-^^

^Ic^
^
THE
MESSENGER OF MATHEMATICS.
EDITED BT
J. W. L.
GLAISHER, So.D.,P.E.8.,
niLIiOW or TRINITT
OOLLiail,
OAKBBIDSH.
VOL. XVII.
[Mat,
1887"
Apkil,1888.]
MAOMILLAN AND 00.
Honlron anlr
Cambtfiige.
1888.
irr
7;
j/^^z".^
10^
lii'h
^fjc,iG^
CAMBRIDGE) :
PRINTED BT
W. MBTOALPK AND
SON,
TRINITY STREET AND ROSE CRESCENT.
CONTENTS.
ARITHMETIC,
ALGEBRA,
AND TRIGONOMETRY.
On the order of
proof
of
the
principal equations
of
spherical trigonometry.
By
M. Jenkins
.------
Note
on a theorem in
Higher Algebra. By
H. G. Dawsok
"
On
a theorem of
Prof. Klein's
relating to
symmetric
matrices.
By
A.
BnCHHEIM
.......
A
new method for the
graphical representation
of
complex quantities.
By
J. Brill
---.---
Note
on
the
anharmonic ratio
equation. By
Prof. Catlet -
Note
on
the
multiplication of nonions.
By
G. G. Morricb .
An
extension of
a
certain theorem in
inequalities. By
L. J. Roobrs
PAGE
30
69
79
80
95
104
145
GEOMETRY OP TWO AND THREE DIMENSIONS.
System
of
equations for three circles which cut each other at
given angles.
By
Prof. Catlet .......
18
On
plane
cubics which inflect
on crossing
their
asymptotes. By
F, Morlbt 61
Note
on
certain theorems
relating
to the
polar circle of
a triangle
and
Feuerbach's theorem
on
the
nine-point
circle.
By
S. Roberts - 57
On the
system
of three circles which cut each other at
given angles
and
hare their centres in
a
line.
By
Prof. Catlet - - - 60
On
systems
of
rays. By
Prof. Catlet - - - - -
78
Note
on
the two relations
connecting
the distances of four
points on a
circle.
By
Prof, Catlet - - - - - - 94
The
""
cosine" orthocentres of
a triangle
and
a cubic
through
them. By
R. Tttokbr
--.-.--
Geometry on a quadric
surface. By
Prof. Mathews
-
97
- 151
IV CONTENTS,
DIFFERENTIAL AND INTEGRAL CALCULUS AND
DIFFERENTIAL
EQUATIONS.
. PAGE
Note on the
Legendrian
coefficientsof the second kind.
By
Prof. Cayley - 21
The transformation of
multiple
surface
integrals
into
multiple
line
integrals.
ByJ.
Larmor --------23
Depression
of differential
equations.By
Lt.-Col. Allan Cunningham - 118
nomographic
inyariants and
quotient
deriyatiyes.
By
A. R. Forsyth - 154
THEORY OF ELLIPTIC FUNCTIONS.
On the tn^nsformationand
derelopments
of the twelve
elliptic
functions and
the four Zeta functions.
By
J. "W. L. Glaishhr - - - 1
On the second solution of the differential
equation
of the
hypergeometric
series,
and the seriesfor
K'fE', "c.,
in
Elliptic
Functions.
By
Brof.
W. Woolsby Johnson ------- 35
Expressions
for 8
(a;)
as a definite
integral.By
J. W. L. Glaishbr - 152
APPLIED MATHEMATICS.
Yortices in
a compressible
fluid.
By
C. Chrbb .... 105
n
MESSENGER OF MATHEMATICS.
ON
THE TRANSFORMATION AND DEVELOP- MENTS
OF THE TWELVE ELLIPTIC FUNCTIONS
AND THE FOUR ZETA FUNCTIONS.
By
J, W. L. Olaisher.
The
principal object
of this
paper
is to
give
the transfor- mations
of the
elliptic
and Zeta
functions,
which
are
due to
the
change
of
q
into -
q^
^
and
q^;
and also the firstfew
terms of the
expansions
of these functions in
ascending
powers
of the
arguments.
I omit the
demonstrations, giving
only
the resultsin
a
form convenient for reference.
This
paper
(in
so
far
as
itrelates to the
elliptic
functions)
may
be
regarded
as a
continuation of two
papers
^^
On
[Elliptic
Functions,"
which
were
published
in vol.
XI.
of the
Messenger, (pp.81-95, 120-138).
In
""
17"19 I have
given
tables of the values of the
elliptic
functions when the
argument
is increased
by Jf,
iK*
or K-\-iK\
and also the values of the functions for certain
special
values of the
argument.
I have found these tables of
such constant use
in
working
with the twelve
elliptic
functions
that I have been
tempted
to include them in this
paper.
The firsttable in
"
17 was
given
in vol.
XI.,
p. 88,
but the
functions
were
there
arranged
in an inconvenient order.
For the sake of
completeness
the
g'-series
for the
elliptic
and Zeta functions
are
also
given(""
2,3).
Notation^ "!"
"
1. The letters
p
and
u are used to
denote
2Jr
,
^Kx
"
and
IT IT
respectively ;
so that
u ="
px.
The
argument
x
is
supposed
to
be
independent
of k. The lettersh and Ji
are
used to denote
k^ and A'*
respectively,
VOL, XVII. B
2
MB.
QLAISHER,
TBANSFQBMATION AND DEVELOPMENTS
The
q^series
forIcp
snu,
"c.,
and
pZ{u)^"
2.
"
2. The
j-series
for the twelve
elliptic
functions
are
:*
hp
snw
= 2*
^ ^n-i
sin
(2n
-
1)
a?,
kp
cnM=
2"
y^7-5n=i
cos
(2"i
-
1)
a?,
pdnttsa 27
. ,n
co82yia?;
p
nsM=
-:;
"
+
2*
T-^-i"=i
sin
(2n
-
1)
x.
'^
sina;
*
1 -
^*^
* ^ ' '
p
dsw
=
-:
2!"
,
^
"n-i
sin
(2n
-
1)
a?,
^
sina?
*
1+2'
'
k'pncu^"
2r
(-p/^'-i
cos (2n
-
1)
a?,
'^
cosa?
' ^ '
1
+
2
Ap
cdtt=
Sr (-)"-
j^^,
cos
(2n
-
1)
x,
Mp
8d
u
=
Sr (-r
^^
"n (2n
-
1)x,
A;"
nd
tt
= 1 -
2^ (-)"-*
j-^
cos 2nar
;
"
Meittnger,
toI,
xvi., pp. 187,
188.
OF THE TWELVB ELLIPTIC AND FOUR ZETA FUNCTIONS.
3
and
those for the four Zeta functions
are ;*
pz (u)
=
sr
Yif^
"^"
'^'^^
i
to
th
2Van"/brmatu"Mof kp
snu,
"c.,""3,
4.
3. The
changes
in the functions
hp
sn
u,
"c irhich
are
dae
The
changes
in the functionswhich
are
due
to the
change
of
2
into
^
and
^
are
shown in the table
on
the next-
page:
t Messenger
f
yoI.
xv., p.
146.
b2
4
MR.
GLAISHER,
TRANSFORMATION AND
DEVELOPMENTS
OF THE TWELVE ELLIPTIC AND POUR ZETA PUNCTIONS.
5
The
quantities dn'^u"
h' which
occur
in the firstcolumn
of results
may
be
expressed
in the forms
(l"*')(cn'iu"*'sn'iM),
and the
quantities (\"1c)
(\^h snV)
which
occur
in the
second column
may
be
expressed
in the forms
dn'tf
"
k cn'tf.
"4.
In the
following
cases
the transformed
quantities
are
expressible
very simply
in terms of
sums or differencesof
elliptic
or
Zeta functions.
Transformations of pZJ(u)^ "
5.
"5.
The
changes
in
pZ{u)^P^,{^)t
"c" dae to the
change
of
2
into "
j*
are :
?. -?.
pZ("), pZ^u),
pZM, pz,{u\
pZ,{u), pZ{u),
6
MR.
GLAISHER,
TRANSFORMATION AND DEVELOPMENTS
The
changes
which
are
dne to the
change
of
q
into
q*
or q^
are
shown in the
following
table
:
The first two
quantities
in the second column
may
be
expressed
in the forms
\{pZ{^u)^.pZl\u)],\\pZSMu)-^pZi\u)}
respectively
;
and the second and third
quantities
in the third
column
may
be
expressed
in the forms
pZ[u)-^pZ,{u), pZ^{u)+pZJ,u)
respectively.
Transformations of
sum,
"c.,"
6.
"6.
It
seems
desirable to
give
also the transformations
of the functions
snte,cnu,
"c.
The
.changes
due to the
change
of
q
into
"q^x^x
OF THE TWELVE ELLIPTIC AND POUR ZETA FUNCTIONS,
7
The
changes
due to the
change
of
q
into
^
and
^
are :
snu
cnu
dnu
nsu
dsu
csu
dcu
ncu
scu
cdu
sdu
niu
^ '^
dn^u
dn^u
dn^u
1
dn^M
l
+
k'
sn^ucn^t^
1
cn'^M-f- A:'8n'^^
l+k'
sn^i^cQ^u
1
cn'^M
" k'
sn^u
cn'^M+ i'sn'^M
cn^^w
"
Aj'sn'^M
dn^M
cn'^u
" k'
sn'^M
(!+*')::::?
sn^ucn^t^
cn*iu
" k'
sn^'^u
cn'^M
" k'
sn'^M
cn'^M+
i'
sn*^M
a+^')rrf
sn
^i"
en
iu
cn*iw+ A;'Bn*^w
dnj^u
cn^^M+
k'
sn*^u
1 - A sn^u
8
MR.
QLAISMER,
TRANSFORMATION AND DEVELOPMENTS
Transformations ofZ{u)y"c.,"
7.
"7.
It
18
UDDecessaiy
to
give
the tranaforinations
of
-2^(m),
as
they
are
deducible at
sight
from
those of
pZJ^u)given
in
"5.
To deduce the transformations of
Z,{u)
from those of
pZ,{u)
itsuffices
to
replace,
in the transformed
results,
p
by
p
in the "
g
column,
1
2"
"
Transformations ofp,
A:p,
i
'p,AA'p, "
8.
"
8. In connexion with the
preceding systems
of formulas
it is convenient to
give
the transformations of
p,
kpy "c.|
which
are as
follows
:
Transformations ofsna;,
"c.,^{^),"c.,"9.
"
9.
The transformations of the sixteen functions
snar,
"c."
Z[x)^"c.,
are
easily
deducible from those of
sum,
"c.,
Z(w), "c.,
in
""
6 and 7
by simply
multiplying
the trans- formed
arguments
in the columns headed "
g,
g^
and
g*by
OP THE TWELVE ELLIPTIC AND FOUR ZETA FUNCTIONS.
9
For
example,by
the
change
of
q
into
-;,
q*
and
^S
sn^
becomes
respectively.
The
transformations
expressed
as
equatwnSy
"
10.
"10. By
the
change
of
q
into
^q
the modalas k is
converted into
p
and JTinto k'K.
By
the
changes
of
q
into
^
and of
2
into
^
the modulus X;isconverted into X and
7^
and
K is converted into A and
F,
where
Denoting
" and "
by
px
^'^^
Py
^^^ ^^^
^7
^
and
Wj
the transformations
given
in the lasttwo sections
may
be
expressed
as
equations
in the form
:
ikpsn
f
A'w,
77
)
=
ilck'p
sd
u^ "c.,
,
_.
--,
sn^Mcniw "
Xpx8n(t;, X)=^A;p
^^^^
,
"c.,
7P
en [w.7)
=
2A:*p "^; /
^ ,
"c.,
sn(x/|)=*'sd(|),
"c.,
X X
sn
;
"
j7
en
sn
(a:, X)
=
(1+
A:')
,
"c.,
Bn(ar,7)=
"^
"C.
10
MR.
aLAISHER,
TRANSFORMATION AND DEVELOPMENTS
Expansions of
the
elliptic functions
in
powers
ofx^%\\.
"11.
The
expansions
of the twelve
elliptic
functions in
ascending
powers
of
x are as
follows
:
8na?
=
;c-(l
+
A)|^
+
(l+
14A
+
A')^
a;'
-
(1
+
135A
+
1 35A'
+
A')
y7
+ "c.,
cna: = 1 -
1^
+ (1+ 4A)
^
-
(1
+
44*
+ 16A')
1^
+
"c.,
dna!
=
l-A^j
+
A(A
+
4)|^j-A(A'
+
44A4l6)^
+
"c.;
n8a!
=
^
+
i(l+ A)a
+
^(7-22A +
7A')|^
+
tk (31
- ISA - 15A'
+
31A')
jj
+
"c,,
dsa;=
i
-
J (A
-
A')" +
sV (7A'+
22AA'
+
TA")
J^.
-
X
Jy (31A" +
15A'A'- 15AA'"-
3lA")^j
+
"c.,
C8" =
^
-
i
(1+ A')
a; +
j'ty (7
- 22A'
+
7A")
^j
-
Ti, (31
- 15A' - 15A"
+
31A'')
^
+
*c-
J
dca:= 1
+
A'
^
+
A'
(A'
+ 4)
|^,
+
A'
(A"
+
44A'
+ 16)
|^,
+
"c.,
nca;
= 1
+
^
+
(1
+
4A')
2l
+
(1
+
44A'
+
ICA")^+"c.,
Bca! = "+
(1+
A')^
+
{1
+
14A'
+
A'')^j
+
(1+
135A'
+
135A'"
+ A")
^j
+
"fec.
;
.-""
cd" = 1 - A'
Jy
+
A'
(A'-4A)
Jy
- A'
(A"-44AA'+
16A')
J-j
+
"c.,
sdar
=
" +
(A
-
A')
^j
+
(A'
- 14AA'
+
A")|^
4
(A*
- 135A'A'
+
135AA" - A")
",
+ "c.,
nda = 1
+
A
^j
+
A
(A
-
4A')
1^
+
A
(A*
- 44AA'
+
16A")
1^
+
"c.
12
MR.
GLAISHEfi,
TRANSFORMATION AND DEVELOPMENTS
gz,aj
= - ;i'
{^
+
2
^i+2X1
+
A')^j +2X2+13A'+2r)f^+
"c.l
,
gz,:p
=
AA'|2 ^j+2'(A-A')|^j
+
2*(2A'-13AA'+2r)^,
+
"c.|
.
"
14. The other Zeta functions
iz^e,
"c.,
differ from
gz"cc
onlyby multiples
of
a?,
so
that with the sole
exception
of the
term
involving
x
the
expansions
of these functions
are
the
same as
those of
gz/c.
The
accompanying
table
gives
the
value of the
term
involving
x
for each of the twelve functions
iz,a;,
gz^, ez;r.
The
corresponding
terms in the
expansions
of
Z^{x)
and
Z {x)
are
shown in the next table.
OF THE TWELVE ELLIPTIC AND POUR ZETA FUNCTIONS.
13
The fanctions
^{x) are
the functions
so desig^nated
in
vol.
XV.
pp.
92-102.
The
quantities
H and H' denote
J (7+
G
+ JS)
and
J (/'+
ff
'
+ "')respectively.
"
15. It will be noticed that the series for
ffza;,gz,ar,gz,x
contain
A, A',
hh'
respectively
as
factors. The
only
other
series which contain
a
factor
common
to
all the terms
are
those for iz^ and
ez^a;,
neither of which contains
a term in
x.
The three series which contain
no
term
in
x are :
izur=-
aJ2J-2"(1
+
A)^
+
2*(2
+13A
+2A")
^j-"c.l
,
ez^=- *'J2^
+ 2-(l+A')|^+2^(2
+13A'
+2A")|^j+iS:c.| ,
gz,a?=
AA'
|2
1^4 2'{A-A')|'j +2*(2A"-13AA'+2A'") |^+"c.l .
"
16. The values of
Z^(x),
when
x
is
small,are
often
required
in
verifying
formulae and for other
purposes.
I therefore add for the sake of reference the
expansions
of
Z,{x)as
far
as
the terms
involving
x^
:
Z{x)=
-^a?-iAa:',
Increase
of
the
argument
byK^ iK\
K-\-
{K\ "
17.
"17.
In
working
with the twelve
elliptic
functions it
is
convenient to have for reference in
a
tabular form the
completesystem
of
changes
which
ar^ produced
in the
14
MR.
GLAISHER,
TRANSFORMATION AND DEVELOPMENTS
functions
by
the increase of the
argument by K^
iK and
These
changes
are
shown in the
following
table
:
The next
table,
which is deducible
at once
from the
last,
gives
the
corresponding changes
of
A;sna;,
A;
en
a;,
"c.
OF THE TWELVE ELLIPTIC AND FOUR ZETA FUNCTIONS.
15
These twelve fanctions form
a
group
complete
in
itself,
viz.
each function is transformed into another member of the
group
multiplied by "
1
or
"
i.
Valttes
of
the
elliptic Junetions
when the
argument
is
OjKjiK'jK+iK'y "18.
"18.
The
following
table
gives
the values of the
twelve
elliptic
functions for the
arguments 0,Ky iK',
K+ iK*.
The letter
a
denotes
zero
and A
infinity,
but the
following
more
precise significations may
be attached to these
letters^
viz. the
arguments
0, K^ iK'^
Jr+
iK'
may
be
regardedas
denoting
a,
K-^a^ iK'+a,K-\-iK'-\-a^
where
a
is
infinitesimal,
and A denotes -
"
a
16
MR.
GLAISHER,
TRANSFORMATION AND
DEVELOPMENTS
Thus the column headed 0 shows
that,
a
beinginfinitesimaly
sna =
a,
n8a
=
-,
dsa
=
-,
"c.
The column headed K shows that
en (Jr+ a)
= -
A?'a,dc(^+a)
=
--,
"c.,
and
80 on.
Digiti
ized
by
Goo
gle|'
18
PROF.
CATLBT,
SYSTEMS OF EQUATIONS
In the last column the values of the functions for
x^a,
a
beinginfinitesimal, are
added.
By
the aid of the three
previous
columns the values of the functions
2K-"ay 2iir'+a,
2^4 2iK'
+ a
for
a
infinitesimal
may
be written down at
sight
by affixing
the
proper
sign
;
for
example
sn(2^+a)
=
-a,
sn(2iX'4a)=
+
a,
sn(2Z'+2iX'+a)=-a,
cn(2Jr+a)=:-l,cn(2iJS:'
+
a)=-l, cn(2JS:+2tJr'+a)=+lj
"c.,
"c.
SYSTEM OF
EQUATIONS
FOR THREE CIRCLES
WHICH CUT EACH OTHER AT GIVEN ANGLES.
By
Prof.
Cayley.
Consider
a
triangle ABC^ angles.4,'B,
(7
(-4
4 B-\-
C^ir)
:
to
fix the absolute
magnitude,assume
that the radius of the
circumscribed circle is
=^1,
the
lengths
of the sides
are
therefore = 2
sin^,
2
sin",
2 sin (7
respectively.
On the three
sides
as
bases,
outside of
each,
describe isosceles
triangles
aBGy bGAj cABy
the base
angles
whereof
are =
a,
/8,7
respectively.
If
we
draw
a
circle
touchingaJ5,
a
(7 at the
points-B,
G
respectively;
a
circle
touchingiC,
IA at the
points(7,
A
respectively
;
and
a
circle
touching c-4,
cB at the
pointsAy
B
respectively
;
then these circlesform
a
curvilinear
triangle ABGy
the
angles
whereof
are -4+)8 +
7,
jB
+
7
+
a,
(7+a
+ )8
respectively. Taking
as
origin
the centre of the
circumscribed
circle,
and
through
this
point,
for axis of
^,
an
arbitrary line,
Its
position
determined
by
the
angledy
I write
for convenience
F=0-]-2By r^e-^Ay
^'
=
^
+ /S+
7,
"?
=
^
+
2jB+2(7, G^e^By B^B+y +
OLy
H=0y
E'=^0^2B+Cy
G'^C
+
a+fi;
then the coordinates of the
angularpointsAy By
G
are
(cos-F, sinjp'), (cosG,
sin
G^)
(cosS, sini7)
respectively;
and
the
equations
of the three circles
are
(sin
(-4
"
a)
"A' / sin
(-4
"
a)
.
-,A"
sin'-4
x +
" \
^cos^')+{y
+
" ^^
ism^'j
=-t-^,
sma
/
V sin a
/
sin a
'
/
8in(5-/S) ^Y
,
/
^sin(j5-/3). ^V
sin'5
[a
+
"
V--5"
'cos"?
+
(y+
" \
^
^
sin
g
)
=-^^
\ sinjS / V sm/tf / sin'/S
'
/
sin(C-7)
TT'V
.
f
8in((7-7)
. ^V
sin'O
{x+
"
-.
i^cosJ?
+
y
+
"
-'
^
sinJ?
=-:r-j-,
V
sin7
J V am
7
/ Bin'7
'
respectively.
FOR THREE CIRCLES. 19
In verification
observe that
we
have
0'H^2ir-2A, "?'-J?=-w+^,
(7-^'=27r-j4,
H-F^A^
H^F^
-25,
W^F^
ir-^-B,
H^-O'^
-5,
F^"^B,
F^G^
-2(7,^'-(?'=-,r+(7,
F^H^
-C7,
Q^H^Oj
hence
(cos
(7-
co8fi^)'
+ (sin
(7-
sm"^)'-
2 - 2
cos((7-fi^,
= 2
(1
-
cos 2^),
= 4 sInM
;
and
we
thus
see
that the sides are
" 2
sin^,
2
sin
5,
2 sin 0
respectively.
The firstcircleshould
pass
through
the
points
(cosO^
sin
CF)j
(cosjB',
sin
J?)
we
ought
tnerefore to
have tor
the firstof these
points
^^^BinM-")
_
.inV-")^BJn^l
Sin a
^ '
sin a sin a
'
that
isy
.
.sin
(-4-a) .
.
sin"(-4-a)
sinM
1+2
"
^cos-4
+ Vt
i
as
. "
,
sin a sma
sin a
Bnd for the second of the
points
the
same equation.
Write
for
a moment
^
vAuA
. sinf^-a)
^
-
X =
"
:
"
,
then
:
'-
9 JC
cosa
*
cos^,
sin a
'
sina
'
and the
equation
is
1
+2'(Xcosa-
cos-4)
cos^
+ (Xcosa- cos-4)'=s-r,
that
is,
1 " cosM =
-3l*
sin*a,
which is
right.
The second
and third circles should
intersect at the
angle
A\
that is
we ought
to
have
\ smp
sin7
/
\ sin/i^ sin7
/
siu^JS
sin'C
^
sin-B sin (7
.,
=
-^-fo
+
"^"T"
+
2
.
^
.
COS^
,
Sin
p
sin
7 sinp8in7 .
'
or reducing
and
for
cos
(
O' "
H') substitutfng
its
value,
= "
cos^,
the
equation
is
sin'(5-/3)
.
8in'((7-7)
.
,sin(5-/3)sin((7~7) .
"
jTB
*" " 5 T
^
i
7i
"
"
eos ^
sin
p
siu
7
sin
p
sin
7
sin*j5 sin'* G
^
sin" sin (7
=
-T-^TD
+
"
"
+2
.
^
.
cos-4.
sin
p
sin
7
sin p
sin
7
C2
20
PROF.
CAYLET,
SYSTEMS OF EQUATIONS
1
Writing
here
sm-B
^
sinJ?
'
^
Binp
'
siny
'
the
equatiop
is
(
Ycos^
"
cos
j5)* + (Zcos7
" cos Cy
+
2
(ycos/8- co8j5)(Zcos7- cos
C)
cos-4
=
r'
+
Z'
+ 2yZcos4',
Tiz.
this
is,
Y*
cos'^+
Z*
cos*7+
2
rZcos^ cos
7
cos^
- 2
Fcos^
(cos
j5+ cos Ccos.4)
-
2Zcos7 (cos
C+
cos-4 cos
5)
+ cos^jB^-cos'C+ 2
cos
-4
cos
5
cos
(7
4=r'
+
Z"
+
2rZcosJ'.
Keducingby
the relation-4
4 5+
(7=
ir,
thisbecomes
" 2
FcosyS
sin-4 sin (7"
2^cos7
sin-4
sin-B+
1 " cos-M
=
Y'
sin*i8 +
Z*
sln'7+
2
rZ(cosud'
-
co3/3cos
7
cos-4)*
Here -4'
=
-4
+ ^ +
7,
and thence
cos^' =
GOB
A
(cos^
cos
7
- sin
^
sin
7)
" Bin-4
(sin7
cos^S+ sin^S
C0S7),
and hence the
right
hand is
"r'8in"/3 + Z'sin"7
" 2
yZ(cos-4
sin^
sin
7
+
sinA
sin
7
cos 13+
sinA
sin^ cos 7)
or
reducingby
FBin)8=
sinJ5, ^sin7=sin(7,
thisis
=
sin'-iB
+
sin*(7"2 sin B sin C
cos
A
" 2 y
cos)8
sln-^ sin C"
2Zcos7
sin-4
sinjB,
and the terras in
Yj
Z
are
equal
to the like terms
on
the
left-hand
;
the whole
equation
thus becomes
- 1
+
cosM
+
8in''jB+
sin'(7- 2 cos^ sinj5 sin (7
=
0,
where
the last term Is
= 2 cos
-4
{cos(54- (7)
-
cos
B
cos
(7},
= " 2 cos*-4- 2 cos
-4
cos
J5
cos
(7,
=
- 2 cosM
+ (cosM+
cos'"
+
cos'(7-
1),
;
= " cosM
+
cos*J?
+
cos'(7" 1
;
PROF.
CAYLEY,
LEQENDRIAN COEFFICIENTS.
21
the
equation
is thus
- 1
+
cos*^
+ sin'J?+
sin'(7- co^'A
+
cos'5+
cos'C- 1
=0,
or
finally
it is -1
+
1+1 "
1=0,
which is
an identity.
The
formulae for the intersection of the third and first
circles,
and
for that of the firstand second
circles, are
of
course
precisely
similar
to the above formula for the intersection of the
second and third
circles;
and the verifications
are
thus
completed.
Cambridge,April7,
1887.
NOTE ON THE LEGENDRIAN
COEFFICIENTS
OF THE SECOND KIND.
By
Prof.
Cayley.
As
regards
the
integration
of the
equation
(l-*')g-2x|+"("
+ l)y-0
(na
positive integer),
it
seems
to
me
that sufficient
prominence
is
not
given
to
the solution
where
P^
is the
Legendrianintegral
of the first
hind,a
rational and
integral
function of
x
of the
degree
tz,
and
Z^
is
a
rational and
integral
function of the
degreen
" 1
;
viz.
we
have here
a
solution
containing
no
transcendental
funqtion
other than the
logarithm,
and which should thus be
adopted
as a
second
particular integral
in
preference
to
the form
y
=
C"
in which
we
have the infinite series
Q^
which is
an unknown
transcendental function.
Moreover,
the
expression usually given
for
Z^jviz.,
^_2n-lp
,
2n-5
,
2n-9
*""
l.n "-"'*"3(n-l) "-""^5(n-2)
"^'^
(to
term in
P,
or
PJ,
is
a
very
simple
and
elegant
one;
but the
more
natural
definition
(and
that
by
which
Z^
is most
readily
calculated)
is
*c +
1
Z^
isthe
integral part
of
iP"log
-,
when the
logarithm
22
PBOP.
CAYLEY,
LEGENDBIAN COEFFICIENTS.
is
expanded
in
descending powers
of
x^
viz. it is the
integral
part
of
"W,h-\^--)
(whence
also
Q^
isthe
portion containing
negative
powers
only
of
this
same series).
The
expressions
for
P^, Pj,...Pj
are given
in Ferrers*
"
Elementary
Treatise
on
Spherical
Harmonica^
cfec."London
1877,
pp.
23-25.
Reproducing
these,
and
joining
to them
the values of
^, Z^...Z^^
we
have
as
follows :
read
P,
=
fee'
"
i,
and
so
in other
cases.
p.
=(""...I) \v
-
W
+ W
-
A
,
P,=("'...x)
vy.
_
^
+
a^ys
_
ij
,
P,
=(*"...!)Vi^
-
AJiit
+ ifji
-
Vi/'
+
P. =(as'...a!)i.Hgi- AJfi
+
io^
-
Xj5i
+
IJI
,
"
II
,
V
+
A
,
z.
=("'...x)
vv
-
-4^
+ W
,
Z,
=(a/'...l) 4^
-
iji
+
w
-
H
.
z. =(x\..x)v^
-
"AV
+ V^ -^4Vy^
,
Z, =(x'...l)ifi"i-
-tflli
+ m^ -H\l^+
iff
.
MR.
LABMOB}
TBAKSFOBMATION OF SUBFACB INTEOBALS.
23
I notice that the numerical values
of
P" Pj
...,
P,,
for
a;
=
0*00, O'Ol,
...,
I'OO
are given [Report
of
the British
Association
for 1879,
"
Report
on
Mathematical Tables
")
;
as
the functions contain
only
powers
of 2 in their
denominatorS|
the decimal values
terminate,
and the
complete
values
are
given.
The functions Z have not
been
tabulated,
^
the
denominators contain
other
primefactors,
and the decimal
values would
not
terminate.
Cambridge,
Mareh
29,
1887.
THE TRANSFORMATION OF MULTIPLE
SURFACE INTEGRALS INTO MULTIPLE LINE
INTEGRALS.
,
By
J. Larmor.
An
integral
extended
throughout
a
volume
can
in various
ways
be
expressed
as a
surface
integral
over
its
boundary.
Many elegant
theorems of this kind have been
givenby
Gauss.*
L But in order that the
integral
over a surface,
of
a
vector
function, meaning thereby
the
integral
of its normal
component
over
the
surface,
may
be
expressible
by
a
line
integral
over
its
contour,
the function must
satisfy
a
certain
condition.
Li fact the
integrals
over
any
two surfaces
abutting
on
the
same contour would then be
equal,
and the two
together
would form
a
closed
surface,
such that the
integral
taken in
the
same
sense over
the whole of it would be
equal
to
zero.
Now ifB denote the
vector,
X, Y^Zita
componentd
parallel
to
the
axes,
and B
cose
itsnormal
component,
//iieo...".///(f.f.f)....
(,)
"
Therefore if this condition of
zero
integral
is to hold for
allclosed
surfaces, we
must have
identically, throughout
the
space considered,
dX
.
dY
.
dZ
^ ,-"
iJ+^+5?="'
(')
as
the condition
required.
*
Theoria
AUraetionit..., Comm, Soe,
GUtvng.^ II,1813,or W^rke, Band V.
24
MB.
ULBMOBy
THE TRANSFORMATION OF MULTIPLE
The truth of the formola
(1)requires
that the vector
should
not become discontinuous
or
its differentialcoefficients
infinite
anywhere
in the
space
in
question ;
for if that
were
not
providedfor,
the
integration
of its
ri^ht-hand
side
might
introduce other terms: cf.Maxwell's
ElectricitVy
Cb. I.
The
proposition
must therefore be
applied
in its
simple
form^only
when the
region
in
question
does
not
contain
places
where the vector is discontinuous
or
itsdifferentialcoefficients
infinite.
If
Xy Yj
Z
are
the
components
of
a
flux
R^
the condition
(2)
isthe well-known
"
Equation
of
Continuity,"
which
secures
that the flux is that of
an
incompressible
substance. Thus in
continuous motion of
incompressible
fluids the flux
through
any
ideal
aperture
is
expressible
as a
line
integral
round its
contour
;
the
reason
for which is obvious.
To determine the form of the
integral
relation in
question,
we
may
firsttake the
case
of
a
small
plane
surface.
Then
j{ada,
+ /3"7y)
^jjdxdy
(g
-
|)
(3)
by
immediate
integration,
the rule of
signsbeing
that the
line
integral proceeds
round the
contour
in the directionfrom
^ to
y
in the first
quadrant.
In the
same
way,
for
areas
in the
planes
of
yz
and
zx^
we
have
/0Si3,
+
,").//*"(|-f)
(4)
/(r"+.i.)-//.i"(|.-|)
(5)
By
what
precedes, expressions
to be
integrated
on
the
right-hand
are
to be taken
as
the
components
normal
to the
coordinate
planes
of the
vector function
B]
and
we
remark
that
theysatisfy (2).
