3.1.

Characterization of Microlens Arrays

A large-area PS microsphere monolayer on a 3-in. Si wafer fabricated by a convective and capillary assembly is shown in Fig. 3(a). Up to a few square inches of PS monolayer could be created. Although grain boundaries are observed between individually close-packed PS domains, hexagonally close-packed PS microsphere arrays are well formed as shown in Fig. 3(b). We observed that PS monolayers could be easily formed for 20- and 100-μm-size PS microspheres. However, it was more challenging to generate a large-area monolayer for 5-μm-size PS microspheres due to the increased impact of the capillary force; therefore, they are not used for the later device studies.

Fig. 3

(a) Convective and capillary assembly is employed to form a large-area PS microsphere monolayer on a 3-in. Si wafer (coated with a thin SiO₂ layer). (b) Optical microscope image of a PS microsphere monolayer. 100-μm-size PS microspheres are used in both (a) and (b).

Figure 4(a) is a cross-sectional scanning electron microscopic (SEM) image showing the completely embedded PS microspheres in the cured PDMS after removing the substrate. The subsequent application of scotch tape could well remove the embedded PS microspheres and PDMS residues as shown in Fig. 4(b), leading to a concave PDMS mold successfully as shown in Fig. 4(c). A large-area microlens array using a PDMS mold is shown in Fig. 4(d) and highly close-packed microlens arrays are also shown in the inset of Fig. 4(d). The measured FF of this microlens array (from 100-μm PS microspheres) is (85 ± 3)%, which is very close to the ideal FF of two-dimensional hexagonally close-packed spheres (91%).

Fig. 4

(a) Cross-sectional SEM image of embedded 100-μm-size PS spheres in cured PDMS. (b) The PDMS mold after (intentional) incomplete removal of 20-μm-size PS spheres. (c) Concave PDMS template for microlens arrays after completely removing 20-μm-size PS spheres. (d) Large-area 100-μm-size microlens arrays subsequently made with the concave PDMS mold. Inset: hexagonally close-packed lens arrays are observed. Scale bar is 100 μm in (a), (c), and (d), and 20 μm in (b).

The same microlens fabrication method could be applied to different sizes of PS microspheres; however, the contact angle θ_c of the PDMS mold is dependent on ϕ_PS as shown in the cross-sectional SEM images of Figs. 5(a)–5(c). As the size of PS microspheres was increased, the shape of PDMS mold was closer to the perfect hemisphere, showing almost hemispherical shape for ϕ_PS = 100 μm. We believe that this is related to the mechanical robustness of the PDMS mold in the narrow regions between individual features. When the PDMS precursor was poured onto the PS microspheres in order to make a concave template, it could completely cover the PS microspheres, penetrating all the way down to the SiO₂ surface through the interstitials among the PS microspheres [see Fig. 4(a)], which after curing form columns connecting PDMS below and above the PS microspheres. When the scotch tape was applied to remove the PS microspheres from the thermally cured PDMS, it is ideal that PDMS breaks along the a-a^′ plane in Fig. 5(d), leaving concave hemisphere features in the mold. We believe that PDMS columns using ϕ_PS = 100 μm are mechanically robust and broken at the narrowest neck area along the a-a^′ plane. However, narrower columns using smaller ϕ_PS are more easily torn off at locations away from the narrowest neck area (for example, along the b-b^′ plane), resulting in height less than the PS microsphere radius and, in turn, lower contact angles for the PDMS mold.

Fig. 5

Cross-sectional SEM images of concave PDMS molds fabricated using (a) ϕ_PS = 5 μm, (b) ϕ_PS = 20 μm, and (c) ϕ_PS = 100 μm. (d) Schematic illustration of mechanically weak points in the PDMS mold when removing PS spheres from the cured PDMS. The PS separation is expected along the a-a^′ plane for 100-μm-size PS spheres, but along a lower plane (such as the b-b^′ plane) for smaller size PS spheres due to the (unwanted) removal of PDMS in the mechanically weak points.

The contact angles of microlens using different PDMS molds clearly showed the difference in θ_c as shown in Figs. 6(a)–6(c). Here, θ_c, averaged over more than 10 lenses, was obtained by drawing a tangential line at the edge of a lens and measuring its angle with the substrate surface. We obtained θ_c = (45 ± 5 deg) with ϕ_PS = 5 μm, which is similar to previous reports.^{18, 19} As the ϕ_PS was increased to 20 and 100 μm, θ_c was increased to (65 ± 5 deg) and (85 ± 5 deg), respectively, as shown in Fig. 6(d). The semicircle drawn in the cross-sectional microlens image for ϕ_PS = 100 μm [see Fig. 6(c)] clearly shows the hemispherical shape of the lens. The FF of the microlens array was estimated by counting the number of lenses per unit area and measuring the base diameter of lens. The FF for the microlens array from ϕ_PS = 20 μm was only (67 ± 3)%, but increased to (77 ± 3)% and (85 ± 3)% for ϕ_PS = 20 and 100 μm, respectively, as shown in Fig. 6(d).

