Introduction

Phosphors, which can convert the colour of photons through sequential absorption and emission, have been actively researched over the years1. In particular, the emergence of phosphor-capped white light-emitting diodes (LEDs)2,3 has enabled and boosted important application sectors, such as solid-state lighting and modern full-colour displays, and thus reinforced the demand for phosphors. The overall performance of a phosphor can be represented by its external quantum efficiency (EQE), expressed as \({\eta }_{{ext}}=A\times {\eta }_{{{{{\mathrm{int}}}}}}\times {\eta }_{{oc}}\), where A, ηint, and ηoc are the absorbance (of excitation photons by the phosphor), internal quantum efficiency (in converting the absorbed excitation photons to emissive photons of different wavelengths inside the phosphor), and out-coupling efficiency (in extracting the emitted photons from the phosphor), respectively. Thus far, phosphors have been developed primarily excavating new materials with higher ηint, which culminated in the synthesis4,5,6 and adoption7 of colloidal quantum dots (CQDs). Currently, progress in the material-oriented phosphor development process has been limited. As a bypass to the stalemate, attention to the structural aspects of phosphors was given to improve the phosphor performance via A or ηoc. Examples include incorporating high-index scattering centres, such as nano-rods8, a mesoporous film9, or metal-oxide nanoparticles10,11,12, and adding a dielectric distributed Bragg reflector (DBR) with its centre wavelength tuned to either excitation13,14 or down-converted emission14. However, the bottom-up approaches in the former are too complex to analyse or design systematically, whereas the top-down approaches in the latter do not have scope for further development.

We previously proposed and demonstrated the concept of structurally engineered phosphors15,16,17,18, in which the phosphor material was carved into a nanophotonic structure to significantly increase A (while ηint and ηoc remained unaltered). More specifically, we incorporated CQDs, which served as the phosphor material, into a prefabricated planar photonic crystal (PhC) backbone such that the Γ-point (\({k}_{\parallel }=0\)) band-edge mode(s) associated with the resultant PhC structure was tuned at the wavelength for CQD excitation. The PhC phosphor thus prepared exhibited greatly enhanced CQD fluorescence because of the resonant absorption of the vertically incident excitation photons. Notably, the concept of structural engineering (for enhanced A) is not bound to any specific phosphor material and is thus compatible with the traditional efforts of developing new phosphor materials (for higher ηint).

In this Article, we propose simple one-dimensional (1D) resonant cavity (RC), which comprises a central cavity layer and two DBR mirrors on both sides, as yet another platform for structurally engineered phosphors. The RC structure, although old-fashioned, has recently been revaluated as a topological photonic system with functional robustness19. Cavity quantum electrodynamics suggests that the optical transition rates of a quantum system embedded in an RC can be significantly altered. The spontaneous emission rate, for example, is known to be modified by the available photonic density of states (PDOS) in the environment20, implying that radiation from an electric dipole inside an RC is either enhanced or suppressed depending on the fulfilment of resonant conditions21; RC-LEDs with improved electroluminescence properties are the most representative photonic devices of this type22. Similar effects are expected for the induced optical transition (stimulated emission or absorption) as well because the associated optical transition rate is still proportional to the PDOS23, albeit it also depends on the number of existing radiation quanta (photons). Consequently, the induced transition rate can be significantly enhanced by placing dipoles at the antinode(s) of an RC mode where the PDOS is at its maximum. Relevant photonic devices include vertical-cavity surface-emitting lasers (VCSELs)24, RC photodetectors, and RC photovoltaic cells25,26,27,28. Nonetheless, RC has never been explored or considered as a platform for phosphors. The RC phosphor (RC-PSP) proposed herein differs from the PhC phosphor previously developed by us15,16,17,18; it utilises a localised vertical cavity mode whereas the PhC phosphor relies on a periodic and extended in-plane band-edge mode. Furthermore, the RC-PSP is superior to the PhC phosphor in terms of both structure and performance. The RC-PSP can be constructed by vertically stacking planar layers one after another, which is far easier than the sophisticated lateral submicron-patterning process required for fabricating the PhC phosphor. As for the performance, the absorption of excitation photons is greater in the RC-PSP than in the PhC phosphor as revealed hereafter.

