Redefining near-unity luminescence in quantum dots with photothermal threshold quantum yield

See allHide authors and affiliations

Science  15 Mar 2019:
Vol. 363, Issue 6432, pp. 1199-1202
DOI: 10.1126/science.aat3803

Superefficient light emission

A challenge to improving synthesis methods for superefficient light-emitting semiconductor nanoparticles is that current analytical methods cannot measure efficiencies above 99%. Hanifi et al. used photothermal deflection spectroscopy to measure very small nonradiative decay components in quantum dot photoluminescence. The method allowed them to tune the synthesis of CdSe/CdS quantum dots so that the external luminescent efficiencies exceeded 99.5%. This is important for applications that require an absolute minimum amount of photon energy to be lost as heat, such as photovoltaic luminescent concentrators.

Science, this issue p. 1199


A variety of optical applications rely on the absorption and reemission of light. The quantum yield of this process often plays an essential role. When the quantum yield deviates from unity by significantly less than 1%, applications such as luminescent concentrators and optical refrigerators become possible. To evaluate such high performance, we develop a measurement technique for luminescence efficiency with sufficient accuracy below one part per thousand. Photothermal threshold quantum yield is based on the quantization of light to minimize overall measurement uncertainty. This technique is used to guide a procedure capable of making ensembles of near-unity emitting cadmium selenide/cadmium sulfide (CdSe/CdS) core-shell quantum dots. We obtain a photothermal threshold quantum yield luminescence efficiency of 99.6 ± 0.2%, indicating nearly complete suppression of nonradiative decay channels.

Photoluminescence (PL), the absorption and reemission of light, is an essential feature of numerous dyes and semiconducting materials. Recent commercial applications include solid-state lighting (1), high-efficiency color displays, and bioimaging (2). Moreover, there are ongoing efforts to commercialize luminescent solar concentrators (LSCs) (3) and spectrum-shifting greenhouses (4). Each of these applications has specific performance requirements, which are assessed by a standard set of measurement techniques. An essential performance metric is the photoluminescence quantum yield (PLQY), which is determined by the competition between radiative relaxation of the excited material and nonradiative losses, typically due to defects. The PLQY is directly related to the electronic and structural quality of materials, with the highest values recorded to date of 99.5 and 99.7%, respectively, in rare earth–doped high-bandgap single crystals (5) and epitaxially deposited thin films (6). However, in many commercial applications relying on PL, nanocrystals have important potential advantages over their molecular or bulk/thin-film counterparts, which include stability; low cost; ability to be selectively placed inside a variety of composites, fluids, polymers, and biological environments; large-area processing compatibility; and absorption and emission tunability. The PLQY for CdSe/CdS, the prototypical core-shell quantum dot in research labs, routinely exceeds 95% (7). But this is insufficient for applications that require an absolute minimum amount of photon energy to be lost as heat. For example, the light concentration of LSCs increases logarithmically as the nonradiative losses of the luminophore vanish (8). Optical refrigeration (9), thermophotovoltaic engines (10), and thermal energy storage in optical cavities (11, 12) all require materials with PLQY ≥ 99% with negligible nonradiative losses (1315).

Inspired by the photon recycling needs of LSCs (13), we aim in this study to reduce the reabsorption losses by keeping the single-pass PLQY as high as possible while retaining a narrow emission linewidth. We achieve this through the growth of a 4- to 11-monolayer CdS shell around a highly monodisperse CdSe core. Our synthesis technique is inspired by the work of Chen et al. (7), with some noteworthy modifications (supplementary text SM1 and SM2). The most notable change is the complete lack of oleylamine, which Chen et al. use to keep a narrow size distribution. We omit amines, as they can aid in the desorption of the Z-type Cd(oleate)2 surface ligand (16). This synthesis strategy helps maintain a high ligand surface coverage, reducing possible surface traps while preserving high radiative efficiency. The resulting PLQY, as measured in an integrating sphere, is plotted in Fig. 1A as a function of CdS shell thickness for different batches of quantum dots grown with similar-sized cores (~3.3 to 3.6 nm). For a particle with a 2-nm-thick shell, the PLQY consistently exceeds 90 ± 2%, and after 4 nm of shell is grown, the PLQY tends to exceed 97 ± 3%. Figure 1B shows standard characterization of representative samples, including the absorbance spectra, the PL spectra, the photoluminescence excitation (PLE) spectra scaled to core absorbance, and transmission electron micrographs. The PL spectrum has a full width at half maximum of 28 to 32 nm for all shown batches of particles. Critical for LSCs, the Stokes ratio increases as the shell thickness increases (17), reaching values of ~100. The wavelength-dependent PLQYs of these samples were measured in an integrating sphere (Fig. 1C) and are indistinguishable within error regardless of excitation wavelength [100 ± 3% and 97 ± 3% for 8 and 10 monolayers (ML), respectively].

