1: \subsection{Projections of reciprocal space}
2: \label{sec:projections}
3: Based on a data-set that have been rebinned to reciprocal space,
4: projections are trivially done as a sum over voxels in the appropriate
5: directions.
6:
7:
8: The general ``first order rebinning'' based projections are performed
9: in the following way:
10: \begin{itemize}
11: \item Define the bins of the final projected intensity distribution:\\
12: The bins are defined using the average pixel position in the
13: selected region, in the $\ve x$ or $\ve y$-directions for the
14: projections onto the radial direction or azimuthal plane
15: respectively.
16: \item For all voxels in the data set, calculate the $\ve q$-position.
17: \item For all voxels find the closest bin in the final projection,
18: add the intensity to this bin, and record that a voxel has been
19: added to the bin.
20: \item For each bin in the projection normalize by the number of added
21: voxels to that bin.
22: \end{itemize}
23:
24: \enlargethispage{1cm}
25:
26: \subsubsection{Azimuthal projection}
27: \label{sec:QxQzProj}
28: Projections onto the azimuthal plane have been done in two ways; the
29: proper projection based on first order rebinning as described
30: above, and an even simpler projection, based directly on the detector pixels.
31:
32: The latter case is based on the fact that the horizontal plane of the
33: detector is approximately in the $\ve q_x$-direction, hence a good
34: approximation to the projection onto the azimuthal direction is
35: obtained by summing over the columns of the detector pixels.
36:
37: In the $\ve q_x$-direction this is very accurate, as the change in
38: $\ve q_x$-position of the pixels over one column is much smaller than
39: the bin size of the final projection. However, the $\ve q_z$-position
40: of a pixel depends directly on the $2\theta$ position of the pixel
41: (according to equation \ref{eq:qSpaceApprox}) and it is found that
42: by just summing over columns some voxels might end up in the wrong bin
43: in the projection. The error is however minor, and no significant
44: differences are observed between the two projection types.
45:
46: Generally the proper projection has been used for the large azimuthal
47: projections (as the one in figure 3 in \mycitet{acta}), and the simple
48: projection for small projections (as the ones presented e.g. for
49: reproducibility in section \ref{sec:Reproducibility}).
50:
51: \subsubsection{Radial projection}
52: \label{sec:QyProj}
53: All integrated radial peak profiles reported in this study (as the one
54: in figure \ref{fig:RadialProfile}) have been calculated using the
55: first order rebinning technique.
56:
57: The radial peak profiles are normally fitted (by least square fitting)
58: using either a split pseudo-Voigt or normal pseudo-Voigt function. The
59: pseudo-Voigt function is a weighted sum of a Gaussian and Lorentzian
60: function (for a discussion of this function see \citep{Enzo1999}).
61: The normal pseudo-Voigt function has one parameter describing the
62: width of the peak (in the version used the parameter is the full width
63: at half maximum). In contrast the split pseudo-Voigt function has two
64: width parameters, one giving the half width of the lower (with respect
65: to the maximum) part, $\Delta_1$, and one giving the half width of the
66: upper part, $\Delta_2$ (see figure \ref{fig:split_speudo_voigh}). This
67: can therefore be used for fitting asymmetric peaks. The two width
68: parameters for the split pseudo-Voigt are normally added
69: ($\Delta_1+\Delta_2$) to give the full width at half maximum, and
70: subtracted ($\Delta_1-\Delta_2$) to give a measure of the asymmetry.
71:
72: As quantitative results are derived from the radial peak profiles, it
73: is important to know if the first order rebinning scheme distorts the
74: peak profile.
75:
76:
77: \begin{figure}
78: \centering
79: \input{AnalysisMethods/figs/split_pseudo_voigt_all.pstex_t}
80: \caption{Illustration of the split pseudo-Voigt function. Indicated
81: are the two ``half width at half maximum'' parameters ($\Delta_1$
82: and $\Delta_2$), maximum position ($x_{\text{max}}$), maximum
83: value ($I_{\text{max}}$), and half maximum ($I_{\text{max}}/2$).}
84: \label{fig:split_speudo_voigh}
85: \end{figure}
86:
87: To investigate this a detector image of a known radial peak profile
88: was calculated. The peak profile is assumed to be constant in the
89: azimuthal directions and to be described by a split pseudo-Voigt
90: function in the radial direction. Typical parameters for the geometry
91: of the setup and peak profile were used. To simplify the calculation
92: the parameters of the peak profile, were converted to angular units,
93: and the peak described as function of the scattering angle,
94: $f(2\theta)$. The intensity in each pixel, $I_{\text{pixel}}$, was
95: calculated as:
96: \begin{eqnarray}
97: I_{\text{pixel}}=\iint_{\text{pixel}}f(2\theta(x,y))\frac{\partial
98: (2\theta,\eta)}{\partial (x,y)}dx dy
99: \end{eqnarray}
100: where the functions $2\theta(x,y)$ and $\eta(x,y)$ are given by
101: equation \ref{eq:pix2ang}. The integration was carried out numerically.
102:
103: The simulated image was treated in the same way as the normal data, and
104: the simulated parameters and parameters
105: found by the fitting were compared:
106:
107: \begin{tabular}{l|p{0.4\textwidth}}
108: Parameter & Relative difference between fitted and simulated
109: parameter \\ \hline
110: Peak position & $1.5\E{-4}\%$\\
111: Full width at half maximum & $0.15\%$\\
112: Asymmetry & $0.7\%$
113: \end{tabular}
114:
115: From this it is concluded that the first order rebinning scheme does
116: not introduce any significant errors in the relevant parameters of the
117: resulting radial peak profile.
118:
119:
120: