Abstract
The log angles of a rotation matrix are three independent elements of the logarithm of the rotation matrix. Nye’s lattice curvature tensor κij is discussed by using the log angles. For the change in a crystal orientation ΔR with the change in a position Δxi, it is shown that the elements of κij are written as κij = Δωi/Δxj using the log angles Δωi of ΔR. The log angles for the crystal rotation given by the axis/angle pair are also discussed.
This Paper was Originally Published in Japanese in J. Japan Inst. Met. Mater. 82 (2018) 415–418.
1. Introduction
To obtain an essential understanding of the strength of materials, it is necessary to know development and non-uniformity of plastic deformation in crystals, which are caused by motion and arrangement of dislocations. Since orientations of crystals change as a function of position when dislocation structures are formed in the crystals, efforts have been made to show dislocation structures after plastic deformation by considering such orientation changes. Pioneering work of the efforts is the theoretical study by Nye.1) Nye has shown that the dislocation density in a crystal can be quantitatively treated when orientation changes in the crystal are evaluated by using the lattice curvature tensor $\boldsymbol{{\kappa}}$. Using the x1 − x2 − x3 orthogonal coordinate system, this lattice curvature tensor $\boldsymbol{{\kappa}}$ describes the orientation change δϕi around the xi axis with the change in a position δxj in the crystal as κij = δϕi/δxj.
As shown by Nye,1) the theoretical framework to discuss dislocation structures using orientation changes in crystal grains have been presented about half a century ago. On the other hand, experimental measurements of changes in crystal orientations for a wide area, with a high accuracy and in a short time have been performed recently only after a rapid development of SEM/EBSD (scanning electron microscopy/electron back scattering diffraction) method. However, the procedure to determine values of κij = δϕi/δxj from data obtained by SEM/EBSD method has not been shown clearly in previous studies. When the orientation change in a crystal is given by the rotation angle ΔΦ around the unit vector ni, Pantleon2) and He et al.3) have shown that the product ΔΦni of ni and ΔΦ is also a vector and the components ΔΦni are the components of the rotation angle around the xi axes. However, it should be noted that, in successive rotations around different axes such as the Euler angles, resultant rotations generally depend on the order of the rotations even if the rotation angle around each axis is the same.4) This means that it is not obvious that the components of ΔΦni are the components of rotation around the xi axes, but the reason is not shown in their papers.2,3) It is hence significant to show the reason if the components of ΔΦni can be treated as the components of the rotation angle around the xi axes.
A crystal orientation can be described by using a rotation matrix R with respect to a reference coordinate system. The matrix R is a 3 × 3 orthogonal matrix whose determinant is 1. The logarithm ln R of R is a 3 × 3 skew-symmetric matrix with three independent elements of real numbers.4–7) The logarithm ln R of R has been discussed in previous studies to discuss crystal orientations.4,6–10) The three elements of ln R are called the log angles ωi by Hayashi et al.4) and Onaka and Hayashi,9,10) which are utilized as the characteristic angles of R. They have discussed geometrical meanings of the log angles ωi and applied these to the analysis of crystal rotations.4,9,10)
When a crystal orientation changes as much as ΔR with the change in a position Δxi, the elements of the lattice curvature tensor $\boldsymbol{{\kappa}}$ are written as κij = Δωi/Δxj using the log angles Δωi of ΔR. We will show this in the present paper. This knowledge enables us to understand the procedure to determine the lattice curvature tensor $\boldsymbol{{\kappa}}$ from data of crystal orientations as those given by the SEM/EBSD method. Measurements of the lattice curvature tensor $\boldsymbol{{\kappa}}$ for plastically deformed metals and evaluation of densities or structures of dislocations from an experimental point of view will be shown in our future work based on the present study.
2. Log Angles
2.1 Logarithmic function
Here we explain characteristics of the logarithmic function before summarizing the log angles. For a real number x, the relationship between x and the exponential function exp x is well-known as11)
\begin{equation}
\exp x = 1 + \frac{x^{2}}{2!} + \frac{x^{3}}{3!} + \frac{x^{4}}{4!} + \ldots = \lim_{p \to \infty}\left(1 + \frac{x}{p}\right)^{p}.
\end{equation}
| (1) |
Using
y = exp
x (
y > 0), this relationship is rewritten by the logarithmic function as
\begin{equation}
y = \lim\limits_{p \to \infty}\left(1 + \frac{\ln y}{p}\right)^{p}.
\end{equation}
| (2) |
When
N is a sufficiently large positive integer, we have
\begin{equation}
y \approx \left(1 + \frac{\ln y}{N}\right)^{N}.
