⚙️ Mechanical Engineering Resources:
The most asked question in mechanical engineering technical interviews is the cantilever beam question.
We wrote a full guide on how to answer it.
💼 Jobs & Internships
Internship season is in full swing! Companies are one by one trickling out their summer 2025 intern positions.
Freshmen - Check out our 4-Year Plan for how to get internships at startups!
Sophomores - Apply To These Positions
ADP Composites Engineering Intern
L3 Harris Project & Component Engineering 11 Intern
Northrop Grumman - Mechanical Engineer Intern - Redondo Beach
Juniors - Full Send Longshots!
ASML Mechanical Engineering Co-op - Spring/Summer 2025
Polaris Core Engineering Intern
Astranis Thermal Engineering Intern - Fall 2024
Seniors & Graduates - It’s Go (Full) Time!
Super Tough:
Skydio Camera Systems Engineer
Anduril Propulsion Engineer
Shield AI System Design and Rapid Prototyping Engineer
More Reasonable:
RTX Project Engineer
Honeywell Mechanical Design Engineer I - Gears, Bearings, Seals, Lubrication Systems
NVIDIA CAD Engineer
Not seeing what you are looking for? Check out our Job Board for more MechE positions!
👶 Meme Of The Week
🙋♂️ Interview Practice Question of the Week
What are the fundamentals of microfabrication and how do they work?
✅ The Answer
Unless it’s a specialty interview, most interviewers actually won’t dive too deep into the world of microfabrication. HOWEVER, if you’re interested in working on anything interfacing with electronics (pretty much everything), it can really help to know the basics.
The basis of most microfabrication processes can be rolled into lithography, etching, oxidation, thin film deposition, ion implantation, diffusion and metallization (among other things).
As an overview:
Lithography is the process by which we form the fundamental building blocks of devices. In this process, we can use a mask to selectively expose patterns onto a light-sensitive material (photoresist). We can then remove part of the photoresist, giving us the ability to coat/etch/oxidize specific patterns into a silicon substrate.
Etching is used to remove material from the surface of a substrate to create patterns or structures. It can be done using chemical solutions (wet etching) or plasma (dry etching) to selectively remove material based on the pattern defined by the photoresist.
Oxidation involves growing a thin layer of oxide on the surface of a semiconductor substrate, typically silicon. This is usually achieved by exposing the substrate to oxygen or water vapor at high temperatures, forming a silicon dioxide layer that can act as an insulator or mask for further processing.
Thin Film Deposition involves depositing a thin layer of material onto the surface of a substrate. Chemical vapor deposition (CVD), physical vapor deposition (PVD), or atomic layer deposition (ALD) are used to create layers of metals, dielectrics, or semiconductors with precise thickness and properties.
Ion implantation is a method used to introduce dopants into a semiconductor substrate by bombarding it with high-energy ions. This process modifies the electrical properties of the substrate by altering its chemical composition and creating specific regions with different levels of conductivity.
Diffusion is the process of moving dopant atoms into the semiconductor substrate at high temperatures to change its electrical properties. The dopants spread out from areas of high concentration to low concentration, forming regions with controlled impurity profiles.
Metallization involves depositing metal layers on the substrate to create electrical connections between different parts of the device. This process typically uses techniques like sputtering or electroplating to form metal interconnects and contact pads essential for the device's functionality.
That was a lot. Today, we’re mainly going to touch on lithography and a little bit of context into the space.
So… you’ve probably been hearing a lot about Nvidia the last few months. Nvidia blew up (in a good way) this year as their GPUs are significantly better optimized compared to any competitors for training AI models. Given the recent uptick in popularity of LLMs, that means that there is virtually unlimited demand for Nvidia chips. But who makes them???
Transistors are the base unit of computers more or less. It’s why Moore’s law is so famous and why compute has increased so rapidly over the last 30 years (if you’ve ever heard that your phone has more compute than the computers we went to the moon with, this is why). As we make transistors smaller, we can make computers smaller. So to get the highest performance computer, you need the smallest transistors.
TSMC (Taiwan Semiconductor Manufacturing Company) makes the vast majority of high performance capable chips. They are a foundry based out of Taiwan (duh) that makes chips for other companies. They do not own the designs, they merely manufacture. And they are the best in the game at it. They own over 60% of the chip manufacturing market. It’s the reason that tensions between Taiwan and China draw so much US interest. US inability to manufacture high performance chips would essentially kneecap it in the event of global conflict. As a result, the US is currently investing billions in bringing chip production back onshore, but I digress. Why does TSMC have so much market share?
It all comes back to how those pesky transistors are manufactured. The smallest transistors are roughly 3 nm right now. TSMC is the only one who can make those transistors, but who supplies the machines that manufacture them?
Advanced Semiconductor Materials Lithography, or as it is commonly referred to, ASML. They are a dutch company that is the SOLE SUPPLIER of EUV (extreme ultraviolet) machines used to make all chips with transistors <5 nm. The machines in question? Lithographic ones!
FYI TSMC coincidentally owns the vast majority of EUV machines. Each machine costs around $370 million and weighs the same as 2 Airbus A320s. Only 5-6 EUV machines are produced annually, so each one has immense value and provides a moat for TSMC’s first mover advantage. Also, setting up a chip fab and getting it to work well with low downtime? Notoriously difficult. Ask Intel…
So now we’ve come full circle as to why lithography is so important, but how do we do it?
There are a few steps. Note that this is a general overview, and not necessarily tied to EUV though the principle is similar.
