While UV-induced accelerated aging methods have been previously proposed [8] , the main objective of this study is to demonstrate its potential to allow accurate lifetime prediction for commercial OLED displays. To do so, we used PMOLED displays (UG-9664HDDAG01) made by Univision Technology Inc. Shown in Figure 1, it has 96 × 64 RGB pixels and an active area of 20.14 mm by 13.42 mm. It is mounted on a Waveshare Electronics PCB support to facilitate its integration and manipulation. Finally, the displays are powered and controlled using an Arduino Uno programmed with open-license libraries. The code was modified to generate a checkerboard pattern shown in Figure 1, as suggested by the ISO 9241 305 2008 standardized testing method. As we will show later, this can allow simultaneous assessment of the influence of the violet light exposure on both the active and inactive regions of the displays. As OLED materials are fairly sensitive to light, the commercial PMOLED displays possess built-in UV filters shown in Figure 2(a). Using an Ocean Optics USB2000+ spectrometer connected to an integrating sphere, we measured the transmittance spectrum shown in Figure 2(b). The measured transmittance at 405 nm is 9.9%.

As shown in Figure 3, high-power LEDs from LED Engin Inc. (LZ1-10UA00-00U8) having 1 W of radiant flux at 405 nm radiation are used to irradiate the displays. The irradiance of the 405 nm LEDs and the PMOLED are measured with a Thorlabs S120VC photodiode with its USB interface which has its own built-in calibration module. The PMOLED emission is measured at blue sub pixel peak value. The 405 nm excitation power can be

manually fine-tuned via an operational-amplifier with Darlington driver. As such, the 405 nm irradiance incident upon the display can be varied between 300 to 1100 W/m² to be consistent with the previously-proposed method [8] .

An external fan is used to maintain the test bench at room-temperature to prevent the device from overheating and sustaining thermally-induced damage coupled with the violet light-induced aging. It also prevents the Darlington transistor from heating. Moreover, the whole setup was enclosed in a black box to avoid additional exposure to ambient lights.

During the accelerated aging, the emission from the display is measured every 24 hours following the test sequence previously proposed in [7] . The displays are considered to fail when they reach less than 50% of their initial power densities.

Féry et al. previously proposed a correlation between half-lifetime (t_0.5) and the initial luminance (L₀) [9] . This relation is given as

where n and C₁ are characteristic parameters. Meanwhile, Seifert et al. also proposed a direct proportionality relationship between the initial luminance (L₀) and the current density (j) transforming this model into

where n and C₂ are different characteristic parameters [8] . Finally, they propose to directly apply the same relationship between the current density across the device (j) and the incident ultra-violet light power density (I_UV) to yield

where n and C₃ are characteristic parameters [8] .

According to Zhang et al., a WOLED satisfies a lognormal distribution and the maximum likelihood principles can be applied for calculation [5] . Here, we are assuming that a PMOLED would also have a similar ageing behavior. As such, the estimated logarithmic mean value () and the estimated logarithmic standard deviation () can be expressed as

and

where the t_j are the failure times at stress levels = () since we have 6 stress levels in this specific experiment. The three failure times at different stress levels (j = 1, 2, 3), are shown in Table 1. The total number of samples at stress level is, and.

Using Equations (4) and (5) with the information in Table 1, we can then calculate the estimated logarithmic mean value and the estimated logarithmic standard deviation extracted for each sample as shown in Table 2.

Meanwhile, the accelerated life equation would be an inverse power law

where α and β are the acceleration parameters and I is the stress intensity [5] . Here, the intensity of the 405 nm light I_V = I. This model is perfectly consistent with the experimental data shown in Figure 7, from which we can extract the acceleration parameters α = 9.8944 and β = −0.7997.

As such, the weighted average of the logarithmic standard deviation, the average life and median life can be given as

and

and

where is the weighted average of values in Table 2, which gives the estimated standard deviation to be. To solve Equation (8) and Equation (9), the logarithmic mean valueµ₀ is approximated by as given by Equation (6).

Obviously, the real device lifetime statistics will depend heavily on the environment in which the devices are used. As such, accurate predictions require an appropriate value for the irradiance I depending on the conditions in which the displays will be used. As an example, let us assume a display device that would be used in an environment subjected to direct sunlight exposure. There, I at 405 nm would be equal to 0.8785 W/m²/nm at direct

normal circumsolar irradiance (according to standard ASTM G173-03) into Equation (6) and thus obtain. Submitting the results of Equation (6) and Equation (7) into Equation (8), we found the average life. Since, the median life from Equation (9) is hours. Naturally, the conditions given in standard ASTM G173-03 can’t be applied to every situation where the PMOLED or OLED device will be used and are used here only as an example.