Transistor Function and the Limits of Moore's Law

As transistors approach 7-5nm and below, quantum tunneling becomes a significant problem. In classical physics, electrons cannot pass through energy barriers, but quantum mechanics allows electrons to tunnel through barriers that would normally block their flow. This phenomenon causes leakage current even when the transistor is supposed to be in the 'off' state, wasting energy and degrading performance.

Quantum tunneling is a quantum mechanical phenomenon where particles can pass through barriers that they should not be able to cross according to classical physics. In semiconductor manufacturing, as transistors become smaller, the insulating barriers between components become so thin that electrons can 'tunnel' through them, creating leakage current. This leakage current cannot be ignored and represents a fundamental limit to continued miniaturization of electronic components.

Quantum tunneling is a quantum mechanical phenomenon where particles like electrons can pass through potential barriers even when they don't have enough energy to overcome the barrier according to classical physics; instead of bouncing back, the electron's wave function extends into and beyond the barrier as an exponentially decaying evanescent wave, giving it a non-zero probability of appearing on the other side, which has important applications in explaining radioactivity and building scanning tunneling microscopes.

Quantum tunneling is a phenomenon where particles can pass through energy barriers that would be impenetrable according to classical physics. This occurs because the wave function can 'leak' through barriers, and whether a particle successfully tunnels depends on probability. This is a purely quantum effect with no classical equivalent.

A tunneling microscope works by bringing a very sharp tip (containing a single atom) close to a sample surface. A voltage is applied between the tip and sample. Current flows even when the tip doesn't physically touch the sample due to the quantum tunneling effect, where particles can pass through potential barriers. The probability of tunneling depends exponentially on the distance between tip and sample. When viewed at sufficient magnification, surfaces appear not as continuous but as collections of atoms. Atoms themselves are not solid spheres but are 'smeared' in space due to quantum effects. The tunneling current is extremely sensitive to distance, allowing the microscope to trace surface topography. The microscope can image CD discs showing pits and lands that store digital information, and can achieve atomic resolution showing individual atoms.

As silicon-based semiconductors reach their physical limits, researchers are exploring alternative materials including silicon carbide, graphene, gallium nitride, carbon nanotubes, and 2D materials like transition metal dichalcogenides (TMDs) such as molybdenum disulfide. Additionally, alternative computing technologies are being investigated, including Josephson Junction transistors for quantum computing, probabilistic computing, photonics, and biological computing, representing potential pathways beyond traditional silicon-based electronics.

Traditional silicon semiconductors are being replaced by gallium nitride (GaN) and silicon carbide (SiC) for power conversion. GaN excels at switching power quickly with very little waste, while SiC handles extremely high voltages and temperatures efficiently. These materials are physically superior to old-school silicon when handling serious power loads—they switch faster, waste less energy on heat, and can handle higher voltages in smaller packages. Companies specializing in both materials can build power chips for every stage of the electricity chain inside data centers.

Gallium nitride (3.4 eV direct band gap) enables blue light technology, earning the 2014 Nobel Prize for blue LEDs. Lattice-matched alloys span the visible spectrum for multicolor detection. Rock salt structures (NaCl-type) provide diverse properties from UV optics to magnetic storage barriers. Perovskites (ABX₃) offer ferroelectricity and 43% theoretical solar efficiency. Van der Waals materials (graphene, transition metal dichalcogenides) enable 2D semiconductors with tunable band gaps across infrared to UV. These emerging classes expand semiconductor possibilities beyond traditional bulk materials, representing frontier research areas with transformative potential for future electronics and optoelectronics.

Carbon nanotubes represent sheets of graphene rolled into cylindrical structures and exist in both metallic and semiconducting forms. They offer three times higher electrical conductivity than silicon. The first carbon nanotube chip containing 14,000 transistors has already been created, though its process node is approximately one micrometer (equivalent to 1980s silicon processes). Future developments aim to reduce nanotube sizes to create faster and more efficient chips.

