Vector-polarisation mode
Synthetic EMU-style three-axis polarisation projections overlaid in the central plot. The P_z trace carries the dominant slow exponential decay, P_x carries a weak transverse oscillation, and P_y is centred near zero with statistical noise — the standard signature of a sample whose local field is aligned along the z axis of the spectrometer.
Vector-polarisation mode treats the muon-spin polarisation as a three-component vector, exposing the \(P_x\), \(P_y\), and \(P_z\) projections separately rather than collapsing the detector counts onto a single forward/backward asymmetry. This is the right analysis path for anisotropic single crystals — where the precession axis is set by the crystallography rather than by the spectrometer geometry — and for any measurement where the local field at the muon site is canted away from \(\hat{z}\), since the off-axis precession is then carried by the \(P_x\) and \(P_y\) components and is lost in a one-dimensional asymmetry. Powder samples have an orientational average that already collapses onto a single non-trivial component along \(\hat{z}\), so the ordinary F-B asymmetry workflow is sufficient there. EMU’s octant geometry is the canonical example; vector mode is activated automatically when grouping names contain the canonical vector pairs:
Pz Forward/Pz Backwardfrom forward/backward detector groupsPy Top/Py Bottom(orPy Up/Py Down) from top/bottom detector groupsPx Left/Px Rightfrom left/right detector groups
so the same vector workflow can be applied to any instrument whose detector layout supports the same six-group naming convention.
Setup
Open Grouping and choose (or create) a profile for EMU.
Open Detector Layout…
Select instrument
EMU.Apply the
Vector Polarizationpreset.Return to Grouping and calibrate each axis’s alpha (see below).
Per-axis alpha
When vector mode is active, the Grouping window switches to a vector table with separate alpha values for each axis:
alpha_xalpha_yalpha_z
Each axis is its own asymmetry projection (see Detector Grouping and Layout, “Per-projection alpha”), so each calibrates independently through the same alpha calibration dialog used for an ordinary two-group run — open it per axis with that row’s Calibrate… button, or calibrate all three in one action with Estimate All alpha. Because each axis’s alpha and provenance are stored on its own projection, switching between axes never mixes up which value came from which calibration run.
Backwards compatibility:
Existing scalar
alphais still supported.Older projects are migrated so vector groupings initialise
alpha_x,alpha_y, andalpha_zfrom scalaralpha.
Display in the main plot
The plot toolbar shows one chip per projection — P_x, P_y and
P_z, each in its own tint. Click a chip to show or hide that projection;
selecting more than one stacks them as subplots, and all selects every
projection at once. At least one chip always stays selected. A run whose
grouping has a single forward/backward pair has only one polarisation
component, so it gets one plain plot and no chips. Each grouping profile
remembers its own chip selection: browse from a vector run showing P_x
and P_z to a single-pair run and back, and both subplots return, while a
run on another profile keeps the selection last made there. A run released
from its profile remembers its own, and an overlay of runs from several
profiles remembers that combination. Renaming a profile keeps its
selection, and the remembered selections are saved with the project.
The chips never make the plot wider. When the toolbar is too narrow for the
full names they switch to short ones (a detector pair such as Top-Bottom
becomes T–B; the full name stays in the tooltip). When even those do not
fit, the chips fold into a single button that summarises the selection — the
selected projection’s short name (P_x ▾), 2 of 3 ▾, or All 3 ▾ —
and opens a menu of the projections with a Show all entry.
Alpha display behaviour:
Single-axis views show the alpha for the selected axis.
Views with several projections hide alpha in the header.
Runs from different groupings
Overlaying runs whose groupings differ shows every projection any of them
declares, and each subplot holds only the runs that measure that projection.
A single-pair run joins the projection whose forward and backward groups hold
exactly its detectors on the same instrument: an EMU Longitudinal run sits
on the P_z subplot beside the vector runs and is absent from P_x and
P_y; a GPS Longitudinal run joins the WEP (spin-rotated) FB
projection. A run keeps one trace colour on every subplot. When a single-pair
run matches no projection (a MuSR Longitudinal run beside the
Transverse (Vector) grouping, whose detector split differs), the overlay
falls back to one plain plot without chips.
A fit on a single-pair run always belongs to the run’s own asymmetry, even when it is selected from a shared projection subplot.
