Main

The challenges facing the semiconductor industry today include the monolithic integration of novel materials into complementary metal–oxide–semiconductor technology, the need for ever-more-rapid growth of high-crystallinity materials and the continuing miniaturization of devices. Alternatives to traditional semiconductor technologies are desired that exhibit tunable electronic and optical properties, and can be used to fabricate low-cost, large-area, flexible devices.

Colloidal quantum dots (CQDs) are solution-processed semiconducting nanocrystals with diameters below 20 nm. They offer a size-tuned bandgap and materials processing compatible with a range of substrates. Since they first emerged in the 1980s (refs. 1,2), CQDs have been created from semiconductors of groups II–VI, IV–VI, I–III–VI2 and III–V. Their solution processing and ease of manufacture has also allowed them to be integrated into electronic devices including lasers3, light-emitting diodes (LEDs)4,5, solar cells6,7 and photodetectors8.

In this Review, we examine the use of CQDs in the development of next-generation electronic devices. We consider the manufacture of CQDs and their deployment in luminescent applications, quantum information processing, sensor technology, memory devices and thermoelectric devices (Fig. 1). We also explore the emerging opportunities for CQD electronics created via materials engineering and identify the challenges that exist on the path towards practical application.

Fig. 1: CQD-based electronics.
figure 1

By harnessing their distinctive electronic, optical, physical and structural properties, the community deploys CQDs in a variety of electronic and optoelectronic devices.

Full size image

CQDs to thin-film semiconductors

CQDs are commonly synthesized using wet chemical methods such as hot injection. Control over the CQD size and composition is achieved by tuning the precursor type and concentration, reaction time and temperature during synthesis (Fig. 2a)9. When the size of CQDs is of the same order of magnitude as the semiconductor’s Bohr radius, spatial confinement of the electron–hole pairs leads to discrete energy levels that shift to higher energies with decreasing size10. The semiconductor core can further be enclosed within a shell of another semiconductor, forming a heterojunction that modifies the wavefunctions of electrons and holes in CQDs. As a result of the synthetic and size flexibility, the bandgap of CQDs, and therefore their absorption and emission spectra, can be tuned over a wide spectral range from ultraviolet to visible to infrared. Size and shape uniformity of CQDs is required for high-quality ensembles that provide a flat energy landscape. Precursor stoichiometry control is critical for obtaining CQDs with narrow absorption linewidth and small Stokes shift11. A recently developed cation-exchange synthesis converted ZnS nanorods to highly monodisperse low-bandgap PbS nanocrystals, showing control over particle size12. Size-selective precipitation is another facile approach to separating polydisperse nanoparticles into fractions of narrower size distributions9.

Fig. 2: Colloidal quantum dots.
figure 2

a, The solution-processed synthesis resulting in colloidal nanocrystal inks. b, Assembly of CQDs. Left: CQDs can be processed from the solution phase in densely packed thin films using solution-processed deposition techniques. Middle: sintering of CQD-constituent films results in polycrystalline solids. Right: self-assembly of CQDs with highly monodispersed size and shape leads to long-range ordered superlattices.

Full size image

The surface structure of CQDs plays a key role in their electronic13 and vibrational14 structures. As-synthesized CQDs are passivated with a layer of electronically isolating molecules whose functional groups interact directly with CQD surfaces. These ligands prevent CQDs from aggregating in solvents but hinder the carrier transport from one dot to another. Replacement of original capping ligands with shorter conductive ones allows for improved carrier mobility, tunable electronic coupling and the systematic manipulation of energy levels15,16. The highly flexible surface structure renders CQDs promising building blocks for a range of electronic applications.

CQDs can be processed from the solution phase in densely packed thin films by spin coating, spray coating, roll-to-roll printing, inkjet printing and transfer printing (Fig. 2b, left). The ease and versatility of CQD manufacturing are particularly advantageous for scale-up. Inkjet printing yields desired patterns without physical masks, rendering it a convenient technology towards full-colour display. However, the spatial resolution of inkjet printing—around 20 to 30 μm—is limited by the droplet size and solvent drying dynamics17. An alternative electrohydrodynamic printing technique employed electric field to create the fluid flows delivering inks to a substrate18. This method allowed for quantum-dot light-emitting diodes (QLEDs) and photodetectors with reduced feature size down to 1 μm (ref. 19). Solvent-free transfer printing technologies utilize soft elastomeric stamps to transfer prepared CQD solids. Desired patterns can be achieved with the aid of a structured stamp20 or intaglio trench21, resulting in well ordered structures and clearly defined interfaces beneficial to the design of full-colour displays. However, the usage of a soft stamp limits printing fidelity, and the gradual degradation of the stamp restricts reproducibility in producing defect-free pixel arrays. A newly developed immersion transfer printing methodology addresses this challenge by using a silicon trench template to pattern CQD arrays and a thermodynamic-driven adhesion switching mechanism to securely detach the patterned film22. This led to full-colour CQD arrays with sub-500-nm pattern resolution.

