**4. Characterization and analysis**

The discrete roll-to-roll printing system was used to check the printing capability, to characterize ink properties, and to study the system design. From system's point of view the continuous roll-to-roll printing processes handled the large area substrate before it was cut into small pieces for discrete roll-to-roll printing processes. From automation's point of view both the continuous and discrete roll-to-roll printing processes were automatically performed but the lamination process was semi-auto for alignment. The final production result of this study was not influenced by any factor of the system or the automation settings. Figure 28 is the pictures for the final demonstrators of (a) 3×3 active matrix array and (b) 21×39 passive matrix array. The 3×3 active matrix array device was examined and evaluated for various characteristics.

Possibilities for Flexible MEMS:Take Display Systems as Examples 399

made with Al and Ag. As predicted by simulation in section 2.2, Ag samples showed better distribution for the three primary colors than Al samples did. Ag samples actually showed distinguishable red improvement, better blue, and comparable green. An averaged larger than 50% *CPD* refinement can be seen in red and blue colors. The *CPD* also implies that green color almost reached the target; blue color has little room to be improved; and red

From Figure 29, around 20% transmittance difference was visible in both Al and Ag samples between simulation and real devices. This is believed to be the offset of simulation and real device. Besides the offset, as high as 40% unbalanced transmittance difference was in all Al colors but there is only less than 10% unbalanced transmittance difference in all Ag colors. Another important factor when judging the transmittance is its intensity. As shown in Figure 29(a)-(b), the intensities of all colors made by Al electrode are weak. This will become a serious perceptual issue if the substrate or extra protection layer absorbs some more intensity or when the backlight is weak. The common goal for all kinds of transmissive display device or color filter is to increase its transmittance intensity. In contrary, Ag

The description in section 2.2 and the structure design implies that any variation of optical property and thickness of each layer will influence the output transmittance and color purity, it is important to understand how serious can these variation be. Table 3 is the simulation results for different Ag thicknesses. Since Ag was originally designed for 20nm and precise control was difficult, here a 25% (5nm) is set for the simulation. From the *CPD* point of view one may guess that 25nm should be the best setting for the smallest *CPD* for all colors. However, 5nm thickness lowered the highest transmittance peaks in red (618nm), green (562nm), and blue (426nm) for 7%, 15%, and 9%, respectively. Thus a 20nm Ag was finally decided as the process target. Table 4 is the simulation results for different isolation thicknesses, the variation is set to be ±40nm from each target. From the tables one can find

samples showed higher intensity from both simulation and real device results.

color still has great space to be refined.

**4.1.3 Transmittance** 

Fig. 30. Color purity comparison of previous and this works.

**4.1.4 Process variation induced color shift** 

Fig. 28. (a) 3×3 active-matrix and (b) 21×39 passive-matrix demonstrators.

#### **4.1 Optical performance 4.1.1 Color purity**

The color purity simulation was done in section 2.2. The main influencing factor for color purity was explained in the same section by the control of metal electrodes. The best performance was concluded with 20nm Ag for its relatively balanced *n* value and relatively smaller *k* value. The color purity in CIE 1931 chromaticity diagram was measured by color tester (Yokogawa Denki, 3298F) with (*x, y*) axis system. A luminescent light (5500k) was used as the backlight. Figure 29(d) is the real performance of Ag electrode devices. The results typically followed the simulations and purer colors can be expected. From the datasheet PEN shows the best cutoff in UV region. Since strong UV light is harmful to human eyes and this MEMS display device claims sunlight as the backlight, a suitable substrate like PEN helps a lot on the UV cut out performance. When considering the visible region (400-700nm) in Figure 29(d), one can easily understand that only a single main peak appears in one color design which advanced each color's purity.

Fig. 29. Optical transmittance of different samples.

#### **4.1.2 Color purity deviation**

A color purity deviation (*CPD*) was defined in Equation 16 and was used to judge the color purity improvement in Figure 13. Figure 30 is the CIE chromaticity diagram for real devices made with Al and Ag. As predicted by simulation in section 2.2, Ag samples showed better distribution for the three primary colors than Al samples did. Ag samples actually showed distinguishable red improvement, better blue, and comparable green. An averaged larger than 50% *CPD* refinement can be seen in red and blue colors. The *CPD* also implies that green color almost reached the target; blue color has little room to be improved; and red color still has great space to be refined.

Fig. 30. Color purity comparison of previous and this works.

