The Cross-Ratio Function

0 Conversations

Let x be a complex number. The formal values of the cross-ratio function are x, 1-x, (1-x)-1, -x(1-x)-1, -x-1(1-x), and x-1 (where x≠0 or 1). The values equal x, f(x), gf(x), fgf(x), gfgf(x), fgfgf(x) where f(x)=1-x and g(x)=1/x. Let h0(x)=x, h1(x)=f(x), h2(x)=gf(x), h3(x)=fgf(x), .... Then h0=h6 and hence hi=hi+6, i=0, 1, 2, .... In this paper, the equality of the functions h0 and h6 is considered to be the defining concept of the cross-ratio. This allows generalization. For example, if f(x)=2-x and g(x)=2/x where x≠0, 1, or 2, then h0=h8, and if f(x)=3-x and g(x)=3/x where x≠0, 1, 2, 3/2, or 3, then h0=h12. However, as will be shown, if f(x)=k-x and g(x)=k/x where k is a natural number other than 1, 2, or 3, then h0≠hi for any i>0. Let x1 and x2 be the roots of x2-x+1=0. Then x1+x2=1 and x1x2=1 (the symmetric equations), and hence x2=1-x1 and x1=1/x2. Therefore if x is a root of x2-x+1, then h0(x)=h2(x)=h4(x)=... and h1(x)=h3(x)=h5(x)=.... So there is a relationship (in some sense) between the equation x2-x+1=0 and the cross-ratio. This is the motivation for seeking "cross-ratios" associated with the general quadratic equation x2-ax+b=0.

Relationship of "h" Functions to a Lucas Series

Let f(x)=a-x and g(x)=b/x where a and b are integers, b≠0. Let Ui be defined by the recurrence relation Ui-aUi-1+bUi-2=0, i=2, 3, 4, ..., U1=1, U0=0 (a Lucas series). Some properties of this series (commonly cited in the mathematical literature) are;

(1) Ui=(x1i-x2i)/(x1-x2) where x1 and x2 are the roots of x2-ax+b=0.

(2) If i divides j, then Ui divides Uj if Ui≠0, or Uj=0 if Ui=0.

(3) Ui2-Ui-1Ui+1=bi-1.

Let h0(x)=x, h1(x)=f(x), h2(x)=gf(x), h3(x)=fgf(x), ....

(4) h1(x), h2(x), h3(x), ..., hk(x) are defined only if Uix≠Ui+1, i=1, 2, 3, ..., k. If i is odd and 0<i≤k, then hi(x)=(Uj+1-Ujx)/(Uj-Uj-1x) where j=(i+1)/2. If i is even and 0<i≤k, then hi(x)=b(Uj-Uj-1x)/(Uj+1-Ujx) where j=i/2.

Proof: The proof is by induction. Suppose hi(x)=b(Uj-Uj-1x)/(Uj+1-Ujx) where i is even, i≥0, and j=i/2. Then hi+1(x)=f(hi(x))=a-b(Uj-Uj-1x)/(Uj+1-Ujx)=[(aUj+1-bUj)-(aUj-bUj-1)x]/(Uj+1-Ujx)=(Uj+2-Uj+1x)/(Uj+1-Ujx). Therefore hi+1(x)=(U[(i+1)+1]/2+1-U[(i+1)+1]/2)/(U[(i+1)+1]/2-U[(i+1)+1]/2-1) where i+1 is odd. Now suppose hi(x)=(Uj+1-Ujx)/(Uj-Uj-1x) where i is odd, i>0, and j=(i+1)/2. Then hi+1(x)=g(hi(x))=b(Uj-Uj-1x)/(Uj+1-Ujx). Therefore hi+1(x)=b(U(i+1)/2-U(i+1)/2-1x)/(U(i+1)/2+1-U(i+1)/2x) where i+1 is even. Also, h1(x)=f(x)=a-x=(U2-U1x)/(U1-U0x) and h2(x)=b/(a-x)=b(U1-U0x)/(U2-U1x).

Necessary and Sufficient Conditions for Equal "h" Functions

(5) h0(x)=hi(x) where i is odd if and only if x=(Uj±b(j-1)/2)/Uj-1 where j=(i+1)/2 and Uj-1≠0, or x=Uj+1/(2Uj) where Uj-1=0.

Proof: Suppose h0(x)=hi(x) where i is odd. Then x=(Uj+1-Ujx)/(Uj-Uj-1x) where j=(i+1)/2 and hence -Uj-1x2+2Ujx-Uj+1=0. Therefore x=[-2Uj±√(4Uj2-4Uj-1Uj+1)]/(-2Uj-1) if Uj-1≠0, or x=Uj+1/(2Uj) if Uj-1=0. Then since Uj2-Uj-1Uj+1=bj-1, x=(Uj±b(j-1)/2/Uj-1 if Uj-1≠0. Conversely, if x=(Uj±b(j-1)/2)/Uj-1 then h0(x)=hi(x), and if x=Uj+1/(2Uj), Uj-1=0, then h0(x)=hi(x).

