Improved Long-Term Reliability of a Silica-Encapsulated Perovskite Quantum-Dot Light-Emitting Device with an Optically Pumped Remote Film Package.

In this study, inorganic perovskite (CsPbBr3) quantum dots are wrapped in SiO2 to provide better performance against external erosion. Long-term storage (250 days) is demonstrated with very little changes in the illumination capability of these quantum dots. While in the continuous aging procedure, different package architectures can achieve very different lifetimes. As long as 6000 h of lifetime can be expected from these quantum dots, but the blue shift of emission wavelength still needs more investigation.

Introduction

Since its first introduction in the past decade, halide perovskite-related materials have gathered intense attention due to their high efficiencies in illumination and photoabsorption.[1,2] Highly efficient perovskite solar cells have achieved 25.1% of power conversion efficiency recently,[3] and the extraordinary light-emitting capability of the perovskite material cannot be overlooked when the candidate for the next-generation light source is considered. While the original organic halide composites (CH3NH3PbX3, X = I, Br, Cl) are effective in providing different colors, a new breed of inorganic metal halide materials, such as CsPbX3, where X = I, Br, Cl, poses a new opportunity for the device.[4] In addition to different colors provided by different compositions, it is also possible to exploit the quantum confinement effect on these materials to realize the tunability in color via different sizes of nanoparticles of metal halide perovskites.[4] Furthermore, the color purity provided by perovskite quantum dots (PQDs) also prompts a much wider color gamut coverage on the CIE color space, which is greatly desired for the next-generation display.[4−6] However, they are very sensitive to environmental factors like temperature, oxygen, light, water, etc., and various methods have been implemented to extend the lifetime of the QDs. One of the most common methods is to coat these PQDs with a protective layer that is made of either polymer, silica, or other substances.[7−11] Among these methods, coating the QDs with SiO2 is one of the potential choices. The coating process can be integrated into the synthesis procedure, and the protection effectiveness of this SiO2 material is well known in the semiconductor industry. At the same time, the silica shell can also prevent anion exchange if a further mixture of different species of perovskite QDs is necessary.[12,13] In this paper, the cesium lead tribromide (CsPbBr3) perovskite quantum dots (PQDs) are coated with SiO2 and continuously tested under extreme conditions to explore the issues of stability. Different package methods are also adapted to alleviate or improve the lifetime of these QDs in this paper. The filling substances in the package can play an important role in heat dissipation, which might affect the lifetime of the PQD eventually. Some previous results have shown that hexagonal boron nitride (BN) can help to reduce the QD composite temperature,[14] and we will explore the effectiveness of using boron nitride particles in this paper as well. Most importantly, the long-term evolution of photonic properties of these QDs will be reported and analyzed.

Results

Performance of PQD Light-Emitting Diodes (LEDs)

The first thing we should examine is the quality of the remote film and in-chip devices. The remote film was made manually in this study, and it is prone to be not as ideal as the machine-fabricated ones. First, the color of the film can be examined. From Figure a–c, no significant color variation can be found in our PQD film, which can be regarded as a good sign of a thorough mixture of PQD and PDMS. The detailed device picture can be seen in Figure a,b, and the side length of the PQD@SiO2 film is around 4 mm. In Figure c, a flat surface of an in-chip-type device can be observed. From the thickness measurement under an optical microscope, we found a remote film thickness distribution, like the histogram shown in Figure d. A Gaussian fitting can be found to perfectly match with experimental results. The mean thickness is 888.5 μm, and the linewidth of the distribution is 115.5 μm. The picture in Figure b indicates that the overall shape of the remote film is close to a rectangle, and the variation in thickness happens randomly across the whole film. Thus, the surface should be rough due to the manual process of its fabrication and the roughness should be close to the linewidth of distribution in Figure d. The controlled factors in a remote film are the quantity of the quantum dots, the total weight of the film, and the volume of the film in this study, and they can be achieved by the aforementioned molding step. Via these factors, we can control the approximate thickness of the remote films in this study and the random roughness of the QD film can facilitate the uniform scattering of incident photons.[15] We believe that the comparison in their own lifetime is valid because the aged result is compared to its own initial value, and it is the comparison of the same device that we analyzed in this paper.

