Analpticae.
mM
Gud
=
=
ea
eS
>
\ ened
4s
WILLIAM SPOTTISWOODE,
OXFORD,
GE,
G
COLLE
BALLIOL
me
»
= re Ma
: ae pt ae
j\ ‘ tal
190 tes
Taal
™~
-
Le ak i
THE UNIVERSITY -
OF ILLINOIS
LIBRARY
5/0
510-4
Sp6I me
MATHENATIOS
DEPArYesgyy
|
The person charging this material js re-
sponsible for its return to the library from
which it was withdrawn on or before the
Latest Date stamped below.
Theft, mutilation, and underlining of books
are reasons for disciplinary action and may
result in dismissal from the University.
UNIVERSITY OF ILLINOIS LIBRARY AT URBANA-CHAMPAIGN
OCT 14 i976
L161—0-1096
=
_” 5
b
4 f —leditattones = Analpticar.
: .
By
WILLIAM SPOTTISWOODE,
OF BALLIOL COLLEGE, OXFORD.
LONDON: ee | eu
1847. : Leg Ae ae
TO THOSE
ee WHO )
LOVE TO WANDER ON THE SHORE
TILL THE DAY
WHEN THEIR EYES SHALL BE OPENED
AND THEY SHALL SEE CLEARLY
“THE WORKS OF GOD |
IN THE UNFATHOMED OCEAN OF TRUTH, ;
These Papers
ARE INSCRIBED.
ek ; 7
SEND te a Se ’ > ae
PREFACE.
Tue following Papers have been written at various periods, as
the subjects presented themselves to notice from time to time.
If leisure had been afforded, an attempt would have been made
to draw some of them up into a distinct treatise; but it was
thought that even in their present form they might interest
some of those who take pleasure in the pursuit of Mathematical
Science. Some of the papers are entirely original; some are
partly taken from foreign Memoires, and these chiefly from the
writings of M. Cauchy.
In the solution of the problems presented to notice analysis
has been almost invariably employed ; the comprehensiveness
and uniformity of that method being sufficient apology for
the exclusion of geometrical proofs. In the form of the
equations symmetry has been preserved, wherever the circum-
stances of the case would permit ; for although such expressions
are sometimes longer than the corresponding unsymmetrical
ones, yet being more readily committed to memory, more
expressive of their meaning, and to one familiar with them
more easily applicable, they have been thought worthy of
attention.
Vi
The following theorems, on account of their frequent
occurrence in the following pages, are transcribed from
Gregory’s Solid Geometry.
Teas ee ae
i Favib-dh bia
each of these ratios is equal to
Ca ea &c.)?
(0? + b2 + b2 + &e.)'
na+ na, + na, + &e.
NO 10 Oe ee
and to
n, 2, N,, &c. being any quantities whatever.
For assuming each of the ratios equal to 7, we have
a=rb, a,=rb,, a,=rb,, &c.
Squaring and adding, |
Gt, bapsh &C. 920 th ber Oat OCC
whence, extracting the root and dividing,
CSO, als te) ey
CREE ee prey b yao:
Again, na=rnb, na,=rnb, n,a,=7n,b,, &e.
By addition,
(na + na, + n,a, + &c.) = r(nb + n,b, + 0,6, + &c.)
Wh na+ na, + na, + &c. LL eth
ee nb +n,b, + nb, + &e. ~~ b= b=
II. If we wish to determine the variables from three simultaneous
equations of the form
AE OY Nee 10 ee, es ee (1),
LUTE YC — Cte eae ie (2),
O06 Me Za, sae te. (3),
instead of eliminating first z and then y, in order to determine x, we
may eliminate both at one operation by the following rule: Multiply
Vil
(1) by dc, —¢,6,; (2) by cb, — be,; (3) by bc, — 6c, and add: it will
be found that the coefficients of y and z are identically equal to zero,
and we have
_ d(b,c, — ¢,b,) + d,(cb, — bc,) + d,(bc,— b,c)
Tinae bc,— €,b,) + a,(cb,— be,) + a,(be, — b,c)’
with similar expressions for the other variables. If d=0,d,=0, d,=0, the
equations contain only two independent variables, since we may divide
all by any one of them; and the condition that the equations should
coexist is
a(b,¢,—c,b,) +a,(cb,— bc,) +a,(bc,— b,c) =0.
We shall frequently refer to this process under the name of cross
multiplication ; and the student is recommended to make himself
familiar with the forms of the multipliers, as a ready use of the process
will be of great service to him.
III. The sum of three squares of the form
(on bayer (em az) + (bz cy)"
may be put i a shape which is very convenient, especially in geometrical
investigations. or if we add and substract from the preceding expres-
sion the three squares |
ine by’, Gee
the expression may be transformed into
aera. tty t 2-) (art bY ez )*,
‘ ‘ « 9 9 9 OV + ) e ae Z 1
or (a + b?+ ca (Ger + Ge + Zz") / l = ve ae Gz) Toa? ON I
cdr ot Ca) (ay ey? eae)
Now, if a, 6, ¢ be taken as proportional to the direction-cosines of
some one line, and 2, y, z of another, the expression
merge eg ea ee
(@+B+e(e2@tyt+2)
is equal to the cosine of the angle between the lines: let this be 6; then
the sum of the squares is equal to
Came Ge a (tates Yat z*) (sin $)’.
Vill
IV. If we obtain as the result of any process that a function of wv is
equal to a function of y in which y is involved in a manner similar to that
in which « is involved in the other, then, as there is nothing to distinguish
one co-ordinate from another when they are symmetrically involved, we
may say that each of these functions is equal to a similar function of z,
and this is the consequence of the general symmetry of our expressions.
Thus, if we have two equations
la + my + nz =0,
Vx + m’y + n'z=0,
and eliminate z between them, we find
(In' — Un) « + (mn' — m'n) y=0,
ay
or = ns
mn —-mn ln—-In’
here the two sides of the equation are symmetrical, one with respect to
x and the other to y. We may therefore say that each is equal to
ior this being the corresponding symmetrical expression with
respect to z.
Balliol College, Jan. 1, 1847.
MEDITATIONES ANALYTIC.
Symmetrical Investigations of Formule relative to Plane Triangles.
In any plane triangle we have the relations
sin A = sin B = sin C (1)
a b c
__ sin Ccos B+sin Bcos C_ sin Acos C+sin C cos A _ sin Bcos A +sin A cos B (2)
~ ecosB+bcosC acosC +eccosA bcos A + acos B i
sin (B + C) i sin (C + A) mn sin (A + B) (3)
~ ecosB + bcos C acosC +ccosA beosA + acosB
at sin A Pe sin B at sin C (4)
~ ecosB+b6cosC acosC +ccosA bcosA +acosB
whence
Sree vec Cree cod Chae cos | A. 28 oa A) S cos B (5)
a a b b of c
whence
(Ee a ee cos B _ cosC
a c b
b a c
eae
( DR: i
c b a
|
Raa erect cor 8 t (6)
|
whence, multiplying by
be, V.caser ab,
2
respectively, there result
b? — ¢
a
cos A = bcos Saad
ae ie
; cos B= ¢eos © — a cos A
2_ 72
# PeosC = acon A — beosB |
c
whence
(b?— c*) cos *A + (c?— a?) cos 7B + (a*— 8°) cos 7C = 0
and
2 2 2 Eee 2_ #2
da cos A + © — cos B + 2% x cos C = 0
and
cos A cos B cos C
a(b? + c’— a’) i b(c? + a*— b*) Ma c(a* + 6? — c?)
also, since
sin AcosA _ sin Bcos B _ sin C cosC
a.cos A). | 4h eos Bde be cose”
“(osu 2AY asin 2Ber ein 2
QacosA 2bcosB 2c cosC
consequently
Bud
a
a? — b?
c
' c—a?.. 7
sin 2A + ~ sin 2B + sin 2C = 0
b?
ae+e—a) (e+a—b) eat e—e)
also by geometry each of the ratios (10) is
il
~~ Yabe
hence
cosA , cosB , cosC_ a+ +e?
a b 1 cies Qabe
») (F,2,.2 2 272) (4 BA at
a cos A 408 cob B Ac cos Ce cere Cae a) a Cah ace!
Qabe
be , ca a) eye Oe
=i eg) ca atin aT
and similarly other formule might be obtained,
(10)
(11)
(12)
(13)
(14)
(15)
(16)
Again, writing
b?74 2&—a?= 2 C+ &— b= p? a+ B—c?= y?
we find by means of (1) and (10)
tan A = p? tan B = v* tan C
also, since in every plane triangle
tan A + tan B + tan C = tan A tan B tanC
vatan Age. tat) Db
—<<_ ——— ee EL
1 l 1 1 1 1
” Fe F ee
_ tan Atan BtanC _ 2? tanA yp? tan By” tan C
ThA LAE PRT he exe at
yang pov + Ur +p
ne be Vv
r® tan 3A a Hw tan °B if ¥ tan °C
a wv? + Pr? + r2u? ai wv? + vr? +? oo pe? + Pr? + 2?
“. tan A = p* tan B = v* tan C =o (wv? + vd? +72n7)
also from (10)
Chst Ate COa bee a COstGu aL
an bp cP abe
sn A sinB _ snC _ W(p?v? +0? +22?)
Bape c es oats cy Fre Qabe
and if we put
b+.¢
a ee
2
we shall find by actual multiplication that
pev? + vr2+ A2u?= 16S(S — a) (S—3d) (S—c)
and the above will agree with the usual formule.
(18)
(19)
(20)
(21)
(25)
On some Theorems relative to Sections of Surfaces of the Second
Order.
I.—oON THE SECTIONS OF THE CONE.*
Tue plane sections of a surface are given by combining the equation
to the surface
F (2yz) = 0 (1)
with the equation to the plane, which will be
e+ py + vz=0 (2)
or
Ma—a) + wly—B) + (2-7) = 0 (3)
if it passes through a point « 6 y.
Instead however of finding the nature of the section by a trans-
formation of coordinates, a process necessarily long and tedious, we will
make use of the distance of a given point from the surface, as that
quantity is independent of the coordinate axes.
Let the equations to a line passing through the point « 6 y be
ges 3 oie ashes. (4)
Then if (4) les in (3) we have the condition
M+ pm + un =0 (5)
Combining (4) and (5) with (1) we shall find the equation to the
plane section in terms of all the values of r in it.
Before proceeding, we will first determine how curves of the second
order, in plano, are distinguished. ‘The general equation of the second
order between two variables is
Ax’ + 2Bry + Cy?+ 2Dzr + 2Ey + F=0 (6)
and this represents an
ellipse <0)
parabola | scorn as B? — AC is | 0
hyperbola > 0
* The first part of this section will be found in Gregory’s Solid Geometry ; it has been
inserted in order to make the latter part intelligible.
5
and if we substitute for 7 and y in terms of r from the equations
z—a_ yf _
age Tata (7)
we find
(AP + 2Blm + Cm?)r? +..... = 0 (8)
and the discriminating condition is equivalent to. saying that the curve
is an ellipse, parabola, or hyperbola, according as the coefficient of 7”,
equated to zero, leads to impossible, equal, or possible values of the
ratio /: mor m: /.
We shall find this condition equally applicable in three dimensions.
The equation to the cone is
Pz? + Py? — P’? =0 (9)
whence
Ce a te ee ere. a ole at (10)
and the discriminating condition depends upon the nature of the
expression
Pi? + P’m? — P”n? = 0 (11)
combined with
Al + pm + m = 0 (12)
eliminating 7 between these we have
(Po? — P/r™2)2 + (Pe? — P"u2)m? — 2P”rplm = 0 (13)
and the discriminating condition becomes
P2442 — (Py? — P’n2) (Pv? — P’p?)
or
~— PP? + P’P/22 + PP2y2
or
(PP? + P’Pu? — PP'r*) (14)
and the nature of the conic section will depend on the sign of the
quantity under the bracket.
Suppose the cone to be a right one on a circular base
a be (15)
and the condition becomes
y{PP”(r2 + w2) — Prt
and since
Ve weal (16)
we have
Py {P”(1 — v’) — Pr}
or
P24 Pi (Pop By}
*, we have an
ellipse > p”
parabola {sont as v “| Ss Pap”
hyperbola
Now, since the equation to the cone is
Pe ap ye ee 0
we have
pp”
‘Be
where COB = semivertical angle of the cone.
And » = cosine of 7 between a normal to the plane and the axis of
x =sin OED, where OED = angle between the axis of the cone, and
the cutting plane.
Hence we have an
tan COB =
ellipse > :
parabola facsoncng as sin ‘own = ne
hyperbola < see (COB
>
1e.as Z on =| Z COB
“
II.._oN THE PRINCIPAL AXES OF SECTIONS OF SURFACES OF THE SECOND
ORDER.*
Ler the surface be a central one, then its equation will be
Brel yi ke eee (1)
and the equations to a line in the plane of section will be
Pete Gd Ne ee 5
l ae ea (2)
and the condition that it lies m the plane, the direction cosines of a
normal to which are
Xs My DY,
will be
M+ pm + vn = 0 ))
from (2) we find
emE+dr y=n+mr GG 4 nr (A)
* The following investigations were partly communicated to me by the Rev. B. Price,
of Pembroke College.
~y
‘, substituting in (1)
PE + Py? + P’S? + 2(PE + P’nm + eat
+ (PP + P’m? + P’n?)r? = H
then in order to find the principal axes of the section, we must make r
pass through the centre, or have the two values of r equal and of
opposite signs; and then find 7? a maximum or minimum. Hence we
(9)
have
PlE + P’mn + P’nt = 0 (6)
and
Rai Eee big ech Sy) (7)
Pl + Pl? + Pn
Now, in order that 7 may be a maximum or minimum, it will be
sufficient that the differential coefficient of the denominator of this
fraction be put equal to zero; /, m, n, being subject to the condition
P4+m+n=1 (8)
Our object now is to eliminate /m mn from the equations (3), (7),
and (8). Hence differentiating as above mentioned, we have
Pidl + P’mdm + P”ndn = 0
Adl + udm + vdrn = 0 (9)
ldl + mdm + ndn = 0
whence, introducing two indeterminate multipliers +, +, we have
Pl+orx.—7l1=0
Pin ten — rm = 0 (10)
P’"n + ov — m=0
multiplying by / m n respectively, and adding, we find
PP + Pm? + PP’? = 7
or by (5)
1 Flies GPE opm + Pe’) (11)
or writing
Da Pear agra ty Co == (12)
hence (10) become
whence
on om ov
—l= 3 = 9 — ’
ve Te
whence, multiplying by « » » respectively, and adding, we find by
means of (3)
r2 pe yp
+ TE =0
pro zs pi Q p” Q (15)
r* 7?
from this equation we may easily show that the two greatest and least
values of r are at right angles to each other, for let 7’ 7’ be two roots
of this equation, then writing down the equation for each of these values
of r, and subtracting, we find,
QQ C03) Pry Py] m=
CHEE CDE F-DEDis
or by (14), if 7 is different from
1 2
lL +mm,+nn = 0 (17)
i.e. the two lines are perpendicular.
