Dirichlets Integrals

DIRICHLET'S INTEGRALS 15
which is exactly what it would be using the theorem.

But the theorem is true for n = 1 and n = 2, so it is true for
n = 3, and hence similarly for all positive integral values of n.
Applications, (i) Expansion for 2 cos n 6 in terms of 2 cos 9. Let
x = cos 6 + i sin 9, y = cos 9 — i sin 9; then x + y = 2 cos 0 and xy = 1.
Also xn = cos n 0 + i sin w # ; yn — cos n 9 — i sin n 0 so that
£" + yn = 2 cos n 9. Substituting these in the theorem, we obtain
N
n (n — r
2cosw0 = S ( — V
v
n — r\ r
r=o
(ii) To find the equation whose roots are the n-th powers of those of
ax2 + bx + c = 0. Let the roots of the given equation be a and /?.
Then the required equation is x2 — (an + /?") x + {a/3)n = 0
i.e.
r =o n—r
or, since a + /8 = — b/a and afi = cla,
[ N •»
(
r
r= .c/ n-
n-r\ r
124 HALBEATH ROAD,
DtTNFERMLINE.
Dirichlet's Integrals.
By L. J. MORDELL, F.R.S.
The integrals referred to here are those given by the following

Theorem :
Let / = f f if(x + y + z) xl~x ym~l z"'1 dx dy dz,
where the integration is extended over
x ^ 0, y^O, z ^ 0, a^x + y + z <^b,
and l> 0, m> 0, n> 0, 0 5S a ^ b < oo .
Then / = J when either,
https://doi.org/10.1017/S0950184300000124 Published online by Cambridge University Press

16 L. J . MOBDELL, F.K.S.
(1) f(t) is continuous for a sS £ <S b (and this includes the case when
f (t) is bounded and has only a finite number of discontinuities), or
(2) f(t) is bounded and monotone for a 5S t ^ b (and this includes the
case when f(t) is of bounded variation in a ^ t ^ b , but the same
proof gives both).
Probably most people have been not quite content with the
usual proof obtained by changing the variables. Either the procedure
seems to require more rigour than that usually given, or seems less
simple when rigorous as, for example, in Whittaker and Watson's
Modern Analysis. There is also the restriction that f(t) should be
continuous. My feeling of dissatisfaction led me to the following
proof. I do not claim any great novelty for it. It is in idea only
the old proofs given by Todhunter and Williamson but put in a
modern form. But still it may be welcome to many.
Lemma:
It is easily proved that the result is true when f(t) = 1 on
integrating successively for x, y, z.
Now divide the interval a rgj t 5g b into sub-intervals by the points
a = tQ, < tlt <t2, • • < tr = b. Denote by Ms, ms the upper and lower
bounds of f(t) in ts_x ^t<Ltg.
Then since / exists as a multiple integral,
/= £
the typical integral now being taken over
<s_i ^ x + y + z ^ ts, x ^ 0, y ^ 0, z :> 0.
Hence by the lemma,
f* *«-H-H.-l dt < / < i Ms

8= 1 J — —
Since J exists as a simple integral
Write D = S [' tl+m+n~l [f(t) - mg(or Ms)] dt.
To show / = J, we need only prove now that

D ->0 when ts — ^.j^ -» 0 for s = 1, 2, . . r and r -> oo .
Now | D | < S (M, — m t ) (tl+m+» - tls±™+n) / ( l+ m + n).

ON DESARGUES THEOREM 17
If f(t) is continuous, take the subdivisions so fine that

Mg — mt < e. Then
| D | < e (bl+m+n - al+m+n) f ( l + m + n),
-> 0 as e -> 0.
If f(t) is monotone or of bounded variation, take the sub-
divisions so fine that £+»+» — tl+™+n < e. Then
r
(I + m + n) \D\<e 2 {M, — ms),
8= 1
< 6 | / ( 6 - 0 ) - / ( a + 0)[, oreJC, say,

according as/(<) is monotone or of bounded variation,
-> 0 as e-> 0.
Hence the theorem is proved. The result clearly also holds for
the limiting cases a = 0 or b = oo when / (t) is not bounded in
a fS t ^ b if / , J then exist as improper or infinite integrals, and / (t)
satisfies (1) or (2) in every closed sub-interval of 0 < t < oo .
T H E UNIVERSITY,
MANCHESTER.
On Desargues Theorem
By J. H. M. WEDDEUBURN.
The usual proofs of Desargues Theorem employ either metrical

or analytical methods of projection from a point outside the plane;
and if it is attempted to translate the analytical proof by the von
Stuadt-Reye methods, the result is very long and there is trouble
with coincidences. It is the object of this note to give a short
geometrical proof which, in addition to the usual axioms of incid-
ence and extension, uses only the assumption that a projectivity
which leaves three points on a line unchanged also leaves all points
on it unchanged. Degenerate cases are excluded as having no
interest.
LEMMA 1. If the triangles ABC, A'B'C are in perspective from
0, and if B lies on A'C, B' on AC, then the triangles are coaxial.
Let X=(BC, B'C), Y=(ACB', A'CB), Z={AB, A'B'), D =
(B'C, OAA'), F=(YZ, OBB'), E=(YZ, BC), E'=(YZ, B'CD);
then YZFE projects from B into YAB'C, which projects into

Dirichlets Integrals

Uploaded by

Copyright:

Available Formats

Dirichlets Integrals

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dirichlets Integrals

Uploaded by

Copyright:

Available Formats

DIRICHLET'S INTEGRALS 15

which is exactly what it would be using the theorem.

The integrals referred to here are those given by the following

Then / = J when either,

https://doi.org/10.1017/S0950184300000124 Published online by Cambridge University Press

f* *«-H-H.-l dt < / < i Ms

Write D = S [' tl+m+n~l [f(t) - mg(or Ms)] dt.

To show / = J, we need only prove now that

Now | D | < S (M, — m t ) (tl+m+» - tls±™+n) / ( l+ m + n).

https://doi.org/10.1017/S0950184300000124 Published online by Cambridge University Press

If f(t) is continuous, take the subdivisions so fine that

< 6 | / ( 6 - 0 ) - / ( a + 0)[, oreJC, say,

The usual proofs of Desargues Theorem employ either metrical

https://doi.org/10.1017/S0950184300000124 Published online by Cambridge University Press

You might also like