Nanotubes and tube–metal interfaces

vibe-viewer builds carbon nanotubes and the cluster models of their contacts with metals, draws structures of up to a million atoms, and analyses the interface a calculation produced. Everything here makes starting geometries: ideal cylinders and unrelaxed contacts to optimise before a calculation, not converged structures.

The browser’s tools are in the Nanotube builder and views card of the side panel; the same functions are in Python as vibeview.nanotube, vibeview.interface_builder, vibeview.structure_views and vibeview.interface_analysis.

Building a tube

Type the chiral indices n and m, the number of walls, the length in Å and the ends, then choose Build nanotube. The tube opens as a new structure along z. As you type, the card reads out each wall’s (n,m), diameter, chiral angle, whether it is metallic or semiconducting by the zone-folding rule, and the spacing between walls:

(5,5) armchair: d = 6.78 Å, θ = 30.0°, metallic
(10,10) armchair: d = 13.56 Å, θ = 30.0°, metallic
Wall spacing: 3.39 Å

Ends	What you get
Open, H-terminated	Dangling atoms removed and every two-coordinated carbon capped with H at 1.09 Å
Open, bare	Dangling atoms removed; the edge carbons stay two-coordinated
Capped (C60 half)	Both ends closed with a piece of C60 scaled onto the tube. Only rims a C60 half fits, such as (5,5) and (9,0), can be capped; others are refused with a message
Periodic along the axis	A whole number of unit cells, periodic along z, in a box with 15 Å of vacuum around the tube

A multi-wall tube grows outward from (n,m): each outer wall is the (n’,m’) whose radius is nearest the previous one plus the wall spacing (3.4 Å by default), with the chiral angle nearest the inner wall’s breaking ties, so an armchair core grows armchair walls. Multi-wall tubes are built with open or hydrogen-terminated ends; walls of different chirality have incommensurate periods, so they are neither capped nor periodic.

In Python:

from vibeview import nanotube as nt

tube = nt.single_wall(10, 10, length=40.0, end="hydrogen")
mwnt = nt.multi_wall(5, 5, walls=4, length=30.0)
rope = nt.bundle(nt.single_wall(10, 10, 30.0), count=7)       # hexagonal, 3.4 Å gaps
forest = nt.forest(nt.multi_wall(5, 5, 3, 30.0), 4, 4, pitch=50.0)
print(nt.wall_info(10, 5))   # diameter, chiral angle, period, atoms per cell

tube.atoms() gives QVF structure atoms, ready for vibeview.qvf_writer.structure_archive.

Adding a metal or a linker

With a tube shown, choose a partner, its options and Add partner to shown tube. The partner is added to the shown structure through the edit overlay, like every builder edit, so one Undo removes it.

Partner	Options
Metal slab	Al, Ni, Cu, Pd, Ag, Pt or Au; (111), (100) or (110); the number of layers. A square patch sized to the contact, cut from the bulk lattice by the slab builder
Nanoparticle	The metal and a radius: the fcc atoms within it, turned so a (111) layer faces the tube
Aminophenyl amide linker	–C(=O)–NH–C6H4– on a carbon of the tube, the linker of diazonium grafting; optionally with a metal slab on its free aryl carbon, the tube–linker–metal contact of the Alvarez group’s Cu and Au electrodes

Site puts the partner under the tube’s lowest open end or against its sidewall at mid-length; a capped or hydrogenated tube has no open end, and the card says so. Registry puts the facing tube atom over a top, bridge or hollow site of the partner’s facing layer. M–C distance is the distance from that atom to the nearest atoms of the site, one for top, two for bridge, three for an fcc(111) hollow; left blank it is the sum of the covalent radii, a bonded contact (2.08 Å for Cu–C, 2.12 Å for Au–C). A placement that would bring the partner within 1.5 Å of another tube atom is refused: at a hollow site a flat open end may need a longer distance than at a top site.

Partners are finite cluster models, so they need a structure that is not periodic; build the tube with open, hydrogen-terminated or capped ends.

Seeing a large structure

Colour atoms by replaces element colours, on atoms and bonds alike:

Mode	Colour ramp	Needs
Coordination number	Sequential, over the structure’s range; counts the bonds drawn, so bulk metal atoms, which get no metal–metal bonds, count only their bonds to C	Nothing
Distance from the axis	Sequential	Nothing
Partial charge	Diverging blue–white–red, symmetric about 0 e	An atom_properties section; IAO, Hirshfeld, Mulliken and Löwdin are used in that order
C–C bond stretch (%)	Diverging, mean C–C bond-length change relative to the builder’s 1.42 Å reference; non-carbon atoms are zero	Nothing
Open ends, coordination defects and adducts	Grey ordinary; blue two-coordinated carbon within 2 Å of a finite axial end; orange other carbon coordination anomalies; magenta carbon and attached heavy non-carbon atoms	Nothing
Bond order to a metal	Sequential; the largest order of each atom to any metal	A bond_orders section

The status line names the range of the ramp. A mode whose section the archive lacks is refused with a message, and the colours stay as they were.