Also clearing (15) of fractions we have
(7-8) (-8)w+ (m8) (Pye (P-9) (PB) roo as
whence
(P°P’A? + P’Py? + PP’r’)r? \ th
—2Q{(P + PY) (PO P)pA ep (Pie Eye er ea
whence, if a 6 be the principal semiaxes of the section, we have by the
theory of equations
ath? = Q?
PepPAte ee Pai PPy4
or
_ H— (Pe 4 Pat 4+ P'C
w= CprpTye + P”PiE + PP8)* 2)
and
2 i ia (P’ + P”)a? + (PZ a P)u? “ (Pe als P)y? (21)
PP? + P’Pp? + PP?
In the case of the cone
ne 0 (22)
and one of the coefficients as P’’ is negative; hence we find real finite
values for a and 6 only when
VPS is Se PEPS + 2 P7P
i.e. only in the case of the ellipse (by a former paper). In the case of
the parabola they become infinite, by the same paper. In that of the
hyperbola they or one of them becomes imaginary.
Also the area of an elliptic section
= mab
Pé? + Pn? — P’C?
= cranes meme a pe y 1g (23)
/PD// jue eT eee 8S Se
ire pent i ae ees Eee
JII.—Ir a sPHERE BE INSCRIBED IN A CONE, AND A SECTION OF THE CONE
_ BE MADE BY A PLANE TOUCHING THE SPHERE, THE POINT OF CONTACT OF
THE PLANE AND SPHERE WILL BE A FOCUS OF THE SECTION.
Ler the equation to the sphere be
Ded sied Yorks 2am (1)
then the curve of contact is given by the equation
ax + by+cz=7 (2)
where abc are the coordinates of the vertex of the cone. Let.é 1% be
the current coordinates of the cone, then the equation to the generator
may be written
ae a Ee ee (3)
subtracting (1) from (2) we have
x(a—x) + y(b—y) + 2(e—z) = 0 (4)
whence, by means of (3)
a(E—2x) + y(n—y) + 2($—z) = 0 (5)
or
Extay+ @=7r* (6)
whence (3)
wee Oo eee Te oT (7)
@+eP+e—r £at+nbt+&e—r
C
10
whence
(fa+ n+ &e—ryP= (24+ C—r\(a? +0? 4+ —7°) (8)
which is the equation to the cone enveloping the sphere (1).
Now if we take the plane of the section for that of € y
f=-—r (9)
hence the equation to the conic section is
{fa + nb + r(c—r)}? = (& +n") (a? +B? +c? —7*) (10)
i.e. the radius vector
p=V(E +7) (11)
is a rational function of the coordinates; i.e. the origin is the focus.
Similarly the point where the plane of section is touched by another
sphere on its other side will be the other focus.
The equation to the enveloping cone may be put in a remarkable
form as follows ; multiplying out we find
E’a® + °b? + bc? — Qr*(Ea + nb + &)
+ 2(nfbe + SEca + Enab)
— ea + b? am Ce 7) + n? (a? +6? + c?— 7?) + C(a?+b2+ ee 7 (12)
— 72(a2 +B? + c2) + rt
or
(P+ C—P)E + (2 +a? =r)? + (P+ BR — 7°)? \ ds)
2(be&m + caf E+ abEn) + 27r?(aE +bn +cf) — r*(a? + a c) = 10
or
{en — bh)? (ab—e8)" 4 Means 2 rye yt bt ee (14)
r r r
a convenient form for finding the distance of any point on the cone from
the vertex.
pd
On the Reduction of the general Equation of the Second Order.
].—ON SURFACES OF THE SECOND ORDER.
Tuer general equation of the second order between three variables is
Ax? + A’y? + A”2’ + 2Byz + 2B’zr + 2B’zy + Cx + Cy + C’2+E=0 (1)
but since this in its present form contains so many terms, it will be worth
while to investigate the possibility of simplifying the expression by a
change of coordinates.
Let the direction of the coordinate axes be changed by the formule
emaktant+al’t y=bE+ ont zackientec’t (2)
if then the system be still supposed rectangular, the nine direction
cosines will be subject to the six conditions
ata + a= 1 P+ 674+57=1 C4+ceeA4+ce"7=1 (3)
be + De’ + Oc’ = 0 ca + ca’ + c’a"=0 ab + a’l’ + a”’b’=0 (4)
whence also the inverse system
@+R + 2=1 ateh2¢c=1 a? +2 4ce7=1 (5)
aa” + Ub" + cc” = 0 a”a + bYb + c’c =0 aa’ + bb’ + c’=0 (6)
may be easily deduced. i
On the introduction of the new variables £4 ¢ by means of (2), the
equation (1) takes the form
AE + Bn? + AC? + QB nt + QHB'CE + 298"En + OE + On + O"$+E=0 (7)
where
A = Aa? + A’b? + A”c? + 2Bbc+2B’ca+2B"ab, A’=...., @’=.... (8)
95 = Aa’a”’ + A’b/'b" + A’c'c’ + B(b’c” +.0"c’) + Bea” + c/a’) + BY’ (a’'b" +00’)
98’ = Aa’a + A'S") + Alec + B(b"c+be") + Bi(c’a+ca”) + B’(a"b +ab") (9)
45” = Aaa’ + A’bb! + A’cc’ + Bibe’ + b’c) + B(ca’ + ca) + B’(ab’ + ab)
@=Ca+Cb4+C%c, G/ =Ca’ + C+ C’e’, EC” =Ca” + Cb” + Cc” (10)
Since however the nine quantities
Ay Dy Cs a bata Be Ci
are subject to only the six relations (3) and (4), or (5) and (6), it is
allowable to assume three more conditions respecting them ; let those be
B=0 W=0 B”=0 (11)
Cc 2
12
When these relations are satisfied the coordinate axes will be callec
principal axes.
It now remains to be proved that when the equations (11) are satis-
fied the quantities @’@’Q” are real, i.e. that the conditions (11) are
possible ; in order to do this, let the last two of (11) be written thus
(Aa + B’ce+ B’b)a’ + (A+ B’a+ Be)d’ + (A”ce+ Bb+4+ B’a)c’=0 \ (12)
(Aa+ B’e+Bb)a” + (Ab + B’a+Be)b” + (ANe + Bb +Bia)e” =0
whence, by symmetrical eliminations,
Aa+ Be+B% _ Ab + Bla + Be _ A’e + Bb + Ba
b’c” — bc’ ca” — cla’ ab” — ay’
But by similar eliminations from the last two of (4) or of (6) there
result
(13)
oe See ee (14)
hence (13) may be written
Aa + Be + B”) rm A’b + B”’a + Be a A”c + Bb + Bla a (15)
a b c
with similar expressions for @’@’’. Now (15) may also be written
(A—A)a + B’b + Ble = 0
B’a + (A’—A)b + Be = 0 (16)
Bia + Bh + (A”—Aye = 0
whence, by cross multiplication,
(A—AV\A—A'(A—A”)—B(A— A) — B(A—A)—B’(A—A”)—2BB’B’=0 (17)
The same equation would evidently have been deduced if instead of
(15) we had taken the analogous system involving a’b’c’Q’, or that
involving a’b"c"’@"". It may thus be shown that the three roots of (17)
are all real; the equation (17) and the two others, which involve @’, @”’
respectively, being all cubics, must each have one real root at least.
If however any two of these roots were equal, there would result
(A—Aja’ + BY’ + Bl’ =0 \ ag)
(A—Q)a” 4 B’b” 4 Be’ = 0
whence by means of (14) there may be deduced the system
and similarly, by the assistance of (4) or (6), the analogous systems.
mo A’—-G B B’ B A”—-QG
Rs i 7 a’ a p” oe Toma, (20)
13
whence by means of (6)
B’(A—@) + B”(A/—@) + BB = 0
and similarly for the other quantities; whence may be deduced the
system
acai,
B’B 4 Ae
sg t Av + A — 28 = 0 t (21)
at A+ A’ - 2@=0 |
or, as they are usually written,
iD, wt a Ae DO i 7 ey DD'y J ’ ie
AS sa’ SPRAY TATA +A" 28 (22)
Consequently, unless the conditions (22) are satisfied, no two of the
roots are equal to one another; but whatever is a root of one of the
cubics is a root of the other two; hence the equation (17) has in general
three real unequal roots.
If none of the quantities AQ'A” vanish, it will evidently always be
possible to destroy the coefficients of the first powers of the variables
by means of removing the origin, without again changing the direction
of the coordinate axes. The equation (1) will then be finally reduced
to the form
Pé + Qr? + RS? = H (23)
It may be observed that in this form of the equation no change is
effected by altering the sign of any one or more of the variables £1 ¢;
from this property the present position of the origin is called the centre
of the surface. If however any one of the quantities @ @’ @’ vanish, the
last transformation is evidently impracticable ; the condition that this
may be the case will be derived from equating @ to zero in (17); there
results
AA’A” — AB? — A/B? — A”B” + 2BB/B” = 0 (24)
this is, therefore, the condition that the surface may be one without a
centre.
It was shown above that (22) were the conditions that (17) should
have two equal roots; when this is the case it is evident that the values
of a’'b'c! would be equal to those of abc respectively ; but as this would
not satisfy (6), the values must be imaginary, and their ratios equal to
the ratios of the values of a 6 c.
14
Hence the result will be of the form
but since by (4) the axis a’b’c must be perpendicular to the plane of the
axes of abe and a’b’c’, the relations (25) represent that, if the axis
a’b’c’ were drawn in a plane perpendicular to its proper plane, it would
coincide with that of abc, which is the usual interpretation of the
symbol “—. It may also be remarked that the first two of (6) will
then be reduced to one equation only, and consequently the plane of
the axis a’’b"c’’ being determined, the direction of it in that plane will
be indeterminate; i.e. it may lie in any direction in that plane. The
surface will in this case be one of revolution, i.e. it may be generated
by the revolution of a plane curve round the axis whose direction is
determinate.
The conditions that (1) may represent a surface of revolution may
also be deduced in the following manner.
Adding and subtracting the three quantities
B’/B” 2 B’B 4? BB’
B iB’ ay "Dy *
(1) takes the form
(a-* Bet + (aS )rt (4-42
(26)
copeyate +) + Cr+ Cy +C'z+E=0
Now, when the surface is cut by a plane perpendicular to the axis of
revolution a circle is formed, and a sphere may always be drawn through
this curve; hence, when the coordinates of the surface satisfy the
equation to the plane they must also satisfy the equation to a sphere.
The conditions that this may take place are evidently
¥Y
a+o+p=0 (27)
and
BBY fe aR De
A‘— ore te Ce =A B’” (28)
hence (13) is the equation to a plane perpendicular to the axis of
revolution, and (14) are the general conditions for a surface of reyo-
lution.
er
It remains for us to determine the position of the principal axes
relatively to the surface ; for this purpose let
Pere Yi Se Ley (29)
be the equations to a principal axis; the substitution of these in the
numerators of (15) gives
sae Se AYR Mca a9)
a
= Bia=2) +Aty=8) + Bem
(30)
= B(x-a) + Bly—-8) + es |
c
which may also be written thus
Ax + B”y+ B’z+C = B’x + A’y + Bz + C’ " B’x + By+ A”z + C” (31)
a b c
if «, B, y be so chosen that
Aa + B’B + By +C = |
B’a + A’B+ By + C= 0
Ba + BB+ A”y + C’=0
From these equations it is easily seen that «, 8, y are the coordinates
of the centre of the surface when referred to the original axes. It is
also observable that in the values of «, 8, y, deduced from these equations,
the common denominator is
AA‘A” — AB? — A/B? — A”B’? + 2BB/B/
and consequently, if the centre of the surface be removed to an infinite
distance, this quantity must vanish, which in fact agrees with the con-
dition (24). Hence the principal axes pass through the centre of the
surface. Moreover, the equation (1) may be written
(Ax+B’y+B/z+C)x + (B’2+ A’y + Bz+C')y+(Ble+By+A’z+C”)+E=0 (33)
and the surface may consequently be considered as the envelope of a
system of a planes, the direction cosines of whose normals are subject
to the conditions (31). Hence when
ae ae yoy ma (34)
i.e. when a principal axis meets the surface, it will be a normal.
Hence there are generally in surfaces of the second order three
principal axes passing through the centre of the surface at right angles
to one another, and coinciding with the normals at the points in which
they meet the surface.
(32)
16
I].—on PLANE CURVES OF THE SECOND ORDER.
Tue above method has the advantage of being immediately applicable
to the equation to plane curves of the second order.
In this case (1) becomes
Az? + A'y* + 2Bay + Cr + Cy + k=O (i)
the transformation of the coordinates is to be effected by means of the
two equations
x=aé+a'n y = bE +04 (2)
and the condition that the new axes shall be rectangular will be |
a@a+a?=1 P+bh?=1
au“+ ab! =0 } 8)
which, as in the case of three dimensions, may be transformed into the
inverse system
a+h=1]1 foe ta)
teats + bb’ cs } (4)
the new coefficients will be given by the equations
@= Aa + A’? + 2Bab
A’ = Aa? + Ab? + 2Ba’d’
93 = Aaa’ + A’bs’ + Blab! +0'b) (9)
C=Ca+ Cb W=Cr + CH
In order then to destroy the coefficient of the product of the variables
we must, as before, have
13 = 0 (6)
which by means of (4) and (5) will give the relations
(Aa + Bb)a’ + (A’b + Ba)b’) = 0
aa’ + bb’ =90 I (7)
the result of which is
Aa + Bb _ A’b + Ba
a b
which may also be written in the form
(A—A)a + Bb= ot
Ba + (A’/-A) =0
the elimination of a and 6 from which produces the quadratic
(A—&)(A’—A) — B? = 0 (10)
at (8)
(9)
17
The condition that the curve may not have a centre will be deducible
from (8) in the same way as in the case of surfaces, by equating to
zero the coefficient of the square of one of the new variables ; this will
give
Aa + Bb=0 A’‘b + Ba = 0 (11)
the elimination of a and 6 from which produces the relation
AA’ — B?=0 (12)
If the two roots of (10) were equal there would result the relations
A@A—-A_B EA a
(15)
b a b a’
which by means of (4) gives
B(@—A) + B(A—A’) = 0 (14)
or combining this with (10)
Bio Bie BAS (15)
This condition may be also deduced in the same way as in the case
of surfaces. For, if the curve represented by (1) is a circle (which is
the case corresponding to surfaces of revolution), on writing (1) in the
form
(A—B)z? + (A’—B)y? + B(a+y)? + Cr + Cy +E=0 (16)
the conditions in question evidently coimcide with (15).
In order to determine the point through which the principal axes are
drawn, and the angle at which they cut the curve, let
x—a _
sali baie a7
be the equation to any principal axis; the combination of this formula
with (8) gives
Arma) + BY A) _ AYA) + Bisa) Aas)
P
which may be written in the form
Ax + By+C _ Bx+Ay+C
b
a
(19)
if « and B be chosen so as to satisfy the equations
Aa+ BB+C=0 Ba + A’‘B+ C=0 (20)
i.e. if « and £ be the coordinates of the centre of the curve.
Again, writing (1) in the form
(Az+ By+C)z + (Br+A’y+C’)jy + E=0 , (21)
and the curve in question may be considered as the envelope of straight
D
18
lines drawn perpendicular to a straight line, whose equation is (18) ;
consequently whenever
Kar yay (22)
i.e. when xy is a point on the curve, or when a principal axis meets
the curve, it is also perpendicular to it.
Hence in plane curves of the second order there are in general two
principal axes passing through the centre of the curve, at right angles
to one another, and coinciding with the normals at the points where
they meet the curve.
It will be worth while here to enter more particularly into the deter-
mination of the quantities a 6, a'b’, «6. For this purpose, multiplying
the equations (9) by 6,a respectively and subtracting, the quantity 4
will be eliminated, and there will result
B(?—a2) + (A—A’)ab = 0 (23)
or writing
a= sind (24)
and consequently instead of the first equation of (4)
sin 70 + cos*0 = | (25)
(23) may be written
. B(cos ?6—sin 70) + (A —A’) sin @ cos @ = 0 (26)
whence
2B
tan 20 = Aran (27)
if the conditions (15) are satisfied the angle § becomes indeterminate,
i.e. any rectangular axes are principal axes.
Again, from (20) the coordinates of the centre of the curve are easily
deduced ; they are
pling Wore i Ae g—BC— Ac
AA’ — B? A A’— B?
and the centre will be removed to an infinite distance if the condition
(1d) is satisfied.
From equation (10) it is evident that the nature of the roots, and
consequently the form of the curve, depends upon the quantity
AA’ — B?
and the curve will be called an ellipse, a parabola, or an hyperbola,
according as the quantity is greater, equal, or less than zero.
(28)
Fo
II].—iInVESTIGATION OF THE VARIOUS KINDS OF SURFACES REPRESENTED BY
THE GENERAL EQUATION OF THE SECOND DEGREE.
It was shown in § I. that, whenever all the roots of the cubic, formed
for finding the coefficients of the squares of the variables, were finite,
the coefficients of the products of the variables might always be
destroyed ; and, moreover, that unless the condition (24) were satisfied,
the coefficients of the first powers of the variables might also be made
to vanish by a suitable choice of the origin of coordinates. When
all these transformations have been effected, the form of the equation
to the surface will be the following,
Px? + Qy? + R2=H (1)
or, writing for convenience,
Hl ae oo
P= 2 3 R= * (2)
ea ai (3)
As however some of the quantities P, Q, R may in certain cases
become negative, the most general form of (3) will be
eS ey (4)
b= *o-
This equation evidently comprises eight others; viz.
One, when of a’b’c’ all are positive.
Three, - - two are positive, and one negative.
SETIEC sm f= - - negative, and one positive.
One, - - all are negative.
These it will be worth while briefly to consider.
(1) Let a?b’c’ be all positive; the equation under consideration will
then be
x2 y? a
a a a ee (5)
When 7 = 0; z= 0 there results 2 = + a
Se UU y=eb (6)
z=0,7=0 =k C
consequently the surface cuts each of the three coordinate axes in two
points, whose distances from the origin are determined by equations (6).
The lengths a, b, c are called the principal axes of the surface.
D 2
20
Again, writing (5) in ae forms
2
oe eo
l (7)
0) CD |
Gre) Hee) Mal
it is evident that, if the surface be cut by planes parallel to the three
coordinate planes respectively, the sections so formed will be all ellipses,
having their centres in the axis perpendicular to which the cutting plane
is drawn; it is also observable that the principal axes of the ellipses
are decreased as the cutting plane recedes from the centre, by quantities
varying directly as the square of the distance of that plane from the
centre. It is also evident from (7) that, when
c= bha y=0 2)
af SE 0 21) =o} (8)
Zi eee Lreats) vi
and consequently that the surface cuts the coordinate axes only in the
points determined by (6) or (8).
Also,
if.2)> ea, i> ee, or itz $+ ¢
the curves (7) become imaginary; and consequently the surface in
question does not extend beyond the limits
rata aim aD ae (9)
this surface is called the ellipsord.
(2) Let one of the quantities a’, b’,c? be negative; the equation
under ate will then be one of the ae
2 22 2 2 2
= MEE Ge | PR eae Nee ite
n+ b Bi Coa i c. as Cie
Taking de first of these equations
When y = 0, z = 0, there results 2? + a? = a
ak (10)
wget Sls 780
me 0,5, 7 = 0, a are
(11)
hence the surface in question never meets the axis of x, but cuts each
of the axes of y and z in two points determined by the equations (11).
21
Again, writing the first cae of the oe (10) in the forms
a) ie 2) my
0-1) C2) ie
a eel
HC) *O=5)
it is evident that, if the surface be cut by planes parallel to the plane
of yz, there will be formed a series of ellipses (as in the case of the
ellipsoid) whose magnitude is extensible ad infinitum, the smallest being
formed in the plane of yz. The sections made by planes parallel to the
plane of zx are hyperbolas, the magnitude of whose principal axes
decrease as the cutting plane recedes from the origin to a distance = +),
when they vanish, and again increase ad infinitum, as the cutting plane
recedes still farther from the origin. It is also observable that the
position of the curve traced on the cutting plane is turned through an
angle of 90°, as the distance of’ that plane from the origin passes through
either of the values + 4. Precisely similar results will be obtained by
considering the sections of the surface made by planes parallel to the
plane of zy. This surface is called the hyperboloid of one sheet. Since
the forms of the surfaces represented by the two remaining equations
of the system (10) are similar to that of the one considered, their
position only being altered, it will be unnecessary to examine them any
further.
(3) If two of the quantities a’, b’, c? be negative, the equation under
consideration will be one of the system
Gini Ss =, ees ee Se eo
ee eek ea ok eee a te is + a 1 (18)
Taking the first of these eee
acest A), ot a Oe y+h=0
z=0, y= 90, 2+c¢ceé=0
When y = 0, z= 0, there results x =+ a
fe
hence the surface in question cuts the axis of x in two points at a
distance of = +a from the origin; but never meets the axes of y or z.
22
Again, writing the first ies of the a (13) in the forms
WH
"G) — 1) e( —- EGaD
MN!
e CD D
a
Gn" 70+) |
it is evident that, if the surface be cut by planes parallel to the plane
of yz, there will be formed a series of ellipses as long as the distance of
the cutting plane from the origin is greater than +a; the ellipse will
vanish when z= +a, and will become imaginary when w is less than +a.
The sections made by planes parallel to the other coordinate planes will
be hyperbolas, the magnitudes of whose principal axes are extensible ad
infinitum. ‘This surface is called the hyperboloid of two sheets. In this
case (as in that of the hyperboloid of one sheet) it will be unnecessary
to examine the remaining equations of the system.
(4) Ifall the quantities a’, 6’,c’ are negative, the locus of the equa-
tion will evidently be entirely imaginary.
If the quantity H, in (1), vanishes, the equation to the surface will
take the form
to
z
c
apigiee Ta (16)
SS.
to
(1) If all the quantities a’, 6°, c’ are positive, the locus of the equation
will be entirely imaginary.
(2) Ifone of the quantities a’, 6’, c’ be negative, and the remaining
two negative, the equation under consideration will be one of the
system
LE ee, = 0 Eh en ee a
Dane a’ TO wc
In all of which the three variables may vanish together, i.e. the
equations
eee y=0 Zo (18)
are consistent with (17). Hence the surface passes through its own
centre.
23
Again, writing the first equation of the system (17) in the forms
2 2 2 a 2
ie os ae or = Weds ae it (19)
a a & b° c a
it is evident that if the surface be cut by planes parallel to the plane
of yz, there will be formed a series of ellipses, the magnitudes of whose
principal axes, varying directly as the square of the distance of the
cutting plane from the origin, may be increased ad infinitum. If the
surface be cut by planes parallel to the remaining coordinate planes,
there will be produced a series of hyperbolas, subject to the same
variations as the ellipses on the planes parallel to the plane of yz.
If the cutting plane be in the plane of yz, then
z
2
r=0and += =0 (20)
which is satisfied only by
=0 z=0 (21)
consequently the trace of the surface on the plane of yz is a point.
If the cutting plane be in the plane of zx, then
2
eared ne 0 (22)
Gay mec
which represents two straight lines mutually intersecting.
If the cutting plane be in the et of xy, then
zee) and % -¥=0 (23)
which again represents two straight lines mutually intersecting.
This surface is called the cone.
If one of the coefficients of the squares of the variables (as that of Q)
vanishes, the equation may be reduced to the form
Pz? + Qy? + Rz = 0 (24)
which may be written, as in aa cases,
Uhre 2 s
1 ea a (25)
This formula includes the two following ones,
ze a z
oe pe ae (26)
The sections of this class of surfaces made by planes parallel to the
plane of xy are evidently, in the first case ellipses, in the second hyper-
bolas, which vanish in the plane of xy, and the magnitudes of whose
24
principal axes, varying directly as the distance of the cutting plane
from the origin, may be increased ad infinitum. The sections made by
planes parallel to the two remaining coordinate planes are parabolas,
the magnitude of whose parameters, varying directly as the square of
the distance of the cutting plane from the origin, may be increased
ad infinitum.
It is observable that in the second case the direction of the parabolas
in the planes parallel to yz and xz are opposed, while in the first case
they are the same. ‘The, first of these surfaces is called the elliptic
paraboloid, the second the hyperbolic paraboloid. |
The remaining forms which the equation to the surface can take are
the following,
6s
0
19
to
= eee eee (27)
Ee ae ee) ere oO ae a (28)
Lb ec aides ota re eer
fact on =) (29)
Seo OM een (30)
all of which are comprehended under the general term of cylindrical
surfaces. 'These, it is clear, being independent of one variable, will
retain the same form, whatever value be given to that variable; they
may consequently be considered as generated by the motion of a straight
line drawn parallel to the axis of that variable whose coefficients vanish
in the equation to the surface.
25
On the Partial Differential Equations of certain Classes of
Surfaces.
].—DEVELOPABLE SURFACES.
Ler the equation to the developable surface be
tex 0 (1)
and the equation to the plane of which it is the envelope
an + by + cz = (2)
where a, b,c are variable parameters.
Differentiating (1) and (2) we have
Udz + Vdy + Wdz = a
adx + bdy + cdz = 0
whence by means of an indeterminate multiplier
Again, on differentiating the first of (3), there results
Ud?x + Vd*y + Wd?z + dxdU + dydV + dzdW = 0 (5)
or by means of (4)
r(ad?x + bd?y + cd?z) + dxdU + dydV + dzdW = 0 (6)
but on differentiating the second of (3), there results
ad?x + bd*y + cd*z = 0 (7)
= Elence
dUdx + dVdy + dWdz = 0 (8)
combining which with the first of (3), by means of an mdeterminate
multiplier, we find
dU _ dV _dW_ (0)
or
BV = w'dx + vdy + u'dz (10)
WwW =v'de + w'dy + wdz
E
wU = udx + w'dy + |
26
Eliminating p, dx, dy, dz, from these equations with the assistance of
the first of (3), we arrive at the usual result, viz.
U? (vw—u?) + V?(wu—v”?) + W?(uv—w’”) \
11
+ 2VW(v'w’ —uu’) + 2WU(w'u' — vv’) + 2UV(w'v’ —ww’) = 0 ee
II.— TUBULAR SURFACES.
Ler the equation to the tubular surface be
L=0 (1)
and the equation to the sphere of which it is the envelope
(x—a)*? + (y—b)? + (z-—c)? = (2)
where a,6,c are variable parameters. Differentiating (1) and (2) we
have
Udz + Vdy + Wdz = 0
(x—a)dx + (y—b)dy + (z—c)dz = 0
Whence by means of an indeterminate multiplier
(3)
r—a —b z—c r
F Udi io, Na eee 4)
where
Pe 14 V2 y= (5)
hence
U
Pdz = rdU — =—dP
ae = rdU P
V
Pdy = rdV — —dP
y =r P t (6)
Pde = rdW — 1 dP
whence eliminating 7, dz, dy, dz, we arrive at the usual result, viz.
P4+rP{u(V? + W?) + 0(W? + U2) + (U2 + V2)—2u’VW— 20’ WU—2w'UV}
+ r°{U?(vw—w?) + V?(wu—v?) + W?2(uv—w’) (7)
+ 2VW(v'w’— uu’) + 2WU(w'u’ — vv’) + 2UV(u'o’— ww’)! = 0
It may here be remarked, that on r becoming infinite this condition
coincides with that given in § I. for Developable Surfaces.
27
On some Theorems relating to the Curvature of Surfaces.
J.—ON LINES OF CURVATURE.
Der. A line of curvature on any surface is a locus of a series of its
consecutive points, such that normals at each point shall meet the
normal at the consecutive one.
Let the equation to the surface be
L=f(%, 42) = 0 (1)
then, writing for convenience,
dL dL, dL
U= _— =o = 2)
dx ‘4 dy nm dx (2)
the quantities U, V, W will be proportional to the direction-cosines of a
normal at any point (2, y, 2). ‘The equations to this normal are
Cs A ral IE Nove Sl
Unit 1a any Wao at | (8)
where r is the distance from the point (z, y, 2) on the surface to a
pot (1, ¢) on the normal and distinct from (2, y,z). Also, the
equations to a normal at any consecutive point on the surface («+ dz,
y+dy, 2+ dz) and meeting the former normal in the point (£, 1, ¢) are
Mere Ctl Cee a ts dite ae 2. eG
U+dG V+d¥V W+dW
whence, multiplying up and subtracting (3) from (4), there result
=r-+dr (4)
dy = Vdr + (r+dr)dV
dz = Wdr + (r+dr)dW
whence, eliminating 7 and dr by cross-multiplication, there results
(VaW—WaV)dx + (WdU—UdW)dy + (UdV—VdU dz = 0 (6)
which, being independent of £, », %, will hold good for any point on the
surface. Hence (6) is the differential equation of a surface, by the
E 2
dx = Udr + wrsana |
(5)
28
intersection of which with (1) the lines of curvature are formed. It is
observable that since
dU AW ede ?L ’L d aL 7
di i
a ay), et -—_ d= d. ee es
DOT Tet aid Ge ae + yds ede
dV dV dV aL aL Ll
V= —_ aE —- =) ae (E sree "
dV = a gpd as aa ed cae (7)
re dW dW a? ?L aL
1W — ah / el eee], Se cee dL: eae eee |
: dx is dy Ye dz dxdz *Y ayde Has dz?"
the equation (6) will be of the second order in the differentials dz, dy,
dz; and consequently will in general give two values for the ratios
dx: dy:dz; in other words, will represent two distinct lines traced upon
the surface (1).
