proof of a lemma regarding Hodge star (CH2)
DOCX · 28.9 KB
Open DOCX file
Phil's written proof of a lemma on the double Hodge star, in a folder on Sjamaar's treatment of forms. Section 1 derives an explicit sign S = (-1)^(a+b+..+q) (-1)^(k(k+1)/2) for *dx_I and checks it on four small examples against Sjamaar page 24. Section 2 gets the lemma by sliding wedge factors and comparing the two defining sign conditions, without using the explicit S.
AI-written summary; may contain errors.
Extracted text (machine-read; may contain errors)
Proof of a Lemma regarding *(*dxI)
Lemma: Show that
*(*dxI) = (-1)kn+k dxI (1)
Take an example
dxI = dxa dxb .... dxq a < b < ... <q a k-form, k dxi factors (2)
Note that this is in Rn, so we have k and n to worry about. Then
*dxI = S Πh dxi where we have a hole for each of dxa dxb ....dxq and S = a certain sign. (3)
Then sign S is selected so that
dxI(*dxI) = +dx1....dxn = dV. (4)
Inserting (2) and (3) into (4),
dxa dxb .... dxq S Πh dxi = dV (5)
and therefore
dxa dxb .... dxq Πh dxi = SdV (6)
1. Now how do we find S?
I did this just for interest. The resulting expression for S is not needed in Section 2 below.
Combine (2) and (3) above to get
dxI (*dxI) = S dxa dxb .... dxq Πh dxi (1.1)
The sign is found by right-sliding each of the dxi listed into the correct position in Π dxi. Write
dxa dxb .... dxq Πh dxi = dxa dxb .... dxq [ dx1 ....dxa ....dxb........dxq ....dxn ] (1.2)
Here red shows the holes. Let's start by sliding dxa into its red hole. We first have to slide dxa through k-1 other dx's to get it to the right, so this gives
LHS(1.2) = (-1)k-1dxb .... dxq dxa [ dx1 ....dxa ....dxb........dxq ....dxn ] (1.3)
Now we have to slide dxa another (a-1) positions to get it into its proper hole. So,
= (-1)k-1 (-1)a-1dxb .... dxq [ dx1 .....dxb........dxq ....dxn ] . (1.4)
Now we slide dxb first to the right, picking up (-1)k-2 and then we pick up (-1)b-1 so
= (-1)k-1 (-1)a-1(-1)k-2 (-1)b-1 .... dxq [ dx1 ......dxq ....dxn ] . (1.5)
We repeat this little process for each of the k elements of the dxI group.
When we get to the last item dxq, since it is all by itself, we don't have a "first sign" since no sliding is necessary. So I think the k-signs are these (** means exponentiation)
(-1)k-1(-1)k-2 .....(-1)k-k = Πi=0k-1 (-1)i = (-1)**(Σi=0k-1 i) = (-1)** (k-1)k/2
= (-1)k(k-1)/2 (1.6)
where I use this classic result
(1.7)
Can this be simplified? Write out some cases:
k (-1)k(k-1)/2
0 +
1 +
2 -
3 -
4 +
5 +
6 - (1.8)
I see no simplification.
The other sign contributions are this set of k factors
(-1)a-1(-1)b-1.... (-1)q-1 = (-1)a+b+..+q (-1)k (1.9)
The total k related factor is then
(-1)k(k-1)/2 (-1)k = (-1)k+k(k-1)/2 = (-1)k(k+1)/2 (1.10)
My conclusion is that
dxa dxb .... dxq Πh dxi = (-1)a+b+..+q (-1)k(k+1)/2[ dx1 ......dxn ]
= (-1)a+b+..+q (-1)k(k+1)/2dV (1.11)
Now recall from (4) and (1.1) and (1.11) that
dV = dxI (*dxI) = S dxa dxb .... dxq Πh dxi = S (-1)a+b+..+q (-1)k(k+1)/2dV (1.12)
The conclusion then is that
S = (-1)a+b+..+q (-1)k(k+1)/2 (1.13)
Example 1: n = 6, k = 2, dxI = dx2dx5 , *dxI = S dx1dx3dx4dx6 (1.14)
Claim of the above is that
S = (-1)a+b+..+q (-1)k(k+1)/2 = (-1)2+5 (-1)3 = (-1)10 = +1
and then we are supposed to have
dxI = dx2dx5 *dxI = dx1dx3dx4dx6
Let's manually check to verify that dxI (*dxI) = dV
dxI (*dxI) = dx2dx5 dx1dx3dx4dx6 = + dx5 dx1dx2dx3dx4dx6 = + dx1dx2dx3dx4dx5 dx6
This example succeeds.