Results

Structural aspects of RC phosphor

Figure 1a shows the schematic of a typical colour pixel configuration used in modern full-colour display29, in which a phosphor layer absorbs shorter-wavelength excitation photons from a micro-LED and converts them into longer-wavelength photons. For phosphor materials, chemically synthesised monodisperse CQDs are the most preferred candidates owing to their various advantages, including convenient colour tuning, high quantum efficiency, broad absorption bands, narrow emission linewidths, and fast modulation response30. CQDs are typically dispersed in a polymer (30–40 wt%), which is then cast into a thick ( ~10 μm) phosphor film31. Despite the large amount of CQDs involved, a substantial fraction of excitation photons passes through the CQD phosphor layer without being absorbed, deteriorating the colour purity that CQD fluorescence could otherwise offer. To eliminate the ‘blue’ leakage, a pigment-based absorbing colour filter with a thickness of approximately 1 μm is often inserted29,32, as shown in Fig. 1a. However, the additional blue filter layer makes the overall pixel configuration bulkier and complicates the fabrication. In this context, a desirable form factor for phosphors should be a thin film containing costly CQDs as few as possible yet capable of absorbing most excitation photons, which would obviate the need for an extra colour filter to absorb blue. We claim that embedding a thin CQD film inside the RC can help accomplish this seemingly impossible task.

Fig. 1: Structure of the RC-PSP.
figure 1

a, b Schematics of the colour pixel configurations based on a thick-film conventional phosphor (Con-PSP; with no RC structure) (a) and the resonant cavity phosphor (RC-PSP) (b). The inset in b illustrates structural details of the RC-PSP. c, d Cross-sectional TEM images of the RC-PSP (c) and αRC-PSP (d). Calculated modal profiles set in the corresponding RC structures are drawn over the images. The prefix ‘α’ stands for ‘augmenting’ or ‘augmented’.

Full size image

Our RC-PSP has an asymmetric 1D RC structure, as shown in Fig. 1b, in which an λ/2-thick cavity layer is sandwiched between two dielectric DBRs: the emission-side DBR (emDBR; adjacent to the substrate) and the excitation-side DBR (exDBR; adjacent to the air). The asymmetry in the cavity was intentionally incorporated to facilitate efficient excitation through exDBR. The overall RC-PSP structure was designed such that cavity resonance occurred at λ0 = 450 nm, assuming a blue LED or laser diode (LD) as a convenient and compact phosphor excitation source. The cavity layer comprises a CdSe-based dense CQD film (dCQD ≈ 40 nm, nCQD ≈ 1.82) and two SiO2 wing layers (dw ≈ 40 nm) on both sides. The two wing layers save costly CQD materials near the nodal planes in the RC mode where the electric field strength is weak. The two DBRs consist of alternating λ/4-thick dielectric layers of TiO2 (dTiO ≈ 49 nm, nTiO ≈ 2.32) and SiO2 (dSiO ≈ 77 nm, nSiO ≈ 1.47), starting and ending with the TiO2 layers for the higher index contrasts to the environmental materials, air and silica. It should be noted that, compared to the conventional phosphor (Con-PSP) structure shown in Fig. 1a, the RC-PSP structure is only ~1/10 of its physical thickness and contains only ~1/100 of the net CQD amount31.

Enhanced absorption of excitation photons by RC phosphor

The transmittance (T), reflectance (R), and absorbance (A) spectra of the RC-PSP were calculated for various combinations of layer numbers in the two DBRs (Supplementary Information S1). The extinction coefficient of the CQD film determined from independent spectroscopic ellipsometry measurements was used in the calculations (Supplementary Information S2). The peak absorbance and resonance linewidth for the RC-PSP were closely coupled and dependent on DBR layer numbers. Therefore, the appropriate RC structure that best fits given excitation requirements—for example, by an LED or LD—must be determined. This study chose the DBR layer numbers as Nex = 5 (or 2.5 pairs) for exDBR and Nem = 13 (or 6.5 pairs) for emDBR because the resulting RC-PSP structure offers high values for all three different absorption parameters—the peak, integrated, and weighted absorbances—with a relatively small number of DBR layers \({N}_{{ex}}+{N}_{{em}}\) (Supplementary Information S3). As shown in Fig. 2a(ii), the RC-PSP exhibits a peak absorbance Amax ≈ 0.87 and a resonance linewidth Δλ0 ≈ 7 nm in the full-width at half-maximum (FWHM); the corresponding cold cavity, having the same RC structure but no absorbing ability, results in a quality factor Q ≈ 90. Conversely, the reference phosphor (Ref-PSP), which corresponds to the cavity layer itself—a 40-nm-thick CQD layer cladded by the two SiO2 wing layers on both sides, exhibits a fairly flat spectral profile with no resonance feature, and the absorbance at λ0 = 450 nm is only A ≈ 0.03—Fig. 2a(i). The absorption enhancement factor (AEF), defined as the absorbance ratio between RC-PSP and Ref-PSP, was deduced as a function of wavelength, as shown in Fig. 2d. The resonantly enhanced absorption, which is a hallmark feature of RC-PSP, was as large as AEFmax ≈ 32 at resonance.