#### **4.1.3 Transmittance**

398 Microelectromechanical Systems and Devices

The color purity simulation was done in section 2.2. The main influencing factor for color purity was explained in the same section by the control of metal electrodes. The best performance was concluded with 20nm Ag for its relatively balanced *n* value and relatively smaller *k* value. The color purity in CIE 1931 chromaticity diagram was measured by color tester (Yokogawa Denki, 3298F) with (*x, y*) axis system. A luminescent light (5500k) was used as the backlight. Figure 29(d) is the real performance of Ag electrode devices. The results typically followed the simulations and purer colors can be expected. From the datasheet PEN shows the best cutoff in UV region. Since strong UV light is harmful to human eyes and this MEMS display device claims sunlight as the backlight, a suitable substrate like PEN helps a lot on the UV cut out performance. When considering the visible region (400-700nm) in Figure 29(d), one can easily understand that only a single main peak

A color purity deviation (*CPD*) was defined in Equation 16 and was used to judge the color purity improvement in Figure 13. Figure 30 is the CIE chromaticity diagram for real devices

Fig. 28. (a) 3×3 active-matrix and (b) 21×39 passive-matrix demonstrators.

appears in one color design which advanced each color's purity.

Fig. 29. Optical transmittance of different samples.

**4.1.2 Color purity deviation** 

**4.1 Optical performance 4.1.1 Color purity** 

> From Figure 29, around 20% transmittance difference was visible in both Al and Ag samples between simulation and real devices. This is believed to be the offset of simulation and real device. Besides the offset, as high as 40% unbalanced transmittance difference was in all Al colors but there is only less than 10% unbalanced transmittance difference in all Ag colors. Another important factor when judging the transmittance is its intensity. As shown in Figure 29(a)-(b), the intensities of all colors made by Al electrode are weak. This will become a serious perceptual issue if the substrate or extra protection layer absorbs some more intensity or when the backlight is weak. The common goal for all kinds of transmissive display device or color filter is to increase its transmittance intensity. In contrary, Ag samples showed higher intensity from both simulation and real device results.

#### **4.1.4 Process variation induced color shift**

The description in section 2.2 and the structure design implies that any variation of optical property and thickness of each layer will influence the output transmittance and color purity, it is important to understand how serious can these variation be. Table 3 is the simulation results for different Ag thicknesses. Since Ag was originally designed for 20nm and precise control was difficult, here a 25% (5nm) is set for the simulation. From the *CPD* point of view one may guess that 25nm should be the best setting for the smallest *CPD* for all colors. However, 5nm thickness lowered the highest transmittance peaks in red (618nm), green (562nm), and blue (426nm) for 7%, 15%, and 9%, respectively. Thus a 20nm Ag was finally decided as the process target. Table 4 is the simulation results for different isolation thicknesses, the variation is set to be ±40nm from each target. From the tables one can find

Possibilities for Flexible MEMS:Take Display Systems as Examples 401

Here *La*, *Lb*, and *Lc* are the three lengths of a triangle; (*xR, yR*), (*xG, yG*), and (*xB, yB*) are the coordinates of each point. The smallest value of *AOFF* happens on the 750nm spacer case within the three trials. Nevertheless, this spacer layer was gravure printing process prepared and its rheology characteristics have been plotted in Figure 27(c) which showed very limited linear region. In order to operate the gravure printing with sufficient process window, a compromised design of 600nm which located in the center of the linear rheology

SiO2=160nm (0.25, 0.22) (0.30, 0.24) (0.23, 0.25)

SiO2=325nm (0.30, 0.25) (0.25, 0.26) (0.22, 0.25)

SiO2=245nm (0.22, 0.27) (0.22, 0.26) (0.28, 0.24) *AOFF* 0.001943 0.000300 0.000198

Even though the MEMS model also predicted several different solutions to lower the operation voltage with Equation 20, a better solution which avoids changing the device's vertical design was proposed in section 2.4. The introduction of air channel is believe to be helpful to reduce the air pressure trapped inside a single pixel when ON. As shown in Figure 31, the Newton's ring means the colorful interference part at the edge of an interferometer. The root cause of Newton's ring was the different optical interference path lengths (Γ). These different optical path lengths in turn represented different output interfered colors. An interesting behavior in the figure is that the Newton's ring's size neither increases linearly nor increases infinitely.

Table 5. CIE coordinates with spacer height skew.

Fig. 31. A schematic plot to explain how Newton's ring took place.

**4.2 Structural performance** 

**4.2.1 Air channel** 

Spacer=450nm Spacer=600nm Spacer=750nm

region was chosen.