Fig. 6

Cross-sectional SEM images of convex microlenses with different sizes of PS spheres: (a) ϕ_PS = 5 μm, (b) ϕ_PS = 20 μm, and (c) ϕ_PS = 100 μm. In (c), the dashed semicircle line is drawn for comparison. (d) The PS sphere size (ϕ_PS) dependence for the contact angle (θ_c) of microlens and the lens array fill factor (FF).

3.2.

Enhancement in OLED Light Extraction due to Microlens Arrays

The light-extraction enhancement by microlens arrays (fabricated from 20- or 100-μm PS spheres) is investigated based on two different sizes of FOLEDs. For small-area (2 × 2 mm²) FOLEDs, Fig. 7 shows the external quantum efficiencies (η_EQE) with and without microlens arrays based on 100-μm PS spheres. The η_EQE was obtained from the far-field luminous intensity measurement method and showed a light-extraction enhancement factor (defined as the light output with the microlens array to that without the microlens array) of f = (1.35 ± 0.04) in the entire current injection range. This enhancement factor, although confirming the applicability of using microlens arrays to enhance the light extraction in OLEDs, is somewhat lower than the previously reported enhancement factors using microlens arrays fabricated using other methods, which is in the range of 1.5–1.7 (Refs. 7 and 14).

Fig. 7

External quantum efficiencies (η_EQE) of small-area FOLEDs with and without microlens arrays. Light-extraction enhancement factor (f) of (1.35 ± 0.04) is calculated as the ratio of the light intensity with the microlens array to that without the microlens array.

Figure 8(a) compares the absolute luminous intensities for a large-area (12 × 12 mm²) FOLED with or without a microlens array (ϕ_PS = 100 μm). The FOLED is operated at J = 5 mA/cm², in both cases, and measured using an integrating sphere system. By integrating the area under the peak in each spectrum, we obtained a light-extraction enhancement factor of f = (1.70 ± 0.07) with the microlens array, which was significantly higher than the result for the small-area FOLED. We believe such an OLED size dependence on f is attributed to the Al cathode size. When photons are reached at the air/microlens interface, some photons may be reflected back toward the cathode instead of escaping the front surface. A larger metal cathode can redirect those backward traveling photons again to the forward directions, enhancing the chance of forward light extraction as illustrated in Fig. 8(b). However, with a smaller metal cathode as shown in Fig. 8(c), there is a significant backward air mode, leading to low light-extraction efficiency into the forward-viewing direction.

Fig. 8

(a) EL intensity difference with and without microlens arrays (fabricated from ϕ_PS = 100 μm) for large-area FOLEDs (12 × 12 mm). EL intensities are measured with the device operated at 5 mA/cm² in both cases using an integrating sphere system. Inset: Photo of the large-area FOLED covered with a microlens array on the glass surface (ϕ_PS = 100 μm). Schematic illustrations for light-extraction mechanism in (b) large-area and (c) small-area OLEDs with a microlens array attached to the glass substrate.

The enhancement factor for large-area devices was also confirmed by the near-field luminous intensity measurement. As shown in the inset of Fig. 9, only half of a large-area FOLED was covered with a microlens array (ϕ_PS = 100 μm). With the device operated at J = 3 mA/cm², the near-field optical intensity was recorded across the entire light-emitting area (between the bare side and the side covered with the microlens array) using an optical fiber that is perpendicular to the device. As illustrated in Fig. 9, we see that the electroluminescent (EL) intensity in the microlens covered region is ∼1.69 times higher than that in the bare side of the OLED. This is in good agreement with the integrating sphere measurement results for large-area FOLEDs.

Fig. 9

Near-field luminous intensity profile across a large-area FOLED. The right half of the FOLED was covered with a microlens array, while the other half was bare. The FOLED was operated at 3 mA/cm², and negligible degradation was confirmed before/after the measurement. The inset shows a picture of device in the measurement setup.

