Optically Excited Lasing in a Cavity‐Based, High‐Current‐Density Quantum Dot Electroluminescent Device

Laser diodes based on solution‐processable materials can benefit numerous technologies including integrated electronics and photonics, telecommunications, and medical diagnostics. An attractive system for implementing these devices is colloidal semiconductor quantum dots (QDs). The progress towards a QD laser diode has been hampered by rapid nonradiative Auger decay of optical‐gain‐active multicarrier states, fast device degradation at high current densities required for laser action, and unfavorable competition between optical gain and optical losses in a multicomponent device stack. Here we resolve some of these challenges and demonstrate optically excited lasing from fully functional high‐current density electroluminescent (EL) devices with an integrated optical resonator. This advance has become possible due to excellent optical gain properties of continuously graded QDs and a refined device architecture, which allows for highly efficient light amplification in a thin, EL‐active QD layer.


DOI: 10.1002/adma.202206613
Colloidal QDs are promising materials for realizing laser diodes. As prepared, they comprise a nanosized semiconductor core encased into a shell of organic molecules. As a result, they combine chemical flexibility of molecular structures with advantages of quantum-confined inorganic semiconductors. In particular, they offer facile spectral tunability, achievable by size and/or composition control, and high (near-unity) emission quantum yields. [9] Further, a wide separation between their quantized states prevents thermal depopulation of the band-edge emitting levels, thereby reducing lasing thresholds and improving thermal stability of lasing characteristics. [10] The availability of lasers based on solution-processable QDs would benefit multiple technologies and, especially, integrated photonics and electronics where the application of existing lasers based on epitaxial materials is hampered by lattice-match requirements and incompatibilities between different fabrication techniques required to prepare the entire device structure.
The primary difficulty for realizing lasing with colloidal QDs is fast, confinement-enhanced nonradiative Auger recombination wherein the recombination energy of an electron-hole (e-h) pair (exciton) is not emitted as a photon but dissipates via Coulomb energy transfer to the third carrier. [11,12] This complication arises from a multifold degeneracy of the QD band-edge levels, because of which the development of optical gain requires that QDs are populated with more than one exciton per dot on average. As a result, optical gain relaxation is controlled by multicarrier Auger recombination which leads to very short (typically sub-100 ps) gain lifetimes (τ g ). [13] This represents an especially serious problem in the case of continuous-wave optical or direct-current electrical pumping. [3,4] In particular, it leads to extremely high lasing thresholds that are often well above damage thresholds of colloidal QDs and/or other materials (e.g., organic molecules) used in a solution-processed device. [14][15][16] Recently, strong suppression of Auger decay has been achieved by employing continuously graded QDs (cg-QDs) wherein a CdSe core is enclosed into a Cd x Zn 1-x Se shell with x varied from 1 to 0 along the radial direction. [17] For improved stability, the CdSe/Cd x Zn 1-x Se cg-QDs are usually overcoated with a thin protective layer of ZnSe y S 1-y . [18] The invention of cg-QDs allowed for several advances of direct relevance to the development QD laser diodes (QLDs). These include the demonstration of band-edge (1S) optical gain with d.c. electrical Laser diodes based on solution-processable materials can benefit numerous technologies including integrated electronics and photonics, telecommunications, and medical diagnostics. An attractive system for implementing these devices is colloidal semiconductor quantum dots (QDs). The progress towards a QD laser diode has been hampered by rapid nonradiative Auger decay of optical-gain-active multicarrier states, fast device degradation at high current densities required for laser action, and unfavorable competition between optical gain and optical losses in a multicomponent device stack. Here we resolve some of these challenges and demonstrate optically excited lasing from fully functional high-current density electroluminescent (EL) devices with an integrated optical resonator. This advance has become possible due to excellent optical gain properties of continuously graded QDs and a refined device architecture, which allows for highly efficient light amplification in a thin, EL-active QD layer.

Introduction
Semiconductor laser diodes based on solution-processable materials have been pursued across multiple fields including organic semiconductors, [1] perovskites, [2] and colloidal quantum dots (QDs). [3,4] Their development has been motivated by prospective applications in lighting, [5] communications, [6] sensing, [7] and data storage. [8] pumping [17] and, more recently, the realization of pulsed QD devices capable of stable operation at current densities (j) of ≈1000 A cm −2 , which was sufficient to achieve optical gain for both the 1S and the higher-energy 1P transition. [4] Another important milestone towards a functional QLD was the realization of optically excited lasing in light-emitting-diode (LED)-like devices with an integrated distributed feedback (DFB) resonator. [19] The reported structures featured a multilayered stack of a standard "inverted" QD LED which lacked only a top hole-injecting electrode. Despite a fairly small thickness of an active cg-QD layer (≈50 nm) and the presence of "lossy" charge-transport layers, these devices showed good lasing performance under pulsed optical excitation. They also exhibited bright electroluminescence (EL) upon the addition of a final MoO x /Al contact. However, this last step resulted in the suppression of lasing, indicating that the overall optical losses in the complete EL device increased significantly due to the top contact and became greater than the modal gain generated by the QDs.
Here we resolve the "quenching problem" by implementing an integrated approach in which optical gain/loss engineering is combined with engineering of charge injection so that the resulting devices exhibit strong gain, suppressed optical losses, and excellent waveguiding properties. The developed devices can generate extremely high current densities (up to ≈ 560 A cm −2 ), sufficient to excite broad-band optical gain spanning 1S and 1P transitions. Importantly, at the temperature of liquid-nitrogen (LN), the generated modal gain outcompetes optical losses, which allows us to realize strong 1S and 1P amplified spontaneous emission (ASE) with optical excitation. Further, we demonstrate optically excited single-mode lasing in a fully stacked high-j EL device, which contains a 2nd order DFB resonator engraved into a conducting metal oxide cathode.

