Thin film transistor liquid crystal display (TFT-LCD) panels, possessing various advantages such as long service life, low power consumption, and high image quality, have been widely used in mobile phones, notebooks, monitors, TVs, and other display devices^[1-3]. LCDs use the birefringence property of liquid crystal molecules at different voltages to modulate the display gray levels^[4]. The color filter (CF) films are one of the most important components of LCDs, which determine the color gamut, brightness, contrast, and other key properties of TFT-LCD products. Their main function is to enable LCDs to realize full-color display.

The current mainstream CFs of TFT-LCDs are composed of red, green, and blue primitives, corresponding to red, green and blue-colored photoresists, respectively. CFs are generally composed of resins, photoinitiators, monomers, pigments or dyes, solvents, additives and other components^[5-8]. CFs can be prepared by the conventional photolithography process^[9]. As an important bearer of LCD image quality, it is a significant task to develop CFs with improved characteristics. At present, research on CFs for LCDs mainly focuses on three aspects: color gamut^[10-14], transmittance^[15] and contrast ratio^[16-19]. Nevertheless, there are few works on the reliability or ease of manufacturing CFs. Bubble defect is one of the most severe reliability issues in TFT-LCD during manufacture and use of display products. The main source of bubbles is gas released from CFs. To reduce the bubble defect, it is inevitable to increase the pumping time, which impairs manufacturing efficiency. In our company, when replacing the reference photoresist 1 (PT-1) with a new photoresist 2 (PT-2), we found that during the one-drop filling (ODF) process, the working time of the vacuum assembly system (VAS) increased by approximately 3.5 s. The increased time results in a significant direct economic loss of nearly 10 million CNY annually. The consequential damage is enormous and intolerable. For this reason, it is helpful to explore why different photoresist materials manifest various gas releases, and on this basis, to propose an effective solution to decrease gas release.

Herein, we focus on the factors that may affect gas storage and release in photoresist materials, such as thermal stability, compactness and hydrophobicity. Thermal gravimetric analysis (TGA) measurement demonstrated water vapor as the main gassource. Polymers and monomers were then adjusted to check the effect of compactness and hydrophobicity. It was discovered that increasing the compactness and decreasing hydrophilicity could effectively reduce the gas release time during the vacuum drying (VCD) process. With this strategy, a new type of green CF (sample F) was successfully demonstrated, which could reduce pumping time and offer comparable optical performance.

There are six basic components of color photoresists studied in this article, as follows:

(a) Color pastes, G58 and Y138, were provided by SANYO Color WORKS (Tokoyo, Japan).

(b) Polymers, P1 and P2, were purchased from Resonac Highpolymer Materials (shanghai, China).

(c) Monomers, G6-AG-001 and G6-AG-002 were purchased from TOAGOSEI New Technology (Jiangsu, China, as shown in Fig. 1). DPEA-12 and DPCA-60 were purchased from Nipponkayaku (Hiroshima, Japan).

(d) The photoinitiator, oxime ester I-1, was purchased from Tronly New Electronic Materials (Jiangsu, China).

(e) The additive, F-554, was purchased from DIC (Tokyo, Japan).

(f) Solvents, 2-acetoxy-1-methoxypropane (PGMEA) and thiosalicylic acid (MBA) were obtained from Merck and Daicel, respectively. The solvents were used without any further purification.

Based on the six materials mentioned above, different formulations of green CFs were designed, as listed in Tab. 1.

A TFT-LCD glass (10 cm × 10 cm, 0.5 mm thick; AGC, Japan) was used as a substrate, and cleaned with an EUV machine (model name SUS720, Ushio, Tokyo, Japan) before use. A tabletop spin coater (MIKASA, Tokyo, Japan) was used to coat color photoresists to form films. A hot plate (AS ONE, OSAKA, Japan) was used to remove solvent by baking film. An exposure (KEYI Technology, Taiwan, China) and a developing machine (KEYI Technology, Taiwan, China) were used to form patterns. The exposure energy was 40 mJ and the developing time was 60 s. The developing time was defined the period during which the photoresist layer and reveal patterns were washed with the developing solution of 0.04% KOH (aq). The developing time follows the condition of mass production. Finally, patterns were baked at 230 °C for 20 min in a clean oven.

The profiles of the resulting curing patterns were observed by optical microscope (OLYMPUS, Japan) and scanning electron microscope (Gemini SEM 300, Zeiss, German).

The film thickness was measured using an alpha-step profiler (ET4000A, KOSAKA, Japan).

The transmittance spectrum of the patterns was collected using a microspectrophotometer (Lambda Vision, Japan).

The photoresist layer was scraped off the substrate and ground into powder. 2 mg of the powder was heated (10 °C/min) under the N₂ atmosphere. The temperature was first increased to 100 °C and held for 20 min to remove water. The samples were then heated to 230 °C for 50 min, 210 °C for 60 min and 230 °C for 60 min, sequentially.

