Colloidal semiconductor quantum dots (QDs) have unique optical, physical, and chemical properties. Rapid development in synthesis methods has led to the production of high-quality QDs, which have been used in efficient photovoltaic devices, light-emitting diodes, photodetectors, and biomedical imaging systems1,2,3,4.

Indium phosphide (InP) QDs are direct-bandgap III–V compound semiconductors (bulk bandgap: 1.35 eV) that can be made to emit from the blue (~465 nm) to the near-infrared (~750 nm) region by simply tuning their size5,6,7. InP QDs have relatively low toxicity, so they have the potential to be commercially utilized, especially in displays4,8,9,10.

InP magic-sized clusters (MSCs) were recently discovered as intermediates during the growth of InP QDs10,11,12. InP MSCs possessing high thermal stability may act as barriers to the growth of InP QDs13. Research on understanding the growth pathway, including InP MSCs is being actively conducted for the synthesis of high-quality InP QDs. We believe that unmet challenges in InP QD synthesis, such as monodisperse QDs and precise morphology control, can be overcome by exploring the nature of InP MSCs. Isolation and characterization of InP MSCs and studies on their reaction pathways would be key to controlling the synthetic pathways. In this review, developments in the synthesis of InP QDs and research on InP MSCs are summarized. We classify research on InP MSCs into (1) synthesis and characterization and (2) applications.

History of synthesis of InP QDs and InP magic-sized clusters

Methods to synthesize InP QDs are not well developed. Moreover, the broad emission profile of InP QDs is a demerit. The band-edge PL of a single InP QD has a full width at half maximum (FWHM) of 14 nm at 613 nm, so narrow emission of ensemble InP QDs is not limited on a fundamental level14.

An inhomogeneous size distribution may be the main reason for the broad emission peak, so methods to synthesize monodisperse InP QDs have been sought5,15,16. To narrow the size distribution of InP QDs, some studies have attempted to control the reactivity of molecular precursors. The commonly used tris(trimethylsilyl)phosphine ((TMS)3P) is too reactive and leads to inseparable nucleation and growth steps, so it has been replaced by less-reactive precursors such as tris(trimethylgermyl)phosphine15,16, phosphine (PH3)17, aminophosphine18,19, and others20,21,22,23. However, the use of these different molecular precursors did not lead to any significant decrease in the size variation of the synthesized InP QDs. To date, the emission properties of InP QDs have not been reduced to a level comparable to that of Cd-based QDs. To the best of our knowledge, the narrowest ensemble emission peak of InP QDs reported thus far is an ~35 nm FWHM, whereas that of ensemble CdSe can be ~20 nm4,24,25,26. This wide FWHM of InP QDs motivates research to understand the mechanism of InP QD growth.

Nucleation and growth stages may be key steps that determine the uniformity of InP QDs. The classical LaMer model of nucleation and growth kinetics does not cover the whole of nanomaterial growth27, and nonclassical growth mechanisms have been recently reported28,29,30. InP QDs grow via the formation of intermediate states, which have been called MSCs, molecular nanoclusters or atomically precise nanoclusters10. Here, nanoclusters can be defined as small-sized nanocrystals in the range of 2 nm or less31,32. An MSC is composed of a specific number of atoms that are arranged to form highly stable structures with a size <2 nm (Fig. 1)31,32,33,34. MSCs are generally identified by the wavelength of the persistent lowest energy electronic transition (LEET). The optical properties of MSCs are clearly different from those of general QDs in that the narrow bandwidth of the LEET is a result of single-sized products.

Fig. 1: Schematic illustration of the relative chemical potential in the small size regime of NPs.
figure 1

The relative chemical potential of crystals/clusters in the extremely small size regime has dependence of the size and configuration. Adapted with permission from ref. 34. Copyright 2002 American Chemical Society.

MSCs are quite difficult to control. The formation and dissolution of MSCs also limit the rate of formation of QDs (Fig. 1)34,35. InP MSCs have been captured as intermediates during the growth of InP QDs, and the mass of the InP MSCs has been detected up to 300 °C13; the results indicated that the high thermal stability of InP MSCs influences the growth of InP QDs. This insight led to the isolation and characterization of various InP MSCs for study to increase the understanding of the growth mechanisms of InP QDs.

Synthesis and characterization of chalcogenide MSCs and InP MSCs

Three general synthesis pathways for MSCs have been developed: (1) the interruption method36,37,38, (2) the low-temperature method39,40,41, and (3) the size-focusing method42,43,44. The interruption method obtains MSCs by interrupting the growth of large QDs by injecting a cold reagent immediately after hot injection. The low-temperature method grows MSCs very slowly, so thermodynamically favorable MSCs form. The size-focusing method includes three steps: growth of large QDs at high temperature, quenching of the reaction to stop the growth of large QDs, and isothermal annealing to shrink the large QDs via size focusing for growth of MSCs.

Several InP MSCs with different surfactants have been synthesized using the low-temperature method. InP MSCs capped with a carboxylate ligand that had a dominant absorption peak at wavelength λ = 386 nm (386-InP MSCs) were discovered using indium acetate (In(Ac)3), myristic acid (HMy), and (TMS)3P in 1-octadecene (ODE) at 100 °C10,35. The 386-InP MSC was identified as In37P20(O2CR)51 (R: alkyl group) by single-crystal X-ray diffraction (SCXRD); because of its definite MSC structure, it has become the most-studied InP MSC system45. MSCs with amine ligands have been synthesized at 70 °C12. Phosphonate-capped InP MSCs have been grown at 370 °C10. Strong binding ligands such as phosphonic acid could be useful for the growth of InP MSCs even at high temperature.

InP MSCs that incorporate heterogeneous atoms have been reported11,46,47. InP MSCs incorporating chlorine (InP:Cl MSCs) or zinc (InP:Zn MSCs) are considered in this review as representatives.

To understand the optical, structural, and compositional properties of InP MSCs, characterization studies are necessary. Several methods to characterize semiconductor MSCs have been developed (Table 1).