Compact Continuously Graded QDs as Optical Gain Materials
The modal gain realized in an LED device stack (G mod ) can be presented as the product of a mode confinement factor (Γ QD ) of an active QD layer and the QD "material gain" defined as the optical gain of an infinitely thick QD solid (G mat ): G mod = Γ QD G mat . Here, we aim to enhance both G mat and Γ QD via, respectively, enhancing gain performance of the QD medium and optimizing the overall structure of the device stack.
First, we discuss our approach to enhancing G mat . This quantity can be presented as G mat = σ gain n QD , where σ gain is the QD gain cross-section and n QD is the QD density in the film. For our type-I cg-QDs, the oscillator strength of the band-edge transition, which defines σ gain , is expected to be virtually independent of QD dimensions. [20,21] Therefore, we seek to boost G mat , by increasing QD packing density (n QD ). In the case of closepacked solids, n QD scales inversely with QD volume (V QD ) and, hence, it can be increased by decreasing the overall QD size. To test the validity of this approach, we synthesize a series of CdSe/Cd x Zn 1-x Se/ZnSe 0.5 S 0.5 /ZnS QDs (see Experimental Section and Figure S1, Supporting Information, for details of the synthesis) with a similar core radius (ca. 2.6 nm) and a varied thickness of the graded layer (Figure 1a). Then, we process them into a 300-nm thick close-packed film on top of a glass slide and characterize their optical gain using a variable stripe length (VSL) technique (Experimental Section).
The VSL measurements reveal the anticipated trend, which is the increase in G mat with decreasing QD dimensions ( Figure 1b). The measured G mat scales approximately as 1/V QD (Figure 1c), as expected for direct proportionality between G mat and the QD packing density. In particular, for the original cg-QDs with R QD = 9.4 nm (refs. [17,22]), the band-edge (1S) gain coefficient is ≈200 cm −1 . It is boosted more than threefold (to 650-800 cm −1 ) for smaller-size, ≈6-nm cg-QDs, which is consistent with the decrease of V Q by a factor of 3.8. Importantly, despite their reduced dimensions, the "compact" cg-QDs (hereafter referred to as ccg-QDs) exhibit strong suppression of Auger recombination. Specifically, their biexciton lifetimes (τ XX of 0.9-1.3 ns; Figure 1b and Figure S2, Supporting Information) are similar to those of the original cg-QDs with a thick graded layer. [17] Because of their strong optical-gain performance, we use the ccg-QDs in the present work. Specifically, for our device-related studies, we prepare samples whose core radius is 2.6 nm and the overall radius is 6.25 nm (Figure 1a). In Figure 1d (left), we present their photoluminescence (PL; red) and linear absorption (α 0 ; black) spectra along with α 0 ″ (blue). As the original cg-QDs, [17,22] the ccg-QDs show an enhanced splitting (56 meV) between light and heavy hole states (abbreviated as "lh" and "hh"; Figure 1d, right), ascribed previously to asymmetric compression of the CdSe core. [22] The lh-hh splitting manifests as a double-peak band-edge (1S) feature observed in both α 0 and α 0 ″ spectra. These spectra also exhibit the third, higher-energy feature due to the transition involving the 1P electron and hole states (Figure 1d, right).

Optimization of Optical Gain/Loss Characteristics via Device Stack Engineering
In the next step, we aim to reduce optical losses and simultaneously increase Γ QD via the optimization of individual elements of the LED device stack and its overall structure. A common feature of a traditional "inverted" QD LED [23] is a bottom electron-injecting electrode (cathode) made of indium tin oxide (ITO) used for its optical transparency and high electrical conductivity. However, standard ITO may inhibit light amplification because of strong free-carrier absorption in the range of visible wavelengths. Further, due to its high refractive index (n ITO = 1.89 at λ = 600 nm, ref. [24]), which is comparable to that of the QD solid (n QD = 1.92, ref. [19]), the ITO layer tends to "pull" the optical field from the active QD layer, which reduces Γ QD and thereby diminishes modal gain. These problems were identified in ref. [19] and were successfully tackled using socalled low-index ITO or L-ITO (1:2 mixture of ITO and SiO 2 ). This material exhibits a considerably lower absorption coefficient compared to standard ITO (160 cm −1 vs 483 cm −1 at λ = 600 nm; Figure 2a) and features a reduced refractive index (1.6 vs 1.89). Based on these favorable characteristics, we use L-ITO in our devices as a cathode.
Another common element of standard inverted QD LEDs is an electron transport layer (ETL) made of n-type colloidal or sol-gel ZnO. [19,25] The Fermi level of this material matches well the conduction band of CdSe QDs, which facilitates electron injection. [26] However, the ZnO refractive index (n ZnO = 2.01) [27] is higher than that of the QD solid which reduces Γ QD. In addition, the ZnO layer is optically lossy [28] and its low thermal conductivity can hamper heat exchange at high current densities leading to device breakdown ( Figure S3, Supporting Information). To avoid these detriments, we exclude the ZnO ETL from our device stack and attempt to inject electrons into the QDs directly from the L-ITO contact. As discussed later in this work, this approach indeed improves waveguiding properties of our devices and, importantly, still allows us to reach very high current densities sufficient for generating strong optical gain.
