How are Microchips Made?

How are Microchips Made?

Show Video

Inside this smartphone are 62 microchips  containing a total of 90 billion transistors.   These microchips are incredibly powerful and  the cornerstone of all technology, but how are   billions of nanoscopic transistors manufactured  into a microchip the size of a tiny ant?  Well, all these microchips were manufactured in  a semiconductor fabrication plant like this one.   Inside it is a clean room which spans the area of  8 football fields and is filled with hundreds of   machines ranging in size from that of a van to  that of a city bus and costing anywhere between   a few million and 170 million dollars. Within  this microchip factory, silicon wafers travel   from machine to machine and undergo around a  thousand processes over a 3-month period. And,   by the end of production, each silicon wafer  will be covered in hundreds of CPU chips   each containing 26 billion transistors. When we zoom in we can see the nanoscopic   transistors at the bottom and over  a dozen layers of wires above. This  

integrated circuit is then cut out from  the wafer, tested, and packaged so that   it can be installed into your desktop computer. In this video we’re going to explore the entire   microchip manufacturing process and show  you how billions of nanoscopic transistors   and an impossibly complex 3D maze of wires  are manufactured in one of the world’s most   technologically advanced microchip factories.  It’s an incredibly complicated process,   so stick around and let’s jump right in! A portion  of this video is sponsored by 

There are two sides to understanding how microchip  manufacturing works. The first is the sequence of   steps and processes needed to build the nanoscopic  transistors and the labyrinth of wires. Whereas   the second is how the semiconductor fab and  multimillion dollar equipment on the cleanroom   floor work, and we’ll be flipping between  these two sides to get a complete picture. 

Let’s start by opening up this desktop computer,  focusing on the CPU and taking a look at what’s   inside. Here we have an integrated circuit, or  die, which we’ll refer to as a chip. This chip   has 24 cores, a memory controller, a graphics  processor, and many other sections. Within one   of the cores, we can see its block diagram  and the various elements. Zooming in on   this multiply block, we find a layout of 44  thousand transistors that physically execute   32-bit multiplication and constitute just  point zero zero zero one seven percent of the   overall 26 billion transistors in the CPU. Zooming in even further, we see layers of   metal wires, or interconnects, and at the very  bottom are the transistors that form the basic   logic gates. Note that these layers of metal  interconnects aren’t floating, but rather,   the empty space that you see is filled with  insulating materials, thus providing structure,   and preventing the metal wire layers from  touching. Furthermore, here we’re only showing  

the transistors at the bottom and five layers of  metal interconnects with vias traveling vertically   between the layers. In actuality there are a  total of 17 metal layers of wires in the CPU,   and each successive set of levels uses larger  and larger interconnects. At the bottom are   local interconnects that move data around this  32 bit multiply circuit. In the middle are   intermediate interconnects that move data around  the core, and at the top are global interconnects   that move data around the entire CPU. You might be wondering how small are   these transistors? Zooming in again and past  the interconnect layers we find FinFets,   which are transistors whose channel dimensions  are 36 by 6 by 52 nanometers with a transistor   to transistor pitch of 57 nanometers. Clearly  the transistors are incredibly small. Here’s  

a mitochondria, a dust particle, and  a human hair for size comparisons.  Now that you have a sense of what the transistors  and labyrinth of metal interconnects look like,   let’s explore how they’re manufactured. We’ll begin with an analogy. Imagine baking   a cake that’s 80 layers tall, with each layer cut  to a unique shape. To make this cake there are   940 steps in the recipe, which takes 3 months  to complete and includes hundreds of exotic   ingredients. And, if any measurement, baking  time, or temperature is more than one percent off,  

then the cake is entirely ruined. That’s  kind of what it’s like to make a microchip,   but microchips are even more complicated. Let’s look at a single layer of this integrated   circuit and run through a simplified  set of steps used to build it. To start,   a layer of insulating silicon dioxide is deposited  on top of the wafer and then a layer of light   sensitive photoresist is spread across the top.  Next, using UV Light and a stencil, a pattern is   applied to the photoresist. Solvents then are  used to remove the areas hit by the UV light,   thus creating a patterned mask layer. Using the  mask, the revealed silicon dioxide is etched away  

down to the previous layer. Next the mask layer is  removed, and a layer of copper is added to cover   the wafer and fill in the areas that were just  etched away. Finally, the surface is ground down   and leveled off to reveal the copper and insulator  patterns. And thus, a single layer is completed.  In order to build the next layer, which is  a vertical set of metal vias, we repeat the   same set of steps, but use a different pattern  for the photomask. Since these layers are all   built using the same set of steps it’s more  effective to visualize the steps as a circle   like a clock. To build all the 80 layers of the  die, this sequence is repeated over and over,  