Red

Green

Blue

serious color shifts along all settings, which means that slight process difference will result in great color change. Since the isolation layer does not have transmittance issue, the thickness selection was made at the best locations for each color so that the smallest *CPD* also took place in these designs.


Table 3. C*PD* values with electrode layer (Ag) thickness skew.


Table 4. C*PD* values with isolation layer (SiO2) thickness skew.

The thickness variation of spacer layer was also performed and summarized in Table 5. Since there is no target value, no *CPD* calculation had been made. However, the data points represent the OFF state color output, thus how to keep these data points as close as possible to let OFF state color the same is very important. One simple way to check how close those data points are is to calculate the triangle area enclosed by those data points. The Heron's formula (or Hero's formula) describes the triangle's area (*AOFF*) by its three lengths:

$$A\_{OFF} = \sqrt{s(s - L\_d)(s - L\_b)(s - L\_c)}\tag{27}$$

where

$$s = \frac{L\_{\text{fl}} + L\_{\text{fl}} + L\_{\text{C}}}{2} \tag{28}$$

and

$$\begin{aligned} L\_{a} &= \sqrt{(x\_{R} - x\_{G})^{2} + (y\_{R} - y\_{G})^{2}} \\ L\_{c} &= \sqrt{(x\_{G} - x\_{B})^{2} + (y\_{G} - y\_{B})^{2}} \\ L\_{c} &= \sqrt{(x\_{B} - x\_{R})^{2} + (y\_{B} - y\_{R})^{2}} \end{aligned} \tag{29}$$

Here *La*, *Lb*, and *Lc* are the three lengths of a triangle; (*xR, yR*), (*xG, yG*), and (*xB, yB*) are the coordinates of each point. The smallest value of *AOFF* happens on the 750nm spacer case within the three trials. Nevertheless, this spacer layer was gravure printing process prepared and its rheology characteristics have been plotted in Figure 27(c) which showed very limited linear region. In order to operate the gravure printing with sufficient process window, a compromised design of 600nm which located in the center of the linear rheology region was chosen.


Table 5. CIE coordinates with spacer height skew.

## **4.2 Structural performance**

#### **4.2.1 Air channel**

400 Microelectromechanical Systems and Devices

serious color shifts along all settings, which means that slight process difference will result in great color change. Since the isolation layer does not have transmittance issue, the thickness selection was made at the best locations for each color so that the smallest *CPD*

SiO2=160nm 0.26 0.21 0.16 0.13

SiO2=325nm 0.22 0.17 0.12 0.09

SiO2=245nm 0.10 0.04 0.00 0.03

Red SiO2=120nm SiO2=160nm SiO2=200nm

Green SiO2=285nm SiO2=325nm SiO2=365nm

Blue SiO2=200nm SiO2=245nm SiO2=285nm

formula (or Hero's formula) describes the triangle's area (*AOFF*) by its three lengths:

The thickness variation of spacer layer was also performed and summarized in Table 5. Since there is no target value, no *CPD* calculation had been made. However, the data points represent the OFF state color output, thus how to keep these data points as close as possible to let OFF state color the same is very important. One simple way to check how close those data points are is to calculate the triangle area enclosed by those data points. The Heron's

> 2 *LLL aac <sup>s</sup>*

2 2 ( )( )

*L xx yy <sup>a</sup> R R G G*

*L xx yy <sup>b</sup> G G B B*

*L xx yy <sup>c</sup> BR BR*

2 2 ( )( )

2 2 ( )( )

Table 3. C*PD* values with electrode layer (Ag) thickness skew.

Table 4. C*PD* values with isolation layer (SiO2) thickness skew.

Ag=15nm Ag=20nm Ag=25nm Ag=30nm

0.39 0.21 0.40

0.34 0.17 0.35

0.16 0.04 0.22

*A ss L s L s L* ( )( )( ) *OFF a c <sup>b</sup>* (27)

(28)

(29)

also took place in these designs.

Red

Green

Blue

where

and

Even though the MEMS model also predicted several different solutions to lower the operation voltage with Equation 20, a better solution which avoids changing the device's vertical design was proposed in section 2.4. The introduction of air channel is believe to be helpful to reduce the air pressure trapped inside a single pixel when ON. As shown in Figure 31, the Newton's ring means the colorful interference part at the edge of an interferometer. The root cause of Newton's ring was the different optical interference path lengths (Γ). These different optical path lengths in turn represented different output interfered colors. An interesting behavior in the figure is that the Newton's ring's size neither increases linearly nor increases infinitely.

