The $200M Machine that Prints Microchips: The EUV Photolithography System

0:00 Inside every modern laptop, smartphone, desktop computer, advanced AI server,

0:06 or practically any other high-tech device are

0:11 cutting edge microchips such as these CPU, GPU, SoC, DRAM, and SSD chips,

0:19 each with tens of billions of transistors inside of them.

0:23 The transistors inside these microchips are incredibly small with the tiniest

0:29 features measuring around 10 nanometers or 45 silicon atoms.

0:33 This feat of science and engineering may seem impossible because on one hand

0:39 each of these microchips is made

0:42 from connecting billions upon billions of transistors together,

0:46 and then on the other hand,

0:48 each individual transistor is only nanometers in size.

0:52 Additionally, these microchips are everywhere and in everything,

0:56 and therefore they must be reliably mass produced.

1:01 So how is manufacturing such a microchip even possible?

1:05 These are photolithography tools,

1:07 and they are the key to manufacturing microchips.

1:11 However, it’s important to note that there are dozens of different

1:15 types of tools used in the various steps for making microchips,

1:18 and each one plays a critical role in the manufacturing process.

1:23 So, to be accurate, photolithography tools are the ones that are

1:27 used to copy and imprint the nanoscopic patterns of transistors

1:31 and layers of wires onto a microchip and therefore a useful

1:36 analogy is to think of these photolithography

1:40 tools as nanoscale microchip photocopiers.

1:44 Photolithography tools have been continuously evolving to copy and imprint

1:48 smaller and smaller transistors and circuitry, and, in this video,

1:53 we’re going dive into this state-of-the-art EUV Photolithography System

1:58 and explore the science and engineering inside of it.

2:02 So, let’s begin with a quick overview.

2:05 To start, the EUV Lithography Machine takes the design of a single layer

2:10 of a microchip on what’s called a photomask and loads it into the machine.

2:15 Next, a 300 millimeter silicon wafer with a set of prior processes

2:20 applied to it is placed onto a wafer carrier inside the machine.

2:25 With both in place, the machine uses extreme ultraviolet light or EUV and a set

2:32 of mirrors to copy the design from the photomask onto a silicon wafer.

2:37 The wafer moves to the next position

2:39 and the microchip design is copied yet again.

2:43 This copying happens over and over until the wafer is filled with a hundred

2:47 or more microchips and then a new wafer comes in, and the copying starts over.

2:53 This is the real-time speed of the lithography machine,

2:57 taking about 18 seconds to duplicate the same microchip design around

3:02 a hundred times across the entire area of a 300-millimeter wafer.

3:07 Let’s take a look at one of these microchips and see what exactly we’re copying.

3:13 Inside this microchip are approximately 30 billion transistors,

3:17 and if you were wondering, it’s the design of a GPU or graphics processing

3:23 unit found in the center of a graphics card.

3:26 When we zoom into a nanoscopic view of this microchip

3:30 we find a 3D maze of transistors and layers upon

3:33 layers of wires with the smallest dimensions of the bottom

3:37 most layers measuring around 10 nanometers or around 45 silicon atoms.

3:44 Specifically, the EUV Photolithography System typically patterns

3:48 the lower layers with the smallest features,

3:52 whereas other photolithography tools are used to pattern the higher layers.

3:56 It might be difficult to fully grasp the level of detail

4:01 and complexity inside a single layer of billions of nanoscopic transistors,

4:05 so let’s use a thought experiment and pretend

4:08 that instead of copying transistors and wires,

4:12 this EUV photolithography system is used to copy the text from a book.

4:18 If the width of each line of a letter is 13 nanometers,

4:22 then the word ‘Cat’ would take up around 155 by 240 nanometers,

4:26 a page of text would be about the size of a red blood cell,

4:32 and a chapter of a book would be a grain of pollen.

4:36 When we zoom out to see the equivalent area of a GPU chip,

4:40 how many pages of text do you think we could fit using these nanoscopic letters?

4:46 Well, we could print all 7 Harry Potter

4:50 books plus every book written by Stephen King,

4:53 the entirety of the text from the English Wikipedia,

4:56 and still have enough space to fit

4:59 every single book from your local public library.