We
are
entitled therefore to assert
for
any
small
plane
contour,
that
/(ada5
+
^dy
+
7cfe)
^JJdS.BcoBB
(6)
where the
components
of B
are
^"dy^dz' dz
dx'^^d^'Ty^'"^^
26
lOL
LAXXOB,
THB TSAaSTOUULTIOV GP ITULTIFLS
We
nutjr
obtain odmr fonns fiirAe theotem
hy
miiXmg
to
the
rigfat-hfliid
nde of
(9)
the line
int^nl
of
anj
exact
differential|
wbidi will add
nothmg
when taken loond the drcnit.
This fomrala
(9)
expnaMs
as a
line
int^^
the flax dae
to a
singlesonroe
o" floid
at
the
orig:in
of
eoordinates,
or
the
induction doe
to a
angleattracting particle
atnated there
;
and from it
any
more general
case
might
be deduced
by
summation. Bat
dcTclopment
in this direction
amply
leads
to
the well-Juiown
theory
of the
Tector
potential
in Electro- dynamics*
IL There is another class of
int^irsls
related to Mathe- matical
Physics
in which the
intends
are
extended
oyer
two
contoms. For
instance,
a
uniformly
luminous
open
surface
emits
a
quantity
of radiation
through
a
giren
aperture
which
depends
only
on
the contours of the sur"ce and
apeiturei
care
being
taken that all
parts
of
one contour are
yisible from all
parts
of the other.
Again,
the mutual
energy
of two
closed
electric currents
may
be
expressed
either
as an integ^
extended
over
their
circuits,
or as a surface
integral
derived
from the
equivalent magnetic
shells.
We
propose
now
to
investigate
the
general
forms of such
relations.
If
a
line
integral
round
a
contour
is to be
expressible
as a
surface
integralover a
sheet bounded
by
the
contour,
by
means
of
(6),
it
must
involve the elements of the contour
linearly.
Therefore the most
generaltype
of double line
integral
in
question
must involve both contours
linearly.
The
function
to
be
integrated
can
only
involve the distance between
two elements of the contours and the mutual inclinations of
the distance and these elements. If
r
denote the distance of
the elements
dsjds'y
and
i^,
B' the
angles
it makes with these
elements,
and
e
the
angle
between the directions of the
elements,
the most
genersd
forms therefore involve
only
//c"c"7(r)cose, (10)
and
JSdsda' ^ (r)
cos^ cos^'
(11)
Of these the latter
clearly
vanishes when either circuit
is
complete.
The
former
where
V^m',
vl
are
the directioncosinesof da'
;
^Jds'ffdB{X\'\- Yfi+
Zv),by (6),
SURFACE INTEGRALS INTO MULTIPLE LINE INTEGRALS.
27
where
\
fi^
v are the directioncosines of the normal to
dSj
and
r-/'(r)?^,
49,y,
z
being
the
components
of
r,
the
originbeing
taken
temporarily
at
the
position
of ds'.
Thus
changing
the order of
integration,
and
transferrins
the
origin
to the
position
of
dS^so
that
we
write "
^',
"
y
,
" "' for
a;,y, ",
we
have
jj,SJfir)
(fj^a.'.^^^dy'
+
'^^^d.-)
^ffdsjjds'{X'\'
+ ry
+
"v),
-;|:{;/'(r)|[-(y"
+
O^V
+ x'X(y"'
+
"V)+...+...]
-?/'(r)[XX'
+ /"/*' + !'"'']
^rj
|1/'
(r)l
[-
cos
*;
+
cos
^
cos
^
-
^/'
(r)
cos
*;
"
;|:{''/'(r)lcos,
+
r|^|i/'(r)|co."?co8^,
28
MR.
LABMOR,
THE TRANSFORMATION OF MULTIPLE
where
17
is the
angle
between the normals to
dS^
dS each
drawn towards the
positive
side of the
surface,
and
d^
ff
are
the
angles
between these normals and
r,
whose direction is
the
same
in both
cases.
Therefore, finally,
\\d8\\dS [r
1
1^/' w}cos"?cos^-l
I{r/'(r)}
cosi;]
^jdsjdsj^r)
cose,..
.(12)
where the
positive
side of the surface is determined
by
the
rule that
a right-handedscrew
in that direction
corresponds
to the direction of the line
integral
round it.
We have
proved
that this result is the
most
general
possible
of itsclass.
Particular
cases
may
be noted
as
follows
:
"
(i)
Make the
two circuitscoincide.
(ii)
Make the two
open
surfaces
coincide,
and
we
express
the double surface
integralby
a
double line
integral
round
the contour.
To avoid
infinities, f'[r)
must not contain
powers
of
r
lower than the inverse firet.
(iii)
Make the surfaces
plane,
so
that
17
isconstant.
(iv)
Make
/'W
=
-^;
then
jjdSJjdS'
"^'^^^r^^
= -
iJdsjds' logr
cose.
...(13)
The left-hand
side is the
expression
for the illumination
from S that is
intercepted by
8' when the
brightness
of 8
is
unity;
and it follows from
elementaryoptical principles
that this
quantity
must be
expressible
as a
line
integral
round the contours
of 8 and 8\
When S
and 8'
coincide,
we
have
27r5= -
^Jdsjdslogr
cose,
(14)
true
only
when 8 is
plane
;
for when 8 is not
plane
the real
optical
interpretation fails,
the
parts
of the surface not
being
in fullview of each other.
8UBFACE IMTEOBALS INTO MULTIPLE LINE INTEOBALS.
29
(v)
Make
/(r)
=
-^,
so
that
/'(r)
=-
J;
then
which is Neumann's well-known
expression
for the mutual
energy
of two
simplemagneticshellsy
or
of two linear
electric
currents.
(vi)
Make
/'(r)
=
Or;
then
/^5/(?/S'co8i7
=
-l/t"/("Vco8e, (16)
thus
giving
a
double line
integral
form for
JWdS^
where n'
denotes the
area
of the
projection
of 8*
on
the
tangent plane
at
dS, It
was
clear h
priori
that such
a
form must
exist,
for this
integral dependsonly on
S and the contours
of
8'^
while the other form
jUd8'
shows that it
dependsonlyon
the
contour of
fi^;
thus the form of the function of
r
that multi- plies
COSE
is all that remained h
priori
to be
determined,
and
that
might
have been found from the
simplestparticular
case.
When
one
surface 8 is
plane,
we
have
fi^n'
=
-i/i5/A'r'cose, (17)
where n' denotes the
projection
of /8'
on
the
plane
of 8.
Where
5,
8 coincide in
one
plane,
we
have
8''
=
-ifdsJd8r^
COBB
(18)
And
comparing
this with
(iv)we
deduce
(47r/8)"
=
{Jdsjds logrcose}
= - lir'
JdsJds
r*
cose...
(19)
for
any
plane
circuit
;
a
striking
result.
The theorems
justgiven
may
be verified
by
direct
integra- tion
when the surfaces
are
planecircles, [and(18)
without
much
difficultly
for the
generalsurface] ;
by applying
them
to surfaces bounded
by
other
curves,
we
obtain evaluations of
a
crop
of definite
integrals
of somewhat unusual form.
III. If elements of three surfaces enter
into
a
triple
integral,
the
components
of the elements of their three
contours must
enter,
each
linearly,
into the
corresponding
line
integral.
The
most
general
form of such line
integral,
independent
of
special
coordinate
systems,
which
gives
a finite
Talue when taken
over
completecircuits,
is
m{r,r\r")
dx
^
dy
J
dz
dx
,
dy'
,
e"'
d7!\dy'\
dz"
30
MR.
JENKINS,
ON THE OBDER OF PROOF OF THE
where
r,
r',
r"
are
the mutaal diatances of the three elements
of
contour
;
and the determinant is
equal
to 30 dsdi
dd\
where
"*=sini(a+
J+c)8ini(i+c-a)sini(c+a"J)flinj(a+i-c),
a, i,
c
ieing
the sidesof the
spherical triangle
determined
by
the directions of
rf", cfo\
cfo".
The
integral
may
therefore
hj application
of the method
of II be
expressed
as a symmetrical triple
surface
integral
;
the
general
formulae
are
long,
but the
degenerate
cases would
probably
be
interesting.
Finally,
there does not
seem
to be
any
reason
why
the
considerations
on
which these theorems
are
founded should be
restricted to the three dimensions
a;, y,
z
of
ordinary
space ;
but the
more
general
results would
probably
be of
only
analytical
interest.
ON THE OBDER OF PROOF OF THE PRINCIPAL
EQUATIONS
OF SPHERICAL TRIGONOMETRY.
By
ilf.
Jenkins^
M.A.
The
principal
formulas of
spherical trigonometry
are
made
to
depend,
wholly
or
partially,
on
three
independent
equations,
which
are
of a
less
simple
character than
most of those which
are
derived from them
;
that is to
say
on
cosa ~
COS"
cose +
sinh sine cos^
with the two
other
equations
of the
same
form.
Independent
proofs
are
given
of the
more
simpleequations
sin-4
^
sin-B
_
sin G
^
sina
""
sinft
""
sine
^
but
as
these constitute
only
two
independent equations,
a
third
is needed for the
completeinvestigation
of the
properties
of
a
spherical triangle.
I
propose
to take the
equation
sin
(A +
B)
_
cosg+cosft
sin 0 ""1
+ cose
(which
in the usual order would be
obtained
by
the multi- plication
of the
expressions
of two of Gauss's
theorems), givean
mdependentproof
of
it,use
the
properties
of colunar and of
polartriangles
to obtain the
equations
of similar
form,
and
then show how these
may
be
applied
to
prove
other formulas.
PAINCIPAL EQUATIONS
OF SPHERICAL
TRIGONOMETRY. 31
Ab a
leroma
take the
equation
-C08a+C0R"
cosm
2cos^c,
m
being
the median
drawn from G to
the
mid-point
of AB,
Let ABG
be
a spherical
triangle,
D the
raid-point
of the
arc AB^
0 the
centre,
Ca, 00,
CS the
perpendiculars
on
OAj
OBj
OD
respectively,
8/i,
hv
perpendiculars
on OA, 0B\
then because
CaO,
G0O
and CiO are rightangles,
the
sphere
on
OG SLB
diameter
passes
through a,
^8,
S,
Therefore
O,a, ffy
S
are
on a circle;
and because aOS =
/905,Sa^Sff,
also
Bfi^Sv]
hence
fia^v/B]
and
/a,
v
if distinct
from a
and
0 respectively
(coinciding
when
GA "
GB)
must
be
on
opposite
sides of those
points
with
regard
to
0,
because
the
angles
OaS,
O0B are
supplementary.
Therefore
cosa + cos
J OoL+ 00 20fi
^
,
=
"
j^
" =
-7^=2cosic.
coswi
Oo
OS
*
We
may
note
that the
same
proof
applies
if 2" be not
the
mid-point
of
AB] except
that,
instead
of Sa
being
equal
to
S0J
we
have
a0
^
aS
^
0S
sin
c
sinAD
""
sinBD
'
also OS.a0^
O0.a8+ Oa./3S,
whence we
have
cos
OD
,
sin
c
= cos a
.
sin^2)
+ cos
b
.
sinBD.
Proceeding
to the
equation
sin
(-44- B) cos a + cos
"
sin C
1
+ cosc
'
let D be the
mid-point
of the arc AB^ joinOD]
producei
making
DK^
GDy
and
join
KA^
KB
so as
to
obtain the
spherical
rhomboid
A GBK.
In
this,
as
in
planegeometry,
opposite
sides and
angles
are equal,
and alternate
angles
are
equal;
therefore
the
angle
GAK=iA +
B.
If -4
4-5"7r,
CAK
is
a
proper
spherical
triangle
and
denoting
OD
by
m,
Bin
OAK
^
sin2m
__sin2m
Bin-4(7jr'"
sin-^JT""
sina
'
sinA CK sin
^c ,
sinA
sin
a
"
-
"
A~
-
-^-^
}
^^^
-^"n
=
^'
"
;
sm^
sm/7i
'
BiuU smc
32
MR.
JENKINS,
ON THB ORDER OF PROOF OF THE
whence,by multiplication,
Bin
{A-{-B)
2 cosm 2 cosm
cos^
Bin
(7
""
2
cos Jc
~"
2
cob'^c
__
cosa+ cos
J
""
1
+ cosc
If A
+
B"Wj
CAK is not a
proper
spherical triangle;
but since CK is also
"
tt,
if
we
cut
off from it CO' =
tt,
and treat the
triangle
KAG'
as
before,
and
note
that
sin
{A +
B)
is of the
same
sign
as
sin
OJT,
we see
that the
previous equation
stillholds.
Next,by
means
of
a
colunar
triangle Ay a
beingunaltered|
B
changed
into
ir
"
Bj "c.,we
obtain
sin
[A^B)
_
cos
J -
cosa
^
sin 0
""
1 -
cose
'
from the
polartriangle, changingA^ir-^ay "c.,
we
have
sin
(q+ h)
_
oo^A
+
cos-B
sine
""
1 -
cos
0
'
and from another colunar
triangle,
sin (a "
I)
cos-B - cos^
,
sin^
+
sin B sinG
\
'-
= "
7=" ,
also
-T
r-j-
=
-: .
sine 1
+
cosU
'
BmaH-Bin6 sine
Hence
X i/"i.r"x
sin^+sinB sina+sini
.
^
sine
tani(^+5)
=
"
^j" -^
= "
inrr"
sm
G.
cos^-FcosjB sine
'8iu(a+6)(l" C08(7)
C08i(a-i)
.,^
co8i(a
+
6)
* '
.!^it/"i .
px_(cos^-co8^)(sin^
+ sin^
which,on
substitutionand
reduction, gives
cos'^(q-ft)_,,
^
j-^
cos
* O.
cos^c
*
Similarly
we
may
obtain the
rest
of
Napier's
analogies
and Gauss's
theorems,
the
sign
in
taking
the
square
root
being
determined
by
the consideration that the
greater
side is
opposite
the
greaterangle
in
a
colunar
triangle
as
well
as
in
the
original triangle,
that is ^
+
"-7r of the
same
sign
as
a +
6 -
TT,
as
well
as
J - JB of the
same
sign
as a
- 6,
CONTENTS.
PA OB
On the transformation and
developments
of the twelve
elliptic
and the four
Zeta functions
(continued).By
J. W. L. Glaisher - - - 17
System
of
equations
for three circles which cut each other at
given angles.
By
Prof. Catlby ..-----
18
Not" on
the
Legendrian coefficientsof the second kind.
By
Prof. Catley - 21
The transformation of
multiple
surface
integrals
into
multiple
line integrals.
By J. Larmor - - - -""
--23
On the order of
proof
of the
principalequations
of
spherical trigonometry.
By
M. Jenkins
--------30
The
following
papers
liave been received
:
Major
Allan
Cunningham,
"
On the
depression
of differential
equations."
Prof. W.
Woolsey Johnson,
"On the second solution of the differential
equation
of the
hypergeometric series,
and the series for
K\ "', Ac,
in
Elliptic
Functions."
Mr. F,
Morley,
"
On
plane
cubics which inflect
on crossing
the
asymptotes.**
Articles for insertion will be received
by
the
Editor, or by
Messrs. "W,
Metcalfe and
Son, Printing Office,Trinity Street,Cambridge
NoTioR. " A plate will be
given
whenever sufficient
diagrams have been
received.
No.
CXCV.] NEW SERIES.
s^
iL
JUL
18
1887
[July,
1887.
THE
-MESSENGER
OP MATHEMATICS.
EDITED
BT
J, W. L.
GLAISHER,
So.D., F.R.S.,
FELLOW
OP TRINITY
COLLEGE,
CAMBRIDGE.
VOL. XVH." NO. 3.
!|
MAOMILLAN
AND 00.
1887.
r.
MBTOALPB \
ASTD BON,
/
Price-One Shilling.
9
PBINCIPAL EQUA^teKSL^OffieitKIAL TBIOONOMETBT.
33
To obtain
cos
Cy
^
28in^co8(7 smB
sm(A+C) + sin(A- C)
cos
(7=
T" ;
:i
s=
r-: r
^:
"
5
-
2sm^ 2 Bin
A sin//
_
sinh fcosa
+ cose cose
"
cosa|
_
cose
"
cosa
cosJ
""2sina(
1
+
cosft 1
"
cos6
j
""
sinasini
For
right-angled triangles,
if
C=i7r,
sin^
may
be obtained
directfrom the
sine-aquation ;
co8c= cosa cos
6,byproducing
^Cto
A\
so that GA'^
CA;
then CB
being
the median of
the
triangle AGA\
COSC+
cosc=2 cos(
cosa,
cos^
".
"
H
=*
cosa,
from the
sine-equation by
a
modificationof
figure
;
or
otherwise,
cos^
_
sin
((7
+ ^)
_
cose -f cosa
_
cosa
cosft
+ cosa
sin5
""
sinjS
""
l
+
cos6 1
+
co86
=
cosa,
. _sin((7+^)
_
sinJg
sin(C-f^)
""
sin0
""
sinC sinjB
__
sinft
cose + cosa
__
sinb
cosa
""sine l
+
cos6
""
sine
^
sinftcos c
_
tan ft
"^
sin
c
cosft
"~
tan c
'
sin^ sin
a
1
+
cosft sin
a
tana
tanil =
sin((7+^)
sinft
cosc+cosa
sinftcosa sinft
'
^A ^n
sin(0+^) sin((7-f- JB)
cosc + cosa cose +
cosft
cot-acotJ?= "
\
p
^
"
1
"
:;
"
^
=
-" "
7
r"
siujB sm-4 1 + cosft 1 + cosa
=
cosa
cosft= cose.
The
equations
fora
quadrantal triangle may
be found in
a
similar
manner.
By eliminating
cose between
Bin(-4+(7)
sinft
cosa+cose ^siniA"C)
sinft
cose-cosa
"
'
A
=
r-,
T-and
" \" -r"^ =
-, r-
sm^
sma
1
+
cosft
sm^a sma
1 " cosft
we
shouldobtain cosft
cos (7s=cot a
sinft- cot^ sinG.
If the
equationcontaining
cos-4
+
cos5
were
deduced
from that
containing
cosa +
cosft
analytically,
instead of
by
the
use
of the
polartriangle,
we should have to determine
cos^,
cos
J?
separately
in terms of the sides. We
may,
however,
prove
the former
equation independently
thus
:
TOL. XVII. 0
34
MR.
JENKINS,
ON SPHERICAL TRIGONOMETRY.
In the
figure
to the lemma
proved
above the
angalar
equation corresponding
to
the lemma is
cos
B
sin
D CA "
cos
A sinBCD = cos
CD A
sin C,
the
signs
being
most
conveniently
remembered
by producing
BA
to
C\
and
measuring
the
angles
all
one
way,
that is
cosBsinjDC^
+ cos
CAC sinB CD
= cos
CD A smBCA.
To
prove
this,
draw the
arc
CN
perpendicular
to
ABy
CE
perpendicular
to ON;
jEa,JBJ8,
Ei
perpendicular
to the
planes OBC^ OCA^
OGD
respectively,
and EK
perpen- dicular
to OG. Then the five
pointsjB,S,
a,
K^ P are
on a
circlewith EK for
diameter, in a
planeperpendicular
to
0K\
a^ : /8S:
aS =
sinaE^ : smlSES :
sinoM
=
sin a
:
sinA CD
:
sin
J?CZ",
also
Ea=^ECcosCEa =
ECcosB; Efi^ECcosCE/3=^
EC
co"
Ay
-BS
= EC
cos
CES = EC
cos
CD
A,
and
J5z./8S=^/3.aS+
^.a/8,
whence
cos
BeinDCA " cos^ sinJ?CZ) = cos
CD A
sin
C,
Let ABC be
a
spherical triangle.
In BC
produced
make
CE=CA] join
EA and draw the
arc
FCD
bisecting
the
angleECA^
bisecting
EA at
rightangles
in
Fj
and
cutting
AB
produced
in D. Then
8in(a
+
b)
_
sinBE
_
smBAE
_
sinBAE
_
nin
DAF
^
sine
~
s'lnBA
"
sinBE A
"
sin CAE
"
sin CAF
*
but
cos^J9Fsin CAF-- cosAFCsin CAD
= cos
A CF sin
DAF,
and
cos
AFC
=0y
therefore
sin
(a+ J)
_
cos
ADF
_
cos
ADF sinA CB
sine
cosAGF"
cos
A CF sinA CB
__
cos5sin^ CZ"4co8i48in5 CZ"
_
(co85+cos^)cos^
(7
""
cos
(^Tr
"
i (7j
sin C
""
sin
i
C sin C
^
cosB+cos^
""
1 "
COS
0
In
a
similar
manner
it could be
proved
that
sin
(a
-
J)
_
cosB- cos-4
sine
"~
lH-cos(7 '.
April,11,
1887.
(
35
)
ON THE SECOND SOLUTION OF THE DIF- FERENTIAL
EQUATION
OF THE HTPER-
GEOMETKIC
SERIES,
AND THE SERIES FOR
K\ E\ "c.,
IN
ELLIPTIC FUNCTIONS.
By
Prof. W.
Wbolsey
Johnton.
1. The
following
solution of the
case
of failure of
one
of
the
ordinary
solutions of
a
linear differential
equation
of the
second order in series
was suggested
to
me by
Mr.
Forsyth's
solution of the
corresponding
case in
Legendre'sequation,
Messengerof Mathematics^
vol.
xvi.,
p.
162. The form of
solution is
interesting as
giving
the series for the functions
K\
E
"c.,
in
elliptic
functions at
once
in the form
givenby
Mr. Glaisher
{Gamh,
Phil.
Proc.j
vol.
v., p.
186),
and
accounting
for the factors which he has called the
adjuncts^
occurring
in the coefficientsof the series
{Camb.
Phil.
Proc.^
vol.
v., p.
240),
2.
Denoting
x-^hy^^
we
suppose
the differential
equation
*(^)y-a^"i(^)y=o (1),
which is the most
general
form for which
a
relation exists
between
two consecutive coefficients of the series. The
equationbeing
of the second order
we
may
assume
and
i"^
(5)
=^
(5
-
c)(^
-
rf),
where
^
is
a
positive
or
negative
constant. We
may
also take
the
exponent
s as
unity;
for
putting
z=x' and
z
-7-
=y,
we
have
sx ax 8
'
and
equation(1)
becomes
which is of the form
(^-a)(^-J)y-2^aj(5-c)(5-cZ;y
=
0
(2).
C2
36
PROF.
JOHNSON,
ON THE SECOND SOLUTION OP THE
We shalltake
equatioD
(2)as
the standard
form,recollecting
that when
8
is
not
unity
the roots
a,
b and
c,
d
are
to be
divided
by
",
and x' substituted for
x
in the final result.
3.
Putting
in this
equation
we
have
2{(m
+
r-a)(m + r-5)^^aj*^
-p{m
+ r'-c)(m
+ r-d) -4^*^*}
=
0,
and since the coefficientof a;"*^must
vanish,
{m+r"a) (m+r" 5)^,"p("n+r"
c"
1)(m+r-rf" l)^^j=0.
This
gives
the relationbetween consecutive
coefficients,
A
_
(m
"
c + r" l)(m-e? + r-l) .
and,
when
r =
0,
(m-a)(m-J).4,
=
0,
whence
w
= a or w
= J. Let
a ^ 6,
then
taking
tti =
a,
and the solution is
^
a
ft
.
("-c)(a-d)
(o-c)(a-c+l)(ffl-rf)(a-";+l)
-]
"^
1.2(a-6+l)(a-6
+
2)
^^*''
"^-J
*
Again,interchanging
a
and
b,
the second solutionis
R
"r,
"
(b~c-)(b-d)
(b-c)(b-c +
l)(b-d)(b-d+l)
1
"^
1 .2
(6-o+l)(i-o+2)
^^*''
"^-
J
"
To conform to the notation of the
hypergeometric
series
let
us
put
a
"
c=a a" c=a \
o-J+l=7
J
(3),
38
PKOF.
JOHNSON,
ON THE SECOND SOLUTION OF THE
The factor a;"**
=
a;".T*
=
a:"(l
+
Alogaj+...),
and
finally
the
remaining
factor
in
equation
^6)
is a
function of h which
we
shall denote
by ^ (A),
and which is such that
by equation (4)
y,=ajXO).
The
complete integral (6)
may
now
be written
y
=
^^.+B,7;.,4~(l + Aloga:+...)a;''[t(0)
+
At'(0)+...]
-^i^,
+
5,7;_.+5y,loga;
+ Sx"'(0)+ (8)
in which A is
put
for the
constant
-4^,
+
-r*
"
and the omitted
terms
are
terms which vanish with h. It remains to
express
"^'(0)
as a
series in
powers
of
x.
5. Let
H^
denote the coefficient of
{pxy
in the series
-^(h)
;
that
is,
let
i7"i ."^
^_(a+A)...(a
+ r-l+A)(/3-h^)...(/3+r-l-hA)
/i-i,anax2-
(i+ A)...(r
+
*)(7
+ A)...(7 + r-
1
+ A)
"
Then
f (A)=S."S: (pa;)',
and
^'(A)=2^
^'
(pa;)'.
Now
_p.r
1 1 1 1 1 1
^^^L"+A
"^a+l+A'""
a+r-l+A
"*"
/3+A'^*"'*'/3h-
r-l+Aj
When A =
0,
this becomes
^s^irj_
.
j^
1
L.1
"
La+
s"^/3
+
s
1
+ 5
7
+ 5J
in which
H^ now
denotes the coefficientof
(^pxY
in the series
FQi^ffj
7, j?a;), equation (4).
Thus
EQUA.TION
OP THE HTPERaEOMETRIC
SERIES,
"C. 39
and, writing
the
completeintegral
in
equation(8)
in the
forms
y
=
Ay^
+
JBy,,
we
have for the two
independent
inte- grals
when
7
is an
integral,
and
,. ly
(7-2)1 (y-1)!
y
, ,v
where
^-'"Ll-7U
+
/i
1 7/^^
1.2.7(7+1)
/I
.
1
.
1
.
1 111
-
1
\
/ M
,
1
(10),
and
Ty_^
denotes the
sum
of the first
7
" 1 terms
of
y,
in
equation(5).
When
7
= 1
we
have of
course 31=
0,
and
when
7
=
2, r,=a;-^^^
= a;-\
Thus the second solution consists of three
parts,
the first
of which is the
product
of the firstsolution
by log
a?,
the
second is
a
finiteseries
beginning
with o?*"^**and
ending
with
the
power
a?""*,
and the third is the
secondary
series
y',
which
is the
same as
y^
except
that each coefficientis
multiplied by
a
factor which
we
shall call its
adjunct^ consisting
of the
sum
of the
reciprocals
of the factors in the numerator taken
positively
arid of those in the denominator taken
negatively.
The firstcoefficientof
y^
which is 1 must
be considered as
having
the
adjunct
zero.
6. The law of the
adjuncts just
stated is the
same as
that
pointed
out
by
Mr. Glaisher in the
case
of the series for
JT',E^ "c.,
in the
paper
cited in
"1, although
the notation
differs from that of the
hypergeometric
series. The
reason
for this
persistence
of the law is
readilyexplained
as
follows,
and is illustrated in the
examplesgiven
in the
succeeding
sections.
In the first
place,
if
a
and
^8are
fractions
having
the
same
denominator
tw, say
a
= "
,
/8
= "
,
the coefficients in
y^,
m m
m m m\m ) m \m J
^TT"'
*
1.2.7(7+1)
40
PROF.
JOHNSON,
ON THE SECOND SOLUTION OF THE
may
be written in the form
m.rny
'
9n.27n.7117 (ray
+
tn)
If the
adjancts
be
now formed
by
the
same
law
as before,
each term of each will have
one
Tnth of its former
value,
and
the resultwill be the value of "
u\
so
that the law holds
good
for the
integral
"
y^j
when the coefficients
are
written
in the
new
form.
Again,
if
we
have
occasion
to
introduce
new
factors into
the numerators
or
denominators of the coefficientsin
y^^
so that,
for
instance,
we
write
j^y\
for the first
solution,
then each
adjunct
ifformed
by
the
same
law should contain the additional
terms "
7 .
If then
we
take
t y,
+
( t)^i
for the
second
integral,
the law of the
adjunct
will stillhold
good.
7. We
proceed
to
apply
the formulae
given
above in the
case
of the differential
equation
satisfied
by
K and
K\
the
independent
variable
being
the modulus k.
This
equation
is
*(l-*0g
+
(l-3*')|-%
= O.
Multiplying by
k and
putting
A
-^=.9,
whence
thisbecomes
yy-A;"(y + 2^+l)y
=
0;
and
putting
a5=i* to reduce it
to
the standard
form,
equation(2),
we
have
Here^^l,
and
a =
0,
a =
0,
6
=
0, a
=
i,
whence
" ,
'
"?=-i.
7
= 1-
:
"=yi=i
+
^.*"+^"**+-,
EQUATION OF THE HTP"RaEOMETRIC
SERIES,
"0.
41
Equations(4)
and
(9)give
then for
our two
integrals
l^.=yilogi"+ ^*(2+2-
1 -
1)
*"
+
^^]l;P^(2+2
+
"
+
|-l-l-4-i)**+....
It is
cnstomaiy
to
write the first
series,
whieh is the
"alue of "
,
in the form
IT
and
accordingly
we
write the value of
^y,,
which is
in which the law of the
adjuncts
is followed. Now
jr'=y.log4
"
iy,
so
that the law holds for the series for K'
which Mr. Glaisher writes in the form
"^'=
log|
+
j!
(log~Hl)*'+^(logf-f+!-|+})A:"+...
8. Consider next
the
equation
satisfied
by
E and
/',
whieh is
*(i-A^)"+(i-"o|+*y=o,
or
3'y-k\y-l)y
= 0.
Fatting
2 =
A^,
thisbecomes
where
again
^=
1,
and
a =
0,
a = 0
6 =
0,
a
=
-i,
whence
d =
-i, 7=1.
42
PROF.
JOHNSON,
ON THE SECOND SOLUTION 07 THE
Thus
y,=i+:itiA.+=iifc|iA*+...,
and
+
^^[^(-2+2+
2
+ 1-1
-l-i-i)A*+....
Bat
y,
which
is
the value of " is
usually
written
2E_
_
1 r.3 1'.3'.5
,j,
in which
we
have
dropped
from the numerators
of the coeffi- cients
the factor 1 which
corresponds
to the firstfactor
(" i)
in the coefficients
as firstwritten.
Accordingly
if
we
write
-^(-i
+
i
+
i
+
J-i-i-i-i)*'"
",
the law of the
adjunct
is not followed because the term - 1 in
each
adjunct
is
now
superfluous;
but
adding
y^
we
get
rid
of these terms
and have the
integral
y.+iy,=i+iy.iogi'-^"(i-i-4)A'
-^(i
+ i+i-i-i-i-i)**-...,
in which the law holds
good.
Now " /' is found to be
y^
log4
-(^^4-iy,),
and thus satisfiesthe law of the
adjuncts.
9. The
equation
satisfied
by
I and E^
is,
when
aj =
^,
in which
a=l,
o =
l,
6
=
0, a
=
i,
,
whence
^ .
^=-i,
7
= 2.
EQUATION
OF TUE HYPEBaCOMETKlC
SEBIES,
"C. 43
Hence
,,.i.[|+iJf^ilJii.H....],
and,
from
equation (9),
Here,
as
before, ^y^
would follow the
adjunct
law when the
coefficients
are
expressed
as
simple
fractions
;
but the
leading
2/
solution which is the value of is here
^y^,
thus
If
now we
write the value of
^y,
with
corresponding
coefficients,
we
have
iy.=iy.iogA"-i+g(i4i-i-i)A"+...,
in which the law is not
observed
because,owing
to the
new
factors,
1 in the
numerator
and 2 in the
denominator,
the
additional terms 1 "
^
are required
in each
adjunct.
But
2/
adding
one
half of the value of to
supply
these
terms,
we
have
iy.+iy.=i.y.iogA"-i+i(i-i)A*
in which the law isfollowed. The value of E' isfound to be
E'=^yi l^g4
"
(lyg
+ iy^),
and thus abo satisfiesthe law of
the
adjuncts.