Fig. 2: Characteristics of the RC-PSP and αRC-PSP.
figure 2

a Calculated transmittance (T), reflectance (R), and absorbance (A) of (i) the reference phosphor (Ref-PSP), (ii) RC-PSP, and (iii) αRC-PSP. The reflectance spectrum of the αDBR for the αRC-PSP is also shown in (iii). b Photoluminescence excitation (PLE) data for (i) the Ref-PSP, (ii) RC-PSP, and (iii) αRC-PSP, all measured with excitation linewidth δλex = 2 nm. Shown as inset in (i) is the schematic of the PLE measurement setup. (Xe: xenon lamp, MC: monochromator, IS: integrating sphere, SM: spectrometer, L: lens, MF: multimode fiber, B: baffle, S: sample). c Colour photographs captured from the top while excited resonantly (λex = λ0 ≈ 468 nm) for (i) the Ref-PSP, (ii) RC-PSP, and (iii) αRC-PSP. In each, the left panel was taken with a low-pass filter (LPF) ( > 575 nm) inserted, whereas the right panel was taken with a short-pass filter (SPF) ( < 550 nm). d Absorption enhancement factor (AEF) spectra for the RC-PSP and αRC-PSP, deduced from the calculated absorbance spectra in a. e Fluorescence enhancement factor (FEF) spectra for the RC-PSP and αRC-PSP, deduced from the experimental PLE data in b.

Full size image

Experimental implementation of RC phosphor

The fabrication of RC-PSP is similar to that of an optically excited CQD VCSEL33. First, the emDBR and one of the wing layers were sequentially vacuum-deposited on a fused quartz substrate. The CdSe-ZnS core-shell CQDs emitting at λR ≈ 620 nm were, after dispersed in hexane, spin-coated to form a 40-nm-thick film on top of the already deposited wing layer. Another round of sequential vacuum deposition for the other wing layer and exDBR completed device fabrication. Figure 1c shows a cross-sectional transmission electron microscopy (TEM) image of the RC-PSP fabricated as described. For comparison, Ref-PSP that lacked of the both DBRs was prepared on a separate fused silica substrate. T and R spectra were measured for both, while A was deduced from the measured T and R using the relationship A = 1 − T − R (Supplementary Information S4). All the experimental results agree well with the calculated ones in Fig. 2a(i) and 2a(ii), except that the RC resonance is spectrally shifted to λ0 ≈ 468 nm and the linewidth is broadened to Δλ0 ≈ 13 nm due to inhomogeneous broadening.

Photoluminescence excitation measurements

Photoluminescence excitation (PLE) measurements were performed to characterise RC-PSP. The inset of Fig. 2b(i) schematically depicts the PLE setup, in which a custom-built tunable excitation source based on a Xe lamp was combined with an integrating sphere, which was employed to capture all the photons emanating from the emission-side hemisphere through the fused silica substrate. The PLE data measured with excitation linewidth δλex = 2 nm are plotted in Fig. 2b(i) and 2b(ii) for Ref-PSP and RC-PSP, respectively. As shown in Fig. 2b(ii), the CQD fluorescence from the RC-PSP is resonantly enhanced at λex ≈ 468 nm. The resonant excitation wavelength matched perfectly with the cavity resonance wavelength λ0 identified from the independently measured T and R spectra, indicating that the enhanced CQD fluorescence was a direct consequence of the resonant absorption of the excitation photons by the RC. In contrast, Ref-PSP exhibited neither resonance nor enhancement of CQD fluorescence—Fig. 2b(i). The fluorescence enhancement factor (FEF) is shown as a function of excitation wavelength in Fig. 2e. It is defined as the CQD fluorescence intensity ratio between the RC-PSP and Ref-PSP, and can be deduced experimentally from Fig. 2b(i) and 2b(ii). Note that FEF is roughly equivalent to the EQE ratio, \({\eta }_{{ext}}^{{RC}}/{\eta }_{{ext}}^{{Ref}}\), as ηint and ηoc are not altered by the presence of RC. The FEF reached its maximum value FEFmax ≈ 29 at λex = λ0 ≈ 468 nm.