The decomposition of CFs during high-temperature processing was confirmed using a thermal gravimetric analyzer (STA449 F5, Netzsch, Germany).

To check the hydrophobicity effect, the photoresist powder was annealed at 230 °C for 60 min to eliminate the effect of the water content and decomposition. Next, the baked powder was placed in an environment of 120 °C, 2 bar pressure, and 100% humidity for 12 hours. After treatment, 2 mg of the powder was tested by TGA.

The VCD test was completed in KOLON Lab (Japan). The photoresist layer was annealed at 230 °C for 60 min to eliminate the effect of the water content and decomposition. Then, the samples were placed in an environment of 120 °C, 2 bar pressure, and 100% humidity for 12 hours. The layer was immediately transferred to the vacuum drying device. The time required to extract the gas to half an atmosphere was recorded.

The hydrophilic properties of different photoresist compositions were tested using a contact angle tester (SINDIN, China). Deionized water was dropped onto the pretreated photoresist layer, and the contact angle between water and photoresists was recorded.

As shown in Fig. 2 (color online), we assumed three main possible routes of gas release during panel manufacturing, which are respectively related to the thermal stability, compactness and hydrophobicity of CFs. As organic mixtures, the CFs may be oxidized and decompose to release gas molecules such as carbon dioxide^{[ 8-9, 20-21]}. In this hypothesis, gas released during fabrication comes mainly from the inner side of the photoresists. Regarding the other two factors, compactness and hydrophobicity are associated with gas storage capability. The gas from the environment first enters the micro-cavities of photoresists. Then, it remains in the photoresist layer, and is finally released during VCD, resulting in a longer vacuum time.

Thermogravimetric analysis (TGA) was applied to determine whether the gas increase resulted from the materials decomposition. The reference photoresist, PT-1, and the new photoresist, PT-2, were placed in the oven under the nitrogen atmosphere. To mimic the baking process of polymeric flat films curing on arrays (PFA), indium tin oxide (ITO) coating and polyimide (PI) curing process, the samples were baked at 230 °C for 50 min, 210 °C for 60 min and 230 °C for 60 min, sequentially (Fig. 3, color online). From 40 min to 215 min, PT-1 exhibited 2.46% mass loss, while PT-2 displayed only 1.50% mass loss (Table 2). The mass loss of photoresists during the baking process seems related to the decomposition of photoresists. CFs may decompose at high temperatures releasing gas molecules such as carbon dioxide^{[8-9, 20-21]}. The results indicate that PT-2 is probably more thermally stable than PT-1 and thus releases less gas, which in principle, should lead to a shorter pumping time in fabrication. However, in reality, the pumping time for PT-2 in the production line is 3.5 s longer than PT-1, which appears inconsistent with the thermal stability of PT-1 and PT-2. Considering that the panels are fabricated in an ambient atmosphere, the gas generated by decomposition should have already escaped during the device transfer before VAS, therefore, the minimal thermal stability difference of PT-1 and PT-2 may not change the pumping time. Therefore, the thermal stability of photoresists should have a negligible correlation with gas release.

It is worth noting that the mass loss trend was different in the two TGA curves, when two photoresists were pre-baked. During the first 30 min, the temperature gradually rose to 100 °C. The mass of PT-1 decreased from 100% to 99.80%, while that of PT-2 decreased to 99.15%. The mass loss at around 100 °C was speculated to be related to the water vapor due to the difference in hydrophobicity between PT-1 and PT-2.

Therefore, the TGA experiment demonstrated that the longer pumping time of PT-2 was not related to the decomposition of photoresists. The results suggested that the gas release was probably related to the water vapor capacity of the photoresist layer, which might be affected by the compactness and hydrophobicity of CFs. Therefore, the compactness and hydrophobicity of the other two main materials of the photoresists (polymers and monomers) were then analyzed.

Since polymer components constitute the backbone of CFs, the compactness of CFs will be mainly influenced by the polymers. The polymer crosslinkers formed micro-porous structures, where the gas is probably stored^[22-23]. Denser pores make it harder for air to enter the photoresist layer.

To verify the above assumption, we choose two different polymers: P1 and another new polymer, P2. P1 is the polymer used in PT-1 (Fig. 4(a), color online). In P1, the repeating unit contains one reactive double bond, a short and a long chains. Therefore, after curing, the double bonds in the polymer are cross-linked (red bars in Fig. 4(b), color online), and P1 forms a porous architecture with large voids (dotted circles in Fig. 4(b)). The voids can absorb and retain the gas from the environment. In terms of P2, there are two double bonds and a long side chain per unit, which enables the formation of much more compact structures than P1. Two different photoresists were prepared with various ratios of P1 and P2. Sample A consisted of only polymer resin P1 and Sample B consisted of mixed resins of 50% P1 and 50% P2.

