Professional Documents
Culture Documents
SoJa*ien:A tittfu experimentati''on with functions wl.rich satiS (n) will likely convince the reader that these functions are not unlike the identity mapping. (For example: powers of the arguments, multiplication by non-trivial constants, and the addition of constants, as functions, all fail the criterion.) In fact, we prove that the property (z-) uniquely characterizes Id(r). The proof will depend on the following lemma; the proof of the lemma is tedious, but I don't know any way around it. Lemma: If / has the property (zr), then /(0) : 0. Proof : Put y - /(0). Take s :0,f :0 in / to show
f@)-a'
0 Takings:U,t-0shows
(1)
f (y' + u) : f @)'
and so substitution from (1) gives f @'-ty)-ya.
However, Ietting s : 0, t : U shows (by (t) again)
(2)
(B)
f(y')-a+az.
If we substitute the RHS of ( ) in the LHS of (3), we obtain
(4)
f u@\) : y4
but letting s : A,t: y2 directly showsthat
(5)
f(f(y')) - 2u'.
(6)
ft
,q or \/r.
So we discover that we must have A4 : 2y2,, which requires U : -1/i,0,
f @ ' + f U @ \ ) ) - f Q a 2 ) : f @ 2 ) +f ( y ) 2
(wherethe far left and midclieare equalUV(6)); but ffiffi$'i y says
(T)
I 1ft,y,t -
2 f ( s a 2-) f ( r t i l ' + y - f ( y 2 )+ y
(8)
Equating (7) and (8), and recalling that f (y2) - A I A2, we then find
f ( y ' ) + f ( i l 2 - f ( a 2 ) 2 + a+
y*a2+a4 -u2+zy3+a4+a +
2 y s- 0 : U in (7)
which is absurd. AssumingA - -rn and settirg s - -{2y,t similarly shows that y + -\/r. We concludethat U : 0. I
With the lemma in hand, we can finish the proof in a more reasonable(and interesting) way. Consider any function / with the property (zr). If we fix s - 0 and let t take any real value, we find that /(/(t)) - t. The reader may recall that this says that f is an 'inuolut'ioz;as such, it constitutes its own inverse. (Clearly, it maps the range of / into its domain.) Because/ is invertible, it must be one-to-one;so / cannot take the value of zero for more than one axgument. Since /(0) - 0, we have that for ,r IR : r + 0,,f (*) 1 0. We can use this fact to show that f is increasing. Proof: Pick r,a lR,r ( y, andsetd - A-r.Bychoosing t - f(r), t : J6, we have
/ ( d + f ( f ( " ) ) ) : f ( 6 r r ) - f ( " ) +f ( 1 6 , +
f ( " + 6 ) - f ( " )- f b f q ,
Since\fr 10, we havef (t/6) f A, andso /(\/5)' > 0; so f (r+ d) - f (*) f (a) - f (") > 0. This showsthat / is increasingon its domain. The reader may prove for himself (it isn't hard!) that f (r) - r is the only increasinginvolution; he is advised to supposef(") > r or f(r) < r for arbitrary r, and derive a contradiction. Since we proved that any function with the property (zr) is an increasing involution, we have thus proved that
e,
f (*) : r is the only function which satisfies (r). I It is quite fortunate that functions satisfying (zr) are increasing; Remark were they not, there would be infinitely many possibilities, the simplest being linear and piecewise linear. In general, the strongest condition one can impose on a continuous involutiorr is that it meet the graph of Id(r) at some pcint; of course, w-ehad no glrarantee that functions with the property (zr) were even continuous!
e,
0,