Performance-augmented RC phosphor

The performance of the RC-PSP can be further improved by a modified RC structure. As a proof of concept, an augmenting DBR (αDBR) composed of 2.5 pairs of TiO2 (dTiO ≈ 68 nm) and SiO2 (dSiO ≈ 104 nm) layers was added on top of the exDBR to reflect the CQD fluorescence emanating from the excitation side back to the emission side13,14. The calculated T, R, and A spectra of the resultant augmented RC-PSP (αRC-PSP), along with the calculated reflectance spectrum of the αDBR, which exhibits R ≈ 0.8 at λR ≈ 620 nm, are shown in Fig. 2a(iii). Note that the overall resonance features near λ0 ≈ 450 nm are unchanged—albeit Amax is lowered and Δλ0 is broadened due to partial reduction in the cavity quality by the addition of the αDBR. However, linewidth broadening is not necessarily a disadvantage because phosphors are typically excited by a light source with a broad bandwidth, such as an LED, for which an RC with a comparable resonance linewidth is preferable to maximise the use of excitation photons. The AEF of the αRC-PSP deduced from Fig. 2a(i) and 2a(iii) is also presented in Fig. 2d.

The αRC-PSP was also experimentally implemented. Its cross-sectional TEM image is shown in Fig. 1d. The T, R, and A spectra measured for the αRC-PSP (Supplementary Information S4) are in excellent agreement with the calculated spectra in Fig. 2a(iii). PLE measurements were repeated for the αRC-PSP; the results are plotted in Fig. 2b(iii). Again, the CQD fluorescence intensity was resonantly enhanced. The FEF spectrum of αRC-PSP, deduced from Fig. 2b(i) and 2b(iii), is shown in Fig. 2e. Despite the lower maximum absorbance in Fig. 2d, the αRC-PSP exhibits better fluorescence performance than the RC-PSP in terms of both maximum FEF (FEFmax ≈ 33) and linewidth (Δλ0 ≈ 26 nm) due to the presence of the αDBR. Further ingenious structural modifications to the overall RC could improve the performance.

Visual demonstration

For direct visual demonstrations, three types of CQD phosphors—the Ref-PSP, RC-PSP, and αRC-PSP—were photographed under excitation at various wavelengths (Supplementary Information S5). Figure 2c shows selected photographs under resonant excitation conditions at λex = λ0 ≈ 468 nm. The left panel for each PSP was captured with a long-pass filter (LPF; λ > 575 nm) inserted in front of the camera to capture only the red CQD fluorescence. The CQD fluorescence intensities of RC-PSP and αRC-PSP significantly exceeded that of Ref-PSP. For quantitative comparisons, both the peak and integrated fluorescence intensities for Ref-PSP, RC-PSP, and αRC-PSP are extracted from the PLE data in Fig. 2b and summarized in Table 1. The photographs in the right panels were captured with a short-pass filter (SPF; λ < 550 nm) so that only the blue leakage was visible. While Ref-PSP exhibited an intense blue leakage (due to low A and high T), RC-PSP and αRC-PSP only showed a dim blue leakage (due to high A and low T). All these visual observations completely agree with the T and A spectra in Fig. 2a and the PLE data in Fig. 2b.

Table 1 Peak and integrated fluorescence intensities for Ref-PSP, RC-PSP, and αRC-PSP
Full size table

Red and green RC phosphors for display application

Although efficient white light generation is certainly a possibility, the most obvious and impactful application sector for RC-PSP should be the development of full-colour displays based on hybrid micro-LED arrays, in which independently addressable red-green-blue (RGB) pixels can be prepared by combining blue LEDs with red and green phosphors29, as depicted in Fig. 3a. It should be noted that the RC-PSP structures for the red and green pixels are nominally identical because their excitation sources and thus their excitation wavelengths are identical. However, to realise both red and green pixels from a common RC-PSP platform, the index contrast between the constituent DBR layers must be reduced. Otherwise, the stopband of the emDBR may extend to the green spectral region, making it difficult for the green CQD fluorescence to escape from the RC-PSP structure. Hence, the high-index DBR material TiO2 was replaced with Ta2O5 (nTaO ≈ 2.08), while the numbers of DBR layers were increased to Nex = 7 (3.5 pairs) and Nem = 31 (15.5 pairs) to compensate for the lowered index contrast. A batch of green CQDs with a peak fluorescence wavelength λG ≈ 530 nm was prepared in addition to the red CQDs used in the concept demonstration.