Again, suppose the equation (6) to have been integrated, and that
the result of the integration is
BS (8)
the differential of which is
dc + Wdy + dz = 0 (9)
where @, G, G1 are quantities analogous to U,V, W. Combining this
with (6) there results the following system,
VdW — WdV _ WdU — UdW _ UdV — VdU
(10)
at oD cet
hence, using the factors U, V, W (as in Theorem I. Pref.) there follows
UG + VP + We =o (11)
whence the following theorem may be enunciated ; ¢f two surfaces cut one
another in their lines of curvature, they cut one another at right angles.
Again, let three surfaces whose equations are
L= Z=0 Pi . (12)
cut one another at right angles; the conditions that this may be the
case are
QuU+0V+Maw=0
UU+VV+WW=0 (13)
U@ + V0 + WU =0
where U, V, W are quantities analogous to U, V, W, or &, ©, Ga.
But on differentiating (12) we have
Udx + Vdy + Wdz =0
Clie + Fay + uo} (14)
Udx + Vdy + Wdz =0
29
from the first and second of which there result
dx i dy s dz (15)
VEi—-DPW WRA-GIU UY-Gv
and using the factors
VdW—WdV, WdU—UdW, UdV—VdU
the numerator of the result will be
(VdW — WdV )dx + (WdU—UdW)dy + (UdV—VdU)dz (16)
and the denominator
(U? + V?+ W?)(@idV + Pav + Wadw) (17)
Hence, writing for convenience
U?+ V?+ W? = P? (18)
each of the ratios (15) is equal to
(VaW— Wav) os + (WaU— aw) + (UdV—VaU) e
. (19)
GAdU + DV dvV + GW
Again, using the factors
Dd-TdY, 3 =9§dA- Aa, = Ad - Pda
each of the ratios (15) would be equal to
(aw —- way) = + Marta”, + May —yay =
pe 3p AP" P (20)
Ud@i + VdV + Wddet
But differentiating the first of (13) we find
Ud&t + VdV + Wd@@i + CAdU + DadV + GAdW =0 (21)
Hence, writing
VdW — WdV = P*dA WdU — UdW = P?dB Udi Vall = P'dC-
VdtA—TGAdV=P-da Gud W— Ud GA = 3p2d 38 adv —DdG=Pd€ | (22)
VdW — WdV = P°dA WdU — UdW= P?°dB UdV — VdU= P*dC
and forming quantities similar to (19) and (20) for U, V, W, we should
have the system
(d&+dA)dx + (d%3+dB)dy + (d€+dC)dz = 0
(d4+dA)dx + (dB+dB)dy + (dC+dC)dz = 0 |
(dA+dA)dzx + (dB+d3B)dy + (dC+d€)dz = 0
(23)
or
dAdx + dBdy + dCdz =0 }
dAdx + diBdy + d&dz = ey (24)
dAdz + dBdy + dCdz = 0
which are the equations to the lines of curvature on the surfaces L, £, L
respectively ; hence if three surfaces cut one another at right angles
30
the lines of intersection of any one of the surfaces with the other two
are its lines of curvature ; which is Dupin’s Theorem.
From this it also follows that the two lines of curvature on any surface
are perpendicular to one another.
There is one other remarkable property of lines of curvature which it
will be worth while to investigate before proceeding further.
Writing for convenience
H =u + yu (Uw —Vo'— Woo! |
V if / / :
K =o + gy (Vo —Ww'— Uw) t (25)
WwW , LENT.
L=w + tv (We Uw — Vv’) |
there results
si fs H? + Km? + Ln? (26) 3
Pp
where , is the radius of curvature of the surface, and
_ de _ dy _. dz Og
is ds Can ds im ds Sh
also from the equation to the surface
Ul + Vm + Wn = 0 (28)
Now in order to find the variations of p corresponding to the variations
of 1, m,n, we have from (26) and (28)
Hidl + Kmdm + Lndn = sa(= ) (29)
Zea We, :
and
Udl + Vdm + Wdn = 0 (30)
but
P+mtn=l (81)
hence
ldl + mdm + ndn = 0 (32)
But on a line of curvature the relation
Hl(Wm— Vn) + Km(Un— Wi) + La(Vi—Um) = 0 (33)
is satisfied, which gives by cross-multiplication from (29), (30), and (32),
P
d{—~)=0 34
() (34)
Hence on a line of curvature the radius of curvature is a maximum
or minimum : or, the lines of curvature pass through the principal sections
of the surface.
31
IJ.—vmBILIct.
Der. An umbilicus is a point on a surface round which (for all
points situated indefinitely near the point in question) the curvature is
the same in every direction. The directions of the lines of curvature
will consequently become indeterminate at an umbilicus.
In order to find the conditions for the existence of such a point,
we have
VdW — WdV = (Vo!— Ww’)dzx + (Vu' —Wv)dy + (Vw— Ww')dz (1)
also
Udx + Vdy + Wdz =0 (2)
therefore eliminating dx
VaW — WdV =— WKdy + VLdz (3)
Modifying the other terms of (6) of § I. in the same way, and arranging
the result, the transformed equation becomes
U(K—L)dydz + V(L—H)dzdz + W(H—K)dzrdy = 0 (4)
hence if
HoK=S 1 (5)
the equation to the lines of curvature is satisfied independently of
dx, dy, dz. But since when either U, V, or W vanish these con-
ditions become indeterminate, it will be necessary to make some
transformations, in order to put them into a determinate form.
Suppose then that U=0. Multiplying (5) throughout by U V W,
and substituting for H, K, L, from (25) of $I., we see that when U
vanishes
Vo’ — Ww’ = 0 (6)
also writing for convenience
Vol — Ww! = O (7)
(5) becomes
Base (O= Un) aioe C0 = Ut} (8)
UW UV
r Wv
Whence, using the factors VW there results
WwW Vv U
—v0 + —w — 2’
po Naan Os, a
Veer V.
Van Ww
32
Hence we obtain the following systems, any one of which are the
conditions for the existence of an umbilicus.
eh Vigo — 2VWw + Ws
6-10 Vo’! — Ww’ = 0 Ve We (11)
me Wu — 2 WUv’ + Uw
2 / 2
We='0) .9 Ua oo i oe set (13)
II].—on THE LINES OF CURVATURE ON AN ELLIPSOID.
Varrous methods have been proposed for investigating the nature
of the lines of curvature on an ellipsoid ; but as all have involved certain
transformations more or less laborious in order to integrate the equation,
it appeared that these tedious operations might be advantageously |
avoided by the following course. ‘The equimomental surface was first
brought into connexion with the problem by Mr. Cayley, in the Cambridge
and Dublin Mathematical Journal, May 1846.
Let the equation to the ame oo
aot y “4 =1 (1)
then the equation to the ee of curvature is
Q at 2 a, Y 2 a, =
-— ia (C7 — ee on is == 9
(b ee a Se A ) 0 (2)
Now if «, 6, y be so chosen that
eo—@=l —-BP=ac—y=h (3)
the equation (2) may be replaced by
oo nee 9 eae: 2 Ohh he pee
(VE + (Fat) £4 (a8) Z=0 (4)
But the equation to the ellipsoid whose lines of curvature are giyen
by (4) is
or substituting from (3)
xv y z
= - . = 6
Beh Fon eee, (6)
Hence the lines of curvature on the ellipsoid (1) coincide with those
on the surface (6), but (4) are the conditions that (6) shall be confocal
33
with (1); consequently, the lines of curvature on a central surface of
the second order are formed by the intersection of that surface with
confocal surfaces of the same order.
Again, if instead of (3) the following system had been employed
&—a®=P—PP=e-y? =k (7)
we should have had instead of (6)
wae Tas Pt cg ATA eS aed 1 (8)
atk +k o+k
Now subtracting (6) from (1), (8) from (1), and (8) from (6), there
result
x 2 2 a 2 2
Her HET Teer? ASEH RED exw = °
mo) F: y? as, Co gee
(2+h)(@+h) (PAP +A | (2+h)(C+A)
Hence the lines of curvature on a central surface of the second order
are formed by the intersection of that surface with cones of the same
order.
Again, from (9) there result by symmetrical elimination
(9)
ot y? 2
a a be Ke Ce
FELINE — CHAFEE — (EDN SINC HA)
E 2 (10)
— (h+h)at(P? —c) +b —a?) +.c#(a? — 07) ] + a'(b?—c*) + 8c? — a?) + (a? 0")
= 1 |
; a‘ (b? —c*) rm b*(c? —a?) a c#(a?— 6?)
whence
‘ rPah+tkt w+ P+ (11)
and (8) may consequently be written thus,
2 ¥ 2 cs
Ea Cy | + et te ep eet t (12)
Hence the lines of curvature on a central surface of the second order
are formed by the intersection of that surface with an equimomental
surface (which in certain cases takes the form of the wave surface).
The perpendicularity of the lines of curvature on an ellipsoid follows
immediately from (9); those equations being in fact the conditions that
the system of surfaces shall be orthogonal.
34
On certain Formule for the Transformation of Coordinates.
Mr. Caytey has, in a paper in the Cambridge Mathematical Journal,
quoted some very beautiful formule for determining the position of two
sets of rectangular axes with respect to each other, employing rational
functions of three quantities only. The geometrical method, however,
by which he has deduced them is somewhat complex; it is hoped,
therefore, that the following analytical investigations will be found as
elegant and more simple.
Let the nine direction-cosines be
iT} mmm”, nn'n”’ ;
then since the two systems are rectangular, they are connected by the
following six formulze of relation,
P+ rtt+nt=1 P28 me ne 1 PS gl? 4 on? — (1)
Ul’ + m'm’ + nn" = 0 ll + mm + n/n = 0 U' + mm'+nnr’'=0 (2)
whence we easily deduce the inverse systems
2s (7? 4+ 77=1 m +m? + m’?= 1 FU ik YE * cok a (3)
mn + mn’ + m’n’= 0 nl + nl +n''l’= 0 lm + I'm’ +1'm’” = 0 (4)
Now subtracting (3) from (1)
m? + n& — 1? —0? = 1? 4 n? —m? — mM? = 1? + mi? — n? — n®= 0
whence
n? — mM? =I"? —n? = m — I? (5)
Let us now assume
r 5 YD La
7m’ Pon m—?l (8)
combining which with (6) we arrive at
(n’ +m) = (Ul +n) = (m+)
or
pv vr Ape
nan! VA gn en ee
35
Now we have introduced three quantities, but have made only two
assumptions concerning them; we are therefore at liberty to make
another ; let this be
(6) = (7) (8)
Then we easily deduce
put rN_ w+ em _AMtVY_ py—-rA_ WAM _dA— (9)
n’ 1’ m m’”’ an a * va I’
whence
eee ge ee 1—P a 4(1—P)
Que Ooty (ev + Ota)? I++ e+) (14
wee ® at 1 — m? a 4(1—m’) 10
(YAP (Apt)? (uyv—r)?+(ud+v)? (L074? +07)? — (14+ ?— YP —r?)? { )
n° n’? 1— nn” 4(1—n’’)
(A—pe (ptAP (A—p)P?+ (wa (LFA +e +P (1+ — pF)?
Now it is obvious that we may from inspection find a solution which
will satisfy these equations, but it will be more satisfactory to solve
them directly ; we shall arrive at the same result.
We will write (10) for convenience thus
i i Hh an ma a) Sig NITi]
l
o 7 a’ B oe B” y yy’
The values of
aa’, BRB", yy
are obvious; our object is to determine
Now since the two systems are rectangular we have the conditions
UIT+ m’m + nn =0 (12)
UN + mm”! + nn’ = 0
W + mm’ + nn’ =0|
whence by (11)
0.0+8'T+y¥u=—aa"
cio Or += — 88 (13)
a’o + Br + Ov =— yy
whence by cross multiplication
o oe T = Uv
aa’ Bry — Bn (BP +7) BY BY a’ yal" (y? +a?) ona" B"— a’ Ba’? +B’) (14)
1
~ al@"y+ a” By
F 2
36
But
a’ Bi'y + al Bry! = (pv +X) (vN+ BAW +Y) + i Re eek (15)
— 2(r2u2v? + w?v? + v?A? + r2u?)
and
aa’ Bry — By (BP + 9°) = (pe —v*) (NV? — pw?) — (wv? —V*) (L +?) (Ww? + a (16)
= (Np? + wy ele m?)(1 +A? — ps? — v”)
whence forming symmetrical quantities for + and v, we have
7 FAM ms p= ee Seed (17)
Lib Am ee ae ope iti ee Dee
whence by (10)
K=14+NV4p'+r (18)
whence we immediately arrive at the system given by Mr. Cayley
Kd=1+N—p’?—v Km = 2AAp+r) Kn = 2(vA—p)
KU = 2(Au—yv) cm =14+—rv?—r? Kn’ = 2(vu+r) f (19)
Kl’ =2(rv+ p) Km’ =2(uv—D) qn’ =14+7?—)d?2— pw?
37
On the Principle of Virtual Velocities.
By the Rev. B. Price of Pembroke College.
Tue object of the following paper being to deduce the equation of
virtual velocities from more abstract principles than is attempted in the
Mécanique Analytique, it may be allowable to make some few remarks
on what Lagrange has there given as the proof of the principle, but
which appears to the writer of the present paper to be nothing more
than an illustration.
Lagrange having acknowledged that the principles of mechanics
adopted by the older mathematicians are nothing more than particular
expressions of the equation of virtual velocities under different points of
view, conceives that the equation depends on a principle different to
that of the lever and the composition of forces, viz. on that of the
pully ; and by means of this constructing and measuring a system of
forces, which act on a body or system of connected particles, proceeds
to give what has been usually termed by English writers, Lagrange’s
Proof of the Principle of Virtual Velocities.
