Bismuth contacts can inject current across just 2 nanometers
At the frontier where matter thins to a few atoms and human ingenuity strains against the limits of the measurable, researchers have done something previously thought impractical: they have directly seen, rather than merely inferred, how far electrical current actually penetrates into a two-dimensional transistor's contact region. Working with bismuth-contacted molybdenum disulfide and a scanning tunneling microscope operated on a live device, a team published in Nature has placed that distance at roughly 2 to 3 nanometers—a finding that redraws the boundary of what miniaturization can still achieve. The result does not end the challenge of shrinking electronics, but it dissolves one of its most stubborn uncertainties, replacing decades of theoretical estimation with direct observation.
- Silicon's physical limits have forced the semiconductor industry to gamble on two-dimensional materials, but a critical unknown—how small metal contacts can actually be—has blocked confident progress toward sub-10 nm devices.
- Previous estimates of carrier transfer length ranged wildly from 9 to 45 nanometers depending on assumptions, leaving engineers designing next-generation chips on theoretical quicksand.
- By cleaving a live bismuth-MoS2 transistor inside an ultrahigh-vacuum chamber and scanning its cross-section with sub-nanometer precision, the team replaced extrapolation with direct measurement for the first time.
- The measured transfer length of 2–3 nm meets the specifications for a 1-nanometer technology node, meaning contact sizes once considered impossible are now experimentally validated as achievable.
- A parallel test on p-type tungsten diselenide revealed a transfer length of ~211 nm, exposing a stark asymmetry that must be resolved before practical complementary circuits can be built at this scale.
The transistor has been shrinking for half a century, but silicon is approaching a hard physical boundary. Two-dimensional materials—semiconductors just a single layer of atoms thick—have emerged as the most promising path forward, offering sharper electronic control and lower power at scales silicon cannot reach. Yet a foundational question has lingered without a clean answer: how small can the metal contacts that inject current into these devices actually become?
For years, the answer came from theory. Estimates of the so-called carrier transfer length—the effective depth over which current flows from metal into semiconductor—ranged from 9 to 45 nanometers depending on the material and the model. That uncertainty is tolerable in a lecture hall but dangerous on a fabrication floor. A new study in Nature replaces those estimates with direct evidence.
The team deposited bismuth contacts onto monolayer molybdenum disulfide, then cleaved the assembled device inside an ultrahigh-vacuum chamber to expose a pristine cross-section. With the transistor actively operating, they scanned perpendicular to the metal-semiconductor interface using cross-sectional scanning tunneling microscopy—a technique capable of resolving electronic structure at sub-nanometer scales. They watched the conduction band edge shift sharply in the first few nanometers from the contact, then flatten. Fitting that decay to an exponential curve, they extracted a transfer length of 2.0 to 2.9 nanometers.
Bismuth was not a random choice. Its electronic structure aligns unusually well with molybdenum disulfide, producing a nearly barrier-free interface and a contact resistance of just 70 ohm-micrometers—among the lowest on record. Together, the short transfer length and low resistance confirm that bismuth-contacted 2D transistors can be scaled to contact sizes well below 10 nanometers, satisfying the requirements of what the industry labels the 1-nanometer technology node.
The methodology itself demanded rigor. Scanning tunneling microscopy does not read carrier density directly; it reads band structure shifts that must be disentangled from artifacts like tip-induced band bending. The team ran simulations and ruled out competing explanations—defects, dopants, instrumental distortions—before attributing the observed shifts to genuine carrier injection.
When they applied the same approach to a p-type system—tungsten diselenide with palladium and antimony contacts—the transfer length ballooned to roughly 211 nanometers. The contrast is sobering. N-type contacts in 2D transistors can now be engineered at the nanometer scale; p-type contacts cannot, and any practical circuit requires both. The work establishes a powerful new tool for probing metal-semiconductor interfaces and offers the clearest experimental evidence yet that contact scaling has room left to run—but it also maps, with unusual precision, exactly where the next hard problem begins.
The race to shrink transistors has hit a wall. Silicon, the material that powered computing for decades, is running out of room. Physicists and engineers have been looking to two-dimensional materials—sheets of matter just a few atoms thick—as the next frontier. These ultrathin semiconductors promise better control, lower power consumption, and the possibility of devices smaller than anyone thought feasible. But there's a problem nobody could quite measure: how small can the metal contacts that feed current into these devices actually get?
For years, researchers have estimated this limit through theory and simulation. A team working with molybdenum disulfide, a popular 2D material, used various metals and mathematical models to guess that the effective current injection region—what physicists call the carrier transfer length—might be somewhere between 9 and 45 nanometers, depending on the metal and the assumptions they made. But guessing isn't good enough when you're trying to build devices smaller than 10 nanometers. You need to see it directly.
