Introduction

High-efficiency mid-wavelength infrared (mid-IR, 3–5 µm) optoelectronics, including light emitting diodes (LEDs) and photodetectors, are attractive for spectroscopy, imaging, and gas sensing1,2,3,4,5,6,7,8,9. Black phosphorus (BP), a narrow bandgap van der Waals semiconductor (Eg ~ 0.33 eV for bulk BP), shows superior photoluminescence (PL) quantum efficiency as compared to conventional III-V and II-VI semiconductors of similar bandgap due to significantly lower Auger recombination rate (over five orders of magnitude lower)10,11,12,13,14,15,16 and surface recombination velocity (over four orders of magnitude lower)11,17,18,19. The unusually low Auger recombination rate arises from more symmetric electron/hole effective masses in BP, while the low surface recombination is due to its self-terminated, layered crystal structure9,10,11. A substantial challenge in BP optoelectronics for practical applications is the ease of oxidation in the air because of the lone pair electrons in P atoms, which constitutes the top of the valence band and favors bonding with oxygen20,21. This effect is dramatically accelerated in LEDs, where high current densities with fast carrier recombination processes lead to rapid material and device degradation.

Various passivation techniques have been employed to enhance the stability of BP using atomic layer deposition (ALD) oxide layers22,23,24,25, h-BN encapsulation24,26,27, PxOy self-passivation23, and parylene capping28. Notably, long-term stability has been demonstrated in BP field-effect transistors (FETs) within the different monitoring times;22,24,25,26,27 specifically, Al2O3 encapsulation via ALD enables a reliable FET operation for up to eight months25. Those approaches effectively prevent device degradation; however, their characterizations are mostly limited to FETs and optical spectroscopies such as PL and Raman, under intermittent electrical or optical stress, if any. The reliability and lifetime of LEDs and FETs for any technology are vastly different. Generally, LEDs require higher current densities than FETs, resulting in more severe electrical and thermal stress. In this regard, measuring the lifetime of BP LEDs and developing pathways to improve them are of significant practical importance.

Here, we present packaging technology for BP-based mid-IR LEDs. Nitrogen sealing was utilized for packaging the BP LEDs together with the ALD Al2O3 passivation layer. Although electroluminescence (EL) intensity from bare BP LED quenches just in few seconds owing to the oxidation in air, packaged LEDs show a robust continuous operation at room temperature (RT) as long as our measurement range for 100 h. To estimate the device lifetime at RT, an accelerated stress test is investigated by heating the devices during the LED operation. Notably, RT lifetime (half-life) is extrapolated to be ~15,000 h with an initial power density of 340 mW/cm2. Our results motivate further efforts to improve the packaging technology to provide practical applications based on BP, not limited to LED but including various functional devices such as IR detectors, gas sensing, and optical communications.

Results

Air stability of optical quality in black phosphorus

Before investigating the BP LEDs, PL stability was studied to see the effect of passivation and sealing. Fig. 1a–c shows different conditions of BP samples: bare, N2 sealed, and with Al2O3 passivation. BP flakes were mechanically exfoliated on SiO2/Si substrate inside a glove box. For the N2 sealed sample, BP was placed on the chip carrier with a quartz cover sealed by an ultra-violet (UV) curable resin in the glove box. An Al2O3 (20 nm) layer was deposited by ALD at 200 °C to encapsulate BP flakes. Recently our study has shown that air exposure has an effect on the PL degradation for relatively thinner BP below 12 nm, while that is not the case in thicker bulk flakes: for thick BP above ~60 nm, PL intensity was constant even with the surface oxidation suggesting the self-limited oxidation with PxOy formation and minor effect of oxide layer on surface recombination velocity11. Here we focus on BP flakes as thin as 10 nm to observe a pronounced oxidation effect on optical performance. A calibrated PL quantum yield (QY) was measured using a 638 nm excitation laser (see details in Methods and Supplementary Note 1). Fig. 1d shows internal PL QY as a function of the generation rate for the control sample of bare BP, as exfoliated and after 1 week air exposure. Typical PL QY in as exfoliated BP was ~0.2% at G = 1 × 1027 cm−3s−1, consistent with previously reported thickness-dependent PL QY in BP11. Power-dependent PL measurement helps us capture the different carrier recombination pathways. In the semiconductor with low background doping concentration (n = p), the ABC recombination model can be expressed as follows,