10. Of the
seven examplesgivenby
Mr. Glaisher in the
paper
cited in
"
1,
we
give
but
one
more,
and that
on
account
of
an
apparent
anomaly
in
one
of the coefficients. The
equation
satisfied
by
/-i-(? and
-S'+
G^
is,
when
x=ik%
3(5-.2)y-a;(5-i)'y
=
0,
44
PROF.
JOHNSON,
ON THE SECOND SOLUTION OP THE
in which
a =
2,
a =
2,
i =
0, a=f,
whence
^ .
rf
=
J, 7=3.
Thus
y.=**[i+a*'+rlli **+"""]
and
y.=y.log*"--^^^ [l+^fe*-^**]
+i*[f|(*+*-^-*"*'+iill4(*+*+"+*-^-^-*-*)**+-]'
the
expression
for
Ty-i
in
equation (9)
in this
case
containing
two terms.
The value of -
^
^^'^^
is
iy"
thus
^
-iy.-2.4*+2'.4.6*+-'
and if
we
write
accordingly
the value of
i^y,
we
have
i^ry.
=
iSy.logA'-2(l-ii')+
ig(i
+ i-i-i)
*"+...
In
which,
in order to follow the
law,
the terms 1
+
I -
i
-
i
are
needed in each
adjunct.Adding
therefore
f
of the
preceding
series
we
have
iVy,+/jyi=iVytlog**-
2
+iA'
in which the law of the
adjuncts
holds for the coefficientof k^
and
higher
powers,
but
not for the coefficientof A;'. But this
term,
as we
have
seen
above,
is
part
of the
expression
for
T^
and not of that for
y'.
The value of JT
+
G^'is
46
PROF.
JOHNSON,
ON THE SECOND SOLUTION OF THE
where
T^
denotes the
sum
of the first
n terms
of the
preceding
series for
y"
and
^-'"[TFn-)('*^)*"'
""
i.2(n
+
i)(/i
+ 2)
(^
"*"*
"^^TTi
"*"
^TTa)
(^"'^^
"^"""J
'
in which the terms
in the
adjunctscorresponding
to the
a-
and
/3-factor8
vanish because these factors
are
infinite. This
solution
agrees
with that of
Hankel,
referred to
bj
Mr.
Forsyth
in the
paper
cited in
"
I.
12. When each of the functions
4"
"nd
4"^
is of the second
degree
we can
derive
a
series either in
ascending
or
in
descending
powers
of
x.
For
example, .Legendre's equation,
(l-a,')g-2xg
+
n(n-H)y
=
0,
may
be written
in which
p
=
1,
and
" = 2
;
thus
a =
i,
a =
i,
* =
^'
.
.hence
-=*(^-""'
rf=-i(n+l),
7
=
1,
and
bj equations (4)
and
(5)
the
integrals are
" ,
r,
,
(1-")(2
+
")
.
(l-")(3-")(2+")(4+")
-j
J^'-'L^"^
2:3
"^
2XI5
-e+'-j,
and
"
_ 1 ,
-n(\+n) -n(2-")(l+")(3 4-")
13. But if
we
write the
equation
in the form
(5-n)(5
+ n +
l)y-i
3(^-1)^
=
0,
EQUATION
OF THE HTPERGEOHETBIC
SEBIES,
"C. 47
we
may
take
"
=
-
2,
and therefore
J =
-in, a=^(n
+
l),
whence
^ , ,
c=0, /8
=
Kn
+ 2),
and the
integrals
are
y"-^
L
2(2w
+
3)
^
(w
+ l)(r"+ 2)(n+ 3)(n+ 4)
"I
"*"
2.4(2n+
3)(2n
+ 5)
""
'^ '"
y
and
_
" fi
w
(n
-
1)
^
w(w-l)(n-2)(n-3)
^
"1
^''-^
L^""2(2n-1)'^
"^
2.4(2n-l)(2n-3)
""
"^-J
'
14. When
n = "
i,
7
=
1,
and the two series
are
identical
;
and when 2n is
a
positive
odd
integer
7
is
an
integer
greater
than
1,
so
that the series
y^
contains infinitecoefficients. In
these
cases,
therefore,
we
have
by equation (9)
or,
taking\y^
as
the
integral, because,as
explained
in
"6,
\y'
will follow the
adjunct
law when the coefficients
are
written
as
in the
expression
for
y^
above,
sy.-yiiogj,
{"("-2)(n-4)...(l-")}'
^o*k^iy^
where
T^^^
denotes the
sum
of the first
" + i
terms of the
preceding
series for
y,,
and
*^
L
2(2n
+
3)
U
+
1
"+2
'
2n
+ 3/
-!)("
+ 2)(n
+
3)(n
2.4
(2"i + 3)
(2n
+ 5)
("
+
l)("
+
2)(n+3)(n +
4)^^^^"^
J^
48
PROP.
JOHNSON,
ON THE SECOND SOLUTION OP THE
where the
symbol(ad)
written after
a coeflSci^nt
denotes its
adjunct.
This solution is the
same as
that
given by
Mr.
Forsyth
in the
paper
before cited.
When 2/2 is
a
negative
odd
integer
the solution
may
be
found in like
manner,
or
at once
by putting
"
n
" 1 in
place
of
n
in the solution
justfound,
the result
being
^
-^-
n
,
-n(l-n) -n(l-")(2-n)(3-n)
1
Vi-'^
L^+2(l-2")*
+
2.4(l-2n)(3-2n)
*
+-J
"
and
iv
-V
Wl
-
{(-2"-3)(-2"-5)...4.2}'(-2"-l)
in which
T_^^
denotes
"^.
r
(n
+
l)("
+
2)
"*
L
2(2"
+ 3)
* +"""
+
("+
l)(""
+ 2)...(-"-3)
1
"^2.4...(-2n-3)(2n +
3)...(-4)(-2)
J'
and
*^
L2(l-2n)^'^^^
+
2.4
(1-2/1) (3-2/1)
^^^^
+-J
"
15. The functions
"^
and
4"^
in
equation (1)beingsupposed
real,
the
case we are considering, namely
that in which
7
is
an integer,
cannot arise when
a
and
",
the roots of
"^,
are
imaginary
;
but
c
and
J,
the roots of
ff"^^ may
be
imaginary,
and
a
and
/3
will then take the forms
a
=
/i
+
tV,
jS^fA--
tV.
In this
case
equations (4)
and
(5)
may
be written in
the forms
yi
-A
1
+
/*'+"'_,,(m'
+
i^'XC/^
+
O'+v'}
1.7
|"J!
+
1.2.7(7+1)
(px)'+...J
W,
(^+1-7)'
+
^'
'""
(2-7).
I
^" y.
=
x--[l
+
(2- 7)
(3-7)1.2
(i'^J+'-J-CS),
and for the
case
in wtucb
y
is
an
integer equation (9)
becomes
(
SEP 26 1887
)
EQUiLTXOM
OF THl!^"K"^|[^"^UHt5sEBIE8,
"C. 49
for the
nes
.f ty
(7-l)l(7-2)t
y
,
,0^
^ ^
i*'-' {(/* +
!
-7)'+"''l-.K/i-O'+F')}
""'^^ '"^ ^'
where
"*"
1.2.7(7+1)
V^"+
v"
(m+O'
+
v'
*
-^-7Ti)(i'-)'+-]-(io')'
These
forms,
with
a
change
of
sign
of
v',are
also
con-
venient when
a
and
/9are
irrationalreal numbers.
16. The form of the solution when
7
is
an integer
undergoes
modificationwhen either a or /3
is
an
integer
less
than
7.
In the first
place,
suppose
one
of them to be a
positive Integer
less than
7.
We cannot now
employ
equation (9),
because the coeflScientof
Ty-i
is
infinite;
in
fact,
B in
equation (7)
isnow
zero,
and the
complete integral
(8)
reduces to
m
which 7 isa finiteseries
containing 7
"
a terms.
It is to
be noticed
that, comparing
thiswith the
generalintegral,
T is
not
equivalent
to
y,,
but
we
have
in which the coeflScient
of
y,
takes the indeterminate
form,
but has
a determinatevalue when
7
and
a are
functions
of a
single quantity
and become
integers
simultaneously.
This
case,
in which the finiteseries is not
the
limiting
value of the infiniteseriesfrom which it is
derived,
occurs
in the solution of Kiccati's
equation
(see
the Memoir
by
J. W.
L.
Glaisher,
Phil.
Trans.^ 1881,
p.
771).
YOL. XVII.
E
50 PBOF.
JOHNSON,
ON THE SECOND SOLUTION "C.
17. In the next
place,
let
a or )8
be
zero or a
negative
integer, say
a
= "
n
;
then
y^
is
a
finite series of which the
last
term
is
.
-^fi(-n+l)..,(^l)/3(/3-H)".(/3-hfi-l)
"
""
fil7(7+
l)...(7+"-l)
^^^
"
In this
case, supposing7
an
integer,
the series
y
in the
second
integral
is
not a
finite
series,although
the series
y^
from which it is derived is finite.
Denoting
the term
of
y^
just
written
by N{pxf^
the coefficientis
"
.
.
/8(y8
+
l)...(g+"-l)
^^ ^
7(7+l)...(7
+ n-l)'
or,
when
fi =
0,
-^^=1.
The coefficient of the next
term,
that is to
say
of the
firstterm in
y^
which
vanishes,
is
the
corresponding adjunct
in
y
contains the term
,
which is
infinite,
the
remaining
terms
being
finite. Thus the
entire coefficientof ""
d^a-)"*^
in
y'
is
(n+l)(7
+ 7i)*
In like
manner, every
succeeding adjunct
contains the
same
infinite
term,
and the entire coefficientin
y
is the
same
that
it would be in
y,
with the omission of the
zero
factor.
Thus,
in addition
to the
part
of
y
formed
by means
of
adjuncts
fit"m
the actual terms
in
y^,
we
have,corresponding
to the
vanishing
part
of
y^,
the infiniteseries
("
+
l)(7+ ")
L
("
+
2)(7
+ n + l)^
1.2.(/3
+
"+l)(j8
+ w + 2)
(n+2)(n
+
3)(7
+ n+
l)(7+n+2)
ipxy+.
52
MR. F.
MOELEY,
ON PLANE CUBICS WHICH INFLECT
With the firstvalue for
m
the cubic is
y
(/
+
^\^y
+
3a,^")
+
36,(y*
+ 3a,"')
= 0
. . . (3).
Equation(1)
now
becomes
-
"8^(y
+
36,:ry" + 3a/y)
- 3
(-\ay
+
{^a^fi^
-
a,') ^'y
+
2"Aa3^y^-.aAy)
=
0,
or
\ {\xy'+
a,^'y)
-
hfl^x''
+
(2a,"3
-
a.)x*y
+
2
J^or/
=
0,
or 3
JgOry'
=
Sajb^x^^
or
y'
=
V'
(4).
Since for 3 real inflexions
a,
must be
positive,
write
a^
=
a*.
Then the lines
to
the other inflexions
are
^
=
"
axy
so
that
they
are
harmonic
to the
axes.
Substituting
from
(4)
in
(3),
or
4y
+ 3J,ar+12J,
= 0
(5),
which is the line of inflexions.
Let
PQR
be the
inflexions,
P
being
on
the
asymptote,
and
let the inflexional
tangents
form
a
triangle
pqr.
Let 0 be
the node and let
FQB
and
Fqr
meet the ar-axisin T and L
"?
.,^
It is calculatedat once
that the
tangent
at P is
3"-ar
+
y
+
3J3
= 0
(6),
and at
Q
and R
3J,x-8(y+3J3)
=
"9^y.
.(7).
Hence Itis manifest that
p
is
on
the x-axis
;
also
putting
y
=
0in(5),(6),(7),
Op="2.0T=8.0t.
ON CROSSING THEIR ASYMPTOTES.
53
Also from
(6),(7),
Or and
Oj
are
y
"
"Zax ......(8),
eo
that if
OQ^ OR^ Oq^
Or meet
the
asymptote
in
OQ,^OIt,^30q,
=
30r^.
The
equation
to
pQ^
is
1=
?*!^
or
3J,
3J,x~8(y
+
3J^-^-,
(Box-y)
= 2
(Sb^x
+
4y
+
12 J
J.
Hence
pQ^^
Or
and
similarly i?5" Og'
meet on P^J?.
These
are
of
course
not
all
indepenaent
results.
2. If the cubic inflects
on
crossing
a
second
asymptote,
say
at
0,
then
Oq
is
parallel
to
the
asymptote,
so
that
y
+
Sax
is
a
factor of
y'+ 3i,^y
+
Sa'x*.
Hence the other factor
is
j/
+
ax^
and
3",
= 4a. Hence
y
+ ox = 0 meets the
curve
at
infinity^
or
H is
at
infinity.
The metrical relations
are now
very
simple.
Let
OAy
OB be
parallel
to
the
asymptotes
at
P, Q.
Let
PFaqj QFffp
be the mflexional
tangents,
then,smce
0p
= 20A=^S0a and
Oq^20B=SO^f
54 MR. F.
MOBLEY,
ON PLANE CUBXCS WHICH INFLECT
aP^pq^
AB
are
parallel,
and
Pi
" 2-45=
34P=
35$,
therefore
BP=PQ=:
QA
;
hence
pB^
OQ are
parallel,
and also
qAy
OP.
Also in the
asymptotes
meet
in
E^
PDxDA^PEx OA^PQxAB^l
:
3,
therefore
PD =
\PQ
= QD.
Also
PQ
iscut
by
OD
in
a
ratio
= op : eJS?=a^ : PQ^^F: FQ,
therefore-Flies
on
OE. And since
EDiDO=^EQ\BO^UZ
and 0F\
FE^Op:
QE^Z\4.
itfollowsthat
DF\ FO^ OF:
FE
or
OF'^DF.EF.
.Clearly
all the linesin the
figure,
and all
got
by joining
the
points
in the
figure,
are
cut
in
very
simple
ratios.
(jriven
0, P, Q,
the
construction
for the
asymptotes
is
:
take on PQ points
A^
B such that BP=PQ^ QA^
then
OA^
OB are
in the directionof the
asymptotes.
For the
tangents
:
draw
Bp parallel
to OQ
to meet AO.
Then
pQ
is the
tangent
at
Q.
The cubic cannot
inflecton
crossing
all
asymptotes
unless
each inflexionis at
infinity,
and the
asymptotes
are
then the
inflexional
tangents.
3. A
non-singular
cubic
has of
course
much more
freedom.
But if its inflectson
crossing
each
asymptote
the
line of
inflexionshas an
envelope
which itis
proposed
to
investigate.
Let
J5r=ar+y+"=0
be the lineat
infinity, L=ax+by-\-cz=0
be the line
joining
the
points
where the
asymptotes
meet
the
curve,
and let
xyz
= 0 be the
asymptotes.
Then the cubic
is
LK^A-^mxyz^O
(9).
The firstdifferential coeiBcientsare
aJ5r*+22iir+
6my", ... .
The second
are,
omitting
a
factor
2,
2aX'+i,
...,
(6+ c)
/iL-t-i
+
3mx, ....
ON CROSSING THEIR ASYMPTOTES.
55
The Hessian formed
.from
these lastsix works down to
2^{(a-J)(a-c)x+...}
+
6mK{(h +
c)yz''
oar'
-f...}
+
37ni(2y"-ar'+...)
+
18m'jry"
= 0
(10).
Letar
= 0 and i
=
0,
therefore 2
(y+ ")'{(ft-c) (ft-a)y
+
(c-a)(c-J)z]
+
6in(y
+
"){("
+
c)y5?- Jy'-c"*}
=
0,
or
writing,
since
iy
-f c"
=
0,
y
=
"
c,
" =
6,
(J-c)'{-c(J-a)-J(c-a)} + 3m(6-c){-Jc(J+c)-Jc'-cJ'},
or
(b
-
c)'{2"C
-
a
(6
+
c)}
+
GrnJc
(J+
c)
= 0
... (11),
"i
" cs=0 would
mean an
inflexionat
infinity].
Suppose
two such conditionshold
:
(a-c)"{2ca-
J
(c+a)}
+ 6mca
(c
+
a)=:0,
(a
-
by {2ba-c(b + a)}+
Qmab
(ai'b)=^
0.
Subtracting
and
dividing by c
-
J,
ic(2a-J-c) + 6ma(6
+
c)+6wa'+ 3a'~2a"(2c+ 2J)
+
2(6VJc + o')a-aJc=:0.
Let
a,
i,
c
be the roots of
x*-px^+
qx-r=iO (12).
Then since
3aic+6ma(a+J+c)+3a'-4a"("
+
"3)f20(^+0*)-Jc("+c)=0,
therefore 3r
+
6map
+
3a*" 4a'
(p
-
a)
+
2a(2"*-2j-a")-^(;"-a)
=
0,
and if the 3 such conditions
hold,
a,
",
e are
givenby
4rar
+
Gmp^'+
5ar*-
4;?x'
+
2ar*
(p"
- 2
j)
-
pr
= 0.
Dividing
this
by (12)
the remainder is
3
(/
-
3j
+
2mp)
x'
+
(yr-;?j)
or,
66
ME. F.
MOBLET,
ON PLANE CUBICS WHic^ INFLECT.
and therefore
the
only
conditions which the
constaats must
satisfy
are
37
" "' Jc + ca +
ai-a'- """ c'
2m =
-*
"
^
s=
7
,
p
a-\-o-\-c
'
and
pq
= dr
(13),
or
a(J-c)'
+
"(c-a)*
+ c(a-")*=b.
Given
the
asymptotes,
the
envelope
of the Hne of inflexions
is the
envelope
of
aar -f
iy-h C2 = 0
subject
to
(13).
It
may
of
course
be
got byeliminating
a,
but the
following symmetrical
method
gives
the result in
a
better form.
Equating
dif- ferential
coefficientswith
regard
to
a,
J,
c,
and
neglecting
a
constant,
;r
s=
2
+
(J
+
c)p
- 9bc
or pa*
-
(2?' +
2
-
^)
a + 9r =
0,
and
a'-2?a'
+
ga-r
= 0.
From these
we
must
eliminate
a.
Multiply
the second
by
9 and add
;
9a*-8pa-(p'-8g'-a:)
=
0,
therefore
{9(p"
+
2
-
a:)
-
8;?'} {(p*+
j
-
x)(^
-
Sg
-
x)
+
72pr}
=
{p(p*
-
82
-
ar)
+
81r}',
or-9a:^+18a:'(/-32)-^^(V+lS/?-9.152*
+
8p*2-722'=0,
or x^'-Fx^+Qx^B^O,
where
i'=2(p*-32),
Q =/ + 2/2
-
152'
=
(/
-
32)(;""
+
52),
5 =
1 (/2-9A
therefore
y'-32
=
iP,
MR. S.
ROBERTS,
NOTE ON CERTAIN THEOREMS. 57
therefore
("-*^) (^^+")
=
^4'
or
(4g-P")(12"?
+
P0
=
288PH
(14).
The
envelope
is thereforea
quartic.
It
meetg
the lineat
infinity, P=0,
on ^ =
0,
which is the
minimum
ellipse
about the
triangle.
It meets
the
sides,
5
=
0,
on
4^
=
P*,
but this touches at
the
middle
points,
and also
on
12
Q -i-
JF*=
0.
Hence ittouches a:=0 at the middle
point
and cuts
itswhere
12y"
+
(y+i5)"
=
0,
"r
y
+
70 =
"4^30.
Hence the
shape
is
as
in the
figure.
Bath
College.
NOTE ON CERTAIN
THEOREMS
RELATING TO
THE POLAR CIRCLE OF A TRIANGLE
AND
FEUERBACH'S
THEOREM
ON THE NINE-POINT
CIRCLE.
By
Samuel BoberU.
1. If with reference to a
triangle
the centres of the cir- cumscribed
and inscribed
circles and the orthocentre are
respectively
0, 7,P,
and the radii of those circles and the
polar
circleare respectively -B,r, p,
then the
following
well-
known
relationshold
:
0P;
=
i?'+2p",
7P"=.2r'
+ p'.
58
ME. S.
KOBERTS,
NOTE ON CERTAIN THEOREMS.
The
proof
of the
second
equation
has been found
more
difficultthan that of the
former,*
while the
comparison
of
the
expressions
indicates that
they
should both be deducible
by
methods
closely analogous
or
differing correlatively.
In order to
obtain such
proofs
I make
use
of two theorems
which form
part
of the
generaltheory
of
conies,
but
which,
in relation
to
circles, strictly belong
to the
geometry
of the
front line and circle.
The theorems
are :
"
(I.)
If
a triangle
be inscribed in
one
circle and self-
conjugate
with
regard
to
another
circle,
any
point
on
the
firstcircle determines and is
a vertex of
a
triangle
inscribed
in the firstand
self-conjugate
with
regard
to the second circle.
(II.)
If
a
triangle
be circumscribed about
one
circleand
self-conjugate
with
regard
to another
circle,
any
tangent
to
the firstcircle determines and is
a
side of
a triangle
circum- scribed
about the firstand
self-conjugate
with
regard
to the
second circle.
2. Consider two circles
(0),(C),
whose
centres are C7,
C7'
and whose radii
are
5,
E. Let the intersectionof the circles
be real.
Draw VW
9k common
tangenttouching (0)
at V
and
(C)
at W.
Then, setting
out with this
tangent,
V is its
pole
with
respect
to
(C),
and from V
we
draw
Ffl,
and
again
VW
tangents
to
(G') (t.e.j
VW counts as two
coincident
tangents).
Then VW and VR constitute
a
limiting
case
of
a
triangle circumscribing
the circle
(G').
If this
triangle
be
self-conjugate
with
respect
to
(0)^
VB must
be
perpendicular
to
CW.
Take this to
be
so,
and let
CW^
C V intersectin 8.
Since
G'Wis
parallel
to
CV^ we
have
l8G'W=^C'8W=iCV8,
and 8 lies
on
the circle
((7).
Therefore CTr=-B
+
-B',
and
this
gives
Again,
letL be
an
intersectionof
(C)
and
(C),
and draw
LT
touching (C)
at L and
meeting((7)again
at T. Then
*
See Mr. Lendesdorf's
paper,
"Proofs of Fentrbach's theorem,"Uestengtr
^MathemaHetj to1"xiii.,p.
116 et
seq.
60
PROF.
CAYLEY,
ON THE SYSTEM OF THREE CIRCLES
It is
easy
to establish
also without direct reference to
conies
the
following
theorem :
"If
A^ By C,D, E^
F denote six
points
or lines,
and that
any
four of the
points
connect
equi-anharmonically
with the
remainingtwo,
or
any
four of the lines intersect
equi-
anharmonically
with the
remainingtwo,
and if
any
five of
the
points
lie
on,
or
any
five of the lines touch
a circle,
the
remaining
point
lies
on,
or
the
remaining
line touches the
same
circle."
Now let ABG
be
a
triangle self-conjugate
with
regard
to a
circle
(/S).
Take
any
point
D on
the circlecircum- scribing
the
triangle
ABG and draw the
polar
of D with
respect
to
(/S).Suppose
that
one
of the intersections of this
polar
with the circle
circumscribing
the
triangle
is E, Then
take F the
pole
of the line
joining i",
E with
respect
to
(/8).
The
triangle
DEF will be
self-conjugate
with
regard
to
(5),
and
by
the above
given
theorems F lies
on
the circle
through
ABGDE^
and must
be the second intersection of the
polar
of D with that circle. This is theorem
(I),
and theorem
(II)
can
be
similarly proved.*
ON THE SYSTEM OF THREE CIRCLES WHICH
CUT EACH OTHER AT GIVEN ANGLES AND
HAVE THEIR CENTRES IN A LINE.
By
Prof.
Cayley,
In the
system
considered in the
paper
"
System
of
equations
for three
given
circleswhich cut each other at
givenangles,"
Messenger
y
t. XVII.
pp.
18-21.,
we
may
considerthe
particular
case
where the centres
of the circles
are
in
a
line. The
condition in order that this
may
be
so
is
obviously
sin
{A
"
a)cosF\
sin
[A
"
a)sini^',
sin
a
=
0,
sin
{B
-
/3)
cos
0\
sin
{B
-
/3)
sin
G\
sin
^
sin
((7-7) cosjH',
sin
((7-7) sini?',
sin
7
that
is,
sin(5-iS)sin((7-7)sina8in((?'-5')+...=
0;
*
See Cremona's Geometrie t.
par
Dewulf,
p.
224. The bare statement that
if
one triangle can
be circumscribed abont or inscribed in a circle and self
-conjugate
with
regard
to a second
circle,
an
infinity
of such
triangles can
be
drawn,
does
not
fulljexpress
the result.
WHICH CUT EACH OTHER AT GIVEN ANGLES.
61
or
since sin
(G'-S"),
^miW-^F')
^m[r
-
G')
are ^sin^,
sinjS,
sin(7
respectively,
this is
sin(jB")8)8in((7"
7)
sin-4 sina
+ ...= 0,
viz.this is
sinA sin
a
sinB sin
yS
sin C^sin
7
_
sin(4-a)
"^
sin(jB-/3)
^
sin(a-
7)""
'
or as
this
may
also be be
written
1 1
2
cot-4 " cota cot
5- cot )8
cot(7" cot7~
Bat
assuming
this
equation
to be
satisfied,
it does not
appear
that there is
any
simpleexpression
for the
equation
of
the line
through
the three centres
;
nor
would it be
easy
to
transform the
equations
so as to have
this line for
one
of
the
axes.
The
case
in
question
(which
is
a
very
important
one
from
its
connexion with Poincar^'s
theory
of the Fuchsian
functions)
may
be considered
independently.
Taking
the line of centres
for the axis of
ar,
and
writing
a,
/8,
7
for the abscissaeof the
centres,
and
P, Q^
R for the
radii,
then the
equations
of the circlesare
(;r-a)'
+
/=P',
(x-/S)"
+
/="2',
and then if the
pairs
of circlescut
at the
anglesA^ B^
C
respectively,
we
have
Q'+ 2QBcofiA
+
B':={l3^y)\
R
+
2iZPcos5 +
P' =
(7
-
a)',
P'
+ 2P(2co8O+0"=(a-.i8)',
which are
the
equations connecting a,
/8,7,
P, Q,
B. See
the
figure.
It is to
be remarked
in
regard
hereto that if
-4,P,
G
are
used to denote
the interior
angles
of the curvilinear
triangle
ABCy
then the
anglesyA/3^olBj^
0Ca are =7r
"
-4,P,
G
respectively ;
whence if
P, Q,
B
were
used to denote the three
radii taken
positively,
the first
equation
would be
g*+
2
(^P
cos^
4-
P' =
(/8
-
7)*,
62
PROF.
CAYLET,
ON THE SYSTEM OF THBEE CIBCLES
as
aboTe
;
but the other two
equations
would be
i?-2i?Pcos5+P' =
(7-a)',
P'-2P(2co8(7+(2"
=
(a-^)';
hence,
in order that the
equations
may
be
as
above,
it is
necessaiy
that P denote the radius of the
circle,
centre
a,
taken
negatively
\
and it in fact
appears
that in
a
limiting
case
afterwards considered the value of P
comes
out
negative.
Similarly
as regards
the curvilinear
triangle ABC]
here
Ay
B{^B^)
and
C(= (T)are
the interior
angles
of the
triangle;
and the radius of the
circle,
centre
o^,
must be
regarded
as
negative.
Considering A^B^Oba given,
we
have an
equation
between
the radii
P,
$,
R. In fact this is at
once
obtained in the
irrationalform
\/(-3r)
+
Vl^O
+
Vl-^j^O,
and
proceeding
to
rationalise
this,
we
obtain
-2V(rz)=r+z-x,
that is
-V{(P'+ 2Pi?co85+i?) (P'+ 2PacosO+ ^)}
= P'
+ P(^ cos (7+
^
cosP)
-
^5
cosX
Hence,squaring
and
reducing,
we
find without
difficulty
0=
(2'i?sin"^ +
i2'P'sin'P+
PVsm'(7
+ 2P'^5 (cos^+
cosP
cos O)+ 2P(^R (cos^+
cos
(7
cos
-4)
+ 2PQR (cos
C+ cos-4
cosP,
or
puttmg
herem
P, Q^
B^
"r-
i "
"
y-
,
*his
la
(cos^
+
cosBcosC cosP+cos(7cos4
^'^'^'
sini^ sin 0
'
sinCsin^
'
cosO+ cos-4cos
sin^ sin P
^(f,i7,5)'
=
Oj
and it
may
be remarked that in this
quadric
form the three
coefficients
are
each less than
1,
or
each
greater
than
1,
according
as-4 + P+
C7"7r,
or -4+P+ C"ir.
Suppose
r,
A-k-B-{-
C"v\
the coefficients
are
here
B
cosX,cos/i, cosK,
the form is
(1,1,1,cosX,
cos
/A,
cosvXf
,
17,
?)*,
that is
(f
+
17
cos F + ?cos/i*)'
+
[if
smV
+ 217?cos\ + ?'
8in",
r{8m'
WHICH CUT EACH OTHER AT aiVEN ANGLES.
63
namely
thisis
,- ^ ,,
(
. ^cosX- cos/Ltcosv)*
(f+ 17COBf+?cos/ia)'+
|i7Bmv+? ^^ 1
(cos
X "
COB
/14
cos
V\
')
^y Jp
where the lastterm is
= -^ fsinV
sinV "
(cosX cos ll cos v)*]
;
sm V
^ ' / J I
where the coefficient
in
{ }
is
= 1 " cos'X "
cosV
" cosV
+
2 cosX
cos/x
cos
v,
namely,substituting
for
cosX,
cos/i,
cosv
their
values,
thisis
=
.
, ^
.
iT"
"
"At
(l-cosM-
cos'-B-cos'(7-2COS-4cos5cos
0)'.
sm
-4
sm
5sm C
'
It thus
appears
that the form is the
sum
of
three
squares,
and isthus
constantly positive
:
it therefore
only
vanishes for
imaginary
values of the radii
;
or
the
case
does
not arise for
any
real
figure.
Hence, 2'',
if the
figure
be
real,
-4
+
5+
(7"7r,
that is
the
sum
of the
angles
of the curvilinear
triangle
is less than
two
rightangles:
the radii
are
connected
as above
by
the
equation
(
cos ^+ cos
J?
cos
(7
cos
g
+ cos
g
cos
.4
'^'^
sin^sinO
'
sinOsin^
"
cos
G
+ cos
A
cos
B\
/sinA sin5 sin
G\
sin.4 sinjB
g\ /sin^
sin5
sInC\*_
in which form the three coefficients
are
each
greater
than 1.
Eestoring
therein
f
,
i;,
f
and
regarding
these
as
rectangular
coordinates,
the
equationrepresents
a
real
cone
which
might
be constructed without
difficulty,
and
then
takingf,
i;,
f as
the coordinates of
any
particular point
on
the conicalsurface
1 T^ ^ ^^
9in-4 sinJ5 sin CJ'
^m
. ,
we
have
P, Q^
-B =
"
^,
,
"
u-
" Obviously, points
on
the
same
generating
line of the
cone
give
values of
P, Q^
R which difierin their absolute
magnitudesonly,
their
ratios
being
the
same :
the
original equations
in fact remain
unaltered when
P, Q, jS,
a,
/3,7
are each affectedwith
any
common
factor.
64
PROF.
CAYLEY,
ON THE SYSTEM OP THREE CIBCLES
SupposingP, Qy
B taken
so as to
satisfy
the
equation
in
question,
then
taking
the radicals
'JiQ'+
2
QR cos
A
+ IT),
"J{IP-{'2RPcosB'\-r)y
V(P*
+ 2PCco8 0+^y
with the
proper
signs
we
have
a sum =0,
and
these
give
the values of
)8
"
7^ 7
"
a,
a
"
^
respectively ;
and the
construction of the
figure
would be thus
completed.