Fig. 3: RC-PSPs composed of Ta2O5/SiO2 DBRs.
figure 3

a Schematic of the RC-PSP based RGB colour pixels for full-colour displays. The combinations of the blue excitation sources and the RC-PSPs constitute the red and green pixels, while the blue excitation source itself serves as the blue pixel. b, c Experimentally determined FEF spectrum of the red RC-PSP (b) and green RC-PSP (c). The reflectance spectrum of the emDBR that constitutes the RC and the intrinsic CQD fluorescence spectrum (with no RC) are also shown in each figure. The excitation linewidth during the PLE measurements (and subsequent FEF determinations) was set at δλex = 20 nm.

Full size image

In Fig. 3b and c, the fluorescence spectra measured directly from the red and green CQDs are compared with the measured reflectance spectrum of the Ta2O5/SiO2 emDBR. As intended, the red and green CQD fluorescence peaks were outside the emDBR stopband, ensuring high transmittance of the red and green CQD fluorescence through the emDBR. PLE measurements were performed on the red and green RC-PSPs consisting of Ta2O5/SiO2 DBRs (Supplementary Information S6). Unlike the previous PLE measurements, the excitation linewidth was intentionally broadened to δλex ≈ 20 nm to mimic an LED-like realistic excitation source34,35. As shown in Fig. 3b and c, the FEFs are as high as ~6.8 and ~5.9 for the red and green RC-PSPs, respectively. Compared with the results shown in Fig. 2e, which were obtained with δλex ≈ 2 nm, the maximum FEFs are much lower, while the resonance linewidths become larger (Δλ0 ≈ 20 nm).

Colour rendering

Figure 4a shows the CIE 1931 colour space chromaticity diagram, in which the RGB colour coordinates rendered by the red and green RC-PSPs (under resonant conditions) and the blue excitation source were identified. The resulting colour triangle defines the colour gamut covered by our RGB sources. For comparison, two standard colour spaces used in the display industry, sRGB and NTSC, are also shown. Regarding the colour space area, our colour triangle covered ~121% of the sRGB and ~86% of the NTSC. The emission spectra of the corresponding RGB chromatic sources, which are shown in Fig. 4b–d, are dominated by distinct red, green, and blue emission peaks. The inset in each figure shows a photograph captured through an open output port on the integrating sphere, which vividly displays the colour purity rendered by each source. Nonetheless, Fig. 4b and c display small but finite blue peaks resulting from unabsorbed residual excitation photons, estimated to be approximately 4% and 13% of the red and green peaks, respectively, in terms of integrated intensity. Further reduction or complete elimination of the blue leakage is a subject for future study. In contrast, the Ref-PSPs containing identical amounts of CQDs allowed the majority ( ~90%) of blue excitation photons to pass through, resulting in the red and green pixels completely overwhelmed by the blue hue, as shown in Fig. 4e–g (and the inset photographs). The resultant chromaticity triangle, plotted in the inset of Fig. 4a, is unrecognisably small. It should be emphasised that all the spectra and photographs shown in Fig. 4b–g were captured without any type of colour filter.

Fig. 4: Colour rendering by the RC-PSPs.
figure 4

a Colour coordinates (and the corresponding colour triangles) rendered by the resonantly excited red and green RC-PSPs and the blue excitation source, plotted on the CIE 1931 colour space chromaticity diagram. Two colour triangles of the industrial standards, sRGB and NTSC, are also plotted for comparison. The colour triangle rendered by the Ref-PSPs, which is tightly localised in the blue corner, is magnified in the inset. b–d Total emission spectra captured by the integrating sphere for the red RC-PSP (b), green RC-PSP (c), and blue excitation source (d). e–g Total emission spectra captured by the integrating sphere for the red Ref-PSP (e), green Ref-PSP (f), and blue excitation source (g). Photographs of an open output port on the integrating sphere, taken for the corresponding pixel configurations, are shown as insets in b–g.

Full size image

Discussion

We investigated a simple 1D RC as yet another platform for structurally engineered phosphors. The RC was designed such that its resonance wavelength could be tuned for the excitation of—rather than for the emission from—the phosphor material inserted in the centre of the cavity. We demonstrated that owing to the significantly enhanced light-matter interaction at the antinode of the RC mode, a 40-nm-thick CQD phosphor layer was sufficient to absorb most of the excitation photons, resulting in a CQD fluorescence enhancement of ~29 times compared with the reference without RC. We also demonstrated that further improvements are possible by ingeniously modifying the RC structure. To assess the applicability of the RC-PSP to full-colour displays, we experimentally improvised and characterised RGB pixels based on the RC-PSP. The resultant colour gamut attained with only ~1/100 of the CQD amount used in Con-PSP-based full-colour display pixels was comparable to, and therefore competitive with, industrial standards. The RC-PSP proposed and demonstrated in this study has a simple planar geometry that enables precise design for tailored performance and facilitates fabrication. Therefore, the RC-PSP platform not only offers superior performance but is also readily applicable owing to its simplicity in structure.