$${QY}=\frac{B({n}^{2}-{n}_{i}^{2})}{\left(A+\frac{2S}{d}\right)\frac{{n}^{2}-{n}_{i}^{2}}{n}+B\left({n}^{2}-{n}_{i}^{2}\right)+2{Cn}({n}^{2}-{n}_{i}^{2})}$$
(1)

where ni is the intrinsic carrier concentration; A, B, and C are Shockley-Read-Hall (SRH), radiative, and Auger recombination coefficients, respectively; and 2 S/d is a surface recombination term, where d and S are material thickness and surface recombination velocity, respectively. SRH process is less dominant than other recombination processes11. Thus, surface recombination is dominant at the lower generation rate, while QY is limited by the Auger recombination at the higher generation rates. There was no decrease in PL QY of as exfoliated BP at a high generation rate, though PL QY increased monotonically, indicating that surface recombination is dominant in this regime (Fig. 1d). This nonradiative surface recombination is probably due to native defects on the BP surface, whose rate increase with decreasing BP thickness (\(\propto 1/d\)) according to Eq. (1). After the sample was exposed to air for one week, the PL QY was quenched in all generation rates. The PL peak position blue shifted from 0.34 eV to 0.43 eV, indicating top layers were oxidized so that effective BP thickness decreased (Supplementary Fig. 1a). BP color difference in optical microscopic images is also caused by the reduction of the effective thickness (Supplementary Fig. 1b). Using N2 sealing or Al2O3 encapsulation, BP flakes show comparable PL QY as the bare sample indicating that optical coupling change and additional defects during the process are negligible (Fig. 1e, f). Even after one month, both samples show no apparent PL QY quenching (Supplementary Fig. 2). Note that there was no PL QY degradation at all generation rates, especially at the lower generation rate, indicating surface oxidation was effectively prevented so that the surface recombination velocity was constant. Unlike the bare sample, no degradation was confirmed from the optical images (Supplementary Fig. 1c).

Fig. 1: Air stability of photoluminescence quantum yield (PL QY) in black phosphorus (BP).
figure 1

a–c Schematic of exfoliated BP with different conditions: as exfoliated bare BP on SiO2/Si (a), with N2 sealed (b), and Al2O3 passivated (c). d–f PL QY in BP as a function of generation rate before (filled symbols) and after (open symbols) air exposure, with different conditions: bare (d), with N2 sealed (e), and Al2O3 passivated (f). The BP thickness was fixed at ~10 nm. All the samples were exposed in the air with a relative humidity of 50 ± 5% under dark condition at room temperature (RT).

Full size image

Packaged black phosphorus LEDs

To evaluate the air stability of optoelectrical devices, we fabricated LEDs based on BP, which operate under more severe stresses, i.e., higher current and voltage. We prepared different BP LED structures: bare, N2 sealed, with Al2O3 passivation, and incorporating both N2 sealing and Al2O3 passivation (hereafter termed packaged). Fig. 2a and b shows schematic and photograph of the packaged device structure of BP LED, respectively. The LED comprised BP (p-type) and MoS2 (n-type) heterojunction. In contrast to the conventional epitaxial materials (e.g., HgCdTe) that are only compatible with a limited platform owing to the lattice matching, van der Waals semiconductors can be transferred on almost any material. MoS2/BP heterostructure was dry transferred onto the optical cavity with gold and Al2O3 layers, enhancing the Purcell factor in BP and light outcoupling for mid-IR emission29. To design each layer thickness, finite-difference time domain (FDTD) simulations were performed with the fixed MoS2 thickness of 30 nm, obtaining optimal BP and bottom Al2O3 thicknesses of 10–40 nm and 360 nm, respectively (Supplementary Fig. 3). In our previous study, the output power density was found to be maximized with the BP thickness at around 40 nm29. To fairly compare LED performances under different conditions, BP and MoS2 flakes were selected with thicknesses of 30–40 nm and ~30 nm, respectively. Their heterostructures showed similar I–V characteristics and LED performances (Supplementary Fig. 4). A 20 nm thick Al2O3 passivation layer was deposited via the ALD process, followed by wire bonding on the chip carrier and N2 sealing with an optical window. Device performances were carefully investigated before and after the processes to confirm no degradation in MoS2/BP LEDs during these procedures, specifically the ALD process that uses the highest process temperature at 200 °C. Fig. 2c shows I–V characteristics in the same MoS2/BP device before and after ALD. Both I–V curves represent a typical Schottky diode feature; no significant change was found after packaging. Also, the EL intensity was confirmed to be sustained after ALD (Fig. 2d). Note that EL measurements were carried out in the vacuum to prevent an unexpected degradation in the air. These results validate our process packages BP without compromising its optical quality.

Fig. 2: Packaged mid-infrared light-emitting diode (LED) based on BP.
figure 2

Schematic (a) and photograph (b) of packaged MoS2/BP LED. Heterostructure of MoS2/BP was fabricated on an optical cavity, consisting of gold mirror and Al2O3. The device was passivated with Al2O3 layer via atomic layer deposition (ALD), followed by N2 sealing with quartz window and ultra-violet (UV) curable resin. The scale bar represents 1 cm. I–V curves (c) and electroluminescence (EL) spectra (d) of MoS2/BP LED before and after ALD process.

Full size image

Improved operational lifetime via packaging

We next investigated time-resolved EL measurements for different BP LEDs. EL signal was measured in a Fourier-transform infrared (FTIR) system by applying a voltage at 5 kHz frequency and 50% duty cycle. In the commercial products of mid-IR LEDs based on semiconductor alloys, the maximum currents are estimated to be 6–14 A/cm2, dividing the maximum forward current by the emission area30,31,32. Our device size and injected current density were fixed at ~400 µm2 and ~75 A/cm2, respectively, which is a sufficiently high current density compared to semiconductor alloys or BP based LEDs9,29,30,31,32. The carrier density was estimated by dividing injected current by the area of heterojunctions so that the actual carrier density can be lower considering parasitic resistances in BP and MoS2 between electrodes and hetero regions. By calibrated EL quantum efficiency (QE) measurements for BP LED, a typical output power density and external QE were determined to be 340 mW/cm2 and 1.4 %, respectively. Those specs are consistent with previously reported values for the BP LEDs with the same structure29. Fig. 3a shows normalized EL intensities (L/L0) of MoS2/BP LED with different conditions operated in the air at RT. The bare sample’s EL was quenched in a few seconds due to the oxidation in the air. To visualize the BP degradation, time resolve optical microscope imaging was performed on bare BP LED (Supplementary Fig. 5 and Supplementary Movie 1). The degradation starts from the air-exposed BP region, severe damage was observed in a few seconds, and the exposed BP layer was entirely degraded after eight seconds. BP oxidation is activated by electrical stress and Joule heating under the relatively high current density of ~75 A/cm2. N2 sealed or Al2O3 passivated devices reveal better endurance than bare ones, though their lifetime is limited to several hours. This discrepancy with the PL measurements, where the PL QY does not degrade after one month (Fig. 1e, f and Supplementary Fig. 2), means that LED operation is much more severe than the atmospheric exposure with no external stress to activate the chemical reactions. No spectrum shape change was observed during the degradation, showing a constant emission wavelength (Supplementary Fig. 6a). After the LED operation for two hours, the Raman signal from BP became poor compared to that from MoS2, indicating only BP layer was damaged during the LED operation (Supplementary Fig. 6b). Even in the carefully prepared sample, oxygen residues can be within N2 sealing, or oxygen can diffuse through Al2O3 layers. To further stabilize BP LED, we performed the packaging using both N2 sealing and Al2O3 encapsulation. The oxidation was effectively suppressed, resulting in the non-degradable EL emission as far as our measurement was carried out for 100 h, as shown in Fig. 3a and Supplementary Fig. 7. The time evolution of EL spectra during continuous operation also shows invariant peak positions and widths (Fig. 3b). These results indicate that it is necessary for robust BP optoelectronic devices to make an inert environment as well as encapsulation with oxygen diffusion barriers.

