Source-linked AI summary
HORACE: software for the analysis of data from single crystal spectroscopy experiments at time-of-flight neutron instruments
R. A. Ewings, A. Buts, M. D. Le, J. van Duijn, I. Bustinduy, T. G. Perring
TL;DR
Large time-of-flight neutron experiments produce multidimensional S(Q, ω) data that require efficient visualization and analysis. Horace combines measurements into extensible data sets, supports cuts and manipulations, and provides simulation and fitting; examples demonstrate its use for magnetic excitations and phonons.
Problem
Time-of-flight neutron spectrometers generate large multidimensional data sets whose visualization, manipulation, and analysis require specialized software.
Method
Horace combines multiple measurements into large data sets, enables multidimensional cuts and manipulations, and supports simulations and fits of S(Q, ω) models.
Results
Horace was used to examine magnetic excitations in RbMnF3 and iron and phonons in URu2Si2, with URu2Si2 data covering a much larger reciprocal-space volume than earlier work.
Takeaways & Limitations
Horace provides an extensible framework for exploratory visualization, algebraic manipulation, simulation, and fitting of combined neutron-scattering data.
Abstract
from arXiv · showhide
The HORACE suite of programs has been developed to work with large multiple-measurement data sets collected from time-of-flight neutron spectrometers equipped with arrays of position-sensitive detectors. The software allows exploratory studies of the four dimensions of reciprocal space and excitation energy to be undertaken, enabling multi-dimensional subsets to be visualized, algebraically manipulated, and models for the scattering to simulated or fitted to the data. The software is designed to be an extensible framework, thus allowing user-customized operations to be performed on the data. Examples of the use of its features are given for measurements exploring the spin waves of the simple antiferromagnet RbMnF$_{3}$ and ferromagnetic iron, and the phonons in URu$_{2}$Si$_{2}$.
I. INTRODUCTION
Time-of-flight spectrometers with position-sensitive detectors enable efficient mapping of S(Q, ω), but the resulting multidimensional data require specialized software. Horace combines measurements across reciprocal space and energy while supporting visualization and analysis on practical computer hardware.
- I. INTRODUCTION: Time-of-flight spectrometers with position-sensitive detectors measure S(Q, ω) over multidimensional reciprocal-space and energy regions.These measurements complement the point-by-point controllability of triple-axis spectrometers.
- I. INTRODUCTION: Known instrument geometry and neutron time-of-arrival determine the scattered wave vector, momentum transfer Q, and energy transfer ℏω.The incident wave vector is selected by the chopper, while detector position and timing provide the scattered-neutron information.
- I. INTRODUCTION: Crystal-fixed coordinate frames and physically motivated projections organize S(Q, ω) for materials with anisotropic interactions.For layered systems, the weakly coupled direction can be treated as an implicit coordinate and projected out.
- I. INTRODUCTION: Coarse-grain sorting and large detector coverage support rapid access to multidimensional subsets despite large data volumes.A MERLIN data representation can contain O(10^7) voxels and about 0.5 GBytes.
- I. INTRODUCTION: Horace combines multiple measurements to map the four-dimensional dependence of vector momentum and energy.An additional scan parameter, usually sample orientation Ψ, supplies the extra degree of freedom needed for full mapping.
- I. INTRODUCTION: Combining runs with different incident energies and sample orientations is generally discouraged because it can produce a multi-valued resolution function.The resulting ambiguity makes interpretation and analysis more difficult.
A. Main Purpose
Horace is designed to make four-dimensional inelastic neutron-scattering data from multiple measurements easy to visualize, manipulate, and analyze on commodity computers.
- A. Main Purpose: Horace provides visualization, manipulation, and analysis of inelastic neutron-scattering data across four dimensions of momentum and energy.The data may be gathered from multiple crystal orientations or incident energies.
B. Architecture and key features
Horace uses extensible data objects and coarse-grained storage to support multidimensional cuts, plotting, algebraic operations, symmetrization, simulation, and fitting.
- B. Architecture and key features: The sqw object stores voxel signals, variances, detector metadata, crystal information, and Cartesian plotting grids.The dnd object omits individual voxel and detector-location information and therefore uses substantially less memory.
- B. Architecture and key features: Horace constructs 4D sqw data from multiple inelastic-neutron-scattering measurements and extracts 1D–4D cuts along arbitrary momentum or energy directions.Cuts can be created recursively from previously cut sqw objects.
- B. Architecture and key features: Horace supports plotting, unary and binary algebraic operations, lower-dimensional replication, and symmetrization of data objects.Unary operations include detailed-balance correction, while replication can support background estimation.
- B. Architecture and key features: Simulation and fitting routines model the scattering function S(Q, ω).The software also provides planning and plotting utilities for reciprocal-space measurements and dispersion relations.
- B. Architecture and key features: Coarse-grain sorting onto a regular 4D grid accelerates access to selected reciprocal-space and energy volumes.The sorted data are saved in a generally large .sqw file.
C. Interface
Horace exposes its operations as Matlab functions that users can run interactively or compose into scripts and functions for more complex analyses.
- C. Interface: Users interact with Horace through the Matlab command line or through their own Matlab scripts and functions.The paper illustrates command-line use with a cut and plots generated during the command sequence.
D. Operation
Horace uses a two-stage workflow to combine multiple run files into large sqw files, then extract, manipulate, visualize, simulate, and fit cuts. Illustrative examples cover RbMnF3, iron, and URu2Si2.
- D. Operation: Horace first combines individual run files into file-backed sqw files, limiting memory requirements for data sets typically ranging from 10 to 500 GB.
- D. Operation: After sqw creation, users can take cuts of any dimensionality, apply corrections, and simulate or fit models to extract physically meaningful parameters.
