<?xml version="1.0" encoding="utf-8" standalone="no"?>
<!DOCTYPE document PUBLIC "-//CNX//DTD CNXML 0.5 plus MathML//EN" "http://cnx.rice.edu/cnxml/0.5/DTD/cnxml_mathml.dtd">
<document xmlns="http://cnx.rice.edu/cnxml" xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="new63">
  <name>Silicon Growth</name>
  <metadata>
  <md:version>2.15</md:version>
  <md:created>2000/08/04</md:created>
  <md:revised>2007/08/21 13:09:54.018 GMT-5</md:revised>
  <md:authorlist>
      <md:author id="wlw">
      <md:firstname>Bill</md:firstname>
      
      <md:surname>Wilson</md:surname>
      <md:email>wlw@madriver.net</md:email>
    </md:author>
  </md:authorlist>

  <md:maintainerlist>
    <md:maintainer id="wlw">
      <md:firstname>Bill</md:firstname>
      
      <md:surname>Wilson</md:surname>
      <md:email>wlw@madriver.net</md:email>
    </md:maintainer>
    <md:maintainer id="lizzardg">
      <md:firstname>Elizabeth</md:firstname>
      
      <md:surname>Gregory</md:surname>
      <md:email>elizabeth.gregory@gmail.com</md:email>
    </md:maintainer>
    <md:maintainer id="jsilv">
      <md:firstname>Jeffrey</md:firstname>
      <md:othername>M</md:othername>
      <md:surname>Silverman</md:surname>
      <md:email>JSilverman@astro.berkeley.edu</md:email>
    </md:maintainer>
    <md:maintainer id="gerardw">
      <md:firstname>Gerard</md:firstname>
      
      <md:surname>Wysocki</md:surname>
      <md:email>gerardw@rice.edu</md:email>
    </md:maintainer>
  </md:maintainerlist>
  
  <md:keywordlist>
    <md:keyword>semiconductors</md:keyword>
    <md:keyword>silicon</md:keyword>
    <md:keyword>wafers</md:keyword>
  </md:keywordlist>

  <md:abstract>How to create pure silicon crystals</md:abstract>
</metadata>

  <content>
    <para id="pa1">
      How is it possible for the IC industry to continue to make such
      gains, and how do they build so many circuits on one chip
      anyway? In order for us to be able to understand this, we have
      to take a look at the <term>monolithic fabrication
      process</term>.  <foreign>Lith</foreign> comes from the Greek
      word for stone, and <foreign>mono</foreign> means one, of
      course. Thus, monolithic construction refers to building the
      circuit in "one stone" or in one single crystal substrate.
    </para>


    <para id="pa2">
      In order for us to do this however, we first of all need the
      "stone", so let's see where that comes from.
    </para>


    <para id="pa3">
      We start out with a natural form of silicon which is
      very abundant (and relatively pure); quartzite or

      <m:math>
	<m:apply>
	  <m:times/>
	  <m:ci>Si</m:ci>
	  <m:ci>
	    <m:msub><m:mi>O</m:mi><m:mn>2</m:mn></m:msub>
	  </m:ci>
	</m:apply>
      </m:math>

      (sand). In fact, silicon is one of the most abundant elements on
      the earth. This is reacted in a furnace with carbon (from coke
      and/or coal) to make what is known as <term>metallurgical grade
      silicon</term> (MGS) which is about 98% pure, via the reaction

      <equation id="eqeqn01">
	<m:math mode="display">
	  <m:apply>
	    <m:tendsto/>
	    <m:apply>
	      <m:plus/>
	      <m:apply>
		<m:times/>
		<m:ci>Si</m:ci>
		<m:ci>
		  <m:msub><m:mi>O</m:mi><m:mn>2</m:mn></m:msub>
		</m:ci>
	      </m:apply>
	      <m:apply>
		<m:times/>
		<m:cn>2</m:cn>
		<m:ci>C</m:ci>
	      </m:apply>
	    </m:apply>
	    <m:apply>
	      <m:plus/>
	      <m:ci>Si</m:ci>
	      <m:apply>
		<m:times/>
		<m:cn>2</m:cn>
		<m:ci>C</m:ci>
		<m:ci>O</m:ci>
	      </m:apply>
	    </m:apply>
	  </m:apply>
	</m:math>
      </equation>
    </para>


    <para id="pa4">
      We have seen that on the order of

      <m:math>
	<m:apply>
	  <m:power/>
	  <m:cn>10</m:cn>
	  <m:cn>14</m:cn>
	</m:apply>
      </m:math>

      impurities will make major changes in the electrical behavior of
      a piece of silicon. Since there are about

      <m:math>
	<m:apply>
	  <m:times/>
	  <m:cn>5</m:cn>
	  <m:apply>
	    <m:power/>
	    <m:cn>10</m:cn>
	    <m:cn>22</m:cn>
	  </m:apply>
	</m:apply>
      </m:math>

      atoms/<m:math>
	<m:apply>
	  <m:power/>
	  <m:ci>cm</m:ci>
	  <m:cn>3</m:cn>
	</m:apply>
      </m:math>

      in a silicon crystal, this means we need a purity of better than
      1 part in

      <m:math>
	<m:apply>
	  <m:power/>
	  <m:cn>10</m:cn>
	  <m:cn>8</m:cn>
	</m:apply>
      </m:math>

      or 99.999999% pure material. Thus we have a long way to go from
      the purity of the MGS if we want to make electronic devices that
      we can use in silicon.
    </para>