Table 1 Characterization methods of semiconductor magic-sized clusters.

Confirmation of characteristic features in absorption spectra is the most common method used to identify MSCs after synthesis. The absorption features of MSCs are very unique because they have a very narrow width at the excitonic transition. In addition, their absorption peaks persistently appear at fixed wavelengths. The optical properties of MSCs can be investigated using steady-state absorption/PL, transient absorption (TA)/PL, and photoluminescence excitation (PLE) spectroscopy11. Especially for the magneto-optical properties of doped MSCs, magnetic circular dichroism (MCD) can be useful48,49.

The structure can be analyzed using SCXRD45. SCXRD analysis upon successful crystallization of MSC samples can reveal the whole structure of MSCs, but the preparation of single crystals fit for SCXRD is difficult for several reasons, including the existence of MSC isomers and the intrinsic low symmetry of MSCs. Powder X-ray diffraction (pXRD), Raman spectroscopy, the pair distribution function (PDF) technique, X-ray absorption spectroscopy (XAS), including extended X-ray absorption fine structure (EXAFS) and X-ray absorption near-edge structure (XANES), X-ray photoelectron spectroscopy (XPS), and solid-state nuclear magnetic resonance (NMR) spectroscopy have been used as tools to determine MSC structures, where each of the analytical tools may provide complementary pieces of information for MSC structures11,49,50,51.

The mass of MSCs can be determined using matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry, electrospray ionization mass spectrometry (ESI-MS), laser desorption/ionization mass spectrometry (LDI-MS), and ion-mobility mass spectrometry (IM-MS)13,52,53,54. Despite the soft ionization method, MSCs can often be fragmented, and broad mass peaks can appear30,46. Samples of MSCs often exist as mixtures and require isolation of specific kinds using, for example, gel-permeation chromatography (GPC), high-performance liquid chromatography (HPLC), and thin layer chromatography (TLC)11,55,56.

The size and morphology of MSCs can also be analyzed using transmission electron microscopy (TEM) or scanning transmission electron microscopy (STEM)11,57. However, the limited resolution for small-sized MSCs and sample degradation issues under electron beam irradiation remain challenges. Synchrotron small-angle X-ray scattering (SAXS) has also been used to attempt to measure MSC sizes58.

The compositions of MSCs can be analyzed using inductively coupled plasma (ICP) spectroscopy, energy-dispersive spectroscopy (EDS), Rutherford backscattering spectroscopy (RBS), and thermogravimetric analysis (TGA)11,59,60. The inorganic components of MSCs are mainly determined using these tools. Surface ligand properties such as quantity and binding mode can be determined using 1H NMR, 13C NMR, or Fourier-transform infrared spectroscopy (FT-IR)11,58.

Density functional theory (DFT) calculations can also help elucidate MSCs61. The atomically resolved structure of MSCs revealed by SCXRD can be used for DFT calculations. For example, the correlation of electronic or geometric structural properties with energetics can be identified, which helps interpret the experimental results62. Unknown MSC structures can also be expected via modeling using such calculations61,63,64,65. Much experimental data, such as UV–vis absorption, PDF, EXAFS, and Raman spectra, can be compared to models calculated from simulated data50,63,66,67,68.

Use of chalcogenide MSCs and InP MSCs

Isolation and characterization of MSCs can yield an understanding of their growth process and suggest methods to control the intermediates11. Chalcogenide MSCs have been used as single-source precursors (SSPs)69. Precisely controlled growth was achieved for 2.5–9.0 nm CdSe and 3.0–5.0 nm ZnSe QDs by using (Li)4[M10Se4(SPh)16] (M = Cd, Zn; SPh = thiophenolate) as SSPs. This synthesis approach has advantages of size control, narrow size distribution, and large-scale production of QDs. Furthermore, high-quality QDs grew even at a low reaction temperature TR < 200 °C because MSCs and their fragments could function as seeds for QDs without a nucleation process70. III–V semiconductor QDs such as InP and InAs have been synthesized using nanoclusters as SSPs10,25,71,72,73,74,75,76. Size-focused InAs QDs have been synthesized using InAs MSC-mediated growth. Rapid dissolution of MSCs may be the key to avoiding a broad size distribution that occurs via Ostwald ripening74. Highly monodispersed InAs QDs (<10% size deviation) can be synthesized using ~1.8 nm InAs nanoclusters as precursors73, which may be more stable than molecular arsenic precursors and are therefore suitable as growth precursors. Variously sized InAs QDs can be synthesized using intermediate nanoclusters as growth precursors to maintain the initial homogeneity75.

Semiconductor MSCs are often intermediates during the growth of nanorods, nanowires, nanoribbons, or nanoplatelets34,38,77,78,79. MSCs have been considered to be useful building blocks for the construction of semiconductor nanostructures (NSs) of diverse morphology (Fig. 2a)34. Semiconductor MSCs or their fragments can grow via oriented attachment in certain directions (Fig. 2b, c)78,79. Branched InP NSs (BNSs) and hyperbranched InP NSs (HBNSs)/dendrimer-like NSs (DLNSs) have been synthesized using InP MSCs as precursors80.

Fig. 2: Application of semiconductor MSCs.
figure 2

a Monomer-concentration-dependent, MSCs-mediated growth of CdSe nanocrystals. Adapted with permission from ref. 34. Copyright 2003 WILEY VCH. b TEM image of CdSe nanoribbons synthesized from CdSe MSCs. Adapted with permission from ref. 79. Copyright 2009 WILEY VCH. c TEM image of CdSe single-layered nanosheets synthesized from CdSe MSCs. Adapted with permission from ref. 78. Copyright 2006 American Chemical Society. d UV–vis absorption and PL spectra of oleic acid/trioctylphosphine-capped CdSe MSCs. Adapted with permission from ref. 54. Copyright 2014 American Chemical Society. e UV–vis absorption and PL spectra of oleylamine capped CdSe MSCs. Adapted with permission from ref. 81. Copyright 2015 American Chemical Society. f Thin film of white-light CdSe MSCs in polyurethane excited by a frequency-doubled titanium: sapphire laser (400 nm) with white light clearly seen reflecting off the table surface (left) and a 5-mm commercial UV LED (400 nm) illuminating a thin coating of white-light CdSe MSCs in polyurethane (right). Adapted with permission from ref. 85. Copyright 2005 American Chemical Society. g Representative differential interference contrast image of a green fluorescence protein transfected fibroblast cell incubated with ~100 nM TPP-treated PEG18-thiolate-protected CdSe nanoclusters (red) and green fluorescence protein transfected cell. Adapted with permission from ref. 89. Copyright 2014 The Royal Society of Chemistry.