Fig. 1 Characterization of highly emissive CdSe/CdS quantum dots.

(A) PLQY for many samples with different shell thicknesses grown with our synthesis procedure. (B) Absorption factor (Abs) (dotted lines), PLE (shaded regions), and PL spectra (solid lines) for 5-, 8-, 10-, and 11-ML CdS shells, respectively. A.u., arbitrary units. (Inset) Representative transmission electron micrographs. Scale bars, 25 nm. (C) Wavelength-resolved PLQY of the 8- and 10-ML samples shown in Fig. 1B. Full characterization (SM4) dataset in figs. S1 to S9.

Once the PLQY is within a couple of percent of unity, existing experimental measurements do not carry the accuracy necessary to provide feedback toward the materials’ quality to further improve synthetic design. The most common and well-characterized PLQY measurement techniques, the relative dye method and the integrating sphere technique, are both radiometry techniques at their core. Both rely on spectral sensitivity calibrations with a spectral radiance transfer standard, introducing at least 2 to 5% uncertainty into the measurement through the uncertainty of the spectral radiance transfer standard itself (17, 18).

In response to these optical sensitivity limitations, we developed a measurement of the PLQY that does not rely on photon flux measurement but rather uses the quantization of light in a process analogous to the photoelectric effect. Instead of measuring the excitation energy threshold for electron emission, this technique measures the excitation energy threshold that results in net positive heat generation. Therefore, we call our technique photothermal threshold quantum yield (PTQY).

In a typical PL event (Fig. 2A), an absorbed photon excites an electron from a ground state into a higher-energy excited state, leaving behind a hole; both charge carriers quickly (approximately picoseconds) relax to the band edges by emitting thermal phonons into the host crystal. At a later time (approximately nanoseconds), the thermalized electron and hole recombine, bringing the material back to its ground state. This final transition can either be radiative or mediated by defects by emitting more heat, i.e., a nonradiative loss. Therefore, the average amount of heat emitted per absorption event is a combination of the intrinsic band-edge thermalization process (often called “blue loss” in the photovoltaic literature) and nonradiative band-to-band relaxation (“nonradiative loss”). Because the blue loss represents heat generated by carrier thermalization, an excitation energy–dependent measurement of the heat per luminescence event allows us to determine the fraction of emitted heat that originates from nonradiative band-to-band relaxation. Determination of the photon energy at which the net emitted heat vanishes (threshold energy, E0) allows us to precisely determine the PLQY (Fig. 2B and SM6).

Fig. 2 Measurement scheme for the PTQY technique.

(A) Excitation light is absorbed by an electron, which moves from a low energy state to a higher energy state. The excited electron and hole emit heat Q to relax to their lowest energy excited states and then can recombine through PL or nonradiative processes (the latter of which emits heat QNR). (B) A plot of the heat emitted per excitation versus the excitation energy shows how the PLQY can be determined from the threshold energy E0 above which heat is emitted and the photoluminescence energy EPL. A near-perfect emitter at the thermal limit (Q + QNR = 0) where the deviation of EPLE0 represents the deviation in PTQY (i.e., PLQY ≤ 1). (C) The instrumental realization of the PTQY technique uses a photothermal deflection spectrometer to measure heat, synchronized to a PL spectrometer, which determines both EPL and the number of excitation events (SM5, SM6, and figs. S17 to S25).