Figure 1

Device under an optical microscope: (a) top view and (b) bird’s-eye view of a remote film device and (c) bird’s-eye view of a fresh in-chip device. (d) Thickness histogram of the remote film. (Photograph courtesy of Yang, Jhen Jia. Copyright 2020).

Figure a shows the current-dependent electroluminescence (EL) spectra of a remote film device with 0.5 mg of BN particles. The two-peak feature is due to the partial conversion of UV photons, and the corresponding photonic conversion efficiency (PCE) can be calculated as follows[16,17]where IexQD and Iexref are the integrated intensities of the UV LED excitation source with and without the PQD layers and IemQD and Iemref are the intensities of the visible band (but excluding the UV band) with and without PQD layers, respectively. In this experiment, the PCE values vary among different types of packaged devices. For the in-chip type, the PCE values are around 41 and 15% for BN = 0 and 2 mg cases, respectively. For the remote file type, the PCE is around 7.3% for the BN-doped devices at best. The detailed information can be seen in Table . The efficiency of the device generally decreases as the current increases due to the increasing temperature in the package. The current-dependent luminous efficiency can be found in the Supporting Information (Figure S1).

Figure 2

(a) Current-dependent spectra of a remote film device with 0.5 mg of BN doping. (b) Normalized efficiencies of no-BN- and 1 mg BN-doped devices vs currents.

Table 1

Summary of Measured Parameters for All Devices

type of package	in-chip		remote film
BN weight (mg)	0	2	0	0.5	1	2
PCE @ 20 mA (%)	41.04	14.94	8.74	7.29	2.64	4.63
luminous efficiency (lm/W)	60.19	26.05	13.48	12.72	4.22	8.01

Another important parameter that can be extracted is the luminous efficiency (lm/W) that also takes the electrical power into account. The best result we obtained is from the in-chip-type device without any h-BN particles, and it is about 60 lm/W. This value is comparable to some of the best devices reported previously.[13] The corresponding efficiency in the remote film-type device is limited due to the installation of the copper tape and limited opening window on top of the lead frame. The hole at the center of the copper tape is 2.5 mm in diameter that can prevent some UV photons from absorbing by the PQD layer but by the copper tape. The remote film-type device with BN doping shows its best result in BN = 0.5 mg case, and it is around 12.5 lm/W. The device without BN doping still has the best efficiency of 13.48 lm/W like their counterparts in the “in-chip” type. The BN particles cause a certain loss in efficiency as we observe the dependence of this luminous efficiency vs doping weight of h-BN particles. The role of h-BN particles is originally designed for heat dissipation, but these particles could also have unexpected optical scattering and absorption in the UV range such that the optical efficiency is eventually lowered.[18] In Figure a, the saturation of the green peak and the increasing blue peak can be seen at high current levels, which indicates that the film becomes transparent due to overexcitation. If we normalize the luminous efficiency using low current values, we can obtain Figure b, where the low-BN-doped device can sometimes have better thermal performance than that of the no-BN sample at low to medium current. However, at higher currents, the accumulated heat due to nonradiative recombination of BN and PQD can really drag down the device performance.

Long-Term Aging Tests

One of the important characteristics of these SiO2-coated PQDs is the ability to withstand the erosion of external water, oxygen, and light degradation. To examine this, two samples of the PQD@SiO2 + PDMS mixture were placed in vacuum and in air, respectively, and they were constantly monitored by the photoluminescence (PL) system for their illumination intensities, emission peak wavelengths, and full width at half-maxima (FWHM) of the spectra. Figure a–c shows the results for a duration of 250 days. The PL intensities of the sample left in air are kept at the same level (95.5% of the initial value) and show almost no changes in FWHM and emission peak wavelength in this long period (the FWHM changes for −0.19 nm and the peak red-shifts for 0.60 nm). The quality of the PQD@SiO2 films is maintained very well throughout these days. It is important to acknowledge this stable long-term property of PQD films in the on-shelf storage condition before putting them in a harsher one.

