What new can appear in the semiconductor industry due to the introduction of multibeam electron lithography? Innovative technology proposed by the Dutch company Mapper Lithography is slowly but surely approaching the level of industrial application - not without the help of Rosnano, which owns a significant share of the company. What new can be done in semiconductor production using multipath electron lithography? Let's get a look.
Introduction
The last fifty years, the development of semiconductor technologies followed the so-called
"Moore's Law" , which was formulated by Gordon Moore in 1965, quite well: "The number of transistors placed on an integrated circuit chip doubles every 24 months." A direct consequence of Moore's law is the reduction of critical sizes (critical dimensions - CD) of transistors, they are called technological processes (technology nodes) used for the manufacture of integrated circuits. Since the formulation of Moore’s law to this day, the CD has decreased 5,000 times: from 50 μm to 10 nm. Although the death of Moore's law has been predicted for quite a long time (somewhere starting at 90 nm), the size reduction continues to this day. On the other hand, it is obvious that this cannot continue indefinitely - the physical limit will come - the size of the atoms is finite.
However, the downward movement continues (often referred to as more Moore). Since the critical process of semiconductor technology is lithography, the main efforts to reduce the size are concentrated there, a vivid example is the Extreme Ultra Violet (EUV)
machine , developed by the flagship photolithography, the Dutch company
ASML . The wavelength of this technology is 13.5 nm (while the wavelength of modern photolithographic machines is 193 nm), which will allow printing CDs for technologies less than 45 nm directly (currently, multipatterning is used for such technologies, which includes several operations of lithography and etching).
Another approach is to retreat from simply reducing the size and adding various new technologies to the standard process technology (this is usually called more than Moore). Examples of such technologies are new materials (for example, dielectrics with low or high dielectric constant instead of silicon oxide, copper instead of aluminum, germanium instead of silicon, etc.); new architecture of transistors (tense channels, three-dimensional gates, etc.); new system solutions (multi-core processors, combining mathematical and graphics processors on a single chip, etc.).
If the EUV technology from ASML is an obvious example of following Moore's law, then the maskless electron lithography technology from another Dutch company
Mapper Lithography will allow it to be used in both cases: more Moore and more than Moore. In today's article we will not dwell on the advantages of high-resolution electron lithography (which is a surface advantage over optical), but argue that this technology can bring innovation to the process of manufacturing integrated circuits.
What will allow to produce multipath electron lithography
By itself, electron lithography is well known and not something special, it is widely used, for example, in the manufacture of masks for photolithography. The main drawback of electron lithography is the depressingly slow speed of work - it takes about a month to expose a 300 mm plate. This is not acceptable for production, where conventional photolithography exhibits plates with a capacity of more than a hundred per hour.
300 mm plate after exposure on the Mapper Lithography and development machine.Multibeam electron lithography uses 13,000 electron beams at the same time, each beam is individually controlled, plus it is further divided into 49 sub rays. Using 13,000 rays, you can print about 10 plates with a diameter of 300 mm per hour. The stiness of such a machine will be 2-3 times less than a modern photolithography setup (a 193 nm scanner with a water layer).
The main advantage of multipath electron lithography is the absence of masks - the pattern on the semiconductor wafer is transferred directly from the computer. If classic photolithography can be compared with film photography — printing a large number of prints from one negative, then multipath electronic lithography can be compared to digital photography — printing prints directly from a computer on an inkjet printer. In the first case, we get high performance with low variability (it is easy to reproduce a lot of prints, but it is difficult to change the negative), in the second case, we get lower performance, but high variability (correct the file on the computer will not work). Another good analogy is the casting and printing on a 3D printer.
Lines with a half-period of 42 nm after electron lithography and etching. To obtain the same structure using classical photolithography, several exposure / etching operations are required.It is worth noting that the key elements of the machine - electronic lenses - are made in Russia in a
small MEMS factory built specifically for this purpose with the support of Rosnano.
Given that the manufacture of masks (photo masks) for modern technologies (~ 20 nm) is a long (several months) and expensive (millions of dollars) process, let's see where the maskless lithography technology can be used.
Rapid prototyping of new products
How is the development of new semiconductor products today? A company that wants to launch a new product on the market first produces a test photo pattern with several options for a future product — several months and several million dollars — makes a number of test chips, selects the best design and orders the final photo pattern according to the best design — these are a few more months and several million dollars.
How can the development of new products occur when using the MEL installation? A test batch of chips with a new product may contain hundreds or thousands of variants of a new chip, and no extra time is required to make it to make photo masks, nor additional costs. That is, the development of new products will be faster (by several months), cheaper (by several million dollars) and more qualitatively (more variants of the new chip).
