SemiAnalysis Profile picture
Oct 26, 2025 11 tweets 7 min read Read on X
Etching is a process used to remove material from the wafer surface to meet the design requirements of an integrated circuit (IC).
There are two types of etching: one is patterning etching, which removes material in specified areas, such as transferring patterns from a photoresist or hard mask layer onto the substrate film. Another type is blanket etching, which removes the entire surface film to meet process requirements, for example, backside wafer etching. (1/11) 🧵
Etching also can be categorized into two types based on characteristics: wet etching and dry etching. Wet etching is typically performed at room temperature, requiring no additional vacuum equipment, RF systems, or gas delivery setup. The process is relatively easy to control, making the equipment significantly cheaper than that used for dry etching. Below, we will introduce each in detail. (2/11)Image
However, the chemical reaction itself has no directional preference—making wet etching inherently isotropic.
Isotropic etching means that material is removed not only vertically but also laterally, leading to an undercut effect. This undercut prevents accurate pattern transfer to the wafer and can cause line collapse in sub-3-micron processes, leading to their gradual replacement by dry (plasma) etching. (3/11)Image
Image
Common wet etching processes are applied to silicon dioxide (SiO₂), single-crystal silicon, polysilicon, silicon nitride (Si₃N₄), and various metals.

⚆ Silicon dioxide etching typically uses hydrofluoric acid (HF) diluted at 6:1 or 10:1, since a 1:1 HF mixture etches too rapidly at room temperature, making it difficult to control the rate and uniformity. A 10:1 HF solution can also be used to remove the native oxide layer on the surface of a silicon wafer.
⚆ Single-crystal silicon etching is used to form isolation regions between transistors, while polycrystalline silicon etching is used to form gate electrodes and local interconnects. It’s typically performed using a mixed solution of nitric acid (HNO₃) and HF. In this process, nitric acid oxidizes silicon into silicon dioxide, which then reacts with HF and dissolves, exposing the underlying silicon to further oxidation by nitric acid. Through repeated cycles of these reactions, silicon is gradually etched away.
⚆ Silicon nitride etching is commonly used in insulation processes. Phosphoric acid (H₃PO₄) is the typical etchant because its reaction products, silicon phosphate (Si₃(PO₄) ₄) and ammonia (NH₃)—are both soluble in water.
⚆ In metal etching, a mixture of phosphoric acid, acetic acid, and nitric acid was traditionally used for aluminum patterning processes. However, as advanced process requirements have evolved, it is now mostly used to evaluate the quality of PVD-deposited aluminum films. (4/11)Image
Interestingly, semiconductor engineers once sought to eliminate all wet processes from chip manufacturing, but the widespread adoption of CMP and CVD has made it difficult to completely remove wet processing. Because wet processes offer high selectivity, etching efficiency can also be used to evaluate film quality, making this another important application of wet process equipment. (5/11)Image
Dry etching uses gaseous chemical etchants to react with materials and produce volatile byproducts that can be removed from the substrate. At the same time, plasma bombardment of the wafer surface causes ion impacts that not only physically remove surface material but also break the chemical bonds of surface atoms, significantly increasing etching efficiency. (6/11)Image
Process:
⚆The etching gas is first introduced into a vacuum reaction chamber
⚆After pressure stabilization, an RF discharge generates plasma
⚆Some etchant molecules are decomposed by high-energy electron collisions, producing reactive free radicals.
⚆These radicals diffuse through the boundary layer to the wafer surface and are adsorbed
⚆Under ion bombardment, the radicals quickly react with surface atoms to form gaseous byproducts, which detach from the wafer surface, diffuse back through the boundary layer, and are eventually exhausted from the chamber. (7/11)Image
Next, we will introduce process control in etching:
⚆ Etch rate, which measures how quickly material is removed during etching. It is calculated as: Etch Rate = (Initial Thickness−Final Thickness)​/ Etching Time
⚆ Uniformity is a key parameter to ensure consistent etch rates and high repeatability across the entire wafer. It is determined by measuring the etch rates at multiple predefined points on the wafer and calculating their variation.
⚆ Etch selectivity refers to the ratio and difference in etch rates among the photoresist, films, and substrate during the etching process. Selectivity = Etch Rate1/Etch Rate2 (8/11)Image
⚆Etch profile/outline is one of the most critical characteristics, as it directly affects subsequent deposition processes. The shape of the profile varies depending on the process, but the ideal form is vertical, since it ensures that the pattern defined in the photoresist is fully transferred without critical dimension (CD) loss.
⚆Over-etching occurs because etch rates and film thicknesses vary across the wafer, often leaving some areas where the film has not been completely removed. The additional etching step used to eliminate these remaining films is called over-etching.
⚆Residue refers to unwanted material that may remain on the sidewalls or wafer surface after etching. Such residue often results from incomplete over-etching caused by the wafer’s complex surface topology.(9/11)Image
Another worth mentioned topic is Atomic Layer Etching (ALE) refers to a process that removes material with atomic-level precision, typically at an average rate of about 0.1 nm/min.
There are two types of ALE processes — ions-first and ions-last — depending on whether ions are introduced in the first or last step.
Ions-last: Plasma such as Cl₂, O₂, or CHF₃ (depending on the substrate) is introduced into the reaction chamber first. The plasma reacts with the surface, forming bonds with the layer in a process called adsorption. Like ALD, this reaction is self-limiting, stopping once the entire surface is occupied. Next, Nitrogen or Argon are used to purge unreacted species. Finally, ions (e.g., Ar⁺) bombard the surface to remove the reacted layer in a step known as desorption.
Ions-first: Ions are introduced first and implanted into the surface through diffusion or heating. Plasma is then introduced to remove the modified surface layers.(10/11)Image
In the etch equipment market, Lam holds the leading position, followed by TEL and AMAT. As mentioned earlier, wet process tools are relatively simple, and with rising cleanliness requirements across advanced nodes, several Taiwanese companies—GPTC, Scientech, SHT and GPM—have entered the competition. (11/11)Image