Figure 3

Measured (a) peak intensity, (b) peak wavelength, and (c) FWHM of PQD@SiO2 + PDMS films stored in air and vacuum conditions against storage days.

The next thing to be verified is the long-term stability of these PQDs under continuous operations. This is an ultimate test before a novel type of LED can be commercialized. Different from the on-shelf test, once the LED was turned on, the PQDs would be shined and heated continuously except during the measurement in the integrated sphere for monitoring. The constant photon wear-out and high-temperature environment can be really challenging toward these nanocrystals. The in-chip samples might experience the worst scenario because the direct contact to the largest heat source, i.e., the LED chip, and the poor thermal conductivity of the slurry makes the situation worse.

On the other hand, in the remote film-type device, which is similar to the remote phosphor package,[19] the PQD film was suspended above the chip and contacted with copper foil for better heat conduction to alleviate the direct heating from the chip. In both types, the BN particles were doped into the slurry when mixing and they were expected to enhance the thermal conductivity of the overall composites.

Figure a–c shows the over 2500 h of data for the remote film devices with different BN particle doping amounts, and the in-chip results can be found in the Supporting Information (Figure S2). To properly evaluate the lifetime of the devices, a criterion called LT50 can be linearly extrapolated from the last 2500 h of data and the degradation of the 50% initial peak intensity can be calculated. For detailed calculations, the formula can be found in the Supporting Information. For the in-chip-type devices, the decrease of the PQD illumination capability is dramatic. In less than 1 h, the LT50 criterion, which is the time span for the efficiency of the device reducing to 50% of the initial value, was breached. Both the peak wavelength and FWHM trend similarly to the intensity: they both saturated quickly in several hours from the beginning of the aging process. The blue-shifted peak and widened spectrum are both observed in the samples with and without BN particles.

Figure 4

(a) Normalized peak intensity, (b) FWHM, and (c) peak shift of remote film samples vs aging time.

As for the remote film samples, the change is more gradual and the stabilized emission power and FWHM of the spectra can be observed after several hundreds of hours into the aging process. However, the emission peak tends to blue shift and shows no sign of stopping even after 2500 h of aging. Higher than 29 nm of blue shift can be observed from the measurement. The resulting LT50 for the remote film-type devices can be as high as 6292 h, and all of the detailed parameters of optical spectra after the aging process can be found in Table .

Table 2

Measured Optical Characteristics and Lifetimes (LT50) of Various Samples

type of package	in-chip		remote film
BN weight (mg)	0	2	0	0.5	1	2
LT50 (h)	0.32	0.35	638	6292	4549	2878
λPeak change (nm)	–15.0	–19.9	–32.6	–25.7	–29.4	–21.7
FWHM change (nm)	19.6	19.7	22.9	17.2	26.4	14.3

Discussion

In addition to intensity monitoring, we also performed several material characterizations to compare between fresh and aged PQD@SiO2 powders. The purpose of this examination is to find out the variation of the PQD material during the harsh testing conditions. To prepare these samples, the PQD powders were aged (both optically and thermally) for at least 24 h prior to any characterization. The length of this preaging process was determined by our previous results that most of the degradation seems to happen in the first 24 h of the aging process.

Fourier Transform Infrared (FTIR) and X-ray Diffraction (XRD) Measurements

FTIR can provide important information of the chemical bonding in the material. In Figure a, both the fresh and the burnt samples were tested for FTIR. From the FTIR spectra, several featured dips can be identified from comparison to the past results: the Si–O–Si mode at around 1070 cm–1, the N–H-related bending mode at 1630 cm–1, and the mixed broadband signal at around 3000 cm–1 for C–H-related mode.[13,20−23] There are no apparent differences between the burnt and fresh PQD@SiO2. The comparison shows that the compositions and bonding conditions in the PQD composites did not change before and after the aging process.

Figure 5

(a) Comparison of FTIR spectra of PQD@SiO2 powder before and after aging. (b) Comparison of the X-ray diffraction pattern of the PQD@SiO2 + PDMS film before and after aging.