Fig. 3: Lifetime measurement of BP LED at RT.
figure 3

a Time dependent EL intensity curves of BP LEDs with different conditions: bare, N2 sealed, with Al2O3 passivation, and packaged (Al2O3 and N2). Normalized EL intensity (L/L0) is obtained by dividing integrated EL intensity by initial intensity. b Time evolution of EL spectrum from BP. Input current density was fixed at 75 A/cm2. All LED measurements were performed at RT in an ambient lab condition with a relative humidity of 40 ± 5%.

Full size image

Discussion

An accelerated stress test was conducted for further quantitative evaluation of the operating lifetime. In situ EL measurement was carried out for BP LED by heating the device from the backside of the carrier chip using a heater. Operation temperature was changed between 80–140 °C, much lower than the melting point of BP33. In the air environment with oxygen, it has been reported that the endothermic process for bulk BP starts at 150 °C, based on the differential scanning calorimetry measurements34, thus we can assume the degradation mechanism in BP LED is identical within the present temperature range. Fig. 4a shows normalized EL intensity decay at different temperatures. The injected current density was fixed at 75 A/cm2 and the measurement time constant was set at 1 s. A stretched exponential decay model commonly used to fit the LED luminous decay35,36,

$$\frac{{L}_{0}}{L}={{\exp }}\left[-{\left(\alpha t\right)}^{\beta }\right]$$
(2)

where t is the LED operation time in the unit of hours; β is a stretching constant independent of initial EL intensity; and α is a temperature dependent parameter, which can be written as follows according to the Arrhenius equation,

$$\alpha=A{{\exp }}\left(-\frac{{E}_{a}}{{kT}}\right)$$
(3)

where A is the pre-factor, Ea is the activation energy, and k is the Boltzmann constant. By using Eqs. (2) and (3), experimental results of the acceleration lifetime test can be globally fitted with three fitting parameters, Ea, A, and β. Based on Eq. (2), the fitting curves well reproduced the experimental results, as shown in Fig. 4a and Supplementary Fig. 8. The fitting parameters Ea, A, and β are determined to be 0.96 eV, 6.6 × 1011, and 0.60, respectively. The activation energy of BP is comparable to that of III-V LEDs (Ea = 0.5–0.9 eV); and higher than organic LEDs (OLEDs) (Ea = 0.2–0.6 eV)37, which is consistent with the fact that the activation energy of oxygen chemisorption on BP (Ea = 0.5–0.7 eV)20,38 is higher than that of chemical reactions in organic materials. When the temperature increased above 160 °C, the activation energies were smaller than that in the temperature range of 80–140 °C, indicating the degradation mechanism differs from the lower temperature regime (Supplementary Fig. 9). From Eqs. (2) and (3), lifetime as a function of L/L0 is derived as follows,

$${t}_{L/{L}_{0}}=A{\left[{{{{{\rm{ln}}}}}}{\left(\frac{L}{{L}_{0}}\right)}^{-1}\right]}^{\frac{1}{\beta }}{{\exp }}\left(\frac{{E}_{a}}{{kT}}\right)$$
(4)
Fig. 4: Accelerated lifetime measurement for packaged BP LED.
figure 4

a Normalized EL intensities of BP LEDs operated at different temperature: 80 °C, 100 °C, 125 °C, and 140 °C. Dashed lines show stretched exponential decay model with a global fitting with parameters of Ea, A, and β, which are determined to be 0.96 eV, 6.6 × 1011, and 0.60, respectively. The coefficient of determination (R2) for each curve is shown in Supplementary Fig. 8. b Lifetime as a function of inverted temperature (T−1) for BP LED with different conditions: bare, N2 sealed, with Al2O3 passivation, and packaged (Al2O3 and N2). The lifetime is defined as the device operation time required for the EL intensity to become 50% of the initial brightness. Lifetime of packaged LED at RT (25 °C) was estimated from the fit as shown in a solid line.

