Haze Defects Limit Life of Photomasks (Part 1)

Haze impacts yield in two ways: Typically, the defect grows large enough that it becomes a point defect and prints on the wafer. More rarely, haze can affect the transmission of light through the mask, causing a CD change. Eventually, the contamination reaches an unacceptable level and a “repell” procedure is needed, where the pellicle is removed, the mask is cleaned, a new pellicle is installed, and the mask is returned to the production line. However, not only is this procedure costly, but also a cleaned mask is never as clean as one straight from the shop, and its lifetime is further compromised with each clean.

Wafer fabs experienced with the impact of haze might track the number of wafers exposed with a given mask, then inspect it for haze. “They may even set up a repell procedure — pull, clean, repell — whether or not they see haze. The more sophisticated fabs will track the cumulative dose on the mask, then pull and inspect the reticle after that dose or a specific period of time,” explained Franklin Kalk, CTO of Toppan Photomasks (Round Rock, Texas).

Three usual suspects

There are three types of haze defects that show up most frequently. Ammonium sulfate was the earliest, and is still the most common. This inorganic crystalline type of defect can grow up to several microns in size, and is caused by residual contaminants from resist strip and clean, storage conditions and/or the environment. The second is the family of oxalic acids, including carboxylic acid and ammonium oxalate. “These defects tend to be limited to a few fabs, and the cause is unknown, but the atmosphere is one source of carboxylic acid,” said Kalk. The third type is organic defects, which are tiny defects caused by volatile organic carbon (VOC) in the pellicle, packaging, storage or exposure equipment.

Life of a mask

Reticle life varies greatly (Fig. 1 ) and can be anywhere between 1000-50,000 wafer exposures and counting, according to Kalk. The lower numbers correspond to reticles that have been repelled. “When you pull it, there's always some residual adhesive on the mask, and it's incredibly difficult to remove without using sulfur-based processes,” he said. The prior art was to use piranha, but since it has sulfur, that had to be removed. This is done by adding ammonium hydroxide to form ammonium sulfate, which gets rinsed off. But even then, sulfur dioxide remains, which, if not chemically bound to the reticle, can be physically bound to the chromium or molybdenum silicide. Better rinses using hydrogenated water, ozonated or hot water help reduce these levels. “The second thrust has been to go to zero sulfate processing. The problem there is it's not as good at removing things like resist, but you can do it without sulfur.”

1. The life of the reticle is fairly unpredictable, but it is definitely much more drastically impacted by haze at the 193 nm wavelength, as evidenced by this data on gate layer embedded phase-shift masks (EPSMs). (Source: KLA-Tencor)

A recent study between KLA-Tencor and Toshiba (Tokyo) compared the cost of direct reticle inspection vs. image qualification (a print check on the wafer), the two primary ways that reticles can be checked for haze. The study determined that since wafer inspection does not provide enough early warning, inspection frequency needs to be higher (mask defects are 4× the size of wafer defects). A 15 wafer start/min fab with five ArF litho clusters will require ~25 image qualifications per day and has a significant opportunity cost (Fig. 2 ) because it occupies scanner time. That fab saves $4M/yr by implementing direct mask inspection.

2. The toll on fab productivity from tying up litho tools to perform wafer inspections to check for haze. (Source: KLA-Tencor)

Find more information on yield management .