The luminescence of chalcogenide MSCs and InP MSCs has strong band-edge emission or broad white light emission that involves band edges and traps (Fig. 2d, e)12,54,81. The emission properties of MSCs have been exploited in luminous applications such as LEDs82. Blue-emitting CdSe MSCs with emission peaks at λ = 420 nm or 450 nm have been reported; their quantum yield (QY) was improved from 5 to 24% by simple surface treatment (Fig. 2d)54,83.

Band-edge emission of MSCs is usually associated with broad trap emission at 400 ≤ λ ≤ 800 nm. White emission with QY = 27% was achieved by additional ligand treatment; the MSCs were used as white-light phosphors (Fig. 2e, f)57,81,82,84,85,86,87. PbSe MSCs with 50 ≤ QY ≤ 90% were achieved with no posttreatment88, and ultrasmall CdSe nanoclusters with QY = 6.3% (in water) were used for intracellular imaging (Fig. 2g)89.

Synthesis and characterization of InP MSCs

InP MSCs with different surface ligands

Ultrasmall InP nanoclusters of size 1.5–2.3 nm were successfully synthesized in 200190. A chloroindium oxalate complex and (TMS)3P were used as synthesis precursors; trioctylphosphine oxide (TOPO) and trioctylphosphine (TOP) were added as colloidal stabilizers/solvents. To synthesize ultrasmall InP nanoclusters, TR was dropped to 100–220 °C instead of 300–400 °C, which is suitable for the growth of InP QDs. In addition, volatile stabilizers such as pyridine, tributylphosphine, hexylamine, or trioctylamine were used instead of the heavy coordinating solvents TOPO and TOP. The low TR inhibits the growth of nanoclusters into large QDs and is, therefore, one of the most important conditions to yield small nanoclusters, including MSCs30,39,85. Optical characterization of synthesized ultrasmall InP nanoclusters showed an excitonic absorption peak at λ ~400 nm that is quite narrow compared to the general absorption features of InP QDs.

As common precursors for the synthesis of InP QDs, an indium carboxylate complex and (TMS)3P have been developed since 20096. 386-InP MSCs were observed at 138 ≤ TR ≤ 208 °C during the growth of InP QDs (Fig. 3a). Subsequent SCXRD analysis demonstrated that 386-InP MSCs have the formula In37P20(O2CR)51 (R: alkyl group) (Fig. 3b)45. 386-InP MSCs have a large FWHM of ~40 nm at λ = 386 nm due to vibronic coupling. The absorption peak at 386 nm narrows significantly when the 386-InP MSC solution is cooled to 100 K (Fig. 3c).

Fig. 3: Carboxylate Capped InP MSCs with Persistent Absorption Peak at 386 nm.
figure 3

a UV–vis absorption spectra of 386-InP MSCs formed with different reaction temperature. Adapted with permission from ref. 35. Copyright 2009 American Chemical Society. b Complete single-crystal XRD structure of 386-InP MSC, In37P20(O2CR)51 (R = Phenylacetate) and c variable-temperature UV–vis absorption spectra of 386-InP MSCs in the solid state. Adapted with permission from ref. 45. Copyright 2016 American Chemical Society. d The [In21P20]3+ core of the 386-InP MSC (P: orange, In: purple), e a view of one InP twistane unit (top) and an excerpt of the cadmoindite unit cell showing structural homology (bottom), and f a view of the In13P14 subunit with T-symmetry derived from the 386-InP MSC illustrating three of twelve interlocking and intersecting polytwistane axes. Adapted with permission from ref. 71. Copyright 2018 WILEY VCH. g Ligand binding modes of 386-InP MSC. Of the 51 carboxylate ligands, 12 are identified as bidentate, one is identified as bidentate + dative, five are identified as symmetric bridging, and 33 are identified as asymmetric bridging by SCXRD. Adapted with permission from ref. 45. Copyright 2016 American Chemical Society.

To produce 386-InP MSCs, various synthesis conditions, including TR, the precursor concentration, and the nature of ligands, have been investigated35. Myristate-capped 386-InP MSCs are generally synthesized at 100 °C and are stable up to 208 °C. Survival at this high temperature indicates that 386-InP MSCs are heavily involved in the growth of InP QDs.

Synthesis of 386-InP MSCs requires a concentration of the indium carboxylate complex that is least two times that of (TMS)3P due to the cation-rich composition of MSCs.

Alkyl chain lengths of carboxylate ligands from butanoic acid (C4) to stearic acid (C18) have been used in the synthesis of 386-InP MSCs. When the alkyl chain is short, 386-InP MSCs form quickly but survive only at low temperature; for example, 386-InP MSCs with hexanoic acid (C6) can survive up to 158 °C, whereas 386-InP MSCs with HMy (C12) can survive up to 208 °C. Prior to the formation of 386-InP MSCs, molecular precursors and monomers with short-length ligands may be unstable in organic solvents, which leads to fast nucleation. After the formation of 386-InP MSCs, long alkyl chain ligands on the surface increase the solubility of the MSCs and thereby help stabilize the structure even at high temperature.

