Seria I: PRACE MATEMATYCZNE X II (1969) ANNALES SOCIETATIS MATHEMATICAE POLONAE Series I: COMMENTATIONES MATHEMATICAE X II (1969)
Z. K areńska (Kraków)
On some functional equation in the theory of geometric objects
In this paper we find all solutions of the functional equation (
0.
1) g(oo-y) = F{ x ) - g( y) + g{x),
where x and у are non-singular
2x
2real matrices, g is an unknown function whose values are 3 x 1 real matrices, and F is a given function of the form
x\x 2 х Х1 х 12 x\2
(0.2) 1
я И = ~ j XlxX21 ffil\X22~\- xx2x2X <^i2x22 2 x 2X x 22 ffi22
where Л is the determinant of x and x lx, x l2, x 21, x 22 are the entries of x.
The general linear group G L (
2,B ) will shortly he denoted by G2. W e do not make any assumptions concerning the regularity of the matrix func
tion g.
Equation (
0.
1) appears in the theory of geometric objects when we want to find the solutions of the system of functional equations
F ( x - y ) = F { x ) - F ( y ) , g{x-y) = F{ x) - g{ y) + g{x)
([2], p. 152) in order to determine the geometric objects of type [3, 2 ,1]
with linear non-homogeneous transformation rule.
The method used below is analogous to that used in [5] and the function (0.2) is similar to that appearing in [5 ]; the present case, however, is more interesting and more complicated, and we get some non-meas- urable solutions whereas all solutions in [5] were measurable.
The main result of the paper is
T heorem
0.
1. Any solution of (0.1) defined on G2 is of the form g{x) = [ F { x ) - E ] - q + g*(x),
(0.3)
220 Z. K a r e ń s k a
where F is given by (
0.
2), E is the unit 3 x 3 matrix, q is a 3 x 1 matrix whose entries &2, &3 are real parameters,
/у» /V»
^ 1 1 ^ 1 2
* 1
9 (я) = — ł <p(xu) ę>(®22) (p(x21) <p{x12)
Л &11 X22 + 21 *^12
<P(®2l) ^(^
2 2)
# 2 1 ® 2 2
«md
9 9is any derivation of R.
B y a derivation of we mean any function <p: R R satisfying the conditions
(0.4) . <p{£+y) = <pW + <p(y),
(0.5)
9?(£??) = w ( £ ) +
for all i, у in R (cf. [9], p. 120 and [3]).
Let
90 be such a derivation. It is well known that (p(0) =
9?(
1) = 0,
<p(— I) = — <p{i) and
(
0.
6) q>{f) = p | p- V ( D
for £ in R and p = 1, 2, ... I f у Ф 0, then <p{lly) = — <p{y)ly2 and
(0.7) J l \ = w W z W ' i ) .
\yf 7)2
Furthermore
9 9vanishes on the algebraic closure Q of the field Q of rationale (in R ) and has a dense set of periods. I t is well known that any measurable derivation of R is trivial (i.e., identically 0), but there are non-trivial derivations of R. W e thus get the following corollary.
C
o r o l l a r y 0.
1. Any measurable solution of (0.1) is of the form g(x)
= [ F ( x ) ~ E] - q, where F , E and q are as in Theorem 0.1.
§1 . Some properties of solutions of (0.1). A straightforward veri
fication shows that the function F defined by (0.2) satisfies the equation F ( x ' y ) = F ( x ) - F ( y ) (cf. [4]) and any function of the form (0.3) satisfies (0.1). W e have thus to show the converse statement. Let us notice that the matrix equation (
0.