Full size image

Thus the half-lifetime (L/L0 = 0.5) can be plotted as a function of inverted temperature. The lifetime at RT (25 °C) is extracted as t0.5 ~ 15,000 h, as shown in Fig. 4b. After the degradation at RT and even high temperature, no emission peak shift or additional peak (0.1–0.8 eV) were observed (Supplementary Figs. 6 and  10). These results guarantee a constant emission center wavelength within the present degradation level. The typical lifetime of commercially available mid-IR LEDs based on narrow gap semiconductor alloys (λpeak = 3.8 µm) is 80,000–100,000 h30,32. Although further improvement is required for BP-based LED toward the practical applications, we emphasize that the lifetime was prolonged from a few seconds to years-long by using a simple and scalable packaging process that is easy to scale up. Besides LEDs, our technique will be helpful for other functional devices such as photodetectors and FETs. Given that the first OLED was only 100 h life39 that eventually reached a sufficient level for authentic commercial products in the 2010s (t0.5 ~ 400,000 h)40, a remarkable improvement is expected in BP LED by developing better packaging technologies. A UV curable resin encapsulation, commonly used for OLEDs and other materials, can be applied to BP LEDs as a less-stress and scalable technique36,41.

We have demonstrated packaged BP LEDs utilizing ALD oxide encapsulation with N2 sealing. High current injection activates the reaction of BP with oxygen in the air, which deteriorates the luminescence efficiency much more quickly than mere exposure to the air without thermal or electrical stresses. Without packaging, the EL intensity was quenched in a few seconds due to the rapid oxidation. BP LED lifetime was dramatically improved with packaging to the operating lifetime t0.5 ~ 15,000 h (extrapolated) at RT. Although BP itself is unstable, our results have proved that BP is optically robust with simple and scalable packaging technologies useful for practical devices in mid-IR optical applications.

Methods

Device fabrication

BP LEDs were fabricated on Al2O3/Au optical cavities. Au (100 nm) and Al2O3 (360 nm) films were deposited on SiO2/Si substrate by electron beam (EB) evaporator and sputtering, respectively. BP and MoS2 (HQ Graphene) were mechanically exfoliated on 50 nm SiO2/Si. BP and MoS2 flake thicknesses were identified by optical contrast and atomic force microscope height profile. MoS2 flake was picked up with poly(methyl methacrylate) (PMMA) and transferred onto the BP flake, followed by a post-baking at 170 °C for 90 s. BP flakes (30–40 nm) were picked up with MoS2 (~30 nm) and transferred onto the optical cavity. MoS2/BP heterostructure covered with PMMA was annealed to improve the adhesion on the Al2O3 surface. PMMA layer was removed by dichloromethane treatment. Contact electrodes of 50 nm Ni were fabricated by EB lithography and thermal evaporation. For the top Al2O3 passivation layer formation, a buffer layer of SiOx was deposited via EB evaporation. Al2O3 (20 nm) passivation layer was deposited via ALD at 200 °C. LEDs were bonded on the chip carrier. To eliminate the absorbed oxygen and water on the device surface, annealing was performed at 200 °C in an ultra-high vacuum with a base pressure of 5 × 10−8 Torr. After the annealing, the device was sealed with a quartz substrate and UV-curable resin. The annealing and sealing were performed inside a glove box with an oxygen level below 0.1 ppm.

Electrical and optical characterization

PL and EL spectra were measured by an FTIR spectrometer (iS50, Thermo Fisher) with a liquid N2-cooled HgCdTe detector was used9. For PL measurement, a 638 nm excitation laser was used. LED devices were operated by applying a voltage at 5 kHz frequency and 50% duty cycle. An accelerated degradation test was performed by heating the substrate from the backside of the chip carrier by using a polyimide flexible heater. The temperature was calibrated by directly measuring the surface temperature of the device. Raman spectra were measured by a Raman microscopic system (Horiba Labram HR Evolution) using a 532 nm excitation laser.