Methods

Calculations of transmittance, reflectance, and absorbance spectra

Although analytical techniques exist for dealing with RC with complex active layers36, numerical simulations based on the finite-difference time-domain (FDTD) method were performed to imitate real experiments as closely as possible, using commercial software (Ansys Lumerical FDTD, ANSYS, Inc.). The complex refractive indices for the dielectric layers (SiO2, TiO2, and Ta2O5) and red/green CQD layers were determined from independent ellipsometry measurements and used in the simulations, while the refractive index of the fused quartz substrate was obtained from a handbook37. Because of the simple planar geometry of the RC-PSP, two-dimensional (2D) simulations were deemed sufficient. Plane-wave sources were used to mimic real excitation conditions, where the excitation photons were incident in the direction perpendicular to the phosphor plane. Planar monitors for transmittance and reflectance were placed after the RC (beyond the emDBR) and before the RC (in front of the exDBR), respectively, and a 2D box monitor encircling the CQD layer was used for absorbance.

Device fabrication

Device fabrication involved three simple steps: two vacuum depositions and one spin-coating in between. First, the TiO2/SiO2 (or Ta2O5/SiO2) emDBR and the first SiO2 wing layer were deposited on a fused quartz substrate using an e-gun evaporator at an elevated temperature of T = 150 °C. During the deposition of the high-index layers, an O2 environment was provided to fully oxidise the metal ions. Subsequently, commercially procured CdSe-ZnS core-shell CQDs (CZO-620H and CZO-530H, Zeus) were dispersed in a cyclohexane solution and spin-coated on top of the first SiO2 wing layer. The CQD concentration (2 wt%) and spin speed (3000 rpm) were carefully selected to obtain the desired CQD film thickness (40 nm). Finally, another round of vacuum deposition of the second SiO2 wing layer and exDBR was performed to complete the device fabrication. The first and last layers of both emDBR and exDBR was a high-index layer (TiO2 or Ta2O5). For αRC-PSP, exDBR and αDBR were simultaneously deposited in a single run.

Measurements of transmittance and reflectance spectra

A halogen illuminator system with a fibre-based light guide (FOK-100W, Fiber Optic Korea) was employed as a convenient white-light source for the entire visible spectral range. The diverging beam from the output tip of the fibre was then transformed into a parallel beam with a diameter of ~1 mm using a combination of a collimating lens and iris diaphragm. To measure both transmittance and reflectance at an incidence angle of 0°, a non-polarising beam-splitting cube with a 50:50 ratio was inserted in front of the sample. The transmitting and reflecting beams were coupled independently to the multimode fibres and sequentially fed into a spectrometer (HR4000CG-UV-NIR, Ocean Optics). The recorded spectra were then normalised using the reference spectra obtained either without the sample for transmittance or with a high-reflection (R > 0.99) broadband (λ = 400–700 nm) dielectric mirror for reflectance.

Photoluminescence excitation measurements

An excitation source with a wide wavelength-tuning range was constructed by combining a Xe lamp (6271 Xenon Arc Lamp, Newport) and a monochromator (CM110, Spectral Products). The excitation linewidth δλex was adjusted over an FWHM range of 2–20 nm using the monochromator slit width. The excitation beam was guided to the sample located at the input port of the integrating sphere using a multimode fibre. The CQD fluorescence spectra from the sample were measured in transmission geometry while the excitation wavelength was scanned. During the measurements, the excitation beam was incident on the exDBR whereas the CQD fluorescence emanating from the other side (through the emDBR and fused quartz substrate) was collected using an integrating sphere (819D-SF-4, Newport). Care was taken to ensure that the sample was at the same level as the input port of the integrating sphere so that only the CQD fluorescence emitted across the upper hemisphere could be captured, whereas the CQD fluorescence emitted across the lower hemisphere and guided through the silica substrate was excluded. The CQD fluorescence collected by the integrating sphere was fed into a spectrometer (Kymera 193i-A with an iVac 316 CCD, ANDOR) using a multimode fibre or photographed through an open output port.