\end{equation}
| (3) |
This equation gives us better understanding on variables describing various changes.
Figure 1 shows the schematic illustration of deformation of a rod-like material along the longitudinal direction. The upper figure shows the state before deformation with the initial length L0 and the state after uniform deformation with the length L = L0 + ΔL is shown in the lower one. The nominal strain e is defined as e = ΔL/L0 and the definition of the stretch λ showing the length ratio before and after deformation is
\begin{equation}
\lambda = L/L_{0} = (L_{0} + \Delta L)/L_{0}.
\end{equation}
| (4) |
This λ of λ > 0 is written as λ = 1 +
e.
From (3), we have
\begin{equation}
\lambda\approx \left(1 + \frac{\ln\lambda}{N}\right)^{N}
\end{equation}
| (5a) |
and this equation is rewritten from
(4) as
\begin{equation}
L \approx L_{0}\left(1 + \frac{\ln\lambda}{N}\right)^{N}.
\end{equation}
| (5b) |
Since the logarithmic strain ε is
\begin{equation}
\varepsilon = \ln(1 + e) = \ln\lambda,
\end{equation}
| (6) |
using a sufficiently large integer
N, from
(6) and
(5b), the relationship among
L0,
L and ε is given by
\begin{equation}
L \approx L_{0}\left(1 + \frac{\varepsilon}{N}\right)^{N}.
\end{equation}
| (7) |
This
eq. (7) shows a characteristic of the logarithmic strain ε as an appropriate measure to describe large deformation. Even for large deformation where |ε| ≪ 1 is not satisfied, if we divide the large deformation into small deformation of
N-times, we can reasonably treat the large deformation as an extension of the concept of infinitesimal deformation. That is to say, ε = (ε/
N) ×
N and this ε is understood as the sum of
N-times repeated small strain ε/
N. This is an interpretation of the logarithmic strain ε given by
(7). A similar interpretation can be put on
(5a): the stretch λ is the quantity describing the deformation as it occurred, but its logarithm ln λ is a measure to show the extent of the deformation.
2.2 Outline of log angles
The relationship between the 3 × 3 orthogonal matrix R and its logarithm ln R is written as12)
\begin{equation}
\mathbf{R} = \lim\limits_{p \to \infty}\left(\mathbf{E} + \frac{\ln\mathbf{R}}{p}\right)^{p},
\end{equation}
| (8) |
where
E is the 3 × 3 unit matrix. This is given from
(2) by expanding the concept of numbers to matrices. The logarithm ln
R is the skew-symmetric matrix with elements of real numbers.
5–7,12) The log angles ω
i of
R are the elements of ln
R and are written as
4,9,10)
\begin{equation}
\ln\mathbf{R} =
\begin{pmatrix}
0 & -\omega_{3} & \omega_{2}\\
\omega_{3} & 0 & -\omega_{1}\\
-\omega_{2} & \omega_{1} & 0
\end{pmatrix}
.
\end{equation}
| (9) |
A calculation method to obtain ln
R from
R will be shown later in this paper.
As well as the discussion made in 2.1, using a sufficiently large positive integer N, from (8) we have
\begin{equation}
\mathbf{R} \approx \left(\mathbf{E} + \frac{\ln\mathbf{R}}{N}\right)^{N}.
\end{equation}
| (10) |
This equation can be rewritten from
(9) as
\begin{equation}
\mathbf{R} \approx (\delta\mathbf{R})^{N},
\end{equation}
| (11a) |
where
\begin{equation}
\delta\mathbf{R} = \mathbf{E} + \left(\frac{\ln\mathbf{R}}{N}\right) =
\begin{pmatrix}
1 & -(\omega_{3}/N) & (\omega_{2}/N)\\
(\omega_{3}/N) & 1 & -(\omega_{1}/N)\\
-(\omega_{2}/N) & (\omega_{1}/N) & 1
\end{pmatrix}
.
\end{equation}
| (11b) |
From
(11a) and
(11b), the meanings of the log angles ω
i are understood as well as the case of the logarithmic strain ε. The third side of
(11b) shows that δ
R is a small-angle rotation matrix since |ω
i/
N| ≪ 1 is satisfied for the sufficiently large
N.
4) This δ
R is interpreted as the result of three successive rotations composed of the angle ω
1/
N around the
x1 axis, the angle ω
2/
N around the
x2 axis and the angle ω
3/
N around the
x3 axis.
4) Since the rotation angles around the axes are small, any orders of the successive rotations give the same result given by the third side of
(11b) by neglecting the second and higher terms of the elements.