    <para id="pa5">
      The silicon is crushed and reacted with

      <m:math>
	<m:apply>
	  <m:times/>
	  <m:ci>H</m:ci>
	  <m:ci>Cl</m:ci>
	</m:apply>
      </m:math>

      (gas) to make trichlorosilane, a high vapor pressure liquid that
      boils at 32°C as in:

      <equation id="eqeqn02">
	<m:math>
	  <m:apply>
	    <m:tendsto/>
	    <m:apply>
	      <m:plus/>
	      <m:ci>Si</m:ci>
	      <m:apply>
		<m:times/>
		<m:cn>3</m:cn>
		<m:ci>H</m:ci>
		<m:ci>Cl</m:ci>
		<m:ci>(gas)</m:ci>
	      </m:apply>
	    </m:apply>
	    <m:apply>
	      <m:plus/>
	      <m:apply>
		<m:times/>
		<m:ci>Si</m:ci>
		<m:ci>H</m:ci>
		<m:ci>
		  <m:msub><m:mi>Cl</m:mi><m:mn>3</m:mn></m:msub>
		</m:ci>
	      </m:apply>
	      <m:ci>
		<m:msub><m:mi>H</m:mi><m:mn>2</m:mn></m:msub>
	      </m:ci>
	    </m:apply>
	  </m:apply>
	</m:math>
      </equation>
    </para>


    <para id="pa6">
      Many of the impurities in the silicon (aluminum, iron,
      phosphorus, chromium, manganese, titanium, vanadium and carbon)
      also react with the

      <m:math>
	<m:apply>
	  <m:times/>
	  <m:ci>H</m:ci>
	  <m:ci>Cl</m:ci>
	</m:apply>
      </m:math>,

      forming various chlorides. One of the nice things about the
      halogens is that they will react with almost anything. Each of
      these chlorides have different boiling points, and so, by
      fractional distillation, it is possible to separate out the

      <m:math>
	<m:apply>
	  <m:times/>
	  <m:ci>Si</m:ci>
	  <m:ci>H</m:ci>
	  <m:ci>
	    <m:msub><m:mi>Cl</m:mi><m:mn>3</m:mn></m:msub>
	  </m:ci>
	</m:apply>
      </m:math>

      from most of the impurities. The (pure) trichlorosilane is then
      reacted with hydrogen gas (again at an elevated temperature) to
      form pure <term>electronic grade silicon</term> (EGS).

      <equation id="eqeqn03">
	<m:math>
	  <m:apply>
	    <m:tendsto/>
	    <m:apply>
	      <m:plus/>
	      <m:apply>
		<m:times/>
		<m:ci>Si</m:ci>
		<m:ci>H</m:ci>
		<m:ci>
		  <m:msub><m:mi>Cl</m:mi><m:mn>3</m:mn></m:msub>
		</m:ci>
	      </m:apply>
	      <m:ci>
		<m:msub><m:mi>H</m:mi><m:mn>2</m:mn></m:msub>
	      </m:ci>
	    </m:apply>
	    <m:apply>
	      <m:plus/>
	      <m:apply>
		<m:times/>
		<m:cn>2</m:cn>
		<m:ci>Si</m:ci>
	      </m:apply>
	      <m:apply>
		<m:times/>
		<m:cn>3</m:cn>
		<m:ci>H</m:ci>
		<m:ci>Cl</m:ci>
	      </m:apply>
	    </m:apply>
	  </m:apply>
	</m:math>
      </equation>
    </para>


    <para id="pa7">Although the EGS is relatively pure, it is in a polycrystalline
      form which is not suitable for device manufacture. The next step
      in the process is to grow single crystal silicon which is
      usually done via the <term>Czochralski</term>(pronounced
      "cha-krawl-ski") method to make what is sometimes called CZ
      silicon. The Czochralski process involves melting the EGS in a
      crucible, and then inserting a seed crystal on a rod called a
      puller which is then slowly removed from the melt. If the
      temperature gradient of the melt is adjusted so that the
      melting/freezing temperature is just at the seed-melt interface,
      a continuous single crystal rod of silicon, called a
      <term>boule</term>, will grow as the puller is withdrawn.
    </para>


    <para id="pa8"><cnxn target="fig05"/> is a diagram of how the Czochralski
      process works. The entire apparatus must be enclosed in an argon
      atmosphere to prevent oxygen from getting into the silicon. The
      rod and the crucible are rotated in opposite directions to
      minimize the effects of convection in the melt. The pull-rate,
      the rotation rate and the temperature gradient must all be
      carefully optimized for a particular wafer diameter and growth
      direction. The &lt;111&gt; direction (along a diagonal of the
      cubic lattice structure) is usually chosen for wafers to be used
      for bipolar devices, while the &lt;100&gt; direction (along
      one of the sides of the cube) is favored for MOS
      applications. Currently, wafers are typically 6" or 8" in
      diameter, although 12" diameter wafers (300 mm) are looming on
      the horizon.
    </para>


    <figure id="fig05" orient="horizontal"><media type="image/png" src="5.05.png"/>
	<caption>Czochralski crystal growth</caption>
</figure>


    <para id="pa9">
      Once the boule is grown, it is ground down to a standard
      diameter (so the wafers can be used in automatic processing
      machines) and sliced into wafers, much like a salami. The wafers
      are etched and polished, and move on to the process line. A
      point to note however, is that due to "kerf" losses (the width
      of the saw blade) as well as polishing losses, more than half of
      the carefully grown, very pure, single crystal silicon is thrown
      away before the circuit fabrication process even begins!
    </para>
  </content>
  
</document>