resulting in 940 steps. One important note is  that the FinFet transistors at the bottom are   even more complicated than the metal wires, and  thus additional steps are needed to fabricate   them. Furthermore, cleaning the wafer to wash away  dust particles that may have landed on the wafer,   as well as inspecting the wafer to make  sure everything is being built properly,   happens frequently and these steps need to  be added to the circle. A different tool is  

used to complete each of these process steps. . Now that we have an understanding of the steps,   let’s take a look at this semiconductor  fabrication plant. This CPU is manufactured   on a 300-millimeter silicon wafer which can  fit 230 CPU chips. In contrast DRAM chips are   considerably smaller and thus 952 of them can fit  on a wafer. These silicon wafers are carried in   stacks of 25 using a container called a front  opening universal pod, or foup. This sealed   plastic wafer carrier is transported around the  cleanroom floor using an overhead transport system   which lowers the foup onto the tool’s landing pad.  Inside the tool, robotic arms transport the wafer  

through vacuum load locks and to different  process chambers where materials are added,   removed, or processed in ways that we’ll explore  later. The wafers are then returned to the foup,   resealed inside, lifted up to the overhead  transport system, and carried to and dropped   onto the next tool, where the next step in the  process is completed. To build the entire chip   composed of 80 different layers it takes 3 months  of traveling from tool to tool where at each   stop one of the 940 process steps is completed. In order to increase the microchip mass production   capabilities of a semiconductor fabrication  plant or fab, typically there are dozens of the   same semiconductor tools organized in rows that  perform the same process. On the cleanroom floor  

there are a total of 435 semiconductor tools  resulting in the fab’s production capacity of   50,000 wafers or 11.5 million CPUs a month. These tools have rather complicated names,   so we’ll start by categorizing them according  to their functionality. There are 6 groups:   making the mask layer, adding material,  removing material, modifying the material,   cleaning the wafer, and finally inspecting  the wafer. We’ve color coded the different  

functional groups to the various tools and  process steps to help you not get lost.  Let’s next look at each of these semiconductor  tools and see how they process the wafer in   various ways. We’ll start with the ones that are  used to make the mask layer or the nanoscopic   stencil on the wafer. These tools include the  photoresist spin coater, photolithography tool,   developer and photoresist stripper. First the  photoresist spin coater applies a light-sensitive   layer to the surface of the wafer and sends  it through a soft bake where the wafer is   heated in order to evaporate the solvent from  the photoresist. Next the wafer goes to the  

lithography tool which shines UV light through a  stencil, which is technically called a photomask.   The light passes through the stencil and is then  demagnified or shrunk down to produce a nanoscopic   pattern on the wafer. Wherever the light from  the stencil touches the wafer, the photoresist is   weakened. The wafer then goes to the developer  and the weakened photoresist is washed away,  

leaving only the patterned nanoscopic stencil on  the wafer. The wafer is then sent through a hard   bake to harden the remaining photoresist.  Next the wafer travels to other tools to   undergo processing, and once these processes  are completed the wafer goes to a photoresist   stripper which uses solvents to dissolve and  remove the photoresist mask layer. And that’s  

how a mask layer is formed and then removed. The photolithography tool is one of the most   important, so let’s take a look at it. Inside  is a UV light source, a set of lenses to   focus the light, a photomask which contains the  stencil, or design of the layer to be patterned,   and a wafer carrier. The photomask is 6 by 6  inches, and, based on the dimensions of the CPU,  

can fit 2 copies of a single layer of  the CPU design. The purpose of using   a photomask with these crazy optics is because  it’s a reliable way to copy and paste a design   for billions of nanoscopic transistors and wires  onto 230 identical CPUs on a single wafer in a   few minutes. After the light passes through the  photomask, the UV light goes to more lenses in   order to shrink down the pattern by a factor of  4 and print a single layer of the design onto the   photoresist. The wafer carrier steps from position  to position, printing the photomask image at each   stop, until all 230 chips are patterned. Let’s clarify one detail. In our previous   examples, we talked a lot about this CPU having 80  layers. Specifically, what we were referring to is  

the number of photomasks and mask layers used to  create all the different layers of patterns on   the wafer. Therefore, one complete CPU chip  uses 80 different photomasks, each costing   300,000 dollars. With only one mask layer being  patterned at a time, this CPU chip will undergo 80   separate visits to the lithography tool. We could  spend another hour talking about photolithography   but let’s move onto the next category of tools. Deposition tools are used to add or deposit  

