How a Rock Becomes a Smartphone CPU

1. Quartz

Your smartphone processor began its journey in the northwest corner of Spain, at Mina Serrabal, a quartz mine near the city of Santiago de Compostela. Quartz—or more technically, silicon dioxide or silica—is the main component of sand. But at Serrabal it can come in huge pieces twice the width of a smartphone. Mine operator Ferroglobe runs an automated system to sort the silica by size. After the pieces are washed and treated, the big ones head to the Atlantic coast for the next step in the journey.

2. Silicon Metal

After an hour by truck, the quartz mini-boulders arrive at Sabón, Ferroglobe’s 125,000-square-meter factory in the coastal province of A Coruña. Here the quartz will be mixed with dehydrated wood chips and heated to 1,500 to 2,000 °C in a trio of electric-arc furnaces that use massive electrodes invented at this plant in the 1990s. Inside the furnace, a reaction takes place that rips the oxygen from the silica and sticks it to the carbon from the wood. The result is silicon metal and carbon monoxide.

3. Purified Polysilicon

The resulting silicon metal is about 98 percent pure, and that’s not good enough. It will need to be at least 99.9999999 percent pure to become a microprocessor, which will require some pretty powerful chemistry. So it’s off to Wacker Chemie, in Burghausen, Germany. Here, the metal undergoes what’s called the Siemens process: It’s bathed in hydrochloric acid and reacts to form hydrogen gas and a liquid called trichlorosilane. Any impurities will be in the liquid, which is then run through a multistep distillation process that separates the pure trichlorosilane from anything unwanted. Once the needed purity is reached, the reaction is reversed: At 1,150 °C, the trichlorosilane is reacted with hydrogen to deposit multiple crystals of silicon, called polysilicon, and the resulting hydrochloric acid gas is sucked away. The polysilicon forms thick rods around heating elements. Once it’s cooled and removed from the reaction chamber, the polysilicon is smashed up for shipping.

4. Silicon Wafers

The ultrapure silicon is made up of many crystals at different orientations. But microprocessors must be made from a single crystal. So the material might migrate to Sherman, Texas, where GlobalWafers recently opened a US $3.5 billion silicon-wafer plant. Here the polysilicon is put through what’s called the Czochralski (Cz) method. In a high-purity quartz crucible, the polysilicon is heated to about 1,425 °C and melts. Then a seed crystal with a precise crystal orientation is dipped into the melt, slowly drawn upwards, and rotated. Do all that exactly right, and you will pull up an ingot of pure, crystalline silicon that’s 300 millimeters across and several meters tall. Specialized saws then slice this pillar of semiconducting purity into wafers less than 1 millimeter thick. The wafers are cleaned, polished, and sometimes further processed, before heading to wafer fabs.

5. Processed Wafers

Now it’s off to Tainan, in southern Taiwan, where TSMC’s Fab 18 will turn these wafers into the latest smartphone processors. It’s an exceedingly intricate process, involving some of the most complex and expensive equipment on the planet, including EUV lithography systems that can cost upward of $300 million each. In Fab 18, each wafer will go through months of exquisitely precise torture to produce the transistors and wiring that make up the processors. Extreme ultraviolet radiation will print patterns onto it, hot ions will ram into its surface, precision chemical reactions will build up some parts one atomic layer at a time, acids will etch away nanometer-scale structures, and metals will electrochemically plate parts and be polished away in others. The result: a wafer full of identical processors.

6. Packaged Chips

As amazing as these processors are, you can’t use them in this form. They first need to be packaged. For our silicon, that’s going to happen at ASE’s facility in Penang, Malaysia. A package provides the chip with mechanical protection, a way for heat to be removed, and a way of connecting the chip’s micrometer-scale parts to a circuit board’s millimeter-scale ones. To do this, the wafers are first diced into chips. Then tiny balls of solder, some only tens of micrometers across, are attached to the chips. The solder bumps are aligned to corresponding parts of the package, and the two parts are melted together. It’s becoming more common for multiple pieces of silicon to be integrated within the same package, either stacked on top of each other or positioned next to each other on a separate piece of silicon called an interposer. Other steps to the process follow, and the packaged part is now ready for its next step.

Global Trade