I look
at
the
question
from
a
different
point
of
viewj
takingQ^
E, 0
"
y
such
as
to
satisfy
the first
equation
^'4
2
"5 cos
id
+
JB'=t
(/8
-
7)%
(that
i"
starting
from the two circles
{x
-
^)'
+
y*= Q'f
{x-yY +
y^^R^
which cut each other at a
givenangle-4),
then the
problem
is
to
find
a
circle
{x
"
a)'
+ y*
=
P*, cutting
these at
givenangles(7,
B
respectively ;
and to
determine
the coordinate of the centre
a,
and radius
P,
we
have th6
remarining
two
equations
iZ''
+
2iilP
cos
P
+
P" =
(7
-
a)",
P'+2P(2cosa+(2"=(a-/S)".
namely,considering
a,
P
as
the coordinates of
a
point(io^
reference to
the
foregoing origin
and
axes),
and for
greater
clearness
writing
a =
x,
P=y
we
have
y*
+ 2yi?
cosP
+
i2"-
(x
-
7)'
=
0,
y'+
2y(2cosa+(2*-(x-/S)'=a,
or a"
these
may
be written
{j
+
R
COB By
-
(x
-
7)'
= - R'
sin*P,
(y4 0 cos
Cy
-
(x
-
P)*=
-
Q^sin'G,
namely^
the firstof these
equations
denotes
a rectangular
hyperbola,
coordinates of
centre
(x
=
7, y
= " P
cos P),
transverse
semi-axes
=P8inP;
and the second of them
a
rectangular hyperbola
coordinates of centre
(x=/8,y=-Qco8(7),
transverse
semi-axes
^QsinC: as
similar and
similarly
situate
hyperbolas,
these intersect in two
points
only
;
namely,
the
points
are
the intersections of either of them with the
common
chord
2y
(R
cosP-
Q cos (7)+
2
(7
-
^) {x
-
i (7+ /3)},+P'
-
Q'
= 0.
It is
possible
to construct a circle
through
the two
points
of
intersection,
and
so
to obtain
these
points
as
the intersections
%
66
PROF.
CAYLEY,
ON THE SYSTEM OF THREE CIRCLES
But
we
have
and this
equation
thus becomes
""-""=
16)8' Q'W {(cos^
+ cos
B
cos C)'
- sin"5 sin'
G]
=16^8'
(^R\-
l+cosM+cos'^+cos'
(7+2cos^cos5cosG).
We have therefore
2
{4/3*
-
[R
cos5-
e cos
C)'}
y
+
{4/9* (5
cos5+ Q cos C)
-
{R
cos5-
^ cos (7)[R"- Q')}
=
"
4/8Q5
V{- (
1 - cosM - cos*J?- cos*(7- 2 cos^ cos5
cos C)}
,
4(5cos5- ^cos(7)y
+
ii!*-
G'
=
4i8x,
or
completing
the reduction
by
the substitution
of the value
of
4/3*,
this is
y
{(^
sin"(7-f
R^
sin'5)+ 2Qfi(cos^
+
cos5cos
G)}
"^
QR{Q (cos5+
cos
G
cos^)
+
R
(cos
(7+ cos^ cos 5)}
=
"4/SQ5V{- (1-
cosM-
cos'J?- cos'(7- 2 cos^
cos5cos(7)},
viz.
we
have thus
two
values of the radius
y
(= P)
;
and to
each of these there
corresponds
a
single
value of the
abscissa
x,
givenby
4)8x
=
^'-
^ + 2(5co3J5-.QcosC)y.
The
two values become
equal
if -4
4 jB"
0=7r;
in
this
case
the three circles
meet
in
a
pair
of
points(x^,y,),(x,,
-
y^).
In
fact, writing
-4
+
5+
C=7r,
and thence
C08-4 =
"
cos
(5+ (7),
=
"
cos
J?
cos
(7
+
sinjB sin
(7,"c.,
we
find
{G"
sin'0+ 2^^
(cos^+
cosJ?cos
G) +
iZ'
sin'P}
y
"\^QR[Q(cosjB
+ cos
C
cos-4)
+
R
(cos
C+
cos^
cos5)}
=
0,
that is
(Q
sin
(7+5
sin
J?)'y
+ QR (^
sin 0+ R
sin5)
sin-4 =
0,
or
throwing
out
the factor
[QAnG-^-R
sin
J?)
this is
[Q
sin(7+ R sin
5)
y
+ $5
sin^ =
0,
WHICH CUT "ACH OTHER AT GIVEN
ANGLES.
67
and
we
then have
- 2
("
COB^
-QcmO)
QB fimA}.
The
term
in
{ }
is here
.
B?{AnB)
"\-B?Q{
8in(7-2
8in^co8J5)
+ 5^
(-
8in5+ 2 sin^
cos
C)
+
i^[-AnC),
which Is
=
i?( sin5)
+ 5'Q
(-
sin
(7+ 2 sinJ?
C08-4)
+ -""'(
8in5-2
8in(7cos-4)
=
(5*+ 0'+ 2i?(2cos^)[R
AnB^Q
sin
(7),
4^(^sin5-^sin(7),
or
finally
'^""^sin^H-
^sinC
_/3(Jg8in^-gsing)
^""
^sinjB+QsinC?
*
In these
equations
y,
x
should be
replacedby
P,
a
respectively;
and in
obtaining
them it
was
assumed that
7
= -
/3;restoring
the
general
values of
/8,
7
the
equations
become
p
"
^jRsin^
""iZsinJS+Csina*
^
i^o..A_4(^-7)(^8mg-"(g8ing)
"-i(/S+7)
i23in"+(2sinO
'
F2
68
PROF.
CAYLEY,
ON THE SYSTEM OF THREE CIRCLES.
viz.this last
equation
becomes
*""
EsinB-{-QainG
'
or
say
a {B
sin5+ Q
sin
C)
-
I3BsmB-yQ
sin (7=
0,
"y^hich
by
means
of the first
equation
becomes
a-^sin^
+ i8iisinjB+7(28ina=0.
It thus
appears
that the two
equations
are
sin^ BinJ5 sin (7
T
^
Q
^
B
=
0,
asin^
8 H\nB
y"inC
^
-p- "*"-"-+
S-=^'
viz. these
equations,
wherein -4
+ -B+C7=7r, belong
to the
case
where the three circles intersect in the
same pair
of
points;
hence if the coordinates
x, y
refer to
the
points
of
intersection of the three
circles,
we
have
simultaneously
the
equations
of the three
circles,
and the three
equations
which
determine the
angles
at
which
theyintersect,
viz.
we
have the
six
equations
(aj-a)'+
y*=PV
Q'+ B'-\-2QBcosA =
{^-'y)%
{x-j3Y+y'^Q\
fi'
+
P'
+ 25Pcos5 =
(7-a)',
(aj-yy'\-y'=-B\
P"+
(?"+ 2P(?cos
C =
(a -/3)V
viz.
from these six
equations,
with the conditionA
+ B-\-
G=irj
it must be
possible
to deduce the last-mentioned
pair
of
equations.
In the
general
case,
where
A-\-B-\-
G"ir,
and the three
circles do not meet
in
a
point,
then
taking
the circles
(x
"
/8)* + y
=
Q^ (x
"
7)*-f y^^B*
to be circles
cutting
each
other at the
angleA,
or,
what is the
same
thing,
the values
Qy By /8,
7
to be such
as to
satisfy
the relation
^ +
5' + 2Q^cos^
=
(i8-7)%-
the two
equations
for the determination of the abscissa of the
centre
a,
and the radius P of the
remaining
circle
give,by
what
precedes
MR.
DAWSON,
A THEOREM IN HIGHER ALaEBRA.
69
2
{(yS
-
7)"
-
{B COS
B-
Q cos (7)*}P
+
{(/8-7)'(^co85+
CcosC)- (Bcos5-
gco8C)(ii'-C'j}
="2(/8-7)C^V{-(l-co8*^-cos'5-cos"(7-2cos^co85cos(7)},
4(i2cosB-(2oo8(7)P+(iZ'-
e2')
=
(i8-7)(2a-i8-7),
viz.
we
have thus the two
circles
(x
"
a)*+
y*
=
P*,
each
of
them
cutting
the circles
(x-/8)*+y*=Q*,
and
(x
-
7)*+
y*
= jB'
at the
anglesC^
B
respectively.
NOTE ON A THEOREM IN HIGHER ALGEBRA.
By
H. O,
Dawson,
B,A.
The theorem
concerns
the
equivalence
of the
operators
and
"o^^
+
2a,^^
+
3a.A
+ "c,
when
applied
to a function of the
quantities
-4
,
A^^A^...A^j
where the transformation
x=pX'^qYy y=p'X
+
qY
trans- forms
the
binary
form
iaja.,..a)
(x.
yY
into the form
This
equivalence,
whicU is
of
importance
in the
theory
of
the covariants of the
form,
is
proved
somewhat
tediously
in
Faa-de-Bruno's Theorie des
formes btnaires^
p.
189,
the
equivalence having
been
previously
stated in Section 77.
The
proof
now given
seems
interesting,
not
only
as
being
expeditious,
but also
as
pointing
out
very
precisely
the
raison (TStre of the
equivalence
in
question, namely,
that it
is
a
necessary consequence
of the linear character of the
transformation.
The
proof
in
question
is
as
follows
:
Let
us
suppose
that the transformation
x={p^Xp')X+{q-^\q')Y, y=/J:+}'r
is
made,
then
we
get
a new
form whose coefficients are
dp dq
,
X
beingsupposed
to be
indefinitely
small.
70 MR.
DAWSON,
NOTE ON A THEOREM
Now, owing
to the linear
character of the
transformation,
we
might
have
proceeded
in
a
different
manner,
thus^
we
might
have
put
fl5
SB u + Xv,
y
=
t?,
and afterwards transform
by
the substitution
The firstof these two transformations
gives
us a
form whose
coefficients are
a^,a^
+
2\a^j
a +
3\a^j
"c.,supposing
X* to be
inappreciable,
whilst the second
^ives
(-4/,
-k/,
...,^/)
{X, F)*,
where
A^\ A^^...yA^
are
just
the
same
functions of
a^,a^
+
2Xa5,
a^
+
3Xag,
"c.,
that
-4^, A^...A^
are
of
a^,a^...a^.
But,remembering
our
former view of the
transformation!
we
have the theorem
in
question,
for
where
y|
+
j'|=8^.
The
proof
is
easily
applied
to other
forms;
take for
an
example
the
ternary
n'" and write it
as
follows
:
Suppose
that when
we
make the
transformation
we
obtain
{A)(Z, F,Z)\
IN HTQHEK ALaEBRA. 71
Let
as now
make the tranformatlon
" =
(X,
+
^A,
+
"f,X,)
X+
(^,+
^A*.
+
^At.)
r+
(v,+ 0v,+ ipv,) Z,
we
thus obtain
a
form whose coefficients
are
where
g;
=
X.
A
+
^,^
+
^,^
,
S.="c.;
again,
the transformation could have been
performed
in
another
order, namely,by
first
using
the substitution
and afterwards
putting
Now if
we
make the first
transformation,
we
easily
see that
the coefficient
a^^a
changes
to
[This
is
easilyobtained,
for 0 and
^
are
of
course
course supposed
to
be
very
small
thoughperfectly arbitrary
;
we are
therefore at
liberty
to
neglect
all
terms
higher
than
the first
(in
0 and
0)
of such
a term as
{U+0V+4"W)'^.
Eesuming,
if
we
perform
the second transformation
we
obt"iin
a
form
[A')
[XYZfy
whose coefficients
are the
same
functions of the
quantities
that
A^, -4,0, ...
"c.
are
of
a^,"r,
"c.,
and therefore
dA
where the 2
applies
to all such values of
Up^^
as can enter
into
Ap^tr*
72 MR.
DAWSON,
A THEOBEM
IN HIGHEB
ALGEBRA,
But we
obtained
before tbe
equation
HeQoe we
have
an
equivalence
betwen
the
operators
and between
forthis
reason,
that 6 and
^
are
perfectly
arbitrary.
By writing
the form in
powers
of
y^
we can
obtain two
operators
which
will be
equivalent
to
^
d d d
^ ^
d
d d
similarly
we
can
escpress
the
operators
^
d
d d
-,
d
d
d
in terms of the coefficients.
The invariants
of the form
satisfying
the six
equations
V
=
0,8/
=
0,8/'
=
0,8/'
=
0,8/''
=
0,8/'^=
0
are
the-common
solutions
of the
equations
thus found.
If
we
take
the
case
of the
quadric
form
(a,J,c,/,g^
h){ocy^Y
our equations
are
d .d
^
d d
^^d
d
Interchange a?,y
and we
get
Interchange y, ssj
'd/^^dh-'^^M^'^dh^^^di^'d^^
mi BQ on
for other forms.
Qhivrt'BCollege,Cambridge,
Vtt/i/ 80,
1887.
74
a
point
on
the
surface,
then
0
is
a
given
function of
(a?, y)
(or
more
generally
a?,y,
z are
given
functions of two
arbitrary
Earameters
u^
v);
and to determine the
congruence
we must
ave
a,
^8,7
each of them
a
given
function of
(a:, y)
or
of
(m,r),
such that
identically
a*
+ )8*+ 7"=
1,
but with
no
other
condition
as to the form of the functions.
2". The
rays may
be considered
irrespectively
of
any
surface
:
here
(a,^,7) are
each of them
a
given
function of
(a?,
y,
"),
such that
always
a*
-f /ff* +
7'
= 1
;
but there
are
further
conditions,
viz. it is assumed that
we
have
one
and the
same
ray,
whatever is the
point(a:,
y,
z)
on
the
ray ;
or
what
is the
same thing(using
p
to denote
an arbitrary distance),
that
a,
/8,
7
regarded
as
functions of
a;,y,
z
remain unaltered
by
the
change
of
a?, y,
z
into
x-^pa^ y+p^j z+py
respectively
;
this
implies
that
we
have
(aS"
+
/SS^
+
78:)a,("",
+
/8S^
+
7S.)/3,(oiS^ +
/3S^
+ yS)y
each
= 0
;
and,conversely,
it
may
be shown that when these
relations
are
satisfiedthen
a,
^,
7
remain unaltered
by
the
change
in
question.
The
equation
a*
+
/S*+ 7"
=
1,gives
aS.a
+
/SS^
+ 7S^7,
aS^a
+
ffSj3
+
yB^y,aS^oL
+ ffS0 + ySy
each
= 0
;
and
combining
with the last-mentioned
system
of
equations,
it follows that
^.^-^y7)
s^7-s.",K^-K^
are
proportional
to
a,
/8,7
;
or
say
=
ka^Je^jhy respectively.
Hamilton considers whether the
rays
are
such that
they
are
cut at
rightanglesby
a
surface
; supposing
this is
so
(say
in this
case
the
rays
are a
rectangular, or
orthotomic
system,
or
that
they
are
the normals of
a
surface),
then if
a;,^,
z refer
to the
point
on
the
surface,
we must
have
(kdx
-^r^dy^ydz
= 0
;
this
implies
a(S./3-S,7)
+ ^(S.7-S.a)+
7(S,a-S^)
=
0,
a
condition
which,by
what
precedes,
is
Aj(a* + i8*+
7*)
=
0;
viz.
we
must
have A:
=
0,
and therefore
KS^\y,
iji'tji,
8,a-S./3
each
=0;
that
is,
adx
+
fidy^ydz
must be
a complete
dif- ferential,
say
it is =dV. And
we
then have
Vssc^
the
equation
of
a
system
of
parallel
surfaces each of them
catting
PROF.
CAYLET,
ON SYSTEMS OP RAYS.
75
the
ra;^8
at
rightangles.Evidently
a,
/8,
7
=
8
F,
8^F,S^V
reBpectiveljy
and the function V satisfiesthe
partial
differential
equation
(8.F)'+
(S,F)'
+ (8.F)'
= 1.
Hamilton in effect considers
only
systems
of
rays
of the
form in
question,
viz. those which
are
the normals of
a
surface
(or,
what is the
same thing,
the normals of
a
system
of
parallel surfaces),
and it is such
a
system
which issaid
to have
the characteristic function V. It is shown that
a
system
of
rays
originally
of this kind remains
a
system
of this kind
after
any
number of reflexions
(orordinaryrefractions)
;
in
particular
if the
rays
originally
emanate from
a
point,then,
after
any
number of reflexionsat mirrors of
any
form what- ever,
they
are a
system
of
rays
cut at
rightanglesby a
surface. And
moreover,
there is
given
for the surface
a
simple
construction,
viz.
starting
from
any
surface which
cuts
the
rays
at
rightangles,
and
measuring
off
on
the
path
of each
ray
(as
reflectedat the mirror
or
succession of
mirrors)
one
and the
same
arbitrary distance, we
have
a set of
pointsforming
a
surface which
cuts at
rightangles
the
system
of
rays
as
reflectedat
the mirror
or
last of the mirrors.
The
ray-systems
considered
by
Hamilton
are
thus the
normals of
a
surface F"
c
=
0,
and
a
large
part
of the
properties
of the
system
are
thus included under the known
theory
of the normals of
a
surface
;
it
may
be remarked that
the
analytical
formulaB
are
somewhat
simplified by
the circum- stance
that V instead of
being (as usual)an
arbitrary
function of
(a;,
y,
z)
is
a
function
satisfying
the
partial
differential
equation(8"F)'+
(8^F)*+
(Sr)*
= l. In
par-
ticular
we
have the theorem that each
ray
is intersected
by
two consecutive
rays
in foci which
are
the centres of
curva- ture
of the normal surface
;
the
intersecting
rays
are
rays
proceeding
from the
curves
of curvature
of the normal
surface,
"c. There
are
other
properties easily
deducible
from,
but not
actually
included
in,
the
theory
of the normals
;
for
instance,
the
intersecting rays
aforesaid
are
rays
proceeding
from certain
curves on
the
mirror,analogousto,
but which
obviously
are
not,
the
curves
of curvature
of the mirror.
The natural mode of treatment of this
part
of the
theory
is
to
regard
the
rays
as
proceeding
not
from the normal
surface,
but from the
mirror,
and
(by
an
investigation perfectly
analogous
to
that for the normals of
a surface)
to
enquire
as
to the intersection of the
ray
by
rays
proceeding
from
con- secutive
points
of the mirror
;
it would thus
appear
that there
76
PBOP.
CAYLEY,
ON SYSTEMS OP RAYS.
are on
the
mirror
two
directions,
such that
proceeding along
either of them to
a
consecutive
point,
the
ray
from the
original point
is intersected
by
the
ray
from the consecutive
point,
but that these directions
are
not in
general
at
right
angles,
"c.
But in
regard
to
such
an investigation,
the restriction
introduced
by
the Hamiltonian
theory
is
altogether
unne- cessary
;
there is
no
occasion to
consider the
rays
which
proceed
from the several
points
of the
pairroras being
rays
which
are
the normals of
a
surface,
and the
question
is
con- sidered
from the
more
generalpoint
of view
as
well
by
Malus
in his
"
Th^orie de la Double
Befraction, "c.,"Paris,1810,
as
more
recently
by
Eummer in the Memoir
"
AUgemeine
Theorie der
gradhnigen Strahlensysteme,"
CrelUy
t. 57
(1860),
pp.
189"
230y
viz.
we
have in Kummer
a
surface of
any
form
whatever
(definedaccording
to the Gaussian
theory,
a;,y,
z
S'ven
functions of the
arbitrary parameters m,
r),
and from
e
several
points
thereof
rays
proceeding according
to
any
law
whatever,
viz. the cosine-inclinations
a,
/S,
7
(or
as
Eummer writes them
f
,
17,
g) beinggiven
functions
(such
of
course
that a*
+ ^8"+ 7*=
1)
of the
same
parametersm,
v.
It
may
be remarked: 1* that
Kummer,
while
considering
the
simplifications
of the
generaltheorypresenting
themselves in
the
case
where the
rays
are
normals of the
surface,
does not
specifically
consider the
case where,
not
being
such
normals,
they
are
(as
in the Hamiltonian
theory)
normals of
a
surface.
2^ that
some
interesting investigations
in
regard
to the
shortest distances between consecutive
rays,
while
naturally
connecting themselves,
with the normals of the
surface,or
with that of the
rays
which
are
normals of another
surface,
do not
properlybelong
to the
"
AUgemeine
Theorie
"
of
a
congruence,
which is
independent
of the motion of rect-
angularity.
It has been
already
remarked that the
system may
be
looked
at
in the two
ways
1' and
2*,
and itis in the former of
these that the
question
is considered
by Kummer;
it is
interesting
to work out
part
of the
theory
in the latterof the
two
ways.
TakingX^ Y,
Z
as current
coordinates,
we have,
for
a
line
through
the
point(a;,
y,
is),
the
equations
Z, F,
Z=aj
+
a/",y
+
^p,
"
+
7p;
a,
/8,
7
are
functions of
(a?,
y,
z\ satisfying identically
the
equation
a*
+ i8^+ 7*
= 1
(and
therefore the derived
equations
in
regard
to
(c,y,
z
respectively);
and also
satisfying
the
PBOP.
CATLET,
ON SYSTEMS OP BATS.
77
eqaations
(a8.+
i88^
+ 7S.)a
=
0, (aS.+
)88^
+
7SJ
/S
=
0,
(aS.+
i8S^
+
7S.)7-0.
It should be remarked that if these
equations
were
not
satisfied,
then instead of
a
congruence
there would be
a
complex,
or
triply
infinite
system
of
lines,
viz.
to the several
points
of
space
(x^
y,
z)
there would
correspond
lines
X, F,
Z^
x +
OLpyy
+ i"p, iS 4
7p
as above,
which lines would
not
reduce themselves to
a doubly
infinite
system.
Suppose
that the line
through
the
pointXy y^
z
is
met
by
the line
through
a
consecutive
point{x+ dx^
y
+ dt/jZ-^ dz)
;
then,
if
X, F,
Z referto
the
point
of intersectionof the two
lines,
we have
dx
+ adp +
pdOf dt/'\rl3dp
+
pd/S^dZ'\-ydp+ pdj=^Oi
and
consequently
dxy duj
d
dy, dp,
fi
dzy dy,
7
as a
relation
connecting
the increments
dx,dy^dz,
in order
that the lines
may
intersect,
viz. this is
a quadric
relation
(^ydxydy,
dzy
= 0 between the increments. In the
case
of
a
complex
this
equationrepresents
a cone
(passingevidently
through
the line dx
: dy :
dz^a:
fi: 7),
but in the
case
of
a
congruence
the
cone must break
up
into
a
pair
of
planes
intersecting
in the line in
quisstion
dx
: dy :
dz
=sai 0 i
y.
To
verify
h
posteriori
that this is
so,
observe that the
differential
equations
satisfied
by
a,
/8,
7
give
as
above
8,7-"A 8.a-S.7,S^-S,a
proportional
to
a,
^,
7,
or
say
=
2Aa,2A)8,2A7
;
and it hence
follows that the differentials
"a,S/3.By can
be
expressed
in
the forms
da "adx
+
hdy
+
gdz
+ JcdSdz
"
ydy),
dl3^hdx+
bdy+/dz
'\-
k
(y
dx -
adz),
dy "gdx+ fdy
+
cdz
+
k(ady -fidx),
where 0 =
aa + A^S+
gy,
0 =
Aa+"i8+/7,
0 =
flra+//3 +
C7,
.
78
PROP.
CATLEY,
ON SYSTEMS OP BAYS.
The
equation
= 0
dxj
doL
,
d
dy,dl3,
j3
dz^ ^7,
7
thus
assumes
the form
(a,J,
c,
/,
ffj
h']j[dxj df/j dz\^dy"fidz^adz"ydxy^dx"ady)
+
*
{(7dy
-
fidzy+ (ac?"
-
7^0?)'
+
(/Srfa?
-
adyY]
= 0.
Write for shortness
^dy
"
^dz^ oidz-r^idx^
^dx-
ady=^^yffj ^]
then
putting
for
a moment adx
+
hdy-^-gdzss"j
from this
equation
and
aa + A^S+
^7
= 0
we
deduce
a0 =3 ^ A
(^0dx
-
ady)
+
g
(adz
-
7^^),
that is^ "
hl^-^-gv
] or,
putting
for 0 itsvalue and
forming
the
analogous equations,
we
have
a (adx^-
Tidy
+ gdz^
= -
Af+
(717,
^Qidx^ hdy -\-fdz)
=
-/f
+ A?,
7
{gdx"\-fdy +
c")
=
-gri
+/f
,
and the
quadricequation
in
(dx^dy^
dz)
thus becomes
|(-A?+^i?)+|(-/f+*'?)
+
^(-i7'7+A)
+
^(r-H"*+n=0,
which,
in virtue of the linear relation
af+ ^i7+ 7f=0
connecting
the
(f
,
17,
(f),
breaks
up
into
a
pair
of
planes,
each
passingthrough
the line
{sOy
17
=
0,
^=0
(that
is
dx
: dy :
dz =^
a: 0 :
7).
We obtain at
once as
the condition that the two
planes
may
be at
rightangles
to each other
k
=
0]
that is
8,7-8A S.a-S,7,S^/3"8^aeach=0;
or,
what is the
same
thing, adx^^dy-\'ydz
=
dVy
a
complete
differential
;
and,
as was
seen,
this is the condition in order
that the lines
may
be the
normals of
a
surface.
It thus
appears
that in
a
congruence
each line is inter- sected
by
two consecutive lines which determine
respectively
two
planesthrough
the line
;
and
further,
that if for
every
line of the
congruence,
the two
planes
are
at
rightangles
to
each
other,
then the consecutive
lines
are
the normals
of
a
surface.
(
79
)
ON A THEOREM OF PROF. KLEIN'S RELATING
TO SYMMETRIC MATRICES.
By
Arthur
JBuchheim,
M,A.
The
theorem,
that if
a
is
a
real
symmetric
matrix the
roots of
I
a
" X
[
= 0*
are
all
real,
is old and well knowD.
Prof. Klein has
given
a
generalisation
of it in his
paper
on
the most
general
linear transformation of line coordinates.
Prof. Klein's theorem is
as follows: Let
a,
b be
two
real
symmetric
matrices.
Then,if,
when
we
reduce
hx^'\
to a sum
of real
squares,
the
excess
of the number of terms of
one
sign
over
the number of terms of the other
sign
is
r,
then the
equation|
a
" a;5
|= 0 has at least
r
real
roots.
I shall
prove
this
by
the method I have used
to
prove
the
particular
case
when 5=1 in
a
former note in the
Messengerof
Mathematics.
I
use
the notations
employed
in
my paper
on
the
theory
of
matrices and in
my
fourth note on the
theory
of
screws
in
elliptic space
in the Proc. Land. Math, 8oc. The
proof
can
easily
be translatedinto
ordinary
notation. I
only
consider
the
case
in which
a
and b
are
not both
vacuous.
Let b
not
be
vacuous,
then if
"
is
a point
such that
(a
"
Xi)6 =
0,
we
have b'^ae=
Xe,
so
that
e
is
a
latent
point
of
J"*a;now
let
8xy
=
xKby^
T^x=Sxx. Then it is
easy
to shew that if
e,
e'
appertain
to differentlatent roots of
a"*6,
we
have See
=
0,
and hence that if
a + */8,a
"
10 are
latent
points
corresponding
to
imaginary
roots T^ol
+ T'jS
=^ 0.
Now let
OL^"i0.oe
the
pairs
of
conjugateimaginary
latent
points,"
e^
the
remaining
latent
points.
Take these
points
of
reference.
Then,
if
a:
= S
(^. " IVi)(a"
"
^/S^)
+
2a?^.
"
.,
we
easily
see
that
T'a;=
22(^."-^/)
T^a.+
Saj/ye..
It
18 now
obvious that the
excess
of the number of
terms
of
one
sign
over
the number of terms
of the other
sign
is
at
most w
"
2r,
where
r
is the number of
imaginary
roots
of
Ia
" aJ
I
= 0
;
but
w
- 2r is the
number of real
roots.
Therefore the number of real roots is at least
equal
to the
excess
in
question.
This is Prof. Klein's theorem.
The Grammar School,Manchester,
June
4,
1887.
"
I write Ia I
for the content of the matrix
a,
t
If ^ = II
^(K
U
^ ^ symmetric matrix,
the
quadiic
bx^ is the
qnadric
Zb
je X
,
(
80
)
A NEW METHOD FOR THE GRAPHICAL
REPRESENTATION
OF COMPLEX
QUANTITIES.
By
jr.Brill.
1. As it is
now more
than
eighty
years
since
Argand
introduced the well-known method for the
graphical
represen*
tation of
complexquantities,
and the
principle
of
duality
has
long
been
recognized by
geometers,
it
seems
strange
that it
should have occun'cd to no one
to
apply
the
same
idea
to
tangential
coordinates
;
in other words to construct
a
diagram
in which
complex quantities
should be
represented by
lines
instead of
points.
It
might
be
anticipated
that the theorems
obtained
by
this method would be
none
other than the
polar
reciprocals
of theorems
similarly
obtained
by
the older
method
;
however,
since metrical
properties
oflFerconsiderable
diflSculty
to
reciprocation,
it
might
be advisable to have
a
method which would
yield
the
reciprocal properties directly*
I have endeavoured
to
supply
such
a method in the
following
paper.
It will be founa that the
theory
is not
an
exact
analogue
of that in which
complexquantities
are
represented
by points.
This is
as might
be
expected,
for it is well known
that
completeduality
does not exist in the
planimetry
of
homaloidal
space.
2. Before
proceeding
to
develope
the
theory
we
will
prove
a
theorem which will be of
use
in the
course
of the investi- gation.
0 is
a
fixed
point,
and
a
straight
line is drawn
through
O
meeting
n
fixed lines in the
points
r^,r,,
...y
r^.
A
pomt
R
is taken
on
the line
through0,
so
that
OB "ft^"^ OT,"^-"^"^-
The locus of S will be
a
straight
line.
To
prove
this take two
rectangular
axes
through0,
and
letthe
equations
of the
given
lines referredto these
axes
be
Mjic
+
Vjy
-
1 =
0,
u^x
+
1?^
" 1 = 0.
..tf^sTj
MR.
BRILL,
ON COMPLEX
QUANTITIES.
81
Then,
if B be the
angle
made with the
axis
of
x
by
the
line
through0,
we
have
TT"
= u.
cos5
+ V.
sin0.
Tr
=
^i
cos^
+
v,
sin
tf,
-pr"
= M
cos^
+ "
sin0,
Therefore the
equation
of the locus of B is
'^""^'^'"*"""'^'^'*=ffl,(^,cosg+t^,8ing)+...+7n",co8g+t^,8mg},
r
Swim
.
Smv
^
.
We shall callthisthe
mean
line with
respect
to
0 of the
n
given
lines for
multiples
Wj, wi" ...,
?n^.
3. Let the
equation
of
a
straight
line be
given
in the
form ux +
vy"l =0,
where
u
and
v are
the
reciprocals
of
the
intercepts
on
the
axes.
We shall denote the
position
of this line
by
the
expression
u 4-
iv,
or
if
we
desire to use
a
single
symbol
by
w.
This
beingpremised,
we see
that the
lineat
infinity
will be
represented by
zero,
and that
any
line
through
the
origin
will be
represented by
an
infinite
complex.
Beal
quantities
will
represent
lines
parallel
to the axis of
y^
and
purelyimaginaryquantities
will
represent
lines
parallel
to the axis of
x.
Further,
any
line
parallel
to
the
given
one
will be
represented by
a
real
mjiltiple
of the
expression
which
denotes the
given
line.
4. Let there be two
givenstraight
lines whose
equations
Bre t"aj +
vy
" 1 = 0 and
ux + v'y
" 1
= 0. Consider the line
m {ux+
vy
-
1)
+ n
[u'x+ v'y
"
1)
=
0,
or as
it
may
be written
mu^nu' mv +
nv*
x^
y-
1 = 0.
Let thisbe
equivalent
to
Kb
+
T^
" 1 =
0,
then
we
have
{m
+ n)U=mu +
nu* and
(m + ")
r" twv +
"v'.
VOL. XTII.
G
82 MS.
BRILL,
A NEW METHOD FOB THE GRAPHICAL
Hence,
if
W=V+iVj
w=u+tv^
and
w'=u'+w'^
we
have
(wt+ n)
W^mw-{-nw\
The line W is what
we
have
calledthe
mean
line of the
lines
w
and w' for
multiples
m
and
n.
If
we
make
m
=
n,
then
we
have 2W=w
+
xo\ In this
case
W will coincide with the harmonic
polar
of the
origin
with
respect
to
the lines
w
and w'. Thus to obtain the line
which is the
sum
of two
givenlines,
we
have to draw the
harmonic
polar
of the
origin
with
respect
to
them,
and then
to draw
a
line
parallel
to this and at one-half the distance
from the
origin.
