Research on graphene and two-dimensional materials (2DM) in science and engineering has been going on for 15 years since it was first proposed. A large amount of available data and high-performance device demonstrations leave no doubt about the potential of 2DMs for applications in electronics, optoelectronics, and sensing. So where are the integrated chips and products using 2D materials? In this article, we answer this question by summarizing the main challenges and opportunities currently hindering the application of 2D materials.
Research on graphene and two-dimensional materials (2DM) in science and engineering has been going on for 15 years since it was first proposed. A large amount of available data and high-performance device demonstrations leave no doubt about the potential of 2DMs for applications in electronics, optoelectronics, and sensing. So where are the integrated chips and products using 2D materials? In this article, we answer this question by summarizing the main challenges and opportunities currently hindering the application of 2D materials.
Manufacturing Technology
In our opinion, this issue can be concluded by comparing the level of manufacturing readiness of 2D materials with that of standard semiconductormaterials. The current lack of solutions for introducing 2D materials into silicon (Si) semiconductor fab lines are "unit processes" that integrate 2D materials with silicon complementary metal-oxide-semiconductor (CMOS) chips at the back or front end of the line. At present, the deposition and growth of 2D materials can be suitable for wafer-scale, but defects and contamination are not yet in line with the needs of large-scale production.
In general, high-quality materials typically require higher process temperatures, which exacerbates the complexity of direct wafer growth, but also makes transfer technology more desirable. In theory, wafer bonding technology could solve this problem, but it is clear that this technology is not at a perfect manufacturing level. At the device level, the challenge for 2D materials is the control of dielectric and contact interfaces. The self-passivating nature of 2D material surfaces requires fabrication methods (e.g., by atomic layer deposition) to achieve dielectric deposition. The resulting non-ideal interfaces limit device performance compared to the best laboratory demonstrations using crystalline 2D insulators (e.g., hexagonal boron nitride).
The same is true for electrical contacts of 2D materials, which only partially meet industrial specifications and are not yet up to manufacturing standards. For 2D materials, pulling or etching materials with high selectivity for the underlying layer is particularly challenging because it requires atomic-level precision, which can only be achieved by specific chemical reactions and dedicated atomic layer etching equipment. The process of developing suitable processes for production is lengthy and tedious because of the wide range of potential 2D materials and their combinations. In general, etch chemistry and other physical process parameters are strongly dependent on the specific situation and each requires a separate solution. Doping, the replacement of atoms in the lattice, is a standard and critical technique required for silicon, and it relies on statistical distributions. In the field of 2D materials, the term "doping" is commonly used to describe the charge transfer from defects or molecular adsorbs in their vicinity to 2D material layers.
Precise and long-term stable control of this "effective doping" is still a challenge, but so is conventional doping. As silicon technology shows, two-dimensional crystal atoms need to be replaced in a deterministic way. Addressing these critical manufacturing bottlenecks is the clear goal of the European 2D materials pilot line. The co-integration of 2D materials with silicon CMOS technology will dramatically improve chip functionality and enable 2D material applications to emerge in the order of their device complexity.
As shown in the figure, materials including copper interconnect, high-k metal gate dielectrics, and FinFETs, as well as architectural innovations, have been adopted over the years to continue to drive Moore's Law forward. But in the future "More Moore" and "continuation of Moore's Law" scaling may require thinner nanosheet transistors, and two-dimensional materials are considered ideal candidates. High-frequency electronic devices integrated on CMOS chips through "CMOS + X" integration, e.g., through "More Than Moore" "beyond Moore" field sensors or electronic devices, substantial performance and functional gains are expected. With the optoelectronic properties of two-dimensional materials, photonic integrated circuits can improve overall system performance and data processing capabilities, and open up spectral sensing applications. Memory computing or memristors make future neuromorphic computing applications possible, and 2D materials may be well suited for integration with silicon CMOS. Even in the laboratory, 2D quantum technology is the least mature, but it will benefit from all the expected achievements as 2D materials enter the semiconductor processing line.
In addition, two-dimensional materials are expected to become the X-factor for CMOS. In the era of heterogeneously integrated scaling, new materials offer unprecedented performance in 3D chip stacks. It is important to note that in the classic "Moore's Law" era, the unit of the Y-axis was "log2". However, in the era of heterogeneously integrated scaling, we suggest labeling it "Performance (A.U.)". Because the performance improvement will be application-specific. It will be determined by (a combination of) factors such as power consumption and efficiency, pattern recognition capability, sensor fusion, etc., which will lead to the creation of some arbitrary units due to the diversity of functions and underlying technologies.
