Inline Monitoring Detects CMOS Image Sensor Colorization Problems

CMOS-based image sensors are widely used in the digital imaging industry, and are rapidly gaining acceptance in consumer products. These sensors are common in digital cameras, cellular phones, webcams, toys and security cameras.

The need for imaging sensors with smaller pixel sizes (<2 µm) has become a commercial reality, but these advanced designs are challenging from an inspection and metrology standpoint. One of the critical process steps is the composition of the color filter array, where the inspection of colorization effects is required.

We describe here a new method for optical inline monitoring of colorization effects on CMOS image sensors in a high-volume fab.

Image sensor technology

Image sensors are coated with different color filters on top of the sensor pixels to capture all three colors with a single exposure of the camera. These image sensors are the most common type used in cell phone digital cameras, and are manufactured through a CMOS process.

CMOS image sensors consist of many p-n junctions, which build the sensor’s photodiodes (Fig. 1). Since silicon-based CMOS sensors are monochrome only, they cannot distinguish between the different wavelengths of the incoming light. A color filter array must be placed in front of the detectors. Usually a primary red, green and blue (RGB) color filter array is used, but secondary cyan, magenta and yellow (CMY) filters are also common. The color filter array is patterned on top of the photodiode using pigmented photoresist, which acts as a bandpass filter — only certain wavelengths can pass the filter and reach the detector. On top of the photoresist, an additional layer with microlenses bundles the incoming light.

1. A CMOS image sensor, as shown in this schematic, consists of photodiodes covered with a color filter array patterned with colored photoresist. Microlenses sit on top of the resist.

Color filter array issues

Typical defects in the manufacturing process are streaks on the color filter array (Fig. 2). They occur after liquid processes, mainly after resist spin-coating steps. The streaks are radial striations with a thickness of ~100 Å. The process problem is amplified if the previous layer has a high topography.

2. Typical defects in the manufacturing process are streaks on the color filter array, shown here with a 10× brightfield system (left) and a 150× DUV system (right).

There are two main causes for streak defects. At every liquid process step that involves drying the wafer, droplets can remain behind. Even when the majority of the liquid is spin-dried, some remaining droplets can hold their position, especially at the wafer’s center, where the centrifugal force is less than at the edge. The remaining droplets contain very small deposits of dye material, and this material will remain on the wafer surface when the droplets evaporate.

Another reason can be the resist itself. Colored resist is used in the composition of the color filter arrays. These resists are doped with color pigments to build the necessary RGB films on the wafer. These color pigments in the resist can agglutinate and build clusters. After spinning, these clusters can appear as streaks of extremely low optical contrast (Fig. 3).

3. A streak effect is shown with an original microscope image (left) and then processed with image analysis software (right).

As a result of these problems, the photoresist will discolor and a non-uniform distribution will occur over the color filter array. These effects reduce the sensitivity, resolution and finally the yield of the image sensor devices.

4. Since streaks are very thin and low-contrast, they cannot be detected by using classical micro defect detection systems.

5. Typical micro scanning tools also have limitations in finding low-contrast defects in color filter arrays.

Extensive investigations and tests to quantify and qualify the streak effect have been conducted. Since the streaks are very thin and have a very low contrast, they cannot be detected by using classical macro defect detection systems (Fig. 4). Even typical micro scanning tools are limited in finding these low-contrast defects (Fig 5).

A new, automatic and fast optical inspection method for colorization effects was developed by Vistec Semiconductor Systems in collaboration with STMicroelectronics. The successful detection of colorization problems became possible by using an automated macro defect inspection system (Vistec LDS3200) with a combined microscope module and special optimized analysis software.

The optical system is optimized (Fig. 6) to get one complete sensor pad into the field of view. Only one inspection shot is necessary to detect all resist problems within one sensor pad. The LDS3200 hardware can be adapted to the user’s specific requirements. It allows optimizing the magnification to fit with the sensor pad size for highest throughput.

In its present form, the tool scans a wafer in <6 minutes and delivers the inspection result for the full wafer. This throughput provides a very high sampling rate for excursion monitoring and fulfills the requirements needed to support high-volume production.

6. An optimized optical system gets a complete sensor pad into the field of view, making only one inspection shot necessary.

High-resolution image analysis software is implemented to detect low-contrast defects. It was necessary to adapt specific algorithms to find all defects of interest within one detection pass. Two major groups of defects have been identified — thin streaks and point-shaped defects; both kinds appeared very broad with low contrast.

7. The LDS3200 with integrated microscope images high-frequency (top) and low-frequency (bottom) defect signals.

With the optimization of the optical system to fit with the size of a sensor pad, the throughput could be increased significantly to allow a high sample rate. The parallel processing of two or more detection methods would provide the means to capture all defects of interest within one scan.

Conclusions

On the way to producing very fast and high-resolution CMOS image sensors, increased performance requirements are placed on today’s inspection and metrology systems. A process excursion must be detected early, and quickly resolved to maximize yield. This requires a reliable in-line process control using cost-effective inspection systems capable of high capture rates and throughput.

As a result of joint investigations with STMicroelectronics, a new optical non-destructive inspection method was developed for the detection of colorization effects on CMOS image sensors. The method can inspect 100% of the wafer surface, and has proven to be highly repeatable and reliable in extensive testing. The method provides substantial data for process control and process optimization while lowering overall production costs. It has helped to speed the learning cycles in development and ramp to production at STMicroelectronics. In addition, response time to excursions was shortened significantly.

With the success experienced at the Rousset site, the new method of inline monitoring to detect colorization problems on CMOS image sensors has subsequently been transferred to other STMicroelectronics manufacturing sites.