Thus the idea of the
mean
line furnishes
us
with
an
interpretation
of the addition of
lines, justas.
the
idea
of the
mean centre furnishes
us
with
an interpretation
of the
addition of
points.
The actual
position
of the mean
Jine,however, depends
upon
the
position
of the
origin,
so
ths^tthe
jinalogy
is not
quitecomplete.
If
we write
??i 4 n =
0,
then TV
becomes a
line
through
the
origin,
apd all that
we can
say
is that
w
" vS is
some
line
{parallel
to this. It will therefore be
necessary
to
investigate
he
case oi
'{jo
"
yi
separately.
The
equation
of thislineia
(u
"
w')a; + (v
-
v')
y
" 1
=? 0,
This is
ohvioqsly parallel
to
(t*
-
w')
a?
+ (v
-
v')
y
=
0,
the line
joining
the
origin
to the intersectionof
vo
and v)\
In fact this latter line coincides
with TT.
Further,
the
equation
of
to
" to'
may
be
written
"iB +
vy
"
1
"
[yix+ v'y)
=
0,
which shews that it
passes
through
the intersectionof w with
H
Une drawn
through
the
originparallel
to w\
Thus,
to
construct
the line
v"
"
w',
we draw
through
the
origina
line
parallel
to
w\
and
through
the intersectionof this with
w we
draw
a
line
parallel
to that
joining
the
origin
to the intersex*
tion of
w
and w\
.
If TF be the
mean
of the lines
"^j,w^, ..., ""^
for
multiples
m^
w"
.,., "i^,
then,by proceeding as
in the
early
part
of
this
article,
we
should obtain
(Wj
+
w,
+
. . .+
?w
J
Tr=
m^^
+
m^^
+.
.
.+
Vijio^^
6,
Let the
straight
linet"aj +
t?y
" 1 =0 cut the axis of
x
fit
the
point-4,
and that of
y
at the
point5;
and let OA^a
l^ndQB"^b, From Q draw OL
perpendicular
to -45. Let
BEPKESENTATION OF COMPLEX
QUANTITIES.
83
OL=p^
and let 0 be the
angle
which OD makes with the
axis of
X.
Then
we
have
1
CO8
0
,
1 sin^
tt=
- =
,
and
v=T
=
.
a
p
^
op
Therefore
1
" .
"*"
w = u
+
iv=i-
(cos5+ isind)
=
"
"
P V
This
formula will enable
us
to discuss the
question
of the
multiplication
of two lines. For let
wlc
=
WjW^
where
w,
fff^j w^
denote linesand
c
denotes
some length. Substituting
for
Wj w^
and
w^
their values
as
givenby
the above formula
we
obtain
cosg
+
tsing
_
cos
0^
+
i sin
0^cos^^+
i
sinS,
^
cos {0^+ gj +
i sin
{0,
+ 0,)
Hence,
we
have
co8g^cos(g^-f
g,)
.
sing
_
sin
(g^
+
g,)
^ PiPt
*"
cj"
'
.
p,p,
'
from which itfollows that
^-P\P%
**^d
0^0^-\-0^.
In
a
similar
manner it
can
be
proved
that if
vf^vi^^
then
P*
'=^PxPt
and 2g =
g,
+
g,.
6. As
a
first
example
of
the
utility
of the method
we
will
investigate
the
analogue
of the well-known
proofby Argand's
method of "uc. I. 47.
Taking
the
figure
of the last article
produce
BO to
B, making
OJS'= OB. Then the line AB
will be
represented by \\a-\-i\h^
and the line AB^
by
\\a
"
ilh. Further,
from the last article
we
have
1
t
_
"*
a 0
p
Also,
the
perpendicular
from 0 on
AB is
equal
in
length
to that from 0
on AB^
and the two
perpendiculars
make
equalangles
with the axis of x on
opposite
sides of
it;
therefore
ah
p
'
G2
84
MB.
BRILL,
A NEW METHOD FOR THE GRAPHICAL
Multiplying together
these two
equations
we
obtain
1
+
1
=
1
which
expresses
a
well-known
properly
of the
right-angled
triangle.
7. We will
now
proceed
to
develope
some
metrical
properties
connected
with the
theory
of the
mean
line.
JLet
w
be the
mean
of the lines
w^
and
w^
for
multiples
971 and
fij
then
we
have
{m + n)w^
mw^
+
nw^.
This is
equivalent
to the two metrical relations
,
.cosS
cos^, cos 9,
(m
+ n)
=
"i -+n
=
P Pi Pf
"
,
V
sin^ sin0.
sin 6^
and (m
+ n)
= m
*
+ n
'
.
P Pi Pf
From these
we
easily
deduce the
two
following
:
p Pi p, PiPf
and
^^^^^
=
-
cos(5
-
^,)+
-
cos(^
- 0X
P Pi
'
Pf
*
Let
AXj ABy
^40 be the lines
respectively
denoted
by
to, w^j w^.
From 0 draw
OL, OM^
01s
respectively
perpen- dicular
to
AX, ABj
AG.
Then,
if 0 lie within the
angle
BAG
or
the
vertically opposite angle,
we
have
Im
+ ")"
Iff?
^
IP?
^
mn
t% a n
but}
if 0 liewithout both these
angles,
we
have
(m +
n)'_
m*
^
r?
^^
mn
^.^
or
^'OW'^ON^'^
OMTON^^
Through
L draw LH and LK
respectively perpendicular
to OM and ON. The second of the above relationsbecomes
OH OK
BEPBESENTAllON OF COMPLEX
QUANTITIES.
85
Both these
properties
admit of
generalization,
but
we
will
only
trouble ourselves with the second. Let
(wij
+
7W,
+"""+
mj
w =
m^w^
+
mju)^
+...+
mjjo^.
This is
equivalent
to
the two metrical
relations
,
, . .
.cos^
cos5, cosS,
.
costf.
P
'
Pi
^
Pt Pn
and
,
. , ,
.
sin^ sin0. sin
S,
sin
0^
P
'
Pi
"ft "A
From these
we
deduce
"
i
"
s s
=
"
5^co8f5-5J+-3cos(^-5,)+...+-*cos(5-tfJ.
i" Pi ft
'
ft
Let
OLj OM^j OM^,
...,
Oif^
be the
perpendiculars
from
0 on
Wy to^yw^, ...,
w^.
Prom Z draw
iJ^^,
LN^^...,
i-ft^
respectively perpendicular
to
OJf,,Oif"
...,
Oif^.
The
above relationbecomes
m.
OK
ON,
OK
...
''0MJ'"'0MJ"'^"'-0M^
or as
it
may
be written
M,K M,N,
. .
M^N^
^
8. The relation
{m-\-n)w^mw^-\-nv)^
may
be written
in the form
e"^ e^i "*^"
(m + n)
"
="wi "
+ n
"
.
\P ft- ft
Hence,
if
r
be
a
positive integeri
^
P Pi Pi ft
2
!
ft ft ft
86 MR.
BRILL,
A NEW METHOD FOR THE GRAPHICAL
If
we
split
thisinto its real and
imaginaryparts,
multiply
the first
by
cosr^ and the second
by
sinr^ and
add,
we
have
^"'+^=^;co8r(^.-5)+r^'^cos{(r-l)(5.-tf)
+ ".-^}
V K
' ' '
Pi P"
a
+...+
"
cosr (5,-^).
p;
It would be troublesome to enumerate
all the
cases
of this
that
might arise,so we
will content ourselves
with
one
particular
case.
Let 0 liewithout both the
angle
BA O and
the
vertically opposite angle,
and let
m
and
n
be both of the
same
sign
so
that AX liesbetween AB and A C.
The above
property
becomes
(m + n)'
w'
(
-DA V)
H-
. . .+
-^^r
cos
{r.
(7-4Z}.
This is the
generalized analogue
of
a
property
of the
trianglepublishedby me
in the Educational Times for
October 1885
(No.8290).
9. Let
"?,
and
w^
denote the lines AB and
-4(7,
the
origin
0
lying
withm the
angle
BAG. Draw AX the
harmonic
polar
of 0 with the
respect
to the lines AB snidAG,
On the
opposite
side of 0 from
AX^
and at a
distance from 0
equal
to one-half of that of AX from
0,
draw
a
line BG.
Then,
if this line be denoted
by
w^,
we
have
w^
+
w^
+
w^
= 0.
This is
equivalent
to
the
two metrical relations
lind
cos 5, cos^, cosS,
^
i-+ 2.+ 5=0
A P. Pz
8in^^sln^_^ slnj93^^
P^ P. P.
'
Fi F, Pz
88
MR.
BRILL,
A NEW METHOD FOR THE GRAPHICAL
Then the
expression s^jw^
"
s^/w^
is
equal
to
fC08(^,- 0,)
-^C08,(^.- dj
+
.-If' 8in(^,- e,)
-^
8in(^,-d,)|
,
Stf Sfa VSfi Sfa J
and thereforeitsmodulus is
Hence,we
have the
following
theorem
:
Let ABC and DEF
be two
triangles having
a common
centroid
0,
and let
OL, OM, ON, OP, OQ,
OR be the
respective perpendiculars
from 0
on
BC,
CA, AB, EF,
FD
DE'y
then
we
have
fON\' (0L\^ ^ON.OL
,o ",
fqL\\/OM^.* ^OL.OM
111
Hopj'^vooj'^op.oe'''''^^~^^--oP'oE^^'o^'
11. Let ABC be
a
triangle.
Take a
point
0 within the
triangle, joinOA, OB, 00,
and draw
OB, OE,
Oi^
respec-
TIT
tively perpendicular
to
BG, GA,
AB.
Through
0 draw OX
parallel
to AB to meet GA in
X,
and
through
X draw JfiV
parallel
to OA.
Through
0 draw OY
parallel
to J5(7 to
meet
AB in
F,
and
through
Y draw
-A/i
parallel
to OB.
^Through
0 draw OZ
parallel
to GA to meet BG in
-2',
and
SEPRESENTA.TION OF COMPLEX
QUANTITIES.
89
through
^draw iilf
parallel
to OG. Draw
OP,
OQ,
OR
respectively perpendicular
to
MN^ NL,
LM,
If the lines
BG^ GAy
AB be denoted
by
ti?,,i/?,,w^
then
MNy NLy
LM will be denoted
by
w^
"
Wj, m?,
"
w^^
w^
"
tr,.
We
shall, therefore,
be able to deduce
a
theorem from
our
figure by
means
of the
identity
If
!",,ft. fti ?n ?" ?,
Btand for
OD, OE, OF, OP, OQ3
OR
J
and
0^,^" dg,^" 0
,
^3
be the
angles
made
by
these
lines with the axis of
x
;
then,proceeding
as
in the
early
part
of Art.
9,
we
shall arrive at the result
Pdi
sin
(^.
-
^a
+
*,
-
*,)
^p,q,
sin
[0^
-
^,
+ ^3
-
"f",)
i.e.
OD.OPsm{BAG-^MLN)=OKOQsm{GBA
+ NML)
= OF.OR
sm{AGB
+
LNM).
12. The
same
figure
also furnishes
us
with
interpretations
of the formulas
and
ll",W?3
{W^
-
W,)
+
W?3"7j
(?/?3
-
tl?J +
W,W?,
{W^
-
"",)
+
("^,
-
^s)K
-
"",) (^i
-
W'a)
= 0.
The firstof these is
equivalent
to the two metrical relations
cos(2g,
+ 0,)
^
co8(2g,+0J cos(2g,-h0,)
Pill
Viit Pliz
and
sin(2g,
+
"^,)
^
Bin(2g,-f "^,)
sin
(2^3
+
^3)
ph^ Ph^ Pilz
,
8in("^,4"^,4-"^.)^^
from which
we
deduce
cos(20,'(t",-4",)
^
008(2^3 -"^3-0J
^
cos(2g,-"",-"^3)
P'ii P/?.
ft'?.
1
+
= 0.
SiMm
90 MB"
BBILL|
A NEW METHOD FOR THE GRAPHICAL
Let
MNj NL,
LM meet
AB^ BCj
GA
respectively
in
X\ Y\
Z\ Then this relation becomes
cos(Zr^-ZZr') cos(Jf^^X~J/ZZ^
Olf.OP
"*"
OE\OQ
cos(NX'Y^NYX')
_
1
OF'.OB
"
OROQ.OR^
Similarly
we
may
deduce from the second formula the
relation
co^{AY"Y-^AZ"Z)
co^(BZ"Z^ BX"X)
OE.OF.OP
"*"
OF.OD.OQ
co"(GX"X^CY"Y)^
OD.OE.OB
""
OROQ.OB^
where
X". Y'\ Z"
are
the
respective
intersections of
MN^
NL,
LM with
BG, GA,
AB.
13. We could obtain
analogues
of all the
properties
of
rectilinear
figures
obtained in
my paper
on Argand's
method.*
We
will, however,only
consider
one more instance^
viz.
the
interpretation
of the formula
Let ABG be
a
triangle
and 0
a
point
within it. Let
MN be the harmonic
polar
of 0 with
respect
to
GA and
AB,
NL that of 0 with
respect
to
AB and
BG^
and LM that of O
with
respect
to BG and GA. Draw
CD, OE, OF, OP^ OQ^
OR
respectively perpendicular
to
BG^ GA^ ABj MN^ NL^
LM.
Then
we
have
cos(LBG'-LGB}
cob(MGA^MAG) cos(NAB-NBAy
OD\OP
"*"
OE\OQ
"*"
OF\OB
cosiMGA'-NBA) cos(NAB-LGB)
coB(LBG-MAGy
*"
OE.OF.OP
"*"
OF.OD.OQ
"*"
OD.OE.OB
*
14. Let 0 be the
origin
and A
a
fixed
point
distant2a
from 0. On OA
as
diameter describe
a
circle,
and
through
A draw
a
line AB
cutting
the circlein L. Draw
a
tangent
LM to the
circle, join OL^
and draw OM
perpendicular
"
MessengerofAfathenuUicSf
rol.
xvi., pp.
8-20,
EEPBESENTATION OP COMPLEX
QUANTITIES.
91
to LM. Then
we
have OM, OA =
0L%
and the
angle
MO
A
is
bisected
by
OL.
Thus,
if
w
represent
the line
AB^
then
LM is
represented by
2at/;*.
Through
A draw another line AB*
meeting
the
circle
in
L\
and
join
OL'. Draw LM' the
tangent
at
L*jmeeting
LM in
T,
and draw OM'
perpendicular
to L'M '. Join LL*
and draw ON
perpendicular
to it. Join M'N and NM^
Then
we see
that the
angle
M'NO^M'L'O^LAO^L'LO^NMO.
Also the
angle
M'ON^ M'L'N= TL'L
= TLL =t MON.
Thus the
triangles
M' ON and MON
are similar,
and therefore
OM. OM' = ON". Now
we
have
OA.OM=Or and O^.OJf'=
Oi'*;
consequently
OL\OL"= OA\OM.OM'=
OA\ON\
and therefore Oi
.
OU = 0^
.
OiV.
Further,
2.AON=^AOM+AOM'^2(^AOL + AOL)*^
that
is,
^ OJV^= ^ Oi
+
^
OL'.
Hence
we see
that liAB be
represented by w
and AB'
by i5"/,
then LL' is
represented by
2ai^tt?'.
Thus
we see
that the circle
can
be utilizedto obtain
interpretations
of formulae
containing
squares
and
products.
The method of
interpretation
may
however be made
more
general
than that
given
above. We have taken
Ay
the
point
diametrically opposite
the
origin,
as our
fixed
point,
and the
diameter OA as our
initial
direction.
Suppose
that
we now
take L for
our
fixed
point,
and OM the
perpendicular
from 0
92 MR.
BRILL,
A NEW METHOD FOR THE GRAPHICAL
on
the
tangent
at L for
our
initial direction. We have
ON'=:OM.OM',
and the
angle
MOM'^^.MON.
Thus,
if
(?Jf=2c,
and LL' be
represented by
w^^
then LM' is
represented by
2cw\
Draw another hne LL"
throughi,
and draw ON'
per- pendicular
to
LL". Let L"M" be the
tangent
at
L\
and
OM" the
perpendicular
on
it from the
origin.
Join
LL"^
and draw UK
perpendicular
to it. Then
we
have
ON'^OM.OM' and
ON'^^OM.OM",
and therefore
0N\ ON" = 0M\ OM'. OM" = 0M\ OK'
;
that
is,
ON. ON' = OM. OK.
Further,
0^ bisectsthe
angleM'OM",
and therefore
2.MOK=:MOM'
+
MOM" =
2(MON-i-MON')\
that
is,
MOK^ MON+ MON'.
Thus,
if LL be
represented by
w^,
and LL"
by
w"
then
iX" is
represented by
2cw^w^.
15. As
a
first
example
of the method of
interpretation
developed
in the last article
we
will take the formula for the
product
of two
polynomials.
This will be
easily
seen
to
yield
the
following
theorem
:
0 and L
are two fixed
pointson
the circumference of
a
circle. A seriesof
m
fixed
points
is taken
on
the circum- ference
of the
same
circle,
and
a
second series
containing
n
points
is also taken. Let LM be the
mean
line for
equal
multiples
with
respect
to 0 of the lines
joining
L
to
the
m
points,
and let LN be the
mean
line for
equalmultiples
with
respect
to 0 of the lines
joining
L to 'the
n
points.
Let LM and LN
meet the circle in M and N
respectively
;
then MN is the
mean
line for
equalmultiples
with
respect
to 0 of the
mn
lines that
can
be obtained
byjoining
any
one
of the
m
points
to
any
one
of the n
points.
16. We will next take the formula for the
square
of
a
polynomial,
viz.
As
before,
let 0 be the
origin
and L another fixed
point
on
the circumference of the circle. Take
n
other fixed
points
on
the
circle,
and let the lines
joining
L
to these
points
be
REPRESENTATION OP COMPLEX QUANTITIES.
93
denoted
by
w^^ w?"
...,
w^.
Then,
if W be the
mean
line for
equalmultiples
with
respect
to 0 of these
lines,
we
have
nlF=w^
+
!"?,+...+"?".
Let PX be the mean
line for
equal
multiples
with
respect
to 0 of the
tangents
at the
n
points,
and
let
QY
he that of the
^n(n"l)
lines that
can
be
obtained
by joining
the
n
points
two and two.
Then,
if
FX and
Q
F be denoted
by W^
and
TT,,
we
have
nW^
= 2w' and
w
(w
-
1)
PF,
=
22z^m?',
and therefore
n
TF' =
TT^, + (n
- 1
) W,.
Thus,
if W cut the circlein
M^
the
tangent
at M is the
mean
line with
respect
to 0 of FX and
QY
for the
respective
multiples 1,
w
" 1.
Further,
it is evident that if
we
make L
moveable,
Jlfwill
remain
fixed,
and the
envelope
of LM will be the fixed
point
jJf. Thus
we are
led
to the
following
theorem
:
Let 0 be
a
fixed and L
a
moveable
point
on the circum- ference
of
a
circle,
and let
n
fixed
points
be taken
on
the
circle. Let PX be the
mean
line for
equal
multiples
with
respect
to 0 of the
tangents
at the
n
points,
and
QY
that
of the secants that
can
be obtained
by joining
the
n points.
Then the
mean
line for
equalmultiples
with
respect
to
0
of the lines
joining
L to the
n
points
passes
through
a
fixed
point
M on
the circumference of the
circle,
and the
tangent
at M is the mean
line with
respect
to 0 oi PX and
QY
for
the
respective multiples 1,
n" 1.
17. As
a
final
example
we
will
give
the
interpretation
of
the formula
Let
0,i, Aj -B,Gyi",-E,
F be
eight
fixed
points
on
the
circumference
of
a
circle. Let iP be the harmonic
polar
of 0
with
respect
to LB and
LG^
LQ
that of 0 with
respect
to LG and
LA^
LB that with
respect
to
LA and
i5,
LX
that with
respect
to LE and
LF,
LY that with
respect
to
LFdJiA
LD^
and LZ that with
respect
to
LD and
LE^
the
six
points
P, Q^ -B,JC,F,
^
lyingon
the circumference of
the circle.
Then the
mean
line for
equalmultiples
with
respect
to 0 of the lines
AX^ BY^
CZ coincides with that
oftheUnesZ"P,^Q,i^i?.
(
94
)
NOTE ON THE TWO RELATIONS CONNECTING
THE DISTANCES OF FOUR POINTS ON A
CIRCLE.
By
Prof.
Cayley.
Consider
a
qaadrilateral
BACD inscribed in
a
circle;
and let the sides
BA^ AG^ GD^
DB and
diagonals
BG and
AD be
=
c, J,A,
"/,a,
"
/ respectively ;
/
is for convenience
taken
negative,
so
that the
equationconnecting
the sides and
diagonals
may
be
A, =a/+J^
+ cA,
=0.
We have between the sides and
diagonals
another relation
F,
=aJc
+ a^A+
M/+c/i/, =0,
as
is
easilyprovedgeometrically;
in
fact,recollecting
that
the
oppositeanglesare
supplementary
to each
other,
the
doable
area
of the
qnadrilateiul
is
=
(ic+ gK)edn^,
and it is
also
=
(bh+ cg)
sin^;
that
is,we
have
(he
+gd)
sinA "
(bh
+
eg)
sinjS= 0.
But from the
triangles
BAD and
BAG^
in which the
anglesD^
G
are equal
to each
other,we
have
c f
c a
sini" sinj?' sin "7 sin^
'
that is
/sin^
+
asin-B=0;
and thence the
required
relation
a
(Jc
+ :7*)+/(** + "y)
= ^"
The distances of the four
points
on
the circle
are thus
connected
by
the two
equations
A =
0,
F=0.
Considering
a,
i,
c,
f^g^h
as
the distances from each other of
any
four
points
in the
plane,
we
have between them the relation
12,
=
ay(-a'-/'
+
6*
+ i7' +
c'
+ A")
+ jy( a'+Z'^J'-Z +
c'
+ AO
-aVc'-ayA'-J'Ay*-cy^", =0;
and it is clear that this
equation
should be
a
consequence
of
the
equations
A =
0,
F=
0. To
verify this, forming
the
sum
fl+ F*, we
have
a+F'=
(a*+/")(-ay'+
6y
+
c'A'
+ 2J(7cA)-
+ (*'+i7*)(- *y +
o'A*
+
ay+ 2cAa/)
+
(c*
+
A')(-c'A'+ay"+jy+ 2a/ft"7);
96 PROP.
CAYLEY,
THE ANHARMONIC RATIO
EQUATION.
or,
multiplying out,
the
equation,
as
is
well-known,
takes
the form
But to effectthe
multiplication
in the easiestmanner we
may
proceed
as follows:
writing
a,i,c=(a-S)(^-7), (/3-S)(7-a), iy-8)(a-fi),
80
that
a +
i
+ c
=
0,
the
equation
is
The
product
of the first
pair
of factors
is
thus the
equation
Is
that
is,
and
recollecting
that
a+
"+ c =
0,
and
writing q^hc-k-ca-^-abj
r^ahc^
the
equation
becomes
(a:+l)*-3(aj+l/a;+f3
+
^)(aj
+ l)'aj"-l
=
0;
that is
(^"+ aj +
l)'+^(aj
+ l)'aj*
= 0.
But,writing
tf= -
,
we
have
(fl*+tf+l)"+^(^+l)'^=0;
or
finally
(^"+ a, +
ly
-
"^,|/;^^^?'
a?(x^.
1)'=0,
the
required
result.
CONTENTS.
PAGB
A new
method for the
graphical representation
of
complex quantities
(continued).
By
J. Brill
81
Note on the two relations
connecting
the distances of four
points
on a
dide. By
Prof. Catlky - 94
Note
on
the anharmonic ratio
equation. By
Prof. Cayley - - - 95
The
following
papers
have been received
:
Major
Allan
Cunningham,
"On the
depression
of differential
equations."
Mr. B.
Tucker,
"
The cosine orthocentres and a
cubic
through
them.*'
Mr. Q-. Q-.
Morrice,
"
Note
on
the multiplication
of nonions."
Mr. C.
Chree,
**
Vortices in
a compressible
fluid."
Articles for insertion will be received
by
the
Editor, or by
Messrs.
Metcalfe and
Son, PrintingOffice,TrinityStreet,
Cambridge.
WA
No.
CXCIX.]
4"
NEW SERIES.
-rff^
[November,
1887.
"^^
MESSENGER OF
MATHEMATICS.
EDITED BY
J. W. L. QLA.ISUELI, So.D.,F.K.8.,
FELLOW OF TRINITY
COLLEGE,
CAMBRIDQE.
VOL. XVII." NO. 7.
MAOMILLAN
AND 00,
HonBon
aui
ffiambrfifge.
1887.
Price" One
Shilling.
MWrnAT.PR AND SON. OAMBRIDOB.
_Eig']^trvv
"^
u^"
-^6^
NOV 15 1887
^J"E "COStNE^* ORTHOCENTRES OF A TRIANGLE
AND A CUBIC THROUGH THEM.
By
R,
Tttcksr,
M,A.
ABG
is a
triangle,
of which
AL^ BMj
CNsre the altitudes
co-intersecting
in the orthocentre H.
LFj
LE'
are
drawn
parallel
io
AG^ AB]
MD,
MF
AB, BG',
tod
NE.NU BG, GA]
the lines
AD^ JBlS^
GF cointersect
in
d-^,
and the lines
AD\ BE\
GF' cointersect in
"r,.
The
equations (In
trilinear
coordinates)
to
AD, BE^
GF
are
respectively
J)3cos-4
=
a7Cos(7, c7CosB= Jacos^,
aacos(7=c)8cos5...(i)j
and to
AD\ BE\
GF'
are
afi
cosB=
cy
cos A
,
by
cos (7= aa
cosJBj col cos
A =
5^ cos
G.
, " (ii),
hence
"^^c^
are
givenby
the
equations
a )8
7
4A
c cos
jB
a cos O b
cos
A a*
+
i*
+
c'
a 0
y
^
r
7y
=
T
= "
"r,
" tanto
0 cos u c COS
A
a cosx"
1
= tan
CD
(iii).
From
(lii)
itis
at once seen
that the Lemoine
point
(It)is
the
mid-point
of the
join
of
o-^^c^
*'(iv)"
If
now
Z"Zj,5
771
,
m^] w"
n are
the
projections
of
o-j,
a
On
the
altitudes,
tnen
Bl^,Um^^ An^
cointersectin
12,
ana
0Z"Am,,Bn^
in a'
(v).
Again Gl^jAm^^ Bn^\Bl^^Gm,^An,
cointersect
respec^
lively
in
Sj
(a
cot
5,
b cot
C,
c cot
-4)
and
6,
(a
cot
C,Jcot-4, ccot-B) *
(vi).
The
mid-point (rf)
of the
join
of
e^,e,
is
(vil),
a': J'
and lies
on
the cifcum-Brocardaxis
;
thisaxisthereforebisects
cr,"r"
12i2'and
SjE,.
VOL, XVII. H
98
MR.
TUCKER,
THE
"
COSINE
"
ORTHOCENTRE"
OF A
We
readily
obtain the
following
results
:
AE =
V
cos
Ajc^
CE =ab cos
jB/c,
AE' = be
cos^/a,
CE' = b'
cos
0/a,
CD = a*
cos
(7/6,
^Z" = ac
cosAjb,
CD' = ab
cosAjCj
BD' =
a'
cos
J5/o,
AF
=
bc
cos
(7/ a,
J5F =
c'
cos
5/a,
AF'^e
cosA/bj BF'^accosCjb.
..(viii).
If
"7,Z,, a,/^produced
meet
the side AC in
Aj,A,',
and ^-B
in
A/,Ag,
then if K be
put
for a*
+
"**
4
c",we
obtain
cr,
Aj
=
2a*6
cos
(7/ JST,cr^A,
=
2a*c
cos
"/Z,
whence
"7,A,
+
o-,A,
=
2a7ir,=(by
I.
x.)*
twice the
intercept
made
on jBC
by
the T. R.
circle.f
Also
cTjAj'
= 2abc cos^
/
JSr=
o-,Ag',
and therefore
a^h^'aji^'
is
a
parallelogram, A/A^'
being
bisected
m
the Lemoine
point
with like
results for the other sides.
..(ix).
^
.
Ah/
"rX'
Agam
^
=
BD' AG CD'
hence
Ah^.
AB =
Ah^.
A
(7,
f.e.
A/A^'
is
an
anti-parallel
to BC
(x).
Since
Ah;^WclK, Ah^^2bc'IK,
therefore
A;A;
=
2abclK(^
2DF =
2ED' = 2FE of
I)...(xi).
Similarly
for the other sides
0 jV,hlk^corresponding
to
A/A,'),
therefore
i.e,these lines
are
diameters of the cosine-circle
...... (xii);
the
equation
of which is
(/*
= 1
+ cos
-4
cos
j?
cos
(7),
=
2(aa-\-...+...)[abc"inBsin
(7cos^
+.,,+. ..]...(xiii).
"
I cite
"
The Triplicate-ratio circle," Quar, Jour.,
yol.
xix,,
No.
76,as I.
t
We
readily
obtain
-~^
=-^
"
*c.
100 MR.
TUCKER,
THE
"
COSINE
"
ORTHOCENTRES OF A
I collect
together
here
a
few resultsof interest.
The
equation
to
a^a^^
is,
if /
= a*
-f
6*
+ c*,
aa(a'K^v')
+
bff( )+
cy( )
= 0*..
.(xix),
to
a^O
is
which
passes
through
5"/a,cVJ, fl'/c,(fi,), )....(xx),
and
l/(c
+
a)a, l/(a
+
J)6, l/(J
+ c)c...
to
a-^O
is
which
passes
through
c'/a,a75, 57c, (m,), *..(xxi).
(a"+J')/a,(b'-hc')lb, ic'
+
a')lc,
and
l/(a
+
J)a, l/("
+
c)i, l/(c
+
a)c,
The
join
of
G^Z,
aa
(J*
-
c')+...+...
=
0, evidently passes
through
the middle
point
of the
join
of
^t" /i,
(x^^Oj
and the
equation
to
fi^f^^
is
aa(a*-JV)+...4...=0 (xxiii).
Since
AFLE', "c., are
parallelograms,
therefore
AL,
FE\ "c.,mutually
bisect each other.
Since
Cj^
= a cot
C
tano), Bj^
= c
cosec5 tan
w,
we
get
the
perpendicular
from "1
on
-4Z
=
tanft)(ccosecJBsina)),
i.e.
X2 is
Brocard-point
of
one
cosine
triangle,
and SI' of the
other
triangle
(xxiv).
The lines
"tjY,aj^
intersectin
Jccos-4,
a"
cos
(7,a'cosjB,(TTj),
which is
evidently
on
AL
(xxv).
We note the
following points
in the
figure,
which will be
of
use
in the
sequel.
[AO^
jBcTj,
on meet
in
7rj(ccosJB,
ccos^,icos^)
'
TT,
(ccosjB,
a cos
(7,
a cos
jB)
TTg
(bcos
(7,
a cos (7,
J
cos
-4)
7r/(6
cos (7,
c
cos^,
b
cos^)
7r,'(c
cos
5,
c cos
-4,a cosB)
7r^(b
cos
(7,
a cos
(7,
a cos
5)
^
(xxvi),
"
It is
readilyprovedby rotating
the
figures
that OM is
perpendicular
to
"r,"r,.
TRIANGLE AND A CUBIC THROUGH THEM.
101
. (xxviii).
IB"r^j
C"T^
in
Tj(tan-4/a,
tan
(7/ J,tan5/c)
C"7j, -4(7,
in
T^(tSLnCla^ tan^B/J, tan-4/c)
-4"7"Aj^
in
T,(tanjB/a,
tan
-4/6,
tan
C/c)
ICcTj,
jBcTj
in
T'j(cot-4/co8^, cotC/cosjB,
cot
5/
cos
C)
4"rj,C"r,
in
T\(cotGj
cos
A
j
cot
B/
cos 5,
cot
-4/
cos C)
Ba-^y Aa^
in
T',(cot jB/cos-4,
cot
-4/
cos JB,
cot
Cj
cos
C)
("vii),
I-4"r"
5(?,
in
v^^cotAja,
cot
C7/6,
cot A
jc)
Ba^j COy
in
i;,(cotjB/a, cotjB/6, cotAjc)
Ca^^
AGj
in
v^(cot Bja^ cotC/6,
cot
Cjc)
iAa^j
CGj
in
i;/(cot-4/a,
cot
-4/6,cotJB/c)
jB"r" ^G^,
in
i;,'(cot O/a,
cot
5/ J,
cot
Bjc)
Ca^,BG,
in
i;;(cot 0/a,cot^/J,
cot
C/c)
If
-4,5j(7
is the firstBrocard
triangle,
then
-4^^,jBB,,
CC,
bisect i^i^',
2^/"',
i"J5'
respectively (and
intersect
as
is
well-known on
Kiepert's hyperbola) (xxix).