Additionally, we also modify the hole injection/transport part of the LED structure. Commonly, hole injection is accomplished using a combination of an organic hole-transport layer (HTL) Figure 1. Structural, electronic, and optical characteristics of ccg-QDs. a) The internal structure of the CdSe/Cd 1-x Zn x Se/ZnSe 0.5 S 0.5 /ZnS ccg-QD (2.6-nm core radius, 6.25-nm overall radius) (top), their exemplary transmission electron microscopy (TEM) image (bottom left), and the approximate shapes of the "graded" electron and hole confinement potentials (bottom right). b) The 1S (red circles) and 1P (green squares) material gain coefficients of the cg-QDs as a function of the overall diameter. Biexciton lifetimes (τ XX ) are indicated next to corresponding data points. c) The measured 1S material gain (red circles) scales approximately linear with inverse of the overall QD volume (dashed black line). d) (Left) The spectra of linear absorption (black line), its second derivative (blue line), and PL (red line) of the ccg-QD sample displayed in "a". (Right) The structure of the near-band-edge electron (1S e , 1P e ) and hole (1S hh , 1S lh , 1P hh ) states along with approximate inter-state energies derived from spectra on left. All experimental data shown in this figure were obtained at room temperature. made of, for example, tris(4-carbazoyl-9-ylphenyl)amine) (TCTA, 50 nm), followed by a thin MoO x (10 nm) hole injection layer (HIL). (Figure 2b, top) [17,18] From the QLD standpoint, the detriment of this design is extremely high optical losses in MoO x (α MoOx = 3153 cm −1 at λ = 600 nm), [29] which results in undesired quenching of waveguided modes. [19] Strikingly, when MoO x is combined with TCTA, the absorbance of the resulting TCTA/ MoO x bi-layer becomes even greater than the sum of α MoOx and α TCTA (Figure 2b, top), leading to additional optical losses. This likely occurs due to the increased free-carrier density in the organic HTL caused by the inflow of holes from MoO x . [30] To mitigate these problems, we exclude MoO x from our devices and instead implement hole injection using an all-organic bi-layer made of 2,2′,7,7′-tetrakis[Nnaphthalenyl(phenyl)-amino]-9,9-spirobifluorene (Spiro-2NPB) and dipyrazino[2,3-f:2′,3′-h]quinoxaline-2,3,6,7,10,11-hexacarbonitrile (HAT-CN). [31] The absorbance of individual layers in this combination is much lower than that of MoO x and, importantly, it is not increased due to interfacial effects when they are joined together (Figure 2b, bottom).
To compare the optical properties of our new device stack with those of the standard inverted LED, we conduct their modeling using a finite element method implemented with a standard COMSOL software (Experimental Section). The structures of the simulated devices, which are assumed to be completed with a top contact (anode) made of silver, are ITO/ZnO/ QDs/TCTA/MoO x /Ag ( Figure 2c) and L-ITO/QDs/Spiro-2NPB/ HAT-CN/Ag ( Figure 2d). Importantly, the presence of the metal anode, leads to strong quenching of transverse magnetic (TM) modes ( Figure S4, Supporting Information). Therefore, in our modeling, we focus on the fundamental transverse electric (TE 0 ) mode.
The results of simulations are displayed in Figure 2c,d. They include profiles of transverse electric field (black line) and optical power losses (red line), as well as the spatial distribution of the electric field along the waveguide formed by the QD layer (red/blue shading) at λ = 600 nm. In the structure with a traditional inverted LED design (Figure 2c  220 cm −1 . Other device elements yield a small additional contribution of 42.5 cm −1 . As a result, the total loss coefficient (α loss ) is 262.5 cm −1 (Figure 2e, blue circles).
In addition to large losses, a further problem of the standard LED architecture is a nonoptimal optical field profile which peaks not in the QD part of the device, but at the interface of the ZnO and ITO layers (Figure 2c, black line), as a consequence of their high refractive indices. This reduces the confinement factor for the QD layer ( Figure 2f) and leads to high threshold values of the QD material gain (G mat,th ) required for lasing. In particular, for the structure in Figure 2c, Γ QD = 0.21, which yields G mat,th = α loss /Γ QD = 1250 cm −1 . Since the maximal 1S gain achievable with ccg-QDs is ≈800 cm −1 (Figure 1b,c), the traditional LED design is not suitable for realizing a QLD.
The newly proposed LED design ( Figure 2d) is much more favorable for light amplification. In this case, the QD film is sandwiched between the L-ITO cathode and the organic HTL whose refractive indices are lower than that of the QDs (n L-ITO = 1.6, n HTL = 1.8). As a result, the peak of the optical field distribution shifts into the QD layer ( Figure 2d, black line), which leads to about twofold increase of the mode confinement factor (Γ QD = 0.4; Figure 2e). Further, thanks to the reduced absorbance of L-ITO, the overall optical loss drops to ≈143 cm −1 (Figure 2d, red line and Figure 2e, green diamonds). As a result of these improvements, G mat,th is reduced to 358 cm −1 , a value easily accessible with ccg-QDs (Figure 1b,c).

EL Properties of LEDs with Optimized Optical-Gain Characteristics
To practically implement the developed design, we fabricate LEDs whose structure and schematic band diagram are displayed in Figure 3a (left and right panels, respectively); see Experimental Section for fabrication details. In our devices, we use an active layer made of 6 ccg-QD monolayers. This is thicker than in standard QD LEDs that usually employ 1 to 3 QD monolayers to ensure spatially uniform injection of electrons and holes across the entire thickness of the active medium. Our devices require a greater active-layer thickness as dictated by the need to boost modal gain to levels sufficient to overcome optical losses. In the course of device optimization, we have determined that a 6-QD-monolayer thickness yields a desired compromise that allows for fairly efficient charge injection and simultaneously provides sufficiently high modal gain.
