Enhanced Room Temperature Infrared LEDs using Monolithically Integrated Plasmonic Materials

Remarkable systems have been reported recently using the polylithic integration of semiconductor optoelectronic devices and plasmonic materials exhibiting epsilon-near-zero (ENZ) and negative permittivity. In traditional noble metals, the ENZ and plasmonic response is achieved near their plasma frequencies, limiting plasmonic optoelectronic device design flexibility. Here, we leverage an all-epitaxial approach to monolithically and seamlessly integrate designer plasmonic materials into a quantum dot light emitting diode (LED), leading to a ~5.6 x enhancement over an otherwise identical non-plasmonic control sample. Devices exhibited optical powers comparable, and temperature performance far superior, to commercially-available devices.

Remarkable systems have been reported recently using the polylithic integration of semiconductor optoelectronic devices and plasmonic materials exhibiting epsilon-near-zero (ENZ) and negative permittivity.In traditional noble metals, the ENZ and plasmonic response is achieved near their plasma frequencies, limiting plasmonic optoelectronic device design flexibility.Here, we leverage an allepitaxial approach to monolithically and seamlessly integrate designer plasmonic materials into a quantum dot light emitting diode (LED), leading to a 5. 6 × enhancement over an otherwise identical non-plasmonic control sample.Devices exhibited optical powers comparable, and temperature performance far superior, to commerciallyavailable devices.
Polylithic integration of semiconductors with plasmonic materials exhibiting epsilon-near-zero (ENZ) and negative permittivity response offers exciting prospects for enhancing lightmatter interactions to increase optoelectronic device performance and realize new functionality [1][2][3][4][5].Unfortunately, with traditional noble metals, ENZ and plasmonic response are achieved only at limited spectral positions near their plasma frequencies, typically in the UV/visible wavelengths [6][7][8][9].This limits plasmonic optoelectronic device design and spectral flexibility, often leading to weaker performance improvements than predicted [10].In the mid-infrared (mid-IR), however, the picture is very different, and the opportunity exists for designer plasmonic and quantum-engineered optoelectronic materials in the same material systems.
Molecular beam epitaxy (MBE) allows for the engineering of so-called "designer metals," relatively low-loss plasmonic materials with plasma wavelengths that span the mid-infrared with the control of doping concentration during growth [11].The epitaxial materials typically employed as designer metals are narrow bandgap materials, such as InAs or InAsSb, whose small effective masses allow for plasma wavelengths as short as λ p ∼ 5 µm, where λ p = 2πc m * o b /e 2 n e [corresponding to the epsilon-near-zero (ENZ) wavelength where the doped semiconductor transitions from a lossy "dielectric" to a plasmonic "metal"], with m * being the electron effective mass, b the background dielectric permittivity of the semiconductor, and n e the electron doping concentration.Additionally, and perhaps even more tantalizing, the InAs(Sb) material system is home to a variety of mid-IR emitting optoelectronic materials whose emission overlaps spectrally with the plasmonic and ENZ behavior of heavily doped InAs(Sb).Quantum cascade lasers, interband cascade lasers, superlattice-based emitters and nanostructured quantum dot materials, provide opportunities