Article pubs.acs.org/cm
Field-Driven Ion Migration and Color Instability in Red-Emitting Mixed Halide Perovskite Nanocrystal Light-Emitting Diodes Parth Vashishtha and Jonathan E. Halpert*,† MacDiarmid Institute for Advanced Materials and Nanotechnology, and School of Chemical and Physical Sciences, Victoria University of Wellington, P.O. Box 600, Wellington, New Zealand S Supporting Information *
ABSTRACT: Perovskite nanocrystals have shown great promise as the basis of a new family of nanocrystal lightemitting diodes (LEDs). However, the external quantum efficiency and color stability of these materials still lag behind those of well-established technologies. Producing stable efficient red emitters with electroluminescence (EL) in the “pure” red range of 620−650 nm is a particular challenge. Here we present mixed halide CsPbBr3−xXx (X = I or Cl) peNC organic LEDs using peNC emitters with photoluminescence across the visible region to produce LEDs displaying EL across the visible spectrum. By focusing on the yellow-orange to deep red (560−680 nm) visible regime, we present evidence that field-driven halide separation in CsPbBr3−xIx peNCs is responsible for the observed red-shifting and splitting of the EL peaks. Greater compositional stability is demonstrated to be the key to higher efficiency, long-lived devices for deep red-emitting mixed halide peNCs with higher compositional concentrations of iodide. emitters. Song et al. were among the first to demonstrate blue, green, and orange peNC-LEDs with EQEs of 0.07, 0.12, and 0.09%, respectively.10 They did not report a fully red device as they claimed that it is not possible to fabricate LEDs using a higher content of iodide because of its metastable state in the cubic phase.10,14,15 Zang et al. implemented a perfluorinated ionomer with their hole transport layer and claimed better results with a 0.06% EQE green-emitting device.12 Pan et al. improved the efficiency of green and blue LEDs further to 3.0% by modifying the NC surface with a ligand to improve QY and conduction but did not report a red LED.16 Red-emitting LEDs were produced by Li et al. using CsPbX3 nanocrystals crosslinked with an atomic layer-deposited (ALD) film of alumina.11 This method produced devices with a deep red 680 nm emitting device reported at 5% EQE and a 620 nm-emitting device reported at 1.4% EQE, as well as less efficient blue and green LEDs.11 However, they also reported significant color instability in their red LEDs at 620 nm and assumed that it was caused by ion segregation, an issue that could derail efforts to produce commercially viable LEDs using these materials.11,17 It should be noted that ion separation and segregated halide domains have been reported previously by many groups working on MAPb(I/Br)3 thin films and solar cells.18−25 Photoluminescence (PL) emission from halide-segregated
S
emiconductor quantum dots (QDs) have proven to be useful materials for optoelectronic applications because of their thin line width of emission, high photoluminescence quantum yield, and size- and composition-tunable emission wavelengths. Over the past decade, significant effort has gone into refining the synthesis and processing of quantum dots to improve the quantum yield (QY) for red, green, and blue emitters to be used as the active layer in quantum dot organic light-emitting devices (QD-LEDs).1−8 To date, the best QDLEDs have used cadmium-based core quantum dots, with various Type I shell coatings, including CdS or ZnS.1−3 These quantum dots can display photoluminescent quantum yields (QYs) near 100% and have very stable emission peaks after processing, allowing the color to be tuned via size and composition control during chemical synthesis.4,5 This has translated into QD-LEDs with maximal reported external quantum efficiencies (EQEs) of nearly 20%.2 Although other material options exist for QD optoelectronics, including III−V and I−III−VI semiconductor nanocrystals such as InP and CuInS2, these materials tend to have lower QYs.6−8 Recently, CsPbX3 (X = Cl, Br, or I) QDs with QYs as high as 90% have been reported, suggesting that these could be a viable candidate material for the active layer in organic LED (OLED) structured devices.9 Since the reported synthesis of CsPbX3 quantum dots, several groups have been successful in producing functional perovskite nanocrystal LEDs (peNC-LEDs).10−13 Color tunability in these materials is usually achieved by altering the I/Br or Br/Cl ratio in the crystal structure to achieve RGB © 2017 American Chemical Society