Clear identification of the structure of 386-InP MSCs allowed characterization of the details of the inorganic core and surface ligand binding mode45,71. The [In21P20]3+ core is located in the center of In37P20(O2CR)51 with pseudo-C2v symmetry (Fig. 3d). The average In-P bond length in the [In21P20]3+ core is 2.528 Å, whereas that in the zinc-blende lattice is 2.541 Å. Interestingly, the structure of 386-InP MSCs is not zinc-blende or wurtzite. 386-InP MSCs contain numerous interlocking polytwistane units71. A twistane of In5P5 has a twisted-boat conformation (Fig. 3e). An In13P14 polytwistane subunit with T symmetry also occurs in 386-InP MSCs (Fig. 3f). The surface of 386-InP MSCs bears an additional 16 indium atoms bound to surface-exposed phosphorus atoms45. The ligand binding modes of the 51 carboxylates also revealed that 12 are bidentate, one is bidentate dative, five are symmetric bridging, and 33 are asymmetric bridging (Fig. 3g)45.

13C NMR and 1H NMR spectroscopies were used to analyze the equilibrium dynamics of surface ligands. Isotopically labeled In37P20(O213CCH2Ph)51 was investigated by NMR, and the interaction between In37P20(O213CCH2Ph)51 and In(O213CCH2Ph)3 was traced76. The free indium carboxylate in solution is dynamically exchanged with the bound indium carboxylate on the 386-InP MSC, even at room temperature.

Fatty amine ligands have been widely used as surfactants for the synthesis of InP QDs or InP MSCs. When fatty amines such as oleylamine were used along with In(Ac)3 as surfactants/solvents and reacted with (TMS)3P, two kinds of InP MSCs formed (Fig. 4a)12, one with the LEET at λ = 365 nm and another with the transition at λ = 395 nm. During heating to 170 °C, 365 nm InP MSCs began to form at 45 °C; the optimal TR was ~70 °C. Application of the Debye-Scherrer formula to the pXRD pattern determined that the 365-nm InP MSCs were 1.1 nm in size. Further heating from 140 °C to 170 °C yielded the formation of 395 nm InP MSCs (Fig. 4a).

Fig. 4: Oleylamine and octadecylphosphonate capped InP MSCs.
figure 4

a UV–vis absorption of aliquots taken at different temperatures and reaction times during growth of two kinds of amine passivated InP MSCs. Adapted with permission from ref. 12. Copyright 2017 The Royal Society of Chemistry. b UV–vis absorption of aliquots taken during growth of phosphonate passivated InP MSCs, F397-InP MSCs, as heating up to 370 °C. Adapted with permission from ref. 10. Copyright 2015 American Chemical Society.

The addition of a fatty amine with carboxylic acid to the In(Ac)3 solution followed by reaction with (TMS)3P precursors yields various InP MSCs with LEETs at λ > 400 nm35,91. Several species of InP MSCs with amine/carboxylate co-ligands can be synthesized by controlling TR and the amount of amines. When 3–6 equivalents of octylamine relative to the In(My)3 complex were used, InP MSCs with the LEET at λ = 450 nm appeared at 158 ≤ TR ≤ 208 °C35. However, at TR = 160 °C, excess octylamine with the In(My)3 complex destabilized the MSCs and led to continuous growth of InP QDs10.

Adding amines to InP MSCs separately also has an impact91. The addition of as little as one equivalent of amine to a 386-InP MSC solution caused a redshift of the absorption peak from 386 nm to 420 nm. When more than six equivalents of amine were added to the 386-InP MSCs, this redshifted absorption peak at 420 nm was maintained without a further shift.

The redshifted absorption peak was assumed to be a result of a change in the electronic structure caused by etching of the initial 386-InP MSCs. Reaction of amines with 386-InP MSCs caused etching that yielded crystals of molecules composed of phenylacetate and benzylamine, In(O2CH2C5H6)3(H2NCH2C6H5)3.

31P NMR spectroscopy revealed selective etching of InP MSCs, even for a small amount of amine reacting with 386-InP MSCs. Especially for InP MSCs, solution-phase 31P NMR is a useful tool for observing the electronic environment of each P in an MSC. 386-InP MSCs showed phosphorus peaks in the chemical shift range between −190 and −250 ppm10,91. When three equivalents of amine were added to a 386-InP MSC solution, a particular peak underwent a large shift in the 31P NMR spectrum92. The peak corresponds to the phosphorus that is nearest to the most acidic indium that is etched by three equivalents of amine.

Phosphonic acid is a ligand that binds strongly with cations in semiconductor QDs and MSCs. Octadecylphosphonic acid (ODPA) with In(Ac)3 and (TMS)3P were used as InP MSC precursors; their reaction in squalane at 370 °C led to the formation of phosphonate-capped InP MSCs10. ODPA-capped InP MSCs have four distinct absorption peaks at 397, 355, 285, and 238 nm, which correspond to four excitonic transitions (Fig. 4b). The wavelength of the LEET is 397 nm, so they are denoted as F397-InP MSCs. Surprisingly, F397-InP MSCs were stable up to 370 °C for three days due to the strong binding affinity of phosphonate on the surface indium of the MSCs. The pXRD pattern of F397-InP MSCs showed a zinc-blende structure, and PDF data supported the structural difference from 386-InP MSCs, which form polytwistane10.

In summary, carboxylate-passivated 386-InP MSCs, several kinds of amine-passivated InP MSCs, amine/carboxylate co-passivated InP MSCs, and phosphonate-passivated F397-InP MSCs have been synthesized (Table 2). All of the InP MSCs were formed using In(Ac)3 as the indium precursor and (TMS)3P as the phosphorus precursor; different ligands, such as HMy, oleylamine, and ODPA, for the indium complex led to different types of InP MSCs. Different ligand binding affinities to the MSC surface affect the thermal stability of InP MSCs. Among the InP MSCs, only the structure of the 386-InP MSC has been revealed (In37P20(O2CR)51) by SCXRD. 31P NMR spectroscopy was used to reveal the selective etching of 386-InP MSCs by amines, and pXRD and PDF analysis showed the zinc-blende structure of phosphonate-passivated F397-InP MSCs.

Table 2 Kinds of synthesized InP magic-sized clusters.