5:03 There’s an unbelievable quantity of nanoscopic lines and details

5:07 that can fit into the area of a microchip,

5:11 and it’s all photocopied by this EUV Lithography System in less than a second.

5:16 It’s no exaggeration to say that every piece of modern

5:21 technology that you use is made possible by this machine,

5:24 and in this video, we’re going to explore

5:29 the key modules inside it and see how they work.

5:37 So, let’s jump right in.

5:41 This video is sponsored by ASML,

5:44 the company that designs and manufactures EUV Lithography Systems.

5:48 Throughout the video all the details and facts were independently researched,

5:54 written, and animated.

5:56 Additionally some aspects are simplified,

5:58 and due to the proprietary knowledge and confidentiality around EUV Lithography,

6:03 some of the details we present are approximated or modified.

6:08 Before we open up and explore this EUV System, let’s first spend a few minutes

6:15 discussing microchip manufacturing and semiconductor fabrication

6:18 plants or fabs for short and the exact role of this machine.

6:24 Inside our example fab are hundreds of machines of which a couple

6:28 dozen or so are the EUV Lithography machines we’ve been discussing.

6:33 To make a microchip, 300-millimeter silicon wafers are stacked inside

6:38 a front-opening universal pod or foup

6:41 and carried from machine to machine using an overhead transport system.

6:45 The foup is lowered onto a machine where each wafer is processed

6:50 in one way or another and, once the machine completes its work,

6:54 the wafers are returned to the foup, the pod is picked up,

6:59 carried to the next machine and dropped off for the next step in the process.

7:04 Microchip manufacturing is incredibly complicated,

7:06 but a simple way to think about it is that it’s

7:11 kind of like spray painting a design through a stencil,

7:14 but instead of art this stencil contains the nanoscopic

7:18 patterns used to build the transistors and wires.

7:21 Inside the microchip factory,

7:23 some tools are used to build the stencil such as the EUV lithography system,

7:28 and many of the other machines such as the deposition

7:32 tools or ion implanters are the spray paint.

7:36 So, let’s take a look at how we build the stencil

7:39 on the wafer which is technically called a photoresist layer.

7:42 To begin, the wafer travels to a machine called a track

7:47 tool where a light sensitive material called photoresist or just resist,

7:52 is poured on and evenly spread across a spinning wafer.

7:57 Next the wafer is heated in order to dry

8:00 and solidify the resist thus forming a flat blank stencil.

8:05 The wafer then moves to the photolithography tool where

8:09 EUV or extreme ultraviolet light is projected onto the photomask,

8:13 which is also called a reticle but typically just a mask for short.

8:18 When EUV light hits the mask,

8:21 the patterned information is imprinted in the light,

8:23 and this imprinted light then bounces off a set of mirrored lenses in order

8:28 to project a focused and scaled down image of the mask onto the wafer.

8:33 Wherever the EUV light touches, the resist is modified and thus the design

8:38 is copied from the mask onto the wafer.

8:41 The wafer moves to the next position

8:43 and the EUV patterning process repeats again

8:46 until the entire wafer is filled with copies of the design from the mask.

8:52 Next the wafer travels back to the track tool,

8:55 where the modified resist is washed away using

8:58 a developing solvent and the wafer is heated

9:01 to form a hardened stencil or completed photoresist

9:05 mask layer on the top of the wafer.

9:08 Now that the wafer is patterned,

9:10 the wafer travels to the other spray paint-like tools

9:13 in the fab which are used to etch away the uncovered areas,

9:17 implant dopants such as boron or phosphor, or deposit a layer of copper,

9:23 tungsten, or other metals,

9:26 thereby building a single layer of nanoscopic structures.

9:30 Note that there are additional process steps that we’re not going to get into.

9:35 Now that we have a basic understanding of how

9:38 the stencil and spray paint like processes form a single layer,

9:42 let’s zoom into a nanoscopic view inside a microchip where we

9:46 can see how the transistors

9:48 and wires are incredibly complicated 3-dimensional structures.