To analyze their difference in compactness, the shrinkage of the photoresist layer was tested by comparing the thickness before (d₁) and after (d₂) UV exposure. Sample A shrank by 0.37% and Sample B shrank by 0.76% (Table 3). More considerable shrinkage demonstrated that Sample B was more compact because of the higher degree of crosslinking. The SEM images showed that Sample B displayed a flatter section with fewer porous structures than Sample A, suggesting a denser polymeric structure and higher compactness (Fig. 4(c), color online). The VCD method was exploited again to simulate the pumping condition during manufacturing. First, the samples were treated with high temperature and humidity, then the extraction time for depressurizing the sample to half an atmosphere was recorded. The pumping time of Sample A was 7 s while that of Sample B was only 5.5 s, demonstrating that compact polymers could effectively reduce pumping time (Fig. 4(d), color online ). Therefore, the compactness of the photoresist layer could significantly affect the gas release by changing the gas storage capacity. Overall, polymers with more compact architecture can be exploited to form a more rigid photoresist layer and reduce the pumping time during fabrication.

The hydrophilicity of the photoresist layer may also affect the gas release and is strongly determined by the nature of monomers. Hydrophilic monomers, such as carboxyl or hydroxyl groups, can absorb water vapor from the environment. In contrast, hydrophobic monomers with long alkyl chains, can repel the proximity of water molecules (Fig. 5(a), color online).

To explore the relationship between hydrophobicity and gas release, a series of samples were prepared (Table 4). Sample C only contained one monomer, M1, which was employed in PT-1. Sample D consisted of 30% hydrophilic monomer M2 and 70% M1. To prepare Sample E, 30% hydrophobic monomer M3 was added to M1. The developing time of these samples was then tested to check their hydrophobicity. Hydrophobic samples were more difficult to wash away during the developing process, thus displaying a longer developing time. As expected, with the addition of M2, Sample D displayed a shorter developing time than Sample C. Sample E with hydrophobic monomer M3 exhibited the longest developing time. In addition, the contact angles of three samples were tested, as shown in Fig. 5(b). Sample E exhibited the largest contact angle, demonstrating the highest hydrophobicity. These samples were later treated with high temperature and humidity before being tested with VCD. The pumping time to half an atmosphere of Sample C was 7 s, slightly shorter than Sample D (7.5 s). That of Sample E was only 3 s (Fig. 5(c), color online). These results showed that the higher proportion of hydrophilic monomers led to more gas release due to higher water content in the photoresist layer. In contrast, hydrophobic monomers could prevent the absorption of water vapor from the environment and reduce gas generation during VAS. Furthermore, these three samples were tested with TGA (Fig. 5(d), color online). After treatment with high temperature and humidity, Sample E showed the lowest mass loss (3.5%), while Sample D displayed the highest mass loss (23.4%), meaning that hydrophobic photoresists could effectively prevent water vapor intrusion. The results were consistent with the mass loss of PT-1 and PT-2 at around 100 °C. Therefore, hydrophobic monomers containing long alkyl chains can efficiently reduce gas release from the photoresist layer and decrease the pumping time during fabrication.

Based on the above experimental results, we can see that the denser the color barrier and the better the hydrophobicity of CFs, the more beneficial it would be to reduce the gas release and the pumping time. In this case, an optimized CF (sample F) containing dense and hydrophobic monomers was designed with the detailed composition listed in Table 1.

The morphologies of sample F were characterized as shown in Fig. 6 (color online), and the optical characteristics were listed in Table 5. It is obvious that both the optical and manufacturing characteristics of sample F are comparable to those of the commercial reference green CF, which could well meet the requirements of the fabrication specification.

Finally, a TFT-LCD prototype was manufactured, adopting sample F as green CF. A colorful photo of different types of fruits (Fig. 7, color online) was created using this demo. These results indicate that the sample F is a suitable candidate for commercial applications.

In summary, the relationship between the compositions of CF materials and gas release was studied to increase the production efficiency of TFT-LCD displays. Preliminary analysis confirmed that the gas release of the CF layer was barely related to the thermal stability of the materials. The main reasons for the long pumping time were the compactness and hydrophobicity of polymers and monomers in CFs. The hydrophilic groups could capture water and gas from the atmosphere and retain the absorbed gas in the porous structure. By utilizing more compact and hydrophobic materials in photoresists, the vacuum drying process could be sufficiently reduced. Finally, a new type of green CF (sample F) with slightly reduced pumping time and comparable optical performance to commercial benchmarks was successfully demonstrated. This strategy can be universally applied to developing new photoresists for the LCD industry.