1) is equivalent to the system of three equations
Yi (a-y) = ~ [ ^hyi ( y) + 2x11x 12y2(y) + x2 12yz(y)] + yl (x),
1
УЛХ' У) = -j[oOu002iyi{y) + {co11^22 + co12xn )y 2{y)-\-x12x 22y3{y)'] + yi (x), 1
1
Уз(х ‘ У) =-^-[ooh ri(y) + 2x2lx 22y2{y) + xl2y3(y)] + y3{x),
where
y x( x ) , y 2( x ) ,Уз(ж) are the entries of the matrix
g ( x ) .Let
g ( x )denote the matrix
\\yik{x)\\,i, к — 1 ,2 where
У11И = У1О»),
y X2( x )= у 21 (ж) = ya(®)> У22И = y 3 (a?).
The above system is equivalent to the matrix equation
1 T „
(1.1) 0(®-y) = - j x - g { y ) - x + g ( x ) ,
where ж2 denotes the transposed matrix and A = deta;. I f g is a solution of (0.1), then g is a solution of (1.1) and conversely. I f g satisfies (1.1), then g(e) = ||0||, where [|0|| is the 2 x 2 zero matrix and e is the unit 2 x 2 matrix. Moreover, if г = fie, where fieB and fi Ф 0, then g(z) = ||0||.
Indeed, since г commutes with any 2 x 2 matrix, we get g( x-z) — g( z-x) and
1 „ T „ «
— x- g{ z) - x + g { x ) = g{x)Ą-g{z), and hence
(1.2) A~ 1 x- g{ z) - xT = g(z).
Substituting x — j, where j denotes the matrix (bik) with bxx — b 22 — 0,
&i
2= b 21 = 1, we get —j - g { z ) - j = g{z). Hence yX 2 (z) = 0 and y 22 (z)
— ~Уи( я) - Consequently,
9(*) = У u («) 0 0 — yxl(z)
Now, substituting x = n in (1.2), where n is the matrix (eik) with cn = — 1, c22 = 1, c 12 = c 21 = 0, we get — n- g( z ) - n = g(z); therefore Уи{у) = 0. This concludes the proof of the identity g(z) = ||0||.
Substituting x = у = j in (1.1) we get
II0II = g(e) = g ( j - j ) = - j ’ § U) - j + g{j)- Hence
(i- 3) 5 (j) =
§ 2. The functions r and h. We are going to find the general form of a solution g( x) of (1.1). W e begin with the following identities used by Kucharzewski and Kuczma ([6], p. 3), which guide us in our consid
erations :
(
2
.1
)x xx 0 1 0 ---- 0 A 1 0
\T
0 1 x 2X 1 0 1 Ж12 ^
®11
222 Z. K a r e ń s k a
provided that x xx Ф 0, and 0
x X2= (3D’j ) - j =
x X2
0
(2.2)
x=
x 2i x 22 X 22 X 21
provided that
x xx= 0 and det
xФ 0. Consequently, in order to find the general solution
g ( x )of (1.1) it is enough to know g( j ) and the functions
r ( e) = 9 Q Ф 0, Щ ) = 9
The identity g(e) = ||0|| proved above implies that r( 1) and h( 0) are equal to the zero matrix. Applying (2.1), (1.1) we get, when x xx ф 0, (2.3) g{®) = — - j x - g ( j ) ’ ®
0 x xi A
A
0 ---- Хц
X11 X21
3D 11 3C2X 1 0 xxl •r\ — )■ 1 A 0 1 X11 1
x 21\ / 3D a X 21 + — 1
Хц 0
’9(3)'
Хц X2x +
X11 3D 21 f 0 1
+
x xx 0
0 1 Ji(x21) Хц 0
0 1 ф г ( х
11)Thus, we have to find r(g), h(r)) and g{j). An easy computation yields
(2.4) га?] ) = 1 0
1 'Г(т)У I 0
0 - 0 1
+ r(£)
