You are on page 1of 4

Deriving the 4 Dimensional Cross Product of 2 Vectors

 W. F. Esterhuyse

October 2009
It is possible to derive the four dimensional cross product in the following way. We first define
the

operation "dimensional superposition" (symbol = &) as the operation to obtain a n dimensional

space from n-k dimensional spaces (n > k > 1). We define the operation to respect both the

(+ - + -) structure of the cross product and the signature of the (flat) n dimensional space. To

obtain a 3D space from 2D spaces we need to superimpose three 2 dimensional planes at right

angles to each other. A 2D plane is defined by three points. Choosing two planes and four
vectors

in them such that each pair of vectors are orthogonal, we form "&" of the two planes by

superimposing two vectors from different planes such that the other two are orthogoal and the
origin

of both pairs coincide. In this way we can form a right handed or left handed 3D space by the

appropriate labelling of the resulting three vectors spannning the 3D space. The three vectors

can be normalised to produce an orthonormal basis.

We use [i] as the i'th unit vector.

Taking the 3D space as right handed we may conceivably choose two vectors u and v in the

flat 3D space such that:

uxv = & uxv|12 & uxv|13 & uxv|23 (1)

= & uxv|12not3 & uxv|13not2 & uxv|23not1

where the RS is defined as:

RS = & [3] | u1 U2 | & [2] | u1 u3 | & [1] | u2 u3 |


| v1 v2 | | v1 v3 | | v2 v3 |

that is:

RS = & [1]x[2] | u1 U2 | & [1]x[3] | u1 u3 | & [2]x[3] | u2 u3 |


| v1 v2 | | v1 v3 | | v 2 v3 |

obtained by taking the vector product in the 2D subspaces as in RS (1). We need the first "&" just
like we would if "&" was "-".

We see that the above construction applies. Superimposing two 2-planes by leaving out the

plane of [2] and [3] is not valid (if considering the 3D space spanned by surfaces) since then

projection onto the 2-3-plane is impossible.

Now, since we know how the 3D cross product is defined, we can work out what the "&"

operation reduces to by comparing RS (1) with this definition.

The RS must thus reduce to:

RS = [3] u1 u2 - [2] u1 u3 + [1] u2 u3 (3)


v1 v2 v1 v3 v2 v3 .

Since (3) can be obained from:

[1] [2] [3] (4)


u1 u2 u3
v1 v2 v3

the rule to take "&" to (+ or -) is development of (4) while respecting the + - + order of the

determinant. We note that this is a flat 3D space with signature: (+++). Thus the construction (1)

is proved valid under the stated "&"- rule.

Taking this process further we let u and v be two vectors in a flat 4D space. Then by the

above analog we have by constructing a 4D space from 3D spaces:

uxv = & uxv|123 & uxv|124 & uxv|134 & uxv|234 (4)

with:

RS = & [1] [2] [3] & [1] [2] [4] & [1] [3] [4] & [2] [3] [4] (5)
u1 u2 u3 u1 u2 u4 u1 u3 u4 u2 u3 u4
v1 v2 v3 v1 v2 v4 v1 v3 v4 v2 v3 v4

by the 3D vector product definition (applied in the subspaces in RS (4)). This is

equivalent to a superposition of all 3x3 subdeterminants of:

[1] [2] [3] [4] (6)


u1 u2 u3 u4
v1 v2 v3 v4

and the "&" operation must reflect development of (6) to obtain (5) with the correct signs.

The process to determine the correct transitions from "&" to "+ or -" requires a bit of a leap

but the results can be tested for fit into the properties of the cross product. The process
to determine this is by explicitly deleting the i'th column in (6), and shifting the remaining

columns left such that all deleted columns are at rightmost while indices still increases to the

right. Count the amount of shifts required (say this is s) and change "&" of the term into "+" if s

is even and into "-" of s is odd. Therefore (6) reduces as:

[1] [2] [3] [4] = & [1] [2] [3] [] & [1] [2] [] [4] & [1] [] [3] [4] & [] [2] [3] [4] (7)
u1 u2 u3 u4 u1 u2 u3 [] u1 u2 [] u4 u1 [] u3 u4 [] u2 u3 u4
v1 v2 v3 v4 v1 v2 v3 [] v1 v2 [] v4 v1 [] v3 v4 [] v2 v3 v4

=+ ... - ... + ... - ...

where the ellipses are the corresponding 3x3 subdeterminant of the term above it, just

after the shifts and with the deleted column left out. This is the correct process to determine

what "&" changes into since in 5D we would have two deleted columns (then assigning a

"+ or - " by just looking what column got deleted is invalid). This method is valid for evaluating

any non-square determinant (by row) having rows longer than columns. This method is equivalent

to first developing (6) by row 1 like an ordinary determinant to obtain unit vectors mutiplied by 2x3

determinants and then using the method to evaluate them.

We state the change if the signature is not (++++) ( the first row signs is + - + - with this

signature). Say the signature is (+++ -) then the first row signs changes to: + - + +, and this is

implemented by developing LS (7) by row 1 untill the 2x3 stage (using + - + -) and flipping the
sign

of the term with [4] in it, or just developing LS (7) by row 1 using the altered pattern. The rule in

general is that the determinant sign stays the same on a signature of "+" and flips on a signature

of "-".

The subspaces in (4) are all assumed right handed. In the case that the terms are a combination

of right handed or left handed we would need another type of signature that does not work the
same

as the metric signature. For a signature of RRRL, (4) would change to:

uxv = & uxv|123 & uxv|124 & uxv|134 & ( -uxv|234 ) (4)

(it seems reasonable by looking at term 1 of (7)). In (3) we may say that the superposition due to

[1]x[3] is negative since right handedness says [1]x[3] points in the -[2] direction. If we considered
the [1]x[3] generated plane as left handed it would have gone to +[2].

Length Condition

The length condition not holding does not disqualify this nD vector product as a candidate for the
generalisation of the cross product since:

any two dimensional hyperplane is defined by three points therefore by two vectors (if we chose
one of the points at the origin of the two vectors), and we may define in 4 dimensions
four 2-hyperplanes by having various restrictions on the vector product of u and v and having
areas:

|[(uxv)|1,2,3]|2 or

|[(uxv)|2,3,4 ]|2 or

|[(uxv)|1,3,4 ]|2 or

|[(uxv)|1,2,4 ]|2

and we may even define a compound 2-plane in 4 dimensions as having area:

star |& (uxv)|1,2,3 & (uxv)|2,3,4 & (uxv)|1,3,4 & (uxv)|1,2,4 |2 = star |uxv|2 = (d-2) [(u.u)(v.v) - (u.v)2]

(5)

where we specify by "&" the operation of "dimensional superposition". We use this definition

since it holds in any dimension (d). The operator is not the Hodge star and this fomula is proved

in another article of mine.

By the formula for the angle between two vectors we have four "angles" between any two vectors

in 4D:

|[(uxv)|1,2,3]|2 / [(u.u)(v.v)|123] = (sin t)2 (6)

etc.

It is clear from (5) and the derivation in 3D that a "compound angle" may exist as:

star |uxv|2/[(d-2)(u.u)(v.v)] = (sin t)2

provided u and/or v is not zero.


_______________________________

Email me for any comments/equiries:

esterhuysew@telkomsa.net

You might also like