Therefore h0≠hi where i is odd (since h0(x)=hi(x), i is odd, if and only if x has a certain value). (Some U values corresponding to the cross-ratio are U0=0, U1=1, U2=1, U3=0, U4=-1, U5=-1, U6=0. Therefore h0(x)=h1(x) if and only if x=U2/(2U1)=1/2, h0(x)=h3(x) if and only if x=(U2±1)/U1=0, 2, and h0(x)=h5(x) if and only if x=(U3±1)/U2=1, -1. Therefore h0(x)=hi(x) where i is odd and a=b=1 if and only if x=1/2, 2, or -1 [since x=0, 1 are not considered to be in the domain of f and g].)

(6) h0(x)=hi(x) where i>0, i is even, if and only if Ui/2(x2-ax+b)=0.

Proof: Suppose h0(x)=hi(x) where i>0, i is even. Then x=b(Uj-Uj-1x)/(Uj+1-Ujx) where j=i/2 and hence Ujx2-(Uj+1+bUj-1)x+bUj=0, that is, Uj(x2-ax+b)=0. Conversely, if Ui/2(x2-ax+b)=0, then h0(x)=hi(x).

There can be up to i/2+1 x values for which hi(x) is not defined if h0(x)=hi(x), i>0, i is even. These x values and the two x values satisfying x2-ax+b=0 are not considered to be in the domain of f and g. Therefore h0=hi where i>0, i is even, if and only if Ui/2=0. If a=0, then U2=0, h1(x)=-x, h2(x)=-b/x, h3(x)=b/x, and h4(x)=x=h0(x). This is a trivial "cross-ratio", so it is stipulated that a≠0 in the following.

Necessary and Sufficient Conditions for Lucas Series Values of Zero

(7) Ui=0, i>0, if and only if b=a2/k where k=1, 2, or 3 and k divides a. If b=a2, then Ui=0, i>0, if and only if 3 divides i. If b=a2/2, then Ui=0, i>0, if and only if 4 divides i. If b=a2/3, then Ui=0, i>0, if and only if 6 divides i.

Proof: First, some lemmas will be proved. Let c(i,j) denote i "choose" j (a binomial coefficient).

(8) If i is even, Ui=ai-1-c(i-2,1)ai-3b+c(i-3,2)ai-5b2-c(i-4,3)ai-7b3+...+(-1)i/2-2c(i/2+1,i/2-2)a3bi/2-2+(-1)i/2-1c(i/2,i/2-1)abi/2-1. If i is odd, Ui=ai-1-c(i-2,1)ai-3b+c(i-3,2)ai-5b2-c(i-4,3)ai-7b3+...+(-1)(i-1)/2-1c((i-1)/2+1,(i-1)/2-1)a2b(i-1)/2-1+(-1)(i-1)/2c((i-1)/2,(i-1)/2)b(i-1)/2.

Proof: The proof is by induction. Suppose Ui=ai-1-c(i-2,1)ai-3b+c(i-3,2)ai-5b2-...+(-1)i/2-1c(i/2,i/2-1)abi/2-1 and Ui-1=ai-2-c(i-3,1)ai-4b+c(i-4,2)ai-6b2-...+(-1)(i-2)/2c((i-2)/2,(i-2)/2)b(i-2)/2 where i is even. Then using the relations Ui+1=aUi-bUi-1, c(i,j-1)+c(i,j)=c(i+1),j), j=1, 2, 3, ..., i, and collecting terms gives Ui+1=ai-c(i-1,1)ai-2b+c(i-2,2)ai-4b2-...+(-1)i/2c(i/2,i/2)bi/2, that is, Ui+1=a(i+1)-1-c((i+1)-2,1)a(i+1)-3b+c((i+1)-3,2)a(i+1)-5b2-...+(-1)[(i+1)-1]/2c([(i+1)-1)]/2,[(i+1)-1]/2)b[(i+1)-1]/2 where i+1 is odd. Now suppose Ui=ai-1-c(i-2,1)ai-3b+c(i-3,2)ai-5b2-...+(-1)(i-1)/2c((i-1)/2,(i-1)/2)b(i-1)/2 and Ui-1=ai-2-c(i-3,1)ai-4b+c(i-4,2)ai-6b2-...+(-1)(i-1)/2-1c((i-1)/2,(i-1)/2-1)ab(i-1)/2-1 where i is odd. Then using the relations and collecting terms as before gives Ui+1=a(i+1)-1-c((i+1)-2,1)a(i+1)-3b+c((i+1)-3,2)a(i+1)-5b2-...+(-1)(i+1)/2-1c((i+1)/2,(i+1)/2-1)ab(i+1)/2-1 where i+1 is even. Finally, if i=1, Ui=ai-1=(-1)(i-1)/2c((i-1)/2,(i-1)/2)b(i-1)/2=1 and if i=2, Ui=ai-1=(-1)i/2-1c(i/2,i/2-1)abi/2-1=a.