9:52 Each of these layers are built one after

9:55 the other starting with the transistors at the bottom,

9:59 moving up to the small wires, and then wider and wider metal layers further up.

10:04 In essence, to build a complete microchip,

10:07 the stencil and spray paint process is repeated over

10:11 and over each time building only a single layer,

10:15 and therefore it’s more effective to visualize these processes as a loop,

10:20 where a single pass of the loop forms one

10:23 layer using a single mask design in the lithography tool,

10:27 and then another layer is built using

10:29 an entirely different mask loaded in the machine.

10:32 To complete a GPU chip like this one,

10:36 the series of process steps or loops is repeated around 80 times resulting

10:41 in around a thousand individual process steps

10:45 and taking four or so months to complete.

10:48 Let’s go back to the nanoscopic view of the microchip.

10:52 Here we can see that the lower layers are incredibly tiny and 13

10:56 nanometer EUV light is used to build the pattern for these layers.

11:02 However, the upper wires are substantially larger and are

11:06 patterned using an entirely different machine called a DUV

11:11 or Deep Ultra Violet Photolithography System which is also

11:15 built by ASML and uses deep ultraviolet wavelength light.

11:21 DUV lithography tools were introduced in the 2000s

11:25 and are still incredibly advanced machines.

11:28 Because DUV tools typically cost less than EUV machines,

11:33 it’s more cost-effective to use EUV tools

11:36 to pattern the transistors and so-called critical layers,

11:41 and then use DUV tools to pattern the upper, wider, less-critical layers.

11:47 Additionally, less advanced chips that don’t

11:50 require the smallest transistors and wires may

11:54 forgo using EUV lithography altogether and only

11:59 use DUV wavelengths, such as 193, 248 or 365 nanometers As a result,

12:08 cutting edge fabs typically utilize different types of lithography tools,

12:15 and all these machines work as an intricate ecosystem to make a microchip.

12:21 With the basics of microchip manufacturing covered,

12:24 let’s open an EUV lithography system,

12:27 explore the incredible science and engineering inside,

12:30 and divide the system into its five key parts:

12:34 The light source, the illuminator, the reticle handler and reticle stage,

12:40 the projection optics, and finally the wafer handler and wafer stages.

12:44 We’ll begin with the light source which produces the EUV light,

12:49 but let’s first answer the question of why

12:52 we even need to use Extreme Ultraviolet Light.

12:55 Well, a simple analogy is to think of the light used to project

12:59 and copy the pattern from the photomask as the tip of a marker.

13:04 And, as you’re probably familiar with markers,

13:06 if you want to draw thin lines then you need a marker that also has a thin tip.

13:12 You can do tricks like angling the marker,

13:14 but if you want to draw lines that are 100 times thinner,

13:18 well then you need to switch pens and use a much smaller, fine tipped pen.

13:25 Likewise, to copy designs with dimensions only around 10 nanometers wide,

13:30 we use 13 nanometer light,

13:33 which is in the extreme ultraviolet light range of the electromagnetic spectrum.

13:38 The more technical answer deals with the wavelike nature of light

13:42 and what happens when light hits these nanoscopic patterns inside the photomask.

13:47 These patterns are made from nanoscopic EUV absorbing blockers

13:51 on top of a surface that reflects EUV light.

13:55 We’ll explore the photomask and how the system

13:58 uses reflective lenses a little later in this video,

14:01 but for now instead of using reflective optics,

14:04 it’s easier to visualize the photomask as through beam optics

14:08 in a setup similar to the well-known double slit experiment.

14:12 However, instead of the double slit

14:15 we’re showing light passing through a complicated

14:18 pattern of nanoscopic slits that represents

14:21 a small portion of the overall photomask.

14:23 So, what happens when we use a wavelength of light

14:28 that’s substantially larger than the 13 nanometer EUV light,

14:32 such as this 450-nanometer blue light.

14:35 Well, when this large wavelength light hits the pattern,

14:40 the pattern is almost entirely lost.

14:42 This is due to the width of the holes in the pattern

14:46 being substantially smaller than the wavelength of light hitting it.