A new study published in Nature describes exactly that: the first direct measurement of carrier transfer length in a 2D transistor. Using a specialized form of scanning tunneling microscopy that can probe electronic structures at the sub-nanometer scale while the device is actually operating, researchers found that bismuth-contacted molybdenum disulfide transistors have a transfer length of approximately 2 to 3 nanometers. The finding is striking because it suggests that metal contacts can be scaled down far more aggressively than previous estimates implied—well within the requirements for the next generation of chip technology.
The team's approach was ingenious. They grew monolayer molybdenum disulfide using chemical vapor deposition, deposited bismuth contacts on top, and then cleaved the device in an ultrahigh-vacuum chamber to expose a cross-section. This prevented contamination and allowed them to scan perpendicular to the metal-semiconductor interface while applying voltage to the device. Using scanning tunneling spectroscopy, they measured how the electronic band structure of the molybdenum disulfide changed as they moved away from the contact edge. The conduction band edge—the energy level at which electrons can move freely—shifted dramatically in the first few nanometers from the metal, then stabilized. That shift, they determined, reflected the injection of carriers from the bismuth into the semiconductor. By fitting the data to an exponential decay function, they calculated a transfer length of 2.0 nanometers for the device with a silicon dioxide gate dielectric, extending to 2.9 nanometers when additional gate voltage was applied.
The measurement required careful interpretation. The scanning tunneling microscope doesn't measure carrier concentration directly; instead, it measures how the band structure appears to shift due to changes in local carrier density. The team had to account for a quantum mechanical effect called tip-induced band bending, where the microscope tip itself influences the electronic structure it's trying to measure. By running simulations and systematically ruling out alternative explanations—defects, dopants, instrumental artifacts—they concluded that the observed band shifts reflected genuine carrier injection from the bismuth contact into the molybdenum disulfide.
Why bismuth? Unlike gold or nickel, which have been used in earlier studies, bismuth is a semimetal with electronic properties that align well with molybdenum disulfide. It creates what physicists call a nearly barrier-free contact, meaning electrons can flow across the interface with minimal resistance. The team measured a contact resistance of just 70 ohm-micrometers, among the lowest reported. This low resistance, combined with the ultrashort transfer length, suggests that bismuth-contacted molybdenum disulfide transistors could be scaled to contact sizes well below 10 nanometers—meeting the specifications for what the semiconductor industry calls the 1-nanometer technology node, a designation that refers to the smallest feature size, not the actual physical dimension.
The researchers also tested their methodology on a p-type semiconductor, tungsten diselenide with palladium and antimony contacts, finding a transfer length of about 211 nanometers—much longer, reflecting the current state of p-type contact engineering. This disparity points to the next challenge: while n-type contacts in 2D transistors can now be scaled to the nanometer regime, p-type contacts lag far behind. For practical devices that need both electrons and holes to flow, this asymmetry is a significant hurdle. The work establishes a new experimental tool for contact engineering and provides direct evidence that the metal-semiconductor interface, long a mysterious bottleneck in device scaling, can be pushed much further than theory alone suggested. What remains is the harder work of making it reliable, reproducible, and applicable to the full range of materials and contact types that future electronics will demand.
Citas Notables
The observed exponential decay of the conduction band edge shift directly correlates with the spatial distribution of injected carriers, providing unprecedented insight into the carrier injection process at the metal–2D semiconductor interface.— Nature study authors
These results provide direct experimental evidence that contact scaling in 2D transistors can extend well into the deep nanoscale regime, consistent with recent industry projections.— Nature study authors
La Conversación del Hearth Otra perspectiva de la historia
Why does it matter that you can measure this directly instead of just calculating it?
Because calculation requires assumptions. When you don't know what's actually happening at the interface, you have to guess about the material properties, the defects, how the metal changes the semiconductor underneath it. Those guesses compound. Direct measurement cuts through that.
But you're still using a microscope tip that affects what you're measuring. How do you know you're seeing the real thing?
You don't, not entirely. That's why we had to simulate the tip effect separately and show that our interpretation—that we're seeing carrier injection—is the only one consistent with the data. We ruled out everything else.
The transfer length is 2 nanometers. Is that small enough?
For the next technology node, yes. The industry needs contacts smaller than 18 nanometers. We're showing it's possible to go to 2 or 3. But that's just n-type. The p-type contacts are still 200 nanometers. That's the real problem now.
Why is p-type so much worse?
We don't fully know yet. The physics is different—holes behave differently than electrons. But also, bismuth works beautifully for electrons. We haven't found the bismuth equivalent for holes. That's the next frontier.
Does this mean we can actually build these tiny transistors now?
It means we can build the contacts. Building the whole device—the channel, the gates, keeping everything aligned and defect-free at that scale—that's still years away. But this removes one major uncertainty from the roadmap.
What surprised you most?
How much smaller it could be than the old estimates. The previous work suggested 9 to 45 nanometers. We're at 2. That's not a refinement—that's a different regime entirely.