Fig. 31. A schematic plot to explain how Newton's ring took place.

Possibilities for Flexible MEMS:Take Display Systems as Examples 403

The main target on the electrical performance of this MEMS display device is to reduce its operation voltage as described in the design part in section 2.3. However, to reduce the upper layer thickness together incorporate with the handling issue in which the electrostatic force is too strong on the <20m PEN. The thin PEN will be easily attracted to the rubber pad and other equipment parts during printing and lamination processes by electrostatic force. The ultra thin upper layer will also induce special concerns on reliability. If extra layer should be added unto the whole structure, the transmittance and the optical performance should also be re-designed. Thus to reduce the thickness of the upper layer is not adequate.

Spacer=400nm Spacer=500nm Spacer=600nm

To reduce the spacer height is not also a proper solution because the spacer height in OFF state also influences the output color as described in section 4.1. Table 6 is the list of simulated *CPD* under OFF state with white target of (0.31, 0.31) on CIE 1931 chromaticity diagram. The design goal not only fell on the small *CPD* but also required a small *CPD* difference between different colors. Both 400nm and 600nm spacer height designs are with smallest *CPD* differences ( 0.10 0.07 0.12 0.09 0.03 ) but from the gravure printing characteristic point of view in Figure 27(c), the original 600nm design falls on the center part of the linear region thus provides more confidence on process control. Since the isolation layer thickness is the key for color interference, to reduce its thickness while keeping the same color design is then inaccessible. The final possibility fell on the pixel size and since this study aims on a large area display device for decoration, a 2000m pixel size was set in section 2.3. Note that even though a 15V operation voltage was simulated in the same section, actual driving voltage was far

Table 6. *CPD* and its CIE coordinate under different spacer height settings.

higher than expectation in previous publication as listed in Table 7.

Red, SiO2=370nm

Table 7. Operation voltage difference between simulation and real device [35].

(0.31, 0.21) (0.19, 0.33) (0.30, 0.24) 0.10 0.12 0.07

(0.25, 0.24) (0.25, 0.26) (0.25, 0.26) 0.09 0.08 0.08

(0.19, 0.31) (0.28, 0.24) (0.22, 0.26) 0.12 0.08 0.10

> Green, SiO2=310nm

Sim. Real Sim. Real Sim. Real

Voltage 65V 153V 55V 118V 40V 101V aperture 19% 19% 22%

Voltage 40V 153V 30V 118V 28V 101V aperture 54% 54% 58%

Voltage 34V 153V 25V 118V 22V 101V aperture 70% 68% 70%

Voltage 28V 153V 25V 118V 18V 101V aperture 76% 78% 78%

Blue, SiO2=240nm

**4.3 Electrical performance** 

Red

Green SiO2=325nm

Blue

SiO2=160nm

SiO2=245nm

Pixel Size Item

200μm

400μm

600μm

800μm

**4.3.1 Factors influencing electrical performance** 

Rather, the increment decreased along the increasing pixel size. A special saturation behavior of the Newton's ring's size appeared in Figure 32(a). This means that the increased pixel size will only make the display aperture looks larger instead of reduce the Newton's ring. This conclusion strongly supports the necessity of a revolutionary structure change. With different spacer coverage designs – 100% (no air channel), 90%, 80%, and 60% – the experimental data showed great amount of improvement. The width of Newton's ring reduced with all the coverage designs in Figure 32(b).

Fig. 32. Different spacer coverage alleviated the operation voltage.

Figure 33 is the picture which explains why saturation took place even with air channel design and why larger air channel did not yield in smaller Newton's ring: When the air channel is spacious enough, high applied voltage will let the lower layer attract the upper layer in the air channel area. Since the upper layer was put on the spacer and both sides (pixel area and air channel area) were competing each other, a see-saw performance showed – The more the air channel area in contact, the less the pixel area in contact. Thus a proper instead of a wide air channel is preferred. In this experiment a 90% coverage showed the smallest Newton's ring. This behavior was also obvious during simulation in Figure 20 and Figure 21: The display area tended to expand to the central part of air channel when the channel was wide enough but the display area tended to expand to the four corners when the channel with the same coverage was divided into two parts and were put aside. A combination of the experimental and simulation data suggested narrow and separate air channels are better. However, consider the resolution of printing process and the function of spacer layer for lamination, a single and large air channel was decided for the final structure.

Fig. 33. Contact areas protruded into spacer areas when design was not optimized.