More Moore
In general, advanced semiconductor technology nodes can be achieved by increasing the complexity of the integrated architecture and the co-optimization of the overall STCO design with the system architecture. At the transistor level, leading semiconductor manufacturers are moving from FinFETs to stacked nanosheet CFET architectures to enable the most advanced CMOS technology nodes. Currently, these nanosheet devices are still based on silicon channels, and various structures of such nanosheets are being used to evaluate future technology nodes, such as the so-called "fork sheet" design, which allows for tighter np spacing or the integration of p- and n-type nanosheets with each other. However, further scaling of the channel length requires reducing the channel thickness by similar factors to ensure sufficient electrostatic control to suppress short-channel effects. Reducing the wafer thickness to the desired value increases the charge scattering at the interface (charge scattering) and leads to a dramatic decrease in carrier mobility in the channel, which disrupts the device performance. Two-dimensional semiconductors would be the ultimate version of nanosheets because they are self-passivated in the third dimension and the carrier mobility is not strongly affected by surface scattering. As a result, the mobility remains high even under thickness constraints.
In principle, this property could enable the practical scaling of multiple technology nodes and motivate the semiconductor industry to eventually consider replacing silicon with 2D materials as transistor channel materials for future advanced nodes. However, this issue brings us back to the fundamental technical and scientific challenges associated with 2D integration. Notably, it is particularly important to identify a suitable gate oxide stack and to find low contact resistance schemes (contact schemes). The former is necessary to maintain the properties of the 2D material and provide adequate electrostatic control while reducing gate leakage current. Two-dimensional hexagonal boron nitride (hBN) has been widely used to demonstrate high-performance devices based on 2D materials, but its bandgap and band order dictates that only one or two single molecular layers can achieve sufficient electrostatic control. This additional boundary condition leads to intolerable device leakage, so other solutions must be found. To maintain the benefits of channel materials in integrated circuits, low contact resistance is required, as high contact resistance dominates and severely limits the performance of integrated devices.
Recently, metal-induced gap states and spontaneous formation of degenerate states in MoS2 have been reported to greatly reduce the contact resistance of MoS2 through the use of semimetallic bismuth. However, more such breakthroughs are needed to reveal and fully utilize the potential of monolayer transistors in CMOS circuits, to reinvigorate the miniaturization rate of transistors and to perpetuate Moore's law.
More than Moore
These types of applications may enter the market first because they are multifaceted. However, they are usually very specific, so defects and large device variations in device performance can be tolerated.
Two-dimensional materials are well suited for gas, chemical, and biological sensor devices due to their inherent high surface/volume ratio and multifunctionalization properties. Thus, any charged particle or molecule in the vicinity of some 2D layered materials can alter their electrical conductivity. However, ideally, 2D materials are chemically inert, which means that the lack of chemical activity will greatly enhance the reactivity of 2D material-based sensors. Therefore, precise defect control is essential to ensure device sensitivity. In addition, the selectivity of the sensor is crucial. It can be achieved by surface functionalization or by forming arrays of different sensors to simulate complex biological systems such as the nose, and combinations of 2D materials with different sensor "fingerprints" can be used with machine learning algorithms for sensor reading.
MEMS typically rely on mechanically movable parts on a chip. Two-dimensional materials with excellent mechanical properties can produce ultra-thin films that translate directly into piezoresistive and opto-mechanical readout methods with extremely high sensitivity, providing efficient signal transmission for MEMS. MEMS applications based on 2D thin films include pressure sensors, accelerometers, oscillators, resonant mass sensors, gas sensors, Hall effect sensors, and thermal radiometers.
Two-dimensional materials offer a range of advantages over existing optoelectronic and photonic technologies, especially outside the spectral range that silicon materials can handle. But even so, many 2D materials have a direct bandgap advantage over silicon when it comes to photoemission. Semimetallic and small bandgap materials, such as graphene, platinum diselenide, or black phosphorus, open up the infrared (IR) regime to compete with expensive III-V semiconductor technologies. While the two-dimensional properties translate into lower absolute absorption in the vertical direction, the combination with IR-sensitive absorption layers brings higher detector responsiveness.
Photonic Integrated Circuits
Photonic integrated circuits are considered the ultimate performance enabler for data transfer on or between computer chips, and connecting them to silicon-based devices via optoelectronic converters at extremely high data transfer rates is a key application technology. Two-dimensional materials, particularly graphene, can be transferred to photonic waveguides and provide broadband optical detection and signal modulation. By eliminating the need for epitaxy, 2D-based photonic integration allows the integration of active device components with silicon optical devices, but also with passive non-crystalline waveguide materials, such as silicon nitride, which opens the door to complex photonics applications on CMOS. The fact that some 2D materials, such as platinum diselenide, can also be grown directly conformally at temperatures below 400°C is a clear advantage in the quest for photonic integrated circuits in conjunction with silicon CMOS technology. With the potential to integrate 2D light sources, 2D materials could eventually enable the convergence of electronics and photonics and bridge the spectrum over the terahertz gap.