Long-range ordered superlattices built up with CQDs as artificial atoms can be achieved by self-assembly (Fig. 2b, right)23. When CQDs have highly monodispersed size and shape, slow evaporation of solvent from the colloidal dispersion allows for the growth of superstructures. By tuning the energy levels of constituent dots, interparticle coupling and structural arrangement, the electronic structure of the superlattice is widely modified24,25,26.

Integrating CQDs into a host medium enables greater tunability of electronic properties. CQD solids normally have a packing density between 64% (random close packing) and 74% (hexagonal/face-centred cubic close packing). The voids or gaps among CQDs are well suited for the diffusion of gaseous precursors used in an atomic layer deposition process. This allows semiconductor oxides to infiltrate pores among CQDs, protecting the surface from oxidation and forming a heterojunction that improves carrier transport27,28. Alternatively, CQDs can be codeposited with a second phase by using a sol–gel method or solution-processed approach. For example, CQD–polymer composites29 have been achieved with wide emission tunability and stability. Lattice-matched perovskite matrix grown epitaxially around CQDs led to enhanced surface passivation and improved carrier mobility, and it stabilized CQD solids against agglomeration and oxidation under elevated temperatures30,31. This materials system has been employed in multiple electronic applications, yielding high-performance solar cells30,32, LEDs33 and photodetectors34.

Luminescent electronic devices

In CQDs, the three-dimensional confinement of charge carriers imparts a strong overlap between electron and hole wavefunctions, contributing to a high photoluminescence quantum yield (PLQY). This—in addition to the broad spectral tunability and narrow emission linewidth—makes CQDs candidates for luminescent electronic devices, such as lasers, LEDs and solar concentrators35,36. So far, CdSe-based core–shell heterostructures are the material of choice for luminescence in the visible range, owing to their narrow and symmetric emission spectral lineshape37 and near-unity PLQY38. These include both quasi-type-II heterostructures such as CdSe/CdS, where the hole is confined to the core while the electron is delocalized throughout the entire volume, and type I heterostructures such as CdSe/ZnSe and CdSe/ZnS, in which both the hole and electron are confined to the core.

To achieve lasing, population inversion is required. This corresponds to the situation in which more electrons are in the excited state than the ground state, and this is the condition for stimulated emission and optical gain (Fig. 3a)3.

Fig. 3: The design and engineering of luminescent devices.
figure 3

a, Carrier dynamics in CQDs. Top: CQDs can be excited by photons with energies larger than that of the CQD bandgap. The excited electron leaves a hole in the valence band. The electron and hole each relax to the lowest energy states by a non-radiative process (carrier cooling), then recombine spontaneously. Top left: radiative recombination results in the emission of a photon. Top right: non-radiative decay occurs when the electron is trapped by an in-gap state; this is followed by a trap-to-band recombination. Middle left: a biexciton forms when two photons are absorbed in an individual CQD. Middle right: multiexciton generation happens when the absorbed photons are at least twice the CQD bandgap and the carrier cooling energy excites the generation of extra electron–hole pairs. The biexciton or multiexciton generation is necessary to achieve population inversion in CQD solids. Bottom: when the population inversion is established, photon absorption is inhibited. Bottom left: optical gain occurs when an incident photon induces the decay of an excited electron and leads to the emission of another photon with the same phase and frequency. Bottom right: the Auger process is a non-radiative loss mechanism that consumes the energy of electron–hole recombination by pumping a third carrier to a remote energy state. b, Strategies to suppress Auger recombination and reduce gain threshold. c, Schematic of the device structure of a CQD laser. d, Schematic of the working principle of QLED devices. HTL, hole-transporting layer. ETL, electron-transporting layer.

