fig.11). This is assigned to the wavefunctions modification in the fused system, which can be considered from a viewpoint of hybridization among the excited states.

Structural characterization of fused CQD dimers
We next consider further the nontrivial fusion stage and its consequences. Analysis by HAADF-STEM confirms that coupled dimer formation is indeed achieved based on fusion of the 1.9/4.0 nm core/shell QD monomers ( Fig. 2). A continuous atomic lattice through the entire structure was formed upon fusing the two QDs shells (Fig. 2a). The core architecture in the coupled structure was maintained as demonstrated by the energy dispersive X-Ray spectroscopy (EDS) line scan measurement ( Fig. 2b-  Underlying our fusion reaction strategy, is the process of oriented attachment -a crystal growth mechanism in which secondary mono-crystalline particles can be achieved through oriented and irreversible attachments of primary particles [24][25][26][27][28][29][30] . PbSe CQD dimers were prepared via oriented attachment in solution, but even under succinct control of ligands, concentration and reaction conditions, only ~30% dimer fraction, a rod-like colloidal quantum system, could be achieved along with monomers and higher order oligomers. In our template based strategy, high control over dimer formation was achieved by firstly forming a connection by molecular linkers. The molecular linkers however constrain the initial relative crystal orientations between the two monomers. With careful tuning of precursor and judicious choice of the fusion condition, we can foresee high potential for this method to serve as a general coupling strategy for other colloidal nanocrystal systems. Here, we studied this special case of "constrained oriented attachment", and  Supplementary Fig.3). In such homo-plane attachment cases, both CQD monomers of a fused pair are projected under the same zone axis. This allows accurate identification of the fused faces at dimers orientated with its fusion axes normal to the projection zone axis (depicted in Fig.3). Hetero-plane attachment orientation is observed at fusion of heteronymous faces: (0002)||(101 ̅ 0) (Fig. 3e), Fig.14a,c). common. This is consistent with an interplay between the relative reactivity, surface passivation and occurrence of the various faces on the monomer QDs. The (0002) facets, while in minority, manifest a Cd rich termination with 3 dangling bonds per atom, that can easily react with thiol linkers 31 . Both (101 ̅ 0) and (101 ̅ 1) facets are plentiful but better passivated 32,33 . However, linking to the (101 ̅ 1) facet is sterically hindered. The hetero-plane attachment statistics is also consistent with these considerations.
The generality of our formation strategy is well manifested also for the other CQD homodimers.    Fig.18 for representative traces and statistics of the lifetimes). Analyzing the lifetimes of the high intensity occurrences (green shaded regions in Fig. 5b,c), yields a significantly shortened average lifetime of 5ns for the single fused dimer compared with the monomer (29ns). It is noteworthy that within the fused dimers sample, we have detected ~15% of particles that have similar fluorescence characteristics as the CQD monomer sample, in line with their fraction from TEM analysis ( Supplementary Fig.19-20).
This establishes that the fusion procedure in its entirety did not change the core/shell CQDs.
We consider two possible mechanisms related to coupling, both leading to shortening the lifetime in dimers. First, resonance energy transfer between the two dots (Fig. 5d), a mechanism nearly equally active for unfused and fused dimers. Second, tunneling of the electron to the other dot (Fig. 5d) as already illustrated in Fig. 1. Tunneling in unfused dimers occurs by collisional electron transfer and is strongly dependent on the linker 36 , while in the fused dimer the potential barrier for tunneling is modified and reduced substantially. Both mechanisms will be enhanced for the smaller CQD dimers, but tunneling is more strongly dependent on the size/distances. Indeed, the large CQD molecules, where tunneling probability is negligible according to our calculations (Fig. 1), show smaller changes in lifetimes with shortening in the case of dimers compared to monomers and little differences before and after fusion (Fig.   5c, Supplementary Fig.17, 21) the notable lifetime shortening that is seen for the small CQD molecules upon fusion is indicative of the enhanced contribution of the tunneling mechanism in this case.
Next, we consider the photon statistics, which in CQDs are strongly influenced by multicarrier effects.
An increment in the g 2 (0) value was found in the case of dimers in general with the possibilities of either two emission centers in the excitation spot or intrinsic properties of coupled systems. Specifically, in the fused dimers, the particles absorb the light as one unit, and at low excitation regime (<N>~0.1) the possibility of emission from one of the centers is rational statistically. The full understanding of the g 2 (0) value requires more rigorous experimental studies while an interesting multicarrier configuration can be realized in these fused CQD dimers. In dimer CQD molecules, a new type of biexcitons can occur, with each exciton occupying a different core (Fig. 5e). The large increase in the value of of [g 2 = (0 ) at low excitation power] observed for the dimers versus monomers can be explained by this new type of biexciton for which the non-radiative Auger decay will be strongly suppressed increasing the biexciton quantum yield (QY) (Supplementary Fig.22). Moreover, the single particle exciton QY will decrease for dimers on account of the tunneling of the electrons reducing the electron-hole overlap.
An additional difference relates to the fluorescence flickering in the dimers rather than distinct on-off fluorescence of the monomers (Fig. 5c, Supplementary Fig.23). This indicates presence of multitude emitting configurations for dimers. Indeed, the lifetime traces for the high/low intensity regions in the dimer are not single exponential. The low intensity region is above background and not off and the lifetime has a ~5ns component. All this indicates to trion formation (positive or negative). In small CQD monomers, the trion states are strongly quenched by the Auger decay yielding an off state behavior. In dimers, which have large volume and the excess carrier may occupy the second dot region forming a new type of trion (Fig. 5e), the Auger rate is suppressed, and the trion can become emissive. Such an effect was reported for large CdSe/CdS core/shell manifesting gray state emission and is also observed by us for the large CQDs 37 . The multitude possibilities for emissive trion formation can explain the larger distribution of observed fluorescence intensities and the lifetime behavior for the dimers. To further address this point, an excitation intensity dependence was performed varying the average exciton occupancy <N> from 0.03 to 0.18 (Fig. 5f, Supplementary Fig.24)
After stirring for 1 min, the solution was stored overnight. The SiO2 NPs were collected by centrifugation and dispersed in 10 mL of ethanol. Then the SiO2 solution was mixed with PVP solvent (0.02 g/mL) for 30min. Finally, the nanoparticles were stored after the cleaning by centrifugation.

