2. NanoDCAL: electronic structure and transport

Note

The calculations here run on a local NanoDCAL installation or on an HPC cluster. For everything NanoDCAL can do, see NanoDCAL support; to continue a transport project across several requests, see Continuing a transport project.

LatticeMind’s NanoDCAL backend brings quantum-transport-oriented calculations into the same plain-language workflow. You select it simply by saying “Using NanoDCAL, …” in your prompt. This tutorial walks through a silicon band structure, then points to the full two-probe transport workflow.

2.1. Silicon band structure

Using NanoDCAL, run an SCF and band structure calculation for silicon.

LatticeMind builds the silicon crystal, copies the LCAO basis files into the project and writes the two inputs. It asks your permission, runs the SCF, then computes the band structure along the standard high-symmetry path from the SCF’s saved result — and plots it referenced to the Fermi level. The answer gives NanoDCAL’s own band gap and the band edges along the path:

Silicon band structure computed with NanoDCAL.

Fig. 2.1.1 Silicon band structure from NanoDCAL, along L–Γ–X–W–K with energies referenced to the Fermi level (dashed line). The indirect gap between the valence-band maximum at Γ and the conduction-band minimum near X is clearly resolved.

2.2. Validate the band-gap interpretation

The plotted zero is the NanoDCAL Fermi energy. The reported silicon gap is obtained from the occupied and unoccupied bands sampled along the full L–Γ–X–W–K path. It is therefore an indirect path-sampled gap, not the energy separation between two states selected at one k-point.

This distinction matters when comparing BandStructure.mat with EigenStates.mat. For example, requesting one state below and one state above the Fermi level on a [1, 1, 1] grid samples two Γ-point states. Their direct separation can be much larger than silicon’s indirect band gap and should not be reported as the PBE fundamental gap.

Before using the value in a manuscript:

  • repeat the SCF calculation with a denser k-grid and keep the plotted high-symmetry path unchanged;

  • record the exchange-correlation functional and NeutralAtomDatabase basis choice, because both affect the numerical gap;

  • confirm that the band edges in the answer (or in the /report evidence report) match what the figure shows; and

  • remember that a converged PBE gap is a method-specific result, not an experimental-gap correction.

Tip

Because the LCAO basis travels with the project, the generated project runs unchanged on another machine — locally or on a remote NanoDCAL server (see Remote & HPC execution). When sharing the generated project, please keep the staged NeutralAtomDatabase files within the scope of your Nanoacademic license.

2.3. Spin-resolved transmission channels

This prompt exercises the transverse-k and spin dimensions in TransmissionChannel.mat:

Using NanoDCAL, run SCF and transmission-channel analysis for
ferromagnetic bcc iron.

LatticeMind plots the channel count on the regular transverse-k mesh and keeps the spin-up and spin-down maps on the same colour scale:

Spin-resolved transmission-channel maps for ferromagnetic bcc iron.

Fig. 2.3.2 NanoDCAL transmission-channel count at the Fermi level for 900 transverse k-points. Separate spin panels preserve the two-dimensional k-space structure instead of flattening it into a sample-index curve.

2.4. How to read and validate the channel map

Each pixel represents one transverse (k1, k2) point at E - EF = 0 eV; the colour gives the number of open transmission channels at that point. Both spin panels use the same colour limits, so their different k-space patterns can be compared directly. The map is not, by itself, a total conductance: a scalar transport value also requires the transverse-k weights and the appropriate energy or bias integration.

Before interpreting the spin contrast, check that:

  • the bcc Fe SCF calculation converged to the intended magnetic state;

  • both panels use the same energy, transverse mesh, and colour scale;

  • the channel pattern and any integrated transport quantity are stable when the transverse mesh is refined beyond the illustrated 30 by 30 grid; and

  • the lead and device calculations use compatible cells, basis data, and spin settings.

The figure is saved in the project with the other plots, and /report writes an evidence report that names TransmissionChannel.mat as its source. With latticemind quick, SVG and PDF exports are written beside the PNG for manuscript editing.

2.5. Going further: quantum transport

The same one-line pattern drives full two-probe transport calculations — LatticeMind builds the lead-SCF → device-SCF → property chain automatically:

Using NanoDCAL, run a transmission calculation along z for a homogeneous
aluminum two-probe device.

Start with the zero-bias transmission and inspect Transmission.mat before requesting a bias sweep. A useful follow-up is:

Using NanoDCAL, calculate an I-V curve along z for a homogeneous aluminum
two-probe device from -0.2 V to 0.2 V.

For every transport comparison, keep the device geometry, lead construction, NeutralAtomDatabase basis, transverse k-grid, energy grid, and electronic temperature fixed unless the prompt explicitly studies one of those choices.

See NanoDCAL support for the complete list of supported transport, electronic-structure, phonon, and mechanical calculations.

2.6. Three useful follow-up checks

The following prompts exercise different NanoDCAL output families without changing the installation:

  1. Using NanoDCAL, calculate total energy, effective potential, electron density, and charge analysis for rocksalt MgO. This produces TotalEnergy.mat, EffectivePotential.mat, TotalElectronDensity.mat, and Charge.mat after one SCF calculation.

  2. Using NanoDCAL, calculate the effective mass for GaAs. This produces EffectiveMass.mat and tests the electron/hole dispersion fit and directional masses in units of \(m_e\).

  3. Using NanoDCAL, run a conductance calculation along z for a homogeneous aluminum two-probe device. This exercises the lead-SCF, device-SCF, and conductance chain and produces ConductanceAndCurrent.mat.

For each project, read the permission question before you answer it. After the runs, ask “Which calculations finished, and did each one converge?” and check that the expected .mat files exist and that the answer raises no unexplained warning. /report writes the evidence report under results/. To redraw figures only, ask LatticeMind to plot the existing NanoDCAL outputs without rerunning the solver.

2.7. What you learned

  • Selecting NanoDCAL is as simple as saying so in your prompt.

  • LatticeMind handles NanoDCAL’s basis files, multi-step dependencies, and device chains for you.

  • Every launch asks first, and the answer ties each number to the NanoDCAL .mat file it came from.

  • /report writes the evidence report; figures are saved in the project.