Full size image

For CQD lasing, the non-unity degeneracy of band-edge states leads to high optical gain threshold, demanding intense excitation to achieve population inversion. This, in turn, aggravates the non-radiative Auger recombination, whereby the energy released in electron–hole recombination is transferred to a third carrier, exciting it to a higher energy level. The ultrafast Auger recombination of biexcitons limits the optical gain lifetime to subnanosecond and increases the gain threshold. Auger recombination also accounts for the photoluminescence (PL) blinking of single dots—random switching of PL between a dark charged state and a bright neutral state39.

Auger recombination can be substantially suppressed by delocalizing the wavefunction and softening the confinement potential40,41. For quasi-type-II core–shell structures, this can be achieved by manipulating the size and shape of CQDs (Fig. 3b)40. Well designed thick inorganic shells are key to suppressing blinking42,43. In type I CQDs, compositional gradient shells have been used to smoothen the interfacial confinement potential, resulting in extended optical gain lifetime and reduced gain threshold41,44. Enhancing the geometry-dependent dielectric screening by coating CQDs with a high-dielectric-constant medium has been recently proposed as another feasible strategy for Auger suppression45.

Efforts to reduce the gain threshold have initially focused on using type II heterostructures, given their spatially separated electrons and holes46. However, the weak electron–hole overlap in type II structure led to reduced emission rate and PLQY. An alternative approach is to grow an asymmetric shell that creates biaxial strain in CQDs47. The biaxial strain causes additional band splitting, concentrating the carriers into the lowest energy levels and decreasing the level of degeneracy48. The biaxially strained CQDs therefore exhibit ultranarrow emission linewidth at a single-dot level. Negatively charged CQDs have also shown reduced gain threshold due to the suppression of ground-state absorption49.

Rapid progress on Cd-based CQDs has led to the realization of continuous-wave lasing with optical pumping (Fig. 3c)48 and electrically pumped optical gain in the visible range44, showing promise on the path to true CQD lasing with electrical injection. This, in parallel with ongoing efforts to demonstrate electrically pumped organic semiconductor lasers50, highlights an important milestone towards future laser diode technology. Low-threshold near-infrared optical gain has also been achieved in HgTe CQDs, paving the way for infrared CQD lasers51.

In addition to lasers, efforts have been devoted to QLEDs and have led to impressive external quantum efficiencies (EQEs) of beyond 20% (refs. 38,52,53). These values approach the theoretical maximum set mainly by the light outcoupling efficiency limit (~20% for a glass substrate with a refractive index of 1.5) and suggest internal quantum efficiencies nearing 100% (ref. 54). One key challenge facing the commercialization of QLED displays is to achieve simultaneously high brightness, high EQE and long operating lifetimes (lifetime of >10,000 h is required for displays). This requires sufficient excitonic confinement in CQDs, high carrier mobility and judiciously engineered device architecture that favours fast balanced charge injection.

The prototypical QLED devices consist of a thin CQD layer, sandwiched between an electron-transporting layer and a hole-transporting layer (Fig. 3d). Typically, CQDs with wide-bandgap shells are used to achieve high PLQY, yet create a large barrier for charge injection, particularly for holes. Inadequate and unbalanced charge injection resulted in charge accumulation near the interface, increasing Auger recombination and decreasing device lifetime55.

Precise control over CQD synthesis has led to CQDs exhibiting a favourable band alignment for hole injection while retaining a high PLQY. This was achieved by growing a narrower-bandgap shell with strain-relaxed core–shell interface38,56. The optimized QLEDs showed extended operating lifetime meeting the requirement for display application57. Replacing isolating surface ligands with conductive ones is another approach to improve carrier mobility and balance the charge distribution58. This leads to QLED devices with reduced EQE roll-off at high injection current density. Inserting an insulating layer between the CQDs and the electron-transporting layer has also been successful in maintaining charge balance and preventing CQD charging59. Heavy-metal-free QLEDs have recently achieved outstanding EQE comparable to state-of-the-art Cd-based devices60; this marks an important step toward the commercialization of QLEDs. High-efficiency infrared LEDs were also achieved using PbS CQD emitters; this has been accomplished by use of a carrier transport matrix33 or ternary CQD blends61.

Inorganic caesium lead halide perovskite CQDs have recently emerged as a new class of luminescent materials62. These CQDs exhibited high PLQYs, reaching 90% over a wide spectral range (400–700 nm), and narrow emission linewidths of 70–100 meV (ref. 63). Room-temperature optical gain has been achieved with low thresholds in the entire visible range64. Progress on surface passivation and ligand replacement65,66 has also led to red and green perovskite QLEDs with EQE above 20% (refs. 67,68). If long-term stability, charge injection and the achievement of efficient blue emission69 can be tackled, perovskite CQDs offer promise in wide-colour-gamut displays.

