Electrically driven amplified spontaneous emission from colloidal quantum dots

Abstract

Colloidal quantum dots (QDs) are attractive materials for realizing solution-processable laser diodes that could benefit from size-controlled emission wavelengths, low optical-gain thresholds and ease of integration with photonic and electronic circuits^1-7. However, the implementation of such devices has been hampered by fast Auger recombination of gain-active multicarrier states^1,8, poor stability of QD films at high current densities^9,10 and the difficulty to obtain net optical gain in a complex device stack wherein a thin electroluminescent QD layer is combined with optically lossy charge-conducting layers^11-13. Here we resolve these challenges and achieve amplified spontaneous emission (ASE) from electrically pumped colloidal QDs. The developed devices use compact, continuously graded QDs with suppressed Auger recombination incorporated into a pulsed, high-current-density charge-injection structure supplemented by a low-loss photonic waveguide. These colloidal QD ASE diodes exhibit strong, broadband optical gain and demonstrate bright edge emission with instantaneous power of up to 170 μW.

Fig. 1. Optical and EL properties of ccg-QDs.

a, The ground-state absorption spectrum (α₀) of the CdSe/Cd_1−xZn_xSe/ZnSe_0.5S_0.5/ZnS ccg-QDs (top-right inset). Vertical arrows mark the three lowest-energy transitions involving 1S and 1P electron and hole states (shown in the top-left inset). b, Transient absorption (TA) measurements of the ccg-QD solution sample conducted using 110-fs, 2.4-eV pump pulses with per-pulse fluence w_p = 1.6 mJ cm⁻² (⟨N⟩ = 25). The TA signal is presented as α(hv,t) = α₀(hv) + Δα(hv,t), where α₀ and α are the absorption coefficients of the unexcited and excited sample, respectively, and Δα is the pump-induced absorption change. The solid black line (α = 0) separates the regions of absorption (α > 0, brown) and optical gain (α < 0; green). The dashed black line is the second derivative of α₀ (panel a). c, Pump-intensity-dependent spectra of edge-emitted photoluminescence (PL) of a 300-nm-thick ccg-QD film on a glass substrate under excitation with 110-fs, 3.6-eV pump pulses. The pump spot is shaped as a narrow 1.7-mm-long stripe orthogonal to the sample edge. The emergence of narrow peaks at 1.93 eV and 2.08 eV (full width at half maximum 35 meV and 40 meV, respectively) at higher ⟨N⟩ indicates the transition to the ASE regime. On the basis of the onset of sharp intensity growth (inset), the 1S and 1P ASE thresholds are, respectively, about 1 and about 3 excitons per dot on average. d, A device stack of the reference LED comprises an L-ITO cathode, a ccg-QD layer and TFB/HAT-CN hole transport/injection layers separated by a LiF spacer with a current-focusing aperture. The device is completed with a Ag anode prepared as a narrow strip. e, The j–V (solid black line) and EL intensity–V (dashed blue line) dependences of the reference device. f, The j-dependent EL spectra of front (surface) emission of the reference device. The EL spectrum recorded at 1,019 A cm⁻² is deconvolved into three Lorentzian bands that correspond to the three ccg-QD transitions shown in a. AU, arbitrary units.

Fig. 2. Guided optical modes in the reference and BRW devices.

a,b Left, cross-sectional structure of the reference (a) and BRW (b) devices, along with the computed distribution of the TE modes (shown as red-shaded profiles). In the reference device, this mode is supported by TIR, whose critical angle (θ_c) is controlled by the refractive-index contrast at the L-ITO–glass interface (sinθ_c = n_glass/n_L-ITO). In the BRW device, the mode angle (θ_m = θ_BRW) is defined by the condition of constructive interference (Bragg condition) of reflections from different layers of the DBR. As a result, the optical-field profile exhibits an oscillatory pattern linked to the periodic structure of the DBR. Right, dependence of front-emitted and edge-emitted light intensities (yellow and red symbols, respectively) on current density for the reference (a) and BRW (b) devices. Owing to large propagation losses, the reference device radiates primarily  from its front glass-cladded surface (the front-to-edge intensity ratio is about 50). By contrast, owing to reduced optical losses (inset in b, right) and strong amplification of guided light, the BRW emits more strongly from its edge (the edge-to-front intensity ratio is about 2 to 3). AU, arbitrary units.