Figure b shows the XRD comparison between the burned and fresh PQD films. The standard XRD peak locations of CsPbBr3 are also marked in the plot. The broadened peak of SiO2 can be seen at around 20°.[13] The upper side of Figure b marks the peak locations of CsPbBr3, Cs4PbBr6, and PbBr2 obtained from the Inorganic Crystal Structure Database (ICSD). Major peaks for PQDs can be observed in Figure b, such as those in 29.68, 36.54, 42.61°, etc.[24,25] However, due to the presence of SiO2 with a large quantity in the sample, sometimes, it is difficult to acquire the highly resolved XRD pattern from the PQD itself. In both cases, the major peaks are kept the same in terms of peak locations and the relative intensities and linewidths. From this experiment, the change in crystallography seems to be negligible during this aging process.

Location-Dependent Photoluminescence

In Figure , the image of a BN-doped in-chip PQD LED after the aging process is shown. A faint and circular mark can be seen at the center. This mark shows up more clearly when UV excitation illuminates upon the device (also shown in the figure). By taking the location-dependent PL, one can observe the dramatic difference in the peak wavelength and the peak intensity of the specified area in Figure . The central area demonstrated a six-time reduction in the peak intensity, while its peak shifts to a shorter wavelength by 7.65 nm. The FWHM of the signal of the center is also widened by 6.74 nm when compared to the side area (31.47 and 38.21 nm for A and C regions, respectively). This revelation demonstrated a possible scenario that the early depletion or degradation of PQD in the central region due to the harsh testing conditions. The remote film-type devices showed a similar situation but with a much lesser degree. The broadening of FWHM and the blue shift in the wavelength are certainly affected by the heat in the package. A comparison between BN-0 and other samples in terms of a lifetime can be a very good example. Without the BN mixture, the heat dissipation could be worse and could cause the rapid degradation of PL of PQD and also the fast broadening of the linewidth in the optical spectrum.

Figure 6

(a) Location-dependent PL signal of an aged in-chip device and (b) pictures of this device under natural light and UV.

TRPL

Time-resolved photoluminescence (TRPL) can be an indicator of the PQD’s internal illumination capability. A comparison experiment was performed between before and after burned-in PQD@SiO2 samples. The time constants of the traces can be expressed as[26−28]

The multiple decaying functions can provide us some insights into the radiative transition of the quantum dots before and after the aging processes. As shown in Figure a, the TRPL trace of fresh PQD@SiO2 has a longer τ1 (3.6 ns) than the burnt one (1.59 ns). A reduction in the decaying time constant can be explained as the increasing portion of the nonradiative recombination inside quantum dots.[28,29]

Figure 7

(a) TRPL profile of a fresh and an aged PQD sample. (b) Normalized and peak-shifted spectra overlay showing the increased tail emission in the short-wavelength range after the aging process. (c) Calculated quantum dot size of the PQD@SiO2 remote film-type devices as they are aged over time. (d) Histogram of the dot side-length ratio of fresh and aged PQD@SiO2 powders. The inset is an illustration of the quantum dot and the definition of x-length and y-length.

Wavelength Shift

While the remote-type package can help with the lifetime of the PQD film, the blue shift of the emission wavelength is significant and cannot be overlooked. Previous research showed that the continuous blue shift under direct laser illumination for several tens of seconds.[30] In the past, many possible reasons have been discussed about the blue shift of nanocrystals. In most cases, oxidation can be a very plausible reason because of the abundance of oxygen during the quantum dot device operation.[31] Other labs found that humidity can enhance the formation of smaller perovskite grains in the larger ones.[32] However, after the aforementioned characterization, we found that it is not very easy to get a definitive answer to our situation. It has been recognized that the emission spectrum can usually reflect the size distribution of the nanocrystals.[33] With optical spectra, one can deduce the possible size distribution of these nanocrystals, and we can extract useful information about this QD ensemble. From the previous section, our PQD@SiO2 can maintain their size and illumination efficiency under the on-shelf test, and this result proves that our nanocrystals are naturally stable when there is no external stimulant that alters the surrounding environment of PQD greatly. However, the continuous and prolonged aging process did change the PQD’s emission wavelength progressively. Even when the intensity is stable, the peak still shifts. On the other hand, both the crystallography (XRD) and chemical binding (FTIR) data seem to be the same between fresh and aged samples. There might be several explanations we can consider: (A) the formation of PbBr2 in the outer shell, (B) the internal fissure of the PQD, and (C) the shape change of the PQD. In the following, we will explain our reasoning and findings.