The
pole
of
o-^o-,
with
regard
to the circum-circle is
a'["c-a*cos(jB"C)],
..., ...,
'
and therefore lies
on
the well-known line
.... (xxx),
5ca
+ ca^ + a Jy
= 0
and of
e^6"
lies
on
a
(6"
+
c*)/a
+...+...
=0,
i.e.
on
the radical axis of the circum- and "T.R"
circles
(xxxi).
Assume
then
whence
and
therefore
and CO
= tan ^
tan
" tan
sin
(B
"
^i)
_
i cos-4
sin"^,
""acosJS*
cot^j
= tanjB+ cotjB-
cot
G,
cot
^1
= tan (7
+
cot C7- cot B
;
cot
(f"^
+
cot
^j'
= tanB
+
tan
(7, 3
cot
9^
+
cot
9j
= taniJ
+ tan
O,
^
t
^1
+
cot^,
+
cot^3=
tan-4+tanjB+tan
C
"
.
A tanBtan 6'= cot
(j"\
+
cot
(j"\
+ cot
^'
J
.(xxxii).
102
MR.
TUCKER,
THE
"
COSINE
"
ORTHOCENTRES
OF A
If
c7,5C=Vr^,cr^C5=V'\,
then
8in(g^'ifr^)
^
acoag
sill
^j
i
cos
C
'
whence cot
-^j
=
2 cot
5
+
cot'JBtan
(7,
]
and cot
^/
=
2 cot
(7
+
cot'C tan
5, [....
(xxxiii).
therefore 2 cot
^ .
cot
^'
= 4 cot'o)- 3.
'
From
(viii)
we
see
that AE,AM=:b^
cos'A
==JN\
thereforecircle
round -ft/j^ilf
touches AB at
N]
similar results
hold
good
for the other
sides (xxxiv).
The
equations
to the circles
LE'M^
LFN are
respectively
a?'
+
y*
=
cy
cos
Cj
sin
vl,
and a:'
+ y*=5iyC08i5/3in4...
(xxxv),
where
CL,
LA
are
the axes
;
whence,
if
D,
U are
their
diameters,
we
have D
+
D' =
B
cos
(B
-
C).
The
trilinear
equations
to the circlesin
(xxxiv)
are
FLNj aC=Jj[b
cos
Ccosu4a
+ o co8"5/8+
6
cos'CV]]
DML^ hC=L\c
cosMa -f c cos
-4
cos B^ + a
cos*
6V][
ENM^ cC=L[b
cosMa
+ a cos'Bff+ a
couB cos
Cy]
and
E'MLy
a
(7= Z/
[c
cos
^ cos
-4a + c cos'^/3+
6 cos'*
C7]
(xxxvi),
J?"-^i/,
J (7=
i
[c
cosMa + a cos
C cos ^/3 + a
cos*
C7J,
D'LN^ cC=^L\h
cos^Aa
+ a cos*jBi8 -f
i
cos 4 cos C7],
if
(7=
0^87+ ^qt
4- ca/8
and i
= aa + JyS+
C7,
From
(xxxv)
we see
that if
/",,
p^,
p, ;
p'^, p'"p\
are
the
two sets
of
radii,
then
P \PtPz
=
^*^ ^^*
-^ ^^*
^ ^^*
"7/
8 "=
PiP\p\^ \
ap^
+
1/)^
+
c/"jj
= 2jB*8in4 sin^ sin
C=^ap\
4
Jp't
+
^pVJ
,.., , ,
(xxxvii).
Some
generalproperties
of the cubics
given by
the
equation
C;E(a"+
J/9+ C7)
(-+
A
+ -)
^k'
''
\a%
bti cy)
we given
in
the
Messengerof
Mathematics
(vol,Jl.,
1864,
TKIANdLE AND A CUBIC THROUGH THEM,
103
p.
116)
and in the
Reprintfrom
the Educational
Times^
(vol.
v., p.
38)
;
we
propose
here
to consider the cubic
Cj
= cot
o)
tan
A tan B tan C==^k\..
(xxxviii),
which
passes
through
the
points
of
reference, through
0-^,cr,,
and
through
the orthocentre.
The
equationcan
be
put
into the form
(aa + J/8+ C7)(hc^y+
ca'^a.
+ aJa^S)
=
A;'aJca^7...(xxxix),
from whence it is
seen
that it touches the minimum circum-
ellipse i87/a
+
7"/c
+
a)8/c=0
at
the
points
of reference.
The centroid of the
triangle
is the centre of this
ellipse,
and
HA^ HB^ HC^
which
are normals
to
the
ellipse,
are
also
normals to
the
cubic,and,
from above
statement,
are
drawn
from
a pointon
the cubic.
The six
pointsH^
t^, t" t,
(xxvii),
and
T,
tan^/a,
tan
0/6, tt^AJc (xl),*
T
,
tan
CJa
J
tan -4
/J,
tan
jB/
c,
are
manifestly
on
(7"
and
they
also lie
on
the
ellipse
abal3+ bcl3y+
cay
a = cot
" cot -4 cot 5 cot 0
=
(aa+J/3
+ (yy)" (xli),
wliich
IS
concentric and
similarly placed
with the above
minimum
ellipse.
The cubic also
passes
through
the six
points
^^, cot^/a, cotjB/J,cotC/c,
'
j8,,
cot
-4
/a, cotC/6, cotJB/c,
/Sj,cot/?/
a,
cot A
Iby cot(7/o,
;9^,
cot
5/a,
cotC7/6,cot-4/c,
ySg,cotO/a,
cot
-4/6, cotjB/c,
)8g,cotO/a, cotJ5/6,cot^/c.
^
which lie also
on
the
ellipse
a6ai8+ 645^87 +
^7a
= tan*"
(aa+ i/8+ "7)*
concentric and
similarly placed
with the minimum
ellipse*
March, 1887"
.(xUi),
*
These
pc^ntsare easily oonstrncted in
a difEerent
way
from that indicated
in
(xxTii): for
o-,
(a,),c- (oj),t (o),
t'
(a')we have the relation between coor- dinates
to be
a^aui
"= a*
{aa^
=
^^Pi
=
'
(
104
)
NOTE ON THE
MULTIPLICATION OF
NONIONS,
By
G. O.
Morrie^.
The
object
of the
present
note
is
to
present
the multi-^
plication
table of nonions in its
proper
form. We have
m
and
n
for
our two fundamental
ternary
matrices with the
condition
nm =
pmny
where
p
is
a
primitive
cube root of
unity,
wi':=n'=l.
Now
in
any
?!uch
multiplication
for
a
group
of
operators,
and in
particular
for substitutions
(videDyck's Gruppen^
theorettscheStudien
//.),
it is
important
to consider the
per-,
mutations of the
elements of the initialrow
which leads to
any
one
of the
followingrows.
In the
present
case we
have
cle?^rly
a
non-primitive
group,
the substitutions which inter-,
change
the triads
(1,
wi,
?"*),(",
wn,
m*n)j (n*,iwn*, mV)
being cyclic,
and also the substitutionswhich
interchange
the letters
within the
cycles^
We
may
call the
substitutions^
1
,
5
,
5"
,
's%
8',
S*s,
8*s\
Let
us
form the matrix
J/,
1
,
m
,
w*
,
n
,
mn
,
m*^
,
w*, 7WW*,
m*n*
5
8
will effect
a
cyclic interchange
of
columns,
S of
rows*
MR.
CHREE,
VORTICES IN A COMPRESSIBLE FLUID.
105
Moreover,
the
simplest
form of
m is
m'=
(1, 0,
0
)
,
0
0.
Pt
0,
0,
P*
m
( h 0,
0, P%
0),
0
0, 0,
p
and
it is
easy
to
verify that,
for
example,
the elements of the
second
row are
those of the matrix formed
by multiplying
m'
by
8
(M)^
the third
row
w'*
by
s^
(M)^
and
so on.
June
21, 1887.
VORTICES IN A COMPRESSIBLE FLUID.
By
CharUB
Chree,M.A.y
Fellow of
King'sCollege, Cambridge.
The
following
paper
contains certain
applications
of the
equations
of vortex motion in two dimensions
to a
compressi- ble
fluid. The
equations
of
vortex motion in
an
infinite
fluid
are*
^^dP
dN_dM\
dx
dy
dz
^^^^dL^dN.
dy
dz dx
^
'*
dP
.
dM
dL
If
dz dx
dy
^
du dv dw
^
.(2),
.(3)
then
F^-^fjj^d^dy'dz'
gives
the
value
of
the function P at
a
point(a?,
y,
z)
at
a
distance
r
from the
point{x\y\ "'),
where 6 has the value
^,
the
integration extendingthrough
all
space.
If
-j
denote
partial
differentiation,
and
^
differentiation
following
the
fluid,
the
equation
of
continuity
is
dp dpu dpv dpw
_ ^
dt dx
dy
dz
"^
^
or
p
ot
"
gee
Lamb's Maion
of Fluids^ pp. 160,151,
for the
meaning
of
X, Jtf, iV,."c.
106
MR.
CHREE,
VORTICES IN A COMPRESSIBLE FLUID.
Thus
P-ijjjl J
I
"i^"^y'"i^ (^).
From the form of this
integral
and of the
equations
of
motion it follows that the direction and numerical
measure
of the
velocity,
at
any
pointdependent
on
the function
P,
are
the
same as
the direction and numerical
measure
of the
resultant force at that
point
in
an
imaginarygravitating
1 7\
system,
whose
density
at
(a/,y',"')equals
j
"
-^"
Thus
the
velocity
due to variation of
density
in the fluid
can
be
deduced from the
theory
of attractions.
Supposenow
all the vortices to be
straight
and
parallel
to the axis of
z^
then the
velocity,
so
far
as
depends
on
the
vortices,
is
independent
of
z^
i.e. is in.
two dimensions. The
functionsL
and M
vanish,
and
N*=-ljji'logrd^dt/'i
or,
if
a
denote the cross-sectionof the
elementary
vortex
filament whose
vorticity
is
^\
and 2 denote summation of
such
elements,
JV=--Slogr?V (5).
7r
The motion
depending
on
P will also be
independent
of
z^
provided
-
^
he
independent
of
z.
This is the
case
i"
w = 0
and
P='Pq/{z)9
where
p^
is the
density
at the
plane
"5=s0,
which
may
be
any
function of
x
and
^,
while
/(")
denotes
any
function of
z
which is
unity
when
" = 0. If
/j'
be the
density
over
the
elementary
cross-section
"/,
the
equation
of
continuity
becomes
a'p'
= constant.
and
we
simplyget,
in
place
of
(4),
^=2^
=
^i^|^
(")"
"
Lamb,
p. 168,Equation(33),
108 MR.
CHREE,
VORTICES IN A COMPRESSIBLE FLUID.
be treated
exactly
as
in
an
incompressible
fluid
by
the aid
of
images.
Each
image
is to have the
same
cross-section
at
every
instant
as
the real
vortex,
and the
consequent
velocity
due to variation of
density
will in
every
case
be
parallel
to the
boundaryplanes.
Consider
now a
singlestraight vortex,parallel
to
oz^
and
let
m
be its
strength
and
a
its
cross-section at
time t
;
then
the
components
of the
velocity
at
a
point
in the
surrounding
fluid,
whose coordinates
are x
and
y
relative to axes
through
the centre of the
vortex
fixed in
direction,
are
givenby
my
\ diT
X
Tir* 2ir dt r^
1 diT
y
mx
V =
5
+
"
,
-"
Trr
2ir at r
.(7).
For
our
special object
we
require
to determine the motion
of two such
vortices,
whose mutual distance is
supposedgreat
compared
to
the diameters of their cross-sections. Let their
strengths
be
m^^ m^,
and
at
time t lettheir cross-sectionsbe
"7j,(7,,
and the coordinates of their
centres
(a?^,
y J
and
(x,, y^)
;
then,
if r*
=
(a?,
-
x^)*
+
(y,
-
yj",
we
have
dt
irr' 27r dt r*
dt
TTT*
"^
27r dt r'
^
^ "
^Jyi-y")
.
1.^
^Sli5l
dt TIT* 27r dt r*
dt TTT* 2ir dt f^
[S).
From these
eqaations
we
get
Thus ifwhen
"s=0,r.
o?"
\^" ^"'
2ir rf"
'
"" ""i
=
."'i"
and
"r,
=
""r"
we
get
r"
=
c'+-("r,+
"r,-,"r,-,"r,). .(9).
These
equations
also
give
d
MR.
CHREE,
VORTICES IN A COMPRESSIBLE FLUID.
109
But if
8
denote the inclinationto ox
of the line
joining
the
centres of the
vortices,
y," y,
= r
sin
e
and
x^-x^
= r
cose,
and the above
equation
becomes
dt
"
Trr'
"
Thus,
if
e
= 0 when
" =
0,
we
find
,^^,r
" _ "
^
(10).
The
case
of
a
vortex before
an
infinite
plane
is included
in this solution. We have
only
to take
^c
as
the distance
of the
vortex from the
plane
at the time ^ =
0,
and make
m,
= "
"ij,a^=a-^^
and
^a'^'=^"r^.
In this
case s
remains
zero.
The
stability
of the circular form in
one
and in
two
straight
vortices has been
considered,
under the title of
Linked
VorticeSj by
Prof. J. J.
Thomson,
in his "Motion
of Vortex
Rings."*
The main
object
of this
paper
is to
extend his
treatment to a
compressible
fluid. To render
more
easy
comparison
with Prof. Thomson's
results,
I have
followed his notation and method so
far
as
possible.
Consider first
a
single, approximately circular,
vortex
whose section is the
same
for all values of
2;,
and let the
radius of
the cross
section
making
an
angle
6 with
a
fixed
direction be
given
at time t
by
jB
= a + a,jC0Sw^
+
/8^sinn^ (11),
the last two terms
being
the
types
of
any
number of
pairs
of
terms,
while
a^,
0^
are
supposed
small
compared
with
a.
Suppose
the
vorticity
and the
density
to be at
any
instant
the
same at
all
points
in the
cross
section. At external
points
we
have
approximately,
as
for
a
vortex of
truly
circular
cross
section,
the functions
'27r
dt
logr.
Assume for the fluidoutside the vortex
i^=(7-^logr
+
{^"cosn^
+
5,sinn")f^y
...(12),
*
Part in.,p.
71.
(13),
110
MR.
CHREE,
YORTICES IN A COMPRESSIBLE FLUID,
and for the flaidinsidethe rortex
^'==^'-
^^
+"co8n^
+
^"inn^)Q"
47ra'
at
^ " *
\aj
where
(7,2",(7',
iX
are
absolute
constants,
while
A^,B^,
"c.
denote,
as
do
a,,
^^^
functions of the time. The values of
N and
N'y
and of P and
P*,
must
agree
at
the surface of the
vortex,
1.
e.,
when
r = a +
a"
cos n^
+
)8^
sinw^. We thus
get,
neglecting products
of small
quantities^
We must also have the
velocities,
radial and
tangentiah
the
same
for
pointsjust
inside and
just
outside the vortex ;
therefore,
when
r
=
a +
a"
cosn^
+ /8^
sin
w^,
IdN
dP^l
dN' dF
r
dd dr
r
dO dr
and -"
1^=-^'
1^'
dr
r
dd dr
r
dd
'
Neglecting
terms such
as
-4^a^,
it is however
obvious,
if
A\
=
A^j"c.,
that at the surface considered
1
^_1 dN^ .
1
d^_l dP^
T
dd "r dd
r
d0 ^r de
'
thus
we are leftwith
dN^dN^ dP^dF
dr dr
^
dr dr
'
Equating
coefficientsof cosn^ and of
sinn^,
we
get
from
these
equations
"
Iran
" "
iran
"^
2iran dt
""^
*
2'iran dJt
'^*'
Thus outside
tjje
vortex
^= (7"
^
logr+
-^
(a
cosw^
+
B
sin
w^)
f-V
TT
"
Iran
^ * * '
\r)
i'-^+,i|H'-,-^|(..c",.""4A"."")(2)"
I
(")"
MR.
CHREE,
VORTICES IN A COMPRESSIBLE FLUID.
Ill
But from
(11)
"
n0(a^Binnff
"
4^^cosnff) (15),
where 0 isthe
angularvelocity
round the axis of the vortex
of the fluidat itssurface. Thus
^
1 dif I dP
m
^
0 =
=- +__--=
" -
4. terms
m a and
p.
r
dr
ir
dd
ira
" *
Thus in
(15)we
may
take 0 = "
|
.
ira
We have also
a
second value for the radial
velocity givenby
dB_ldN
dP
dt"
r
de"^
dr'
Noticing
that
57
==
-57 "
this last value of
-^
can
^
2ira dt dt
^
dt
be reduced to
i
(a"
sinnff-
pn
cosn^)
+
-j- ,
iTQ, at
which must be identicalwith
(15). Equating
the coefficients
of cosn^ and of
sinrad,
we
find
^-= -/3,("-l)^.
rf/3. .
,.
m
I
" ('^^"
_.=
".(n-l)-.
whence
a/+
^^'=
constant
(17).
Also,
eliminating 13^
from
(16),
we
get
,
d*a
.
da
rfa
.
" /
^ \s /^
112
MR.
CHREE,
VORTICES
IN A COMPRESSIBLE FLUID.
Multiplying
up
by
2
-^ ,
and
integrating,
we
find
a"(^y+wi'(n-l)'a"'=
constant
(18).
fin
If then
a^
and
-^-f
vanish
initially
theyalways
do
so,
the
same
is
obviously
true
of
)3^.
Supposeinitially
"^
=
0,
and
so
from
(16)^,
=
0,
while
a"=oa,,
and
so
-^/=o"h
^^.
i
where
a,
is the initialvalue
of
a
;
then
from
(18)
""
=
oa"cos
'"("-1)
-\'i
^'11
(1^).
similarly /S^
=
^^a^
sin
j
w
(w
"
1)
1 "\
Thus the form of the cross-section at time t
is
givenby
^ = a +
,a"
cos
[nd
^m{n^ I)
T -I
(20).
The section thus remains
approximately
circular,
and the
disturbance in
shape
travels round the cvlinder in the
grai-'
duallyvarying
times
givenby
m
(n
"
1)
I
"
=i2tV,
where
i
is
an
integer.
If
we
suppose
^
= constant =
^7,.
we
get
i? =
a4.a.co,|"^-^i^=illog(l+7")}.
The
period
of the ith revolution of the disturbance
is
givenby
Ti
=
-
\
e^ff-l)- 1
U
fn{n-l)
,
and
so
increases
with
t
If
7
be
positive,
i,e. if the cros^-
sectionof the vortex
be
increasing.
If
"r
= constant =
Tra'*,
and m^coTra^ and the time be
properlychosen,
the result
(20)
is identical with that of
rrof. Thomson on
his
page
74,
114 MB,
CHREE,
YORTICES IN A COMPRESSIBLE
FLUID.
Since we
suppose
afc
small we
require
to retain to our
present
degree
of
approximation
onlyKj,a"
13^^
and
0^.
Let iR denote the radial
velocity
of
a point
on
the second
vortex,
and 60 the
velocity perpendicular
to
the radius
vector,
both
being
taken relative to the centre
of that vortex.
Then,
from
(21),
-w"
(a'^
Binnff -
P\ cosnO')
(25),
where to
the
present
degree
of
approximation
^
Try-
But we
have
also,putting
r'
=
J after
diflferentiation,
r
du dr r
dff
ttc
^ '
^
dP.
1 d(T.
,-,
,
The terms in
m^
and
in -7^ are
introduced as
it is the
velocity
relative to
the centre
of the second vortex
that is
being
considered.
We thus
get,
the terms
in
n
being
of
course merelytypical,
" =
.|^.(a"inn^./S;cosn^')+^j|^
-^"8in2((?'-e)...
971. a
+
"
*
-^i+j
{(a^cosns
+
iS^sinwe)
sin(^'
"
b)
TT C
+ (/8^
COS we"
a^
sin
we)cos(5'" e)...|
- - " -,
-^
h cos2
{ff-e).
"*"
2"^
"^
?^
^^"*
^^^'^^ "*"
^*"^"^^^
^^ ^^'
"
^^
-
()8^
cos
ne-a"
sin
we)sin(^-
e)...}
(26).
MB.
CHREE,
YOKTICBS IN A GOKPBESSIBLE
FLUID. 115
These two values for It must
be identical. The terms
independent
of ff
are so
since
-j
=
27r6
-r*
"
and the
identity
extends to the coefficients
of
every
sine and cosine of
mul- tiples
of 0^
Equating
coefficients of
cos^,
we
get,
after
reductioni
"S"
"
S
^^^
^"2b-
a,
sin2e)+
g^.-J
(a,cos2e+/3,
sin2s)
;
but to our
present
degree
of
approximation
terms
of orders
"
^
or
-J
-7-
are
negligible,
so
-^
=
0; similarly
we
get
-^
= 0.
Thus
a'j, /8\
if
originally
zero
remain
so,
whether
the fluid be
compressible
or
not.
To the
same degree
of
approximation
we find,
from
equat- ing
the coefficientsof cos2"' and of sin20' in
(25j
and
(26),
rfa'
m,
^,
tn}"
.
^ .
1 " dfT. "
^ ,^^.
^_
!".
a',
+
^co82e+J^
^.
^'
8m2.=0
(28).
dt irb
'
TTC
27r c ai^
^ '
It is
scarcely
likely
that these
equations
admit of
a
complete
solution,
but
general
ideas
of the motion can
be
deduced.
Supposing
for an
instant
there
were no vorticity,
but
only
two
columns
of fluid of
varying
density,
we
should
have
nij
=
wij
=
8
=
0,
and
so
e?a',
_
\ h
d"T^
_
ah da
~dt 27rc''di
V dt
^
dt
Thus
/8',
would be
wholly
unaflected,
while
a'^
would alter
from its
original
value
^a
,
according
to
the law
""""""-/;?"'*
p')-
This shews that
a,
would increase or
decrease
according
as a were decreasing
or
increasing.
Thus,
if both columns
'Vere
diminishing
in
cross"*3ection,
and
so
approaching,
there
would be
a
decided
tendency
in both
cross^'sectionsto assume
12
116 MR.
CHREE,
VORTICES IN A COMPRESSIBLE FLUID.
an
elliptical
sort of
outline,
the
major
axes
coinciding
with
the line
Joining
their
centres.
It is
pretty obvious,taking
into consideration the existence of
vorticity,
that the vibrations
will become
more
important
if the vortices
are
approaching,
and will
not
take
place
about
a
truly
circular form.
If the
density
of the vortices
vary very
slowly
itis
com- paratively
easy
to trace
the effect
on
the vibrations.
Suppose
and let
m^^
=
^a'o"",
"w,
=
irb'^m.
It what follows
we
suppose
7*
and
yt
to remain small
during
the time considered.
Now,
when the fluid is incom-
;res8ible,
a,
and
/S',
experience,
as
is shewn
by
Prof,
'homson,*
two forms of
vibration,
the shorter
periodbeing
27r/"'
and the
longer7r/n,
where
n={aya\+a"'b\)l"^^ (30).
If then the
period
to which
our
equations
are
applicable
be
supposed
to be
at
least several times
greater
than the
longerperiod
of
vibration,
we must have
70*^
and
7'c'o
small
compared
to
cDa*^
+
o"'6'^
;
thus terms in
7
or 7'
must be
neglected
when terms
in
cd or
"'
occur.
Terms in
(7^)*or
{y'tYare
negligible,
and terms in
n are
small
compared
to
terms in a" or a"'.
Removing
the terms
which
are
negligible according
to
the above
hypothesis,
we
get
for the
equations
in
a,
and
/8',,
from
(27)
and
(28),
^'-Ko'(l-7'0i8',-"a',V"'(l + i70 8in2n^=ol
f'
)...(31).
?Ji
-
a,'(l "ry'") a',
+
""a'^Vo'* (^+i7'0
cos2n"=0
These
give
^^
+
ft)"
(1
-
27V)a',
=
a\b^ioc;' (ft)' +
2w -
^myt)
cos2nf,
^
+
""
(1
-
277)/3\
=
a'Aa)c/(ft)'
+
2n -
la"Wt)
sin2nt.
*
Motion
0/
Vortex
Rings^p.
77.
MR.
CUBEE,
YOSTICES IN A COMPRESSIBLE
FLUID.
117
Suppose
a'y
= 0 =
)8',
when e =
0,
then to the
same
degree
of
approximation
the solutions of these
equations
are
";=^T-^^j {(l
+ !7'0cos2n"-(l
+ i'/0cos"'"
+ i
^,(1
-
fn^e)
sin"'fl
(32),
'^'"-oT^^
^(^+ "'y''^ sin2ne-(l+ i'/0
"in"'e
+
i"7Vcos"^} (33).
Putting"/sO we
obtain Prof. Thomson's results
on
his
page
77.
The
case
m^s- 7n"
a =
J,
which
applies
to a
single
vortex
Earallel
to an
infinite
wall,
has
not
been
specially
treated
7
Prof. Thomson. It
presents
certain
peculiarities
which
seem
worthy
of notice. First
neglecting
the
compressibility,
the
equations (30)
become
^-"V.-"'AV"j' = 0.
dt
Here
o"', 5,e are
constants; thos,
if
oC^"O"ff^^
when
"=0,
the solutions
are
a"-6V(l-cosa,'0]
,o.^
/8;=
iVsIna"'"
j
^^*^-
The form of the cross-sectionat time t
is thus
givenbj
5'
=
6 - iV* cos2^
+
6V'
cos(2^-o)'0 (35),
the wall
beingperpendicular
to
the line ^
= 0 at
the distance
^c.
Thus the diameter of the vortex
perpendicular
to the
wall suffers
a
shortening 2b^c^^
and that
parallel
to the wall
an equal
lengthening,
and the vortex has
a
single disturbance,
of
period2irjm\
about this altered
position.
In the
same
case,
considering
the
compressibility alone,
we
have
^^
=
constant,
= 0
say,
and from
(29)
118
COL.
CUNNINGHAM,
DEPRESSION OF
Thus
"',
=
y._J^' (36),
neglecting higher
powers
of
(blc^Y.
The diameter
perpendicular
to
the wall would thus become
c:
'
and that
parallel
to the wall would become
2(6-a',)=26-2/,
+ J^^.
Supposeoa'g=:0,
then the diameter
perpendicular
to the
wall is
obviously
the
greater
if b be
decreasing,
i.e. if the
vortex be
approaching
the wall
;
the
reverse
is the
case*
if
the
vortex
be
receding
from the wall.
Considering
then either the
vorticity
alone
or
the
com- pressibility
alone,we come
to the conclusion that
a single
vortex in
presence
of
an
infinite wall will not
retain
a
truly
circular
cross-section,
and that the deviation from the circular
form varies
inversely
as
the
square
of the distance from the
wall. When both
vorticity
and
compressibility
are considered,
the deviation from the circular form will still
vary
inversely
as
the
square
of the distance from the
wall,
but the exact
change
in the form of the cross-section could
only
be deduced
from
a
complete
solution of the
equations(27)
and
(28)
for
the
case
6==0,
and this I have been unable to
obtain.
DEPRESSION OF DIFFEEENTIAL
EQUATIONS-
By
LU'CoL Allan
Cunningham^R,E,f
Fellow of
King's
Coll.,Lond.,
"o.
[BeferenceBto Boole's Treatise on Differential Equations^
2nd.
Ed.]
1. General Notation. As follows
:
"
a?,y
are
always
the variables of the
original equation,
f,
u
"
the variablesof
a
^'depressed
"
equation.
(^11ydy(",jyi)j(".iy.))(^4?yJ
^^e
the variables of
the
first, second,third,
and fourth
depressed equa- tions
formed in succession.
DIFFERENTIAL
EQUATIONS.
119
Lagrangian
notation is used for differentialcoefficients
where
compactness
is
required
;
thus
:^
y\y",...y^''^
denote differentiation with
respect
to
a?,
u\u'\...u^r)
^^ ^^ ^^ ^^
t^
2. Boole's
DepressioriB.
It is shown in Boole'a"Differen- tial
Equations,"
Chap.
X. that
a
differential
equation
admits
of
depression
by
one
order in each of the
following
cases :
"
i. When not
involving
y.
ii. When not
involving a?.
iii.When
homogeneous
in order
1,
and when
homogeneous
in order
fit.
iv. When
homogeneous
in order oo
.
But it is not shown whether these
depressions
can
be either
carried out in
succession^ or
repeated.
3.
Object ofPaper.
It is
proposed
here to
investigate
the
possibility
of
successive and of
repeated
depressions,
thus
greatlyextending
the
power
of the
process
of
aepression.
As to successive
depression
the
followingimportant
result
will be shown "
^'
A differential
equation
admits of successive
depressions
as follows
:
"
i.
By
r orders when the
r
quantities
y,
y',y",...y^*^''
are
all absent.
ii.
By
one
order when x
is absent.
iii.
By one
order when
homogeneous
in
any
one
order.
iv.
By
one order when
homogeneous
in
any
other
order."
Thus an
equation
not
Involving a?,y,
y',y","."y^'"*\
and
also
homogeneous
in
two
orders,
admits of
depression by
(r+3)
orders.
To
prove
the above it will sufficeto show that the
success
sive
depressions
" if
applied
in suitable order" do not
affect
the
remainingsingularities,
i.e.
producedepressed equation"
still
possessing
those
singularities.
120
COL.
CUNNINGHAM,
DEPRESSION OF
It will also be shown that in certain
cases one
of the
depressions can
be
repeated
without
affecting
any
of the other
singularities;
and
that in certaincases
the
application
of the
depressions ii., iii., or
iv.
produces
an
equation possessing
one
of the
singularities i.,
ii.
not
present
in the
original,
and
therefore
susceptible
of further
depression.
4,
Homogeneity,
It is convenient firstto
investigate
some
properties
of
homogeneity.
Supposing
a;,^
to be
quantities
of
degreesX, fi respec*
tively,
then the
following
scale of
degrees
obtains
:"
*
Quantities...
a?,y,
y', j/\ f\ ...y^
Degrees X, /*, /*
"
X, /*
"
2X, /x
"
3X, "../x
" rX.
Also the
degree(N)
of
any
"
term
"
Q consisting
of the
product
of several of these
quantities
is
clearly equal
to the
sum of
the
degrees
of its
component
factors
;
thus,
if
then
-AT^i^X
+
j/t
+ a
(/i
-
X)
+ iS
(/A
-
2X)+...p (/A
-
rX)
s=2/.X
+
Jf./tt,
where,
for
shortness,
icr^
-
(a
+ 2/8+ 37 +...rp),
if=j-|.(a+ iS+
7
+..../)).
Now,
it is obvious " from first
principles
" that all "terms"
with coefficientsof
equaldegree
connected
by
the
signs
+,
^,
=
in
any
sort of
"
equation
"
must
necessarily
be
^
same
degree{N) throughout
the
equation
;
hence the theorem
:
"
"
Every
differential
equation
with coefficientsof
equal
degree
in each term
is
homogeneous
in
Xj y, y\y",
"c.,
and
may
therefore be
depressed
one
order."
A
homogeneous
differential
equationbeing
then
a sum
of
terms of form
Q
of
equaldegree
N
may
be written
s((2)=o,
and the
homogeneity
is
expressed by
the
following,
which
may
be called the
"equation
of
homogeneity,"
L.\
+
M.fi
=
N(9,
constant for
every
term).
Def. The ratio
v^fi;\
of the
degrees
of
y
and
x
is
called the
"
order of
homogeneity,"
and the
quantity
N is
calledthe
"
degree
of
homogeneity,"
122
COL.