InP MSCs with incorporated heterogeneous atoms

The incorporation of heterogeneous atoms into semiconductor QDs and MSCs has been widely adopted to modify the optical properties of original nanomaterials or to develop functional materials40,93,94,95,96,97. Heterogeneous atoms can be introduced directly at the beginning of the synthesis reaction or indirectly by ion exchange or doping47,98,99. For MSCs, the number or position of heterogeneous atoms in their structures can be precisely controlled48,53,100,101,102. In InP nanomaterials, zinc, gallium, cadmium, copper, and tin have been introduced as heterogeneous cations; halides and sulfides have been incorporated as heterogeneous anions to increase the brightness and stability of InP nanomaterials103,104,105,106,107,108,109,110. InP MSCs with incorporated heterogeneous atoms have been reported11. Here, we consider Cl and Zn as heterogeneous atoms in InP MSCs.

InP MSCs with incorporated chlorine

Cl has been widely utilized to control the morphology of semiconductor nanomaterials due to the strong binding affinity of Cl to certain facets during growth109. In addition, the introduction of Cl to the surface of semiconductor nanomaterials by posttreatment enables an increase in the carrier mobility in nanomaterial devices such as field-effect transistors and solar cells111.

Two kinds of InP:Cl MSCs were successfully synthesized, which were named F360-InP:Cl MSCs and F399-InP:Cl MSCs according to their LEETs (Fig. 5a, b)11. The “F” in the MSC names categorizes the group of MSCs that have the same absorption spectrum as the family until a single distinct characteristic is identified. MSCs with incorporated heterogeneous atoms may be heterogeneous as a result of their isomers in both the inorganic crystal structures and surface-binding ligand configurations.

Fig. 5: InP MSCs with incorporated chlorine atoms.
figure 5

UV−vis absorption spectra of aliquots taken over time during (a) direct synthesis of F360-InP:Cl MSCs from indium acetate, HMy, InCl3 and (TMS)3P at 50 °C, (b) direct synthesis of F399-InP:Cl MSCs from indium acetate, InCl3, HMy and (TMS)3P at 80 °C, (c) conversion synthesis of F360-InP:Cl MSCs from as-synthesized 386-MA InP MSCs using 56 equivalents of InCl3 at 80 °C, and (d) conversion synthesis of F399-InP:Cl MSCs from F360-InP:Cl MSCs by heating. TA spectra of e 386-InP MSCs, f F360-InP:Cl MSCs, and g F399-InP:Cl MSCs. h Plot for normalized LEET bleach signals of 386-InP MSCs and F399-InP:Cl MSCs with the estimated absorption cross section σabs (cm−1). i Schematic illustration showing direct or conversion synthesis pathway for Cl-incorporated InP MSCs. Adapted with permission from ref. 11. Copyright 2019 NATURE PUBLISHING GROUP.

F360-InP:Cl MSCs and F399-InP:Cl MSCs were synthesized using In(Ac)3, InCl3, HMy, and (TMS)3P as molecular precursors. The fate of these types of InP MSCs is determined by TR and the precursor ratio. F360-InP:Cl MSCs form at low TR ~ 50 °C, whereas F399-InP:Cl MSCs grow at a relatively higher temperature of 80 ≤ TR ≤ 110 °C. In addition, F360-InP:Cl MSCs are optimally formed when the amount of Cl is less than the amount of HMy (In(Ac)3:HMy:InCl3:(TMS)3P = 1:3:0.6:0.5), whereas F399-InP:Cl MSCs are synthesized with the highest yield when the amount of Cl is high (In(Ac)3:HMy:InCl3:(TMS)3P = 1:3:1:0.5).

Interestingly, two InP MSCs were synthesized by the other route, i.e. the conversion process. 386-InP MSCs that did not have Cl were used as well-defined starting scaffolds for conversion synthesis. Reaction of as-synthesized 386-InP MSCs and 56 equivalents of InCl3 yielded F360-InP:Cl MSCs (Fig. 5c). The conversion mechanism of F360-InP:Cl MSCs from 386-InP MSCs was revealed to be etching by HCl in situ generated by the reaction of InCl3 and remaining carboxylic acid. A proton-coupled 31P NMR spectrum showed that the etching process released PH3. The conversion reaction also occurred when HCl was added directly to the 386-InP MSC solution. However, the isosbestic point of the absorption spectrum was not visible during the etching reaction from 386-InP MSCs to F360-InP:Cl MSCs; this absence indicates that the reaction is not one-to-one but involves intermediate species such as fragments of MSCs. The concept of fragmentation of MSCs and fragment reassembly was recently proposed47,112,113.

F399-InP:Cl MSCs were also synthesized by a conversion route (Fig. 5d). When F360-InP:Cl MSCs were heated to a temperature >110 °C, the absorption spectrum of F360-InP:Cl MSCs disappeared completely, and a sharp absorption peak emerged at 399 nm. These changes can be understood to represent thermal fragmentation of F360-InP:Cl MSCs and reassembly from fragments. pXRD and Raman spectra showed that F399-InP:Cl MSCs have a zinc-blende structure, whereas 386-InP MSCs and F360-InP MSCs are polytwistanes; i.e., the structure evolved as conversion proceeded from 386-InP MSCs to F360-InP MSCs to F399-InP MSCs.

MSCs may be expected to have characteristics that are intermediate between those of molecules and nanocrystals. Depending on their sizes or structures, MSCs might have physical and chemical properties that resemble those of a molecular system or a nanocrystal system. For example, the transition from molecular vibrations to nanocrystal phonons was investigated for a series of three CdSe MSCs66. Phonon confinement and molecular vibration differ in how their vibrational peaks depend on temperature. Variable-temperature Raman spectra of three CdSe MSCs, Cd35Se20X30L30, Cd56Se35X42L42, and Cd84Se56X56L56 MSCs (X: benzoate, L: butylamine), revealed a transition of MSCs from molecular characteristics to nanocrystal characteristics.