While PbBr2 is one of the chemicals to form the PQD, it is also possible that the PQD can be decomposed and return back to PbBr2.[34] The heat and external photons provide a hotbed leading to this path. Once the outer shell is transformed to PbBr2, it loses its illumination capability. Thus, the effective size of the quantum dot reduces and causes the blue shift of the wavelength. However, the XRD result before and after aging did not show many differences, especially for the PbBr2 peak locations. Another possibility is to have a transformation between Cs4PbBr6 and CsPbBr3 due to various external influences.[35,36] Either from the result of PbBr2 addition or H2O desorption, the transformation of the CsBr–PbBr2 composite system can provide either highly luminous CsPbBr3 or photonic-inactive Cs4PbBr6.[35,36]

The second possibility is the internal fissure of the PQD, meaning that it might possible that the PQD broke up during the aging test. When the dot breaks, the size becomes smaller and the wavelength can move toward the higher-energy side. Because this breakage is random, the linewidth of the spectrum, which is the distribution of the size of the PQD ensemble, can be widened. A similar phenomenon was reported before in PQD,[29] and it was attributed to the Ostwald ripening of the PQD during the storage.[29,37] A significant tail at the shorter-wavelength side, as shown in Figure b, can be an indicator of this situation. (The detailed spectra can be found in Figures S3 and S4 of the Supporting Information.)

To estimate the quantum dot size, the peak emission wavelength is used for the effective transition energy due to quantum confinement. Our calculation follows the effective mass theory,[4,38] and the expression iswhere Eg is the band gap of CsPbBr3 (equal to 2.32 eV), me* and mh* are the effective mass of CsPbBr3, r is the average size of the dot, and Eexciton is the binding energy of the exciton in this material (equal to 40 meV).[4]Figure c shows the calculated dot size of the corresponding emission peaks during the aging process. Both samples started with similar size distribution of dots and ended up with very different size distribution at the end of 2700 plus burn-in hours. The final estimated sizes are 4.83 nm (no BN) vs 5.48 nm (with BN 2 mg).

The third hypothesis is the shape change (sintering or morphological change) of the dots. In the previous work, the researcher reported the elongated shape of the dots after aging, and this led to the blue shift of the wavelength.[39] The sintering process could happen due to the extreme heat and UV excitation conditions that were imposed on these nanocrystals.[39] In our experiment, by comparison of transmission electron microscopy (TEM) pictures of limited samples, we could have a slight increase of elongated dots after the aging process. As shown in Figure d, if we plot the length ratio of dots in the x-direction over that in the y-direction, the fresh sample has a more concentrated distribution around 1, while the samples after aging have a wider range of the size ratio. The TEM pictures are provided in Figure S6 of the Supporting Information.

From these observations, we could say that the continuous change of wavelength is an important subject to be solved in this type of PQDs because it did change the color provided by this LED. To summarize what we found in various experiments and theoretical analyses, two possible scenarios can be proposed: the reduction in the size of the quantum dot core and the change of the shape of quantum dots. Both of them can be supported from the past literature works, TEM pictures (in Figure S6), and the broadening of the optical spectra of the aged quantum dots. Meanwhile, other factors can be also the cause of this phenomenon. More investigation will be necessary before the conclusion can be drawn.

Thermal Properties of Samples

The thermal characteristics of these samples are one of the important parameters that we need to find out. To do that, the surface temperature of the sample is an easy way to start. Figure a–c shows the thermal images taken at the later stage of this stressed aging process (>1000 h of continuous tests). The in-chip sample without BN particles showed the highest surface temperature, while the other two were about 15° lower.

Figure 8

Thermal images of (a) a remote film-type device with BN = 0 mg at 1083th hour, and the max. temperature is 45.2 °C. (b) Remote film-type device with BN = 2 mg at the 1126th hour, and the max. temperature is 42.3 °C. (c) In-chip-type device with BN = 0 mg at the 1066th hour, and the max. temperature is 59.4 °C.