to monolithically integrate -with the atomic precision enabled by epitaxial growth -quantum-engineered emitters with these designer plasmonic materials [12][13][14][15][16].In fact, a variety of quantum cascade lasers have leveraged noble metal waveguides to confine optical modes in lasers, and have even employed doped semiconductors to compensate unwanted positive group velocity dispersion [17][18][19].InGaSb type-II quantum dots grown in an InAs matrix are a particularly intriguing option due to their excellent temperature performance, a result of decreased Auger recombination rates, and increased radiative transition rates, stemming from their low dimensionality [20,21].The wavelength flexibility of both the designer metals and In(Ga)Sb quantum dots allow for an epitaxially-integrated architecture for cavity enhancement, providing a highly-engineerable mechanism for improved output and efficiency across the mid-IR.Mid-IR LEDs have long lagged behind lasers in the same wavelength range.This is partially because the most efficient mid-IR lasers are weak sub-threshold emitters, making mid-IR LEDs (that utilize mid-IR laser active regions) extremely inefficient [22].
Two LED structures, each containing five layers of In(Ga)Sb quantum emitters embedded in an InAs PIN-junction, were grown by molecular beam epitaxy on doped n-type InAs 'vir- tual substrates': the first, a control sample, was grown on a moderately-doped n-InAs backplane (n-doped ∼2 x 10 18 cm −3 ), and the second, our n ++ device that monolithically integrates electrically pumped material and long wavelength metal, on heavily doped n ++ InAs (active concentration ∼1.2 x 10 20 cm −3 ).In both cases, the top of the virtual substrate lies 100 nm below the first In(Ga)Sb emitter layer; band diagrams and layer structures of the devices, as well as bright-field transmission electron microscope (TEM) images of the quantum emitters, are shown in Fig 1 .Each active region was capped with a lattice-matched p-AlAs 0.16 Sb 0.84 barrier, which has a significant conduction band offset with InAs, confining electrons injected into the active region and mitigating parasitic surface recombination [23].To promote current spreading and decrease contact resistance the barriers were capped with a p ++ InAs layer (doped ∼1 x 10 19 cm −3 ).Both samples were fabricated into mesa emitters (700 µm x 600 µm) with window contacts for electroluminescence experiments.
The reflectance spectra of both devices were measured [Fig.2(a)] and fitted using a transfer-matrix method, treating the doped semiconductor layer as a Drude plasmonic material with permittivity, with the plasma frequency (ω p = 2πc/λ p ) and scattering rate (γ) of the virtual substrate as the fitting parameters, from which we extracted the plasma wavelength and scattering rate of the n ++ virtual substrate of the n ++ monolithic device (λ p = 6.25 µm, γ = 5 x 10 12 Hz).Qualitatively, the heavily doped sample showed the spectral features expected of a three-layer air-dielectric-Drude metal system: the strong reflectance at long wavelengths and the noticeable reflection dip at 6 µm corresponding to a leaky λ/4n cavity mode set up by the three-layer system [24].
The low temperature photoluminescence (PL) spectrum of the control device [Fig 2(b)] was dominated by the broad continuous QD emission spectrum centered at λ = 5 µm, with PL spectrum cut off at λ = 3.6 µm by a long pass filter, used to block the InAs bandedge and pump laser.The PL from the highlydoped (n ++ ) device showed a markedly different spectrum [Fig

