Gas permeation through nanoscale pores is ubiquitous in nature and has an important role in many technologies1,2. Because the pore size is typically smaller than the mean free path of gas molecules, the flow of the gas molecules is conventionally described by Knudsen theory, which assumes diffuse reflection (random-angle scattering) at confining walls3,4,5,6,7. This assumption holds surprisingly well in experiments, with only a few cases of partially specular (mirror-like) reflection known5,8,9,10,11. Here we report gas transport through ångström-scale channels with atomically flat walls12,13 and show that surface scattering can be either diffuse or specular, depending on the fine details of the atomic landscape of the surface, and that quantum effects contribute to the specularity at room temperature. The channels, made from graphene or boron nitride, allow helium gas flow that is orders of magnitude faster than expected from theory. This is explained by specular surface scattering, which leads to ballistic transport and frictionless gas flow. Similar channels, but with molybdenum disulfide walls, exhibit much slower permeation that remains well described by Knudsen diffusion. We attribute the difference to the larger atomic corrugations at molybdenum disulfide surfaces, which are similar in height to the size of the atoms being transported and their de Broglie wavelength. The importance of this matter-wave contribution is corroborated by the observation of a reversed isotope effect, whereby the mass flow of hydrogen is notably higher than that of deuterium, in contrast to the relation expected for classical flows. Our results provide insights into the atomistic details of molecular permeation, which previously could be accessed only in simulations10,14, and demonstrate the possibility of studying gas transport under controlled confinement comparable in size to the quantum-mechanical size of atoms.
Access optionsAccess options
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work was supported by the European Research Council, Lloyd’s Register Foundation, the EU Graphene Flagship and the Royal Society. B.R. acknowledges a Leverhulme Early Career Fellowship, a L’Oréal Fellowship for Women in Science and EPSRC grant EP/R013063/1. F.C.W. acknowledges support from the National Natural Science Foundation of China (11772319 and 11572307) and the Shanghai Supercomputer Center. S.J.H. and A.P.R. were funded by the European Research Council under the European Union’s Horizon 2020 research and innovation programme (grant agreements ERC-2016-STG-EvoluTEM-715502 and DISCOVERER-2017 737183), the US Defence Threat Reduction Agency (HDTRA1-12-1-0013) and the EPSRC (EP/P009050/1 and EP/K016946/1).Reviewer information
Nature thanks L. Bocquet and the other anonymous reviewer(s) for their contribution to the peer review of this work.
Extended data figures and tables
a, Thin crystal was transferred to cover an opening in a SiN x membrane. b, The opening is extended through the bottom crystal. c, Spacer stripes were deposited on top of the bottom layer and etched from the back side. The top crystal is then transferred on top to fully cover the rectangular opening. d, Left, optical image of a single-channel capillary device made entirely from graphite; N = 5. The SiN x /Si wafer is seen in dark green; the SiN x membrane appears as a light-green square; the top graphite layer shows up in bright yellow; and the rectangular opening (lighter green) is indicated by the black arrow. Centre, atomic force micrograph near the channel entry, where the top graphite does not cover the spacer layer (the scan area is shown by the red contour). The height profile taken along the dotted white line is shown on the right, indicating h ≈ 1.7 ± 0.1 nm. The side cavities perpendicular to the 2D channel were made to prevent contamination bubbles25 across the main channel. e, A gold mask is placed on top of the trilayer assembly for final etching, to define L and to unblock the channel entry. f, Optical image of the final capillary device in the transmission mode. The SiN x membrane is fully transparent (bright). The Au mask is partially transparent, and both the top graphite and the rectangular hole in SiN x can be seen underneath the Au, as indicated by the arrows.
a, Array of 2D slits made entirely from MoS2, as imaged in bright-field STEM. For guidance, the edges of one of the channels are indicated by red marks. The vertical stripes are from the curtaining effect caused by ion milling32. b, High-magnification STEM image of a 2D channel with the top and bottom walls made from MoS2 and bilayer graphene as the spacer (right panel). The channel is white in the bright-field image, and atomic layers of MoS2 can be seen as dark lines running parallel to the channel. The left panel shows a contrast profile across the region indicated by the red rectangle. Cross-sectional images of 2D slits made entirely from graphite crystals can be found in ref. 13.
a, Schematic of our experimental set-up. b, Helium flow through round apertures of various diameters as measured by our He-leak detector (symbols). Red line, expected Knudsen flow through these apertures (no fitting parameters). Inset, optical image of one of the apertures. The error bars are from measurements using different devices.
Electron-density profiles near graphite, h-BN and MoS2 surfaces are shown. Schematics of the atomic structures are shown on top. The red curves indicate the thermal exclusion surfaces.
a, Sketch of our simulation set-up. b, c, Energy of a PMMA molecule (M = 40,000) for slits with N = 4 (b) and N = 12 (c). The axes show the position of the centre of mass with respect to the entrance edge. The origin of the axes is shown in a. The colour scales to the right show the relative free energy (potential of mean force, PMF).
MD results are shown for the apparent (circles) and gyration (squares) heights of PMMA on graphene. The shaded areas indicate the standard errors using the data for M ≥ 40,000. Error bars show the standard error from our simulation runs lasting 10 ns.
If an incident atom of diameter d hits one of the edges of the channel, it can be reflected. To avoid this, the centre trajectory should be d/(2cosθ) away from the edge, effectively reducing the entry aperture to h*(θ) = h − d/cosθ.