Mention submodule in README
[qpms.git] / qpms / quaternions.h
blobeed1ce879befc56ad3e87b796ec0d4d771361299
1 /*! \file quaternions.h
2 * \brief Quaternions and Wigner matrices
3 */
4 #ifndef QPMS_WIGNER_H
5 #define QPMS_WIGNER_H
7 #include "qpms_types.h"
8 #include "vectors.h"
9 #include "tiny_inlines.h"
11 /// Just some arbitrarily chosen "default" value for quaternion comparison tolerance.
12 #define QPMS_QUAT_ATOL (1e-10)
14 /// Conversion from the 4*double to the 2*complex quaternion.
15 // TODO is this really correct?
16 // I.e. do the axis from moble's text match this convention?
17 static inline qpms_quat_t qpms_quat_2c_from_4d (qpms_quat4d_t q) {
18 qpms_quat_t q2c = {q.c1 + I * q.ck, q.cj + I * q.ci};
19 return q2c;
22 /// Conversion from the 2*complex to the 4*double quaternion.
23 // TODO is this really correct?
24 // I.e. do the axis from moble's text match this convention?
25 static inline qpms_quat4d_t qpms_quat_4d_from_2c (qpms_quat_t q) {
26 qpms_quat4d_t q4d = {creal(q.a), cimag(q.b), creal(q.b), cimag(q.a)};
27 return q4d;
30 /// Quaternion multiplication.
31 /**
32 * \f[ (P Q)_a = P_a Q_a - \bar P_b Q_b, \f]
33 * \f[ (P Q)_b = P_b Q_a + \bar P_a Q_b. \f]
35 static inline qpms_quat_t qpms_quat_mult(qpms_quat_t p, qpms_quat_t q) {
36 qpms_quat_t r;
37 r.a = p.a * q.a - conj(p.b) * q.b;
38 r.b = p.b * q.a + conj(p.a) * q.b;
39 return r;
42 /// Quaternion addition.
43 static inline qpms_quat_t qpms_quat_add(qpms_quat_t p, qpms_quat_t q) {
44 qpms_quat_t r;
45 r.a = p.a+q.a;
46 r.b = p.b+q.b;
47 return r;
50 /// Quaternion substraction.
51 static inline qpms_quat_t qpms_quat_sub(qpms_quat_t p, qpms_quat_t q) {
52 qpms_quat_t r;
53 r.a = p.a-q.a;
54 r.b = p.b-q.b;
55 return r;
58 /// Exponential function of a quaternion \f$e^Q$\f.
59 static inline qpms_quat_t qpms_quat_exp(const qpms_quat_t q) {
60 const qpms_quat4d_t q4 = qpms_quat_4d_from_2c(q);
61 const double vn = sqrt(q4.ci*q4.ci + q4.cj*q4.cj + q4.ck *q4.ck);
62 const double ea = exp(q4.c1);
63 const double cv = vn ? (ea*sin(vn)/vn) : ea; // "vector" part common prefactor
64 const qpms_quat4d_t r4 = {ea * cos(vn), cv*q4.ci, cv*q4.cj, cv*q4.ck};
65 return qpms_quat_2c_from_4d(r4);
68 /// Quaternion scaling with a real number.
69 static inline qpms_quat_t qpms_quat_rscale(double s, qpms_quat_t q) {
70 qpms_quat_t r = {s * q.a, s * q.b};
71 return r;
74 // quaternion "basis"
75 /// Quaternion real unit.
76 static const qpms_quat_t QPMS_QUAT_1 = {1, 0};
77 /// Quaternion imaginary unit i.
78 static const qpms_quat_t QPMS_QUAT_I = {0, I};
79 /// Quaternion imaginury unik j.
80 static const qpms_quat_t QPMS_QUAT_J = {0, 1};
81 /// Quaternion imaginary unit k.
82 static const qpms_quat_t QPMS_QUAT_K = {I, 0};
84 /// Quaternion conjugation.
85 static inline qpms_quat_t qpms_quat_conj(const qpms_quat_t q) {
86 qpms_quat_t r = {conj(q.a), -q.b};
87 return r;
90 /// Quaternion norm.
91 static inline double qpms_quat_norm(const qpms_quat_t q) {
92 return sqrt(creal(q.a * conj(q.a) + q.b * conj(q.b)));
95 /// Test approximate equality of quaternions.
96 static inline bool qpms_quat_isclose(const qpms_quat_t p, const qpms_quat_t q, double atol) {
97 return qpms_quat_norm(qpms_quat_sub(p,q)) <= atol;
100 /// "Standardises" a quaternion to have the largest component "positive".
102 * This is to remove the ambiguity stemming from the double cover of SO(3).