We spin-coat the ccg-QDs on top of an L-ITO substrate and then use thermal evaporation to deposit the spiro-2NPB HTL, the insulating LiF layer with a 50-µm-wide gap, the HAT-CN HIL, and the Ag anode prepared as a narrow (~300-µm wide) strip orthogonal to the gap in the LiF membrane (see top-view and cross-section scanning electron microscopy (SEM) images in Figure 3b; left top and bottom, respectively). A combination of the specially-shaped Ag contact and the aperture in the insulting LiF layer leads to 2D confinement of the injection area (Figure 3b, right), commonly described as "current focusing", which helps improve heat exchange of the emitting volume with the environment and thereby boost the maximal current density (j) achievable before device breakdown. [4] Direct contact of the ccg-QDs with the conductive L-ITO cathode could, in principle, lead to emission quenching due to nonradiative pathways associated with charge and/or energy transfer. However, time-resolved PL measurements indicate that PL dynamics of the ccg-QDs prepared on top of the L-ITO substrate are virtually identical to that of dots on the glass substrate, indicating the absence of additional nonradiative decay channels ( Figure S5, Supporting Information). This is likely due to the shielding effect of the graded Cd x Zn 1-x Se layer and the ZnSe 0.5 S 0.5 /ZnS shell that isolate the emitting CdSe core from the electrode. [18] Next, we examine electro-optical characteristics of the fabricated LEDs under both room temperature (T = 300 K) and cryogenic LN cooling (T = 80 K). To reduce overheating of the devices at high j, we drive them using short-pulse excitation (0.5 µs pulse duration, 100 Hz repetition rate). [4] In Figure 3c, we present the current-density-voltage (j-V) characteristic (black line) and the V-dependent EL intensity (blue dashed line) of the LNcooled LED. As V is increased, the j-V curve features a transition from ohmic conductance (j ∝ V 1 ) to charge transport in the "trap-filled limit" (j ∝ V 4 ). [32] During the second phase, the current density rapidly increases reaching j of 361 A cm −2 , which is well above the optical-gain threshold. [3,4] The realization of such high values of j is facilitated by the implemented "currentfocusing" LED design and the use of pulsed bias, the measures applied previously to reduce the amount of generated heat and to improve heat outflow from the active device volume. [4,33,34] Due to high j, the device produces intense EL at j = 169 A cm −2 (Figure 3c, blue dashed line). At higher j, the EL intensity saturates and then exhibits a slight drop. This behavior, however, is reversible, indicating that it is not due to device degradation but rather due to temperature-induced EL quenching. [35] The developed devices also exhibit strong performance at room temperature. In particular, they reach current densities up to 557 A cm −2 ( Figure S6a, Supporting Information), which exceeds maximal j obtained at the LN temperature. Due to higher j, we are able to achieve the regime of trap saturation marked by the transition to the j ∝ V 2 dependence (observed at V > 20 V), which is typical of "trap-free space-charge-limited" conductance. [32] Due to the high current densities accessible with our devices, they allow us to realize the unusual regime of multiband EL. In Figure 3d, we present spectrally resolved EL recorded at the LN temperature for increasing current density. At j = 15 A cm −2 , we observed a standard single-band EL due to the band-edge transition, which couples the 1S e electron and 1S hh heavy hole states (Figure 3d, left). The spectrum recorded at j = 68.1 A cm −2 exhibits a pronounced broadening on its higher-energy side, which is due to the emergence of emission involving the 1S lh state (Figure 3d, middle). The corresponding band obtained via a two-band Lorentzian fit (shown by orange shading) is separated from the band-edge feature (pink shading) by 57 meV. This is consistent with the light-heavy hole splitting observed in the absorption spectrum (Figure 1d). [17] The two-band EL structure also manifests in the 2nd derivative of the EL signal (blue line in Figure 3d, middle), which exhibits two peaks separated by 57 meV.
At j = 68.1 Acm −2 , the EL spectrum exhibits further broadening to higher energies indicating the emergence of the third As a result of this "current-focusing" design, the emitting area is confined to ca. 290 µm by 50 µm (image at right). c) The j-versus-V (black line, left axis) and EL intensity-versus-V (blue line, right axis) characteristics of the fabricated LED measured at LN temperature. The device is driven using electrical pulses with a 500-ns duration and a 100-Hz repetition rate. d) Normalized EL spectra measured at LN temperature (black lines) and the spectra of their 2nd derivative (blue lines) for current densities of 15 (left), 68.1 (middle), and 204.5 (right) A cm −2 . These spectra can be accurately described using 1 to 3 Lorentzian bands (colored shading) that correspond to the 1S e -1S hh (red arrows), 1S e -1S lh (orange arrows), and 1P e -1P hh (green arrows) transitions (insets). These bands emerge one after the other when the QD is excited with more than 0, 2, and 4 e-h pairs, respectively (electrons and holes are shown in the insets by solid and open circles, respectively). band (Figure 3d, right). It is well pronounced in the 2nd derivative of the EL spectrum, based on which we infer that it is separated from the band-edge feature by 125 meV. This is consistent with the spacing between the 1S e -1S hh and the 1P e -1P hh transitions observed in the linear absorption spectrum (Figure 1d), suggesting that the third EL band is due to carriers occupying the 1P electron and hole states.
These observations can be explained by the progressive filling of higher energy states with increasing QD occupancy. If the number of the e-h pairs per dot (N eh ) is 2 or less, all injected carriers reside in the twofold-degenerate 1S e and 1S hh states, which leads to single-band EL (inset of Figure 3d, left). At higher occupancies, the carriers are forced into higher energy 1P e and 1S lh states, which leads to the emergence of the 1S e -1S lh feature (inset of Figure 3d, middle). When N eh exceeds 4, the holes start to fill the 1P hh level, which opens an additional radiative channel due to the 1P e -1P hh transition (inset of Figure 3d, right). This analysis suggests that the current densities realized in our devices are sufficient to completely fill the 1S e and 1S hh states, which requires N eh = 2. Further, since at the highest j, the 1S e -1S lh feature also reaches saturation, this implies that the number of e-h pairs per dot exceeds 4. These results indicate that the developed LEDs allow us to achieve saturation of the band-edge gain (occurs when N eh ≥ 2), [3,4] and, as a result, realize gain coefficients comparable to those obtained with short-pulse optical pumping (Figure 1b,c).