14:49 This limit to the resolving power of the lithography

14:53 machine is described in Rayleigh’s Criterion Equation,

14:57 which we will explain later.

15:00 So, let’s switch to 13 nanometer EUV light.

15:03 As the EUV light hits the pattern, the light passes through and the pattern

15:09 of the photomask is imprinted into the light,

15:12 and the light diffracts similar to the double slit experiment.

15:16 This imprinted light then passes into projection

15:19 optics mirrored lenses which are used

15:21 to focus and scale down the pattern and project it onto the wafer.

15:26 So now that we understand the need for EUV light,

15:29 the next question is how do we produce it?

15:32 To start, two high-powered laser pulses run

15:36 through multiple amplifiers below the cleanroom floor.

15:39 These laser pulses grow in power and then

15:43 travel from the sub-fab using a pathway of mirrors

15:47 and up through the bottom of the tool

15:49 and into a chamber called the source vessel.

15:52 The first laser, called the pre-pulse,

15:54 is around 5 kilowatts in power and is targeted

15:57 at a droplet of tin using a set of actuated mirrors.

16:02 This pre-pulse laser turns the tin droplets into a pancake-like shape.

16:07 The second approximately 25-kilowatt main laser pulse which is more

16:12 than 10 times stronger than the lasers used to cut steel, hits the tin pancake,

16:19 instantly vaporizing it and turning it into glowing plasma.

16:22 Within each of the tin atoms, some electrons are ejected,

16:26 and others are kicked up to higher energy states.

16:30 When electrons drop back down from the 4F to 4D orbitals,

16:35 13-nanometer EUV light is produced.

16:38 Shooting two laser pulses at a droplet

16:41 of tin might seem like a rather obscure process,

16:45 however EUV light doesn’t naturally occur on Earth and it’s one

16:49 of the few ways to efficiently produce over 500 Watts of EUV Light.

16:55 Additionally, the reason for using tin is that its plasma produces

17:01 a wide range of wavelengths with a clear peak at 13 nanometers.

17:06 So, where does the tin droplet come from?

17:09 Well over here a solid ingot of ultra-pure tin is melted

17:14 and fed into a storage tank and then piped towards a microscopic nozzle.

17:19 A piezo electric transducer squeezes the tip of the nozzle,

17:23 and due to high-pressure nitrogen inside the storage tank,

17:28 a droplet of tin is forced out at a speed of 100 meters a second.

17:33 Next, high speed cameras measure and calculate

17:36 the trajectory of the droplet and feed the data

17:39 to a set of actuated mirrors in order

17:42 to angle the laser pulses to precisely hit the tin.

17:46 To control the amount of EUV light,

17:48 sometimes droplets are skipped by the lasers,

17:51 and these droplets are captured over here.

17:54 This process of producing high speed tin droplets and then shooting them with 2

18:01 laser pulses to generate EUV light happens at a rate of 50,000 times a second.

18:07 Now that we have EUV light, the first mirror called the collector,

18:12 focuses all the EUV light to a small hole called the intermediate focus,

18:17 which only EUV light can pass through.

18:20 The light next enters into the illuminator,

18:23 which is composed of the field facet mirror,

18:26 the pupil facet mirror, and another set of mirrors.

18:30 These mirrors are so perfectly shaped that there’s

18:33 less than an atom’s deviation from the surface.

18:37 The illuminator takes this EUV light and shapes it into a thin ribbon

18:42 that has equal uniformity across a well-defined

18:45 range of angles before it hits the photomask.

18:48 Using an equal uniformity of light at all angles is critical to imprinting

18:53 a perfect nanoscopic pattern from the mask via the light and onto the wafer.

18:59 We want to take a short detour and mention that this has

19:03 been a rather challenging video to make simply because there’s

19:06 a mountain of science and engineering inside these machines built

19:10 by ASML and this video only explores the tip of the iceberg.

19:15 Essentially, a lot of the details had to be

19:18 cut in order to keep this video a manageable length.