It seems, however, to the writer of the present paper, that no advan-
tage is gained by the introduction of the pully. ‘The pully is not nearer
to the first principles of mechanics than the lever, the inclined plane, or
the parallelogram of forces; whereby Lagrange, in this proof, is not
nearer to mechanical axioms than was Archimedes or Varignon. Re-
actions of various kinds, tensions, rigidity of cord, friction, and so on,
are forces introduced in the system of pullies and blocks of which no
account is taken; and no principles or axioms are previously laid down
which authorize us to neglect these. Lagrange seems to make the
principle, which is nothing short of every problem in mechanics, to
depend on a single problem, viz. that of the pully and the block. He
deduces the general type of every mechanical equation from an equation
which he, in a manner, proves to be true in the case of forces acting
through a particular combination of the pully. It seems to the writer
of the present paper, that if Lagrange intended his problem to be a
38
proof of the principle, he is guilty of the fallacy of concluding the
general equation from the particular instance.
In the method pursued in the following pages the three laws of
motion are assumed as axioms, and upon them the reasoning is founded.
It is conceived, however, that they are not empirical laws, but general
truths, adapted to the principles of mechanics, which we may thus
state :—
I. “ There is no effect without a cause ;” which may be thus adapted,
‘a particle or body at rest remains at rest, and a particle or body in
‘¢ motion continues to move uniformly in a straight line, unless acted
‘¢ on by some force external to itself.”
II. “‘ Every cause produces its own effect;” and we may thus state
the adapted mechanical law, “ a force acting on a body under the
“ action of other forces produces its own effect equally as if the other
“ forces did not act.” |
III. “* Action is accompanied by an equal and opposite reaction ;” that
is, ‘* whenever a mechanical force acts there is a reaction simultaneous,
‘‘ equal, and opposite.” In what manner action is to be estimated must
be derived from experience ; and on this subject more will be said below.
It may be also worth while to notice that the methods of measuring
velocity and accelerating force by means of differentials (not differential
coefficients) are considered to be known; that D’Alembert’s principle
is assumed as axiomatic; and that wherever the differential calculus is
introduced it is considered to be a calculus of differentials or infinitesi-
mals. But before proceeding it 1s necessary distinctly to understand
what is meant by virtual velocity, in what manner virtual velocities
are to be estimated, with what signs to be affected, and then formally
to enunciate the principle.
99
I. Meaning of the expression “ Virtual Velocity.”
By virtual velocity of a point or of a particle is meant, that velocity
with which the point or the particle would begin to move supposing it
to be disturbed from its position; or it is that velocity with which, at
the first instant of its motion, any point or particle moves, supposing
such a motion to be possible. Hence, when a number of forces act on
a body or a system of points connected in any manner so as to produce
equilibrium, were a slight motion consistently with the geometrical
relations of the points to be given to the system, the small spaces
39
described by the point of application of each force relatively to the
direction of the force in a very short time @ is a correct measure of the
, os
virtual velocity of the point, that is of the force, inasmuch as _ is the
velocity of the point ; and if @¢ be the same for all the points, the virtual
velocity must vary as és; and relatively to this direction of the force,
because the force can produce motion or pressure only in its own direc-
tion, and therefore whatever small motion the point of application of a
force may have in any other direction, that is due to some other force or
forces ; as, for instance, suppose P, Q, R to be three forces acting on a
particle A, and to be so related in intensity and direction as to keep it
at rest ; conceive the point A to be moved over a very small space to A’,
the ahaohike distance over which A has moved is
AA’; but the virtual velocities of the forces are
proportional to the several projections of AA’
on the several directions of the forces: viz. the
virtual velocity of P is measured by Ap, the
virtual velocity of Q by Aq, and of R by Ar;
the displacement being so very small that we
may consider the directions of' the forces still to be parallel to their
former directions in the undisturbed system; and the virtual velocity is
to be considered positive when the projection of the absolute space over
which the point of application of the force is moved is towards the
point whence the force acts, as is the case in P and R, and negative
_when it falls in the direction produced backwards, as in ase .
* As the parallelogram of forces is with great facility deduced from the diagram given
in the text, it may be worth while to anticipate the equation of virtual velocities for the
purpose of exhibiting the several steps.
Let? AAP = @ Lu ess, RAQ—a, PA =), QAP =y,
the equation of virtual velocities is in this case P. Ap —Q.Ag+R.Ar=0
and substituting for Ap, Ag, Ar their values, we have
P. cs cos 6 + Q. 6s cos (06+y) +R. és cos (B—6) = 0
dividing this by cs, expanding the cosines of the multiple arcs, &c. we have
P+Q cos y + R cos 6 + (R sin B—Q sin y) tan 6= 0
whence as @ is indeterminate
P+Q cos y=—R cos (1)
Q sin y= R sin B (2)
Squaring (1) oe (2), and adding R2= P24 Q?+ 2PQ cos y, the parallelogram of forces;
and from (2) ——— Q pall from the symmetry, the triangle of forces.
sina sin inp sin y
40
II. Enunciation of the principle of Virtual Velocities.
The principal of virtual velocities is this,—when any number of forces
are so combined in both intensity and direction as to produce equilibrium,
the sum of the products of each force and its virtual velocity is equal to
zero; or if any number of forces acting on a body or system of particles
are in equilibrium, the forces are inversely proportionate to their virtual
velocities ; whence if P be the general type of a force, and ¢p the
general type of the small distance relatively to its own direction over
which the point of application of P moves, owing to a slight displacement
of the system, then
> Rop = 0
and we may thus formally state the theorem :
If any system of points or of bodies acted on by any number of forces
is in equilibrium, and if a small motion is given to the system con-
sistently with its geometrical conditions, by means of which the points of
application of the forces move through small spaces, these spaces rela-
tively to the directions of the forces are proportional to the virtual
velocities, and the algebraical sum of the product of each force and its
virtual velocity is equal to zero; the velocities being considered positive
when the spaces described are towards the points whence the forces act,
and negative when in direction of the force produced backwards.
III. Explanation of the Principle.
This principle involves two considerations,
Ist. In what manner are forces producing pressure or motion to be
estimated ? How are their effects to be measured ?
2nd. What do we mean by equilibrium ; that is, in what particular
forms and in what proportion, as well as to intensity as to
direction, are forces to be combined to produce equilibrium ?
As to the first, daily experience shows us that the effect of a force
depends on two circumstances, the mass of the moving body which
contains and is acted on by the force, and the velocity with which it
moves; the effect we perceive to be greater the greater the moving
mass, and the greater the velocity ; whence, in mathematical language,
we say the effect is a function of the mass and the velocity ; but what
function? What we may call the “ proof experiment” of Attwood shows
that pressure producing motion or moving force is to be measured by
the product of the mass moved and the accelerating force; that is,
+t
in order to estimate the statical effect of a moving mass, we must
multiply the abstract accelerating force by the concrete mass ; or, as we
may state the case under another point of view, in order to impress
on a given mass a certain velocity in a certain direction (7. e. in
order to produce a certain effect) a force must be applied, the intensity
of which must be proportional to the product of the mass to be moved
and the velocity with which it is to move, and the direction of course
the same as that in which the body is to move; this product of the mass
and the velocity is called the quantity of motion. Or again, supposing
a body to contain P particles, each equal to m, m being the unit
particle, and suppose the body to move in a given direction with a
certain velocity measured by p, the distance over which it passes in
a short time, then the effect of each particle is equal to mp, and there-
fore the effect of the body being the aggregate of all the effects of the
several constituent particles is equal to Pp; for although it may appear
that some of the effective forces, that is, some of the motion, is lost,
owing to the mutual connexion of the particles, as for instance the
molecular forces, yet by the principle of D’Alembert, as long as the body
is rigid, and there is no motion of the particles relatively to one another,
these forces neutralize each other, and the body notwithstanding produces
a force of precisely the same intensity as the sum of the forces due to
each particle m, were each to move independent of and separate from
the others ; whence then it appears that the correct measure of the effect
of a force arising from a moving body is the product of the mass moved
and the velocity with which it moves ; and this being so,
The second consideration comes in; in what manner must forces
measured as to their effects in the way we have just explained be
combined so that a body under the action of them may be at rest?
Forces are, I say, in equilibrium, when the absolute sum of' their effects
measured as above is less than in any other state, when the system is
slightly deranged :—for conceive a system of bodies or of points which
is in equilibrium to be slightly disturbed by the introduction of a new
force external to the system, the equilibrium of the former system is
thereby destroyed, but the new system will arrange itself in a position
of equilibrium, and the resultant of all the forces in the former system
will be such in intensity and direction as exactly to counterbalance the
force which has been introduced; thus the sum of all the forces of the
G
42
deranged system is greater than the sum under the former balanced
system by twice the resultant, that is by twice the force which we have
introduced in deranging the system ; whence it appears that if the body
or the system be at rest, the sum of the forces is less than if it be
disturbed, and therefore the sum of their effects measured as above is a
minimum for a position of equilibrium.
Let therefore P,, P2, P3, . . . P, be m forces acting on a system,
and let us conceive (as we always may do) that these forces act from
or towards certain points in the lines of their directions, which points
we will, after Lagrange, call centres of force, and let p,, ps, Psy» + + Dn
be the distances between these centres and the points of application of
the forces; it is plain then that the several forces P,, P., P;, . . . P,
remain the same in the deranged as in the undisturbed system, so
that the only quantities which vary owing to the displacement are
Fis Dern Dove Pavepscn
t= SPp:
then in a state of equilibrium H is a minimum, and therefore
oll =0
soe Lop =O (1)
that is, the sum of the products of all the forces and the virtual
velocities is equal to zero.
IV. Other circumstances of equilibrium.
Are there not however other ways in which the forces may be com-
bined so that =. Pap = 0, and yet these not be such states of equilibrium
as the one we have supposed? ‘The theory of infinitesimals, as applied
to the discussion of curves, shows that 811 = 0, when [1 is a constant, as
well as when II is a maximum or a minimum-—a maximum or a minimum
I say, because in either case we have such a position of equilibrium as
has been supposed in the last section; the absolute sum of the effects
of the forces having been taken, without reference to change of sign ;
for (to adopt the ordinary language of mathematicians) if a body or a
system of particles is in equilibrium, whether stable or unstable, an
additional force is required to keep it in a position slightly deranged
from its position of equilibrium, and therefore the absolute sum of the
effects of the forces is, as before stated, greater in the displaced than in
the undisturbed system by twice the force which is required to keep
43
the system in its displaced position; but suppose II to be a constant, the
criterion of equilibrium is satisfied, and the body or the system is in
equilibrium in other positions of slight displacement; for although by
the first law of motion a force is required to change the position in
which the body rests, yet as II is constant, not admitting of increase or
decrease owing to any small displacement, we have no force to call into
action to be equal to and counterbalance the external force we have
applied to move the system, and therefore the system or body would
continue changing its position until some other force (as friction,
resistance of the air, &c.) acts to neutralize the applied force, and
thus to produce rest; whence it is manifest that as soon as such
a force does act the system or body is brought to rest and there
remains; and as this counteracting force may be brought into action
at any instant of the body’s motion, it follows that there are an infinite
number of positions in which the body or system can rest under the
action of the impressed forces. Such a state is what is usually called
“ Neutral Equilibrium,” but what may well be named “ a position of
stationary action.”
V. Method of applying the formula.
The difficulty in applying the equation of virtual velocities to the
solution of statical and dynamical problems consists in the calculation of
the virtual velocities %p,,%p,, ... ép, for the several points of application
of the forces. The simplest method, however, is as follows, — to
consider two different positions of the system which are consistent with
the geometrical relation of the parts, one of equilibrium, the other
slightly deranged, and to express the virtual velocities of the pomts of
application of the forces in terms of the coordinates of the points of
application and of arbitrary quantities which have been introduced in
deranging the system, and as these latter are indeterminate, equating to
zero their several coefficients, we shall have a sufficient number of
equation to determine the position and conditions of equilibrium.
Let then P,, P,, P;,... P,, be the forces acting on the system, 7, y; 2,
Lo Yo 2 - » +» Ln Yn%, be the coordinates to their points of application,
A, 6, C,, A, b, Cy, . +. A, 5,C,, be the coordinates to their centres, defined
as in Art. III., and let their direction-cosines be cos «, cos 6,, COS ys
COS a2, COS Fx, COS Y2) «++ COS %y COS B,, COS yn, and let the distances of their
G 2
44
centres from their points of application be p,, p., ps... P,» so that we
have the following system of equations,
Pr = (%-HYP + (%-hY + (4-4)
Px = (%.— 4)” + (Yo— 5g)? + (2-62)? (2)
Pa = (®_= Gn) + (Ya—5n) + (Zn on)”
it is manifest that in any small derangement of the system, such as we
have conceived, we may suppose the centres of force to remain fixed, and
the points of application to move; but before we proceed further we
must make some remarks of great importance.
VI. Analysis of motion into that of translation and that of rotation.
Suppose a body or a system of points of invariable form to change
its position in space, and consider two successive positions of the body
independently of the forces acting and of the time consumed in passing
from one position to the other, it is manifest that there is an infinite
number of different ways in which the change may have taken place.
Which, then, of all these is the most simple? Which is the best adapted
to the methods we have adopted of determining position ?
Whatever displacement a system may have undergone, we may con-
ceive the motion to have taken place as follows: Ist, All the points
may have moved over equal and parallel lines in the same direction ;
2dly, Considering some one point in the body or system to be fixed, and
to be a centre round which the body turns, though this point may be
continually changing, yet we may give such a revolution to the body or
the system, that the several parts of it shall after these two motions hold
relatively to space that position which the body has after its real motion ;
and 3dly, There may be a motion of the particles of the body or system
relative to one another, such as to cause a change of the mutual position,
influences, and distances on each other, as is the case with gases, elastic
fluids, machines in which one part slides on another; this however not
being the case with rigid bodies, to which alone it is the object of the
present paper to apply the principle of virtual velocities, and as the
position of a body in space, with reference to fixed axes and planes, is
not affected by this third kind of motion, we shall at present confine
ourselves to the first two, into which, though they be of distinct character,
all motion, even the most general, may be resolved.
45
And as the principle of virtual velocities is true for the most general
motion conceivable, the velocities are to be estimated in their utmost
generality, and therefore the system must receive a motion both of trans-
lation and rotation, though not simultaneously but at successive times,
as the principle of superposition of small motions, which is equivalent to
that of infinitesimals, authorizes ; and =.P%p is to be calculated for both
these motions ; we proceed, therefore, 1st, to give a small motion of trans-
lation to the whole system, along each of the three rectangular coordinate
axes, whereby all the points of application of the forces will be moved over
equal small spaces parallel to the axes; and 2dly, we will turn the body
or system successively round the three axes through small angles, the
effect of which two motions will be that the points of application of the
forces will be moved in the most general way possible, and we shall
derive from the results all the conditions of equilibrium.