Neuromorphic Computation
Neuromorphic computing aims to provide brain-inspired computing devices and architectures for artificial intelligence applications to enable energy-efficient hardware. At the device level, requirements for neuromorphic computing include merging memory with logic to implement memory computing and memory device features that emulate synapses and neurons. The former can already be implemented with conventional memory technologies, while the latter translates into threshold switches and non-volatile memristors with a wide range of programmable resistive states. Although this technology is relatively new, 2D memristors have shown promising properties, including Joule-scale switching energy, sub-nanosecond switching times, dozens of programmable states, and wafer-scale artificial neural network prototypes that enable sensor systems and edge computing applications, such as through pre-processing of sensor data or on-chip sensor fusion. In addition to neuromorphic computing, 2D memories have been shown to provide a wide range of non-computational functions, including physical unclonable functions for security systems, and RF switching for communication systems.
From a scientific point of view, resistive switching phenomena in two-dimensional devices arise due to ion transport, defect formation or phase transition effects. Despite these fundamental aspects, two-dimensional memristor switching is still a topic that is increasingly discussed and studied. At the device level, a fundamental challenge is to increase the number of resistive switches, the so-called durability, which requires further research on the underlying mechanisms of aging effects. Similarly, improving the homogeneity of materials will be critical in order to achieve arrays of massively connected devices capable of mimicking the hyperconnectivity and efficiency of the brain. It is exciting to note that over 12 2D materials have demonstrated memory effects to date, and this number is likely to continue to grow in the coming years. As a result, algorithms are increasingly needed to guide experimental studies and optimize memory elements for maximum performance.
Quantum Technology
The various properties of two-dimensional materials and associated van-der-Waals van der Waals heterostructures also make them highly tunable quantum materials for spintronics and future quantum technologies. Two-dimensional material systems enable not only artificial states of quantum matter but also many of the promises of solid-state quantum computing as a key component of quantum communication circuits or to allow interesting quantum sensing schemes. Indeed, 2D materials are a promising platform for quantum dots in the solid states, such as the long-recognized topological quantum computing elements, and coherent sources for single-photon emitters.
Quantum computing based on semiconductor quantum dots (DQs) uses individual spin states of captured electrons. Among other things, it relies on long spin coherence times that play an important role in the host material, which makes graphene a very interesting material for spin quantum dots because of its weak spin-orbit coupling (carbon atoms are very light) and weak hyperfine coupling (carbon 12 is spinless nucleus). With the progress of research on single-electron confinement in gate-controlled quantum dots (QDs), the first spin quantum bits are forthcoming. The possibility of making spin quantum bits in 2D materials will also allow evaluating additional valley degrees of freedom as possible quantum bit states; interesting proposals for the valley and spin valley quantum bits exist.
In addition, stationary quantum bits in two-dimensional materials can be coupled to photonic quantum bits realized in single-photon emitters (SPEs), for example in nearby wide-bandgap hexagonal boron nitride or semiconductor transition metal dihalides (e.g., WSe2). In these two-dimensional materials, SPEs have been shown in recent years to open the door to distributed quantum networks in which photonic quantum bits can act as interconnections, allowing distant stationary quantum bits, such as spin quantum bits, to be entangled. Such robust, bright, indistinguishable single-photon emitters are essential for creating photonic (flying) quantum bits to enable efficient quantum communication.
Furthermore, two-dimensional heterostructures are promising materials for topological quantum computing, where quantum states may be better (i.e., topologically) protected against disorder than in standard quantum computing. For example, modulation of quantum anomalous Hall insulators or graphene to tilted antiferromagnetic quantum Hall phases in combination with s-wave superconductors is a promising platform for applications in topological quantum computing. In short, these advances make two-dimensional materials and their heterostructures an exciting platform for future quantum technology applications in many ways.
Conclusion
2D materials offer superior performance advantages at the device level compared to existing technologies and can also be easily integrated with silicon CMOS technology, making them a prime candidate for extended functionality on silicon chips (also known as "CMOS + X"). We believe that 2D materials will increasingly become an x-factor in future integrated products, depending on the target application and that bottlenecks in 2D material-based heterogeneous electronics will be broken through to reach the required level of mass manufacturing.
Related News
1、MediaTek, Qualcomm announce joining Russia sanctions
2、Automotive chips rose across the board!
3、Apple M1 Ultra -- The Technology Behind the Chip Interconnection
4、Foxconn Announces Investment of $9 Billion to Build A Chip Factory in Saudi Arabia
5、Japanese Companies Increase Investment in Power Semiconductors