Momentum and energy dependent resolution function of the ARCS neutron chopper spectrometer at high momentum transfer: Comparing simulation and experiment

Inelastic neutron scattering at high momentum transfers (i.e. Q≥20A˚), commonly known as deep inelastic neutron scattering (DINS), provides direct observation of the momentum distribution of light atoms, making it a powerful probe for studying single-particle motions in liquids and solids. The quantitative analysis of DINS data requires an accurate knowledge of the instrument resolution function R_i(Q,E) at each momentum Q and energy transfer E, where the label i indicates whether the resolution was experimentally observed i=obs or simulated i=sim. Here, we describe two independent methods for determining the total resolution function R_i(Q,E) of the ARCS neutron instrument at the Spallation Neutron Source, Oak Ridge National Laboratory. The first method uses experimental data from an archetypical system (liquid ⁴He) studied with DINS, which are then numerically deconvoluted using its previously determined intrinsic scattering function to yield R_obs(Q,E). The second approach uses accurate Monte Carlo simulations of the ARCS spectrometer, which account for all instrument contributions, coupled to a representative scattering kernel to reproduce the experimentally observed response S(Q,E). Using a delta function as scattering kernel, the simulation yields a resolution function R_sim(Q,E) with comparable lineshape and features as R_obs(Q,E), but somewhat narrower due to the ideal nature of the model. Using each of these two R_i(Q,E) separately, we extract characteristic parameters of liquid ⁴He such as the intrinsic linewidth α₂ (which sets the atomic kinetic energy 〈K〉∼α₂) in the normal liquid and the Bose–Einstein condensate parameter n₀ in the superfluid phase. The extracted α₂ values agree well with previous measurements at saturated vapor pressure (SVP) as well as at elevated pressure (24 bars) within experimental precision, independent of which R_i(Q,y) is used to analyze the data. The actual observed n₀ values at each Q vary little with the model R_i(Q,E), and the effective Q-averaged n₀ values are consistent with each other, and with previously reported values.