Two families of extreme materials are transforming electronics: gallium nitride (wide bandgap semiconductors) and two-dimensional materials (graphene, molybdenum disulfide). Gallium nitride enables efficient power conversion at high voltages and frequencies, addressing the critical problem that 40-50% of microprocessor energy is lost as heat in interconnects. Two-dimensional materials offer exceptional properties including ultra-high carrier mobility and outstanding electrostatic control due to extreme thinness. These materials can be directly integrated onto fully fabricated silicon wafers using straightforward processes. For RF applications, gallium nitride achieves ~10x silicon performance, diamond offers 2-3x improvement, and vacuum represents the theoretical limit with 20x potential. Nano-vacuum tubes use gallium nitride field emitters with 10-50 nm tips to extract electrons for high-frequency operation, combining vacuum tube principles with microfabrication technology.

Three emerging paradigms challenge traditional computing: neuromorphic computing emulates biological spiking neural networks with Intel's Loihi and SpiNNaker achieving million-neuron scales; quantum computing promises revolutionary capabilities for physics, cryptography, and machine learning despite coherence and error correction challenges (IBM's roadmap projects thousands of qubits); optical computing leverages light propagation through silicon waveguides with Mach-Zehnder Interferometers for theoretically unlimited-speed computation. Each paradigm represents fundamentally different approaches requiring new programming models and mathematical frameworks. Neuromorphic computing requires understanding temporal spike dynamics; quantum computing demands deep quantum mechanics expertise; optical computing faces manufacturing challenges at nanometer scales. These technologies remain in early research or limited deployment phases, with practical viability years away, but represent fundamental shifts in how computation might be performed in the future.

Russia has successfully developed a 50-kuubit universal quantum computer using ion-based technology and created the domestic neuromorphic processor 'Alтай', positioning the country as a leader in emerging quantum and neuromorphic computing technologies that promise revolutionary advances in computing power and efficiency.

Intel Labs pursues transformative computing paradigms including neuromorphic and quantum computing. Neuromorphic computing emulates brain processing with two primary application paths: small configurations for energy-constrained environments enabling on-the-fly learning near sensors (such as vision sensors for autonomous vehicles), and larger clustered configurations for constraint satisfaction problems or similarity searches across large datasets. The Loihi chip demonstrates neuromorphic principles with 120,000 neurons at 14nm, scaling to 100 million neurons consuming approximately 300 watts. Scaling faces challenges primarily from SRAM memory costs rather than interconnect limitations. Quantum computing research focuses on spin qubits due to practical requirements for millions of fault-tolerant qubits with quantum error correction, aligning with Intel's core competence in building small, scalable devices. The Horse Ridge controller addresses cryogenic control electronics challenges. Research focuses on enabling spiking neural networks to learn with less data and adapt more quickly than traditional training approaches.

Two revolutionary computing technologies are converging to create a new framework for artificial intelligence. Neuromorphic computing draws inspiration from biology and neuroscience, modeling neural networks after how the human brain functions. Within a decade, this could enable 100 billion programmable neurons and 1 trillion synapses, creating virtual entities downloadable on-demand. Simultaneously, quantum computing offers a fundamentally different architecture, potentially achieving speeds one hundred million times faster than current supercomputers. Together, these technologies promise to collapse millennia of processing into mere minutes, enabling capabilities that transcend traditional bit-based computation by harnessing hyperdimensions, time, and space.

To transcend biological limitations, researchers are combining evolved neural algorithms with quantum hardware. Neurons can be replaced with atoms (different firing states correspond to electronic states), and synapses with photons (enabling communication through photon emission/absorption). This creates quantum neuromorphic systems including quantum associative memories with enhanced capacity, robustness, and recall, as well as quantum optimizers solving problems through fundamentally new mechanisms. This hybrid approach opens entirely new computational paradigms that evolution could never discover.