Single-photon emitters

The discrete electronic structure of CQDs coupled with high quantum yields has long inspired the prospect of employing them as coherent single-photon emitters for quantum information processing. Their facile and scalable synthesis and ease of integration into optical cavities present advantages over commonly studied solid-state coherent single-photon emitters, such as colour centres in diamond70 and epitaxially grown quantum dots71. For quantum information processing, the optical coherence time (T2) should approach twice the spontaneous emission lifetime (T1), that is transform-limited coherence with T2 = 2T1. However, CQDs typically suffer from strong electron–phonon coupling, which shortens the coherence time72,73,74,75. This, combined with their typically long emission lifetime—a result of weak optical coupling of the lowest-energy optical transition76—results in T2 ≪ 2T1 for most CQD systems. In addition, spectral diffusion—an environmentally induced fluctuation in the emission energy of single CQDs—and PL blinking both lead to further optical decoherence and inhibit their application in quantum optics77,78,79.

Mechanisms to suppress Auger recombination, and to improve PL stability, are needed in the field. One recent report demonstrated Cd-based core–shell CQDs with non-fluctuating emission energy and stable intensity47. Perovskite CQDs have also been intensively studied with this aim. Their luminescence properties are governed by the formation and radiative recombination of band-edge excitons, whose fine structure remains a topic of debate80,81,82. Perovskite CQDs exhibit a remarkably short radiative lifetime and long optical coherence time, which is coupled with strongly suppressed PL blinking and spectral diffusion83,84; however, these still suffer from limited stability at ambient conditions. A recent study demonstrated T2/2T1 ratios of ~0.2 in CsPbBr3 CQDs at cryogenic temperatures, rivalling those measured in epitaxial quantum dots85. Further control over surface structures to reduce electron–phonon coupling and increase coherence time86 and development of encapsulation strategies to improve long-term stability87,88 are promising avenues toward their use as quantum light sources.

CQD-based photodetectors and sensors

Present-day photodetection and sensing technology rely primarily on crystalline inorganic semiconductors such as silicon and III–V and II–VI compounds. Emerging applications such as self-driving cars, security systems and medical monitoring require new miniaturized detection tools to carry out rapid and real-time analysis. CQDs, thanks to their tunable optical absorption and low-temperature solution processing, allow for monolithic integration with front-end readout electronics and therefore have been appreciated as a low-cost complementary metal–oxide–semiconductor-compatible alternative to conventional photoactive materials (Fig. 4a). Of particular interest is the development of infrared CQD photodetectors that compensate silicon by extending the sensitization wavelength beyond 1,100 nm (cutoff wavelength of silicon). Following the early demonstration of infrared photodetection and photovoltaic effect from CQD–polymer hybrid materials29, rapid progress has been made in both CQD-based photodetectors and solar cells89,90,91.

Fig. 4: Broadly tunable absorption enables efficient optoelectronic devices.
figure 4

a, Solar spectral irradiance and the absorption ranges of different semiconductors. b, Schematic of typical CQD-based phototransistor devices. c, Schematic of photovoltage field-effect transistors. d, Schematic of CQD photodiodes. TCO, transparent conducting oxide. e, Schematic of a hybrid photodetector integrating a CQD photodiode with a graphene phototransistor.

Full size image

CQD-based photodetectors have been developed with three basic device geometries: photoconductor, phototransistor and photodiode. Photoconductor devices work on the principle of circulating one type of carrier under the influence of an external electrical field until they recombine with the opposite carriers, which remain trapped in the gap states. The photoconductive gain is proportional to the ratio of the minority carrier lifetime to the majority carrier transit time. It can be considerably leveraged by introducing trap states that capture minority carriers and prolong carrier lifetime92, or by enhancing the trapping efficiency through multiexciton-generation induced photoionization93. Surface ligand exchange plays an important role in influencing trap state densities and carrier mobility94. Controllable surface treatment and oxidation of CQDs has delivered responsivities greater than 103 A W−1 and impressive detectivities of up to 2 × 1013 Jones at 1.3 μm at room temperature8.