provided that ф 0. Since the matrix g( x) is symmetric, r(| ) is also symmetric, i.e., r 12{£) = r 21(£), where rik(£), i, к = 1 , 2 denotes the entries of r(|). Hence
(2.5) = Г 12 {£) + Г 12 {
г}),
if irj ф 0 . Applying (2.4) to the identity r(irj) = r(r)tj) and comparing the corresponding terms of matrices we get
- - l W 2(|)
П I
( l — i ) r n ( v ) = (v — i ) r n U ) ,
for all £,r) different from 0. I f g = 2, then
r lx{£) = A J ! — 1) and r22(£) = — 1), where Kx = ru (2) and 13 =
(2.6) r(|) =
— 2r22(2). Thus, if f # 0, Ях( | - 1 ) rla( f )
^*12 ( £) ^3 1 j Now, since g is a solution of (1.1), we obtain
(2.7) &(£ + »?) = l о I i •*to)-
i I
o 1 f *(*)•
Arguing as above and taking into account the symmetry of the matrix Ji(rj) we infer that £hxl(T)) — ф х1{£) and
12^
и(^) + 2 ^ 12(^) = У 2 Ьп (%) + 2 ф Х2(£)-
In particular, if rj — 1, then tjhxl( 1) = hxx(£) and 12йп (1) + 2|Л12(1)
= ^ u ( l ) + 2Л12(£). Thus,
(
2
.8
)W ) =
2
tl £ ^l^2+ ^ 2 l 'rl^ 2+ T2l ^22 (£)
where r x = \hxx{ 1), r2 = hx 2 { l ) — p u (l), and £ is any real number.
Substituting rj = — I in (2.7) and applying (2.8) we get h22(£) + /i22( — £)
= 2r2 £2. In virtue of the identity
1 0| 1 0 1 0 1 I 0
0 1 1
1 Sy 1 Г) 1 0 1 and equation (1.1), we get
1 7
i о
0 1 • й (£>?)•
I о
0 1 ■r(£) I1 0
I V 1
•r(|) 1 ?7
0 1 + h(rj).
Substituting rj = 1, applying (2.6) and (2.8) and comparing the corresponding terms of matrices we infer that
r ^ - l J ^ O ,
(£ — i ) [ T'i ( £ + i ) + T2— Л-
i] = о
and A22(£) = £Л22(1) + Лг£(£ — l ) + 2 fria( f ) for all f # 0. Consequently, Ti = 0, r 2 = Ax, and
(2.9) * ..(* ) = 2£[r12(£ )-< 9 ] + r 2£2,
224 Z. K a r e ń s k a
where 0 — | [т2— й22(1)]. Combining the above results with (2.6) and (2.8) we get
(
2
.10
)if £ yŁ 0, and ( 2 . 11 )
r ( i ) =
t
2( £ - 1 ) ria(£)
W )
* w (f) Mr-)
0
t2£
t
2£
<Mfor all £. I f £ ^ 0, then the relation (2.9) between r12 and /г22 may be written as r12(£) = (2£)~1ft22( ! ) — -|
t2£ + <9- On the other hand, substi
tuting (2.11) to (2.7) and comparing the corresponding terms we get
” ^22 ( £) "I" ^22 (??) "H 2
t2 £>7 .
Let £Г(£) = &22(£) —
t2£2. An immediate computation shows that i f ( £ + ?
7) = И (| ) + Я (^ ) for all £ and rj. Let
9?(£) = IT (£) + 2(9£. Then
9 9
(
1) =
2£r12(£) if £ Ф
0. Consequently, by (2.5),
9 9is a derivation of B, (2.12) fl2( £ ) = A l A and * „ ( £ ) = 99(£)-2<9£ +
t2£2.
2£
W e thus get a description of functions in (2.10) and (2.11).
§ 3. Computation of g{ j ) . The symbols j and n have the same meaning as in § 1. Let gu, g12, g21, g 22 be the entries of the matrix g ( j ); by (1.3), g 12 = g21. In virtue of the identity n - j - n - j = j - n - j - n , (1.1) yields
g (j - n -j - n) = g { n - j - n - j ) = g ( { n - j - n ) - j ) = g { - j - j ) = g { - e ) = ||0||.