We implemented the PTQY technique (Fig. 2C and SM5) using a transverse photothermal deflection spectrometer (19) to measure the nonradiative heat and a PL spectrometer to measure the above-gap PL excitation spectrum. By measuring the ratio of nonradiative to radiative emission components, this technique enables removal of the spectral radiance dependence in the PLQY measurement (figs. S10 to S16). By extrapolating a thermal limit, the experimental technique relies solely on the measurement of photon energies [i.e., excitation energy (Eexc) and photoluminescence energy (EPL)], values that can be determined considerably more accurately than quantifying spectral radiance. A spectral radiance transfer standard is required only to determine the correct shape of the PL spectrum, bringing an uncertainty limit of 6 parts per million (Fig. 3D). Our apparatus has an experimental uncertainty of 0.02%, limited by the detection spectrometer wavelength accuracy, which was used both to characterize the excitation lamp energy and average emission energies. Typical measurements are further limited by finite averaging, commonly yielding uncertainties of 0.1 to 0.2% (Fig. 3D), whereas the best uncertainty we achieved was 0.04%. Our method has a 100× improvement in uncertainty compared with 4% for relative and integrating sphere measurements (18). Measurement of heat generation can be performed very sensitively; even optically heterogeneous thermally thin samples (<50 μm) present little complication (see SM3 and SM7 and figs. S26 to S33 for details ensuring measurement fidelity).

Fig. 3 Detailed optical characterization of CdSe/CdS quantum dots and PTQY measurement uncertainty.

(A) PTQY of the best CdSe/CdS (core-shell) particles (8.5 ML). The inset displays probability distributions and uncertainties associated with PTQY uncertainty (blue and black shading), and integrating sphere PLQY (Gaussian, red shading). (B) High-resolution PTQY of high-quality CdSe core (96.2 ± 0.1%) but with poorly passivated CdS shell (93.6 ± 0.2%) with inset of respective integrating sphere PLQY and PTQY uncertainty (for detailed uncertainty, see SM6 and figs. S10 to S16). (C) Top plot shows PL lifetimes of CdSe/CdS core and shell excited lifetimes, and bottom plot shows PLQY and PTQY of the same near-unity particles with measurement uncertainty plotted against particle Feret diameter. The circle markers represent a constant CdSe core size and varied CdS shell thickness, and the inset diamond markers represent the highest PTQY particles measured thus far, with a slightly larger 3.6 ± 0.2 nm CdSe core and 3.7 ± 0.3 nm (8.5 ML) CdS shell. (D) Table representing the main sources of uncertainty and the general uncertainty budget for the calibrated PTQY spectrometer dependent on photon energy (figs. S17 to S26). CCD, charge-coupled device. (E) TRPL data (log-log scale) taken at 407.1-nm excitation. Despite near-unity PLQY of 4- to 11-ML samples, significant differences in long lifetime components arise across the shelled series. (F) PL lifetime excitation wavelength dependence with varied shell thickness normalized to each particle’s respective core lifetime.

The sensitivity of PTQY enables us to investigate high-quality CdSe/CdS quantum dots where PLQY surpasses 90% (Figs. 1A and 3C). The highest PTQY value, shown in Fig. 3A, is 99.6 ± 0.2%. This value was measured for excitations originating in the CdS shell of an 8.5-ML (core diameter: 3.6 ± 0.2 nm; shell thickness: 3.7 ± 0.3 nm) CdSe/CdS core-shell sample. This PL efficiency compares favorably with the internal radiative efficiency estimated indirectly in well-passivated GaAs [99.7 ± 0.2% extrapolated from photon recycling (6)] and single-crystal Nd:Y3Al5O12 [99.5 ± 0.2% measured via laser cooling (5)]. One sample, shown in Fig. 3B, exhibited small radiative variations between the shell and core that are barely discernible by traditional methods. In this sample (Fig. 3B, inset), the PLQY of core excitations is 96.2 ± 0.1%, whereas the PLQY when exciting the shell is 93.6 ± 0.2%. We ascribe this decrease to shell excitations naturally leading to more trapping and nonradiative processes (see SM8 for further discussion).