CQD-based phototransistors, governed by the same gain mechanism, use an additional electrode (gate electrode), along with the source–drain contacts, to tune the majority carrier concentration and shift the Fermi level (Fig. 4b). This provides for a higher degree of control over carrier mobility and, thus, the transit time. A variety of inorganic ligands, represented by metal chalcogenide complexes such as Sn2S64− and CdTe22−, have been used to provide strong electronic coupling in CQD solids15. Upon thermal treatment, metal chalcogenide complex ligands form a crystalline inorganic phase that bridges CQDs, enabling solid-state field-effect transistors with high electron mobility of above 10 cm2 V−1 s−1 (ref. 95). The integration of CQDs with high-conductivity channel materials, such as two-dimensional materials96 and amorphous-oxide semiconductors97, has been exploited to compensate for limited carrier transport in CQDs. In these geometries, photogenerated electrons were ejected from CQDs and circulated through the channel, resulting in considerably reduced transit time and high gain. A remarkably high responsivity approaching 109 A W−1 was achieved in a graphene–PbS CQDs hybrid structure98. Wide-bandgap transparent oxides were adopted for Cd-based visible-sensitive CQDs99. The demonstrated ultrasensitive photoresponse, with detectivity above 8 × 1013 Jones, opens new avenues for multispectral photodetection in the visible spectrum.

However, for photoconductor and phototransistor devices, the prolonged carrier lifetime—necessary for a high gain—limits their temporal response. By contrast, photodiodes relying on a p–n or p–i–n junction, where photogenerated electrons and holes are extracted by an internal built-in electric field and collected by two electrodes, have usually fast response but low gain (EQE is smaller than unity unless the avalanche effect or carrier multiplication is exploited). They can be operated as photodetectors under reverse bias, or as solar cells with zero bias (Fig. 4d). The carrier transport—via diffusion and drift—is limited by the modest carrier mobility and curtailed by the high trap-state density. Noise current, a key determinant of detectivity in CQD photodetectors, arises due to low-frequency flicker noise (1/f), generation–recombination noise, thermal noise and shot noise. Flicker noise—which dominates at low frequency—can arise from mobility fluctuation and carrier number fluctuation, while generation–recombination noise originates from photocarrier trapping and detrapping processes. The noise current is affected by the architecture and temperature of photodetectors, the doping levels and the trap state density of CQD solids.

The development of printable CQD inks, based on a solution-phase ligand exchange method, has substantially advanced carrier transport100. This method allowed for the single-step deposition of densely packed CQD films with reduced energetic heterogeneity101, and it enabled stable mixtures that contained CQDs with different surface ligands. When packed into ensembles, the mixed CQDs yielded donor–acceptor domains facilitating the charge separation102. The combined benefits delivered CQD solar cells of up to 12.5% power conversion efficiency and opened up new avenues for photodetection. The exchanged CQDs can be embedded into an epitaxially grown perovskite matrix, which features high mobility. Photocarriers tunnelled into the matrix under high electric field and yielded photoresponse extending from visible to the short-wave infrared range, with superior detectivity beyond 4 × 1012 Jones (ref. 34). The improved surface passivation provided by the compact perovskite shell further enhance the stability of CQDs and associated devices30.

New architectures were designed to tackle the trade-off between photoconductive gain and bandwidth. A hybrid architecture integrating a CQD photodiode with a graphene transistor has been explored (Fig. 4e)103. The device can be operated as both photodiode and phototransistor, achieving high gain through the graphene channel and fast response driven by the photodiode operation. In addtion, a photojunction field-effect transistor has been demonstrated by combining a CQD photogate with a silicon field-effect transistor (Fig. 4c)104. A photovoltage arose at the Si:CQD interface upon light absorption, modulating the width of the depletion region and the carrier concentration within the silicon channel. This results in a gain–bandwidth product of the order of 109 Hz—a figure of merit outperforming that of conventional photodetectors.

Mid-wave and long-wave infrared photodetectors are of growing interest for thermal imaging of room-temperature objects. Progress in the synthesis of narrow-bandgap semiconductor (such as HgTe) CQDs has led to steadily tunable interband absorption from the near-infrared to long-wave infrared region105,106. Photon management strategies using interference or plasmonic structures enhanced the infrared absorption and photon collection efficiency in HgTe CQDs, resulting in a detectivity greater than 1011 Jones at 5 µm at cryogenic temperature107. A dual-band infrared photodetector was developed using two rectifying HgTe CQD photodiodes, each sensitive to different infrared wavelengths, oriented in a back-to-back configuration108. This results in a bias-switchable spectral response in two distinct bands.

