Science1 publisherNot yet confirmed elsewhere2 min readPublished
Oxygen released from an oxide barrier pins each MoS2 crystal's nucleus to its pattern's centre
Researchers reporting in Nature pin each MoS2 crystal's nucleus to its pattern's centre with an oxygen etching flux, reaching mobilities up to 117 cm2/Vs. In principle, a 2D-semiconductor maker can now decide where each single crystal starts when drawing the pattern.
The Scientist · Science desk
What happened
- In conventional area-selective growth the whole patterned region can nucleate, so crystals start at random sites and their size is capped.
- The method grew single-crystal molybdenum disulfide at the 10-micrometre scale.
- Demonstrations included line-shaped single crystals, several transistors inside one large common crystal, and aligned lateral heterostructures for self-aligned contacts.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- capability A circuit layout can put a transistor channel on single-crystal MoS2 by design, because the surviving nucleus sits at a position fixed by the pattern's geometry.
- decision The number of nuclei per site becomes something chosen when the pattern and barrier are designed, using the authors' rules for the zero, single and multiple regimes.
- precedent Area-selective growth is already an established deposition step for silicon-germanium, tungsten and cobalt, so a nucleation rule built on it gives 2D materials a plausible way into existing process flows.
The oxide barrier in this process does two jobs. Like any area-selective barrier, it marks out where MoS2 may grow. It also releases oxygen that etches sideways into the growth region [3]. Conventional 2D area-selective growth gets its selectivity from surfaces with different binding energies, and that contrast treats every point inside the growth region the same [12]. Earlier confined-growth work did produce single-crystal transition-metal dichalcogenides, but only small ones [10].
The lateral flux makes the probability of nucleation uneven inside the pattern [12]. It suppresses nuclei near the pattern boundary, so the only place a nucleus can survive is the geometric centre [3]. The authors report design rules for three regimes, giving zero, one or several nuclei per pattern [4]. I think the zero and multiple regimes are the most useful controls in the set. A method that only ever showed one crystal per pattern might just have been lucky with seeding. A method that can be tuned to suppress every nucleus, or to let several survive, shows that the flux sets the count [4].
The mobility in the headline is a maximum: field-effect mobilities of "up to" 117 cm2 per volt-second [6]. The abstract does not report a median mobility, the number of transistors measured or the fraction of patterns that ended in the single-nucleus regime [6]. For a fab, that fraction sets yield. A patterned wafer is useful only to the extent that its sites each grow exactly one crystal. The authors do claim large-area uniformity and compatibility with existing processes [7].
The authors wrote that their approach expands semiconductor deposition "from controlling where materials grow to controlling where and how crystals can begin to form" [14]. So far the work covers one material, molybdenum disulfide, grown as single crystals at the 10-micrometre scale [5].
What to watch
- Wafer-scale counts of how many patterns land in the single-nucleus regime, reported alongside median mobility and the number of devices measured.
- Whether the same oxygen-flux design rules hold for transition-metal dichalcogenides other than MoS2.
- Whether crystals grown from a single centred nucleus can be scaled well past the 10-micrometre scale.