We study the origin of efficiency roll-off (also called ''efficiency droop'') in colloidal quantum-dot light-emitting diodes through the comparison of quantum-dot (QD) electroluminescence and photoluminescence. We find that an electric-field-induced decrease in QD luminescence efficiency-and not charge leakage or QD charging (Auger recombination)-is responsible for the roll-off behavior, and use the quantum confined Stark effect to accurately predict the external quantum efficiency roll-off of QD light-emitting diodes. DOI: 10.1103/PhysRevLett.110.217403 PACS numbers: 85.60.Jb, 85.35.Be, 85.60.Bt Quantum-dot light-emitting diodes (QD-LEDs), which capitalize on the excellent color saturation and high photoluminescence efficiency of colloidal QDs, offer the prospect of a new generation of display technologies [1]. However, these devices suffer from decreasing efficiency [measured as the ratio of the number of photons emitted out of the QD-LED to the number of electrons injected into the device, per unit time, known as external quantum efficiency (EQE)] at high current densities. This behavior, termed efficiency roll-off or efficiency droop, is a problem that affects most types of LEDs [2][3][4]. The origin of the efficiency roll-off continues to be a topic of debate and understanding its cause is essential to developing highbrightness, high current density QD-LEDs. In this Letter, we investigate the origins of the roll-off behavior in QD-LEDs by performing simultaneous measurements of electroluminescence (EL) and photoluminescence (PL) intensities of a QD-LED, which pinpoint the cause to be a decrease in QD luminescence efficiency. Comparison of EL and PL spectra reveals that strong electric fields are responsible for the reduced QD luminescence, and the quantum confined Stark effect (QCSE) and transient PL measurements consistently explain the observed phenomena.The device structure investigated was a QD-LED with organic-inorganic hybrid charge transport layers that has recently attracted attention owing to its record high EQE and brightness [2]. The device was fabricated on a glass substrate coated with indium tin oxide and has the structure: ITO ð150nmÞ=ZnO ð50 nmÞ=QDs ð30 nmÞ=4, 4-bis(carbazole-9-yl) biphenyl (CBP) ð100 nmÞ=MoO 3 ð10 nmÞ=Al (100 nm). ZnO was radio-frequency sputtered, QDs were spin-cast out of chloroform, and CBP, MoO 3 , and Al were thermally evaporated. We used CdSe-ZnCdS core-shell QDs with a peak PL wavelength of 610 nm, provided by QD Vision, Inc. Current density and normalized EQE for a typical device are shown in Fig. 1(a). The EQE peaks at 2% for 4 V applied bias and rolls-off by 50% by 8 V. The energy band diagram of the device is shown in the inset and is based on literature values [4][5][6][7].The decrease in EQE at high biases may be a result of either charge carriers leaking out of the QD layer or a reduction in QD luminescence efficiency. To identify which of these two mechanisms dominates, we perform a simultaneous EL-PL experiment and monitor the relative PL efficiency of t...