Intraband transitions occurring in the first levels of the conduction band109,110,111 and plasmonic transitions112,113, each observed in degenerately doped CQDs, extend the absorption into the terahertz region and thus provide additional access to broadband photodetection. HgSe CQDs are intrinsically n doped and exhibit a low-energy intraband photoresponse. Introducing HgSe CQDs into a high-mobility HgTe CQD matrix, to create a band structure where the 1Se and 1Pe states of HgSe are located within the bandgap of HgTe CQDs, combines the advantages of each material, resulting in a fast mid-infrared photoresponse114.

In addition to photodetectors, CQDs can be used as sensor materials for the detection of gas, pressure and temperature. CQDs have a large number of surface dangling bonds, hence showing high reactivities with several chemical species at room temperature. This allows doping and functionalization of CQDs when they are exposed to the target gas. Depending on the specific surface chemistry, a variety of CQDs (such as PbS, PbSe and SnO2) have been employed for the detection of gases including NO2, H2S, CH4 and NH3115,116,117. The strain- and temperature-dependent optical properties of CQDs make them promising for new sensor devices118,119. Pressure induces changes in both the CQD bandgap and the dynamics of the surrounding matrix material, leading to shifts in PL emission wavelength, intensity and decay rate118. Future studies are expected to provide detailed analysis and interpretation of the diverse behaviour of CQDs under pressure and temperature and the effect of surrounding environment on their optical response.

Non-volatile memory devices

Non-volatile memories are playing an important role in the semiconductor market thanks in particular to flash, which is widely used in portable electronic devices. Future electronic systems demand non-volatile memories in nanoscale dimension, with high throughput and fast speed to access data.

The floating-gate transistor is the basic building block for flash memory. Conventional floating-gate memories employ a floating-gate layer, surrounded completely by a dielectric layer, to store a bit by the presence or absence of a charge. The dielectric provides tunnelling barriers for memory operations to prevent lateral charge losses. The conventional structure, which employs continuous films, is approaching the physical scaling limitation: further downscaling requires thinner tunnelling oxide, which is risky for current leakage and storage information loss.

To address this challenge, a discrete charge-trapping floating gate was designed using metallic or semiconducting nanocrystals embedded in a charge-tunnelling matrix. The surface ligands or shell components can be selected and engineered to play the role of a charge-tunnelling layer120. This contributed to discrete charge-trapping centres without the aid of an extra tunnelling matrix. For instance, the use of a type I ZnSe/ZnS CQD floating gate in a graphene transistor led to a large memory window and good endurance (Fig. 5a)120.

Fig. 5: CQD-based memory devices.
figure 5

a, Schematic of charge-trapping, retaining and detrapping processes in CQD-based floating-gate memory devices. HOMO, highest occupied molecular orbital. LUMO, lowest unoccupied molecular orbital. b, Schematic of operational principle of CQD-based PCM devices, exploiting Joule heating to reach local crystallization or melting of PCM material. c, GeTe nanoparticles as solution-processable PCM materials128. GeTe nanoparticles exhibit amorphous structure, narrow size distribution, and size-dependent crystallization temperature and crystallization kinetics. Horizonal error bars for GeTe diameter indicate standard deviation of size distributions, measured by TEM; vertical error bars for crystallization temperature stem from linear extrapolation fits of Bragg reflection intensity profiles, derived from high-temperature X-ray diffraction. Dashed curve in the right panel is a guide for the eye.

Full size image

The use of a CQD-based floating gate lends additional benefits to light-erasable memory devices, where photoactive channel materials are used121,122: the photoresponse of CQDs upon optical illumination facilitates the photon-induced recovery, resulting in a faster erasing time. A pentacene field-effect transistor using a perfluorinated-thiol-capped CdSe CQD floating gate was recently reported123. The device exhibited stable switching behaviour, high memory ratios (ratio of drift current in ON/OFF states) over 105, and fast light-induced erasing within 1 s; this was partially due to the careful selection of surface ligands, which enhanced hole diffusion in CQDs and modulated the energy level at CQD/pentacene interfaces.