We are also able to achieve high per-dot occupation factors at room temperature ( Figure S6b,c, Supporting Information). In particular, at the highest current densities, we observe the intense 1P e -1P hh feature indicating that N eh exceeds 4. In fact, in this case, the realized values of N eh are higher than those at the LN temperature as indicated by the greater ratio of the amplitudes of the 1P and 1S features (0.6 vs 0.3; Figure S7, Supporting Information). The external quantum efficiency (EQE) of the device reaches 1.5% at 86 A cm −2 ( Figure S6d, Supporting Information), which is comparable to EQEs of high-j devices with a traditional "optically lossy" MoO x /Ag anode. [36]

Optically Excited ASE from Fully Stacked EL-Active Devices
Next, we demonstrate that in addition to delivering strong optical gain, the developed high-j LEDs are capable to produce optically excited ASE in a complete, EL-active device. In these experiments, the EL-active part of the device is excited through the transparent ITO electrode using a femtosecond pulsed laser (<190-fs pulse duration, 343-nm excitation wavelength) whose beam is tightly focused so as to illuminate QDs exclusively within the injection area defined by the "current-focusing" elements of the device (Figure 3b, top and Figure 4a). This type of excitation allows us to re-create the situation of electrical pumping and thereby accurately evaluate the effect of losses arising from all elements of a fully stacked LED.
In Figure 4b, we display pump-intensity dependent PL spectra collected from the cleaved edge of the device cooled to the LN temperature. As per-pulse pump fluence (w p ) is increased, we observe a clear transition from spontaneous emission to ASE, first at the band-edge 1S transition and then at the higher energy 1P transition. The development of ASE is evident in both a sharp growth of the PL peak amplitude (I PL , Figure 4c, open symbols) and pronounced line narrowing (Figure 4c, solid symbols). In particular, the emergence of ASE is accompanied by the change in the log-log slope of the I PL -versus-w p dependence from 0.86 to 1.57 for the 1S feature (Figure 4c, left), and from 1.54 to 5.51 for the 1P band (Figure 4c, right). The ASE polarization corresponds to the TE waveguided modes ( Figure S8, Supporting Information). This is in agreement with our modeling which predicts strong quenching of the TM modes ( Figure S4, Supporting Information).
Based on the conducted measurements, the 1S and 1P ASE thresholds are w 1S,th = 56.4 µJ cm −2 and w 1P,th = 88.2 µJ cm −2 . If defined in terms of the average QD occupancy, 〈N eh 〉, they translate into 〈N eh 〉 ≈ 4.5 and 7, respectively. These two values are higher than the "ideal" optical gain thresholds (1 and 6 e-h pairs per dot; Note S1, Supporting Information), which is a result of optical losses arising from various elements of the LED. Importantly, however, in the developed devices, the modal gain is sufficiently high to overcome optical losses, which leads to the desired effect of light amplification. This is a direct result of the conducted optimization of the optical field profile across the device stack, improved optical-gain properties of the ccg-QDs, and elimination (or replacement) of lossy device components based on ZnO, MoO x , and standard ITO.
To analyze the interplay between optical gain and optical losses, we investigate ASE characteristics of partial (incomplete) device stacks that comprise selected sub-units of a full LED (Figure 4d-f and Figure S9, Supporting Information; T = 80 K). In the electrode-free structure, which contains ccg-QDs sandwiched between the glass substrate and the organic HTL/ HIL (Figure 4d), ASE develops at low pump fluences of ≈6 (1S) and ≈28 µJ cm −2 (1P), due to the lack of appreciable optical losses in either the underlying substrate or the top organic layers (Figure 2b).
In the device with the L-ITO layer inserted between the QDs and the glass substrate (Figure 4e), the ASE thresholds remain virtually unchanged (8.8 and 28.2 µJ cm −2 for the 1S and 1P features, respectively). The situation is different in the device wherein instead of the bottom L-ITO layer, we deposit a top Ag electrode (Figure 4f). In this case, the ASE thresholds exhibit a considerable (about twofold) increase (to 21.2 and 56.4 µJ cm −2 , for the 1S and 1P bands, respectively), observed simultaneously with the decrease of the log-log slope characterizing the ASE growth. These observations indicate that the metal contact is a dominant source of optical losses in our devices, while the specially engineered L-ITO electrode only contributes to a small fraction of the device loss.

Effect of "Double Pumping" on ASE of a Fully Stacked EL-Active Device
Since optically excited ASE is observed in a fully stacked EL device, we are able to conduct experiments with "double pumping" when the sample is excited by synchronized electrical and optical pulses. Figure S10, Supporting Information, displays the results of the measurements for the regime when the device driven by 500-ns electrical pulses is simultaneously illuminated from the L-ITO side by 190-fs optical pulses each of which is aligned with the center of the electrical pulse. These results are compared to the measurements with purely optical excitation. These experiments reveal an appreciable enhancement in the ASE signal for both the 1S and 1P transitions, indicating that electrical excitation indeed helps increase optical gain in the device.

Effect of Temperature
While our devices exhibit ASE at T = 80 K, the ASE is quenched at room temperature ( Figure S11, Supporting Information). Since ccg-QD gain is virtually the same at T = 80 and 300 K ( Figure S12, Supporting Information), the observed quenching likely occurs due to temperature-induced increase in optical losses. In particular, the absorbance of an Ag layer is known to increase with temperature as a result of the effects of electron-phonon scattering. [37] The increase in temperature can also raise effective doping levels of the L-ITO electrode and the organic layers and thereby boost free-carrier absorption. [38] Although our devices are capable to generate large modal gain with electrical pumping, they do not exhibit electrically excited ASE or lasing, which is likely due to considerable overheating at high current densities. Indeed, although in the case of cryogenic cooling, the nominal temperature of our LEDs is 80 K, the actual temperature can be much higher due to heat accumulation at high j. [4] This effect is evident in the progressive red shift of the EL spectrum with increasing j ( Figure  S6b, Supporting Information). Based on the magnitude of the observed shift ( Figure S13, Supporting Information), the device temperature near breakdown reaches 180 K, which is 100 K higher than the nominal temperature. As discussed earlier, the increase in the temperature leads to increased optical losses which apparently become so high that they overwhelm optical gain.