1. Photomask and Photoresist
Photomask: A photomask is a template that contains the desired circuit patterns. It is typically made of a glass or quartz plate with an opaque layer that defines the patterns. The photomask acts as a stencil during the exposure process, allowing light to pass through only the clear regions.
Because photomasks are so difficult to make, each one costs between a few hundred thousand and tens of millions. A device may require anywhere between 10 and 100 of these photomasks to get all the steps working. This is primarily why it’s so expensive to fab a new set of chips. The typical paradigm with hardware is to manufacture, test, then iterate. With an iteration costing millions of dollars at a minimum, it becomes crucial to get it right the first time (or at least as early as possible). This was actually one of Nvidia’s early differentiators, trusting and investing in software checks to make sure their chips were functioning before being sent to the fab!
Even though these photomasks are extremely pricey, depending on the exposure (aka printing) method, they can last forever once made.
Photoresist: Photoresist is a light-sensitive material applied to the silicone substrate (aka the wafer). There are two types of photoresists:
Positive Photoresist: Becomes soluble when exposed to light, allowing the exposed regions to be washed away during development.
Negative Photoresist: Becomes insoluble when exposed to light, allowing the unexposed regions to be washed away during development.
2. Coating
The substrate is coated with a uniformly thin layer of photoresist using a process called spin coating. During this process, a few drops of photoresist are placed on the wafer, which is then rapidly spun to spread the photoresist evenly across the surface, forming a uniform film. The boundary layer (if you recall fluid dynamics) acts as an amazingly consistent mechanism for getting uniform layer thickness of a fluid across the wafer.
3. Exposure
In the exposure step, the coated wafer is aligned with the photomask. Light, typically ultraviolet (UV), is shone through the photomask onto the photoresist-coated substrate. This exposure transfers the photomask's pattern to the photoresist. Different types of lithography techniques use different light sources which will ultimately contribute to the minimum achievable feature size. The shorter the wavelength, the smaller the minimum feature size. EUV (extreme ultraviolet) lithography uses light with wavelengths around 13.5 nm, just on the border of X-ray range!
Check out the above picture. Those walls weren’t 3D printed, that’s an extremely zoomed in image of photoresist. Those grooves? STANDING WAVES from the light developing the photoresist. So cool.
There are a few different printing techniques.
Contact printing/exposure: places the mask directly against the substrate. This can damage the mask and the wafer, and limits the number of times the mask can be used, but is simple and has high accuracy.
Proximity printing/exposure: puts a slight gap between the mask and the substrate so that no damage is incurred. However, accuracy takes a hit due to light reflecting around.
Projection printing/exposure: is expensive and difficult, but allows a significant improvement in accuracy. By using a series of lenses (and mirrors sometimes), we can shrink the image. As such, even if we are limited by the resolution of our mask manufacturing, we can just optically shrink the light before it hits the substrate, effectively increasing the resolution. As you can imagine, this opens up a world of possibilities. But now that the mask is larger than its projection on the wafer, it also means we need to step over and re-expose another part of the wafer so that it doesn’t go to waste. This results in significantly higher resolution, but lower units per hour (UPH).
4. Development
After exposure, the substrate undergoes a development process where a chemical developer solution is used to dissolve the soluble parts of the photoresist, leaving behind the desired pattern on the substrate. This step reveals the underlying substrate in the pattern defined by the photomask.
5. Etching
The developed pattern on the photoresist serves as a mask for the etching process. Etching removes the unprotected areas of the substrate, transferring the pattern into the substrate material. There are two main types of etching:
Wet Etching: Uses liquid chemicals to dissolve the exposed substrate material.
Dry Etching: Uses plasma or reactive ions to remove the exposed material, providing better precision and control.
There’s a lot more on etching to get into, but we’ll save that for another time.
6. Photoresist Removal
After etching, the remaining photoresist is removed using a process called stripping or ashing. This leaves the patterned substrate ready for further processing, such as doping, metallization, or additional lithography steps.
That’s the basics of how lithography works! Hope you enjoyed it!
If you’re interested, here’s a fun story on how nanotech got started (more or less) and Feynman Challenge:
The physicist Richard Feynman (professor at Caltech and later a nobel prize winner) gave an after dinner lecture on nanotechnology in 1959, famously titled “There’s Plenty of Room at the Bottom.” It was allegedly somewhat “off the cuff” and no copies of the speech existed until someone who happened to bring a tape recorder transcribed it (a version edited to take out Feynman’s jokes was later released in Caltech’s Engineering and Science Magazine). In it, Feynman issued two challenges:
Build a functioning electric motor within a cube 1/64” (0.4 mm) on each side.
Inscribe a book page on a surface area 25,000 times smaller than its standard print (the entire Encyclopedia Britannica could fit on the head of a pin at this size).
The prize for each? $1,000 (a little over $10,000 inflation adjusted today).
The first challenge was vanquished 11 months later by William McLellan. What’s crazy is that he was just a meticulous guy. He didn’t use any new technology and literally just made the world’s tiniest motor with 13 parts weighing 250 micrograms capable of spinning 2000 rpm.
The second challenge, despite hundreds of attempts and shams being shown to Feynman, went uncompleted for 26 years. Tom Newman at Stanford completed the challenge in 1985 by transcribing the first chapter of Dickens’ “A Tale of Two Cities” onto a piece of polymethyl methacrylate at 1/25,000 scale. He essentially ran an electron beam over the material to break organic molecules making the exposed area more soluble to developer solution. Sound familiar? (hint: photoresist).
PS: Here are some pictures from the wafer I had the opportunity to fabricate during an exceptionally cool microfabrication course back in college.