19:22 For example, EUV light is incredibly difficult

19:24 to work with because it’s absorbed by atmospheric molecules,

19:29 and therefore the entire light path and wafer

19:32 carrier stage is connected to vacuum pumps, which remove all the air.

19:37 Additionally, EUV light is absorbed

19:39 by glass and practically all other materials,

19:42 and therefore to focus and transport the light,

19:46 this system uses mirrors rather than transmissive lenses.

19:50 However, these mirrors called Bragg Reflectors,

19:53 are nothing like the mirrors in your bathroom,

19:57 but rather they’re composed of dozens

19:59 of alternating layers of silicon and molybdenum,

20:02 each only a few nanometers thick.

20:05 When EUV light hits the surface of this Bragg Reflector,

20:09 only 3% is reflected at each boundary layer while the rest passes through.

20:15 But with so many layers the cumulative

20:18 3% reflections add together using constructive interference,

20:22 resulting in a total of 70% being reflected for a single mirror,

20:27 while 30% of the light is lost and absorbed.

20:31 However, with more than 10 mirrors in the optical system,

20:35 and only 70% reflection at each one,

20:37 the final light hitting the wafer is less than 10

20:40 percent the brightness of the light emitted by the tin plasma,

20:44 which is why the initial light from the source

20:47 vessel needs to be as bright as possible.

20:50 Another example of the incredible engineering inside this machine

20:54 is that this field facet mirror is assembled from hundreds

20:58 of independently controlled mirrors that can be angled to direct

21:02 the light onto specific regions of the segmented pupil facet mirror.

21:07 Together, these two mirrors take the cone of EUV light

21:11 from the intermediate focus and turn it into a complex pattern of illumination.

21:16 For example, this is called annular illumination,

21:21 here’s dipole illumination and then here’s quasar illumination.

21:25 You’re probably wondering why we

21:28 require such complicated patterns of illumination.

21:31 Well, when we look back at the microchip,

21:34 one layer of wires is running mostly horizontally,

21:37 the next layer is a set of cylinders called vias,

21:41 and then the following layers have wires that run vertically,

21:45 and each layer uses a different mask.

21:47 Earlier we said that the EUV light is kind of like the tip of a fine tipped pen.

21:54 Having different patterns of illumination is

21:56 like holding the marker at different

21:58 angles with respect to the lines or circles that are being patterned.

22:02 Specifically, annular illumination is best used to pattern the layers containing

22:09 vias and is like holding the marker straight up and down,

22:13 whereas dipole illumination like this is

22:16 best used to pattern lines running horizontally, and then we rotate the dipole

22:21 illumination for patterning the vertically oriented wires.

22:24 Imagine being at the forefront of this groundbreaking science and engineering.

22:30 Then picture ASML, whose work powers the innovations

22:35 that solve some of humanity's toughest challenges in energy,

22:39 mobility, and healthcare.

22:41 ASML is a leader in photolithography systems,

22:44 serving as the backbone for the world's leading

22:47 chipmakers and enabling the technology that drives our future.

22:52 With over 44,000 talented individuals and growing,

22:56 ASML is headquartered in The Netherlands,

22:59 with major R&D and manufacturing sites in the U.S.

23:03 and Asia.

23:03 Their sprawling campus is not just a workplace;

23:07 it’s an exceptional environment where cutting-edge technology comes to life.

23:11 To keep pushing the boundaries of what's possible,

23:15 ASML seeks exceptional talent.

23:17 They are looking for scientists and engineers

23:20 ready to design the next-gen lithography systems,

23:24 technicians and logistics experts eager to build,

23:27 ship, and support these groundbreaking systems,

23:30 and software developers passionate about working in a world of nanometers.

23:35 ASML is the next step for those ready to make an impact in an inspiring setting.

23:41 Together with their suppliers, partners, and customers around the world,

23:45 they are committed to powering technology forward.

23:48 Visit their website using the link in the description

23:52 to learn more and start a journey with ASML today.

23:56 Let’s move onto the next part of this EUV

24:00 Lithography tool and explore the photomask or mask

24:04 which is also called a reticle and contains

24:07 the entire design of a single layer of a microchip.