Quantitative Optical Gain/Loss Analysis
In the next step, we quantify modal gain and overall optical losses in our LEDs using a combination of VSL measurements and calculations (see details in Figure S14, Supporting Information). First, we apply a VSL method to a ccg-QD layer within a fully assembled LED. In this case, the active area was extended to 1.5 × 1.5 mm 2 to allow for a wide range of excitation stripe lengths. Such a measurement yields the net gain of the device (G net = G mod − α loss ). Figure 5a presents a scheme of the measurement as well as the clear progression of the ASE as the stripe length is increased. Based on these measurements, the net gain reaches ≈5 (1S) and 220 cm −1 (1P) at the highest pump fluence w p = 248 µJ cm −2 . (Figure 5b,c). The net gain coefficients at four different pump fluences are estimated in Figure S15a,b, Supporting Information, and marked as green squares in Figure 5d,e for the 1S and 1P bands, respectively.
Second, we infer the material gain by conducting pumpfluence-dependent VSL measurements on a thick ccg-QD film assembled on top of a low-loss glass substrate ( Figure S15c,d, Supporting Information). Based on the measurements and the computed mode confinement factor (Γ QD = 0.92), we obtain the material gain (G mat ) of our ccg-QDs plotted in Figure 5d,e (black diamonds and dashed lines) for the 1S and 1P bands, respectively. These data indicate a quick growth of G mat with increasing w p , followed by saturation. The saturated material gain coefficients are 543 (1S) and 1071 cm −1 (1P). Using these results and Γ QD computed for our devices (Figure 2f), we obtain the w p -dependent 1S and 1P modal gain coefficients G mod (Figure 5d,e; blue diamonds and dashed lines). Based on these data, the maximal modal gain realized in our devices is 190 cm −1 for the 1S transition and 407 cm −1 for the 1P transition. Using these values along with previously derived net gain coefficients in the device, we obtain that the loss coefficients (α loss = G mod − G net ) are 185 (1S) and 187 cm −1 (1P) bands.  Based on the standard VSL analysis (dashed lines), the net 1S and 1P gain coefficients are 5 and 220 cm −1 , respectively. d) The 1S material, modal, and net gain coefficients (black diamonds, blue diamonds, and green squares, respectively) obtained from the pump-fluence-dependent VSL measurements of the thick ccg-QD film on a glass substrate and the complete device; dashed lines are guides for the eye. The difference between the modal and the net gain coefficients (yellow shading) yields a total loss coefficient of 185 cm −1 . e) A similar set of data for the 1P transition yields α loss of 187 cm −1 . All measurements were performed at LN temperature. (Figure 5d,e and Table 1) These values are in close agreement with our simulations (Figure 2e), according to which α loss is around 150 cm −1 for both 1S and 1P transitions. The higher value of α loss obtained from the measurements (≈40 cm −1 difference) is likely due to additional losses arising from light scattering, not accounted for in the modeling.

Optically Pumped Lasing in an LED with an Integrated Cavity
Next, we study LEDs with an integrated 2nd-order DFB cavity. To prepare these devices, we engrave a 1D grating with the period Λ DFB = 370 nm in a bottom L-ITO electrode via laser interference lithography (Figure 6a); see Experimental Section for fabrication details. The 2nd order Bragg resonance of this structure (λ 2 = Λ DFB n eff = 617 nm; n eff = 1.67 is the effective refractive index) falls within the 1S ASE band of the ccg-QDs (Figure 5a). Spectrally and angle-resolved Fourier-plane measurements of the fabricated devices reveal an X-shaped pattern of photonic bands, typical of 2nd-order DFB resonators (Figure 6a, bottom). [39] It results from the 1st-order nearnormal Bragg scattering of two waves propagating in opposite directions in the plane of the DFB grating. The coupling between these waves, required for the lasing action, manifests as a sizable "stopband" (Δ sb = 4 nm or 13 meV) near ≈617 nm, which separates the higher-and lower-energy branches of the photon dispersion curve. The photonic modes at the edges of the stopband require the lowest optical gain for the lasing action. Therefore, one of them is usually excited at the onset of the lasing effect. [40] The edge modes are also preferred in the case of the ASE action, as they correspond to the longest effective interaction path of the propagating light with the gain medium. [41] The fabricated devices show good lasing performance at T = 80 K under excitation with short optical pulses (Figure 6b,c). At lower pump intensities (w p < 80 µJ cm −2 ), the surface-emitted PL exhibits a pronounced modulation on its longer-wavelength side with a dip within the DFB stopband and two weak peaks at its edges (Figure 6d, bottom). When w p reaches ≈82 µJ cm −2 , a nearly linear growth of the shorter-wavelength edge peak switches to a fast super-linear growth, which signifies a transition to ASE (Figure 6c and  6d, middle). At w p of ≈115 µJ cm −2 , this peak sharply narrows (from 2 to <0.6 nm) and jumps in intensity, which indicates the transition to single-mode DFB lasing (Figure 6c and 6d,  top). These observations suggest that the developed lowoptical-loss EL devices can support both the ASE and the lasing regimes at T = 80 K.

