SK hynix’s BSI technology: leading the way in the global mobile market

//php echo do_shortcode(‘[responsivevoice_button voice=”US English Male” buttontext=”Listen to Post”]’) ?>

A CIS (CMOS image sensor) is a sensor that converts the color and brightness of light received by a lens into an electrical signal and passes it on to a processing unit. Those image sensors therefore function as the eyes of mobile devices such as smartphones and tablets. Recently, CIS technology has gained attention as a key technology in the Fourth Industrial Revolution alongside virtual reality (VR), augmented reality (AR) and autonomous vehicles. It is expected that the technology will not stop at simply becoming the eyes of a device, but will even make further developments in its capabilities.

It has been 15 years since SK hynix launched a task force to develop CIS products. In addition to its core semiconductor memory business represented by DRAM and NAND flash, SK hynix has also developed and produced the non-memory semiconductor CIS to increase its competitiveness. SK hynix has developed numerous device and process technologies to close the technology gap with competitors, and the company has now reached the point of producing ultra-high resolution CIS products with 50 million pixels or more with a pixel size of just 0.64 μm (micrometer). This article introduces BSI (Backside Illumination) technology, a key element of CIS, based on the contents of the 10th SK hynix Academic Conference, held in November.

FSI technology and its limits

The pixels of early CIS products have a Frontside Illumination (FSI) structure that superimposes an optical structure on top of a CMOS1 process based circuit. This technology applies to most CIS solutions with a pixel size of 1.12 μm or larger and is used in a variety of products including mobile devices, CCTVs, dash cams, DSLR cameras and vehicle sensors.

1) Complementary Metal Oxide Silicon (CMOS): A complementary logic circuit consisting of pairs of n-channel and p-channel MOSFETs. CMOS devices consume minimal power and are used in DRAM products and CPUs because they are capable of extremely large-scale integration regardless of their complex processors.
Figure 1. FSI structure and unit pixel diagram

A powerful image sensor should be able to display clear images even in low light, and this requires an increase in quantum efficiency (QE)2 of the pixels. Therefore, the design of the metal wiring in the pixel’s lower circuit should be based on the FSI structure to minimize light interference.

2) Quantum Efficiency (QE): The measure of an imaging device’s effectiveness in converting incident photons into electrons. A sensor with 100% QE exposed to 100 photons would produce 100 signal electrons.
Figure 2. QE equation and FSI structure diagram

In general, however, diffraction3 of light occurs when continuous light waves pass through an opening or around objects. In the case of an aperture, as the size of the hole gets smaller, more light is scattered as the diffraction increases.

3) Diffraction: The propagation of waves such as sound and light as they pass through an obstacle or an opening. In the case of light, diffraction occurs when the size of the obstacle or aperture is equal to or smaller than the wavelength of the passing wave.

Likewise, diffraction is unavoidable even when external light reaches a single pixel. In the case of the FSI structure, it is more vulnerable to diffraction because it is affected by the metal wiring layer in the lower circuit. Even when FSI pixel sizes are reduced, the area covered by the metal remains the same. Consequently, the area through which the light passes becomes smaller and the diffraction becomes more intense, resulting in mixing of colors in the image.

Figure 3. Diffraction of light (top) and intensification of diffraction by pixel shrinkage (bottom)

Still, it is not impossible to control the diffraction of pixels. To improve the diffraction in a single area, the distance between the microlens and the silicon (Si) can be reduced according to the diffraction calculation formula. To this end, a BSI process was proposed that eliminated the metal interference by flipping the wafer to use the back side. At SK hynix, the introduction of BSI technology started with products with a pixel size of 1.12 μm or smaller.

Birth of BSI based pixel technology

In 2011, Apple introduced the iPhone 4, which featured the first CIS applied with BSI. This BSI-based CIS would capture relatively higher amounts of light compared to existing FSI technology and thereby reproduce higher quality images.

The BSI process used by Apple and now throughout the industry is shown in the flowchart below. In the case of BSI technology, all parts of the circuit are first produced on one side of the wafer before it is then flipped over and flipped upside down to create an optical structure that collects light on the back. As a result it is possible to remove interference caused by metal wiring in FSI and widen the area for light to pass through to provide higher QE.

Figure 4. BSI Process Flow Chart
Figure 5. Comparison of the distance between the microlens and photodiode (PD) according to structure

Through such BSI technology it is it became possible to use a pixel size of 1.12 μm or smaller and a market for high-resolution products of 16 million pixels or more emerged. Unlike the FSI structure, which suffered from interference caused by the wiring, the optical process was able to have a higher degree of freedom. As a result, various optical pixel structures have been developed, such as BDTI, W Grid and Air Grid, and these structures are used to increase the QE of products.

  • BDTI (Backside Deep Trench Isolation) process.

While it is possible to have high QE with only a BSI structure that has overcome light refraction, an additional pixel-division structure was needed to support the ever-shrinking pixel size and declining F-number4 of smartphone cameras. A good example of an additional grading structure is the BDTI that promotes total internal reflection (TIR).5 in areas where light enters diagonally from the outside of a CIS chip, amplifying the signal. Currently, this technology is applied to most BSI based CIS products.

4) F-number: A value that determines the brightness of the aperture. The lower the camera’s F-number, the wider the aperture opens to capture more light, allowing a camera to take brighter pictures in dark places while reducing image noise.
5) Total Internal Reflection (TIR): A total reflection of light in a medium, including water or glass, from surrounding surfaces back to the medium. TIR occurs when the angle of attack is greater than the critical angle.
Figure 6. Conventional BSI structure and the BDTI process as additional pixel division structure
  • Isolation structure for color filters

Parallel to the BDTI structure, another technique to improve the performance of BSI-based pixels is to place physical barriers between color filters. Because the distance between the microlens and the photodiode6 could not be reduced after BSI, this structure prevents diffraction caused by pixel shrinkage. Representative structures of the color filter isolation method include a W-shaped grid formation, or SK hynix’s patented Air Grid structure. Unlike the W Grid, which is a simple light-blocking structure, the Air Grid, which uses TIR, is expected to be a next-generation technology as it can actually increase QE.

6) Photodiode (PD): converts the light received by the CIS sensor into an electrical signal.
Figure 7. W grid structure and Air Grid structure

Bright future for SK hynix’s BSI-based pixel technology

Since BSI-based CIS products first appeared in the iPhone 4 in 2011, the gap between the best-performing CIS companies and the rest has widened, leading many CIS sensor companies to withdraw from the mobile market. However, SK hynix quickly secured the BSI technology through its own capabilities and developed core element technologies, including BDTI and Air Grid, by applying them to products with a pixel size of 1.12 μm or smaller.

SK hynix’s BSI technology is constantly evolving. Recently, SK hynix succeeded in developing hybrid bonding technology that applies ‘Cu-to-Cu bonding’ to stacked sensors based on TSV (via silicon via), laying the foundation for greater competitiveness in chip size and expansion of multi-layer wafer bonding technology . These technological achievements are expected to contribute to the expansion of the market by being used in the development of various sensors that support AI, medical devices, AR and VR in the future.

Learn more


Leave a Reply

Your email address will not be published. Required fields are marked *