The rotating frame
With both transverse projections on show, a Lab | Rotating switch sits to
the left of the chips. Rotating turns \(P_x\) and \(P_y\) exactly
into a frame rotating at the RF generator’s frequency about
\(B_0 \parallel z\), relabelling the chips P′_x and P′_y and leaving
P_z unchanged, which is how an RF nutation or spin-locking experiment is
read and fitted. The frame’s setup and per-run fields, Auto-detect…, the
handling of two-period runs and fitting of the rotated curves are described in
Rotating-frame projection.
Persistence
Per-axis alpha values are persisted in:
project files (schema v4+)
dataset grouping state
This preserves axis-specific alpha values across save/load cycles and across axis switching in vector mode. Each run’s rotating frame, when it has one, is saved on the run as well (schema v26; see Project files).
Transverse-field dual grouping
The same projection workflow generalises beyond EMU’s three-axis vector mode.
A forward/backward asymmetry is the muon polarisation projected onto the axis
joining that detector pair, so any preset that exposes more than one such pair is
a set of projections. MuSR and HiFi each ship a combined Transverse (Vector)
preset that exposes two transverse projections of the same run:
MuSR —
Top-BottomandFwd-BackHiFi —
Left-RightandTop-Bottom
Apply it from Detector Layout exactly as for EMU vector mode (select the
instrument, apply the Transverse (Vector) preset). The projection chip bar
then shows one chip per transverse projection; selecting two stacks them as
subplots, and clicking a subplot makes it the fit target with its own
per-projection single fit — identical behaviour to the EMU \(P_x\)/
\(P_y\)/\(P_z\) projections, including the tinted Fitting: <label>
echo and save/load persistence. Unlike EMU’s octant model, the two transverse
pairs use four distinct detector groups so both coexist (the legacy split
presets reused the same group IDs and were mutually exclusive).
Detector group composition
EMU vector mode follows octant-style detector composition. The detector layout reference in Detector Grouping and Layout should be used for instrument-consistent verification of assigned groups.
EMU has no facility-documented “vector-polarisation” grouping — the EMU User
Guide and the Mantid EMU instrument definition only describe the physical
detector numbering (Section 8.1), not a Px/Py/Pz preset. The Vector
Polarization preset is therefore an Asymmetry construct: it is verified
internally consistent (each octant selection matches the geometric
half-plane of the layout’s own detector angles) but is not itself a
published EMU convention.
Assumptions and limitations
The vector treatment rests on a small set of geometry assumptions; read the projections with them in mind:
Each projection is a forward/backward asymmetry along one detector-pair axis. A forward/backward asymmetry measures the muon polarisation projected onto the axis joining that detector pair, so \(P_x\), \(P_y\), and \(P_z\) are the three orthogonal projections only insofar as the three detector-pair axes are mutually orthogonal and aligned with the spectrometer frame. On a real instrument the octant groups approximate those axes; the reconstruction is exact only for an idealised orthogonal layout.
Calibration and counting independence are distinct. Each axis retains its own alpha, but projections can intentionally reuse detectors across disjoint F/B pairs. Their counting covariance depends on the retained acquisition bins, rates, corrections, and reference reuse. The rotating-frame operation propagates that covariance while keeping calibration and frame errors conditional; see Rotating-frame projection. Shared-background noise also survives display packing. Corrected marginal errors do not certify an independent joint-projection or overlapping-group MaxEnt objective.
Powder samples do not need it. An orientational average collapses the polarisation onto a single non-trivial component along \(\hat{z}\), so the ordinary F–B asymmetry workflow is sufficient; the vector projections add information only when the local field is canted away from \(\hat{z}\), as in an oriented single crystal.
The EMU Px/Py/Pz preset is an Asymmetry construct, not a published EMU convention (see Detector Group Composition): it is verified internally consistent against the layout’s own detector angles, but should be cross-checked against the facility detector numbering before quantitative use.
Compared with other programs
Mantid RRFMuon v1 rotates two supplied transverse spectra after matching axis, phase, and unit conventions. WiMDA 1.326 and musrfit 1.12.1 scalar rotating-frame workflows use different inputs and smoothing or packing. The shared-counting propagation for Asymmetry’s overlapping EMU ratios is justified by retained observation reuse, not verified reference-program error equivalence. See Rotating-frame projection for equations, sources, and evidence limits.
References
S. J. Blundell, R. De Renzi, T. Lancaster, and F. L. Pratt, Muon Spectroscopy: An Introduction (Oxford University Press, Oxford, 2022) — detector geometry and the polarisation-projection observable.