Received: April 19, 2017 Revised: June 20, 2017 Published: June 21, 2017 5965
DOI: 10.1021/acs.chemmater.7b01609 Chem. Mater. 2017, 29, 5965−5973
Article
Chemistry of Materials
Figure 1. (a) Schematic and (b) energy diagram for the ITO/PEDOT:PSS/pTPD/peNC/TPBi/Al LEDs investigated here along with (c) a photo of peNCs in hexane excited by a UV lamp and (d) a transmission electron micrograph of CsPbBr3 NCs with a diagram of the perovskite NCs (inset). Energy levels relative to vacuum are shown for reference.12,27,28
for QD-OLEDs, wherein ITO/glass is coated with PEDOT:PSS and then poly-TPD as a hole transport layer, a spin-cast QD active layer, and evaporated TPBi as an electron transport layer with an Al metal electrode.29 Photoluminescence peaks and corresponding electroluminescence peaks taken at early times near the turn-on voltage (∼3.5−6.0 V) are displayed in Figure 2, with performance comparisons provided in Table 1 (see also Table S2). As one can see, the EL peaks are nearly identical to the initial PL peaks with a slight red-shift of a few nanometers that is ascribed to the transition from solution to thin films, wherein Fö rster resonance energy transfer (FRET) occurs to neighboring NCs with a lower band gap. For populations of NCs with small distributions of size and composition, as estimated by full width at half-maximum (fwhm) of ∼30 nm, this shift is generally slight. Devices investigated here are described in Table 1. Blue (B499) and green (G518) LED measurements are provided for reference, while orange, orange-red, and deep red devices are investigated for trends in color instability. For the greenemitting LEDs (G518), there is no evidence of a spectral shift at higher applied voltages (∼8−9 V).30 These devices showed no shift in EL peak over time, indicating that despite an increased rate of heating, from power dissipation during operation, and the presence of an electric field, the emission from the NCs is stable. No EL was observed from the constituent organic semiconductors used as the hole and electron transport layers, poly-TPD and TPBi, respectively.5,31,32 Red-emitting perovskite NCs were produced using a mixture of bromide and iodide precursors to create CsPbBr3−xIx, with x varying from 0 to 3. These mixed
domains, dubbed the Hoke effect, has been observed via PL spectra from films of MAPb(I/Br)3 under intense excitation (i.e., high charge density).25,26 Red is a critical color for any display, and “pure” red is generally produced from emission that peaks between 620 and 650 nm. Lower wavelengths appear orange-red, and higher wavelengths appear less bright, because of the sensitivity of the human retina to red photons that falls off rapidly at higher wavelengths. Li et al. and others attributed the color instability to the unstable crystal structure of CsPbBr3−xIx with large quantities of iodine.11 Although rapid advances are being made in the overall efficiency of peNC-LEDs using a wide range of device architectures,13 the color instability of red-emitting peNCs remains a challenge. While some aspects of color instability and poor device performance in red-emitting peNCLEDs have been sporadically described, here we are able to systematically investigate this phenomenon to better understand how composition affects stability. By comparison to similar effects in MAPI thin films, and known properties of NCs in LEDs, we can then infer plausible drivers for changes in the electroluminescence (EL) spectrum. In addition, we use trends in longevity and color stability that suggest next steps for improving performance in devices using these materials.
■
DEVICE FABRICATION AND PERFORMANCE METRICS QD-OLEDs were constructed from CsPbX3 nanocrystals (X = Cl, Br, or I) grown via a colloidal synthesis procedure adapted from the literature.9 As seen in Figure 1, devices were constructed using a commonly reported device architecture 5966
DOI: 10.1021/acs.chemmater.7b01609 Chem. Mater. 2017, 29, 5965−5973
Article
Chemistry of Materials
Figure 2. Normalized photoluminescence spectra (top) taken at 400 nm excitation of a solution of CsPbBr3−xIx for varying values of x from 0 (green) to 2.75 (deep red). Blue NCs consist of CsPbBr3−xClx (x = 0.75). Normalized electroluminescence spectra (bottom) from ITO/ PEDOT:PSS/pTPD/peNC/TPBi/Al LEDs for these same peNCs. EL spectra were recorded immediately after turn-on (prior to red-shift). As an exception, the orange-emitting sample from the PL spectra did not produce an orange-emitting LED (gray dashes to mark the absence) as discussed in the text. At the right are shown unprocessed photos of LEDs.
