Why is the chip so small? How are more than 10 billion transistors packed in?

Chips are hidden in electronic devices that can be found everywhere in the city. Smart phones, computers, home appliances, etc. cannot be separated from their control.

A tiny chip integrates a huge circuit .

If you zoom in on the chip, you can see that there are densely packed circuits inside it, like densely interwoven highways, as if an orderly circuit city has been built on a very small scale.

Chip structure diagram (naked eye and microscopic) | Image source: pixabay

How small is the inside of a chip? Today, the smallest chip process used in industry , that is, the smallest size that humans can create , has reached 3nm , and tens of billions of transistors can be integrated inside a chip.

The “multi-layer” approach to chip manufacturing

Countless nanoscale electronic components are arranged on the chip. Are each component made in advance and then placed one by one?

Image source: pixabay (top); Searchmedia - Wikimedia Commons (bottom)

No! We can look at this problem from another angle. If we observe carefully in the vertical direction, we can find that the chip is made up of layers of sheet structures with different patterns stacked vertically . If we make each layer in advance and then stack them vertically, the two-dimensional structure can be stacked into a three-dimensional device, and finally form a chip with rich functions.

Vertical observation of the internal structure of the chip | Image source: Searchmedia - Wikimedia Commons

Now our goal has become how to make a sheet structure with a specific pattern. First, we need a sheet material that can be used to print circuit diagrams , which is the silicon wafer we often hear about. This is a very pure silicon that is processed and cut into smooth, extremely thin discs.

Silicon wafer | Image source: pixabay (left); Searchmedia - Wikimedia Commons (right)

Next, like a carpenter, we need to find the right tools to carve patterns. To make chips with complex internal structures and extremely small sizes, the size of the processing tools must be extremely high.

We are smart and have found the chisel of light. Because light has rich wavelengths , we can use short-wavelength light to achieve extremely fine processing.

The rich wavelengths of visible light (invisible light is even richer) | Image source Searchmedia - Wikimedia Commons

We hope to transfer the circuit pattern designed on the drawing to the silicon wafer through optical exposure , but light cannot affect silicon materials, so we need to use an intermediate material, which is a photoresist that can directly interact with light.

Photoresist spin-coated on a silicon wafer (uniform coverage due to centrifugal force of rotation) | Image source: Searchmedia - Wikimedia Commons

To enable light to transmit pattern information , light can be completely blocked or completely passed to produce light and dark patterns. When light passes through a light-blocking plate (mask) with a circuit pattern , the pattern information of the mask can be copied , and finally, after interacting with the photoresist evenly covered on the surface of the silicon wafer, the pattern information we need appears on the silicon wafer.

Photolithography exposure process | Image source: Searchmedia - Wikimedia Commons

Photoresist is the main carrier medium for photolithography imaging and is divided into positive photoresist and negative photoresist. The exposed area is more easily dissolved in the developer is the positive photoresist, and the exposed area is less likely to dissolve in the developer is the negative photoresist.

Two results of the exposure process (positive and negative) | Image source Searchmedia - Wikimedia Commons

Assuming that positive photoresist is used, when the exposure process is completed, the developer can dissolve the photoresist exposed to light. Then, chemicals are used to dissolve the exposed silicon wafer , and the photoresist left on the surface of the silicon wafer can protect the silicon wafer. This is the etching process.

Now we have achieved our goal and obtained a silicon wafer with a specific circuit pattern. In this whole process, the general idea is actually quite smooth, but chip manufacturing, a precision engineering project that represents the pinnacle of human wisdom, contains countless stringent requirements.

What are the limitations on the internal size of a chip?

The main component of a chip is the transistor. A large chip can have tens of billions of transistors. The smaller the transistors we can make, the more components the chip can accommodate and the lower the power consumption of the transistors.

In chip manufacturing, we hope to use light to create circuit patterns on a small scale , so why can light achieve this effect? ​​What is the limit of light carving?

1

diffraction

The main reason that affects the level of light engraving is the diffraction effect of light . Light is an electromagnetic wave, and diffraction is inevitable during the photolithography process, so the exposure range has a minimum feature size. The resolution of light, that is, the ability of photoresist to reconstruct patterns based on light irradiation, is limited .

Diffraction during exposure | Image source: Searchmedia - Wikimedia Commons

As shown in the figure below, when a beam of parallel light passes through a slit, the light will interfere with each other in the form of countless sub-waves during the propagation process, forming a diffraction pattern of alternating light and dark.