Fig. 3. Electrically driven ASE in the BRW device.

a, A BRW device is built on top of a DBR made of ten pairs of Nb₂O₅ and SiO₂ layers. The device contains an ITO cathode, a ZnO ETL, a ccg-QD gain medium (three QD monolayers), a TFB HTL, a LiF interlayer with a current-focusing slit, a HAT-CN HIL and a strip-like Ag anode. b, Edge-emitted EL spectra of the BRW device as a function of current density tuned from 0.8 to 1,933 A cm⁻². The device was excited using pulsed bias with τ_p = 1 μs and pulse-to-pulse separation T = 1 ms. The EL spectra show a transition from broad spontaneous emission observed at low j to sharp 1S and 1P ASE bands at high j. c, Top, the j-dependent EL intensities at the peaks of the 1S spontaneous (black) and ASE (red) bands indicate the ASE threshold j_th,ASE ≈ 13 A cm⁻². Bottom, the dependence of 1S emission linewidth on j indicates progressive line narrowing from 82 to 39 meV. d, Polarization characteristics of edge-emitted light of the BRW device in the case of electrical (left, j = 650 A cm⁻²) and optical (right, 110-fs, 3.6-eV pulses, w_p = 85 μJ cm⁻²) excitation. Owing to strong damping of TM modes, the 1S and 1P ASE bands are not present in TM-polarized emission (blue) and exhibit nearly perfect TE polarization (red). The spontaneous 1S band is not polarized (black) and, as a result, is present in both TE-polarized and TM-polarized emission. e, The VSL measurements of the optically excited BRW device (inset) indicate the development of the 1S and 1P ASE features with increasing stripe length. These measurements used 110-fs, 3.6-eV pump pulses with w_p = 90 μJ cm⁻². The sharp ASE bands are similar to those observed in the EL spectra (panel b). AU, arbitrary units.

Fig. 4. Characterization of the output of the BRW device.

a, Left, the output of the electrically excited BRW device was characterized using a standard laser-lab power meter. The measured average power (P_av) was converted into the instantaneous output power (P_out) using P_out = (T/τ_p)P_av. For τ_p = 1 μs and T = 1 ms used in our experiments, P_out = 1,000P_av. Right, photograph of the device operating at j = 170 A cm⁻² under ambient conditions in room light clearly shows edge emission that appears very bright despite the small on-time fraction (τ_p/T = 0.001) and the small size of the emitting spot (the nominal area is about 9 μm²). Scale bar, 10 mm. b, The dashed blue line shows the j-dependent instantaneous output power. At the maximal current density (j = 1,933 A cm⁻²), P_out reaches 170 μW. On the basis of the measured output power, we determine the EQE (red circles), which is compared with that of the reference device (black triangles). Owing to the efficient ASE, which leads to the increased QD emission rate and enhanced power extraction from the inverted QD medium, the EQE droop is much less pronounced in the BRW device. In particular, j_½ is about four times higher than that for the reference device (1,933 versus 500 A cm⁻²).

Extended Data Fig. 1. Characterization of the reference electroluminescent device.

a, The EL spectra of the reference device (Fig. 1f) are deconvolved using three Lorentzian bands that correspond to the 1S_e–1S_hh, 1S_e–1S_lh and 1P_e–1P_hh transitions. This procedure is applied to the EL spectra measured at j = 41 A cm⁻² (left), 632 A cm⁻² (middle) and 1,019 A cm⁻² (right). b, The average QD excitonic occupancy, ⟨N⟩, in the active device volume as a function of current density. Symbols are experimental values of ⟨N⟩ obtained from the measured ratio of the 1P and 1S EL intensities. The line is calculations conducted using the model of ref. . According to this analysis, ⟨N⟩ reaches 7.4 at j = 1,019 A cm⁻². This is above the gain thresholds for both the 1S and 1P transitions^,. a.u., arbitrary units.

Extended Data Fig. 2. Comparison of edge-emitted and surface-emitted EL spectra of the reference and the BRW devices.

a, The spectra of edge-emitted and surface-emitted EL (top and bottom subpanels, respectively) of the reference device (insets illustrate the configuration of the measurements) do not show any apparent qualitative distinctions (the device is operated at 920 A cm⁻²). Furthermore, the edge emission is about 50 times weaker than surface emission. b, By contrast, the BRW device shows a substantial difference in the spectral shape of edge-emitted and surface-emitted EL (top and bottom subpanels, respectively); j = 730 A cm⁻². Furthermore, edge emission is about two times more intense than surface emission. The spectrum of edge-emitted EL is dominated by 1S and 1P ASE features, whereas the surface emission comprises a single narrow peak at 2.05 eV owing to a vertical Fabry–Pérot cavity formed by the bottom DBR and the top silver mirror. a.u., arbitrary units.