The transition from a molecular system to a nanocrystal system was investigated for 386-InP MSCs, F360-InP:Cl MSCs and F399-InP:Cl MSCs11. The TA spectra of F399-InP:Cl MSCs were significantly distinct from those of 386-InP MSCs and F360-InP:Cl MSCs (Fig. 5e–g). The TA spectra of 386-InP MSCs and F360-InP:Cl MSCs showed similar spectral features, such as a broad and long-lived photoinduced absorption signal in the lower-energy region of band-edge bleaching around 386 and 360 nm, respectively (Fig. 5e, f). The significantly broad and long-lived signals are attributable to the transition from the lowest excited electronic state to higher excited electronic states, which is typically observed in molecular systems114. The TA spectra of F399-InP:Cl MSCs showed strong band-edge bleaching with a narrow photoinduced absorption signal in the low-energy region (Fig. 5g); this signal is typical of the TA spectral features observed in well-defined QDs and describes the relaxation dynamics of excitons in quantum-confined systems115,116,117. The absorption cross section at the LEET was 14 times larger in F399-InP:Cl MSCs than in the 386-InP MSC counterparts (Fig. 5h). This difference indicates that their different electronic structures lead to a distinct difference in the absorption cross section as well as their optical properties. Information about the size and structure suggests that large F399-InP:Cl MSCs should evolve toward a zinc-blende structure close to the bulk thermodynamic equilibrium that has high symmetry; i.e., evolution of the character from molecule-like to QD-like was observed for the series from 386-InP MSCs to F360-InP:Cl MSCs and F399-InP:Cl MSCs (Fig. 5i).

InP MSCs with incorporated zinc

Zn is important for InP QDs and InP MSCs because bright InP-based QDs typically have Zn-chalcogenide shell layers (e.g., ZnSe, ZnS). Alloy formation at core-shell interfaces (InZnSe; InZnS) and interdiffusion of Zn into InP cores strongly affect the emission properties of QDs104,118,119. Similar to the series of InP:Cl MSCs, InP:Zn MSCs can be synthesized from molecular precursors or by a conversion method11. Three kinds of InP:Zn MSCs, F360-InP:Zn MSCs, F408-InP:Zn MSCs, and F393-InP:Zn MSCs (Fig. 6a, b), were produced by heating a reaction solution containing an In-ODPA complex, zinc stearate (Zn(St)2), and (TMS)3P at TR up to 300 °C. Selective synthesis of pure InP:Zn MSC species can be achieved by controlling TR and the amount of ODPA. F360-InP:Zn MSCs were synthesized at 180 °C using twice the amount of ODPA relative to indium (Fig. 6a), with fixed amounts of the other precursors at In:Zn:(TMS)3P = 1:1:0.5 for all kinds of InP:Zn MSCs. Before crystallization, ODPA has high binding affinity to the indium cation, so a high ODPA content relative to indium slows the reaction rate. The growth temperature was 300 °C for both F408-InP:Zn MSCs and F393-InP:Zn MSCs, but 1.5 times and 1.85 times ODPA relative to indium, respectively, were used. F360-InP:Zn MSCs grew under the slowest reaction conditions.

Fig. 6: InP MSCs with incorporated zinc atoms.
figure 6

UV−vis absorption spectra of aliquots taken during direct synthesis of F360-InP:Zn MSCs from In(Ac)3, ODPA, (TMS)3P and Zn(St)2 precursors at 180 °C and (b) conversion synthesis of F393-InP:Zn MSCs from as-synthesized F408-InP:Zn MSCs at 300 °C. c Steady-state absorption and d PL spectra of F397-ODPA InP MSCs and F393-InP:Zn MSCs. e Normalized TA spectra at 2 ps of F397-ODPA InP MSCs and F393-InP:Zn MSCs. (Inset: normalized TA decay profile of F397-ODPA InP MSCs at 397 nm and of F393-InP:Zn MSCs at 393 nm). f Plot for normalized LEET bleach signals for F397-ODPA InP MSCs and F393-InP:Zn MSCs with the estimated absorption cross section σabs (cm−1). g Schematic illustration showing direct or conversion synthesis pathway for Zn-incorporated InP MSCs. Adapted with permission from ref. 11. Copyright 2019 Nature Publishing Group.

When the TR of the F360-InP:Zn MSC solution was increased to 300 °C, F408-InP:Zn MSCs grew by a conversion mechanism. Further heating at 300 °C for 24 h converted F408-InP:Zn MSCs to F393-InP:Zn MSCs (Fig. 6b). The evolution of the UV–vis absorption spectra was quantized during the conversion of F360-InP:Zn MSCs to F408-InP:Zn MSCs, whereas the evolution of the UV–vis absorption spectra was continuous during the conversion of F408-InP:Zn MSCs to F393-InP:Zn MSCs; the process showed a continuous blueshift from 393 nm to 408 nm. 1H-NMR and FT-IR data revealed the conversion of F408-InP:Zn MSCs to F393-InP:Zn MSCs as the surface ligand was changed. ODPA and stearate were used with In, Zn, and P precursors, so two kinds of ligand could attach to the surface of InP:Zn MSCs. F408-InP:Zn MSCs have both of these ligands, with a 5.9:1 phosphonate:carboxylate ratio. As conversion proceeded, carboxylate ligands were removed from the surface of F408-InP:Zn MSCs, followed by the formation of F393-InP:Zn MSCs, which were passivated by only phosphonate ligands.

Zn incorporation affects several optical spectra of InP MCSs11. Among the series of InP:Zn MSCs, F393-InP:Zn MSCs are solely passivated by phosphonate ligands, so they can be compared with F397-InP MSCs that are capped with phosphonate ligands. In the steady-state absorption spectra, F397-ODPA InP MSCs showed a very narrow LEET at 397 nm with a second electronic transition at 356 nm, whereas F393-InP:Zn MSCs showed a slightly broader and blueshifted LEET at 393 nm with a second transition at 354 nm (Fig. 6c)11. The blueshifted absorption spectrum of F393-InP:Zn MSCs indicates a decreased effective size of the quantum-confined core; i.e. Zn cations might be mostly incorporated at the surface of the MSCs.