Electroluminescence from a DFB-Cavity-Based Device
The cavity-based LEDs also exhibit a clear signature of a photonic stopband in the EL spectra. For the device shown in Figure 6e, it is located at ≈ 619 nm and the stop-band width is ≈5 nm (see also Figure S16a, Supporting Information). The EL signal outside the stopband grows approximately linear with j (Figure 6f, red; feature EL 1 in Figure 6e). In contrast, the EL at the edge of the stopband exhibits a fast superlinear growth proportional to j 2.6 (Figure 6f, blue; feature EL2 in Figure 6e). This is accompanied by a pronounced change in the EL spectrum whose peak shifts towards the photonic stop-band edge (Figure 6e, middle). These observations are likely manifestations of optical gain at the stopband-edge mode which leads to both superlinear intensity growth and reshaping of the EL spectrum. The development of optical gain might be assisted by QD charging, [42] common in devices featuring an "inverted" LED architecture. [18] The features observed in the EL regime are broader than those recorded in the case of optical pumping (Figure 6d). This is due to overheating of the active volume at high current densities, as discussed earlier. Thermally-induced optical losses also likely preclude the realization of electrically excited DFB lasing. In fact, the increase of j above 164 A cm −2 leads to the redshift and broadening of the EL spectrum (Figure 6e, top), clear signs of device overheating. The detrimental influence of thermal effects is also indicated by the fact that neither ASE nor DFB lasing is observed at room temperature ( Figure S16b,c, Supporting Information).

Conclusion
In conclusion, we have reported high-j QD EL devices with an integrated DFB cavity that demonstrate optically pumped lasing at LN temperature. This advance is a result of the integrated approach that combines optical gain/loss engineering with the optimization of charge injection. In particular, the use of novel ccg-QDs allows us to boost material gain and simultaneously achieve strong suppression of Auger decay. Further, we engineer a cross-section profile of the refractive index so as to boost the mode confinement factor for the active QD layer. In addition, we reduce optical losses by removing (or modifying) strongly absorbing charge-transport/injection layers such as the ITO cathode (replaced with the L-ITO layer), a ZnO ETL (eliminated completely), and the TCTA/MoO x HTL/HIL (replaced with the all-organic Spiro-2NPB/HAT-CN HTL/HIL). As a result of these concerted efforts, we realize LEDs wherein the QD layer acts as an efficient waveguide amplifier with a large net optical gain of >200 cm −1 at the 1P transition. Importantly, the conducted modifications preserve good electrical characteristics which allow for obtaining ultra-high current densities (up to ≈ 560 A cm −2 ) sufficient for realizing strong (saturated) optical gain.
While the developed devices do not show lasing with electrical pumping, they do exhibit a j-dependent super-linear EL intensity growth at the edge of the photonic stopband. This might be indicative of the onset of the ASE effect which, however, is suppressed at higher j due to device overheating. A Adv. Mater. 2023, 35,2206613  quick increase in the active-volume temperature at high current densities and an associated increase in optical losses is the primary obstacle toward electrically pumped lasing.
This problem could be tackled by, for example, the reduction of device serial resistance (e.g., via introduction of conductive inter-dot "linkers"), improved heat management, and/or c) The plot of the intensity of surface emission versus pump fluence exhibits two thresholds. One is due to the transition to ASE (w th,ASE = 82 µJ cm −2 ) and the other, due to transition to DFB lasing. d) Three representative surface emission spectra illustrating the transition from spontaneous emission (bottom) to ASE (middle) and then DFB lasing in the case of optical pumping with increasing pump fluence. e) When j increases from 79 to 159 A cm −2 , the EL spectrum exhibits a pronounced change in the line shape due to fast (super-linear) growth of the EL feature at the edge of the stopband of the DFBstructure (EL 2 ). This might be indicative of the onset of ASE. A further increase in j leads to broadening and a redshift of the EL spectrum which are the signs of device overheating. f) The "edge-mode" EL intensity (EL 2 in panel "e") shows a super-linear dependence on j (I EL2 ∝ j 2.6 ), which is in contrast to a near-linear dependence of the EL 1 feature (I EL1 ∝ j 1.08 ) not affected by the photonic effects.
further optimization of the optical mode profile to boost net optical gain.

Experimental Section
Materials: Cadmium acetate dihydrate (Cd(OAc) 2 ·2H 2 O, 98%, Aldrich), zinc acetate (Zn(Oac) 2 , 99.99%), oleic acid (OA, 90%, Alfa Aesar), 1-octadecene (ODE, 90%, Aldrich), trioctylphosphine (TOP, 97%, Strem), sulfur (99.999%, Alfa Aesar), selenium (shot, 2-6 mm, 99.998%, Alfa Aesar), chloroform (99.9%, Fisher), ethanol (EtOH, Alfa Aesar), toluene (anhydrous, 99.8%, Aldrich), hexane (anhydrous, 95%, Aldrich), 1-propanol (anhydrous, 99.7%, Aldrich), and octane (anhydrous, 99%, Aldrich) were used as received. Synthesis: The growth of CdSe/Cd x Zn 1-x Se/ZnSe 0.5 S 0.5 /ZnS QDs was conducted using a multi-step procedure illustrated in Figure S1, Supporting Information. To grow CdSe cores, a mixture of 6 mL ODE and 0.2 mL Cd-oleate was loaded into a three-neck flask and degassed at 120 °C for 15 min. The reaction was then saturated with nitrogen and heated to 310 °C. When the temperature reached 310 °C, 0.1 mL, TOP-Se was swiftly injected into the reaction flask. After 30 s, 1 mL TOP was added dropwise over 20 s. After 2 min, 1 mL of solution (A) was added dropwise to the reaction at the 5 mL h −1 rate over 12 min. At the end of the injection, CdSe QDs with the band-edge (1S) absorption peak at 615 nm were obtained. To start the growth of the compositionally graded Cd x Zn 1-x Se layer, 2 mL of Zn-oleate solution was injected at once and 5 mL of solution (B) was added continuously over the course of 75 min at the 4 mL h −1 rate. During this stage, Zn-oleate solution was added shot-wise via three quick injections in the amounts 2, 4, and 2 mL at ca. 18.75, 52.5, and 67.5 min after the start of the graded-layer growth. These time intervals corresponded to 1.25, 3.5, and 4.5 mL of solution (B) fed into the reaction. To grow the ZnSe 0.5 S 0.5 layer, 2 mL of solution (C) was continuously injected with the 1 mL h −1 rate over 120 min. Zn(OA) 2 was added in several shots in the amount of 2 mL per shot for every 0.25 mL of solution (C) fed into the reaction (time intervals between consecutive injections were approximately 15 min). After addition of solution (C) was complete, solution (D) was continuously injected over the course of 30 min at the 1 mL h −1 rate. During this time, 1 mL Zn(OA) 2 was added shot-wise every 0.25 mL of (D) (that is, every 15 min). The last reaction step produced the final ZnS protective layer. To complete the reaction, the heating mantle was removed and the reaction products were cooled down to room temperature. This synthesis resulted in ccg-QDs with the following dimensions: 2.6 nm (core radius), 2.4 nm (graded Cd x Zn 1-x Se layer thickness), 1 nm (ZnSe 0.5 S 0.5 layer thickness), and 0.25 nm (final ZnS shell thickness).