24:11 The mask starts in a doubly sealed pod and is

24:15 loaded onto the machine using an overhead transport system.

24:18 The outer protective carrier is opened,

24:21 and a robotic arm picks up the inner pod and carries it to a vacuum load lock.

24:27 The chamber is sealed and pumped down to a vacuum, and the inner door opens.

24:32 Next the inner pod opens up and a separate robotic

24:36 arm carries the mask and base to an inspection station.

24:39 Each mask has half a dozen different marks

24:42 including a bar code as well as fiducials,

24:45 which are designs used to align the mask with sub-nanometer level accuracy.

24:50 The mask is carried over to and loaded

24:53 onto the reticle stage which moves back and forth

24:57 across the EUV beam with incredible accuracy

25:00 and at high speeds with more than 7Gs of acceleration.

25:04 This mask’s surface is built from the same

25:08 Bragg reflector surface mentioned earlier but with a pattern

25:12 of absorbers on top that locally blocks the light

25:15 in order to create the detailed microchip layer pattern.

25:19 This 6 by 6-inch mask has a pattern area of 104 by 132

25:25 millimeters and an absorber pixel resolution of below 10 by 10 nanometers.

25:32 In the beginning of this video we showed

25:35 a variety of different chips with different sizes,

25:38 and shortly after we showed a GPU being patterned across the wafer.

25:43 The pattern on the mask is 4 times larger than the microchip and this GPU

25:48 chip is close to the maximum size chip that can fit on the mask,

25:53 and therefore only one copy fits,

25:56 resulting in 90 GPU chips fitting onto a 300-millimeter wafer.

26:01 However, CPU Chips are typically smaller and therefore,

26:06 in the following example,

26:08 we can fit 2 copies on the mask and a total of 185 chips on the wafer.

26:13 When we look at even smaller DRAM chips,

26:17 12 copies can fit on the mask, yielding 978 chips on the wafer.

26:23 Technically, the exposure field is one scan of the mask onto the wafer,

26:28 and an exposure field can have anywhere from 1

26:31 to a dozen or more die patterns on it,

26:34 yielding around a 100 to a thousand or more chips on a single wafer.

26:38 This mask contains an incredible amount

26:41 of information and, as mentioned in the intro,

26:44 it has the equivalent amount of detail as all the text

26:47 of Wikipedia plus all the books in an average public library.

26:52 This mask, which can cost around 300,000 dollars,

26:55 must be so perfect that using our analogy

26:59 there can’t be a single grammatical error,

27:02 spelling mistake, or even an extra curve

27:06 on a letter across 21 million pages of text,

27:10 otherwise it would damage every chip on the wafer.

27:14 Also, if you’re curious, here are the calculations we used

27:18 for the transistors to text and book conversions.

27:21 Pause the video to work it out.

27:24 The next topics we’ll explore are the projection optics

27:27 and how the wafer is moved around the machine.

27:30 But first we’d like to mention that this video topic

27:34 is incredibly complicated and took hundreds of hours to research,

27:39 write, model, animate, and edit, totaling over 1100 hours.

27:44 So, if you could take a few seconds to ‘Like’ this video,

27:49 subscribe, comment with a quick message below, and most importantly,

27:52 share it on social media and with a friend, family,

27:56 or work colleague, it would help far more than you think!

28:00 Additionally, we have a Patreon page with AMAs and behind the scenes footage,

28:05 and, if you find what we do useful, we would appreciate any support.

28:11 Thank you.

28:12 So, let’s move on to the projection optics.

28:17 These optics are composed of a series of mirrors that are used to project

28:21 and focus the patterned EUV light onto the wafer with extremely high accuracy,

28:28 while minimizing wavefront aberrations,

28:30 and shrinking the image by a factor of 4.

28:34 These mirrors are designed and manufactured by Zeiss,

28:37 who is a longstanding partner of ASML and has been a vital

28:41 collaborator in the development

28:43 of the optics systems inside photolithography tools.

28:46 To understand the projection optics,

28:48 we have to discuss what determines exactly how small these wires can be,

28:54 and for this Rayleigh’s Criterion Equation is used.