Single slit diffraction | Image source: Searchmedia - Wikimedia Commons

That is to say, when considering the propagation of light on a microscopic scale, the light area is no longer clearly distinguished from the dark area, but a fuzzy zone appears . After the light emitted by an ideal object point passes through the edge of an obstacle, it will deviate from the characteristics of geometric optics straight-line propagation and no longer form an ideal image point.

This is because when the slit width is comparable to the wavelength of light , the wave effect of light comes into play. Light can use the wave effect to bypass obstacles and diffuse in space , forming a diffraction effect of light divergence, causing the exposure area to no longer be accurate and the resolution of light to have a limit .

Wave effect of light (comparing linear propagation and wave effect) | Image source Searchmedia - Wikimedia Commons

2

Resolution

In the field of optical imaging, resolution is a measure of the ability to separate the images of two adjacent object points . Ideally, we want each object point to produce a sharp image point , but due to diffraction, the actual result is a light spot of a certain size . If the two light spots (diffraction patterns) overlap too much, the image points will be difficult to distinguish.

Rayleigh proposed an effective criterion, and the resolution calculation formula is:

This resolution expression describes the limit position that can be resolved when two spots are seen - when the maximum position of one spot coincides with the first zero point of the other spot. Where λ is the wavelength of the illumination light.

The limiting cases where the light spots are indistinguishable and just resolvable | Image source: Searchmedia - Wikimedia Commons

NA is the numerical aperture, which describes the ability of the lens to focus light . It is specifically expressed as the degree of deflection of parallel light after it is incident (focused). The calculation expression is:

Numerical aperture (n is the refractive index) | Image source Searchmedia - Wikimedia Commons

The Rayleigh criterion is often used to evaluate imaging quality, and the photolithography system forms images in photoresist. Photoresist is a high-contrast imaging medium . Under certain exposure conditions, although the optical resolution has reached below the resolution limit of the Rayleigh criterion, photoresist can still present good imaging results and achieve the processing goal.

The resolution of photolithography is:

Rlitho is the graphic period that can be resolved by the lithography system; k1 is the process factor.

3

Lithography

Photolithography is the most complex, expensive and critical process in chip manufacturing. A projection lithography system is usually used to project the circuit structure of the mask onto the surface of the silicon wafer.

Optical lenses can gather diffracted light to improve imaging quality . In photolithography, in order to obtain the smallest possible pattern, a projection imaging objective lens with a reduction ratio is used between the mask and the photoresist.

Projection lithography system | Image source: Internet

How do I polish this carving knife?

We now know that the minimum processing scale (resolution) of light determines how small the chip can be. How can we make the chip smaller? We need to make the resolution stronger and the circuits on the chip more functional.

Based on the three terms in the photolithography resolution formula, we have three options for polishing the carving knife.

Increase the numerical aperture of the lithography system

The larger the numerical aperture of the projection lens in the lithography imaging system, the better the resolution. The specific operation is to design an immersion lithography machine , that is, to fill a high refractive index medium between the wafer and the last lens of the projection lens.

Shorten the wavelength

The wavelength of light in the photolithography process has gone through the development of the deep ultraviolet band of G line (432nm), I line (365nm), KrF (248nm) and ArF (193nm). Currently, the extreme ultraviolet lithography machine (EUV) with a wavelength of 13.5nm has been put into use.

Reduce process factors

The lithography resolution can also be improved by optimizing the lithography process parameters , such as improving the lighting conditions, photoresist process and mask design. These methods can reduce the process factor k1 and are called resolution enhancement technology (RET).

Light is an electromagnetic wave , so it contains information such as amplitude, phase, polarization state and propagation direction . Photolithography resolution enhancement technology is to obtain finer graphic structures on photoresist by regulating the above four types of light information. For example, off-axis illumination technology can change the amplitude and phase, optical proximity effect correction technology can change the amplitude of light waves, and light source-mask joint optimization can change the propagation direction, amplitude and phase of light waves.

Relationship table between various process nodes and lithography technology | Source: Sacco Micro Semiconductor official website, ASML, Zhongtai Securities Research Institute

Looking at the development history of lithography machines, we are indeed running along the path of continuously shrinking the wavelength. Observing the data in the table, when the wavelength of the light source is the same, we are still continuously shrinking the process , which is the result of numerical aperture, process factors and other complex technologies.

References

[1] Wei Yayi. Computational lithography and layout optimization[M]. 1. Publishing House of Electronics Industry, 2021.

[2] Stephen A. Campbell. Micro- and Nanoscale Manufacturing Engineering[M]. 3. Publishing House of Electronics Industry, 2010.

Planning and production

Source: Institute of Physics, Chinese Academy of Sciences

Editor: Wang Mengru

Proofread by Xu Lai and Lin Lin