CUNNINGHAM,
DEPRESSION OF
But these
give,introdacing
an arbitrary multiplier kj
L
(Xj
+
*X,)
+
3f
(/ij
+
kii^
=
(N^
+
kN^^
also
a
constant,
and this result
exjresses
that the
equation
has
homogeneity
of order
v^
=
(/i^
-f
ft/*,)
:
(X^
+
A\))
and therefore of
every
order,
since k is
arbitrary. Q.E.D.
As
to the
possibility
of such
multiple homogeneity,
the
necessary
and suflScient conditions
are easilyseen
to be
L and
My
both constant
for
every
term,
or, writingout,
;?
-
(a
+ 2^ +
37 +...+ r/o)
=
ir,(a
constant
for
every
term)^
2+
(a+ P +
7
+.""+
p)=Jf,(a
constant for
every
term)^
and these conditions
can
be satisfied
by
more
than
one
distinct
system
of values of
^, j, a,
/S,
...p
only
when at
least three
of these
quantities
enter
into
every
pair
of
systems,,
or
(which
is the
same
thing)
when
at
least three of the
quantities
a, y,
y\ ...y^^
appear
in
every pair
of
terms:
hence
^^
Multiple homogeneity
is
possible only
in
an equation
containing
at least three of the
quantities a;,y,
y^^
"."y^^
in
every pair
of terms,"
The fact that
duplexhomogeneity
involves
multiple
homo- geneity
bears the
following
consequences
:
"
^'
A differential
equation
which has
multiple homoge-
geneity
cannot in
general
be
depressed
more
than
two orders in virtue
solely
of
homogeneity."
^'The
depression-formulse
suited
to
any
order of
homogeneity
that is convenient
may
be
applied
to
depress
a
differential
equation
which has homo- geneity
of
any
two orders."
Examplesofmultiple homogeneity.
It is worth notice that
multiple homogeneity
occurs* in allfinaldifferential
equations
of rational
algebraic equations(formedby
elimination of all
the
constants).
6. Depeession i. Absence
of
y,
y',y","..y^'^'^
(r
quan- tities).
The
depression-formulae are
(Boole,
Ch.
X.,
Art.
1)
*
ThiB
property
will be
proved
in another
paper.
DIFFERENTIAL
EQUATIONS.
123
Thus this
depression
introduces the
new independent
variable
(t)only
in
place
of the old
one (x)
when
present
in the
original.
As to the effect
on
homogeneity,
it is clear that if
a?, y
be reckoned of
degreesX,
/x,
then
*,
u must be reckoned like
iu, y**^,
i.e. of
degreesX, /*
" rX
(Art.4).
Thus
a;, y^'"\
y**\ """y*^
are
replaced
by
the
quantities ",w,
w',
...i*^"**'^ of
degreesX, il/, (ilf
"
X),(Jf-2X),...{il/- (n- r)X},writing
if=(/Lt
"
rX)
for shortness
;
i.e.
by quantities
of like
degree
with themselves. Thus the
degree
of each term in the de- pressed
equation
is the
same as
that of the
corresponding
term
in the
original.Hence,
if the
original
had
single
homogeneity
of
any
order
fi
:
X and
degreeNj
the
depressed
equation
has
single homogeneity
of order M
:
X and of
same
degree
N,
Similarly
ifthe
original
has
multiple homogeneity,
so
also has the
depressed equation.
Thus this
depression
does not
destroy
the
remaining singu- larities
ii.,iii.,
iv.
;
but it introduces the
new
dependent
variable
(w),
so cannot be
repeated (unless y*^,y"*\
"c. be
also
wanting).
As this
depression
is much easier of
appli- cation
than
depression ii., iii., iv.,
it would
usually
be
applied
fio as to
depressbi/
as
many
orders
as
possible
at
one
step,
and
therefore could not be
repeated,
because
y^*'^
is
herebysupposed
present
and would introduce
u.
7. Depeession ii. Absence
of
X.
The
depression-formulae
are
(Boole^
Ch.
X.,
Art.
1),
y^t,
y'^Uy y''
=
uu% y'''
=
uW'-^uu'%
y"
=
m'w'"
+
4m'm V
+ wm",
"c.
;
and,
in
general, ^/*V
=
(^
"
^^Y **"
As th"se
depression-formulae
thus
generally
introduce
the
new dependent
variable
(m),
with
one
exception
noted here- after
(Art.21,
wherein
u
cancels
out)depression
i.
cannot
be
applied
after the
present (except
in the
case
reserved).
Also this
depression always
introduces the
new
independent
variable
(t)
in
place
of
y,
and therefore cannot be
repeated
when
y
was
either
originally
present
or
when introduced
by
depression i,(see
remarks at
end of last
Article).
As to the effect
on
homogeneity,
the
depression-formulae
fihew that if
a;,y
be reckoned of
degreesX, /a,
then
tj
u
must be reckoned
like
y,
y
of
degrees
fi^
fi-^X respectively,
124
COL.
CUNNINGHAM,
DEPRESSION OF
80
that
the
quantities
ty tt,
t*', tt", t*'",
...
t*"^"
are
of
degrees
fly
(/i
"
X),
-
\,
" X "
/A,
" X "
2/A,
..."
X "
(r" l)/".
On
referring
to the
depression-formulaB,
it will be
seen
that
y,
t/\t/'\"""y^'^ are
replacedby homogeneous
functions
of
t,M,
w',
...w^*""*^ of
same
degrees
as
their
own. Hence,
if
the
originalequation
have
homogeneity
of
any
order
fjb
:
X
and of
degree
N
(so
that each of itsterms
are
of
degree-AT),
the
depressedequation
has
homogeneity
of order
(/li
"
X)
:
/*"
(the
ratio of the
degrees
of
m,
f),
and of the same
degree{N}
as
the
original.
Hence
also,
if the
original equation
haa
multiple homogeneity,
the
depressed
will also have
multiple
homogeneity.
8. Depression iii.
Homogeneityof
order
v.
Here
v
may
be
zero,
but
not
infinite. The
depression-formulae
are
(Booky
Ch.
X.,
Art.
3,
Class
II)
X:=tJuj y^x.ty
BO
that
D^^'y
= (e~/?
. uDtY
{s^"^'^/?
(m
+ vt)].
The
Quantity
a? or
eJ
"
will be found to cancel out of the
depressea
equation,
which
may
be formed
directly by
re- placing
a:
by 1,
y
by f, y'by {u-\-vt\
f by {uw'+ (2i/-l)M+ v(v-l)f},
f by {uV +
ttw"
+
3
(v
-
1)
uu'
+ (3v'
- 6v
+ 2)
m
+ y(y-.l)(v-2)f},
"c.
As these
depression-formulae
thus
generally
involve both
of the
new variables
(tyu)
with certain
exceptions
considered
hereafter
(Arts.
21,22) (wherein
f,
u
cancel
out) depressions
i.,
ii.
cannot be
applied
after the
present (except
in the
cases
reserved).
As to the effect
on
homogeneity,
the
depression-formulas
give
DIFFERENTIAL EQUATIONS.
125
from which it follows
that,
if
a?, y
be reckoned of
degrees
X, /A ;
then
^since
v
=
/a
: X) f,
u must
each
be reckoned
of
"qual
zero
aegree,
or
in other words of
equal
infinitesimal
degree
i,
so
that,bj
Art.
4,
the
quantities f,w,
t*', m", u"',...
u^**'
are
of
degreest, ",
0,
-t,
"
2t,...-(r-l)t.
On
referring
to
the
depression-formulae,
it will be
seen
that
X
is
replaced
by
a
quantity
of
zero degree,
whilst the
rest
of the
quantities y,
y',y",
"c.
are replaced
by
homo- geneous
functions of
^ w,
w',m",
"c. allof
equal
degree
i.
Hence this
depression destroys
the
homogeneity
of
an
equa- tion
possessing singlehomogeneity
of order
v
(notinfinite),
and therefore cannot
be
repeated
in
equations
of
single
homogeneity.
But
an
equation
possessing
multiplehomogeneity
has
necessarily homogeneity
of order oo Tby
Art.
5),
wherein
x
is reckoned of
degreezero,
and
y, y ,
y",
"c. all of
equal
degree.
But these have been shewn
above to be the
degrees
of the functions
(off, w,
u\ u'\"c.)
which
replace
them.
Hence,
in cases
of
original
multiple
homogeneity,
this
depres- sion
produces
an
equation
with
singlehomogeneity
of
same
degree(N)
as
the
original,
and of
firstorder
(sincet^
u are
equaldegree).
Hence
an equation possessing
multiple homogeneity
admits
of two
successive
depressions
(and
no more)
in virtue
thereof,
viz.
1st.
One
depression
by
the formulae for
homogeneity
of
any
arbitrary (notinfinite)
order
(v)
:
this leaves
singlehomogeneity
of firstorder.
2nd. One more
depression
by
formulae for
homogeneity
of firstorder :
this
destroys
the
homogeneity.
9. Depression
iv.
Eomogeneity
of
order oo
.
The de- pression-formulae
are (Booh^
Ch.
X.,
Art.
3,
Class
III)
x^t, y
= e^"^', so
that
i);"y
= 2"/(e^"*.t/),
and the
quantityy
or
e^"
is found to cancel out
of the
depressed equation,
which
may
be formed
directly by replacing
xhj t,
y
by 1, y' by
",
y"
by (m'
+
m'),
y'"
by (w''+
3mu'
+ u"),
y'-by (w'"
+
4mm"
+
3u"'
+
6mV
+ u*),
"c.
126 COL.
CUNNINGHAM,
DEPEESSION OF
As these
depression-formulsB
thus
generatlj
iDtrodace tite
new
dependent
variable
(w),
with
an
exception
noted hereafter
(Art.21) (wherein
u
cancels
out),depression
i. cannot be
applied
after the
present(except
in the
ease
reserved).
Again,
as
they
introduce the
new
independent
variable
(")
only
in
place
of the old
one (a;),
this
depression
does
not
affectthe
applicability
of
depressionii.,
which
can therefore
be
applied
after the
present
if
x was
originally
absent.
As to the effect
on
homogeneity,
the
depression-formulaar
give
from
which itfollows
that,
if
x
be reckoned
as of
degree
" f
y
and
y,
y\ y'\
"c. all of
equal
infinite
degree(see
Art.
4)^
then t must
be reckoned
(likea?)
of
degree
"
I,
and
u
of
degree1,
so
that
(by
Art.
4)
the
quantitiest^ w,
u',w","'",
...
u^^
are
of
degrees"1, 1, 2, 3, 4, ...(r+1).
On
referring
to the
depression-formulas
it will be
seea
that
oj,y,
y',yf\
"e.
are
replacedby homogeneous
functions'
of
t^w,
u\ v!\
"e. of
degrees
as below
:
"
Degree
of function
-1, 0, 1, 2, 3, .,. r,
replacing oj, y,
y\ y",f\ ...
y('*^".
Hence this
depression destroys
the
homogeneity
of
a"
equation
possessing singlehomogeneity
of order
go
,
and
therefore cannot
be
repeatedon
such
an
equation.
But
an
equation
possessing multiplehomogeneity
has
necessarily
homogeneity
of order
zero (Art.5),
wherein
x
is reckoned
of
degree
-
1,
and
y,
y',y",
"c. of
degrees0,1,2,
"c. But
these have been shewn above to be the
degrees
of the functions
(off, tt,
u\ "c.)
which
replace
them.
Hence,
in
cases
of
original multiplehomogeneity,
this
depressionproducesan
equation
with
singhhomogeneity
of
same
degree(N) as
the
original
and of order " 1
(sincef,
u are of
degrees
"
1,
1).
Hence an
equation possessing multiple homogeneity
admits
of two
successive
depressions (and
no more)
in virtue
thereof,
viz.
Ist* One
depression by
formulas
for
homogeneity
of
order go :
thisleaves
singlehomogeneity
of order " 1.
2nd* One more depression by
formulsefor
homogeneity
of order - 1
:
this
destroys
the
homogeneity.
DIFFERENTIAL
EQUATIONS.
127
10.
Depression^FormuhB.
Table II. shews the functions
t)f
",w,
w',m",
"'" that
are to
be substituted for
a:,y,
y',y",y'"
in
performing
any
one depression,
and also the
general
sub- stitution
for
y^^K
The substitutions for
homogeneity
are.
not
ihe actual values of
a?,y,
y',"c.,
the functions of
a:, y
which
would
necessarily
cancel out
of the
depressed equation being
t)mitted.
11. Elevatton-FormulcB. Table III. shews the functions of
a?,y,
y',y",
"c. that
are
to be substituted for
t,u,
"',m",
"c.
in
raising
the order of
any
differential
equation
in
t^ Uj
also
the
general
substitution for u^K These substitutions
are
actual
equivalencies.
Depbession-Formul^.
Table II.
\xf
y
the
original
variables
; t,
u the variables in
depressedequation].
N.B. In these Tables the
sign
H
should be read
as
becomes^
or
may
be
changed
into.
Depression
i.
y,
/, y",...y^"'^ wanting.
x =
t^ y^'-^
=
",
y'^
=
"',y^'^^w","c., y('"^)
=
tt""".
Depression ii.
x absent,
y
=
t, y'=^u, y"^uu\ y'"==uV'+ttM",
y'-
= uV"
+
4ttVtt"
+
""'",
Depression iii.
Homogeneity of
order
y ss " 1.
ajEl, yE^, y^'Eu-t^ y''
=
uu'-Su
+
2t,
y"'
=
uV"
+
ttw"- 6mw'
+
llw -
6^,
y"'
=
ttV4
4MVw''+ttM''-
10(uV'+um'')+ 35wm'-50m+24^,
128 COL.
CUNNINGHAM,
DEPRESSION
OP
Depression iii.
Homogeneity of
order
f
= O,
y"'
=
u
V
+
uu'^- Sue/
+
2tt,
Depression iii.
Homogeneity of
order
v
=
I.
y'"
=
t""M'"+4wVm"+
tttt"-2
(mV'+ mO
- Mtt'-f
2tr,
Depression iii.
Homogeneity of
order
y
= 2.
OJEI,
y
=
f, y'
=
w +
2^, /'
=
wi*'
+
3m
+
2",
y'"
=
m'm"
+
wu'"
+
3wu'
+
2m,
3^
=
uW+ 4mVm"+ wm'"+ 2
(mV'+ ttO
- ttw'-
2i^^
y(r)
=
(g-/?
e^J3
J'
{s/f
(m
+
20).
Depression iii.
Homogeneity of
order v,
xEl,
y
=
tj y'Eu +
vty y"
=
wM'
+ (2v- l)w+ v(K-l)f,
y'"EwV'+WM"+3(v-l)t*u'+(3v'-6v+2)ti+v(v-l)(v-2)f,
+ (4v'-18v" + 22v-6)m + f(v-1)(f-2)(v-3)",
y(-)
E
(e-/?
uDX'
{e""'^/-
("
+
vf)}.
Depression iv.
Homogeneity of
order oo
.
a;=f, y
=
l, y'Ew, y"
=
w'-fM*,
y"-
E
w'"
+
4mm"
+
3m'"
+
6mV
+ u\
CONTENTS.
PAGB
Vortices in
a compressible
fluid
(continued). By
C. Chreb - - - 113
Depression
of
differential
equations. By
Lt.-Ck)l. Allan Cunningham - 118
The
following
papers
have been received
:
Mr.
Bnchheim,
"
Note
on
Matrices in Involution."
Prof.
Oayley, "Analytical formulae in
regard
to an octad of
points," "A
correspondence
of confocal Cartesians with the
right
lines of
a
hyperboloid,*
"Note
on
the relation between the distances of five
points
in
space."
Prof.
Mathews,
"
Geometry on a quadric
surface."
Articles for insertion will be
received
by
the
Editor,
or by
Messrs.
Metcalfe and
Son, PrintingOffice,
TrinityStreet,
Cambridge.
ISVS'
""jr;
7cfA
No.
CO.]
NEW SERIES.
[December,
1887
THE
MESSENGER OP MATHEMATICS.
EDITED BY
J. W. L.
GLAISHER, So.D.,F.R.S.,
FELLOW OP TRINITY
COLLEGE,
CAMBRIDaE.
VOL. XVII "NO. 8.
MACMILLAN
AND CO,
^I'iHonOonani
"ambrilige*
1887.
Price" One
Shilling.
METCALFE
AND SON,
CAMBEIDGB,
-AAfl'
diff"iiektial
equations.
129
Elevation-Foemul^.
Table III,
[t,
u the
original variAbles
; a;,y
the variables of the raised
equation].
Elevation i.
y,
y\ /',...y^'" wanting,
t^x, u:=y^'\
u'=
y(^'\m'=3^''%"c.,u(""=y('"**".
Elevation ii.
x
absent
^^.._y^y-^yyy"+3y
3^ y
Elevation iii.
Homogeneity of
order vsO
. .
aw".
"'"'=ft^.)"^'.
Elevation Iii.
Homogeneity of
order
v"
1.
a;'
^
a'
xy -y
'
Xxy-y V
"
VOL, XVII.
130
COL.
CUNNINQHAM,
DEPRESSION OF
Elevation iii.
Homogeneity of
order
v.
t "
"p
.
U^
p
y
U
7
"
Kj
oj
'
a;
^y -^yy
\xy
-
yy
V
a"
Elevation iv.
Homogeneity of
order
oo
,
y y
"
=
-3
"
u(^)=Dp(yj
12. Successive
Depression.Summing
up^
it is
seen
that
to
apply
the four
Depressions
in
succession;
the
following
must
be attended to
:
"
Depression1. This should
always
be
appliedfirst, as
it does not affect
the
remainingDepressions,
whereas
they
" if
applied
first" will
generally
prevent
its
application (byintroducing").
'Depression
ii. This should
precede Depression
iii.
(because
the latter
in
general
introduces
();
it
changes
the order of
homogeneity (when single)
from i/ = /Li:\toi/ =
fi
"
\;/Li.
Depression
iii. This in
general
introduces
",
and reduces
multiple
homo- geneity
to
single
of order
v = 1, so cannot
precedeDepressions
ii.
or iv.
Hence in cases of
multiplehomogeneity
the
procedure
is
;
"
(1)
When not
precededby Depression
iv. it
may
be
appliedonce
with
any
convenient value of
Vy
and
again
with the value
v
= 1.
(2)
When
precededimmediatelyby Depression
iv. it must be
applied
with value
i; = " 1.
(3)
When
precededby Depressions
iv.and ii.in turn it must be
applied
with value v=2.
Depression
iv. This
may precede
either
Depressions
ii. or
iii.,
but
cannot follow
Depression
iii.
(because
the latter leaves
homogeneity only
of order
v
= + 1);
it reduces
multiplehomogeneity
to
single
of order
v = -l.
DIFFERENTIAL
EQUATIONS.
181
The
following
Table shews the order
in
which the
Depressions
should be
applied
to
equations containing
two or
more of the
singularities
detailed.
Thus in
certain cases
there
are
two or more
Courses
open,
viz.
2 Couwes in
eqnations
with two of the
singnlarities
Nos. ii., iii.,
ir.
8 Conises in
equation!
with
sinffolaritieB
Nos. ii.,
iii
,
ir.
K2
132
COL.
CUNNINGHAM,
DEPRESSION OP
The
particular
Course most
advantageous
will of
course
depend
on
the nature of the
question
;
but it will
probably
generally
be
advantageous
to use
Depression
iii.with the
general
value of
Vy
whenever
available,
as
thisleaves
a constant
to
be hereafter determined in such
away
as to
simplify
the
integrations.
13. Combined
Depresston-Formulm,
The resultsof substitu- tion
by
the several
Depressions applied
in succession
may
be
readilycomputed
once tor all
so as to shew the result after
any
number of successive
Depressions.
These
are
exhibited in
Tables
V, VI,
in which
a:, y,
y, y% "c.,belong
to the
original equation,
^v Vv Vx'i
yd ^^'J
'" " "
^^^^
depressed equation,
^
J
y,"Vi^
yi\ """"
" " "
second
" "
"3" ^8^y
\ yd "c-"
" y yi
^^^^^
" w
^vyA^yl^yd^^'t
" " "
fourth
" "
By
the aid of such
a
Table the
first, second,third,or
fourth
depressed equation
may
be formed at once
without
the
labour of
forming
the intermediate
depressions by simply
substituting
the functions in the Table
belonging
to the
required depressed equation
for the
original
a;,y,
y\
("c.
Successive Depressions.
Table V.
Depression
i.
Step
I in all
cases
" .
oD "
x^j y
"
y,j y "y\^y
"
yi " y\ "y\ jotu...
Two Successive
Depressions,
Depressions
1.and
ii.,
combined,
yi=^,i
y/=y,jyi'=y,y2i2^r=3^,(y,y,"+y/'))
yr-y,Q/:yr^^y.y:y.''^yry
Depressions
i.and iii.
(v
=
0),
combined.
x^=l,y,
=
a;,,
y,'
=
y"
y"
=
y,
(y/
-
1),
3'."'2y;(y.i/."'
+
*:^.'y;'- 6^.")+.y.Cy.'- i)(y.'-2)(y;- 3).
134
COL.
CUNNINGHAM,
DEPRESSION OF
Depressions i.,li.,
and
iv.,
combined,
y.
=
ar"
y,'
=
1,y,"
Hy,,
y/"
=
y.'
+
2y;,
yrsy."+7y.y,'+ 6y/.
Depressions \.y
ili.
(v
=
0),
am? iii.
(v= 1),
comhined.
a'l^l.
yi
= i"yi'=a!"y,"
s
a;,
(y,
+ ",-!),
y.'"=
a'.V. (/.+ 1)+
a',
(y.
+
a'.
-
1)(y.
+
a;.
-
2),
y.'"
=
"y, {x,(y.y,"
+
y/'
-
1)+
(y.'
+ 1)(4y.
+
4x.
-
6)}
+
",
(y.
+
a;.
-
1)(y,
+
ar,
-
2)(y,
+
a;,
-
3).
D^ressionsi.,
iii.
(v= 1),
and iii.
(v= 1)
combined.
".
=
!.
3',
=
1"yi'= a?,+
l,y,"=a!,(y,
+
a!,
+
l),
y/'-
=
a;."y,{ar.(y.y;'
+
y,'"-l)
+
3(y,+ a!.) (y.'
+
l)}
+
a',
(y,+
",
-
2){a;,y, (y,'
+
1)+
(y,
+
a?^'
-
1}.
Degressions i.,
iii.
(v= v),
an"Z iii.
(y"l),
combined.
or.
=
1,
y,
=
I,y,'
=
",
4
v,
y,"
=
a;,y,+(a;,+ v)(a;,+
v
-
1),
y,'"
=
a;,y,
Ky',
+
y,"
+
3
(x,
+ 1^
-
1)}
+
(a;,
+
")(a'.
+ "- 1)(",+"-2),
y."= "y^{y^y:'
+
y,"
+
4y'.
+
6)
+
a^.'y, {(4y,
+
4v -
e)y;
+
7y,
+
6(2i'-3)}+x,y,{y,'+(4v-6)y, + (6v"-18i'+ll)}
+
("'.+"') ("",+
"'-
1)(aJ,
+ i'-2)(a;,+
y- 3).
Depressions i.,
iv.ane^ ii.comhined.
yi
=
i"yi'=a'"y."sy,+a;,",
y."'=y.y.'+3a'.y,+",
yr =
y.
(y^y."
+
y" +
^,y"
+ e^,'
+
3y.)
+ ar,\
Depressions i.,
iv.ant? iii.
(v
=
"
1)
combined.
".
=
!"
y.
= l.y.'^ar,, y,"
=
y,
+
a5.(a!,-l),
y."' = y.(y.' +
3a-.-3)
+
a:.(a;,-l)(x,-2),
y/'=y. {y.y."+ y,"+^^.y.'- 6y;+3y.-
1
+
e
(x,-
1
)(x- 2)}
+
ar.(x.-l)(a;.-2)(x.-3).
DIFFERENTtAL
EQUATT02(S.
135
Four Successive
Depressions.
Depressions i.,ii.,
iii.
(v
=
0),
and iii.
(y" 1),
combined,
y,
E
1,y.'
=
1,y,"
=
x"
y.'"
=
x^
(y,
+
2x,
-
1),
Depressions
i.,
ii.,
iii.
(v= 1),
am? iii.
(v
=
1),combined,
Depressions i.,ii.,
iii.
(!"
=
v),
an(? iii.
(y=i1),
"^mbined.
y,=iiy,'^i,y,"
='"'"+"',
y."'=a;,y,
+
(a;,
+ v)(2a:, +
2^
-
1),
yi"
=
"'.y,
(^"y/
+
y"
+
'^r*
+7 v
-
3)
+
(a'"
+
"') (.^^4
+
3f -
2)(2x^
+2v- 1).
Depressions i.,ii.,
iv.awrf iii.
(y
=
"
1),
combined.
y,
=
l,y/=l, y."
=
x"
y/"=y,-ir,
+ 2x/,
y."=y. (y/
+
'^r,
-
S)+
x,
(2a;,
-
1)(Zx,
-
2).
Depressions i.,iv.,
ii.anc?iii.
(y
=
2),
combined.
y.=i, y.'^i,y,"=^.+i,
y."'='^4y.+(-"4+i)(2a;,+i),
y."'=a'/(y"y;+7y.+i7a;,+6)
+
1.
14.
RepeatedDepression.
The several
depressions
con- sidered
above
may
in
some cases
be
repeated.
15.
Depression
i.
repeated.
The substitution"=
f,y^''^=Uj
applicable
when the
r terms
y,
y',y,
...,
y*"*^
are
all
wanting,
and which
depresses
r
orders at one
step,
is
evidently equiva- lent
to r
repetitions
of this
depression, depressing
one
order
at
each
step,
viz.
by
the substitutions
but,
as
the
same
final result
may
also be obtained
by
the
single
substitution
(x
=
tj^^'^^=w),
there is
no
advantage
in
this
repetition.
16.
Depression
ii.
repeated.
On
referring
to the
depression-
formulae
(Art.7),
itis
seen
that
(in
the absence of
x)
y,
y^
136
COL.
CUNNINaHAM,
DEPRESSION OF
become
t,m,
and
t does not
recur in
the substitutions for
y, y'",
"c.
Hence,
if
y
be absent from the
original,
t will
be absent from the
depressed equation,
so
that
depression
il.
may
be
applied
a
second time.
The h
priori recognition
of the further
applicability
of this
depression
is
more
diflScult.
Thus,
if
u
be absent from the
first
depressed equation (as
well
as
^),
this
depression
could be
applied again,
inasmuch
as
y'
=
w,
and,as further,u
enters
into the substitutions for
y'^
and all
higher
differential
coefficients
(Art.7),
it is clear that the absence of
u
is secured
only
when
y',y'',y,
"c.
occur
in the forms
(given
in
Table
III)equivalent
to
u\ u^\"c.,(and
in
no other).
The results of the double
application
of this
depression,
applicable
when
a?, y
are
both absent from the
original
equation,
are
shown in Table
VII,
with the
new variables
of
Art. 13. It is
probable
that this double
application
will not
often be
advantageous,
as
the
depression by
two orders
might
in the
same case
(a?,
y
both
absent)
be
generally
more
simply
efi^ected
by
the
use of
depressions i.,
ii.
applied
in succession.
The results of further
application
of this
depression
cannot be shown in
an
equallycompendious
way
with
the last.
As
by
Art.
7,
a
singleapplication
of
depression
li.does
not
destroyhomogeneity(merelychanging
the
order,
when
single),
it follows
easily
that
repeatedapplication
will not
destroyhomogeneity,merelychanging
itsorder if
single.
17.
Depression
iii.
repeated.
The
a
priorirecognition
of
repeated applicability (with
same
value of
v)
is
by
no means
simple.
The
original
equation
must of
course
possess
homo- geneity
of
some
finite order
(v)
in
a:, y.
Besides which the
depressed
equation
must also have
homogeneity
of the
same
finiteorder
(y)
in
tyU]
this involves that the
original equation
should contain
a?, y, y',y","c.,only
in the forms
given
in
Table III
as
the
equivalents
of
t, w,
u\ w","c.,
and should
possess
homogeneity
of order
v
in those functions
;
a
condition
so
complex
as to make the h
priorirecognition
of the
repeated
applicability
of this
depression
difficult.There
is,however,
one
case
of
easy
recognition,
viz. that of
multiple homogeneity,
in which
it
has been shown
(Art. 8)
that this
depression
may
be efi'ectedtwice with the value
v = 1 in each instance.
By reasoning
similar to that in Art.
8,
it is
seen
that
the
repeatedapplication
of this
depressiongenerallydestroys
"the
applicability
of
depression
ii.
(by introducing
the inde- pendent
variable)
;
also,
that in
an
equation possessing
DIFFERENTIAL
EQUATIONS.
137
inultiple homogeneity
as
well
as
the
property
above
mentioned,
one
application
of this
depression
will stillleave
multiple
homogeneity.
18.
Depression
iv.
repeated.
The h
priori
recognition
of
repeatedapplicability
is
by
no means
simple.
The
original
equation
must of
course
have
homogeneity
of order oo
in
a;,
y ;
and the
depressed equation
must
have the
same
in
t^
u.
This involves that the
originalequation
should involve
y,
y, y, "c., only
in the forms
given
In Table III
as
equivalent
to
w,
u\v!\"c.,
and should further be
homogeneous
in those functions reckoned of
equaldegree.
Ex.
/
r^ -
^]
= 0
is of this kind.
As this
depression
introduces the
independent
variable
only
in
place
of the
original
a;,
if
present
in the
original equation,
it
IS
clear that its
repeatedapplication
does not
affect the
applicability
of
depression ii.,
which
can
therefore be
applied
after
repeated application
of
depression
iv. if
a; were
originally
absent.
Similarly
In
an
equation possessing multiple homogeneity,
the
repeatedapplication
of this
depression
leaves
finally
(as
in
Art.
9) singlehomogeneity
of order " 1.
19.
RepeatedDepression
Formulae.
Table Vll
gives
the results of
a
double
application
of
each of the
depressions ii., iii.,
iv.
Repeated DEPRESSiON-FoRMULiE.
Table VII.
a;,y
the
original
variables
;
^1) Vx 5 ^j? y%
^^ variables of the firstand second
depressions,
Repeated Depression ii. Absence
ofx^
y.
Repeated Depression Iii.
Homogeneity
of
order v"\
in
aj,y
and in
ar^ y^.
dj=l, yEl, y=a;,+ l; /'=
a;,
(y,
+
a^,
+
1),
y''="y,K(y,y/'+y"-i)
+
3(y,+ar,)(y;+i)}
138
COL.
CUNNINGHAM,
DEPRESSION OP
Repeated
Depression iv.
Homogeneityof
order oo in
Xj y
and in
x^^y^.
x=:x,, y=l, /
=
1, y"
=
y,+ l,
y"'
=
y/'
+
3y^/+ 4y;
+
y/
+
7y/+
6y,+
l.
20. Other
depressible cases.
Inasmuch
as
Depressions
ii., iii.,
iv. introduce in
general
both the
independent
and
dependent
variables
(f,u)
in
place
of
a?,y,
y\
"c
,
it
may
happen
that certain functions of
a;,y,
y\
("c. in the
original
give
rise
on
application
of those
depressions
to a
depressed
equation
free from
one or
other of the
new
variables
(m,"),
and therefore
depressible by Depressions
i.
or ii.,although
the
original equation
may
have contained
y
or
a?,
and
was
not
therefore
depressible by Depressions
i.
or
ii.
21. Absence
of
w,
"c.
(Depression i.).
On
examining
Table III. it is at
once seen that"
**
A differential
equationwanting
x,
or
homogeneous
in
any order,
and
also
involving
", y, ^',
y",
"c.
only
in the forms
equivalent
to
",t*'*"^, t*"'""^*",
Ac.
of the
Depressionli,iii,
or
iv, (as
the
case
may be)
will" after that
depres- sion"be
found free from the r
quantities ",
t*', u",...tt^*"-",
and be therefore
further
depressible by Depression
1."
22.' Absence
of
t
(Depressionii.).
On
examining
the
depression-formulae
of
Depression
iii.
(Table
II.),
it is
seen
that t enters in the first
degreeonly
into several of the substi- tutions
for
y,
y',
y^\
"c.,
and will therefore
disappear
from
certain
simple
functions
thereof, depending
on
the value of
Vj
as
follows
:
"
DIFFERENTIAL
EQUATIONS.