Another important thing is to observe the temperature variation versus time. There are two kinds of data: short-term and long-term records. First, we can take a look on the short-term response or transient response. In Figure a, we measure the temperature variation versus time right after the current is turned on. The increase of package temperature can be stabilized within the first 15 min, and after that, the variation would be small (within 1% of the average temperature). This number gave us a guideline when measuring the surface temperature of these samples. Then, the long-term effect on the surface temperature is presented in Figure b, which demonstrates a slow but increasing trend of the temperature across aging time, from the early stage to the late stage of aging. The possible reason is due to the degradation of the quantum dots, and this situation leads to increasing nonradiative recombination at the later stage of the aging test. Also, in general, the samples with BN particles showed less heating on the surface. If the average temperatures are put together, the in-chip samples have much higher surface temperatures than their remote film counterparts in every stage of the aging process, as shown in Figure c. From these results, we can see that the gap between in-chip and remote grows when the quantum dots deteriorate (this gap can be as large as 16°), and the differences in temperature between with and without BN samples are about 3°.

Figure 9

(a) Transient temperature variation in a 5 h experiment after current injection started. (b) Long-term surface temperature of our remote film devices under tests. (c) Average temperatures of remote film samples and in-chip ones at different stages of the aging tests.

To simulate these phenomena, we built a COMSOL model with similar physical sizes to the actual devices and using a LED chip at the center as the heat source. A 1 W of heat energy was set in the chip location. As shown in Figure a, the direct contact of the PQD composite with the LED surface could induce a much higher temperature inside the QD layer and eventually lead to fast degradation of the quantum dots. Meanwhile, the temperature distribution of the remote film device was more uniform than the in-chip device and as high as 14° of temperature difference could be seen at the bottom of the QD layer in Figure a,b. Although the surface temperature was not that much different (3–5° according to different physical settings), the interior temperatures of the in-chip and remote film samples could be a decisive factor that causes the different aging results we obtained in this study.

Figure 10

COMSOL calculation of temperature distribution of (a) in-chip sample and (b) remote film sample.

Conclusions

In conclusion, we demonstrated highly stable hybrid perovskite quantum-dot LEDs with boron nitride nanoparticles incorporated in the package. While the PQD@SiO2 powders can be stored in air for more than 250 days without significant degradation, the long and continuous operation in the LED package is still proved to be difficult. In this study, after 2500 h of aging, the intensity of PQD can be stabilized, while its wavelength of emission blue-shifts continuously. From the material characterization, we see very few pieces of evidence about chemical changes, but from the optical measurement, a significant change in the PQD size distribution can be observed. Three hypotheses can be proposed but need more scrutiny. Once solved, there is no doubt that inorganic perovskite quantum dots are suitable for the future adaptation of the solid-state lighting devices that involve nanoscale fluorophores.

Experimental Section

Perovskite Quantum Dot Sample Preparation

The colloidal CsPbBr3 PQDs in this study were prepared in two steps: one is the Cs-oleate precursor material preparation, and the other is the synthesis of CsPbBr3.[40][40] For the Cs-oleate step, we put 0.814 g of Cs2CO3, 2.5 mL of oleic acid (OA, 90%), and 40 mL of 1-octadencene (ODE, 90%) into a 100 mL three-necked bottle and mix them fully. The mixture is heated to 120 °C in vacuum for 1 h, followed by heating at 150 °C in N2 until all Cs2CO3 completely reacts with OA. For the step of synthesizing CsPbBr3, we put 0.752 mmol lead(II) bromide (PbBr2, 99%) and 20 mL of ODE into a 100 mL three-necked bottle and mix them fully and then the mixture is heated at 120 °C in vacuum conditions for 1 h. Then, we heat them at 120 °C with N2, inject 2 mL of OA, and 2 mL of oleylamine (OAm, 80–90%) until PbBr2 reacts with OA and OAm totally. Finally, we increase the temperature to 180 °C and inject 1.6 mL of Cs-oleate solution wait for 5 s then cooling it. After the synthesis of nanocrystals (NCs) is finished, (3-aminopropyl)triethoxysilane (APTES) would then be added to the PQD solution and stirred for 2 h and then tetramethoxysilane (TMOS) would be added to further stirred for another 38 h. The −NH2 functional group of APTES is used to silylize the surface of PQD, and the hydrolysis reaction of TMOS can lead to the formation of SiO2 around the quantum dots. After the reaction is completed, the solution is centrifuged at 9000 rpm and then evacuated to extract the PQD@SiO2 powder. Figure shows the dried PQD@SiO2 powder under a UV lamp, and Figure b shows the transmission electron microscopy (TEM) picture of a CsPbBr3 NC.

