Overlap between fabs and sdl

<p><a class="wikilink" href="/fabs/">Fabs</a> (highly integrated places where semicon chips are manufactured: <a class="wikilink" href="/semiconductor_industry/">semiconductor industry</a>) are in essence a well oiled <a class="wikilink" href="/sdl_(self-driving_labs)/">SDL (self-driving labs)</a>. </p> <p>In a semiconductor Fab, there's a tight integration of fabrication and analytical instruments that span an extremely broad range of techniques. From <a class="wikilink" href="/gas_sensing/">gas sensing</a> monitoring the inlets and outlets of the machines all the way to metrology of surfaces, automated system checks, and of course manufacturing steps. </p> <p><a class="wikilink" href="/making_a_wafer/">Making a wafer</a> requires cycling through different production steps, each generating a wealth of data that is analyzed close to real-time and stored for further inspection. </p> <p>Metrology systems (such as those of <a class="wikilink" href="/kla_(keep_looking_ahead)/">KLA (Keep Looking Ahead)</a> and <a class="wikilink" href="/nearfield_instruments/">Nearfield Instruments</a>) are obvious data sources and have a good overlap with what one would expect in an <a class="wikilink" href="/sdl_(self-driving_labs)/">SDL (self-driving labs)</a>: <a class="wikilink" href="/optical_microscopy/">Optical Microscopy</a> and <a class="wikilink" href="/atomic_force_microscopy/">Atomic Force Microscopy</a>. </p> <p>But they are not the only ones. Each system generates data about its own operation. There are systems used to monitor the quality of the input chemicals, or to monitor the output of a cleaning step (are there unexpected substances found). </p> <p>This entire process is running 24/7 with massive infrastructure for data handling, retrospective analysis, and potentially real-time decision-making. </p> <p>One interesting caveat is that <a class="wikilink" href="/semicon_has_defined_a_strict_interface/">semicon has defined a strict interface</a> on wafers, which massively speeds up the opportunities compared to other fields. </p> <p>This can perhaps be manifested in biology-oriented problems, where having DNA as the "interface" (and constraint) already allows for quite a high degree of interoperability. </p> <p>In a broader context of <a class="wikilink" href="/material_discovery/">material discovery</a>, such interfaces are not developed at all. From polymers to catalysts, samples can be liquid, solid, or gas. They can be deposited on substrates with ill-defined properties (stiffness, mounting points, or heat resistance). </p> <p>That may be the next frontier.</p>