VII. Application of the principle to motion of translation.
Recurring to formule (2) Art. V., and taking the variations, we
have i
Tie a) it 91 mel
dp, = oa da, + Pe oy; tape oz)
and substituting the values of the direction-cosines, we have
dp, = cos a, O27, + cos By dy, + cos y, dz,
and similarly
Sp. = Cos a2 Sx, + cos By Sy + COs ¥ dz. !
1 (3)
SPn = COS An O52, + C08 By OY + COS Yn SZq
where @p,, 9p. . . . 8p, are the distances relative to the direction of the
action of the forces over which the points of application of the forces are
moved.
Which may also thus be shown —
Let 8s be the absolute displacement of the point (7, y, 2,) owing to
the motion of translation, then the direction-cosines of this line és are
= Z, a and if ¢ = the angle between the line és and P’s direction,
dp = 6s cos ¢
= be{ cos @ + by J cos B + 5 cos 7}
os 8s
= dx cosa + dy cosB + 8z cos ¥
46
Applying these values in the equation (1) of virtual velocities, and
remembering that all parts of the system move over equal and parallel
distances in the same directions, and therefore
O2p ice 08) O23 Sie t of Ot
Sy) = Deiat eee ae = *| (4)
62; = 62, = 023 = De ch phigh OS a
the equation >. Pip = 0 becomes
>. P {cosa dx + cos B dy + cosy dz} = 0
or, as we may write the equation on account of conditions (4),
=>.P cosa dr -+ =.P cos B dy + =.P cosy dz = 0 (5)
ax, dy, 'z being outside of the sign of summation, and putting
00s aut 2 COS po oY ae 00S 7
Xdx + Yoy + Zdz = 0 (6)
If then the system or body is entirely free to move in space, we have
no other relation given between z, y, z, and their variations, besides (5)
or (6), and therefore we have
Sai Pecos 10 =. P.cos BR = 0 ee Pacisays— (7)
that is, the sum of the parts of the forces resolved severally along the
three axes of coordinates must separately be equal to zero.
For the cases where the system is not free, but confined to move on
a given curve, and for other deductions from (5), reference may once
for all be made to the Mécanique Analytique.
If the forces P, P, . . . P, act in parallel directions, we have
Cy Se Og es Se i Be oe Se tL,
PB} =B =h3=.-++-+ = Bn
V1 =— Yo = Ya = 2 2 ee ee = Yn
and conditions (7) are reduced to the single one,
Pe tet Ps tetbs beter ar as ca eens le oP ee eo ee (8)
VIII. Application of the principle to motion of rotation.
The theory of rotatory motion, as laid down by Poinsot, authorizes us
to conclude that every motion or tendency to motion round any axis
may be resolved into three motions round three rectangular axes, passing
through some point in the axis about which the body turns or has a
tendency to turn; and hence it follows, that if we turn a system through
small angles round each of three rectangular axes, the resultant of these
several motions is equivalent to the most general motion of rotation that
aN
the system can receive round any axis passing through the origin of
coordinates.
Let 2 Y 2) %2 Yo 2% +++ Xn Yn % be the coordinates to the points of
application of the forces P,, P., P;,... P, 3 and let
2,7 + yy + 2," = n°
2 2 2 ay
oh bier ”
zt ie Yr? 4 a =_ 72a
and let the projections of these several radii vectores be on the plane of
yz Tie Tor Tae © + «© Tne
Me eli Oh} aT ay
LY Vig Tox T3z 26 0 8 Try
Let us first turn the system through a small angle #4 round the axis
of x, 6 being the angle between 7, and the axis of y,
y = 7, cos 8 2 =a ty sO
sy = — r, sin 080 dz = 7, cos 060
oe oy = — 260 éz = yo0
taking the variations of y, z, 6, which are the types of yy... Yn
Pera eetcing Ui Us et Uae
Similarly let the system be turned through a small angle % round the
axis of y, ¢ being the angle between 7, and the axis of z, and we have
dz = — xdh dz = 25h
and let the system be turned through a small angle #) round the axis
of z, / being the angle between 7 and the axis of x, and we have
dx = — yd dy = xd
ZY, 2,9, and ¥ being, as before, types of the several quantities which
correspond to the several forces.
And by the principle of superposition of small motions, being autho-
rized to neglect these quantities which are variations of variations,
inasmuch as they become infinitesimals of a higher order, and therefore
inappreciable in an expression which involves infinitesimals of a lower
order, the total variation of these several points of application of the
forces is equal to the sum of the variations due to the several rotations
round the several axes of coordinates; hence we have the following
typical values of the variations
dx = 26h — yor
dy = xd — a0
dz = ¥59 — «dh
(10)
48
and since the expression (3) gives
dp = cos abx + cos Bdy + cos ydz
substituting the values (10) we have
dp = (y cos y—z cos 8)60 + (z cos a—=z cos ¥)6p + (x cos B—y cos a)oy
whence substituting again in (1), remembering that 2%, , @) are the
same for all the forces, and therefore may be written outside the signs
of summation, we have finally
=. P(ycosy—zcos8)d0+ >. P(zcosa—x cosy)op + =>. P(x cos8B—ycosa)iy=0 (11)
and if the system of the body be entirely free to revolve round any of
these axes, that is, if we have no other relation between 4, ¢, and J than
equation (11), then 4, ¢, } being independent of each other, putting
=.P(y cos y—z cos 8) = L
=. P(z cos a—2 cos y) = mt
=.P(x cos B—y cos a) = N
we have L=0, M=0, N =0 as the conditions of equilibrium of rotation ;
that is, since L, M, N are the sum of the moments of the forces tending
to turn the system round the three rectangular axes of coordinates then
of x, y, z respectively, we conclude that if there is equilibrium the sum
of these moments must severally and separately be equal to zero.
Thus then having deduced from the equation of virtual velocities
the equations (6) and (11), the six equations of equilibrium which are
ordinarily given in treatises on mechanics follow immediately in the
manner we have indicated.
(12)
Addition to the foregoing Paper.
Suppose the forces which we have been considering in the last paper
to act in parallel direction, then we have
Oy == 0 10 ee
Roem eey (1)
YY as ae a ae
and the equations (7) are reduced to the single one,
Pret P. + Pee ces ee etn 10
and equations (12) become
cosy >. Py —cosB =.Pz =0
cas. Peco. Pe= 0} (2)
cos8 =.Pr—cosas.Py=0
49
whence we have
=.Pr_ =. Py_ =. Pz
cos a cos 8 cos y
Now conceive the parallel forces and their pomts of application to be
so arranged that
Sree Pee aa asa (4)
then «, 6, y become indeterminate, and there is no pressure of rotation,
no tendency to make the system turn round any axis, whatever be the
direction of the forces relative to the coordinate axes.
Whence it appears that if a system of parallel forces subject to the
conditions expressed in equations (1) and (4) act on a body, the system
is in equilibrium, whatever be the direction in which the forces act; and
this leads to an easy method of determining the centre of parallel
forces.
Conceive a number of forces of which P is the type to be acting on a
body, «yz being the types of the coordinates of their points of appli-
cation, and conceive the forces to be so arranged that we have one of
them P, the coordinates of whose point of application are «yz, to be such
in intensity and position that the conditions (1) and (4) are fulfilled,
viz.
=” {(5. Pzr)?+(S.Py)?+(=. Pz)*} (3)
P=P,4+P.4+ a wore batik
Pre = Piz, + Po t+... + Pit, = 3S. Pz
Py = Pi Paysite 20s 2) + Pap = S.Py
Deted Picola Peta bide pd Tub ney > 2S Pz
then we have
oe Fon: Ps Popa Py tne ob
este! me Pe at asp
Whence it is easy to determine the coordinates to the centre of gravity
of a body, for inasmuch as it is that point at which if the whole weight
of the body be collected the moment of rotation about a given axis is
the same as the moment of the body constituted as it is, we have a case
precisely similar to the one supposed above, and therefore if wy z are the
coordinates to the centre of gravity, we have
~ fxdm ~ _ fydm ~ _fedm
7 = fdm I= fdm a= fam
50
On infinitesimal Analysis.
Tue calculus of infinitesimals appears to be a branch of analysis to
which much attention has not been paid by English writers. The
method of limits has probably been supposed more tangible and more
satisfactory to minds previously unacquainted with considerations of this
nature ; and indeed it is not without difficulty that the mind first forms
a conception of a system of quantities differing infinitely from one
another in magnitude, and yet all distinct from an absolute zero, and
learns that many mathematical expressions are formed only approxi-
mately by means of neglecting terms of certain orders of infinitesimals.
Now although this is the case, there still exists a difficulty of a not
entirely different nature in the doctrine of limits, so that the latter
method is perhaps not so eminently superior to that of: infinitesimals as
to demand the complete exclusion of the latter. In fact its frequent
employment in applied mathematics, as in the theory of curves of double
curvature, of small oscillations, including the undulatory theory of
light, and other cases, appears to call for some explanation of its funda-
mental principles. Moreover, the two methods obviously leading to the
same results must in fact come to the same thing; and it may not be
inappropriate here to mention the point at which the coincidence takes
place. The limit of any continually increasing or decreasing quantity
or ratio is that quantity or ratio to which it always tends, but to which
although it may approach nearer than by any finite quantity or ratio, it
never becomes actually equal. The difficulty then consists in passing
from any given value of the quantity or ratio under consideration to
the limit, for in doing this (however near we may approximate to the
limit) a saltus must finally be made; because the essence of a limit
consists in this, that by no continuous approximation is it possible ever
to arrive at it. Now it is obvious that nothing short of the limit itself
will give accurate expressions in the result, and, consequently, it is
necessary either to reach the absolute limit or to remain satisfied with
ot
approximations. ‘The former course is pointed out by the method of
limits, the latter by that of infinitesimals. In order to compare the two
methods it will be well to consider a simple example. In both cases a
curve is considered to be approximately a polygon, the degree of ap-
proximation increasing with the number of the sides ; and in the method
of limits it is said that the two figures coincide accurately when the
number of sides becomes infinite, or, in other words, that in the limit
the magnitudes of the sides absolutely vanish ; while, in the infinitesimal
theory, it is supposed that the magnitudes of the sides always have an
existence, and that they themselves are indefinitely divisible. Now, if
it be required to consider the deflexion of the curve from the tangent,
that is, the angle between any two consecutive tangents, it will, in the
former theory, be necessary generally to consider the second tangent as
drawn through some point at a small distance from the first point, and
afterwards again to have recourse to the limit, by making the distance
between the two points actually vanish. Hence, geometrical representa-
tions can be formed of an infinitesimal no less than of a finite quantity,
because they differ from one another only in magnitude ; but since in
the limit all magnitude is lost, no complete geometrical representation
can be formed: and, in fact, (returning to the example noticed above, )
as long as any appreciable angle exists, in general between the two
tangents (the two points of contact actually coinciding) the curve
becomes a polygon, and we find ourselves employing the imfinitesimal
method. It therefore appears that the method of limits is only a
particular form of that of infinitesimals, and that, from the geometrical
indeterminateness arising from the notion of the limit, the infinitesimal
quantities are not actually destroyed or eliminated, but only lost
sight of.
Perhaps no better authority for the definition of infinitesimals and
their use can be quoted than Carnot, who makes the following remarks :
“ Jappelle quantité infiniment petite, toute quantité qui est considerée
“comme continuellement décroissante, tellement qu'elle puisse étre
‘¢ rendue aussi petite qu’on le veut, sans qu’on soit obligé pour cela, de
“faire varier celles dont on cherche la relation. Lorsqu’on veut
“ trouver la relation de certaines quantités proposées, les unes con-
“* stantes, les autres variables, on considére le syst¢me général comme
“ parvenu a un état déterminé que lon regarde comme fixe: puis on
H 2
52
“compare ce systéme fixe avec dautres états du méme systéme,
“ lesquels sont considérés comme se rapprochant continuellement du
‘“ premier, jusqu’a en différer aussi peu qu’on le veut. Ces autres états
‘du systéme ne sont donc a proprement parler eux-mémes, que des
‘“‘ systémes auxiliaires, que lon fait imtervenir pour faciliter la com-
‘“‘ paraison entre les parties du premier. Les différences des quantités
‘“ qui se correspondent entre tous ces systémes peuvent donc étre sup-
‘“‘ posées aussi petites qu’on le veut, sans rien changer aux quantités
‘“‘ qui composent le premier, et qui sont celles dont on cherche la
“relation. Ces différences sont done de la nature des quantités que
‘““ nous appelons infiniment petites : puisqu’elles sont considérées comme
‘ continuellement décroissantes, et comme pouvant devenir aussi petites
‘“‘ quon le veut, sans que pour cela, on soit obligé de rien changer a
‘‘ Ja valeur de celles dont on cherche la relation.” “ L’analyse infinité..
‘“ simale nest autre chose que l’art demployer auxiliairement les
‘“‘ quantités infinitésimales, pour découvrir les relations qui existent
‘entre des quantités proposées.”* In addition to which let it be
observed that the ratio of an infinitesimal of the (7+1)th order to one
of the mth is an infinitesimal of the first order, and the ratio of any
two infinitesimals of the same order is finite.
A quantity or function is said to vary continuously from one value
to another when it is capable of receiving any value between the given
ones as its limits. In this it is to be observed, that any values (which
lie between the limits) chosen arbitrarily and in any order of succession
may be given to the quantity or function without violating the Jaws to
which it is subject ; and consequently the difference between any two
consecutive values may itself have any value not greater than the dif-
ference between the two limits. This difference may, however, obviously
be made as small as is desired by taking the two consecutive values
sufficiently near to one another ; and by taking the same value twice in
succession, it may be made absolutely zero. Hence a series of these
differences, considered as a system of quantities, are as arbitrary as the
values themselves, and their differences may consequently have any
value not greater than the greatest of the first differences. These new
* Carnot, Reflexions sur la Metaphysique du Calcul Infinitésimal, Chap. I.
53
differences (or, as they will be called, second differences) may also
obviously be made as small as the difference between the first differences
whose values approach nearest to one another in value ; and as small as
is desired, and even zero, by a proper choice of the first differences, if
the latter are independent, i.e. subject to no conditions. This process,
being evidently subject to no other limitations than those above given,
may be continued as long as is desired; and a difference of any order, of
the mth for example, is consequently as arbitrary in magnitude as the
first assumed values, provided always that no conditions exist which can
affect the differences of any order inferior to the ath.