Cut-away removes part of the structure so you can see inside it:

Cut-away	Keeps
Half wall	The half of the wall away from the viewer of the side view
Cross-section slice	A slice 5 Å thick across the axis, at Cut position
Cut at a point on the axis	Everything below Cut position

Atoms close to the axis, an encapsulated molecule or a filled core, are always kept. Down the axis and Side view point an orthographic camera along the tube or at the face a half cut opens, for figures. The axis is the principal axis of the carbon atoms, so a metal slab under the tube does not tilt it.

A structure of 500 atoms or more is drawn with instanced glyphs: one sphere and one cylinder mesh, repeated on the graphics card for every atom and bond. A scene of 10 000 atoms or more also draws coarse spheres and bonds while you turn it and the full ones once you let go. Large scene construction uses vectorised atom, bond and neighbour batches. The client performance page lists the measured limits.

Interface analysis

Interface analysis reads the archive the calculation wrote and reports:

• every metal–C pair within the sum of the covalent radii plus 0.45 Å, with its length and, if the archive has a bond_orders section, its order and method;

• the projected DOS of the interface atoms, summed separately over the metal atoms and the carbons in those contacts, when a dos.projected section labels its channels with atom indices;

• the first volume.difference or volume.potential along the tube axis: a density difference integrated over each plane perpendicular to the axis, in e/Å, with the charge transferred across each plane accumulated along it; a potential averaged over each plane.

Each part appears when its section is present, and the report says which are missing. In Python:

from vibeview.qvf import QVFReader
from vibeview import interface_analysis as ia

with QVFReader("cnt-cu.qvf") as reader:
    structure = reader.read_structure()
    for row in ia.metal_carbon_contacts(structure, reader):
        print(row["metal_label"], row["carbon_label"], f"{row['distance']:.3f}", row["order"])
    profile = ia.axial_profile(reader, "density_difference", structure)
    print(profile.cumulative[-1], "e transferred")

None of this needs a new QVF field: the analysis reads bond_orders, dos.projected and volume sections as the format defines them.

Growth and reaction-path playback

Trajectory and reaction-path frames now reuse sphere and cylinder prototypes; only their instance positions, scales and directions change. Live frames draw bonds inferred from each displayed geometry, including bond formation and breaking. These are covalent-radius display contacts, not computed bond orders. Periodic reaction paths use their per-frame cells and periodic images. Returning from a volume-series frame removes its intermediate atoms.

Automatic playback strides over intervening frames when a render exceeds its 0.2 s budget; the last frame remains reachable. Manual steps, the frame slider and the NEB energy profile still select individual stored frames. This reduces rendering work, not archive memory: QVF coordinate arrays are still decoded in full. At 100,000 atoms the filmstrip samples four frames instead of sixteen. Large molecular thumbnails reuse glyph inputs; explicit-bond, periodic and wireframe thumbnails retain the established structure renderer.

The coordinate editor displays 100 atoms per page. Previous atoms and Next atoms change the visible rows; Apply table applies the complete edited table in one transaction, including edits on other pages.

Figure cues and capture

Depth cue (browser figure) blends distant fragments toward the background in the vtk.js fragment shader. It follows the camera and works with Down the axis and Side view. Toggle it off for unmodified property colours. It is a browser figure effect, including browser screenshots; headless capture does not apply it.

Bond stretch is a geometric proxy, not a strain tensor or a force calculation. Site highlights are coordination heuristics, not ring-topology or chemical assignments: three-coordinated pentagon/heptagon defects are not identified. Hydrogen termination does not count as a heavy-atom adduct; periodic structures have no highlighted open rims. Inspect and verify sites before modelling them.

Cut-aways and property colours also work in command-line and Python captures:

vibe-viewer capture tube.qvf --cutaway half --atom-color-mode local_strain -o tube.png

from vibeview.api import capture_structure
capture_structure("tube.qvf", "tube.png", cutaway="section", cutaway_position=0.3,
                  atom_color_mode="carbon_sites")

cutaway, cutaway_position and atom_color_mode have the same meaning in vibeview.headless.add_structure and render_structure_frame. The defaults remain an uncut structure in element colours.

Terminal scale

At 10,000 atoms and above, the terminal rasterizer uses one depth-tested point sprite for an atom whose projected radius is at most 0.75 pixel. It keeps the nearest atom at each pixel, rather than discarding an arbitrary stride of atoms. Off-screen atoms are culled. Subpixel bonds covered by their endpoint atoms are omitted; sticks-only views retain their usual bonds. Zooming in restores shaded spheres and resolved bond cylinders automatically. The million-atom forest still takes roughly 0.64 s per orbit step in the latest diagnostic; this does not promise smooth million-atom interaction.