The synthesis of CdSe/CdS@SiO2
1mL of SiO2 NPs (0.0079 g/mL) was mixed with 0.5 nmol CdSe/CdS NPs using vortex for 5 min. Then 5 mL of ethanol was added into the vails to precipitate and remove the unattached NPs. After three washing cycles the final SiO2@CdSe/CdS NPs were redispersed in 5 mL of ethanol.

The synthesis of SiO2@CdSe/CdS@SiO2
The CdSe/CdS@SiO2 was dispersed in 5 mL of ethanol. Then 330 μL of ammonia solvent (28.5% wt %) was added into the solution with stirring for 5 min. Thereafter, 50 μL of TEOS was added dropwise into the solution. After stirring for 10 h, the resulting solvent was centrifuged (6000 rpm, 5 min) and redispersed in 5 mL of THF.

The synthesis of Dimer-CdSe/CdS@SiO2
A tetrathiol linker ,pentaerythritol-tetrakis ( This process is repeated iteratively until the electron and hole energies converge. In most cases, three iterations are sufficient to obtain a convergence. The emission energy was calculated by: Where (1.76 eV) 3 is the energy gap of CdSe. is calculated in a way which avoids the consideration of the coulomb potential twice (both for the electron and both for the hole).  where σ is the absorption cross section of the particle at the excitation wavelength (375 nm in our case), J and EPhoton are the pulse energy and photon energy of the excitation light, respectively, and A is the spot area. The absorption cross-section was calculated according to the procedure reported elsewhere. 6 The value of the absorption cross-section for the 1.4/2.1 nm CQDs and the corresponding fused dimer at 375nm was found to be 2.3×10 -15 cm 2 and 5.7×10 -15 cm 2 , respectively.

Formation of coupled CQDs molecule dimers
We utilized silica nanoparticles as a template 9 for the fabrication of coupled CQD molecules. The detailed route-line is depicted in Supplementary Figure 1. This was performed by the following steps: 1. Fabrication of SiO2 nanoparticles, coated by MPTMS. This kind of SiO2 particle presents thiol groups on its outer surface, which are later used for the binding of the CQDs.
2. Core/shell CQDs binding to the SiO2 particle surface: mixing a solution of the chosen core/shell CQDs with the SiO2 nanoparticles allows their binding to the available thiol sites on the silica surface.
3. Growth of a secondary thin layer of SiO2 on the QDs@ SiO2 for masking: In this manner the CQDs are immobilized and cannot rotate or reorient while only a top hemisphere is remained exposed for further chemical functionalization of the CQDs.
4. Selective surface decoration of the CQDs by linker groups: Chemical grafting of a functional structure based on thiol group as linkers is then applied which selectively reacts only with the exposed NC hemisphere. Here, a tetrathiol ligand was added as a linker by a ligand exchange reaction on the exposed CQDs surface (for example oleyamine). Additional optional purification (size selective separation) of the dimers from free monomer and higher linked oligomers is possible in between steps 6 and 7, or after the fusion step. Step 1: SiO2 nanoparticles synthesis.
Step 2: core/shell CQDs binding to the SiO2 particle surface. Step 3: masking by a secondary thin layer of SiO2 growth on the QDs@SiO2.
Step 4: selective surface decoration of the CQDs by linker groups.
Step 5: dimer geometry formation on the silica surface.
Step 7: fusion to form coupled CQD molecules.