Figure 11

(a) PQD@SiO2 powder under UV and (b) PQD@SiO2 by TEM.

Next, we used the dried powder of green 530 nm PQD@SiO2 and the powder was mixed with PDMS at the weight ratio of 1:9 to form the filling substances in the experiments. The pumping source in this study was the UV LED chip made by Epistar Corporation. The UV chips were first driven at 250 mA for at least 24 h to stabilize their output power. Since the normal bias current level was lower than 250 mA, this burn-in step could ensure a constant photoexcitation intensity in the following aging process. There were two different structures considered in our study. In the first structure, the PQD@SiO2 powder was mixed uniformly with PDMS slurry and placed around the UV LED chip, and this type of device was named “in-chip” type. For the second structure, called the “remote film” type, the PQD@SiO2 + PDMS film was first mixed, dried, and then placed on the copper tape with a 2.5 mm diameter hole at the center and attached on top of the 5070 lead frame. One of the important steps in our manufacturing of these remote films was the molding step. The PQD + PDMS mixture was filled into an empty 5070 lead frame and baked. Then, the film would be peeled out after it became solid. With a precise control on the weight of the PQD + PDMS mixture and the molding step, an approximate shape (height and width) can be maintained from device to device. There were no other filling materials in the remote film style package, and the PQD film was suspended above the UV chip. The purpose of this experimental design was to verify that our remote film package can further improve the reliability of the PQD device even under continuously stressed aging situation. Using the in-chip style package could benchmark our PQDs with the traditional method, and the on-shelf lifetest in the next section could verify our good material quality.[17,41−43] For both types of samples, we placed the composites on the holder and then cured them at 70 °C to solidify the slurry. To improve thermal dissipation, PQD@SiO2 + PDMS composites were mixed with certain weights of hexagonal boron nitride (h-BN) nanoparticles and reference samples without BN particles were also made at the same time. The h-BN powder was purchased from Sigma-Aldrich, and its average particle size was 1 μm (product number 255475-50 G). For the in-chip structure, we had two samples: one was BN-0 mg (no BN), and the other was BN-2 mg (with 2 mg of BN powder added). For the remote film structure, four samples, BN-0 mg, BN-0.5 mg, BN-1 mg, and BN-2 mg, were produced according to the weights of BN particles inside the PQD@SiO2 + PDMS composites. Figure a,b shows the process flow and the packaged structures of the two types of devices.

Figure 12

Process flows for (a) in-chip devices and (b) remote film devices. (c) On-chip device with current off (up) and on (down).

Measurement Methods

The devices were measured in an integrated sphere coated with the BaSO4 material. Keithley 2400 source meters are used to supply suitable currents to the LEDs. The photoluminescence of the thin films, which was analyzed by the on-shelf storage test, was measured by a Princeton Instruments Acton 2150 system with calibrated samples for the long-term stability test. Fourier transform infrared (FTIR) spectroscopy was performed by a Thermo-Fisher Scientific Nicolet iS10 FTIR spectrometer to determine any possible variation in the chemical bonding before and after PQD-aging tests.

After the fabrication of six devices was finished, these devices were driven by a continuous current at room temperature and the electrical currents were provided by a Keithley 2400 source meter. In this process, their electrical and optical properties were extracted from the integrated sphere system. The lifetime tests were conducted in the constant current mode with the consideration of the constant photonic pumping intensity of 1 W/cm2. Figure c shows an “in-chip” device with and without current injection.