We have also studied the light-extraction enhancement in FOLEDs using microlens arrays fabricated from 20-μm PS microspheres. As summarized in Fig. 10, enhancement factors of f = (1.25 ± 0.04) and (1.45 ± 0.05) were obtained for small-area and large-area devices, respectively. In addition to the device area dependence as observed with larger microlenses, we also see that the absolute enhancement was somewhat lower in this case, as compared to using microlenses from 100-μm PS microspheres. We attribute this to the lower contact angles of microlenses from 20-μm PS microspheres rather than similar FF as shown in Fig. 6. Light-extraction efficiency can be enhanced significantly at larger angles than ∼41.8 deg [critical angle for TIR at the air (n = 1.0)/glass (n = 1.5) interface with the planar glass] with the attachment of microlens arrays on the surface of glass substrate, and previous results showed that higher contact angle microlens can extract photons more effectively than by the lower contact angle microlens^{7, 14}, which is consistent with our results here.

Fig. 10

Comparison of the light-extraction enhancement factor (f  ) for small-area (2 × 2 mm) and large-area (12 × 12 mm) FLOEDs and PHOLEDs, using microlens arrays fabricated from two different sizes of PS microspheres (ϕ_PS = 20 and 100 μm).

Figure 11 shows the angular emission patterns for the OLED with the microlens attachment (ϕ_PS = 100 μm). High θ_c of microlens is obtained using standard procedure (see Section 2) with low-viscosity optical adhesive (NOA61, 300 cps), whereas low θ_c is intentionally fabricated by dropping high-viscosity optical adhesive (NOA 68, 5000 cps) onto the glass surface of the OLED and instant curing after putting down the PDMS mold. Although bare OLED showed a slightly wider angular emission pattern than a Lambertian source due to the measurement setup, slightly forward directional and a negligible pattern change were observed with high and low microlenses, respectively, compared to that of bare OLED. We believe tall height of high θ_c microlens arrays blocks the light emission at larger angles and redirects it forward, which is opposite what Sun and Forrest demonstrated using 7-μm microlens arrays with the enhanced luminous intensities at larger angles (especially at >40 deg).¹⁴

Fig. 11

Normalized optical power intensity as a function of the viewing angle for a bare FOLED (red dashed line) and FOLEDs with low and high contact angle microlens arrays (blue open triangles and green solid squares, respectively), as well as that for a Lambertian source (black solid line).

The close-packed hemispherical microlens arrays (ϕ_PS = 100 μm) were also attached to the large-area PHOLEDs, and the light-extraction efficiency was examined using the near-field luminous intensity measurement method. Interestingly, somewhat lower f = (1.56 ± 0.06) was observed for both green and blue-green PHOLEDs compared to that for green FOLEDs (f = 1.69 ± 0.04) as shown in Fig. 10. We attribute this difference to the more significant optical modes confinement for the PHOLEDs with higher refractive indices of the emissive layers (n_CBP = 1.83 at λ_peak = 510 nm and n_mCP = 1.81 at λ_peak = 460 nm) than those of neighboring layers (n_NPB = 1.81 and n_BCP = 1.74, n_TAPC = 1.73 and n_3TPYMB = 1.66), measured by ellipsometry (see Fig. 2). Therefore, it is expected that less amount of photons could reach the glass substrate for the PHOLED, eventually leading to lower light-extraction enhancement by microlens arrays. The two PHOLEDs do exhibit nearly identical light-extraction enhancement factor, suggesting that the light-extraction mechanism we described here is not sensitive to the wavelength of emission; therefore, it could be potentially applied to white LEDs without altering the emission spectrum appreciably.

Note that the microlens arrays can be fabricated on the glass substrate prior to the deposition of the organic multilayer stack, which is on the other side of the glass substrate. Hence, it does not affect any possible degradation processes inside the OLED structure. The increased light-extraction efficiency does suggest that the OLED can be operated at a lower current and voltage in order to achieve the same operating brightness, which in turn prolongs the lifetime of the OLEDs.^{20, 21} We have also tested the microlens array fabrication on the surface of the polyethylene terephthalate (PET) substrate for application to flexible OLEDs. Shown in Fig. 12 is a photograph of a PET substrate with a microlens array attached using the same procedure as described above for fabricating microlens arrays on glass substrates. Although we have not fabricated OLEDs on the PET substrate and, therefore, the light-extraction enhancement factor in flexible OLEDs due to the microlens array is unknown, this clearly shows that the microlens array fabrication process described here could be potentially integrated with roll-to-roll fabrication of OLEDs on flexible substrates.

Fig. 12

Photograph of microlens arrays (ϕ_PS = 100 μm) on a flexible PET substrate. The microlens fabrication method applied to the glass substrate could be also compatible with the flexible PET substrate.