CdSe@CdS Core/Shell NCs synthesis and characterization:
Supplementary Figure

Structural characterization
The core/shell structure of the monomer is directly identified by atomic resolution STEM-HAADF imaging followed by fast Fourier transform (FFT) analysis (Supplementary Figure 3). Supplementary   Figure 3a, b shows raw and Fourier filtered images of CQD monomers, respectively, viewed under curve -integrated intensity of SAED (k)).

Step I -Silica nanoparticles characterization
The SiO2 nanoparticles (step 1) were prepared as described in the methods section 11 and characterized by TEM (Supplementary Figure 4). These nanoparticles are inherently covered by thiol linkers from the MPTMS precursor allowing QDs binding at the next step. During the washing step, small amount of diluted base was utilized to avoid cascade and aggregation and to ensure a uniform binding on each SiO2 sphere surface in the next step.
Supplementary Figure 4. TEM images of SiO2 nanoparticles prepared by MPTMS precursor acquired at different magnifications. Step

II -Characterization of the Silica-QDs conjugates
The Step IIIformation of a silica masking layer The secondary masking silica layer provides two functions: firstly, coverage of the inherent surface thiol groups of MPTMS in order to avoid the adsorption of additional CdSe-CdS CQDs in the dimerization of step 5. 11 This enhances the efficiency of the dimer formation versus monomers binding.
Secondly, this immobilizes the CdSe-CdS CQDs such that they cannot rotate, exposing a hemisphere which emerges in the solvent and can be modified selectively by the chemical grafting of a functional structure/group.
In order to control the thickness of the secondary SiO2 masking layer, an optimized mixture of PVP and TEOS was necessary. If the PVP amount is too low, it usually results in inefficient masking.
Conversely, if it is too high it could lead to a full QDs masking by the SiO2 layer, which prevents the dimer formation in the next step. Additionally, the optimization of the TEOS amount was performed as described below. According to the calculation, 50 L of TEOS leads to a hemisphere mask. Three Step

IVlinker binding
For the dimer formation step, the chosen linkers bind to the exposed region of the anchored CQDs (step 4). A Tetra-thiol molecule was used as a bi-dentate linker molecule (Supplementary Figure 7).
The thiol groups strongly bind to the CQD surface and can displace the existing surface ligands of the exposed CQD hemisphere. In order to enhance the conjugation of the linkers, the surface modification procedure was performed under Argon flow, at 60 °C overnight. The excess linker molecules were removed by precipitation and centrifugation. Here, the cleaning step after the linker addition was significant for achieving high dimer formation yield.

Step V -Dimer formation
For the preparation of homodimers a CQDs, a ratio of 1:1. Step

VI -Dimers release
Precisely controlled dimer CdSe-CdS CQD molecules were successfully achieved as follows. The release and separation of the CdSe-CdS CQDs dimers from the SiO2 spheres was performed by selective etching process of the SiO2 using an HF/NMF (10%) etching solution. The free dimers are shown in Fig. 1c. The freed dimers were separated by centrifugation decanting the supernatant and repeated ethanol precipitation/centrifugation cycles for three times.
Step VII -Fusion to form the coupled NC molecule The fusion procedure plays a significant role in reduction of the potential barrier in the coupled CQDs molecules. The choice of correct temperature, precursor amount and ligands lead to the fusion of two CQDs without ripening and collapse.

Absorption spectrum measurement for 1.4/2.1 nm CQDs
The features in the absorption spectrum change significantly after the fusion of two CQDs. To further demonstrate the changes, we have normalized the data at the bulk absorption regime (300nm). It can be observed that the band-edge absorption feature of the unfused dimer is retained and similar to the monomers. Upon fusion broadening is seen, along with a red shift and lower absorption. The excited state features are also broadened significantly upon fusion. Upon normalization at the band-edge, the significant relative change in absorption at higher energies upon fusion is emphasized (Supplementary Figure 11a). All these aspects indicate coupling effects upon fusion and formation of the CQD molecules.

Spectral characterization of the various fused CQDs dimers
The red shift of the fluorescence was enhanced with a size decrease in the CdSe-CdS CQDs composing the fused dimers. That is, the smaller core-shell CQDs molecules present strong coupling properties. Additionally, upon fusion, the full width at half maximum (FWHM) measured from the fluorescence spectrum was obviously increased compared with that of the monomer sample. For the atomic model, the Cadmium atoms are marked in brown and Sulfur atoms in blue.

nm CQDs and the corresponding dimers
The fluorescence decay of a single 1.4/2.1 nm CQD follows mostly a mono-exponential decay of ~30ns or a bi-exponential decay with a small contribution from a 5 ns component. The average lifetime for the single dimer particles is clearly quenched and follows mostly a tri-exponential decay.
The distribution of the average lifetime for dimer particles is shifted to short values, which is more pronounced for the fused dimers than for the unfused ones.  (b) Intensity distribution of the traces in (a). Narrow distribution is observed with decreasing laser power.