2(b)]
, with a clear enhancements (suppressions) of the control PL spectrum commensurate with the reflection minima (maxima) observed for the n ++ device (which, again, come from the leaky cavity formed by the air/dielectric/Drude metal system).The most prominent occurrence of this is at the long wavelength tail of the QD emission, where the modulation of the n ++ emission included a strong enhancement of emission, by a factor of approximately 3.9 × in photoluminescence.
To elucidate this effect, we modeled our system using a Dyadic Green's function formalism, incorporated into a transfer matrix method solver [25][26][27], positioning our emitter at the five QD layer positions.These calculations provide us with the position-and wavelength-dependent Purcell enhancement (P(λ)), as well as the Purcell-corrected Poynting flux representing the total energy emitted by the point dipole to the far field (S z (λ)), shown in Fig. 3(a).Note the anti-correlation between the strongest Purcell enhancement (at the surface plasmon wavelength) and our S z (λ), a result of the bound nature of the surface plasmon modes at the n ++ /undoped InAs interface which dominate the strong Purcell enhancement.
While Purcell enhancement modifies the quantum efficiency of our QD emitters (the fraction of recombination events in a QD resulting in emitted photons), qi = Pq i /[(P − 1)q i + 1] with q i being the quantum efficiency of an isolated QD, only a fraction of emitted photons escape the leaky LED cavity and the detector in the far field.Therefore, the total measured emission enhancement can be estimated by comparing the product qi (λ) • S z (λ) in the plasmonic structure to its counterpart in the control structure.The overall enhancement of QD emission results from the interplay of the short wavelength tail of the Purcell enhancement (which peaks at the wavelength of the surface plasmon Fig. 3. Spontaneous emission enhancement from In(Ga)Sb for a series of five In(Ga)Sb emitters 100 nm above the n ++ /undoped interface.From this we extracted (a) the Purcell enhancement factor as a function of wavelength (red) as well as the z-component of the Poynting vector, in black, as a function of dipole emitter wavelength.(b) Control EL emission (dashed) for three possible intrinsic quantum efficiencies, (green) q i = 1%, (blue) q i = 5%, (red) q i = 25%, and modeled overall far field emission of the highly-doped device (solid).
wavelength of the n ++ /undoped InAs interface) and the cavity formed by the layered system.Fig. 3(b) illustrates the evolution of the emission spectra of the plasmonic and control devices for the range of intrinsic quantum efficiencies from 1% to 25%.For each quantum efficiency, the spectrum of the plasmonic device is formed from the product of total emission enhancement (defined above) and the spectrum of the control counterpart.The experimental and predicted emission is shown in Fig. 4(c) (for q i = 5%).The calculations that predict the observed EL enhancement of the emission are quite accurate when compared to experimental results.
Transitioning a cavity-enhanced emitter architecture from optical pumping to an electrically-driven device has traditionally proven problematic.In the case of an optoelectronic material (semiconductor emitter) coupled to a plasmonic structure, the electronic requirements of the emitter (contact layers, minimizing current density and Ohmic heating, etc.) often conspire to significantly limit the possible enhancement, in addition to adding significant parasitic effects detrimental to device performance [28].In the device described here, however, the monolithic nature of the device architecture, combined with the longer wavelength operation, allow for placement of the emitters in the near field of the plasmonic material without compromising device operation.Temperature-dependent electroluminescence (EL) of the control sample [Fig.4(a)] showed a significant decrease in emitter efficiency, expected for narrow bandgap emitters [29].The n ++ device, however, showed remarkably different temperature dependence, as shown in Fig. 4(b).While the short wavelength feature of the n ++ device (centered around 4.5 µm) decayed in intensity in a similar manner to the control sample's emission, with intensity decreasing 3 × from 78 K to 298 K, the long wavelength emission peak from the n ++ device showed only a 13% decrease in emission intensity between the low temperature and room temperature measurements, which is highly unusual for mid-IR emitters.Applying our enhancement model to the control sample emission and assuming q i = 5% we can accurately predict the room temperature EL spectrum of the n ++ device.The electroluminescence showed a strong enhancement of emission at peak wavelength, increasing by a factor of approximately 5.6 ×.The calculations that predict the observed EL enhancement of the emission is quite accurate when compared to experimental results.
Light-current-voltage (LIV) measurements of the two emitters demonstrated (Fig. 5) a near uniform enhancement of emission across all pump currents measured, as well as peak upper hemisphere emitted power of 1.45 µW, values already comparable to state-of-the-art commercial mid-IR LEDs at 6 µm [30].
We report the first all-epitaxial integration of a plasmonic "designer metal" and emitter to produce an electrically-pumped, cavity-enhanced LED.The use of a plasmonic virtual substrate resulted in a significant spectral modulation of emission intensity, little to no decrease in emission intensity over a range of 200 K, and a strong enhancement of emission, which we modeled and determined to result from a combination of Purcell enhancement and the leaky cavity formed by our three layer system.The plasmonic virtual substrate device showed upper hemisphere powers of 1.45 µW, comparable to state-of-the-art mid-IR LEDs [30].The device architecture presented here offers a new approach to enhance mid-IR optoelectronic devices, taking advantage of our ability to engineer both our designer metals and our quantum emitters in the same epitaxial material platform.In addition to the potential practical applications for mid-IR optical systems, the devices introduced here offers unprecedented opportunities for exploring and engineering lightmatter interactions and new device architectures leveraging monolithically-integrated designer plasmonic materials.

Fig. 1 .
Fig. 1.Band diagrams for unbiased (a) cavity-enhanced In(Ga)Sb LED and (b) control In(Ga)Sb LED.Insets: layer structure of each device.(c) Representative bright-field TEM of the cavity-enhanced In(Ga)Sb LED.Arrows indicate strain fields from discrete quantum dots.

Fig. 4 .
Fig. 4. Temperature-dependent EL spectra of the (a) five layer control device and the (b) cavity-enhanced device with the n ++ InAs backplane (c) Modeled (dashed, assuming q i = 5%) and experimental room temperature comparison of n ++ (black) and control (red) devices showing strong cavity enhancement around 6 µm.