Example 2: n = 4 and k = 2, dxI = dx1dx2, *dxI = S dx3dx4.
Claim of the above is that
S = (-1)a+b+..+q (-1)k(k+1)/2 = (-1)1+2(-1)3 = (-1)6 = +1
Let's manually check to verify that dxI (*dxI) = dV
dxI (*dxI) = dx1dx2dx3dx4 = dV
This example succeeds and agrees with Sja page 24 A.
Example 3: n = 4 and k = 2, dxI = dx1dx3, *dxI = S dx2dx4.
Claim of the above is that
S = (-1)a+b+..+q (-1)k(k+1)/2 = (-1)1+3(-1)3 = (-1)7 = -1
Let's manually check to verify that dxI (*dxI) = dV
dxI (*dxI) = dx1dx3(-dx2dx4) = dV
This example succeeds and agrees with Sja p 24B
So this is the best result I can get
S = (-1)a+b+..+q (-1)k(k+1)/2 note that S2 = 1 (1.13)
dxI = dxa dxb .... dxq a < b < ... < q a k-form (2)
Example 4 n = 4 and k = 1, dxI = dx1, *dxI = S dx2dx3dx4
Claim of the above is that
S = (-1)a+b+..+q (-1)k(k+1)/2 = (-1)1 (-1)1(1+1)/2 = (-1)1+1 = +1
Let's manually check to verify that dxI (*dxI) = dV
dxI (*dxI) = dx1dx2dx3dx4 = dV
This example succeeds and agrees with Sja p 24C
2. What is the next step?
Recall that
dxI = dxa dxb .... dxn (2)
(*dxI) = S [ dx1 ....dxa ....dxb........dxq ....dxn ] = S Πh dxi (3)
where S is the sign which makes this be true
dxI(*dxI) = dV (4)
Now consider
*(*dxI) = *( S [ dx1 ....dxa ....dxb........dxq ....dxn ] ) (2.1)
= S *{ dx1 ....dxa ....dxb........dxq ....dxn } (2.2)
= S *{ S' dxa dxb .... dxq }
= SS' dxa dxb .... dxq
= SS' dxI (2.3)
where S' is the sign which causes this to be true
(*dxI)*(*dxI) = dV (2.4)
Our goal is to show that SS' = (-1)kn+k since then (2.3) gives the desired result (1). (2.4a)
Now write out (2.4) making use of (3) for (*dxI) and (2.3) for *(*dxI) to get
S [ dx1 ....dxa ....dxb........dxq ....dxn ] S [ S' dxa dxb .... dxq ] = dV (2.5)
Since S appears twice, we get
S' [ dx1 ....dxa ....dxb........dxq ....dxn ] [ dxa dxb .... dxq ] = [dx1......dxn] . (2.6)
Now let's start by moving dxa all the way to the left and sits then after S'. It has then to pass through n-k dxi factors, so we pick up (-1)n-k. We then have
S' (-1)n-k dxa [ dx1 ....dxa ....dxb........dxq ....dxn ] [dxb .... dxq ] = dV (2.7)
Now do this next with dxb and pick up another (-1)n-k. Do this k times and we then have
S' (-1)kn-kk [dxa dxb .... dxq ] [ dx1 ....dxa ....dxb........dxq ....dxn ] = dV (2.8)
or
S' (-1)kn-kk dxa dxb .... dxq Πh dxi = dV (2.9)
or
dxa dxb .... dxq Πh dxi = S' (-1)kn-kk dV (2.10)
Finally, we use our previous result (6),
dxa dxb .... dxq Πh dxi = SdV (6)
Comparing the last two results gives
S' (-1)kn-kk dV = SdV (2.11)
so that
S'S = (-1)kn-kk (2.12)
Now what can you say about
(-1)kk ?
If k is even, then k2 is certainly even 22 = 4 42 = 16
If k is odd, then k2 is odd. 12= 1 32= 9 52 = 25
Therefore
(-1)kk = (-1)k ! (2.15)
and then
S'S = (-1)kn-k
Now multiply by 1 = (-1)2k to get
S'S = (-1)kn+k
and we have then achieved our goal as stated in (2.4a) and thus (1) is proved.
Comments: Try a summary of the above derivation. S is defined by (4) and S' by (2.4). From (2.3) we find that we have solved the problem if we can show that SS' = (-1)kn+k. We have this result
S' (Πhdxi )dxI = dV (2.6)
After sliding dxI to the left, this becomes
dxI (Πhdxi ) = S' (-1)kn-kk dV (2.10)
But our earlier result (6) said
dxI (Πhdxi ) = SdV
and this then gives the solution S'S = (-1)kn-kk. We never have to use the explicit formula for S which was developed above in Section 1.