Extended Data Fig. 3. Photonic modelling of the BRW devices developed in the present study.

a, Contour maps of a TE field of the fundamental TIR mode (top) and the BRW mode (bottom) supported by the structure with the transverse DBR-Ag cavity. The optical field of the BRW mode is confined primarily in the device layer (waveguide core), whereas the TIR mode concentrates at the device–DBR interface and leaks into the DBR. b, The calculated ω–β dispersion (ω is the photon angular frequency and β is the modulus of the wavevector) of the TE modes allowed in the BRW structure (depicted in the inset). In this case, the highest (n₂) and lowest (n₁) index materials of the waveguide are Nb₂O₅ and SiO₂, respectively. There are no waveguided modes for n_eff (= βc/ω) > n₂, which corresponds to the ‘cut-off’ regime. In the range n₁ < n_eff < n₂ (red-shaded area), several TIR modes are supported by the waveguide owing to reflections from various layers of the thick DBR stack. The range n_eff < n₁ corresponds to a photonic bandgap or a stopband defined by the reflection spectrum of the DBR (purple line). A BRW mode (blue line) is located in the stopband of the photonic structure. c, A comparison of guided mode parameters between the TE₀ TIR (pink) and BRW (orange) modes of the DBR-based structure (Fig. 2b) and the TE₀ TIR mode (red) of the reference device (Fig. 2a). The calculated parameters include the effective refractive indices (n_eff), the modal angles (θ_m), the mode confinement factors for the ccg-QD layer (Γ_QD) and the optical-loss coefficients (α_loss).

Extended Data Fig. 4. The calculated mode parameters as a function of photon energy.

The spectral dependences of the effective refractive index (a), the modal angle (b), the mode confinement factor for the ccg-QD layer (c) and the optical-loss coefficient (d) for the reference device (denoted ‘Ref’) and the TIR and BRW modes (denoted ‘TE₀’ and ‘BRW’, respectively) of the BRW device.

Extended Data Fig. 5. Evolution of edge-emitted and surface-emitted EL spectra with increasing current density.

a, Edge-emitted EL spectra of the BRW device as a function of j. The bottom subpanel shows ‘raw’ (not normalized) experimental spectra. The top subpanel shows the normalized spectra scaled so as to match the amplitude of the 1S spontaneous emission feature (owing to a high noise level below the ASE threshold, we present the measured spectra using two-band Gaussian fits). The normalized spectra clearly show the emergence of a sharp 1S ASE band. b, Similar sets of data for surface-emitted EL of the same device. The EL is dominated by a vertical cavity mode at 1.84 eV. It is red-shifted versus that of the device shown in Extended Data Fig. 2b, because of a larger thickness of the device tested in the present measurements. The feature around 1.77 eV results from light leakage through the DBR (see Supplementary Fig. 4). Unlike edge-emitted EL, the surface-emitted signal shows spectrally uniform growth with increasing j. a.u., arbitrary units.

Extended Data Fig. 6. The j-dependent spectra of edge-emitted EL of  the BRW device measured with a TE polarizer.

Owing to a ‘lossy’ character of TM modes (strongly attenuated by the silver electrode), ASE occurs because of low-loss TE modes. The spontaneous emission does not exhibit preferred polarization. Therefore, the use of a TE polarizer allows us to emphasize narrowband ASE features over broadband spontaneous emission. As a result, the EL spectra measured at high current densities (j > 200 A cm⁻²) are dominated by narrow 1S and 1P ASE peaks. The marked difference of these spectra from those of the reference devices (Fig. 1f and Extended Data Fig. 1a) is yet another confirmation of the ASE effect realized in our BRW devices. a.u., arbitrary units.