28:58 This equation states that the smallest dimension or critical dimension,

29:03 is equal to k1 times the wavelength of light

29:07 or Lamda divided by the numerical aperture, or NA.

29:11 The wavelength of EUV light is 13 nanometers.

29:15 K1 is the process factor,

29:17 which relates to the various illumination settings created

29:21 by the field and pupil facet mirrors that we discussed earlier,

29:26 along with the photoresist,

29:27 and other factors and is close to .3 for this machine.

29:32 Finally, numerical aperture or NA is a measure of the angle and amount

29:37 of light the mirrors in the projection

29:39 optics can capture and focus onto the wafer.

29:43 Numerical aperture isn’t just about increasing the brightness of the EUV light,

29:47 but rather it’s more of a measure of the angles

29:51 and amount of constructive interference wave paths that hit the mask,

29:55 and then are projected onto the wafer.

29:58 In short, with a larger numerical aperture or NA,

30:02 which corresponds to a larger angle

30:04 between the projection mirrors and focal point,

30:07 we can achieve a smaller resolution.

30:10 This tool has a numerical aperture of 0.33,

30:14 however the next generation of EUV lithography systems called high NA,

30:21 increases this to 0.55 resulting in an 8-nanometer critical dimension.

30:27 Increasing the numerical aperture to 0.55

30:30 requires significantly larger mirrors which results

30:34 in a redesign of the entire optics system and other parts of the machine,

30:41 thus considerably increasing the size and cost of the system.

30:47 We could spend an entire video discussing the next generation High NA tool,

30:52 but instead let’s move on to discuss the wafer transport system and wafer

30:57 stage and see how a wafer makes its way to the EUV exposure station.

31:02 Let’s start with a wafer that’s carried

31:05 in a foup on the overhead transport system.

31:08 This foup lands on the lithography cluster,

31:11 which is a combination of a wafer track tool and a lithography tool.

31:17 The wafer first enters the track tool where

31:20 a layer of photoresist or resist for short, is evenly spread across the wafer.

31:26 The wafer moves to another area inside the track tool

31:30 where it’s heated in order to dry and solidify the resist.

31:34 Next, using robotic arms the wafer is carried from the track

31:39 tool into a vacuum load lock inside the EUV tool.

31:43 The pneumatically actuated doors close and the chamber

31:46 is pumped down to a vacuum.

31:49 Next the back doors of the load lock open up and a different

31:54 robotic arm carries the wafer to one of the wafer stages.

31:58 This system is called a TWINSCAN because there are

32:01 two complete wafer stages that concurrently move two wafers around.

32:05 The key idea is that while one wafer is actively being patterned,

32:10 a second wafer is being loaded onto

32:13 the wafer stage and measured under an alignment sensor.

32:17 Nanometer level accuracy is crucial with these machines

32:20 and one key philosophy of ASML is “Meten is Weten” which is Dutch

32:27 for “To Measure Something is to Know something”.

32:30 The reason for requiring this level of perfection is

32:33 that when we look at the nanoscopic layers of the microchip,

32:37 which has wires and holes that are only 10 to 20 nanometers wide,

32:41 if one layer is more than a couple nanometers off the previous layer,

32:46 then the electrical connections won’t conduct electricity correctly,

32:49 and if an entire layer is off,

32:53 then every single chip will be catastrophically destroyed.

32:57 To make sure that the layer being

32:59 patterned is perfectly aligned with the previous layer,

33:02 the entire wafer is thoroughly measured by the alignment sensor.

33:06 On the wafer are hundreds of alignment marks,

33:10 which are reference patterns that assist in determining

33:13 the exact position of the earlier layers of patterns.

33:17 The alignment sensor meticulously measures the X

33:20 and Y positions of every alignment

33:23 mark on the wafer and builds a highly accurate 2D map from the results.

33:29 This map shows some regions of the wafer being biased

33:33 in one direction by a few to dozens of nanometers,

33:36 and another region being biased in a different direction.