In the manufacture of small series MEL can be used in subsequent mass production, for large-scale production, you can order a photomask with the best design and print the chips already from it.
Making small series of chips
CubeSat. Source: Wikipedia CC BY 1.0 , LinkIf you want to make a chip with a million copies - no problem, the cost of the photomasks will be spread on a huge number of chips and each chip will not cost so much. But if you need a hundred or a thousand chips? For example, you want to push all the electronics of a nanosatellite into one chip of your own unique design - the cost of such a chip will be huge, since the cost of a photomask (millions of dollars) will be divided into a small number of chips. However, if you do not need to make a photo mask, then no one bothers to make a small number of chips - if the rest of the technology, with the exception of photolithography, does not change much, then the cost of the chip will not change much - on a 300 mm plate one chip will cost from tens to hundreds dollars depending on the size.
Making unique chips
If you make a small series of unique chips using conventional photolithography, although very expensive, but in principle it is possible, then it is impossible to make each chip unique in principle. What may need unique chips? They can be used for security purposes (protection variability is not created at the software level, but at the hardware level) or for identification purposes (a unique chip is too difficult to fake). Quite a lot of customers became interested in making unique chips, so the Dutch created a
special website from Mapper Lithography.
Extending the life of aging factories operating on 200 mm plates
200 mm factory. Source: InfineonCurrently, most semiconductor manufacturing (about 60%) uses 300 mm wafers and related equipment. However, the share of semiconductor production on 200 mm wafers, although decreasing, is still more than 20%. Such factories are less technologically advanced than 300 mm and usually can not produce plates on technological processes less than 90 nm. The key equipment that determines the technical process is the installation of photolithography, which is also the most expensive. In principle, the remaining 200 mm process line can pull and more advanced process technology (45 nm - 65 nm), but it all comes down to lithography, and replacing it with a more advanced one (this is 300 mm machines) will cost too much. In this case, multipath electron lithography can help - the equipment costs several times less than a modern photolithography machine, however, it will allow manufacturing plates using more advanced technologies, even if they are not very large batches, which will prolong the life of the obsolete 200 mm factories.
Manufacturing large light-sensitive matrices
Photomatrix. Source: Wikipedia By Filya1 - Own work, CC BY-SA 3.0 , LinkAs is known, the physical size of the matrix has a greater influence on the image quality than the number of megapixels. The size of the matrix is determined by the maximum field of view of the photolithographic unit (at one time the modern photolithographic unit prints the pattern corresponding to the field of view, then moves to the next section, prints the same pattern, etc.). Today, the maximum matrix size is approximately 20 mm x 20 mm, which corresponds to the field of view of scanners, which is unlikely to increase in the near future. In fairness, I note that ASML has the technology of stitching some fields of the scanner into a single chip, but with it everything is not so simple.
Since the principle of operation of multipath electron lithography is similar to an inkjet printer printing a picture strip by strip from edge to edge, and not to a photo enlarger printing the picture entirely step by step (as in photolithographic scanners), the size of the picture obtained on a multi-beam electron lithography setup is limited only by size semiconductor wafer, which is transferred to the figure (at the moment it is 300 mm, in the future may be 450 mm. But this is not accurate.). Thus, using multipath electron lithography, it is theoretically possible to create photomatrices with the size of a semiconductor plate (300 mm in diameter). It is clear that it is not necessary for the mass consumer, but, for example, for space telescopes or some other applications where image quality is important, and size and price play a secondary role, such matrices will be irreplaceable and some companies are greatly interested in this technology.
Conclusion
Multibeam electron lithography will open a new chapter in semiconductor manufacturing. It is similar to 3D printing versus casting and digital photography + an inkjet printer versus photographing on film and printing from a negative.
I have heard many times the opinion that multipath e-lithography cannot compete with classical photolithography (including EUV) and Mapper Lithography cannot compete with ASML, from which it was concluded that MEL is doomed to failure. If I agree with the first part of this statement, then not with the second. If you look at the story with MEL and classical photolithography a little on the other side, MEL can be compared with a helicopter, and classical photolithography with a main plane. It seems that both technologies transport passengers and cargo by air, but at the same time there is a huge difference between them. If you need to transport several hundred people across the ocean, then your choice is the main liner. And if you deliver a shift of oilmen to the offshore platform, the plane will no longer help you. Yes, the production of helicopters will never reach the scale of production of aircraft and will not compete with them. But to build a successful business in the production of helicopters is quite possible. So over time, multipath electron lithography will occupy its niche in semiconductor manufacturing, just as helicopters have taken their place in air transport.