We present a series of particles with the same size CdSe core but with different CdS shell thicknesses (4 to 11 ML). We can systematically determine the origin of the peak in PLQY at 8 ± 0.5 ML using PTQY and time-resolved PL (TRPL) decay measurements. With shell thicknesses less than 8 ML, insufficient passivation, consistent with surface hole traps, is displayed (20). This effect is apparent in TRPL decays as a tail past ~100 ns (Fig. 3E), resulting in significantly lower efficiencies (Fig. 3C). At shell thicknesses ≥8 ML, the excited-state decays are well-described by a biexponential decay, but >8 ML, the excited-state decays have substantially longer excited-state lifetimes. The increasing excited-state lifetime can be understood by considering the quasi–type II band alignment whereby a thicker shell results in reduced overlap between a delocalized electron in the shell and core-localized hole (SM8 and SM9). Despite this extended lifetime, ensembles still display PTQYs of 99.6%, suggesting very slow nonradiative processes, possibly as slow as 2 to 20 μs (figs. S34 to S36). Moreover, excitations of larger shells allow the electron to be more easily trapped at the CdS surface in shallow traps (Fig. 3F). Thus core-shell particles can tolerate shallow surface traps that only affect one charge carrier (21). These results suggest an alternative, defect-tolerant path toward the future development of quantum dots with 99.9% or better quantum yields, in addition to the well-known but challenging approach of eliminating all defects and trap levels in nanocrystalline samples, which necessarily have an enormous surface area. The combination of more-accurate PLQY measurements and increased TRPL dynamic range allows us to better track rare events that occur in quantum dot ensembles at excitation intensities that single particle luminescence experiments cannot achieve (fig. S37). Such rare events under low flux excitation, N1, where N is the average number of excitons created per pulse per particle, are relevant to further optimization of the particles for numerous energy-related applications.

The true limit of colloidal quantum dot luminescence efficiency is currently unknown. Improvements in both process engineering and luminescence efficiency quantification can help drive the design, synthesis, and understanding of optimized quantum dots, opening the door to tests of fundamental theory and applications of these important materials. In this work, we have demonstrated a synthetic method that produces particles with external luminescent efficiencies >99.5% and a technique that is capable of measuring the efficiencies of these ensembles with nearly 100× less uncertainty than traditional measurement techniques. Our results demonstrate that there is no fundamental impediment to synthesize nanocrystals with nonradiative decay channels at least as low as, if not lower than, those of the best single-crystal semiconductor materials, and provide an important platform for future development of near-lossless materials at the thermodynamic limit.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S37

Tables S1 and S2

References (2328)

References and Notes

Acknowledgments: We thank C. J. Takacs, V. Klimov, W. Jackson, R. Prasanna, and B. Guzelturk for valuable advice, discussions, or reading of the manuscript. Funding: This work is part of the ‘Photonics at Thermodynamic Limits’ Energy Frontier Research Center funded by the U.S. Department of Energy (DOE), Office of Science, Office of Basic Energy Sciences, under award DE-SC0019140. Work at the Molecular Foundry by A.M.S. and L.M. (transient absorption measurements) was supported by the DOE, Office of Science, Office of Basic Energy Sciences, under contract DE-AC02-05CH11231. Y.v.d.B. (modeling and analysis) acknowledges funding from the European Research Council under the European Union’s Horizon 2020 research and innovation program (grant 802615). K.T. (synthesis methodology) acknowledges funding from JXTG Nippon Oil & Energy. Author contributions: D.A.H., N.D.B., and B.A.K. contributed equally to this work. D.A.H. and N.D.B. conceived of the idea and directed the team. N.D.B., D.A.H., B.A.K., and Y.v.d.B. co-wrote the manuscript. D.A.H., N.D.B., Y.v.d.B., and K.V. co-designed the PTQY experimental setup. B.A.K., N.D.B., and K.T. designed and developed the new synthetic method for CdSe/CdS nanoparticles used in this manuscript. B.A.K. and Z.N. synthesized the near-unity nanocrystals used in this manuscript. D.A.H., N.D.B., B.A.K., Z.N., J.K.S., A.M.S., L.M., and Y.v.d.B. all contributed to characterization of the samples in this paper. All authors contributed to the discussion and analysis of the results. Competing interests: The authors declare no competing interests. Data and materials availability: All processed data are available in the main text or in the supplementary materials. All software and calibrations datasets are archived in the Dash repository (22).

Stay Connected to Science

Navigate This Article