material onto the wafer. A lot of times we  use the mask layer from the photolithography   step to add materials to the areas uncovered  by the mask layer, kind of like spray painting   through a stencil. Due to the wide range of  elements and compounds used to create the layers,   deposition tools have a wide range of variations  with complicated names and acronyms for each   variant. But essentially there are 3 key groups  of materials that are added or deposited onto   the wafer: metals such as copper or tantalum,  insulators which are typically called oxides,   and crystalline layers of silicon. Each group of  different materials uses different physics and  

chemistry principles to deposit the material  on the wafer and therefore has a different   technical name for the tool that deposits the  material. Deposition tools typically have a   central wafer handling chamber, with the various  chambers attached to the edges, each one dedicated   to adding just a single element or compound. The next category of machines do the opposite,   which is to remove material. There are 2  key methods. The first is etching. Etchers   use either corrosive chemicals or high energy  plasmas to react with and remove materials from   the surface of the wafer. They are typically used  with the mask layer stencil in order to remove the   material exposed by the mask, thus creating a hole  that can be later filled by a deposition tool.  The second method to remove material is CMP,  which is chemical mechanical planarization. CMP  

applies slurry and uses abrasive pads to grind  and polish away the top surface of the wafer,   making it perfectly flat. CMP levels off the top  layers of the wafer and is typically used as the   last step in a cycle of processes in order to  prepare the wafer for another layer to be added.  The fourth category are tools that modify the  silicon and are called ion implanters. These tools   use the photomask stencil to bombard the unmasked  regions with phosphor, boron, or other elements   in order to make the P and the N regions required  to form the transistors themselves. Therefore, ion   implanters are only used in the front end of line.  You might think that this is adding material.  

However, ion implanters only add around one atom  of phosphor or boron for every 10,000 atoms of   silicon. Additionally, while other machines spray  paint a layer on top of the wafer, ion implanters   hurl atoms deep into the silicon lattice, kind  of like a cannon launching a baseball 6 feet into   a concrete wall. This process typically damages  the silicon lattice, which is why the following   step is to repair the silicon by heating the  wafer using a separate tool called an annealer.  The fifth category of tools are used to clean  and remove any contaminants or particles from the   wafer. These wafer washers use ultra-pure water to  clean the wafer and then dry it with nitrogen or   hot isopropyl alcohol. Cleaning the wafer happens  rather frequently in order to remove any stray   particles that may have fallen onto the wafer. And finally, sixth are tools that inspect the  

transistors and metal layers for defects and  are called metrology tools. A common metrology   tool uses a scanning electron microscope with  nanometer-level resolution to take pictures of the   top surface of the wafer and determine if there  are defects such as improperly patterned layers   or particles on the surface. When fabricating  an integrated circuit that takes 3 months to   complete, it’s important to repeatedly monitor the  progress and make sure that each of the processes   is being executed with nanometer-level precision. Now that we’ve covered each of the categories,   here are the color coded process steps  along with the layout of the tools in the   semiconductor fabrication plant. Let’s run  through the complete set of steps used to   manufacture a single metal interconnect layer. First a layer of insulating silicon dioxide is  

deposited onto the wafer. Next photoresist is  spread across the surface and the wafer is sent   through a soft bake to remove the solvent. The  wafer then travels to the photolithography tool   where the design from the photomask is transferred  to each of the chips on the wafer by weakening   the areas of photoresist hit by the light. The  wafer next goes to the developer to wash away  

the sections that were hit by the light from the  lithography tool and then through a hard bake to   harden the remaining photoresist. With the mask  layer built, the wafer goes to an etching tool,   where a plasma etcher removes a vertical column  through the exposed silicon dioxide until it   reaches the previous layer’s metal vias. Next the  wafer is sent to a photoresist stripper where the   mask layer is removed. The wafer then travels  to a physical vapor deposition tool where a  

sequence of metals fills in the exposed pattern  and coats the wafer in metal. Finally the wafer   is sent to a chemical mechanical planarization  tool where the metal is ground down so that all   that remains is a flat layer of insulating silicon  dioxide and conductive copper interconnects that   match the pattern from the photomask. A single  metal layer is now completed, and the wafer is   ready for the next cycle to begin where insulating  silicon dioxide and the vias will be added. Note  

that cleaning and wafer metrology or inspection  steps occur in between many of these other   steps. Furthermore, the process steps to make the  transistors are less straightforward and utilize   the ion implanter, and thus we’ll cover them in a  separate video on transistor physics and design.  These steps are for building the  integrated circuit on the wafer, however,   there are additional steps in manufacturing a  microchip which we’ll explore in a little bit. 