33:41 Additionally, the leveling sensor uses grazing

33:43 incident light to measure the exact

33:46 height of the wafer and builds a topological map of the wafer

33:50 which is critical for later bringing the wafer stage to the position

33:54 such that the EUV light is perfectly focused onto the wafer.

33:59 Now that we’ve measured and built the exact

34:02 alignment and height map for the wafer,

34:05 the wafer stage next moves to the EUV exposure station.

34:09 As the wafer is being patterned,

34:11 the wafer stage moves in perfect synchrony with the reticle stage,

34:16 but only a quarter of the distance due to the 4 to 1 reduction.

34:20 At the same time, the stage makes nano-scale adjustments using the alignment

34:25 map so that the new layer perfectly aligns with the previous layer.

34:30 When the wafer stage moves from one exposure field to the next,

34:35 it’s important that no EUV light hits the wafer,

34:39 and thus a shutter positioned up here, near the reticle stage closes.

34:44 Once the wafer stage is positioned to pattern the next microchip,

34:49 the shutter opens and the wafer stage

34:52 and reticle stage move in perfect synchrony again.

34:55 This process repeats until the entire wafer

34:59 is patterned taking around 18 seconds in total.

35:03 So then, what actually happens as EUV light hits the photoresist?

35:09 Well, resist is a polymer mixed with a Photo Acid Generator.

35:14 When high energy EUV photons hit the resist,

35:17 the light ionizes it releasing high energy electrons.

35:21 These electrons then hit the photo acid generator producing

35:25 an acid that breaks apart the polymer making it weaker.

35:29 As a result, the areas hit by the EUV light become soluble

35:34 and are washed away by a developing liquid in the subsequent process step.

35:40 One detail is that the resist has an extremely high contrast,

35:44 meaning that at a certain level of EUV light,

35:47 the entirety of the resist hit by that light is broken down.

35:52 This is critical in producing sharp patterns and walls on the resist.

35:57 Let’s next explore how the wafer and wafer stage move around.

36:02 Specifically, the wafer stages levitate on a large

36:06 magnetic table composed of more than a thousand magnets.

36:10 Electromagnets on the underside of the wafer stage move it

36:14 along this magnetic table both quickly and with micrometer level

36:19 accuracy while interferometers on the top of the stage measure

36:23 its exact position and this setup is called the long-stroke stage.

36:27 In order to secure the wafer,

36:29 it’s placed on a plate which is technically called an electrostatic clamp.

36:33 The clamp cycles zones of high voltage across

36:36 the backside of the wafer to keep it in place,

36:40 a phenomenon similar to sticking a ballon to a wall using static electricity.

36:45 To reach nanometer level accuracy the plate is independently

36:50 moved using smaller motors which is called the short-stroke stage.

36:55 By combining the long stroke and short

36:58 stroke stages along with measurement encoders,

37:00 the machine can quickly move the wafer as it’s being patterned and maintain

37:06 an accuracy of less than 1 nanometer or approximately 4 silicon atoms.

37:12 Once all the microchip patterns are copied to the wafer,

37:15 the stage moves back towards the robotic arms where the wafer

37:19 is unloaded and placed into one of the vacuum load locks,

37:23 pumped back to atmosphere,

37:25 and then a separate robotic arm brings the wafer back to the track tool,

37:30 where the patterned and modified resist

37:32 is washed away using a developing liquid.

37:35 Finally, the wafer is heated again to further harden the remaining resist.

37:40 The patterned wafer is then loaded back into the foup which is picked

37:45 up and brought to a different tool to undergo processing in other ways.

37:50 Let’s close this tool and that’s it for our journey into photolithography.

37:56 If you have any questions, feel free to ask them in the comments below.

38:00 We’re thankful to all our Patreon

38:03 and YouTube Membership Sponsors for supporting our videos.

38:07 If you want to financially support our work,

38:10 you can find the links in the description below.

38:13 This is Branch Education, and we create 3D animations that dive deeply

38:18 into the technology that drives our modern world.

38:22 Watch another Branch video by clicking one

38:24 of these cards or click here to subscribe.

38:28 Thanks for watching to the end!