- D. Operation: Examples use RbMnF3, iron, and URu2Si2 data to illustrate Horace functionality.
- D. Operation: Figure 2 demonstrates extracting and plotting a 2D cut, symmetrizing it in two planes, and reproducing the operation through the Horace GUI.
- D. Operation: URu2Si2 data agree with published phonon work while covering a much larger reciprocal-space volume through the paper’s collection method.
V. USE OF THE PROGRAM
Horace supports experiment planning, metadata-based run combination, computational acceleration, benchmarking, and incremental updates as measurements accumulate.
- A. Planning an experiment: The scan planner calculates reciprocal-space coverage before measurements from instrument, incident-energy, scan-range, lattice, and orientation inputs.
- B. Creating an sqw file: Users combine selected runs into an sqw file while manually supplying incident energy, sample orientation, lattice parameters, and lattice angles.
- B. Creating an sqw file: C++ mex routines accelerate sqw creation and computationally intensive operations by using compiled and multithreaded processing.
- B. Creating an sqw file: A 142 GB sqw file comprising 231 runs took 150 minutes to generate on the stated Windows 7 workstation.
- B. Creating an sqw file: Chunk-size options allow Horace to operate on lower-specification commodity PCs when sufficient disk space is available.
- B. Creating an sqw file: Planned runs can be binned onto the same coarse grid and inserted into an existing sqw file without regenerating the entire file.
C. Extracting and visualizing data
Horace extracts and visualizes flexible multidimensional cuts from large neutron-scattering data sets, including reciprocal-space slices, one-dimensional trajectories, and high-symmetry dispersions.
- C. Extracting and visualizing data: Users can sample 3D volumes, 2D slices, and 1D cuts along arbitrary Q and energy trajectories, with coarse-grained sorting accelerating access.
- C. Extracting and visualizing data: Horace provides customizable marker plots, colormaps, and 3D multiple-colormap visualizations for 1D, 2D, and 3D cuts.
- C. Extracting and visualizing data: The sliceomatic tool enables rapid exploration of large 3D reciprocal-space sections by moving visible slice planes.
- C. Extracting and visualizing data: The RbMnF3 slice shows dispersive magnetic excitations from 0 ≤ E ≤ 9 meV, with 1D cuts at L = 0, 0.15, and 0.3 averaged over ±0.05 r.l.u.
- C. Extracting and visualizing data: The spaghetti plot displays URu2Si2 phonon dispersion around Q = (2, −2, 0) and reveals acoustic-mode splitting along Γ–Σ plus differently dispersing higher-energy modes.
D. Manipulating data
Horace supports algebraic and smoothing operations on data objects, including background subtraction by extending lower-dimensional cuts across higher-dimensional regions.
- D. Manipulating data: Unary operations modify intensities through corrections such as Bose-Einstein population factors or magnetic form factors, while binary operations subtract one object from another.
- D. Manipulating data: dnd data can be smoothed with dimensional Gaussian or hat-function convolutions, whereas sqw data cannot be smoothed without losing detector-voxel information.
- D. Manipulating data: A 1D energy cut from a background-only region can be tiled across Q and subtracted to isolate intrinsic S(Q, ω).
E. Simulating and fitting
Horace supports simulation and fitting of measured neutron-scattering data using coordinate-based or physically meaningful S(Q, ω) models. Its global fitting framework handles arbitrary-dimensional cuts while accommodating varying local backgrounds, although instrumental resolution is modeled only crudely.
- Model fitting: Horace fits and simulates data at the measured (Q, ω) values using generic coordinate functions or physically meaningful S(Q, ω) models.Generic functions can parameterize features such as dispersion peaks, while direct S(Q, ω) models provide a more physical description.
- Model fitting: Voxel-level sqw fitting evaluates models across contributing voxels before binning, reducing systematic errors when S(Q, ω) varies within a bin.This is more accurate than evaluating only at bin centers, but requires greater computer processing.
- Global fitting: Horace fits arbitrary numbers of cuts at any dimensionality with global parameters, separating shared foreground models from cut-specific background functions.Global parameters can be fixed, varied, or linked by fixed ratios, and fit functions may use lookup tables or numerical parameters.
- Limitations: Instrumental resolution broadening can be included only crudely, such as through Gaussian energy broadening, because Horace lacks a specific resolution model.The limitation applies to the current fitting implementation.
VI. SUMMARY
Horace combines multiple time-of-flight neutron measurements into large, extensible datasets for visualization, manipulation, simulation, and fitting. It supports global foreground models with independent backgrounds and provides documented installation and hardware guidance.
- VI. SUMMARY: Horace combines multiple time-of-flight neutron inelastic-scattering runs into a single dataset that can reach hundreds of GB.The suite is intended for large multiple-measurement datasets rather than isolated runs.
- VI. SUMMARY: The extensible framework supports sophisticated data manipulations, fast visualization of subsections, simulations, and S(Q, ω)-based fitting.Coarse-grained sorting of detector voxels by (Q, ω) coordinates enables access to relevant subsections of large files.
- VI. SUMMARY: Multiple datasets can be fitted with shared global foreground parameters and independent background models and parameters.This organizes fitting around the physical origin of the measured signal while accommodating dataset-specific backgrounds.
- VI. SUMMARY: Horace requires at least 8 GB of RAM, with additional disk space determined largely by the instrument and stored file sizes.The software was tested on Windows, RHEL, Ubuntu Linux, and Mac OS X systems listed in the paper.
- VI. SUMMARY: Compiled Matlab code, source code by request, installation instructions, and a full manual are available through the Horace website.Installation involves unzipping the download, adding its directory to the Matlab path, and running the initialization routine.