104 static inline qpms_quat_t qpms_quat_standardise(qpms_quat_t p, double atol) {
105 //assert(atol >= 0);
106 double maxabs = 0;
107 int maxi = 0;
108 const double *arr = (double *) &(p.a);
109 for(int i = 0; i < 4; ++i)
110 if (fabs(arr[i]) > maxabs + atol) {
111 maxi = i;
112 maxabs = fabs(arr[i]);
114 if(arr[maxi] < 0) {
115 p.a = -p.a;
116 p.b = -p.b;
118 return p;
121 /// Test approximate equality of "standardised" quaternions, so that \f$-q\f$ is considered equal to \f$q\f$.
122 static inline bool qpms_quat_isclose2(const qpms_quat_t p, const qpms_quat_t q, double atol) {
123 return qpms_quat_norm(qpms_quat_sub(
124 qpms_quat_standardise(p, atol),
125 qpms_quat_standardise(q, atol))) <= atol;
128 /// Norm of the quaternion imaginary (vector) part.
129 static inline double qpms_quat_imnorm(const qpms_quat_t q) {
130 const double z = cimag(q.a), x = cimag(q.b), y = creal(q.b);
131 return sqrt(z*z + x*x + y*y);
134 /// Quaternion normalisation to unit norm.
135 static inline qpms_quat_t qpms_quat_normalise(qpms_quat_t q) {
136 double n = qpms_quat_norm(q);
137 return qpms_quat_rscale(1/n, q);
140 /// Logarithm of a quaternion.
141 static inline qpms_quat_t qpms_quat_log(const qpms_quat_t q) {
142 const double n = qpms_quat_norm(q);
143 const double imnorm = qpms_quat_imnorm(q);
144 if (imnorm != 0.) {
145 const double vc = acos(creal(q.a)/n) / imnorm;
146 const qpms_quat_t r = {log(n) + cimag(q.a)*vc*I,
147 q.b*vc};
148 return r;
150 else {
151 const qpms_quat_t r = {log(n), 0};
152 return r;
156 /// Quaternion power to a real exponent.
157 static inline qpms_quat_t qpms_quat_pow(const qpms_quat_t q, const double exponent) {
158 const qpms_quat_t qe = qpms_quat_rscale(exponent,
159 qpms_quat_log(q));
160 return qpms_quat_exp(qe);
163 /// Quaternion inversion.
164 /** \f[ q^{-1} = \frac{q*}{|q|}. \f] */
165 static inline qpms_quat_t qpms_quat_inv(const qpms_quat_t q) {
166 const double norm = qpms_quat_norm(q);
167 return qpms_quat_rscale(1./(norm*norm),
168 qpms_quat_conj(q));
171 /// Make a pure imaginary quaternion from a 3d cartesian vector.
172 static inline qpms_quat_t qpms_quat_from_cart3(const cart3_t c) {
173 const qpms_quat4d_t q4 = {0, c.x, c.y, c.z};
174 return qpms_quat_2c_from_4d(q4);
177 /// Make a 3d cartesian vector from the imaginary part of a quaternion.
178 static inline cart3_t qpms_quat_to_cart3(const qpms_quat_t q) {
179 const qpms_quat4d_t q4 = qpms_quat_4d_from_2c(q);
180 const cart3_t c = {q4.ci, q4.cj, q4.ck};
181 return c;
184 /// Rotate a 3-dimensional cartesian vector using the quaternion/versor representation.
185 static inline cart3_t qpms_quat_rot_cart3(qpms_quat_t q, const cart3_t v) {
186 q = qpms_quat_normalise(q);
187 //const qpms_quat_t qc = qpms_quat_normalise(qpms_quat_pow(q, -1)); // implementation of _pow wrong!
188 const qpms_quat_t qc = qpms_quat_conj(q);
189 const qpms_quat_t vv = qpms_quat_from_cart3(v);
190 return qpms_quat_to_cart3(qpms_quat_mult(q,
191 qpms_quat_mult(vv, qc)));
194 /// Versor quaternion from rotation vector (norm of the vector is the rotation angle).
195 static inline qpms_quat_t qpms_quat_from_rotvector(cart3_t v) {
196 return qpms_quat_exp(qpms_quat_rscale(0.5,
197 qpms_quat_from_cart3(v)));
200 /// Wigner D matrix element from a rotator quaternion for integer \a l.
202 * The D matrix are calculated using formulae (3), (4), (6), (7) from
203 * http://moble.github.io/spherical_functions/WignerDMatrices.html
205 complex double qpms_wignerD_elem(qpms_quat_t q, qpms_l_t l,
206 qpms_m_t mp, qpms_m_t m);
208 /// A VSWF representation element of the O(3) group.
