README.md

December 23, 2023 · View on GitHub

{% include head.html %}

Geometric Product

IdentityComment
ab=ab+abab = a \cdot b + a \wedge bGeometric product of two vectors aa and bb
ab=12(ab+ba)a \cdot b = \frac{1}{2}\left( ab + ba\right)Inner product of two vectors as symmetric part of geometric product
ab=12(abba)a \wedge b = \frac{1}{2}\left( ab - ba\right)Exterior product of two vectors as antisymmetric part of geometric product

Multivectors

IdentityComment
a1ar=1r!K(1)ϵak1akra_1 \wedge\dots\wedge a_r = \frac{1}{r!} \sum_{\mathcal{K}} (-1)^\epsilon a_{k_1} \dots a_{k_r}Exterior product of rr vectors. K\mathcal{K} is the set of permutations of $1 \dots randand\epsilon$ is even for even perms and odd for odd
a1a2a3=16(a1a2a3+a3a1a2+a2a3a1a1a3a2a2a1a3a3a2a1)a_1 \wedge a_2 \wedge a_3 = \frac{1}{6} \left( a_1a_2a_3+a_3a_1a_2+a_2a_3a_1-a_1a_3a_2-a_2a_1a_3-a_3a_2a_1\right)Example of the above for r=3r=3. Note ab=12(abba)a \wedge b = \frac{1}{2}\left( ab - ba\right) is example for r=2r=2
aAr=aAr(1)rAra()a \cdot A_r = a A_r - (-1)^rA_r a \qquad (*)Prove by repeated application of ab=2abbaab = 2a\cdot b - ba to aa1ara a_1\dots a_r for orthogonal {ai}\{a_i\}
aAr=aAr+(1)rAra()a \wedge A_r = a A_r + (-1)^rA_r a \qquad (**)Extends definition of \wedge to multivectors
aAr=aAr+aAraA_r = a\cdot A_r + a\wedge A_rCombine above two formulae. aAra\cdot A_r has grade r1r-1, aAra\wedge A_r has grade r+1r+1
Ar(BsCt)=(ArBs)Ctr+st()A_r\cdot(B_s\cdot C_t) = (A_r\wedge B_s)\cdot C_t \quad r+s\leq t \qquad (\triangleleft)Prove by grade projection
Ar(BsCt)=(ArBs)Ctr+ts()A_r\cdot(B_s\cdot C_t) = (A_r\cdot B_s)\cdot C_t \quad r+t\leq s \qquad (\triangleright)Ditto
a(a1ar)=k=1r(1)k+1aak  a1aˇkara\cdot (a_1 \wedge\dots\wedge a_r) = \sum_{k=1}^r (-1)^{k+1} a\cdot a_k \; a_1 \wedge\dots\wedge \check a_k \wedge\dots\wedge a_rThe check on aˇk\check a_k means "omit that term from the exterior product"

Vectors and Bivectors

IdentityComment
a(bc)=ab  cac  ba\cdot(b\wedge c) = a\cdot b\;c - a\cdot c\;bSimplest example of above identity, r=2r=2
a(a1a2a3)=aa1  a2a3aa2  a1a3+aa3  a1a2a\cdot(a_1\wedge a_2\wedge a_3) = a\cdot a_1\;a_2\wedge a_3 - a\cdot a_2\;a_1\wedge a_3 + a\cdot a_3\;a_1\wedge a_2Next simplest example, r=3r=3
(ab)B=a(bB)(a\wedge b)\cdot B = a\cdot (b\cdot B)Application of ()(\triangleleft) in the case r=s=1r=s=1 and t=2t=2
(aB)b=a(Bb)(a\cdot B) \cdot b = a\cdot(B\cdot b)Application of ()(\triangleright) in the case r=t=1r=t=1 and s=2s=2
(ab)(cd)=bc  adac  bd(a\wedge b)\cdot(c\wedge d) = b\cdot c\; a\cdot d - a\cdot c \; b\cdot dExpand into geometric products and use ab=2abbaab = 2a\cdot b - ba
a(bB)=ab  Bb(aB)a\cdot(b\wedge B) = a\cdot b\; B - b\wedge(a\cdot B)Expand into geometic products using ()(*) and ()(**)

Commutator Product

IdentityComment
M×N=12(MNNM)M \times N = \frac{1}{2}(MN-NM)Definition, for multivectors MM and NN
L×(M×N)+N×(L×M)+M×(N×L)=0L\times(M\times N) + N\times(L\times M) + M\times(N\times L) = 0Jacobi identity for multivectors LL, MM and NN
B×a=12(BaaB)=BaB\times a = \frac{1}{2}(Ba-aB) = B \cdot aSpecial case for bivector BB and vector aa using ()(*)
B×Ar=B×ArrB\times A_r = \langle B\times A_r\rangle_rCommutator product with bivector is grade-preserving for blade ArA_r and hence multivector MM
BAr=BAr+B×Ar+BArBA_r = B\cdot A_r + B\times A_r + B\wedge A_rTherefore 12(BAr+ArB)=BAr+BAr\frac{1}{2}(BA_r + A_rB) = B\cdot A_r + B\wedge A_r
(ab)×B=(aB)b+a(bB)(a\wedge b)\times B = (a\cdot B)\wedge b + a\wedge (b\cdot B)Expand into geometic products using ()(**) and def'n of commutator product

