Enabling Large Format Deep Pixel Detectors Through 3D Quilt Packaging

This proposal would develop the next generation of interconnects and assembly technology by making use of Quilt Packing® (QP) interconnects for the integration of semiconductor chips to extend systems into the third dimension. Quilt Packaging is a unique and versatile edge- interconnect technology that utilizes “nodule” structures extending out from the vertical facets along the edges of QP-processed chips. This technology enables multiple die to be assembled together into a single quasi-monolithic “metachip,” giving rise to a variety of innovative, novel, and versatile configurations. Most microsystem interconnections are formed using wire bonds or bump bonds, both of which have enabled cutting-edge science such as that associated with LHC, Fermi, LCLS, CDMS, etc. Wire bonding is an intrinsically one-dimensional process, while bump bonding is two- dimensional, but is limited to connecting a pair of substrates. Several national laboratories have been jointly pursuing a partially 3D interconnect technology. However, this 3D technology is effectively limited to connecting 2-4 substrates. While increasing the amount of circuitry per pixel by a corresponding factor of 2-4 is potentially significant, it clearly represents a marginal gain given the increased cost. In contrast, the QP technology is less expensive, simpler, and can increase the circuitry per pixel by a factor between 10 and 100. It can also provide direct, high-bandwidth, low-impedance connections between sensors or circuit chips in spatially separated planes. One example application involves bridging between multiple layers in a tracker to identify high- momentum tracks and use this to generate a prompt event trigger. An IC connected to two or more layers of a tracker or vertexer system could collect and correlate hits from different layers and tag ones that are compatible with stiff tracks. A second example where QP technology might prove useful is in a high-granularity tracking detector (HGTD). In this application, each IC-sensor pair time stamps every hit with a temporal resolution approaching 20 ps. This allows tracks originating from different primary interaction vertices to be separated. If most tracks are associated with different primary interactions, then this would be indicative of a “stochastic” jet and not a physics jet. In practice, it is likely that track-timing histograms would need to be assembled from many IC-sensor modules, as a jet would be distributed over a relatively wide angle. QP interconnects would enable HGTD modules to be connected on all four edges to form an interwoven fabric where information could be exchanged quickly.