4) This shows that the log angles ω
i are treated as the characteristic angles and the components of
R in the sense that the angles ω
i are the sum of the divided rotation angles around the
xi axes.
4)
3. Change of Orientation with Change of Position in Crystal
As shown in Fig. 2, we assume that a crystal orientation changes as much as ΔR with the change of a position from xi to xi + Δxi. Moreover we assume that the change ΔR is written as ΔR ≈ (δR)N and ΔR is given by the N-time successive rotations of δR. This treatment means that the change of orientation between the positions xi and xi + Δxi is assumed to be uniform and δR corresponds to the orientation change with the position change as much as δxi = Δxi/N between xi and xi + Δxi.
When the small-angle rotation δR is composed of the rotation angles δϕi around the xi axes as shown in Fig. 2, using the log angles Δωi of ΔR, from (9) and (11b) we have
\begin{equation}
\delta\phi_{i} = \Delta\omega_{i}/N.
\end{equation}
| (12) |
Hence the relationship between the changes δϕ
i and δ
xi is given by
\begin{equation}
\delta\phi_{i}/\delta x_{j} = \Delta\omega_{i}/\Delta x_{j}.
\end{equation}
| (13) |
This relationship means that using Δω
i of Δ
R with the position change from
xi to
xi + Δ
xi, the average lattice curvature tensor
$\boldsymbol{{\kappa}}$ for the region is given by
\begin{equation}
\kappa_{ij} = \Delta\omega_{i}/\Delta x_{j},
\end{equation}
| (14a) |
which is rewritten as
\begin{equation}
\boldsymbol{{\kappa}} =
\begin{pmatrix}
\Delta\omega_{1}/\Delta x_{1} & \Delta\omega_{1}/\Delta x_{2} & \Delta\omega_{1}/\Delta x_{3}\\
\Delta\omega_{2}/\Delta x_{1} & \Delta\omega_{2}/\Delta x_{2} & \Delta\omega_{2}/\Delta x_{3}\\
\Delta\omega_{3}/\Delta x_{1} & \Delta \omega_{3}/\Delta x_{2} & \Delta\omega_{3}/\Delta x_{3}
\end{pmatrix}
.
\end{equation}
| (14b) |
4. Log Angles for Rotation Matrix Given by the Axis/Angle Pair
The matrix R for the rotation given by the rotation angle Φ (0 ≤ Φ ≤ π) around the unit vector ni is written as9,13)
\begin{equation}
\mathbf{R} =
\begin{pmatrix}
(1 - n_{1}{}^{2})\cos\Phi + n_{1}{}^{2} & n_{1}n_{2}(1 - \cos\Phi) - n_{3}\sin\Phi & n_{3}n_{1}(1 - \cos\Phi) + n_{2}\sin\Phi\\
n_{1}n_{2}(1 - \cos\Phi) + n_{3}\sin\Phi & (1 - n_{2}{}^{2})\cos\Phi + n_{2}{}^{2} & n_{2}n_{3}(1 - \cos\Phi) - n_{1}\sin\Phi\\
n_{3}n_{1}(1 - \cos\Phi) - n_{2}\sin\Phi & n_{2}n_{3}(1 - \cos\Phi) + n_{1}\sin\Phi & (1 - n_{3}{}^{2})\cos\Phi + n_{3}{}^{2}
\end{pmatrix}
.
\end{equation}
| (15) |
This is the representation of
R by using the axis/angle pair,
ni and Φ. When the rotation angle is small and satisfies |δΦ| ≪ 1, we have
\begin{equation*}
\sin\delta\Phi \approx \delta\Phi\ \text{and}\ \cos\delta\Phi \approx 1.
\end{equation*}
|
Using these approximations, the rotation matrix δ
R of δΦ around
ni is written as
\begin{equation}
\delta\mathbf{R} \approx
\begin{pmatrix}
1 & -\delta\Phi n_{3} & \delta\Phi n_{2}\\
\delta\Phi n_{3} & 1 & -\delta\Phi n_{1}\\
-\delta\Phi n_{2} & \delta\Phi n_{1} & 1
\end{pmatrix}
.
\end{equation}
| (16) |
When the relationship between Φ and δΦ is given by
\begin{equation}
\delta\Phi = \Phi/N,
\end{equation}
| (17) |
R of Φ around
ni is the same with the
N-times successive rotations of δ
R. Then, using the relationship
R ≈ (δ
R)
N and
(9),
(11b) and
(16), we have the logarithm ln
R of
R given by
(15) is written as
5,6)
\begin{equation}
\ln \mathbf{R} =
\begin{pmatrix}
0 & -\Phi n_{3} & \Phi n_{2}\\
\Phi n_{3} & 0 & -\Phi n_{1}\\
-\Phi n_{2} & \Phi n_{1} & 0
\end{pmatrix}
.
