Taking AIM at Overlay Control for Aluminum Metallization

Laura Peters, Senior Editor -- Semiconductor International, 6/1/2006

Advanced image metrology (AIM) targets have proven to offer better process stability than box-in-box (BiB) overlay targets for poly to shallow trench isolation (STI) active and other critical overlapping layers. However, with DRAM aluminum back-end processes now running at their technology limits, insufficient overlay control caused by measurement noise has become a cause of rework and yield loss.
 
Detlef Hofmann from Infineon (Dresden, Germany), along with colleagues at KLA-Tencor (Dresden and Migdal Haemek, Israel), showed that AIM targets can provide the requisite overlay control between contact 2 and metal 2 (M2) layers for 110 nm DRAM processes. In addition, benefits of improved matching between post-lithography and post-etch measurements were realized. Most importantly, AIM targets improved the overlay metrology tool capability and tool-to-tool matching. They presented their work at the SPIE Microlithography symposium in February.

In high-density DRAM devices, shrinking design rules require tight process control, with little margin for metrology noise. Lot disposition and stepper correctable data are generated by fitting overlay data from multiple sites to an analytical model. For aluminum processes, the modeled residuals, i.e., discrepancies between modeled and measured data, are typically larger than at any other process layer because of the spatial noise associated with the graininess of aluminum, in addition to process-induced rotation and magnification effects caused by the sputtering process (referred to as wafer-induced shift [WIS]). As the sputtering target wears, the magnitude of these variations changes.

Aside from such sputtering-specific parameters, the overlay target design also affects the magnitude of variation. Sputtering of the M2 layer is generally over a trench structure, which varies in width, creating different rotation and magnification effects that need to be compensated for to achieve stable process control.

To compensate for WIS effects, overlay is measured after litho and after etch. The overlay difference is used to compensate for the post-litho WIS effects. To maintain reliable WIS correction, post-litho and post-etch overlay differences must be stable.

Figure 1 shows the limitations with traditional BiB overlay metrology targets for M2 layers. Aluminum grains cause rough edges on the outer bars and, depending on target age, asymmetry results. In the AIM target (Fig. 2 ), multiple bars in each direction allow better averaging and provide more precise and less noisy results.

1. The outer bars of the traditional box-in-box overlay metrology target suffer from large grains, resulting in noisy data (left). The outer bars are built of trenches filled with aluminum. Its cross-section (right) shows asymmetry, which is a function of sputtering target lifetime.

Smaller trench size compared with BiB bar widths reduces the WIS effect caused by aluminum sputtering. The engineers optimized the AIM target size and duty cycle (line/space) to reduce target noise related to the aluminum grain size. The initial design still proved to be too noisy because of the fact that some aluminum grains grew larger than the outer bar trench width, partially connecting adjacent bars. The redesigned AIM target featured higher duty cycle and reduced overall target size.

2. The initial advanced image metrology grating target on metal 2 (left) features multiple bar design and better averaging, but the measurements were still too noisy. The improved design (right) has increased duty cycle (line/space) and smaller footprint.

The Infineon and KLA-Tencor engineers tested the AIM and BiB targets in controlled experiments and long-term pilot runs to verify performance in real manufacturing conditions. The AIM metrology demonstrated a factor of two better dynamic measurement repeatability compared with BiB metrology. The WIS effect (post-etch and post-litho data) of modeled rotation and wafer magnification from single-lot experiments was smaller with the AIM target than BiB because of multiple bar design and reduced trench size.

Long-term runs looked at process control across multiple sputtering tools, exposure tools and variations within sputtering tools. The switch from BiB to AIM targets resulted in tighter WIS distributions for modeled wafer rotation and magnification, resulting in more stable run-to-run control. Overlay residuals were also improved on the switch to AIM targets. Use of AIM targets on new products enabled faster identification of process challenges and shorter time to achieving stable process control.

Finally, the effect of overlay target change on lot disposition performance was monitored. The AIM target achieved lower maximum overlay error across the wafer after aluminum etch.

Find more information on yield management.