Table 1. Photo- and Electroluminescence Metrics for Champion Devices with peNCs with Varying Br Content LED label
formula equivalent of Br (3 − x) in NCs
PLsoln peak (nm)
fwhm PL (nm)
ELinitial peak (nm)
fwhm ELinitial (nm)
initial CIE-x
initial CIE-y
Vt (V)
B499
2.25a
492
27
499
28
0.043
0.491
6.0
G518
3.00
516
22
518
19
0.092
0.797
4.0
O558 R607 R621 R636 R642 R653 R665 R671
1.88 1.375 1.25 0.875 0.75 0.625 0.50 0.375
558 603 621 632 640 651 667 673
31 29 62 33 36 39 45 50
653b 607 621 636 642 653 665 671
45 36 61 66 35 55 45 32
0.711 0.632 0.654 0.664 0.708 0.698 0.720 0.728
0.289 0.368 0.346 0.336 0.292 0.302 0.280 0.272
5.0 5.0 5.8 6.2 6.0 6.0 4.0 3.9
not observed not observed 1.4 × 10−1 2.1 × 10−1 6.0 × 10−2 7.1 × 10−2 2.4 × 10−2 1.3 × 10−1 2.9 × 10−2 1% EQE using this device architecture and display no shift in EL peak even after prolonged use at a high current density (>10 mA cm−2).5 Maximal EQEs for red devices listed in Table 1 using the ITO/ PEDOT:PSS/poly-TPD/peNC/TPBi/Al devices ranging from 5967
DOI: 10.1021/acs.chemmater.7b01609 Chem. Mater. 2017, 29, 5965−5973
Article
Chemistry of Materials
Figure 3. Peak-normalized electroluminescence spectra for example device (a) R665 and (b) R607 showing spectral shapes and red-shifting of the peak during J−V measurement from near their respective sample PL peak to ∼680 nm. The effect is noticeably stronger for the high-bromide content peNC-LEDs. (c) J−V curves for several devices from Table 1, showing higher current density for the less stable peNC samples. (d) Plot of EL peak position vs voltage for each device (R607−R665), in which points are fitted with a spline curve to show the relation.
Figure 4. EL peak movement at different constant voltages with respect to time in (a) R665 and (b) R607 devices. Higher-iodide content devices showed better color stability over time for (roughly) x values of 2.25−2.50.
voltage for each device (at ∼8−10 V) prior to burnout. As is apparent from the table, the electroluminescence peaks of the red-emitting peNC-LEDs (devices R607−R665) shift significantly to redder wavelengths as the voltage increases. The peak shifts did not appear to be immediately reversible under these operating conditions. However, after several hours (∼16 h), the photoluminescence peaks did return to their initial positions, as
0.02 to 0.2% EQE are similar to those previously reported for untreated films of red-emitting CsPbX3 NCs in peNCLEDs.10,28
■
PEAK SHIFTING IN RED ELECTROLUMINESCENCE
Table 1 shows the starting and final electroluminescence peaks for each device during operation from 0 V to the maximal 5968
DOI: 10.1021/acs.chemmater.7b01609 Chem. Mater. 2017, 29, 5965−5973
Article
Chemistry of Materials
Figure 5. Photoluminescence spectrum (left chart) of a solution prior to thin film formation and of the finished ITO/PEDOT:PSS/pTPD/peNC/ TPBi/Al device before (orange line) and after (green line) device operation. The electroluminescence from the device after operation at 7 V for 2 min is included for reference (dashed gray line). After the initial PL measurements, the voltage for the LED was raised to the turn-on voltage of ∼5 V until EL was observed and then switched to 7 V to ensure observation of ionic separation. Electroluminescence spectra (top center chart) and photos (top right) of the same LED were taken at the turn-on voltage and at 7 V and then at 2 min intervals until the spectrum had shifted completely. The bottom panel shows a diagram of ion migration creating green- and red-emitting NCs, as well as charging of red NCs at later times.
one can see in Figure 3, increasing color instability also seems to be correlated with the higher current in the device (Figure 3c), which suggests that the high-Br content NCs give a higher overall charge mobility. Deeper red devices also appear to be more stable in that the onset of the peak shift occurs at a higher applied bias. This indicates that a high iodide content in fact produces more stable peNC-LEDs and that high-iodide NCs appear to be more resistant to field-induced separation. Finally, moderately red devices (R642−R665) were generally found to produce longer-lived devices with better color and EQE stability. Similar trends were observed at a constant applied voltage (Figure 4) where the rate of the peak shift was found to be highly dependent on the peNC composition and voltage (see also section S10 of the Supporting Information). These data allow us to discount several possible factors in the emission peak shift. It is clear from Figure 3 that the band shift is too extreme to be due to a Stark shift or a similar electrostatic change in the band energies of individual NCs.30 Emission from the organics or interfacial excitons would be clearly distinguishable from NC emission by the width of the peaks.5,32,33 Another possible mechanism for the peak shift is the growth of peNCs in the film when it is heated during device operation. This could be caused by three possible events: (1) incorporation of unreacted precursors present in improperly purified NC solutions and deposited in the thin film, (2) Ostwald ripening, in which large NCs could grow at the expense of their neighbors because of the chemical activity of ligands, or (3) fusion of neighboring NCs, an effect that has
indicated by PL measurements (Figure S9). Deep red LEDs (R671 and R688) were not observed to shift but were observed to be very inefficient (EQE < 10−2%). The extent of the shift is determined by the initial PL peak and thus the composition of the peNC sample used in the NC layer of the device. From plots of the EL spectra during each scan (Figure 3 and Figure S3), the peaks appear to move uniformly to the red, prior to device burnout, to values between 670 and 680 nm. This redshifting starts at the turn-on voltage and ends around 8−9 V for deeper red-emitting devices like R665 and at 6−9 V for orangered and medium red devices such as R636 and R642. Orangered, bromide rich LEDs (R607−R642) consistently performed poorly, in terms of color stability and lifetime, in this study compared to moderately red-emitting LEDs (R653 and R665), which is consistent with previous reports.17 This limits the value of composition tuning in these materials and represents a significant challenge to the production of red peNC-LEDs for commercialization.
■
STABILITY AND COMPOSITION Devices R665 and R607 were tested for field stability by slowly scanning from the turn-on voltage (∼4−5 V) to 9 V at a rate of ∼5 mV s−1. As illustrated in Figure 3, devices with NC films with a high Br/I ratio shifted earlier and at field strengths lower than those with films with a high I/Br ratio. Because the thickness of the organic and NC layers was similar, presumably the field strength at the interface is similar for each LED. As 5969
DOI: 10.1021/acs.chemmater.7b01609 Chem. Mater. 2017, 29, 5965−5973
Article
Chemistry of Materials been observed previously in perovskite material.34−39 For peNCs with >600 nm PL, the peak shifts to the red; however, the size (10−15 nm) of the peNCs used here is too large to observe this much contribution from the confinement effect.40,41 As such, the change in band gap due to the increased size is be very slight and cannot explain the large shifts in EL, all of which end at ∼670−680 nm.42 In addition, fusion or growth of the particles should be an irreversible process. Similarly, the fwhm and wavelength of emission discount exciplex emission or emission from charge transfer states in the NCs.43 Without size changes or electrostatic effects, the best explanation is that there is movement of the ions within the NCs and that the composition of individual emitters within the NC thin film changes during operation.25,44
surrounding organic layers, and emitting at 518 nm (Figure 5).49 The remaining red emission appears to originate from CsPbBr3−xIx NCs that have lost some Br and gained some I, giving them an I/Br ratio higher than that present in the homogeneous thin film prior to turn-on. This mechanism is present in all the devices studied, R607−R688. However, the green emission peak is apparent only in device O558 because of the very large quantity of Br present in the initial sample. Ion separation here creates a quantity of iodide free NCs large enough that many of these are unable to FRET transfer energy to their now absent red-emitting neighbors. Instead, excited states formed on green-emitting NCs can radiatively relax with the characteristic green emission of a large diameter (≥10 nm) CsPbBr3 NC. Furthermore, given the very small dimensions of the NCs, fast internal energy transfer via FRET makes it improbable that green light could be emitted from a single NC with an iodide rich domain. This suggests that halide ions must be exchanged between NCs during purification, a process that has been observed in reverse in solution for inorganic peNCs.46 In devices with an iodide content greater than that of O558 (e.g., R607−R688), there may be green-emitting NCs, but never enough for green emission to be observed. Instead, their excited states are transferred via FRET to neighboring deep red emitters, or peNCs with deep red domains, in the films that emit uniformly near 670−680 nm once the maximal degree of ion separation has occurred. The low efficiency of these types of devices can be ascribed to such FRET losses, which are severe in films with a wide band gap distribution and a QY of