In the PL emission spectra, F393-InP:Zn MSCs had a brighter and broader LEET than F397-ODPA InP MSCs (Fig. 6d)11. The emissive features suggest that the incorporated Zn cations facilitate carrier trapping and intensify the trap emission in F393-InP:Zn MSCs.

The TA spectra of F397-ODPA InP MSCs and F393-InP:Zn MSCs are strikingly similar, with clear spectral features that arise from the exciton relaxation dynamics in quantum-confined systems (Fig. 6e)11. F393-InP:Zn MSCs had a broader photoinduced absorption signal in a lower-energy region of band-edge bleaching with a longer-lived TA decay profile compared to F397-ODPA InP MSCs. These changes are consistent with the increase in the density of trap states after the incorporation of Zn. This incorporation could reduce the symmetry of F393-InP:Zn MSCs and thereby reduce the absorption cross section and widen the absorption and emission spectral features (Fig. 6f). To summarize, the series of InP:Zn MSCs undergo conversion processes similar to the series of InP:Cl MSCs (Fig. 6g), and incorporation of Zn reduces the symmetry of InP MSCs to yield trap-induced optical properties.

Applications of InP MSCs

Using InP MSCs as precursors for QDs and understanding the MSC-mediated growth mechanism

Research on intermediates such as MSCs has revealed the growth process of QDs and provides the possibility of controlling intermediate species. Isolated and characterized InP MSCs have been used as synthesis precursors for InP QDs and nonspherical InP NSs. InP QDs have been synthesized using 386-InP MSCs as precursors10,76,120. The size distribution and sharpness of the excitonic absorption peak of InP QDs synthesized from 386-InP MSCs are superior to those of conventional InP QDs synthesized from molecular precursors120.

In the synthesis strategy, the size of InP QDs is determined by controlling the concentration of 386-InP MSC precursors10. The size of the InP QDs increases as the concentration of 386-InP MSC precursors is increased.

Representative MSC-mediated growth mechanisms have been divided into two types: (1) monomer-driven growth and (2) MSC assembly121. Monomer-driven growth can occur by monomer deposition onto MSC seeds, quantized cluster growth, or complete dissolution/renucleation for each MSC population. MSC assembly occurs by attachment among MSCs.

386-InP MSCs showed both of these growth pathways to form QDs depending on the additives and TR71,76. When 386-InP MSCs were reacted with HMy or In(My)3 as additives, the yields of grown InP nanoparticles (NPs) were very different at TR = 130, 150, 200, 250, and 300 °C (Fig. 7a). At TR = 130 °C, only 386-InP MSCs dissolved. Surprisingly, NPs did not grow after the dissolution of MSCs, even if TR was increased. This result indicates that the dissolution of 386-InP MSCs does not provide a monomer reservoir for nucleation (off-path decomposition). At TR = 150 °C, agglomerates that had no particular shape or size formed, and the chemical yield was low (Fig. 7b). The agglomerates might originate from the assembly of 386-InP MSCs or their fragments, and the low chemical yield may be a result of competition with off-path decomposition of 386-InP MSCs. However, TR > 200 °C yielded uniform spherical InP QDs at high yield as a result of complete dissolution/renucleation of 386-InP MSCs (Fig. 7c).

Fig. 7: Mechanism studies on growth from InP MSCs to InP NPs.
figure 7

a Intensity change at 500 nm during the evolution of 386-InP MSC at different temperature. TEM images showing (b) agglomerates synthesized from MSCs at 150 °C and (c) spherical particles synthesized from MSCs at 250 °C. d Proposed competing pathways for MSC conversion to InP QDs dependent on temperature and additives. Adapted with permission from ref. 76. Copyright 2019 American Chemical Society. e In situ UV-vis spectra of the reaction between 386-InP MSC and 23 equivalents of (TMS)3P injected over 1 h at 100 °C. (Inset: TEM image of synthesized InP NPs). f XRD patterns of the 386-InP MSCs, synthesized InP NPs from the reaction between 386-InP MSC and 23 equivalents of (TMS)3P injected over 1 h at 100 °C, annealed synthesized InP NPs and zinc-blende InP reference. g A proposed reaction between the 386-InP MSCs and (TMS)3P. Adapted with permission from ref. 71. Copyright 2018 WILEY VCH.

386-InP MSC conversion to InP QDs depending on TR may be a result of competing pathways (Fig. 7d)76. As TR is increased, 386-InP MSCs convert to monomers and fragments before the growth of InP QDs. The source of InP QDs could be monomers, fragments, and other monomer species generated from fragments. If TR is high (>200 °C) enough to almost produce monomers, then monomer-driven InP QDs grow; otherwise, fragments of 386-InP MSCs participate in growth to produce agglomerates. At low TR, off-path decomposition of 386-InP MSCs is predominant.

The use of additives with MSC precursors that include carboxylate leads to fast dissolution of 386-InP MSCs followed by monomer-driven growth of InP QDs76. When 386-InP MSCs were reacted with (TMS)3P additives at TR < 125 °C, spherical-shaped InP QDs grew by assembly of MSCs. Due to the 386-InP MSC assemblies, the grown InP QDs retain the polytwistane phase of the initial structure (Fig. 7f). Several spectra indicate that (TMS)3P drives the surface decomposition of 386-InP MSCs to yield MSC assemblies (Fig. 7g). 386-InP MSCs do not dissolve entirely, so their core-dominated structural properties could remain.

Using InP MSCs as precursors to achieve diverse morphology InP nanostructures

InP MSCs have also been used as precursors to control the morphology of InP NSs. Branched nanostructures (BNSs) and hyperbranched nanostructures (HBNSs)/dendrimer-like nanostructures (DLNSs) have been synthesized (Fig. 8) using 386-InP MSCs and molecular additives as precursors80. At TR = 300 °C, without additives, 386-InP MSCs grew into spherical InP QDs (Fig. 7c), whereas with In(My)3 or In(Ac)3 molecular additives, 386-InP MSCs grew to form InP BNSs or InP HBNSs/DLNSs (Fig. 8).