Extended Data Fig. 7. Polarization measurements of edge-emitted EL.

a, Edge-emitted EL spectra of the BRW device operated at about 100 A cm⁻² collected through a linear polarizer whose polarization direction is varied from 90° (vertical polarization; TE mode) to 0° (horizontal polarization; TM mode). The dashed black lines at 1.93, 1.98 and 2.08 eV correspond to the maxima of the 1S ASE, 1S spontaneous emission and 1P ASE features, respectively. Polarization-dependent changes in the EL spectral shape indicate that both the 1S and 1P ASE bands are strongly TE polarized. b, The amplitudes of the 1S ASE (red circles) and 1P ASE (green circles) bands as a function of polarization angle (α). The measurement of one data point took around 1 min and it took approximately 36 min to obtain the entire dataset. There was some degree of signal degradation during the measurements, which led, in particular, to a drift of a baseline. c, The same measurements plotted using polar coordinates. The experimental data (symbols; same colours as in b) are in good agreement with the theoretical sin²α dependence expected for TE-polarized light (dashed black line). a.u., arbitrary units.

Extended Data Fig. 8. Coherence measurement of edge-emitted EL conducted using Michelson interferometry.

a, Interferogram of the 1S ASE band of the BRW device driven at about 100 A cm⁻². The edge-emitted light is fibre coupled into a Michelson interferometer consisting of a cube beam splitter, a fixed arm and a variable-length arm controlled by a piezoelectric stage. The interference signal is detected with a Si avalanche photodiode. 1S ASE is isolated using a 620-nm (or 2-eV) long-pass filter and a linear polarizer set to the ‘vertical’ (TE) polarization (Fig. 3d, left). b, Fringe visibility extracted from the interferogram in a (black symbols) as a function of the time delay between the two arms of the interferometer. An exponential fit of the experimental data (red line) yields coherence time τ_c of 47 fs. c,d The interferometric measurements conducted for the ‘horizontal’ (TM) polarization in which ASE is suppressed (Fig. 3d, left) yield a much shorter coherence time of 16 fs. The observed polarization-dependent threefold difference in τ_c indicates the presence of two subsets of photons with distinct temporal coherences. As expected for the ASE effect, the photons produced by stimulated emission (they dominate the edge-emitted TE-polarized light) exhibit a longer τ_c compared with photons produced by spontaneous emission (they dominate the edge-emitted TM-polarized light). The measured coherence times are consistent with those inferred from τ_c ≈ (πΔν)⁻¹, in which Δν is the emission linewidth in the frequency domain. In particular, based on the linewidths of the 1S spontaneous and ASE bands (82 to 30 meV, respectively, for the sample used in the studies of coherence), the estimated coherence times are 16 fs and 44 fs. Both values are in excellent agreement with the results of the interferometric measurements (16 fs and 47 fs, respectively).

Extended Data Fig. 9. Studies of operational stability of a BRW device.

a, The dependence of the current density (left axis, open diamonds) and the output power (right axis, blue circles) of an unencapsulated BRW device on time of operation under excitation with 10-V, 1-µs pulses at 1 kHz. The data were recorded at intervals of 1 min. The data points shown in the plot (separated by 10-min intervals) are averages over ten measurements. During the first 2 h, the output intensity decreased by only 9%. A slow decrease in the output power and the current density continued for the next 2 h, at which point device failure finally occurred. On the basis of these measurements, the characteristic T₇₅ time was 2.7 h (T₇₅ denotes the time at which the device output decreases to 75% of its original value). b, The EL spectra at the beginning of the test show distinct 1S and 1P ASE features. c, These features are preserved following 2 h of continuous device operation. a.u., arbitrary units.

Extended Data Fig. 10. Reproducibility of characteristics of BRW ASE devices.

a–d, Characteristics of four devices from four different chips (each chip contains eight devices). The data are organized into four columns, one column per device. The top row shows the j–V curves (black lines) and the V-dependent edge emission intensity (right axis, blue lines). The middle row shows EL spectra that feature prominent 1S and 1P ASE bands. The bottom row shows the dependence of the 1S bandwidth on j, which exhibits ‘line narrowing’ typical of the transition to the ASE regime. There is excellent consistency between all shown datasets. e, The analysis of device-to-device variability of a turn-on voltage (left), a 1S ASE threshold (middle) and a 1S ASE linewidth. The histograms were obtained on the basis of the measurements of 11 devices. The average values of the measured parameters and the standard deviations are indicated in the figure. In the case of the turn-on voltage and the ASE linewidth, the standard deviations are 22% and 25% of the average value, respectively. The larger deviation observed for the ASE threshold (about 52%) can be attributed to high sensitivity of j_th,ASE to device-to-device variations in propagation losses and a varied degree of charging of an active QD layer. a.u., arbitrary units.