But before we get there, one important thing  to note is that the semiconductor industry   is incredibly secretive regarding the exact tool  layout and the process steps and recipes used to   make the transistors. We wanted to make the best  video on how microchips are made and it took us   180 hours of scouring the internet and textbooks  for information and reference images and, using   what we found, we spent 205 hours modeling each  of these tools, the many layers of the integrated   circuit, and the semiconductor fab. Furthermore,  writing the script took about 100 hours, and then   animating all these visuals took more than 825  hours. As a result, this video took over 1300   hours to make, and it’s entirely free to watch.  We want to make more videos like this one where we   explore computer architecture and how transistors  work, and we can’t do it without your help. The   best way you can help is by taking a few seconds  to scroll down, write a comment below, like this   video, subscribe if you haven’t already and then  share this video on social media or send it to a   friend or colleague. Truly, just a few seconds  of your time helps far more than you think.  

Additionally, we have a Patreon page where  we’ll be releasing behind the scenes footage   of our work and updates for upcoming  videos. If you find what we do useful,   we would appreciate any support. Thank you. So then, what are the additional steps in   manufacturing a microchip? Before  chip manufacturing at the fab,   we first have to manufacture the silicon  wafers by refining quartzite into pure silicon,   and then growing a monocrystalline ingot  and cutting it into wafers. For reference,   these 300-millimeter wafers are around  three-quarters of a millimeter thick,   they have a barcode on the side and a small  notch in them to indicate the direction of the   crystal lattice. Furthermore, these wafers are  incredibly delicate, and shatter into hundreds   of shards when broken. A single wafer costs  around a hundred dollars, but after being  

populated with CPUs it’s worth closer to a hundred  thousand dollars, making it quite literally ten   times more valuable than its weight in gold. Moving onto the steps after chip manufacturing.   The completed wafer is sent to a separate building  where each of the CPUs undergoes rigorous testing   to figure out if it works as intended. If a CPU  works, that’s great. But frequently a particle   or photomask defect has damaged a section of  the integrated circuit, rendering that section   defective. These semi-functional circuits are then  categorized, or binned, based on what still works.   These Intel Thirteenth Gen processors are sold as  an i9, i7, i5, or i3, depending on how many cores   are functional with different product lines of  CPUs whose on-board integrated graphics sections   are defective. These wafers are transported  to another building where the chips are cut   out using a laser, flipped over, and placed on an  interposer which distributes the connection points   to a printed circuit board while a protective  heat conductive cover is placed on the back   side. The printed circuit board holds the landing  grid array that interfaces with the motherboard  

as well as various electrical components. Next  an integrated heat spreader is mounted on top,   and the entire assembly is tested one last time  before being packaged for sale. Finally, the CPU   is now ready to be mounted onto the motherboard  and installed into your desktop computer.  It’s important to understand that chip  manufacturing requires an incredible   amount of science and engineering and there’s a  free and easy way to learn the basic principles   inside each of these complex tools and that’s with  this video’s sponsor,! Brilliant   reimagines how courses are taught. Instead  of boring hour-long lectures or textbooks  

that put you to sleep, Brilliant uses fun and  interactive modules inside thousands of lessons   from basics to advanced topics – and new lessons  are added every month. Whatever your skill level,   Brilliant customizes its content to fit your  needs and allows you to learn at your own pace.  We use Brilliant daily. We’re working on  videos on how AI and Chat GPT Works, and   so each of our animators is progressing through  their lessons on How Large Language Models work. 

Because you’re watching this video, you probably  enjoy learning about how technology works,   and fortunately for you, Brilliant just added  a course on this very topic. In it they have   lessons such as How GPS Works, How Computer  Memory Works, and how Recommendation Algorithms   such as those used by YouTube work. If  you’re looking to advance your career,   Brilliant is the go-to resource for leveling up  your skills and staying up-to-date on the latest   concepts behind world-changing technology. For the viewers of this channel, Brilliant  

is offering a free 30-day trial with access to  all their thousands of lessons. Additionally,   Brilliant is offering 20% off an  annual subscription. Just go to The  link is in the description below.  Microchip Fabrication is a massive topic, and  thus, we have two more equally complex videos   that we’re working on. The first will be an  in-depth 3D animated factory tour and the   second will explore transistor physics, FinFets,  and the next generation of transistors. We’re  

also working on a series of videos on GPUs and a  separate one on CPU architecture so make sure to   subscribe so you don’t miss any of our videos. We’re thankful to all our Patreon and YouTube   Membership Sponsors for supporting our work.  If you want to financially support our work,   you can find the links in the description below. This is Branch Education, and we create 3D   animations that dive deeply into the technology  that drives our modern world. Watch another Branch  

video by clicking one of these cards or click  here to subscribe. Thanks for watching to the end!

2024-05-24 23:38

Show Video

Other news