While continued research on floating-gate techniques should extend the current capability of flash technology, increasing interest lies in new mechanisms and materials. Among emerging non-volatile memories, phase-change memory (PCM) is a promising technology. PCM relies on this working principle (Fig. 5b): a memory cell is locally heated with a voltage pulse to reach either the crystallization or the melting point of the material124. This transforms the material between its crystalline (low-resistance, logical unit) and amorphous (high-resistance, logical zero) states, which are retained upon cooling to room temperature. Materials exhibiting the basic requirements for PCM are found among the metal chalcogenides. Comparing with conventional silicon-based flash memory, PCMs are notably faster in reading and writing, and show longer endurance especially at elevated temperatures (that is >100 °C)125.

CQDs hold particular promise to overcome the scaling challenges present in conventional top-down manufacturing processes. Despite the wide compositional range of CQDs, until recently colloidal synthesis of known phase-change materials had not been achieved, as a result of their complex composition and non-equilibrium stoichiometry. To date, GeTe CQDs are the only candidate that has been successfully synthesized: both crystalline and amorphous GeTe CQDs with a wide range of size have been reported126,127,128.

In the nanoscale regime, however, most phase-change properties deviate from those of the bulk, including materials structure129, phase-transition temperature and kinetics. Downscaling leads to an increase in crystallization temperature and a decrease in melting temperature, which can be explained by the negative entropy change upon crystallization, as opposed to melting-point depression. For amorphous GeTe CQDs having a diameter of 4.5 nm, the crystallization temperature reached 253 °C, more than 80 °C higher than that of bulk GeTe128. This suggested enhanced amorphous phase stability and a narrowed operation window for GeTe CQDs (Fig. 5c). The investigation of size-dependent phase-transition kinetics, with the help of CQDs, is of great importance130.

Advances in CQD-based PCM devices are currently in the preliminary stage. The first CQD-based PCM device was reported in 2011, with the employment of drop-casted GeTe CQDs131. The device exhibited improved performance, distinct electrical contrast, and promising switching endurance of up to 100 cycles. Though the size of the device was still on the micrometre scale, this proof-of-concept demonstration presented the potential of CQDs for high-performance, nanoscale PCM devices.

Although the binary CQD presents the essential characteristics of a phase-change material, the exploration of ternary and more complex composition systems is required to systematically tune the phase-change properties and stability. Among prospective future research directions are to expand a library of phase-change CQDs, to develop effective ligand removal approaches, and to reach ultrasmall device dimensions with solution-based deposition methods.

Thermoelectric devices

Thermal energy harvesting plays a prominent role in the energy ecosystem as a result of the immense amounts of waste heat ejected into the environment. Thermoelectric devices manage thermal energy by either converting heat into electricity or providing active cooling under current flow. The performance of a thermoelectric material is characterized by a dimensionless figure of merit, ZT = S2σT/κ, where S, σ, T and κ are the Seebeck coefficient, electric conductivity, absolute temperature and thermal conductivity, respectively.

Electrons and phonons are both capable of transporting heat, and the overall thermal conductivity of a material can be written as a sum of the two contributions, κ = κe + κl. The electronic contribution, κe, is directly proportional to the electronic conductivity as described by the Weidmann–Franz law. For bulk single-crystalline materials, the contribution from phonon transport, κl, is linked to κe, a result of the joint dependence of the electronic and phononic properties on the symmetry and chemistry of the crystal lattice. Nanostructuring of a material can, however, alleviate this constraint, decoupling a material’s thermal and electronic properties132,133,134. This enables the rational optimization of ZT through independent optimization of electronic and thermal transport and as a result renders CQD solids promising candidates for thermoelectric applications135.

Thermal transport in CQD solids proceeds through a combination of interfacial phonon-vibration scattering at the CQD–ligand interface136,137,138 and propagation of acoustic phonons of the CQD superlattice (Fig. 6a)139,140. Thermal conductivities in CQD solids can be tuned and are typically orders of magnitude lower than in bulk semiconductors at room temperature135,137. High CQD–ligand interfacial densities can be achieved by reducing the size of CQDs141, imposing limits to phonon mean free paths and constraining thermal transport to the timescales of interfacial scattering138: timescales that can be engineered through ligand chemistry via binding group and vibrational structure137,142. The phonon band structure of superlattices can be tailored through the size of the CQDs, their periodical arrangement, and the interparticle mechanical interactions presented by the ligand. This contributes to decreased phonon group velocities and phononic bandgaps, which hinder thermal transport in the CQD solids143. Enhancing the electronic conductivity of CQD solids utilizes many of the same tuning degrees of freedom, such as CQD size, ligands and ordering, and additionally depends on the free-carrier concentration, which can be increased by both substitutional and remote doping144. Optimization of this set of interrelated design parameters can therefore be used to boost ZT within CQD solids.