210 * TODO more doc.
212 complex double qpms_vswf_irot_elem_from_irot3(
213 const qpms_irot3_t q, ///< The O(3) element in the quaternion representation.
214 qpms_l_t l, qpms_m_t mp, qpms_m_t m,
215 bool pseudo ///< Determines the sign of improper rotations. True for magnetic waves, false otherwise.
219 static inline int qpms_irot3_checkdet(const qpms_irot3_t p) {
220 if (p.det != 1 && p.det != -1) abort();
221 return 0;
224 /// Improper rotation multiplication.
225 static inline qpms_irot3_t qpms_irot3_mult(const qpms_irot3_t p, const qpms_irot3_t q) {
226 #ifndef NDEBUG
227 qpms_irot3_checkdet(p);
228 qpms_irot3_checkdet(q);
229 #endif
230 const qpms_irot3_t r = {qpms_quat_normalise(qpms_quat_mult(p.rot, q.rot)), p.det*q.det};
231 return r;
234 /// Improper rotation inverse operation.
235 static inline qpms_irot3_t qpms_irot3_inv(qpms_irot3_t p) {
236 #ifndef NDEBUG
237 qpms_irot3_checkdet(p);
238 #endif
239 p.rot = qpms_quat_inv(p.rot);
240 return p;
243 /// Improper rotation power \f$ p^n \f$.
244 static inline qpms_irot3_t qpms_irot3_pow(const qpms_irot3_t p, int n) {
245 #ifndef NDEBUG
246 qpms_irot3_checkdet(p);
247 #endif
248 const qpms_irot3_t r = {qpms_quat_normalise(qpms_quat_pow(p.rot, n)),
249 p.det == -1 ? min1pow(n) : 1};
250 return r;
253 /// Test approximate equality of irot3.
254 static inline bool qpms_irot3_isclose(const qpms_irot3_t p, const qpms_irot3_t q, double atol) {
255 return qpms_quat_isclose2(p.rot, q.rot, atol) && p.det == q.det;
258 /// Apply an improper rotation onto a 3d cartesian vector.
259 static inline cart3_t qpms_irot3_apply_cart3(const qpms_irot3_t p, const cart3_t v) {
260 #ifndef NDEBUG
261 qpms_irot3_checkdet(p);
262 #endif
263 return cart3_scale(p.det, qpms_quat_rot_cart3(p.rot, v));
266 // Some basic transformations with irot3 type
267 /// Identity
268 static const qpms_irot3_t QPMS_IROT3_IDENTITY = {{1, 0}, 1};
269 /// \f$ \pi \f$ rotation around x axis.
270 static const qpms_irot3_t QPMS_IROT3_XROT_PI = {{0, I}, 1};
271 /// \f$ \pi \f$ rotation around y axis.
272 static const qpms_irot3_t QPMS_IROT3_YROT_PI = {{0, 1}, 1};
273 /// \f$ \pi \f$ rotation around z axis.
274 static const qpms_irot3_t QPMS_IROT3_ZROT_PI = {{I, 0}, 1};
275 /// Spatial inversion.
276 static const qpms_irot3_t QPMS_IROT3_INVERSION = {{1, 0}, -1};
277 /// yz-plane mirror symmetry
278 static const qpms_irot3_t QPMS_IROT3_XFLIP = {{0, I}, -1};
279 /// xz-plane mirror symmetry
280 static const qpms_irot3_t QPMS_IROT3_YFLIP = {{0, 1}, -1};
281 /// xy-plane mirror symmetry
282 static const qpms_irot3_t QPMS_IROT3_ZFLIP = {{I, 0}, -1};
284 /// versor representing rotation around z-axis.
285 static inline qpms_quat_t qpms_quat_zrot_angle(double angle) {
286 qpms_quat_t q = {cexp(I*(angle/2)), 0};
287 return q;
290 /// versor representing rotation \f$ C_N^k \f$, i.e. of angle \f$ 2\pi k / N\f$ around z axis.
291 static inline qpms_quat_t qpms_quat_zrot_Nk(double N, double k) {
292 return qpms_quat_zrot_angle(2 * M_PI * k / N);
295 /// Rotation around z-axis.
296 static inline qpms_irot3_t qpms_irot3_zrot_angle(double angle) {
297 qpms_irot3_t q = {qpms_quat_zrot_angle(angle), 1};
298 return q;
301 /// Rotation \f$ C_N^k \f$, i.e. of angle \f$ 2\pi k / N\f$ around z axis.
302 static inline qpms_irot3_t qpms_irot3_zrot_Nk(double N, double k) {
303 return qpms_irot3_zrot_angle(2 * M_PI * k / N);
306 #endif //QPMS_WIGNER_H