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with SemiAnalysis

SemiAnalysis Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @SemiAnalysis_

Jan 31
Backend networking architecture is instrumental to networking ownership cost analysis. How does networking architecture matter for networking ownership costs? In general, we observe the following trends:

🟠Hyperscalers typically get preferential terms on networking equipment compared to neoclouds;

🟠Infiniband-based networks are more expensive than Ethernet-based networks for the same networking type and cluster size;

🟠3-Layer networks would be more expensive than 2-Layer networks on a per rack server (1 rack server = 72 GPUs in the case of GB200 and GB300) basis; and

🟠All else constant, G300 networks running on CX-8 NICs are expected to be more expensive than GB200 networks running on CX-7 NICs on a per rack server basis because of the doubling of per GPU bandwidth.
(1/6)🧵
Before we dive into several examples of how architecture influences networking costs, we define the following terminology used to describe networking clusters:

🟠Network layers: Number of switch layers required to connect all GPUs within the same cluster, typically 2 or 3 layers

🟠Rails: Number of pathways you can split a server tray, which is equal to the GPUs per server tray or 4 in the case of GB200 or GB300 deployments

🟠Planes: Number of pathways you can split a NIC

🟠Attach Rate: GPU per Networking Component (or the reverse)
(2/6)
A 36,864-GPU Infiniband cluster consisting of 4 rails and 3 layers can support 9.76 GPUs per QM9700 Switch (25.6T per Switch). For the same cluster size and type, assuming that we swap all GB200 chips for GB300 chips, the number GPUs that can be supported per QM9700 Switch theoretically halves from GB200 to GB300 because the bandwidth of GB300 is double. (3/6)Image
Read 6 tweets
Jan 29
Ever wonder how an Nvidia GB200 NVL72 gets from the factory to the data center? You can’t just throw a state-of-the-art 3,000 lb rack (compared to ~500 lb for CPU-based server) on a standard FedEx truck. Moving AI infrastructure is more like transporting a heart for transplant than shipping electronics. (1/9) 🧵Image
A fully loaded AI rack can weigh up to 3,700 lbs. That’s the weight of a Ford Explorer concentrated into a 2x4 foot footprint. Standard pallets would crush instantly. These AI racks require custom-engineered reinforced bases with shock-absorbent foam just to keep them from tearing through the floor. (2/9)Image
Unlike Ford Explorers that are made for ruggedness and the great outdoors - AI servers are packed with thousands of costly and sensitive components (think accelerator, HBM stacks, liquid cooling loops etc.) that are hyper-sensitive to any sudden movement. A single hard "thud" can dislodge a single component, rendering the rack inoperable and making maintenance necessary.