Fig. 8: Controlling morphology of InP NPs using InP MSCs as precursors.
figure 8

TEM images of (a, b) InP BNSs, which were synthesized by the reaction of 386-InP MSCs with In(My)3 at 220 °C, c InP HBNSs/DLNSs which were synthesized by the reaction of 386-InP MSCs with In(Ac)3 at 220 °C. Adapted with permission from ref. 80. Copyright 2019 The Royal Society of Chemistry.

To understand how MSCs mediate the shape evolution, the process of InP BNS growth from 386-InP MSCs was investigated. Upon heating, 386-InP MSCs fragmented before further evolution. Fragmentation was significantly impeded by extra indium complexes such as In(My)3 but accelerated by extra fatty acids such as HMy. Co-use of extra In(My)3 and HMy with 386-InP MSCs resulted in InP BNSs (Fig. 8a, b). InP BNSs may grow from 386-InP MSCs when the concentration of monomers released by MSC fragmentation reaches a critical concentration for nucleation; the new nuclei may grow rapidly to form tetrahedral seeds and further to form BNSs while consuming monomers and fragments, possibly via oriented attachment80.

When In(Ac)3 was used instead of In(My)3 as an additive for the reaction with 386-InP MSCs, BNSs aggregated to form HBNSs/DLNSs (Fig. 8c)80. The InP HBNS/DLNS has a three-dimensional morphology with many branches, and its size can be controlled in the range of 20–30 nm by changing the total concentration of precursors. Initially formed BNSs partially capped by short acetate ligands interact with each other via the van der Waals force.

In syntheses of anisotropic chalcogenide NSs, structure-directing agents are commonly used78,79,122. The synthesis of InP BNSs and InP HBNS/DLNSs does not require such structure-directing agents. This difference demonstrates that sophisticated control of the growth pathways offers NSs that cannot be attained using conventional synthetic protocols.

Developing the luminescent property of InP MSCs

Semiconductor MSCs have luminescent properties and therefore have applications in luminous devices such as LEDs82,87. To develop the luminescent property of InP MSCs, amine-passivated InP MSCs with LEET = 365 nm have been used as the core to synthesize shell layers12.

The use of a ZnS shell overcoating to improve the luminescent property of InP MSCs achieved only 4% QY (Fig. 9a)12. The QY after shelling may be low because the HOMO-LUMO levels of 365 nm InP MSCs are positioned at high energies due to the extremely small size, so exciton confinement to the core by a wide-bandgap shell is difficult. When five layers of ZnS were overcoated on 365 nm InP MSCs, a redshifted PL peak of 480 nm emerged from nondetectable band-edge emission (Fig. 9b, black line). The large redshift was explained as exciton movement to the ZnS shell from the large bandgap InP MSCs. When 20 layers of ZnS were deposited, a new emission peak at 560 nm corresponding to interfacial traps appeared without a further increase in the QY (Fig. 9b, red line). The PL lifetime was measured for the emission at 480 nm and 560 nm (Fig. 9c). When five layers of ZnS were coated on the InP MSCs, the PL lifetime was 34.2 ns at 480 nm. When 20 layers of ZnS were used, the PL lifetime was 62.3 ns at 480 nm and 248.5 ns at 560 nm. The extension of the lifetime is consistent with trap emission profiles originating from interfacial traps or the outer surface of the InP/ZnS MSCs.

Fig. 9: Developing the luminescent property of InP MSCs.
figure 9

a Potential energy diagram for InP, ZnS, and ZnSe bulk materials. b UV–vis absorption, PL spectra, and (c) PL lifetime of ZnS coated InP MSCs. Adapted with permission from ref. 12. Copyright 2017 The Royal Society of Chemistry.

Conclusion and perspective

This paper has reviewed the history of InP QD synthesis and research on InP MSCs, which were identified as intermediates during the growth of InP QDs. We classified the research on InP MSCs into two parts: (1) synthesis and characterization and (2) applications. InP MSCs passivated with carboxylate, amine, or phosphonate ligands or that incorporate heterogeneous atoms (Cl or Zn) were synthesized by lowering TR or introducing strongly bound ligands such as phosphonic acid. Among InP MSCs, the structure of 386-InP MSCs has been shown to be In37P20(O2CR)51 with a polytwistane phase. This knowledge of the structure accelerated the study of the intermediate characteristics of MSCs that are at a size near the boundary between molecules and QDs. TA characterization successfully revealed the optical evolution from a unimolecular to colloidal-QD-like character in the InP:Cl MSC or InP:Zn MSC series.

InP MSCs have been used as synthesis precursors for high-quality InP QDs or diverse morphology InP NSs. Elongated InP BNSs, HBNSs/DLNSs, and spherical QDs have been successfully formed from InP MSCs by using MSC-mediated growth mechanisms such as monomer-driven growth or MSC assembly. Future research on InP MSCs could use various kinds of InP MSCs as precursors to further develop homogeneous or functional nanomaterials. Until now, research has been limited to experiments that used 386-InP MSCs as precursors. Various InP MSCs differ in their structure, thermal stability, and ligands; the development of methods such as preservation/dissolution or attachment/detachment in the reaction environment to control InP MSCs might allow the synthesis of high-quality and functional InP NSs. The information gained might enable a further understanding of the nucleation and growth mechanism of III–V compound nanocrystal systems.

Finally, isomerization of II–VI semiconductor MSCs has been recently observed50. Reversible structural isomerization in the MSC system bridges the difference between molecules and nanocrystals. We expect isomerization phenomena in III–V semiconductor MSCs and the existence of a difference between III–V compound semiconductor MSCs and II–VI compound semiconductor MSCs. The two groups have atomically precise structures, so the fundamental nature of the ionic and covalent characters in the lattice could also be clearly defined.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.