Let these differences be represented in the following manner ; suppose
that wis any quantity or function, and wu wu’ two values of wv, then their
difference
uo —u = du;
similarly if du, (dw)’ be two values of dw, then
(du)’ — du = ddu;
and similarly the successive differences of uw may be thus represented
du, ddu, COCA as fee
Now it is clear that the superior limit of the magnitude of any
difference of the nth order depends upon the magnitude of the greatest
difference of the (n—1)th order, this again on that of the (n—2)th, and
so on until that of the (n—n)th, i.e. until the given values themselves ;
but the inferior limit of the magnitude of any difference of the nth
order depends upon the distribution of the magnitudes of the differences
of the (n—1)th, these again on those of the (n—2)th, and so on until
the (n—n)th, i.e. finally upon the distribution of the original values
chosen between the given limits. Consequently when the given limits
differ infinitely from one another the differences of any given order, as
for instance the mth, may be made infinitely great, finite, or infinitely
small; while in every case they may be made infinitely small. These
remarks will be sufficient to explain the general theory of the various
orders of infinitesimals.
It was seen above that by choosing the primary arbitrary values of
the given quantity or function sufficiently near one another, the first
differences may be made as small as are required ; in other words, they
may be made infinitely small in comparison with the values themselves ;
in this case they will be called differentials. Again, by repeating the
54
process the second differentials may, in accordance with the remarks
made above, be made infinitely small in comparison with the first
differentials ; and so on without limit.
It must however be remembered that these differentials, although
infinitesimals of any order, however great, are still actually different from
zero; because it is only in the case where two consecutive values of
the (n—1)th differences have been taken actually equal to one another
that the nth differences vanish ; which is a case distinct from the one
indicated above. ‘These remarks are evidently applicable whatever be
the order of infinitude of the primary quantity or difference under con-
sideration. Hence, of any quantity, whether infinite, finite, or infinite-
simal, there may be an infinite number of differentials, each differing from
one another by any required order of infinitude. Now these differences
or differentials of the orders 1, 2,3,.. . . exist, as has been seen, when
the function under consideration varies continuously from one given
value to another between its given limits. Similarly those of the orders
2, 3, 4,.... exist when the first differential varies continuously between
its proper limits; and so on for others. But on the other hand, these
differentials evidently may exist also in the case when the function varies
no longer continuously but per saltus, never receiving any other values
than those contained in a given system. For the differences of these
fixed values will be the differences of the first order, and the differences
of these first differences (if the first differences be unequal, as must be
supposed in the general case, ) those of the second order.’ The differences
will however no longer be arbitrary, as in the case of continuous varia-
tions, but will be determinate as soon as the saltus of the primary func-
tions are given. But in order to make a differential of any given order,
as the mth for instance, vanish, it will be obviously necessary to perform
upon the differentials of the (—1)th order an operation which is only a
particular case of the general one above considered, and differing, not in
nature but in tensity (if it may be so termed), from that which was
performed upon those of the (n—2)th. Since however this would intro-
duce much complexity, especially as im the various problems which
present themselves it would be convenient to employ a larger or smaller
number of orders of differentials according to circumstances (which
would destroy that uniformity of method which it is always desirable to
combine with an unity of principle), the method usually adopted, and
65
the one which appears most desirable for general use, is the following ;
that the differentials of any given orders, as of m,n,p, ... . should
coincide in magnitude with infinitesimals of the same orders respectively ;
hence, the precisely same operation which converts a differential of the
mth order into one of the (n+1)th will convert one of the (n+1)th into
one of the (n+2)th; and so generally r operations of the same nature
will convert a differential of the mth order into one of the (7+7)th.*
* An example of this is the theorem of Taylor for the expansion of f(#+dz) in a series
whose well known form is,
fete) =f(x)$ fey +P" oes.
the number of terms being infinite. The usual expression for the remainder after n terms is
dx”
——— f(a + Oe:
eae nt Ea
where @ is some quantity between 0 and 1. If however
f (2)
represent that the (z-+1)th differential of f(x) is to be made to vanish, the series in
question might be also written,
2 NAB. pit gy a de We,
nae Ce ee et to tea de)
which however is not so convenient as the former, since a less clear idea of the remainder
after m terms is given by the new operation ‘f than by the quantity 0.
Another instance of various symbols of operation is one proposed by M. E. Lamarle in
Liouville’s Journal de Mathématiques, Juillet 1846 ; it is as follows ;
« Lorsqu’on écrit
S(a@+h) — f(x) = hf'(x+6h), (1)
on se borne, en général, a faire observer que 6 désigne une quantité comprise entre 0 et 1.
On sait cependant que les premiéres notions de l’analyse algébrique permettent de fixer
d’une manitre extrémement simple et tout a fait précise le sens de l’équation (1).
Ajoutons qu’en l’écrivant sous la forme que nous venons d’indiquer, on est forcé d’en
restreindre l’application aux cas ou la fonction et sa dérivée demeurent continues dans
Yintervalle que l’on considére.
« Cette remarque suffira sans doute pour justifier la préférence que nous accordons a
la formule suivante:
S(a+h) —f(®) = h§$asr’s (2); (2)
ou la caractérisque J# se trouve définie par l’équation de condition
SA’ (x) = lim lz (2+4)4p(et 2) +(e ae)t +f (e+ a
a ” ”~ a ”
” “ " a *
n
ns
a
a
a
“
{aoe
“ et qui peut s’énoncer en ces termes: ‘ Dans tout intervalle ow la fonction demeure con-
“ tinue, son accroissement a pour mesure l’accroissement de la variable, multiplié par la
“< valeur moyenne de la fonction dérivée.’ ”
56
There are however certain particular cases in which this process, viz.
of making the differentials of certain orders vanish, is used with great
advantage ; as for instance, if there are (m—1) equations involving
n variables and their differentials of various orders, any one of the
differentials of any of the variables may always be determined in terms
of the remaining differentials of the (»—1) variables, the differentials of
the mth variable, and the variables themselves. Consequently one of
the differentials of the mth variable is arbitrary, i.e. subject to no restric-
tion; it may therefore receive any value at pleasure ; the value usually
assumed is zero; and the differential equated to zero is usually the
second. The variable subjected to this arbitrary condition is called
the independent variable; and it is easily seen that any one of the
n variables may be so treated. It is also observable that unless second
differentials are involved there is no clue by which we can discover the
independent variable. It may further be observed that the second
differential has in preference to all others been subjected to the con-
dition above stated, because it is of the lowest order which we are at
liberty so to treat. This will easily be seen when we consider that from
the manner in which these differentials have been formed it appears
that the giving the differentials of any order (the mth for instance)
any arbitrary value (such as zero) is a condition which affects not only
the differentials of the mth and all higher orders, but also those of the
(m—1)th order, inasmuch as the values of the differentials of the latter
order to be chosen for the formation of those of the mth are no longer
arbitrary, but restricted to equidistant values. Hence the first dif-
ferentials could not be made to vanish without subjecting the values of
the variables themselves to limitations which the conditions of the
problem would not justify. It is also clear that if the second dif-
ferentials vanish, the third, fourth, . . . . and all others do so also. This
is the whole theory of the independent variable.
In actual practice it will of course be impossible to retain an infinite
number of orders of differentials, as we should in that case have an
infinite number of quantities to contemplate and take account of at the
same time, and our faculties are such that this is beyond their power ;
hence we are obliged to have recourse to a method of approximation,
and the one naturally adopted is to neglect those small quantities which
do not affect the truth of the results. This may at first sight appear
57
an unsatisfactory method of operation, and one which overturns all
the preconceived notions of the absolute truth and correctness of
mathematical results ; but it is nevertheless the method actually adopted
(or at least equivalent to it), although often brought to notice under
different aspects. But although this is the case, it is still equally true
that the error committed by this method can never become appreciable,
for the neglect of infinitesimals of any order, as those of the mth for
instance, can for each term so discarded produce no errors of an order
lower than the nth; but no sum of a finite number of infinitesimals of
the nth order can produce a quantity of the (n—1)th; similarly no sum
of a finite number of infinitesimals of the (7 + 1)th order can produce
a quantity of the mth; and so on throughout all orders. Suppose now,
for example, that above the (x + r)th order none but finite sums exist,
or in other words that by neglecting differentials of this order no
quantities of lower orders are discarded, (7m and 7 may, of course, have
any value from — «© to+ oo). Now the only case where an error of an
order lower than the (7 + 7)th can be committed by neglecting terms
of this and higher orders is that in which there are an infinite number
of terms of this order; but as in no expression which can occur in
practice is it possible to write down these at full length, the terms in
question must themselves (if they exist at all) appear under some other
form (such as a sum or product, for example), which in fact will present
them as a quantity of the (7 +7)th or some lower order; but as terms
of the latter order have by hypothesis been retained, no error of an
order lower than the (7 +7)th will have been committed; and as terms
of this order have been considered as inappreciable, the possible errors
will themselves also be inappreciable. On the other hand, if after having
retained no quantities of orders higher than the (n +7)th it be thought
desirable to introduce those of the (rn +7+ 1)th, a similar train of
reasoning will be applicable, and no theoretical difficulty will arise. It
has now been shown that the usual method of differentials 1s a conve-
nient one, that if it be adopted terms of certain orders must be
neglected, but that in neglecting these terms no error of any order
lower than the lowest neglected order can be committed ; the extent of
the accuracy of the investigations is consequently always determinate,
and the method when carefully pursued is free from danger.
When the orders of differentials follow those of the infinitesimals, as
indicated above, they will be denoted by the symbol d, so that the
I
58
successive differentials of any quantity or function wu will be thus
denoted
la eid, te Se
or as they are more usually written,
du, Pu, Bus... s. (1)
Since we are best acquainted with quantities which are called finite,
and less so with those which are indefinitely greater or indefinitely less
than these, it has been thought advantageous to commence the notation
of the powers of the symbol d from this as their zero point. Now since
the powers of any symbol (as in the case of d) may be increased in a
negative as well as a positive direction, without limit, it will be necessary
to determine the meaning of quantities when affected with the symbol
of operation d raised to any power.
It must here be recollected that we have not proved, nor do we
assume, that the law of the operation indicated by the symbol ¢ follows
in general the index or any other law; but since it is known 4 priori
that any mathematical symbol (be it one of quantity, quality, or any-
thing else) is liable to be subjected to any algebraical operations, the
same must be the case with the symbol now under discussion, namely d ;
the question therefore now is, what operation does the symbol dz repre-
sent, (1) when 2 is a positive integer, (2) when it is a negative integer,
(3) when it is fractional. Now the first case is easily determined, for from
the manner in which the differentials have been formed, i. e. from their
very nature, we see that the symbol d does follow the index law when
the index is a positive integer; and this also because we have not given
the term differential or differential coefficient any meaning which will
require to be afterwards further elucidated (as, for instance, differential
coefficient has been defined to be the coefficient of the first power of h
in the expansion of f(x+h) in a series of ascending powers of h),
but have shown the actual connexion with the operations of differen-
tiation with the fundamental operations of all algebraical processes.
For this purpose it is easily seen, that in order to pass from the dif-
ferential of any quantity of the mth order to that of the same quantity
of the (n+1)th order, it is sufficient merely to write dd*u, i.e. d**'u, for
du, i.e. multiply the differential of the mth order by the symbol d.
Hence, in the method now adopted, the raising of the index of the
symbol d by unity turns an infinitesimal of the mth into one of the
{n+1)th. Similarly, m order to pass from the differential of any
59
quantity of the nth order to one of the (n—1)th, it is sufficient merely
to divide the differential of the nth order by the symbol d. Hence, in
the method now adopted, the decreasing of the index of the symbol d by
unity turns an infinitesimal of the mth order into one of the (m—1)th.
In order however to explain more fully the nature of the operation
mdicated by the symbol d7’ or s, it will be necessary to revert to the
original formation of the quantity du. It was above seen that the
various values which can be given to du are formed by taking the
differences of the arbitrary values of the given function which lie
between those limits within which the function remains continuous.
This being the operation indicated by d, it is now required to determine
the nature of the inverse operation indicated by d~’. But since by the
operation d the quantity under consideration, w, was divided into a
number of indefinitely small portions, whose general type was dw, and
the sum of which was the difference of the values assumed by w at the
given limits, the required operation will be to deduce from any given
differential, as du; for instance, the quantity w,; this will obviously be
effected, if to the inferior limit of w (w,) there be added the sum of so
many of the differentials whose form is du as will be equal to the dif-
ference between wu, and u,;; or, if more convenient, by subtracting from
the superior limit w, the sum of so many of the differentials as will be
equal to the difference between wu, and uw, This result may also be thus
analytically represented :
Suppose uw to be any function of x, so that
u = f (2)
f'(2) = f(e+ de) — f(x) = f(z)
f'(@) =f(e + 42) — f(z)
f(atdz) =f(x+2dzr) — f(r+dz)
fi (a2+2dz) = f(x+8dz) — f(x +2dz)
also let
* But
f(eti—ldz) = f(x+idr) — f(x+i— dz)
Consequently
S'(z) +f'(2+dz) +... + f'(«+i—1dz) =f (z-+ tdr) —= f(x)
hence | ines
f(z +idz) = f(z) + f(z) +f (e+dz) +. . ++ f(eti—1dz)
12
60.
Now if from the superior limit of f(x) (f(#,) suppose) there be sub-
tracted the inferior limit (f(2,) suppose) the result viz. f(7,) —f(.«,) is
called the definite integral of f’(.7); suppose moreover that
Doe ae ees Ae
then
ST (an) — f (20) =f (to) + f'(%o + dz) #.. 2 +f" (zt. + n= 1dz,) (2)
that is, the definite integral of any function is equal to the sum of the
values assumed by the differentials of that function between the given
limits, inclusive of the inferior and exclusive of the superior limit. Such
then is the nature of the operation indicated by the symbol d~’; and it
is also clear that the same may be repeated any number of times without
limitation; hence it will not be necessary to consider further the nature
of the operations indicated by the symbols whose general type is d~”,
where m is any integer. A similar train of reasoning would show that
the symbol d may be raised to any positive or negative integral power,
and any root of such power be taken, and that the result will always have
an intelligible meaning; thus, for instance, if d be raised to the pth
power the result willbe d@’, and if the gth root of this result be taken
the final result will be dt, and if
13
{|
3
we should have
dy =d’
that is, the result will be the same as if we had arrived at d’ by a direct
process.
The meaning of the operations indicated by the symbols d* d~”, where
nm is any integer, having been determined, the following question also
suggests itself; what interpretation is to be given to a symbol of the
form d=? In the first place it may be remarked that a quantity affected
with the symbol d raised to any power (the mth for instance) is of any
order of infinitesimals one degree higher than the same affected with the
symbol raised to the (n—1)th power. Now this it must be remembered
was a particular method which it was thought desirable to employ.