\end{equation}
| (18) |
This shows that the log angles ω
i of
R are
\begin{equation}
\omega_{i} = \Phi n_{i}.
\end{equation}
| (19) |
Equation (18) is satisfied even if the rotation angle Φ of
R is not small.
The above discussion shows that the log angles ωi of R with Φ around ni are given by Φni. Hence, as shown in the papers by Pantleon2) and He et al.,3) when the rotation angle is ΔΦ, the product ΔΦni can be treated as the components of the rotation angles around the xi axes in the sense that these are the sum of the divided rotation angles around the axes.
From (15) and (18), the relationship between the rotation matrix R and its logarithm ln R is given by
\begin{equation}
\ln\mathbf{R} = \frac{\Phi}{2\sin\Phi}(\mathbf{R} - {}^{t}\mathbf{R}),
\end{equation}
| (20) |
where
tR is the transposed matrix of
R. The proof of
(2) is shown in Ref.
14). The angle Φ is given by the tarce Tr
R of
R as
13)
\begin{equation}
\cos\Phi = (\text{Tr}\mathbf{R} - 1)/2.
\end{equation}
| (21) |
The logarithm of matrix is generally obtained by the procedure starting from the diagonalization of the matrix. Recent computation programs such as
Mathematica have a command to calculate the logarithm of matrix. However, for the logarithm ln
R of the rotation matrix
R,
(20) and
(21) are convenient to calculate the values of ln
R.
5. Conclusions
The log angles ωi of a rotation matrix R are three elements of the logarithm ln R of R. Using the log angles ωi, we have discussed the operation to obtain the lattice curvature tensor $\boldsymbol{{\kappa}}$ from position-dependent changes of crystal orientations in a grain. When the crystal orientation changes as much as ΔR with the change in the position Δxi, using the log angles Δωi of ΔR, the elements of the lattice curvature tensor changes $\boldsymbol{{\kappa}}$ are given by κij = Δωi/Δxj. When crystal-orientation change is given by the rotation angle ΔΦ around the unit vector ni, the log angles Δωi of this orientation change satisfy Δωi = ΔΦni. Both of Δωi and ΔΦni are considered to be the components of the rotation angle around the xi axes.
Acknowledgment
Funding from the JSPS KAKENHI (Grant Number JP16K06703) is gratefully acknowledged.
REFERENCES
- 1) J.F. Nye: Acta Metall. 1 (1953) 153–162.
- 2) W. Pantleon: Scr. Mater. 58 (2008) 994–997.
- 3) W. He, W. Ma and W. Pantleon: Mater. Sci. Eng. A 494 (2008) 21–27.
- 4) K. Hayashi, M. Osada, Y. Kurosu, Y. Miyajima and S. Onaka: J. Japan Inst. Metals Mater. 79 (2015) 9–15.
- 5) M. Saito: Senkeidaisu Nyumon (Introduction to Linear Algebra), (University of Tokyo Press, Tokyo, 1984) pp. 203–223.
- 6) J.E. Marsden and T.S. Ratiu: Introduction to Mechanics and Symmetry, Second edition, (Springer, New York, 1999) pp. 283–308.
- 7) N.R. Barton and P.R. Dawson: Model. Simul. Mater. Sci. Eng. 9 (2001) 433–463.
- 8) C.-S. Man, X. Gao, S. Godefroy and E.A. Kenik: Int. J. Plast. 26 (2010) 423–440.
- 9) S. Onaka and K. Hayashi: J. Math. Chem. 54 (2016) 1686–1695.
- 10) S. Onaka and K. Hayashi: Scanning 2017 (2017) 4893956.
- 11) K. Yoshida, K. Kitahara and T. Nishimura: Bibunsekibungaku no Kiso (Fundamentals of Calculus), (Rigaku Shoin, Tokyo, 2003) pp. 1–52.
- 12) B.C. Hall: Lie Groups, Lie Algebras, and Representations: An Elementary Introduction, (Springer, New York, 2003) pp. 27–62.
- 13) H. Grimmer: Acta Crystallogr. Sec. A 40 (1984) 108–112.
- 14) A. Morawiec: Orientations and Rotations, Computations in Crystallographic Textures, (Springer, New York, 2004) p. 63.