Pseudoscalar

IdentityComment
I=e1enI = e_1 \wedge\dots\wedge e_nFor orthonormal vectors {eii=1n}\{e_i \| i=1\dots n\}
I2=±1I^2 = \pm1Sign depends on metric but is typically -1 in physics applications
IAr=(1)r(n1)ArIIA_r = (-1)^r(n-1)A_rIFor blade ArA_r in nn-dims. II always commutes with even-grade multivectors
a(ArI)=aArIa\cdot (A_rI) = a\wedge A_r\, IUsing the above commutation rule together with ()(*)

Linear Algebra

IdentityComment
b=F(a)b = \mathsf F(a) Linear function (whose components form a matrix)
F(μa+νb)=μF(a)+νF(b)\mathsf F(\mu a + \nu b) = \mu\mathsf F(a) + \nu\mathsf F(b)Linearity for vectors aa and bb and scalars μ\mu and ν\nu
F(ab)=F(a)F(b)\mathsf F(a\wedge b)=\mathsf F(a) \wedge \mathsf F(b)Action on a bivector
F(μM+νN)=μF(M)+νF(N)F(Ar)=F(Ar)r\mathsf F(\mu M + \nu N) = \mu \mathsf F(M) + \nu \mathsf F(N) \\ \mathsf F(A_r) = \langle\mathsf F(A_r)\rangle_rLinearity and grade-preservation for multivectors MM and NN, and blade ArA_r
aF(b)=F(a)ba\cdot\overline{\mathsf F}(b) = \mathsf F(a)\cdot bDefinition of F\overline{\mathsf F}, the adjoint (transpose) of F\mathsf F
AF(B)=F(A)B()\langle A\,\overline{\mathsf F}(B)\rangle = \langle\mathsf F(A) \, B \rangle \qquad (\bullet)Multivector version of above definition of adjoint
ArF(Bs)=F(F(Ar)Bs)rsA_r\cdot \overline{\mathsf F}(B_s) = \overline{\mathsf F}\left(\mathsf F(A_r)\cdot B_s\right) \qquad r\leq s \qquad Combine definition of adjoint with ()(*), eg for Ar=aA_r = a and Bs=bcB_s = b\wedge c
F(Ar)Bs=F(ArF(Bs))rs()\mathsf F(A_r)\cdot B_s = \overline{\mathsf F}\left(A_r\cdot \overline{\mathsf F}(B_s)\right) \qquad r\geq s \qquad (\dagger)Combine definition of adjoint with ()(*), eg for Ar=abA_r = a\wedge b and Bs=cB_s = c
R=nm=exp(B^θ/2)=cos(θ/2)B^sin(θ/2)R = nm = \exp(-\hat B\theta/2) = \cos(\theta/2) - \hat B\sin(\theta/2)Rotor RR from vectors mm and nn where mn=cosθm\cdot n = \cos\theta, B^=mnsin(θ/2)\hat B = \frac{m\wedge n}{\sin(\theta/2)} and B^2=1\hat B^2 = -1
F(M)=RMR~\mathsf F(M) = RM\tilde RMultivector MM rotated by angle θ\theta in B^\hat B plane

Determinant and inverse

IdentityComment
F(I)=det(F)I\mathsf F(I) = \det(\mathsf F)\, IDefinition of determinant, as volume scale factor for F\mathsf F
det(FG)I=det(F)det(G)I\det(\mathsf F\mathsf G)\,I = \det(\mathsf F)\det(\mathsf G)\, IBecause det(FG)I=FG(I)=det(G)F(I)=det(F)det(G)I\det(\mathsf F\mathsf G)\,I = \mathsf F\mathsf G(I) = \det(\mathsf G)\, \mathsf F(I) = \det(\mathsf F)\det(\mathsf G)\, I using the above definition
det(F)=det(F)\det(\mathsf F) = \det(\overline{\mathsf F})Because det(F)=F(I)I1=IF(I1)=F(I1)I=det(F)\det(\mathsf F) = \langle\mathsf F(I)I^{-1}\rangle = \langle I\overline{\mathsf F}(I^{-1})\rangle = \langle \overline{\mathsf F}(I^{-1})I\rangle = \det(\overline{\mathsf F}) using ()(\bullet)
F1(M)=1det(F)IF(I1M)\mathsf F^{-1}(M) = \frac{1}{\det(\mathsf F)} I \overline{\mathsf F}(I^{-1}M)Inverse of F\mathsf F. Holds also when FF\mathsf F \rightarrow \overline{\mathsf F}. From definition of determinant, multiply by arbitrary multivector NN: det(F)IN=F(I)N=F(IF(N))\det(\mathsf F)IN = \mathsf F(I)N = \mathsf F\left(I\overline{\mathsf F}(N)\right) using ()(\dagger) and that F(I)N=F(I)N\mathsf F(I)\cdot N = \mathsf F(I)N and IF(N)=IF(N)I\cdot \overline{\mathsf F}(N) = I\overline{\mathsf F}(N). Now let M=INM = IN and apply F1\mathsf F^{-1} to both sides.

CC BY 4.0

This work is licensed under a Creative Commons Attribution 4.0 International License.

CC BY 4.0