CalculateUMatrix v1

../_images/CalculateUMatrix-v1_dlg.png

CalculateUMatrix dialog.

Table of Contents

Summary

Calculate the U matrix from a peaks workspace, given lattice parameters.

Properties

Name Direction Type Default Description
PeaksWorkspace InOut PeaksWorkspace Mandatory An input workspace.
a Input number Mandatory Lattice parameter a
b Input number Mandatory Lattice parameter b
c Input number Mandatory Lattice parameter c
alpha Input number Mandatory Lattice parameter alpha
beta Input number Mandatory Lattice parameter beta
gamma Input number Mandatory Lattice parameter gamma

Description

Given a set of peaks (Q in the goniometer frame, HKL values), and given lattice parameters (a,b,c,\alpha,\beta,\gamma), it will try to find the U matrix, using least squares approach and quaternions 1. Units of length are in in \rm \AA, angles are in degrees.

The algorithm calculates first the B matrix according to Busing and Levi.

Given a set of peaks in the reference frame of the inner axis of the goniometer, \rm Q_{gon}, indexed by (h_i, k_i, l_i), we want to find the U matrix that maps peaks in the reciprocal space of the sample to the peaks in the goniometer frame

(1)\rm U \rm B \left(
                            \begin{array}{c}
                              h_i \\
                              k_i \\
                              l_i \\
                            \end{array}
                          \right) = \rm Q_{gon,i}

For simplicity, we define

(2)\rm Q_{hkl,i} = \rm B \left(
                            \begin{array}{c}
                              h_i \\
                              k_i \\
                              l_i \\
                            \end{array}
                          \right)

In the real world, such a matrix is not always possible to find. Therefore we just try minimize the difference between the two sets of p

(3)\sum_i |\rm U \rm Q_{hkl,i} - \rm Q_{gon,i}|^2 = \sum_i \left(|\rm U \rm Q_{hkl,i}|^2 + |\rm Q_{gon,i}|^2 -2 \rm U \rm Q_{hkl,i} \cdot \rm Q_{gon,i}\right)

In equation (3), \left|\rm U \rm Q_{hkl,i}\right|^2 = |\rm Q_{hkl,i}|^2, so the first two terms on the left side are U independent. Therefore we want to maximize

(4)\sum_i \left(\rm U \rm Q_{hkl,i} \cdot \rm Q_{gon,i}\right)

We are going to write the scalar product of the vectors in terms of quaternions 2. We define q_{hkl,i} = \left(0, Q_{hkl,i}\right), q_{gon,i} = \left(0, Q_{gon,i}\right) and the rotation U is described by quaternion u = \left(w,x,y,z\right)

Then equation (4) will be written as

(5)\sum_i \left(\rm U \rm Q_{hkl,i} \cdot \rm Q_{gon,i}\right) = 0.5 \cdot \left(u q_{hkl,i} u^*\right) q_{gon,i}\ + 0.5 \cdot q_{gon,i} \left(u q_{hkl,i} u^*\right)

We define matrices

(6)H_i= \left(\begin{array}{cccc}
           0 & -q_{hkl,i,x} & -q_{hkl,i,y} & -q_{hkl,i,z} \\
           q_{hkl,i,x} & 0 & q_{hkl,i,z} & -q_{hkl,i,y} \\
           q_{hkl,i,y} & -q_{hkl,i,z} & 0 & q_{hkl,i,x} \\
           q_{hkl,i,z} & q_{hkl,i,y} & -q_{hkl,i,x} & 0
       \end{array} \right)

and

(7)S_i= \left(\begin{array}{cccc}
           0 & -q_{gonl,i,x} & -q_{gon,i,y} & -q_{gon,i,z} \\
           q_{gon,i,x} & 0 & -q_{gon,i,z} & q_{gon,i,y} \\
           q_{gon,i,y} & q_{gon,i,z} & 0 & -q_{gon,i,x} \\
           q_{gon,i,z} & -q_{gon,i,y} & q_{gon,i,x} & 0
      \end{array} \right)

Then, we can rewrite equation (5) using matrices 3, 4:

(8)\sum_i \left(\rm U \rm Q_{hkl,i} \cdot \rm Q_{gon,i}\right) = \left(\begin{array}{cccc}
                                                               w & x & y & z\end{array} \right)  \sum_i H_i S_i \left(
                            \begin{array}{c}
                              w \\
                              x \\
                              y \\
                              z
                            \end{array}
                            \right)

The problem of finding \left(w,x,y,z\right) that maximizes the sum can now be rewritten in terms of eigenvectors of HS= \sum_i \left(H_i S_i\right) . Let \epsilon_j and \nu_j be the eigenvalues and corresponding eigenvectors of HS, with \epsilon_0 > \epsilon_1 > \epsilon_2 > \epsilon_3. We can write any vector (w,x,y,z) as a linear combination of the eigenvectors of HS:

(9)\left(w,x,y,z\right) = \delta_0 \nu_0 +\delta_1 \nu_1 +\delta_2 \nu_2 +\delta_3 \nu_3

(10)\left(\begin{array}{cccc}
       w & x & y & z\end{array} \right)  HS \left(
                             \begin{array}{c}
                               w \\
                               x \\
                               y \\
                               z
                             \end{array}
                           \right) = \delta_0^2 \nu_0 HS \nu_0 + \delta_1^2 \nu_1 HS \nu_1 +\delta_2^2 \nu_2 HS \nu_2 +\delta_3 \nu_3 HS \nu_3

(11)\begin{split}
                                &  = \delta_0^2 \epsilon_0 + \delta_1^2 \epsilon_1 +\delta_2^2 \epsilon_2 +\delta_3 ^2 \epsilon_3
\end{split}

where u is a unit quaternion, \delta_0^2  + \delta_1^2 +\delta_2^2 +\delta_3 ^2=1 (12)

Then the sum in equation (11) is maximized for \epsilon_0 =1, \epsilon_1 =0,  \epsilon_2 =0 \epsilon_3 =0

Therefore U is the rotation represented by the quaternion u, which is the eigenvector corresponding to the largest eigenvalue of HS.

Categories: Algorithms | Crystal