139
Hence it follows that"
''A differential
equationhaving homogeneity
of
any
order
v {v not
infinite),
and either not
containing
the
{y + 1) quantities
y^
y'j\f\ '""t^^^t
or
else
containing
them
only
in the forms shewn above for the value of
v,
will after
Depression
iii.leave
an equation
free from
ty
and therefore further
depressible by Depression
ii.'*
In the
case
of
an
equation
with
multiple homogeneity,
there is thus a
large
range
of
applicabilitj,
as
the value of
v
maj
then be chosen at will.
23.
Preparation of
equations.
An
equation
not
possessing
the
singularities
here considered
may
sometimes be trans- formed
into
one
possessing
them
by
a
suitable
change
of
variables.
Removal
of
x.
The well-known results.
When
X "
e*,
xD^^D,,
f(xD:)=f(D.\
a;'.i";
=
A(A-l)(A-2)...{A-(r-l)}.
When
a +
bx =
"*",
f[(a
+ bx)DJi^f(D.\
(aH-6^X2";
=
A(A-5)(^.-2")...{Z),-(r-l)J},
enable the variable
x
to be removed from
equations
which
contain
x
and the diflferentials
only
in forms
xD^^f(xDJ,
aT.D;^
or only
in forms
(a+ix)Z"^,/{(a+ia:)i)J, (a-\-bxyD;.
24.
Integration,
The
use
of these
depressions
is
chiefly
as a
help
in the
integration
of
high-order
non-linear difi^eren-
tial
equations.
The result of the
depressions
is either
a
diff^erential
equation
of lower order
or an
equation
free of
difi^erentials
(when
the order of the
original
is
equal
to the
number of its
singularities)
:
in the former
case
the
depressed
difi^erential
equation
must
be
integrated.
In either case
this
equation,
free of
differentials,
is the
startingpoint
for
a
series of
ascending
steps
in which the
successive
depressions
are reversed one
by
one.
Each
reversal
of
a
depression gives
rise to a
first-orderdifferential
equation
which is to be
integrated
before
passing
on to
the next
Step
:
these
correspond
to the
depressions
for
homogeneity
and
140
COL.
CUNNINGHAM,
DEPRESSION OP
absence of
x.
Lastly
there will be
a
series of
r
simple
iiitegratious, correspoodiug
to
the
depression
for absence of
if'^y^y
"
"""y
25. Conditions
forsuccess.
The conditions for the
practical
solution of
a
high-order
differential
equationby
this
process
are
"
1". The final
depressedequationmusl
be either free of
difEerentials, or
else a
solvible difEei-ential
equation.
'
2". The first-order differential
equationsarising
must be
separately
Bolyible.
8". The final
simple integrations(r
in
number)
must be
separately
possible.
26.
Example.
As
an
example
of these
principles,
the
equation
"
9yY-45yyV + 4""'
=
0,
which is the well known differential
equation
of a
conic
may
be taken. It wants
a:, y^
y\
and has
multiple
homogeneity,
so
may
be
depressed
five orders in
all,
the result
being
an
algebraic equation.
As shewn in Art.
12,
there
are
three Courses
open,
viz.
the
Depressions
may
be taken
in
any
of three orders
as below,
(beginningalways
with
Depression
i.
applied
to
the utmost
extent),
The results
are shewn,
step
by
step,
in Table VIII. :
the
finalresult
(StepIV.)
of each Course
might
have been written
down at
once
by using
the formulae of Table
V.,
without
shewing
the intermediate
steps.
The
integration
of the three final
depressed equations
of
each Course is shewn in Table IX.
It has been
thought
sufficientto indicate the
leading
steps
without
shewing
the
actual details of the
integrations (which
would
cover
several
pages).
It will be
seen
that" in this
particular example
" it has
been
possible
at
each
step
to solve the first-oiderdifferential
equationsalgebraically^
so as to
exhibit the differentialcoeflS-
cients,
as
explicit functions,
upon
which the solution
can
be
effected
by separation
of variables. In
particular
in Course
l"*,
142
COL.
CUNNINGHAM,
DEPRESSION OF
Course V. Step IV.
Depression
iii.
(v= 1).
(the
final
depressed equation,
wherein
v
is still
arbitrary).
Course 2*. Step II.
Depression
li.
9^;(M," + yn
-
45:r,y,y/
+ 40y/
=
0,
(anequation possessing multiple homogeneity).
Step III.
Depression
iv.
9^/(y/
+
2y3")
-
450^3^,
+
40 =
0,
^
(anequation homogeneous
in order
v = -
I).
Step IV.
Depression
iii.
y4
+
2(a?,-t)(a^,-|)
=
0,
(the
final
depressed equation).
Course 3^ Step II.
Depression
iv.
9y;'-i8y,y/+
4y,'=o,
(an equation wanting
or,,
and
homogeneous
in order
v = "
1).
Step III.
Depression
ii.
(anequation homogeneous
in order
v
=
2).
Step IV.
Depression
Iii.
(i'
=
2).
'r"y.+2.(x,-|)(x,-i)=o,
(the
final
depressed equation).
Differential
Equation of
Conic
Table IX.
Integration offinaldepressed equation of
Course
1",
2*
or
3*.
Course 1*.
Depressions i., li.,
iii.
(v=
v\
and iii.
(v= 1).
Final
depressed equation,
"^4y4
+
2(a:,
+
v-|)(a.,H-v-i)
= 0.
Digiti
ized
by
DIFFERENTIAL
EQUATIONS.
143
Course V, Step L
Depression
iii.
(v=l)
reversed.
FormuIcB, x^=^
, y^
=
x^-T-^.
Jbguattonm
a?,,a?^.
cr/ =
-^^-^^
" "
i .
-^.5. In these lattertwo results
v is
arbitrary.
Step II.
Depression
iii.
(varbitrary)
reversed.
Equation
in
a?"y,
solved
for
y^.
Make
V =
I
or
f
,
therefore
^3
=
"
ija?,.
doc
Formulae.
x,
=
jr/^,)
Vs
=
^.
^J^
"
Equation
in
a;,,a?,,
ca:/^
= (7"
a?,*.
Equation
in
a?"y,. y,
=
x^i^{A^x^\
+
B^xf^).
Step
IIL
Depression
ii.reversed.
Formul(B.
x,
=
y,,
Vt^j^*
Equation
in
a?j,yj. y,
=
(i^j
+
5^0?^
+
^i^j*)""^*
Step IV.
Depression
i.reversed.
Formulce.
x^
=
x^
y^^y'\
Final
equation
in
x^ y. y
= Ax
-V B"\/(^
Clc*
+
Bx
+
2").
"which is the
requiredcompleteprimitive,
and is the
algebraic
equation
of
a
conic.
Course 2\
Depressions i.,ii., iv.,
and iii.
(v
=
"
1).
-FVnaZ
depressed equation^
y*
+
2
(j:*^
"
f)(x^
-
1)
= 0.
144
COL.
CUNNINGHAM,
DIFFERENTIAL
EQUATIONS.
Step I,
Depression
lii,
(y
= "
1)
reversed,
JEqitation
in
a*,,x^.
cx^
=
-*
"
|.
Equation
in
^3,y,. ^jjr,
=
J
"
^^
" -"
+
f.
Step
II.
Depression
iv. reversed.
Formulce.
x^
=
x., y,
= "
-/^
.
Equation
in
^" y,. y,
=
x^lVC^a^a*
+
B^x^^).
Steps
III and IV. Same as
in Coubse
T,
j.r.
Course
3'.
Depressions
i.,
iv.,
ii.and iii.
(v
=
2).
Final
depressed equation,
ar^y^
4
2
(^^
"
|)(^^
"
J)
= 0.
Step
I.
Depression
iii.
(v
=
2)
reversed.
Formuliz.
^4
=
p"
y4
=
^s^*-
Equation
in
^3,jr^.
cx^
"
-^
"
^^,
.
v^4
""
t/
Equation
in
x^^y,.
fCy^
= 1
+
Cr,*" V(l
+
(^^D*
Step II.
Depression
ii.reversed.
Formulce.
x^^y,, y,=^-^
.
^2.a^i.n
tn
x" y,. y"
=
^^
^
^ig,;
+
g,^
'
Step III.
Depression
iv. reversed.
Formulae.
x^
=
x., y,
= " -r^
.
Equation
in
x^^y^. yi
=
(-^i
+
A^i
+
^i^i*)"
Step.
IV. Same
as
in Course
1%
j.v.
CONTENTS.
PAGH
Depression
of differential
equations (continued). By
Lt.-Gol. Allan
Cunningham
- - - - - - -
- 129
The
following
papers
have been received
:
Mr.
Bnchheim,
"
Note
on
Matrices in Inyolution."
Prof. Cayley, "Analytical
formulae in
regard
to
an
octad of
points,"
"A
correspondence
of confocal Cartesians with the
right lines of
a hyperboloid/'
'^Note
on
the relation between the distances of five
points
in
space."
Prof. Mathews,
"
Geometry on a quadric
surface."
Mr. L. J.
Bogers,
"An extension of the A. M. and G-.M. theorem in
inequalities."
Articles for insertion
will be received
by
the
Editor, or
by
Messrs.
Metcalfe and
Son,PrintingOffice,TrinityStreet,
Cambridge.
.
'XPn'
No.
CCIL]
NEW SERIES.
[Februaiy,
1888.
"^
ir,-,...v"-i.y
"
tSE-
MESSENGER OP
MATHEMATICS.
EDITED BY
J. W. L.
GLAISHER, So.D.,F.R.S.,
FBLLOW OT TRINITY
OOLLBGB,
CAMBRIDGB.
VOL. XVIL" NO. 10.
MAOMILLAN
AND 00.
^l-i
Uontton
anS
Cambriftfie.
1888.
Price"
One
Shilling.
Digiti
MRTCALFB Ain" BOK. CAKBRIDQB
'm
146
MB.
BOGEKS)
AN EXTENSION OF A
Firstly,
let
a,,a,, ...o^
be
integers.
Then
we merely
have
a
particular
case
of the well-known
theorem,
wherein
we have
a,
quantities,
each
equal
to
J^,"c.,
the whole number of them
being
a^
+
a,
+...
a^.
Secondly,
let
a,,a,,
...
be fractional.
Let N be the least
common measure
of their denominators
and let
Na^^A^^ ^a,
=
-4""c.,
then
we
get by
what is
proved
above
\
^,
+
^,+...
J
' *
Taking
the real
positive
N^^ root of each side
we
get
after
reducing
the bracketted
fraction,
the
inequality (1).
Thirdly,
letthe a's be incommensurable.
Then
we
may
substitute for each of these
quantities
fractions,
which
may
differ from them
by
less than
any
assignedquantities,
and since the theorem ia true
for the
substituted
fractions,
we
may
assume
italso true
for the
given
incommensurables.
Hence
we
may
consider
(1)
as
established.
It willbe found
conveniently
brief
to
write
s^
for
d^+a^-^"..j
as we
shall do henceforth.
Let
J,="
for all values of
r
from 1 to
n,
then from
(1)
/5A*i 1 1
("'
+
"'+"""
"")"'^""'"a".'"
(2),
a
well known result.
Write
flj'
for
a,, "/
for
a,,
c"c,
and let
5i
=
"i"''"
J,
=
a^'^j
where
m"r.
Then
(1) gives
Again,
let
b^
=
a*'' where
t "r.
Then
Q'^CO^^'^'-'-r^^
Combining
these results
we
get
fer"a)
CERTAIN
THEOREM IN
INEQUALITIES.
Taking
the
sj^
root and
reducing
we
get
147
provided
Let "=0.
.(3),
(0,
*""V""^""'
m"r"U
(y'"(rj
wbich for r =
1,
so
that m "
1,
we
get
n \
n /
'
a
well known
relation,
if m"U
If
tn
=
1,
^s=
0,
then f-M "
-^
,
where
r
is
a
positive proper
fraction.
""=",,=-,,
("^)-"'^,
and if
"ii=0,
re*
"
1,
then
(-=^)"
-=*.
Again,
from
(3)
we
may
deduce
many
other results.
If
t"Uf
we
also have
r "" t '
and ifw
+ w
= r +
^,
we
get
on
multiplication
or
V""V*
provided m"r"f"w,
and
wi + w " r4-".
The result
(5)
can
be obtained without
using(1).
If
we
multiply
the two
arrays,
(5),
1,1,
and
we
get
the relation
=2aX(ar-"-)("r-"r'),
the
right
side of which contains
onlypositive
terms.
Therefore
sjf^
"
sjs^
L2
148 MB.
ROGERS,
AN EXTENSION
OF A
Let
u
= 0 ia
(5),
so
that
^^f^^
"
8^8^J
we see
then that
in
the
same
way
and
so
on^
or as we
may
better write
it,
J^"J!:I
y.
(6),
a
resultwhich admits
of
easy
extension to n
suffixes
a,
^,7,
... .
We shall
now
pass
on to
applications
of the above results
to
Integral
Calculus.
"
2. In the
inequality "
1
(3)
let
where
^j
+
wA =
^,.
We then
get,
after
multipyling
5
,
5,,s^
by A,
and
putting
f{ai)^yy
and
making
h decrease
indefinitely,
{jy-dxrwdxr'"
wdxr*
o),
where
m"r"t^
and the limits
are
such that
y**,jf^
and
y*
remain
finiteand
positive
for all values between these limits.
As an
example
of this
we
may
put y
E
",
whence after
changing
m +
1 to
97Z,
"c.,
itfollows that
er(")'(?)'""
".
where m"r"t
Here
we
take fbr limits 1
and 0.
From
(1)
we
may
observe that itis
impossible
that
ra
[a*
I u'^dxy, I v'dx^l
(3),
where
the limits
are
independent
of
m
and taken
so
that
the
functions
w,
t;
should remain
positive
between their
re- spective
limits.
For
let
rwV^
=
^(7w).
Then,by (1),
CEKTAIN THEOREM IN
INEQUALITIES.
149
But if
(3)were
true
we
should have the
reverse
inequality
also
true,
which is
impossible.
Hence
(3)
cannot
hold
good.
As
an
example
we
have
/,
0
w +
1'
80
that
no
function
u can
be found such that
/;
a
h
where
a,
h
are
subject
to the afore-stated conditions.
As
we
deduced
(1)
from
"
1
(3),
so
may
we
draw from
"1 (4)
that
where
m"T^
and the limits
are
under conditions
as
before.
If
y
=
0?,
we
get
(r+ir"(7n+iy (5),
where m"r.
"
3. If
we
treat the
inequality "
1
(3)
in the
same
way
as we
deduced
"
1
(0
fi*^" the well-known A.M. and G.M.
relation, we
shall
get
(SaJT' (2ay)"^"(SaJT' (1).
From
"
1
(5)
we
get
2a6^2a5""2ay.Say
(2),
and from
"
1
(6)
^aV^^^'^'"
2aJ" ^af
,.,
2^^
^"2^
"26"
W-
These
give
results
similar
to
"
2
(1),
viz.
Uyv''dxn!yv'dxr-"{jyv^dxT- (4),
jyv^dxjyv'dx"jyv'^dx jyv*dx (5),
jyv'^^'dx
^
jyv'dx
jy^dx
^
.
Jydx Jydx Jydx
^ ^^
where
y,
v are
functions of
Xy
and the limits
are
under the
same
conditions
aa
before.
150
provided
the sines
are
all
positive.
^e
may
also
get
similar
inequalities
from
(5)
and
(6).
As in
"
2
(3),
we
may
also show that if
I
yv^dx
=
^ (rn\
where
o,
i,
y,
v are
subject
to
the
same
conditions,
then
we
cannot
have
with conditionsas
before.
The
following equations
are
thereforeabsurd
:
I
M"
dr =
-p- ,
or
-^"r^
" -.
if w""* is
alwayspositive
between a: =
Xj
and
a: =
x"
and
ar^,ar,
are
independent
of
m.
"
4. We
may
also obtain
a
few
inequalities
from
taking
logarithms
in
"
1
(1),
whence
,
SaJ SaloffJ
Fron;i
this
may
be deduced
. Jvydx
fvlogydx
^""^
Jvdx
^
Jvdx
"
with restrictions
as
before
as
to limits.
These last
inequalities
do not
appear
to lead to
very
interesting
results.
-
Ozfcod,
Ab9.
1,
1887.
(
151
)
GEOMETRY ON A
QUADRIC
SURFACE.
By
Prof. Mathews.
From
a
fixed
point
P of
a
quadric
surface
project plane
sections of the surface
upon
a
plane
which is
parallel
to
the
tangent
plane
at
P;
then* the
projections
will all be similar
and
similarly
situated.
This is
easilyprovedanalytically;
or
geometrically, by
straining
the
stereographic
projection
of
a
sphere-
In
particular
suppose
tnat P is
an
umbilic
Uj
then the
projections
become
circles,
reducing
to
straight
lines for
sections
passingthrough17,
and
conversely
to
every
straight
line
or
circlein the
plane
of
projection corresponds
a
plane
section of the surface.
To
a
coaxal
system
of circles
corresponds
the
system
of
conies in which the
quadric
surface 8 is intersected
by
an
axial
pencil
of
planes.
If the axis of the
pencil
meets
8 in
two
real
points
A and B
every
conic of the coaxal
system
will
go
through
A and
P; if,on
the other
hand,
A and P
are
imaginary,
no two
conies of the coaxal
system
will intersect
in real
points.
In this latter
case
there
are two
"point-
conics
"
of the
system,
viz. the
points
of contact of the
two
planes
of the
pencil
which touch 8,
The
plane
of the
pencil
which
passes
through
U meets 8
in
a
conic,
which
may
be called the radical axis of the
system.
The axis of the
pencil
of
planescorresponding
to
a
coaxal
system may
be called the
"polar
axis
"
of the
system.
Let
a"
be the
tangent plane
to o
at the
point17,
apd
let
g
be
any
straight
line in this
plane.
Then the coaxal
system
of
which
g
is the
polar
axis
projects
into a
system
of circles
having
their radical axis at
mfinity;
that
is,
a
concentric
system.
Hence
we
get
a
construction
for what
may
be called the
"
centroid
"
of
any
conic drawn
upon
8.
Namely,produce
the
plane
of the
conic
to meet
cd
in
a
straight
line
g ;
then the
point
of contact
of the other
tangentplane
to
8^
which
can
be drawn
through
^,
is the
centroid in
question.
Or
again,
join
U to the
pole
of the section
;
this line will meet 8 in the
required
centroid.
The
centroids of
a
coaxal
system
lie
on a
conic
passing
through
U,
152
PROF.
MATHEWS,
GEOMETBY ON
A
QUADRIC
SURFACE.
Let
g
be
the
polar
axis of
a
coaxal
system; then,
if
g'
be
conjugate
to
g
with
regard
to
/S,
the coaxal
system
of which
g'
IS
the
polar
axis
may
be called
conjugate
to the other
system.
The centroids of each
system
he
on
the radical axis
of the other.
Any
two
conies
upon
8 determine two
points corresponding
to the
centres of similitude of the circlesinto which the conies
project.
They
may
be found
as
follows. Find
the centroids
of the two conies and
through
thfem draw
two
planes
inter- secting
0) (the
tangentplane
to 8 at
TJ)
in the
same straight
line.
Suppose
these
planes
meet the
given
conies in
A^
B and
Cy
D
respectively
: then,
if the conies UA
(7,
UBD meet in 0
and
VAD^
UBC
m 0\ 0,
0'
are
the
points
required.
It is needless
to
go
into further
detail,
for itisevident that
every
known theorem in the
geometry
of the
straight
line
and circle
givesa
corresponding
theorem in the
geometry
of
conies
upon
8. For
example,
two conies P and
Q
upon
8
may
be such that
a
finite number of other conies
may
be
drawn
upon
8,
each
touching
P and
Q
and
two
adjacent
conies of the
series;
and this
can
be
done,
if at
all,
in
an
infinitenumber of
ways,
"c.
Or
again,
as an
example
of metrical
theorems,
let
Aj
B be
any
two
fixed
points
upon
/S,
and U
an
umbilic. Draw two
conies
upon
8
passingthrough17,
A and
U^
B
respectively,
and
intersecting
at a
givenangle
in
U]
then the locus of
their other
point
of intersectionis
a
conic
passingthrough
A and B.
Again,
to a
conic in the
plane
of
projection corresponds
upon
8
a
quartic having
a
conjugate point
at
Z7;
and
we
have
upon
8
a
kind of
projective geometry
by
which
any
such
curve
may
be derived from
a
plane
section of
8^
and its
properties
investigated.
EXPKESSIONS FOR
0(a:)
AS A DEFINITE
INTEGRAL.
By
J. W. L. Olaisher.
In Vol.
V.
(1876),
p.
173 of the
Messenger^
I
gave
without
proof
an
expression
for 0
(x)as a
definite
integral.
This
expression
contained several
errors
which
were
corrected
in a
paper
in Vol. ill.
(1887),
pp.
61-66 of the
Proceedings
of the
i3ambridge Philosophical Society.
This
paper
contained
six
expressions
for the function 0
(a:),
with
an
explanation
of the
(
154
)
HOMOGRAPHIO INVARIANTS AND
QUOTIENT
DERIVATIVES.
By
A. J2.
Forayth.
1. The
investigations
contained in the
present
paper
were
originally begun
with the
purpose
of
finding
the
relation,
in
which
a
class of functions called
quotient
derivatives stand
to
reciprocants.
These
quotient
derivatives
have,
among
other
properties,
that of
being
covariantive for
homographic
trans- formations
of the
dependent
and the
independent
variables,
when
applied simultaneously
and therefore also when
applied
separately.
But it is evident from their forms that their
aggregate
does
not constitute the
complete
series of such
functions
;
and
my
firstaim has been to obtain these
complete
series for each of the combinations of
homographic
trans- formation.
The character of the invariance isless
general
than
that of
M.
Halphen's
Differential
Invariants,
which
reproduce
them- selves,
save as to a
factor, by
the substitutions
ax +
by
+ c
a'x
+
b'y-^c'
a"aj
+
6"y+
c"*
Functions of the kinds herein considered have been
previously
suggested by
Mr. L. J.
Rogers,*
who
has,
except
in the
case
of the first
kind,
limited his
investigations
to the deduction
of
the
partial
differential
equations
which
are
satisfied
by
the
functions. His aim
was
the derivation of
homographic
reciprocants.
It is
by
a
comparison
of the
quotient
derivativeswith these
homographicreciprocants
that the desired relation has been
obtained. The cubic derivative has been
expressed
in terms
of
them,
but the combination is not
legitimatef
for the
preservation
of
reciprocal
invariance.
The
relation thus
suggested
is
proved
to be
general.
2.
Perhaps
the
simplest
way
of
obtaining
the
quotient
derivativeais
as
follows. Without
reproducing
the
general
investigation
in which
they arise,
consider for
example
a
"
"Homographic
and Circular
Reciprocants"(first paper),
Proc. Land, MaJth.
Soe.,
vol.
XVII. (1886),
pp.
^20"231,
t Sylvester, "Lectures on the
Theory
of
Bedprocants,''
Amer. Joum,
MaHh,^
vol. viii.
(1886), p.
212.
MB.
FOESTTH,
HOMOGBAPHIC INVABIANT8.
155
cubic
equation
of wbich the
primitive
is
and let
y
be the
quotient
of two differentsolutions
w,
and
w,
of this
equation,
so
that
Since the third and
every higher
derivative of
u
vanish,
we
have
"y"+3/v+ 3y"' =0,
and therefore
eliminating
w^,
m^',
w ",
which
are
linearly
independent
of
one another,
we nave
(now indicating
differentiation
by subscript integers)
\j/i^l^
= 0.
!/sy3y" By,
This is a* differential
equation
of the fifth
order;
its
primitive
is
_Aj^Bx+^
y'^D-\-Ex
+
Fx''
The function
on
the left-hand side is called the cubic
quotient
derivative.
The
quadratic equation
--p^
= 0 leads
similarly
to the
well-known Schwarzian
the
qnarticequation
^4
= 0 leads to a
similarly
formed
quartlc quotient
derivatire
[y""]i'
Vi,
4y." %,. ^Vi
Va
5^4*10y" lOy,
156
MR.
FORSYTH,
HOMOORAPHIC INVARIANTS
and
60 on.
And the
property
of
homographic
invarlance
already
referredto is constituted
by
the
equation
Some other
properties
will be obtained
later,
one
of them
in particular ("45)shewingwhy
in these derivatives the
highest
differential coefficientsof
y
which
occur are
only
those defined
by
alternate
integers.
Invariants when the
independent
variableis
subject
to hom"H
graphictransformation.
3. Let
"f" (y,x)
be
a
function
which,
when the
independent
variablex
is
transformed, reproduces itself,
save as to
a
power
of
-T"
,
BO
that
we
may
write
The
exponent
m
iscalledthe index of the Invariant.
4. Then the
following properties
of the
general
forms of
such functionsare
easily
derived
if
(i).
The
independent
variable does
not
eocplicitly occur.
For a: = " -f
c,
where
c
is
an arbitrary
constant,
is
a
possible
transformation
;
if
x occurs on
the left-handside of the above
invariant
equation,
there
will,
after the substitution for
a?,
arise
a term or
terms
in
c,
which do
not
occur on
the
right-
hand side.
(ii).
The irreducibleinvariantsdo not
explicitly
contain the
dependent
variable.
For,
if
a
given
invariant
^
contain
y,
it
can
be
arranged
in
powers
of
y
in the form
If the transformed value of
(f"
be
4",
then
we
may
write
*
Proc.
Royal 8oe^
12th
Jan.,
1888.
t
These
oorrespond
to the
propositionB girenby Halphen,
Thdse
^
Sur Ic*
inTariantft
diff6rentielB,"
p.
21.
AND
QUOTIENT DERIVATIVES..
157
and since
^
is
an invariant,
we
have
Henoe
*.-(|)"t.+y{*,- (|)"*,}
+....0.
Now
"f"^
and
"I"^
do not
explicitly
contain
y,
which is
subject
to no
variation
;
hence
Similarly
*t-(^)
^i
=
^"
and
so
on;
from which
it
follows that
"^^, 0" ^^j"""
^^"
invariants of index
m.
And
they are
all
explicitly
tree from
the
dependent
variable.
(iii)
Invariants
are
ofuniform
grade^
that
is,
the sum
of
the orders of diflferentiation of the
dependent
variable in the
factors of
any
term of an
invariant is the
same
for all the
terms of that invariant. For
a
possible
transformation
is
z ="
axj
where
a
is
a constant
;
and then
dx dz^
so
that the effect of the transformation
on
any
term is to
multiply
it
by
a
power
of
a
equal
to the
grade
of the
term.
But the effect of this transformation on
the invariant is to
multiply
it
by
a"*,
whence the result.
From this itat once
follows that the index
of
an
invariant
of
the class at 'present
considered is
equal
to
its
grade,
(iv).
Irreducibleinvariants
are homogeneous
in the
diffe- rential
coefficients of
the
dependent
variable. For let
an
invariant
^
of index
m
be
arranged
in the form
where
"l)p
is the
aggregate
of terms
in
yfr
of the
degree
p ;
then we
have
Now
a
change
of the
independent
variable in
any
differential coefficient of
y
gives
a
quantity
which is linearin
158
MB.
FOBSTTH,
HOMOQBAPHIC INYABIANTS
the
new
differentialcoefficients
;
hence the
degree
of
a
term
in the invariant is unaltered. The last
equation
therefore
shews that the
aggregate
of
terms
of
any
degree
transforms
into
a
corresponding aggregate
of the
same
degree,
and that
this
aggregate
is
an
invariant.
It
IS
evident that allthese results hold for
any
transforma- tion
of the
independent
variable.
5. Now for
any
transformation the
general
law of
differentiation is
where
Cr,
J
= coefficientof
p'
in
]p^i
+
^
p\
+
q-j
p\
+"""
["
J
dsi d'z
and
z,
=
-^,z,
=
^,....
One method of
obtaining
the characteristicdifferential
equations,
satisfied
by
the
invariants,
is
to
consider the effect
on
the invariant
equation
consequent
on an
arbitrary
infinitesimal
change
in the
independent
variable. Such
a
change
may
be taken in the form
where
8
is
an
infinitesimaland
constant^
and
/i
is
an arbitrary
function of
a;
;
we
then have
and,
for
r "
1,
Hence
we
deduce
Cr.r=l+
re;A"
and,
for5 "
r,
n
"
^
and therefore
d' d'
_
"f r! d-' r!
.
(T*
rl d
-"^'^^OlVld.
"r rl d^' d
when
quantities
of
only
the firstorder
are
retained.
AND QUOTIENT
DERIVATIVES.
159
6. In the
case
of
a homographictransformatioD,
say
z =
so
that
i2?,
=
^05+
A'
^
eh-fg
{gx^hj
2 9
it
is
sufficient,
for reduction
to what
precedes,
to take
e
=
A,
y
=
0,
(jr
= "
^e
J
and then
"^
= 1
+
so;,
so
that
fi^
=
x^
and
higher
differentialcoefficientsof
ii
vanish. The
pre- ceding
general
formula
now
becomes
d d d
^
.
^
d^'
Applying
this
to
the
equation
=
(1+W2sa;)0(y,
"),
we
have
on substituting
for the differentialcoefficientsof
y
on
the
left-handside
an
equation
=
(l
+
iwsaj)"^(y, "),
where in terms
multiplied by
s we
may
take
" or a?
indifferently
as
the variable. Since this
equation
is
to
be
identically
true,
we
have
an index-equation,
or
say
the
grade-equation
;
and
2r(r-l)y..g"0,
a
form-equaiion.
When the form of the function isdetermined
by
the latter
and the index is inferred from the
form,
then the
grade
equation
is
identically
satisfied.
160
MR.
FORSYTH,
HOMOGRAPHIC INVARIANTS
7. These
equations
coincide with the
equations
otherwise
obtained
by
Mr.
Rogers (I.e.); and,
as
he has
given
a
succession of functions
satisfying them,
it is not
proposed
here
to do
more
than
merely
to state results.
An
important
proposition,
which admits of
easy
proof
and
willbe
subsequently used,
is the
following
:
If ^
he an invariant
ofgrade-index
"n,
then
y^
^
"
fny/f"
is an invariant
ofgrade-index
m +
2.
For
tM_thi$^
\dx) \dz)
so that
dx\y:)-^^dz\/d^^\'
or
-J-
[4^y^)
is
an
invariantof
grade-
mdex
unity,
and there- fore
y, -7
"
wiy,"^
is
an
invariantof
grade-index
w +
2.
It is
by
the
use
of this
proposition
that Mr.
Bogers
obtains his succession of educed
functions, beginning
with the
customary
Schwarzian
y,y,"
|y,*;
from and after the second
educt,
however,simpler
forms
can
be
given
as
follows.
8. The functionof the first
degree
is
The functionsof the second
degreeare :
f.^Vxy.
-
^^y.y,
+
losy.y.
-
^^y:^
and,generally,
-
J^[
2jg
+ l!2jg! )
"^''^""^r'^'^"2^"'-'^"22?--5-fll2p~"!5+lUlJ*
where
e,
=
i(-l)'j
and for
*s=0,1,2,
...,p-
1,
CONTENTS.
PAGE
Homographic
invariants and
quotient
derivatives
(continned). By
A. R.
Forsyth -
- - - - - - - 161
The
following
papers
have been received
:
Hr.
Buchhoim,
"
Note on matrices in involution."
Prof.
Cayley, "Analytical
formulse in
regard
to an
octad of
points/ "A
correspondence
of confocal Cartesians with the
right
lines of
a hyperboloid/'
*'
Note on
the relation between the distances of five
points
in
space."
Articles for insertion will be received
by
the
Editor, or by
Messrs.
Metcalfe and
Son, PrintingOffice,
TrinityStreet,Cambridge.
ffSTS-
No.
CCIIL]
NEW SERIES.
[March,
1888
THE
MESSENGER OP
MATHEMATICS.
EDITED BY
J. W. L.
GLAISHEE, So.D.,F.E.S.,
FELLOW OF TRINITY
COLLEGE,
CAMBRIDGE.
VOL. XVII." NO, 11.
MAOMILLAN
AND CO.
Hon"on
an" Cambttlnge*
1888.
Price" One
Shilling.
'SiSiSX
MTCTflALPE AND SON. CAMBRIDGE
"tSS^