2D materials may not be as far-fetched as we imagined, and it’s a bit of a clue—not only did Intel demonstrate at IEDM last year that it applied 2D materials to GAAFET transistors; but also BEOL of some old processes is already trying 2D materials, mainly MEMS, radio frequency, and photoelectric transceiver. After all, these devices may not have as stringent quality requirements as transistors.
What makes 2D materials become favor?
Several important nodes in the development of modern transistors, whether in terms of materials or major changes in structure, are nothing more than because transistors are always facing various problems in the process of shrinking. Various changes in engineering are aimed at improving the performance and efficiency of transistors, including reducing leakage current and increasing switching speed.
For example, in the era of Planar planar transistors, trying to stretch (strain) the channel of the transistor, silicon is laid on a layer of SiGe (silicon germanium). As the atoms of the upper silicon layer align with the SiGe layer, the bonds between the atoms of the silicon layer are stretched. In other words, the farther apart the silicon atoms are, the less atomic forces interfere with the movement of electrons. Then, the electron mobility can be greatly increased, allowing transistors to switch faster.
In terms of device structure, the transformation of the FinFET structure in the 20nm period, and the imminent transition of the device structure to GAAFET in the 3nm/2nm period are well known. The nature of device structure changes is also similar, that is, as the device size continues to shrink, the difficulty of controlling the current continues to increase, so some innovations must be made in the structure.
After 2nm, new materials are still needed to seek new breakthroughs, because as devices continue to shrink, the electrical mobility in materials such as silicon and germanium will drop significantly. And the next solution may be 2D materials.
At last year’s IEDM conference, Intel talked about several future-oriented technologies in its speech, including an advanced packaging process called QMC, a 3D stacked FeRAM, and 2D materials. What Intel showed was the use of 2D materials on the channel of GAA structure transistors to achieve low leakage and more ideal switching performance. It is said that it is a big step towards the vertical stacking of transistors in the future.
Why 2D materials become another development trend?
The so-called “2D” in “2D materials” should be the opposite of 3D crystals; 2D is a layer as thin as the atomic level. The most well-known 2D material should be graphene which is a single layer of atomic matter arranged in a hexagonal grid. However, graphene has no bandgap.
The bandgap refers to the energy required to excite an electron from the valance band, which cannot conduct electricity, to the conduction band, which can conduct electricity. For transistors, there are two states of on and off, so a clear band gap is needed to distinguish them.
For transistors, the favored 2D material is TMD (transition-metal-dichalcogenide, two-dimensional transition metal sulfide), such as MoS2 (molybdenum disulfide), WS2 (tungsten disulfide) , WSe2 (tungsten diselenide). TMD serial materials have a relatively ideal band gap and good electron mobility when the channel thickness is <5nm.
Compared to the legendary 1D carbon nanotube (carbon nanotube, CNT), which is still far away and faces huge technical challenges, 2D materials will be easy to achieve. At least for now, the manufacture of 2D materials will be relatively easier.
The 2D material (MoS2) in Intel’s publicity demonstration is only 3 atoms thick and is applied to the channel of the GAAFET transistor to replace silicon. On imec’s roadmap, the legendary 3D folded transistor CFET structure also has a similar solution, in which both nFET and pFET channels are based on single-layer TMD.
What challenges are faced in 2D materials nowadays?
Since 2D materials have many advantages, why haven’t they been fully popularized? 2D materials are generally obtained by CVD (Chemical Vapor Deposition) growth. In addition, more recent research can also use the ALD (atomic layer deposition) method. According to the control of substrate and related variables, 2D material growth can be made into single layer or multilayer.
For example, the most mature single-layer graphene is grown by CVD on copper foil or thin film substrate. However, there are many variables in growth, and the consistency of wafer-to-wafer is difficult to achieve. Existing technology can lead to some production defects, such as grain boundaries (grain boundaries, a kind of surface defect, which means that the periodic arrangement law in the crystal structure is broken).
The electron mobility of CVD graphene with crystal defects will drop greatly. If the variation of wafer or material is very large, metrology/inspection will also become difficult. This is also the reason why the graphene market has always been small.
In addition to growth, another technical difficulty is the transfer of 2D materials. Because the growth of 2D materials is generally completed on substrates such as copper or sapphire at a high temperature of >600°C, then the grown 2D materials need to be transferred to the final wafer. It is said that under the existing CMOS manufacturing process, the method of transferring 2D materials to silicon device wafers is still quite inefficient.
The conventional 2D material transfer technology includes wet etching the copper substrate, and then transfers the 2D material to the target substrate with PMMA (polymethyl methacrylate). However, during this process, PMMA will remain on the surface of graphene, which will also affect the electrical properties of the material. Current 2D material transfer methods are feasible for certain types of applications (such as sensors/displays), but cannot meet the needs of CMOS production in terms of quality, throughput, etc.
Another way of thinking is naturally that 2D materials are grown directly on silicon. The problem is how to obtain a low-temperature, high-quality growth solution. The ALD method mentioned above can be implemented at lower temperatures, but throughput is a big problem. In addition, it is said that methods such as MOCVD (metal organic chemical vapor deposition) have various defects that are also problems, mainly organic pollutants, sulfur vacancies, etc.
Therefore, the high-quality growth process and the high-throughput transfer process may still be separated and decoupled—this is also a common idea in terms of variable control and process optimization. In this way, growth and transfer can be performed asynchronously to achieve greater production capacity. Therefore, SemiAnalysis experts generally prefer to adopt the transfer solution instead of growing on the original silicon base, because this has advantages in heterogeneity, stacking, and configurability.
In general, the future value of 2D materials to the industry is unquestionable, but there are still major challenges in the mass production of 2D materials. Most of the above studies use wet transfer technology to transfer 2D materials from the growth substrate to the final wafer, which hinders mass production due to problems such as polymer residue and low throughput. However, judging from the frequency and quantity of IEDM’s annual publications on 2D materials, 2D materials are still very clear as the path for the future semiconductor industry.