Fig. 6: Design of CQD-based thermoelectric devices.
figure 6

a, Schematic illustration of phonon transport and scattering in CQD solids at an atomic and superlattice level. b, Schematic of the assembly of CQDs for thermoelectric applications. The solution-processed CQD solids are annealed and pressed into compact, bulk-like pellets, which can be assembled into a thermoelectric generator. c, Room-temperature thermal conductivity of a nanograined CdSe film as a function of the underlying quantum-dot size141. Horizontal error bars indicate the standard deviation of size distributions; vertical error bars indicate the uncertainty in thermal conductivity, derived from the uncertainty in measurable input parameters. d, High-angle annular dark-field scanning transmission electron microscope image of a PbS–Ag nanocomposite. Panel c reproduced with permission from ref. 141, ACS. Panel d adapted from ref. 150, under a Creative Commons licence CC BY 4.0.

Full size image

CQD solid thermoelectric devices are limited to operation at low temperatures to avoid sintering of the CQDs. The use of CQDs provides a promising playing field, particularly towards flexible thermoelectrics135,145,146 or low-ΔT applications, such as wearable electronics or thermoelectric sensing.

Moving beyond these applications, CQDs can be used as building blocks or additives in the fabrication of polycrystalline materials for high-T thermoelectric devices (Fig. 6b). In polycrystalline materials, κl is proportional to the phonon mean free path, which is affected by phonon scattering at defect centres, grain boundaries and interfaces and with other phonons. Embedding quantum dots into bulk thermoelectric materials as scattering sites for phonons is an effective approach to decrease κl (refs. 147,148). Bottom-up fabrication of polycrystalline CQD-derived solids, on the other hand, results in highly controllable crystalline domains given by the size of the original CQDs, reducing considerably the phonon mean free path length due to the scattering from grain boundaries (Fig. 6c)141,149,150,151. By using CQD sizes of the order of the mean free paths of charge carriers, typically tens of nanometres, these materials can have κ reduced by orders of magnitude with σ similar to those in their single-crystal counterparts135. Moreover, the use of doped CQDs152,153 or inclusion of metallic nanoparticles150 modulates charge carrier concentrations in the resulting polycrystalline solids, offering additional benefits for σ and increasing ZT values (Fig. 6d).

Outlook

CQD-based electronics has progressed due to advances in materials and device optimization. The engineering of chemical and electrical properties has led to state-of-the-art performance metrics for a range of applications, including light emission, lasing and photodetection. The emergence of new classes of CQDs also paves the way for single-photon sources, thermoelectrics and PCM devices. Further improvement in device performance and stability requires continued advances in materials design, enhanced control over hybrid interfaces, and improved understanding of underlying properties.

A key area for future development is the fabrication of monodispersed CQDs with greater compositional and structural complexity. CQDs with asymmetrical strain have achieved minimized spectral diffusion in single-dot emission47. Taking this concept further, carefully designed asymmetrical CQD-in-rod or rod-in-rod heterostructures could provide interesting emission properties obtained by tailoring the electron and hole wavefunctions154. Coupled CQD molecules can be also envisioned to provide greater flexibility to tailor the potential energy landscape and tune the coupling strength155.

Novel nano- and mesoscale CQD assemblies open up new possibilities for electronic devices. The self-assembly of multicomponent CQD arrays using two or more different types of CQD has revealed a diverse library of superstructures. Further exploration of packing principles and growth kinetics underlying the formation of superlattices could lead to patterns and geometries with targeted functionalities24,26,156. Hybrid materials combining CQDs with other semiconductors that exhibit exceptional optoelectronic properties and overcome limitations in single-material systems, such as CQD-in-perovskite solids, have also emerged. Colloidal systems consisting of nanocrystals in molten inorganic salts have been developed, where the colloids are stabilized when chemical bonds are formed between nanoparticles and solvent ions157. This suggests that there are opportunities to introduce CQDs into a variety of inorganic hosts.

These developments, and further advances in device performance, require robust characterization methodologies that probe charge transport and energy transfer across heterointerfaces. Advanced spectroscopies over a broad spectral range—from terahertz to ultraviolet—can be used to elucidate optical, electronic and phononic properties at single-entity and ensemble level. Such detailed studies could provide quantitative information on charge and exciton dynamics, uncover the mechanisms influencing ultrafast processes and guide the modification of materials158,159.