On the other hand, combining the identity r2 0 g ( n) = /( — 1)
with (1.3) we get
g ( j - n - j - n ) = g( { j - n) - { j - n) ) =
0 Я3
2 [^ u + (T 2+ ^3)] 0
0 2
[
0n -f-(т
2+ Я 3)]
Consequently, gn = g 22 = — (r
2+ Я3). Further, by (1.1),
— 2<7i2-f~ Я3 + 20 2gr12 2(9 r 2
^ 9 l 2 2© T2
— 2fi(j To — Яо-t- 4©
/ 1 1 0 Д /
1 0Д
= 5 •j H •1
\ 1
- 1 1I \
1 1/_
l i
- 1 0
v
1 0— 2© + Я3 gX 2 —
t2
l y
0 1)
1 1g
1 2 r 2 ^ 2^3
Hence g x2 = 20. Gathering the above equations we infer that
9(3) =
— (
t24- Я3) 2 0 2 0 - (т24- Я3) J
§ 4 . The concluding p art o f the proof. Using (2.3) and the above formulae for r(£), /i(f), g (j) and the corresponding matrix terms, and applying (0.4) and (0.5) we get the following relations, when xxx Ф 0
(4.1) ' yxx{x) = yx{x) = | i [ r а®?! — 20жи я?12 + Я8я?;2] — raJ +
9?(
ж12) 'ii
(4.2) yia(a?) = y2(®) [т 2 ХххХ 21 — в {хххх 22 + х Х 2 х2Х) + % 2 Хх 2 Х22\ + в\ф 1
2 Л
4>(хп) (Р(Х 22 ) +
<р{х2х) <р{хХ2)
$9.1 $Л9
(4.3) У2г(ж) = Уъ(х ) — \У 2 Х 2 \ х г\х 22 ~\~ h$X22\ ^з| ~Ь д>(х2Х) <р(%22) +
'21
I t remains to determine the form the function g( x) for matrices . with x xx = 0.
Let
' О X 10
e(x 2 (X 12X21 b)
X =
0 x X2 00 2X x 22
and x = x-j . Then g{x) = g{x-e) = g(oc-j-j) = g [ ( x - j ) - j ] = g( x- j ) . Com
bining (1.1) and the above results we get (if x xx = 0) yxx{x) — yi(x) — ^3Ж12 U
1 1
Уi
2(x ) = Уг(ж) ^ ~pp[ 0 xX 2 x 2 X-j- Л 3 х Х 2 х22] -j- 0 -b
У гг^ ) = Уз(^г) = [г 2ж21 2 0 ж 21 ж 22+ Я3 ж 22] ^з| рр
9»(®2l) ^ (^
1 2)
'2 1 X i
9? (#2l) (P(X 22 )
X 21 X 22
This means that the formulae (4.1)-(4.3) are also true for x xx = 0.
W e thus get (0.3); the entries of the matrix q turn out to be equal
to T „ - 0 , A 3.
326
Z. K a r e ń s k a R e fe r e n c e s[1] J. A c z e l,
Y orlesungen iiber Funktionalgleiclmngen und ihre Anwendungen,Basel und Stuttgart 1961.
[2] J. A c z e l und S. G-ołąb,
Funktionalgleichungen der Theorie der geometri- schen Objekte,Warszawa 1960.
[3] W . V. D. H o d g e and D. P e d o e ,
Methods of algebraic geometry,Vol. I, Cambridge 1947.
[4] Z. K a r e ń s k a ,
Ogólne rozwiązanie równania funkcyjnego F { x - y ) — F ( x )x
x F ( y ) , F macierz3 x 3 ,
х , у<=Grl(2,
B )(to appear).
[5] —
O pew nym równaniu funkcyjnym mającym zastosowanie w teorii obiektów geometrycznych,Zeszyty Naukowe Politechniki Krakowskiej 9 (1965), pp. 3-25.
[6] M. K u c h a r z e w s k i and M. K u c z m a ,
O n the functional equation F (A • Б )= F ( A ) - F ( B ) ,