(9) Ui=0, i>0, only if √(a2-4b) is imaginary.

Proof: Ui=0 only if (x1i-x2i)/(x1-x2)=0 where x1=[a+√(a2-4b)]/2 and x2=[a-√(a2-4b)]/2. Now {[a+√(a2-4b)]/2}i-{[a-√(a2-4b)]/2}i=(1/2i)ad(c(i,1)ai-2+c(i,3)ai-4d2+...+c(i,i-1)di-2) if i is even, or (1/2i)d(c(i,1)ai-1+c(i,3)ai-3d2+...+c(i,i)di-1) if i is odd, where d=√(a2-4b), d=x1-x2, therefore Ui=0 only if d is imaginary and hence Ui=0 only if b>0.

(10) Ui=0 where i>0, i is even, only if b=a2/q where q is a positive divisor of i/2. Ui=0 where i is odd only if b=a2.

Proof: Suppose Ui=0 where i>0, i is even. (Note that U2≠0.) Then by Lemma (8), b divides ai-1 and a2 divides c(i/2,i/2-1)bi/2-1. Let k=g.c.d.(a, b) and k1=g.c.d.(a, b/k). Then a{(a/k1)i/2-1(a/k)i/2-1-c(i-2,1)(a/k1)i/2-2(a/k)i/2-2[b/(kk1)]+c(i-3,2)(a/k1)i/2-3(a/k)i/2-3[b/(kk1)]2-...+(-1)i/2-2c(i/2+1,i/2-2)(a/k1)/(a/k)[b/(kk1)]i/2-2+(-1)i/2-1c(i/2,i/2-1)[b/(kk1)]i/2-1}=0. Now g.c.d.(a/k, b/k)=1 and g.c.d.(a/k1, b/(kk1))=1, therefore (a/k1)/(a/k) divides c(i/2,i/2-1), that is, a2/(kk1) divides i/2. Also, b/(kk1) divides 1. Therefore kk1=±b and hence b=a2/±q where q is a positive or negative divisor of i/2. By Lemma (9), b>0, therefore b=a2/q where q is a positive divisor of i/2. Now suppose Ui=0 where i is odd. Then (a/k1)(i-1)/2(a/k)(i-1)/2-c(i-2,1)(a/k1)(i-1)/2-1(a/k)(i-1)/2-1[b/(kk1)]+c(i-3,2)(a/k1)(i-1)/2-2(a/k)(i-1)/2-2[b/(kk1)]2-...+(-1)(i-1)/2-1c((i-1)/2+1,(i-1)/2-1)(a/k1)(a/k)[b/(kk1)](i-1)/2-1+(-1)(i-1)/2c((i-1)/2,(i-1)/2)[b/(kk1)](i-1)/2=0. Therefore (a/k1)/(a/k) divides c((i-1)/2,(i-1)/2), that is, a2/(kk1) divides 1. Also, b/(kk1) divides 1. Therefore kk1=±b and hence b=a2.

√(a2-4b) is imaginary, b=a2/q, only if q=1, 2, or 3. Therefore Ui=0, i>0, i is even, only if q=1, 2, or 3. Now U3=a2-b, U4=a(a2-2b), and U6=a5-4a3b+3ab2=a(a2-3b)(a2-b), therefore Ui=0, i>0, i is even, if q=1, 2, or 3.

References

E. Lucas
Comptes Rendus Paris, 82, 1876, pp. 1303-5.


Bookmark on your Personal Space


Conversations About This Entry

There are no Conversations for this Entry

Entry

A44722569

Infinite Improbability Drive

Infinite Improbability Drive

Read a random Edited Entry


Written and Edited by

Disclaimer

h2g2 is created by h2g2's users, who are members of the public. The views expressed are theirs and unless specifically stated are not those of the Not Panicking Ltd. Unlike Edited Entries, Entries have not been checked by an Editor. If you consider any Entry to be in breach of the site's House Rules, please register a complaint. For any other comments, please visit the Feedback page.

Write an Entry

"The Hitchhiker's Guide to the Galaxy is a wholly remarkable book. It has been compiled and recompiled many times and under many different editorships. It contains contributions from countless numbers of travellers and researchers."

Write an entry
Read more