To protect delicate components from damage during transit, these racks sit on integrated heavy-duty shock absorbers and specialized shock-mount bases designed to neutralize impact. This setup is paired with custom foam layers for advanced vibration dampening, which effectively mitigates vertical G-forces. By allowing the pallet to "bounce" and absorb the energy from road hazards like potholes, the packaging ensures the high-value silicon and interconnects remain perfectly stationary and secure. (3/9)Image
Read 9 tweets
Jan 28
When people think about semiconductor manufacturing, many picture highly automated robotic arms and overhead transport systems. However, the true unsung heroes behind high‑volume chip production are metrology and inspection. If you cannot see it, you cannot manufacture it; and if you cannot measure it, you cannot achieve yield. (1/4) 🧵Image
Although these two terms are often discussed together, they actually refer to two distinct domains.

Metrology:
The core of metrology is accuracy and consistency.
It involves ultra‑precise measurements of critical dimensions, film thickness, surface topology, and overlay alignment in advanced, highly scaled manufacturing. Metrology ensures that interconnects and billions of transistors conform precisely to their intended physical and design specifications

Inspection:
The core of inspection is defect capture and yield improvement.
It relies on optical or e‑beam scanning to identify particles, scratches, pattern defects, or electrical anomalies on the wafer surface. Inspection determines whether a fab can catch defective wafers at the earliest possible stage—before hundreds of thousands of dollars of processed silicon turn into scrap. (2/4)
Why have metrology and inspection become so critical in advanced nodes?Image
Why have metrology and inspection become so critical in advanced nodes?

1. Shrinking margin for error:In advanced process nodes, the alignment requirements between layers are extraordinarily tight. Even a slight deviation in metrology can cause interconnects to either break or short, directly impacting device reliability and overall yield.

2. Increasing structural complexityAs transistor architectures transition to FinFET or even GAA, traditional optical inspection tools are no longer sufficient. Metrology must evolve to effectively “see through” these highly complex gate‑all‑around structures and verify that every nanosheet layer maintains precise thickness and spacing uniformity.

3. Yield is profit: shortening the learning curveIn advanced manufacturing, yield directly determines profitability—good yield versus poor yield can be the difference between earnings and losses. High‑throughput inspection systems can flag process excursions within hours, enabling engineers to quickly adjust parameters. This rapid feedback loop is essential to accelerating yield ramp. (3/4)Image
Read 4 tweets
Jan 22
The U.S. wants 40% of chips made onshore. But quietly, the equipment that makes those chips is moving offshore fast. This gap matters just as much as the chips themselves. (1/6) 🧵 Image
In 2022 we flagged Lam Research expanding in Malaysia. Today, most of its high-volume production is there, not the U.S. That trend has only accelerated. (2/6) newsletter.semianalysis.com/p/lam-research…
Applied Materials is following. Its Singapore 2030 plan aims to double manufacturing capacity in the country over this decade. (3/6) Image
Read 6 tweets
Jan 3
Rolling into the new year, 2 of the Six Tigers quietly filed their IPO prospectuses and will start trading in early January if all goes well. We finally get a glimpse into audited financials of foundation model labs. TLDR: Building Machine God Ain't Cheap. (1/5)🧵 Image
MiniMax (0100 HK) and aka Knowledge Atlas fka ZhiPu (2513 HK) both give a glimpse into the economics of an AI Lab, demonstrating strong product momentum as well as a flagrant disregard for profitability. (2/5) 🔥📉Z.aiImage
Image
Similar to OpenAI and Anthropic, both labs have a “bundle” subscription biz typically consumer facing and a “variable” API biz for enterprises, including major U.S. SAAS/Internet companies like Netflix, Adobe, and others. Z.ai gross margins are more similar to the US labs while MiniMax's is much lower, likely due to the large subsidization of free Talkie users. (3/5) 🤑💸Image
Image
Read 5 tweets
Dec 23, 2025
If you want to power a datacenter off the grid, a gas turbine is the "obvious" choice. But it might not be the best option! Many developers select reciprocating engines for a reason. (1/4)🧵 Image
A recip is more modular than turbines, happier at partial loads, and more comprehensible to maintain. You're mostly changing lubricants, whereas a turbine requires no maintenance...until it needs a massive overhaul. (2/4)
More importantly, recips are less exotic technology. They rely less on rare alloys and critical minerals, and there are WAY more vendors to choose from, speeding up lead times and purchaser power. (3/4)
Read 4 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(