Purification: Purification of the synthesized ccg-QDs QDs was carried out by diluting the content of the reaction flask with 35 mL chloroform and, then, adding 70 mL of ethanol as an antisolvent. This mixture was centrifuged at 6000 rpm for 10 min and, then, the precipitate was dissolved in 10 mL toluene. These solution-based ccg-QD samples were used in spectroscopic studies. For the fabrication of devices, QD samples were further purified using a multi-step procedure. Specifically, 4 mL of ccg-QDs in toluene were mixed with 30 mL acetone. The mixture was centrifuged at 7000 rpm for 15 min. The precipitate was dissolved in 2 mL hexane and, then, after addition of 20 mL of acetone, the mixture was centrifuged at 7000 rpm for 15 min. The precipitate was dissolved again in 1 mL hexane. This was followed by the addition of 15 mL of acetone and centrifugation at 7000 rpm for 15 min. The dry precipitate was weighted and the purified ccg-QD sample was dispersed in octane to obtain the desired concentration (typically, 50 mg mL −1 ).
Fabrication of LEDs: Electrodes made of low-index ITO (L-ITO) deposited onto a glass substrate were purchased from Thin Film Devices Inc. To fabricate an LED, the L-ITO electrode was cleaned via sequential 10-min sonication steps using isopropyl alcohol, acetone, and ethanol. After this procedure, the solvents were removed by "baking" the electrode in a hot oven at 120 °C. Afterwards, 20 µL of ccg-QD solution (50 mg mL −1 ) were spin-coated onto the L-ITO electrode at 2000 rpm for 30 s to form 2 monolayers of the ccg-QDs. This procedure was repeated twice to produce a 6-monolayer-thick ccg-QD film. The QD thickness was experimentally optimized to ensure efficient carrier injection and sufficient optical gain. During the deposition, the film was allowed to dry between consecutive spin-coating steps. Afterwards, a 50-nm HTL of Spiro-2NPB was deposited by thermal evaporation under vacuum (<10 −6 torr) using the evaporation rate of 0.3 Å s −1 . To prepare "currentfocusing" LEDs, a 60-nm thick LiF interlayer was thermally evaporated onto the Spiro-2NPB HTL using a shadow mask with a 50-µm gap. Then, a 10-nm-thick HIL of HAT-CN was deposited by thermal evaporation (0.2 Å s −1 rate). The device was completed with a 100-nm-thick Ag electrode prepared by thermal evaporation (1 Å s −1 rate) through a shadow mask with a 300-µm slit orthogonal to the opening in the LiF interlayer for additional "current-focusing".
Integration of a DFB Cavity: To realize 2nd-order DFB lasing, a 1D grading was engraved in an L-ITO DFB electrode using laser interferometric lithography following the procedures reported elsewhere. [19] The grove depth was 40 nm, the grating period was Λ DFB = 370 nm, and the duty cycle was 0.3. The L-ITO electrode with the integrated DFB structure was used to fabricate an LED following the procedures from the previous section.
Characterization of LEDs: The LEDs were characterized at both room and LN temperatures. For LN measurements, the devices were loaded into a cryostat (Janis ST-100) adapted for electro-optical measurements. A function generator (Tektronix AFG 320) was used to produce squareshaped voltage pulses (0.1 -3.5 V amplitude) at a desired repetition rate. The pulses were amplified by a high-speed bipolar amplifier (HSA4101, NF Corporation) with a 20X gain. The applied voltage was measured using a Tekronix (TDS 2024B) oscilloscope connected to the monitoring port of the amplifier. The current was monitored using the same oscilloscope by measuring a voltage drop across a 10 Ω load resistor on the current return. The device EL emitted through the L-ITO/ glass electrode was measured using an imaging system, consisting of two lenses, a Czerny-Turner spectrograph (Acton SpectraPro 300i), and an LN-cooled charge-coupled device (CCD) camera (Roper Scientific).
Optical Absorption Measurements: Optical absorption spectra were measured using an integrating sphere module of a UV/Vis spectrometer (Lambda 950, Perkin Elmer). ITO and L-ITO samples were measured as purchased. All HTL samples were prepared on bare glass substrates by thermal evaporation at the 0.2 -0.3 Å s −1 rate. For calibration, bare glass was measured and the acquired spectrum was used as a background signal. (2) where the integration volumes extended over the selected layer (numerator) and the entire space (denominator).
Optical Loss Coefficients: Two approaches were used to obtain total loss coefficients. In one approach, the effective refractive index (n eff ), obtained from the boundary mode analysis, was related to the total loss coefficient (α loss ) of a multi-layered stack by In the second approach, the waveguide transmittance (T) was obtained based on the signal at the output port. Then, using the waveguide length (L), the total loss coefficients were calculated based on the Beer-Lambert law from The results obtained by these two methods were different by less than 0.2%.
Spatial Profile of the Electromagnetic Power Density Loss: A crosssectional profile of the power-density loss (optical loss) within a waveguide was calculated from where ω,ε 0 , and ε r ″ were the angular frequency of the propagating wave, the vacuum permeability, and the imaginary part of the relative permittivity, respectively.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.