Suppose, however, that the symbol d denote the operation of differen-
tiation when there is no limitation respecting the connexion of the
61
orders of differentials and infinitesimals, then, as before, the results of
successive operations of this nature on any function wu will give the
series
isn Gate vOGdzs ct. ass
or as they may also be written
Cue Ce United chek ts (8)
in which any conceivable relation may exist between the symbols
d and d; but among all of these the one with which we are now con-
cerned is that in which 7 operations of the nature d are equivalent to m
operations of the nature d, i. e.
da = 0"
or as it may be written symbolically
di=d™ or d = d*4, 0rd" = d (4)
Now it is observable from equation (4) that by performing the
operation d 7 times any given quantity w is changed from a finite
quantity to an infinitesimal of the mth order, or that by performing
the operation d once wis changed from a finite quantity to an infini-
tesimal of the — order. This is what was required. Now, although
we have arrived directly at infinitesimals of fractional order, yet it
must not on that account be expected that instances of such quantities
can be given; for on account of our inability to determine the exact
limits of finite and infinite (whether infinitely large or infinitely small)
quantities, we cannot determine whether any given quantity lies at
“th or 1th part of the distance from one to the other. But the same
m
reasoning which shows the possibility (mdglichkeit) of the various
orders of infinitesimals involves also the necessity (nothwendigkeit) of
all orders, and consequently of fractional as well as integral. Thus
analysis, being the representation of real laws, gives as some of its
results these partial infinitesimals, as they may be called; which, since
they lie between two given orders of infinitesimals, must have a real
existence, although there is an apparent indeterminateness about them.
But this indeterminateness is itself a result of the absolute magnitudes
of the various orders of infinitesimals which have been chosen as near
together as was possible without destroying the general distinctness of the
various orders. _ From the actual existence of these partial infinitesimals
62
it follows that they may be used in any analytical expression with no
jess propriety than the total infinitesimals; and although the former
are not in such general use as the latter, yet there are some cases in
which they may be employed with great advantage.
[The following is an example; in vanishing fractions (i.e. when for
a given value of x, as
the fraction
£(a)
S(@)
takes the form © ) certain cases present themselves in which, however
many times the numerator and denominator may be differentiated, the
value of the fraction still appears indeterminate, so that
df (zx) d’f (x)
Vl) ae ee uO
df(x) — 0 d(x) a, 0’ 4. e+e fe he oe
dx dx?
The readiest way to evaluate these is to find the general forms of
d"f (x) d"f(z)
ieee a ies
and then to give n some fractional value, the result will then be found
determinate. ]
From the manner in which differentials and. integrals have been
formed, viz. by subtraction and addition, it is obvious that the order in
which these are performed when any function is affected with symbols
of both operations, is indifferent, so that
anpru = frau (5)
Hitherto functions of only one variable have been considered ; but as
there are some points in which functions of several variables differ from
those of one variable, it will be worth while briefly to notice them.
Consider then the function of any number of variables
FT (is Ys Estee eee
the total differential of this is obviously
f(at+dz, yt dy, z+dz,..-.) —fl%y, 2%». -)
But as in certain cases it may
Une. «: -
63
happen that only one of the quantities
. may be considered variable while the rest remain constant,
the partial differentials so formed will be represented by the symbols
JS (Fy Ue eek one)
(a+dz, yt+dy, z».. +) —f(x, ytdy, z -.+ +)
eee L(rs Ue Ze oh on)
doh wih (ts Ya Zs « + +)
- -) —f(e+dz,y, z .. »)
-) + f(x,y, 2 « » +)
dj dy des
the suffix denoting the quantity which is considered variable:
Now,
d,f (a, Y;% » » +) = f(z+ dz, y, 2; .
also
df(t, I, % -- =f
— fir ar, Y Ze
Again,
dy f (as Ys 2 > © +) =f ytdyy2s -
Ady f(x, Ys 2 + + +) = f(xtdz, ytdy, Zz, -
— f(a y+ dys 2%» +
hence
dd. f (23 Y; 250% +
oP df (X45 2 « ++)
Similarly it may be proved that
Addy « «+ fl%s YZ
or symbolically
(EE LEAT EL
Oaee rat, a Las Ue Suh ote- s)
(6)
(7)
where P represents the continued product of all the symbols taken in
any order whatever.
written
f(x + dz, y+ dy, z + dz
= f(z + dz, y + dy, z+ dz
+ f(t, y+ dy, z+dz...)
+ f(a, y,2+ dz...)
a ea ae a ee ee ee ee eee
But
S(2+dz,ytdy,z+dz.
=d[ f(x, y+ dy, z+ dz.
consequently
f(a + dz, y+dy,z+dz.
—f(x« + dz, yz+dz.
= d,| fle + dr, y,z + dz.
Again, the total differential of f may be thus
oe) —f (2 Ys % - re}
~~ f(t, y+ dy,z+dz...)
— f(t, y,2+ dz...)
— f (Xs Ys 2% + + +)
Se &5 @. 6) Of 6. 6. 6
— f(%s Ys Z + + +)
..)—f(, y+ dy, z+dz...)
}
(8)
—
~)—f(e y+ dy z+dz. mat
-)tf(ay2z+dz...)
- )—-alf(z,y 2+ dz...)]
= dd, f(t, f, 2 + dz. -)
which is an infinitesimal of the
second order; consequently repeating
64
the process for the other variables, it would be at length found that the
expression
f(z + dz, y+dy,z+dz...)—f(z,y + dy, z+ dz...)
= df fl2,.y + dy, 2i4,dz ."..)]
differs from |
fe t+ dn Qe) — fly, 2e ee) = alfa ee
only by infinitesimals of orders higher than the first, and similar results
would be all deducible for the other terms of (8); consequently, as far
as infinitesimals of the first order
f(at+dtytdyz+dz...)—f(%y2z.. -)
= f(x + dz, ey, Ets) — J (te tecr ie)
+ Rep y dy, zit ai as paz ae e)
or writing
uF (x, 9, So0 s )
du=du+tdu+t+dut+...
or multiplying and dividing the several terms by dz, dy, dz, .
respectively, there results the usual formula
du du du
mt beth, ath Fa ae a te eate Che
Du ae z+ dy ly + a, +
‘ du ,
where D represents the total differential and dg: the ratios of the
partial differentials of w (considering as variable only those quantities
whose differentials appear in their respective denominators), and the
differentials of x,y, 2. . ., or as they are usually called the partial
differential coefficients of wu with respect to 7, y, 2...
65
Examples of the Application of the Infinitesimal Calculus.
Tue following are some miscellaneous examples of the application of
the above methods to actual practice.
i
To find dy in the following cases, calculating the expressions as far
as infinitesimals of the first order only.
yrHarer,
*, dy = (a+a+dz) — (a+z) = dz.
y=a—?,
“. dy = (a—u—dz) — (a—z) = —dz.
Yaa ae,
“. dy = a(x+dz) — av = adz.
a
ie ae
Me Tepe es a adx
ORS TE dat ge a(a + da)
adx dz adx
=— —_(]—— oe re Ve mh ce
x ( mi ) ao
eee
“. dy = («+dz)*— x*
= axt—"dz + GE a eae ia = ON.
Time Os 3
o's Gif = atte gt = af(a*—1)
2
= a*(log .adz Sees +...) = log.a.a‘dz.
ae diy = TE a G% = a (alt 1)
2
= e(de+ + OY dr,
K
66
y = sin 2,
.. dy = sin (x + dr) — sing
= sin x (cos dxr—1)+ cosa sindr
= —tsinz sin*dx +...+cosz sindz
= cosx sin dx = cos zdz,
since cos dx = /1—sin*dx = 1—4sin®”dr +. . ,,
and sin dx = dz, approximately.
y = cos@,
.*. dy = cos (w1+dz) — cos x
= cos x(cos dx—1)—sin z sin dx
= — sing sindr = —sin zdz.
y = tana,
.'. dy = tan (v+dz) —tanz
_ tana + tandz
~ I—tana tan dz
1 + tan *z
1—tanz tan dx
— tan z
= tan dx = (1 + tan*z) dz.
S
Il
Ns
_a+dx x _ zdx—adz _ zdx—xdz
2+ den) eit 2(z+ dz) io eae
I].—rTAYLor’s THEOREM.
Ler
Ua (ey, ce ae)
be any function of the variables xyz... .; and let
Ly Ys Sys
receive increments
AL, dYpleys = 5
respectively, and let
wu’ = K(a+dz, y+dy, z+dz....)
If then we consider the increments
dx, dy, dz, .
as infinitesimals of the first order, and
Du
67
a corresponding quantity relative to the function wu, calculated as far as
the first order only, we shall have as a first approximation
uo —u= Du (1)
or
u=u+ Du (A)
But since the function w’ may involve the products of the differentials
dG dy dz 4+. - , the quantity Dw will not represent accurately the
difference wu’ — uw, but will require certain terms of higher orders to be
added to it. Hence as a second approximation we shall have, in a
similar manner
wu —u— Du=D(Dz) = Du
or
u =u+ Du + Du (2)
Now neither of the values of w’ given by (1) or (2) is a correct one,
and that given by (2) may be either too great or too small (according
to the signs of the succeeding terms) ; hence we must take as the most
probable approximation the average of the values given by the several
approximations: Hence adding (1) and (2) and dividing by 2 (the
number of approximations), we have for the next value of w’
Dit Dx
iad ehagairse @)
Treating these in a similar way we shall have next
Du _ Du _p Duy Lu
eae OW ANT O/, 12
or
; Dien e tu one
= | —
u ut Tae aio a 1) (3)
and, as before, adding together (1) (2) (3), and dividing by 3, we find
Du , D?u D®u
OT Me ae See ere (©)
Similarly the next value would be
A? Dai D*x D®u D*u
ere) MO sl 5.5.4 Oo
until we should have as the mth approximation
ae Die Du D*u D"u
I Ue otis a oS * 1.8.90. sn Gu
Kk) 2
68
Moreover, carrying on the approximations in a similar manner ad
infinitum, the remaining terms of the series giving the accurate value of
w’ in terms of w and its differentials ae be
: De ape Be eee te \
n
Lee Dis): ky OD a (n+ 2) (n+3)
which quantity is evidently
1 n+
AOKI OST
and
1 n+1,,/
SON Pama (TORE
1G.
1 n+1
1 D3 cs oe Uhal) s Oke
and
“¢ D*t!F (2+ dz, y+dy,z+dz....)
1.2.8 :.4006 nda)
Hence if 6, 6’, 6” ... . be some quantities between O and 1, the
remainder will be accurately represented by
]
a See Sy 6 6’ De shee.
[2.8 A ee eee
and the correct value of w’ in terms of w and its differentials will be
u’ = F(r+dz, y+dy, z+dz... .)
eal So af, 2 one at + DE (sy, 2 ea) DF (2,9, 2 Set) Ova
oe th | aes
]
typ t te ee ee ac
os D+F (2+ dz, y+ 6’dy, 2+6’dz .. .)
leet ee (nd)
Where by the usual rules of differentiation
du du du
oe Pedy shoe
Di = ie gag
4 PU 72 ch
ome i ey ody
Diu = — de" + 7
du du du
OU dydz + dede +.
Wyte | a. ideds eal -)
ee
69
tee
From the above formule we may deduce an extended form of
Maclaurin’s theorem ; putting
r=0,y=0,2=0....
Chie ae Oo A at Ug 2 mat aL oe a
and
d d d
Pint Eada sich TOO
we have
] 1 t
Fay z...)=FO)+ VF) + 7, VF) +...
1 1
—— "F(0 ee eee eI MAIR (Oa by: OS,
De sree a Meno. 8s TORE F(Gx, Wy, 82. « .)
L[V.—LAGRANGE’S THEOREM.
To expand F(y) in a series of ascending powers of «, where
pas tery) (1)
(1.) As a first approximation, let
p= (2)
this gives
eo (3)
and consequently
F(y) = F(z) (4)
(II.) In order to make a second approximation, let the value of y
given by (3) be substituted in (1); this gives
y =2 + af (2) (5)
and consequently
F(y) = Fiz + of(z)} (6)
or by Taylor's theorem
Fly) = P(z) + of (2) F(z) (7)
which gives the value of F(y), as far as terms of the first order.
(III.) In order to make a third approximation, let the value of y
given by (8) be substituted in (1); this gives
yo2t f{z+ af(z)} } (8)
=2+4+ 2f(z) + vf(z)f')
70
expanding by Taylor’s theorem as far as terms of the second order.
Consequently, instead of (7) we shall have |
Fly) = Fe + af(z) + of (2)f"(2)5 |
= Fl) + of Fe) + fe) @P +7 5l/l@)PF'2) (9)
expanding by Taylor’s theorem as far as terms of the second order.
This may also be written thus,
F(y) = F(2) + f P(e) + DALE}. (10)
(IV.) In order to make another approximation, let the value of y
given by (8) be substituted in (1); this gives
y=zt af{z t+ af(z) - aS (2)f (z)} ;
= z+ xf(z) + 2f(z) f(z) + ef (z)(f(z))? + 3 (flz)) 7h" (2) (11)
== + ole) + EY + *D.AMDFE) |
expanding by Taylor’s theorem as far as terms of the third order.
Consequently instead of (10) we shall have
= Fle + af(z) + 2%fl2f'(e) + a Lf) (2)} |
= F(z) + af (z) F(z) + xf(z ue (z) F(z) +, D,[( (fer (z)]E"(z)
+ a ))?E"(2) + PEPE) +1 MDPOE 2)
= Fe) + ofl)F(2) + 55 Serta 2 _paxriaF@) |
and soon. Suppose that by continuing the approximations there had
been found
3? e L(f(2) )SF’(z)]
(13)
(12)
Fy) = + 5f(2)F( (z) + aa D, AS (2))F')1+5
Vee) of cas: "17 ( f(z) a
then for the next approximation we should have the general form of the
series (5), (8), or (11); viz.
sDLMNF (2)] +. + -- cl
(14)
y=2zt af(z) + x f(z) f(z) +
ct ep Bite 2 oa
71
and this being substituted in (1) will give
Fly) = Fe) + 24 fe) +S feyf'@ + 2 DSO FE!
teed DPE (OIL PE)
+ Sf re) + FrOso +4 DLL FO)
Derr rah Fe
feet 5S DI rorr ny F(z)
D.LFENFED
DF) £0) ay Fer#N(2)
ye ea ) +2 fle)f
Feee +
|
EP yal | (15)
+p aa SO +FfOS') + DAGON! |
xz”
Le Dackans
or writing for convenience
J (2) =u
du aw dyn!
F(y)=F = gene, Poe a Og OT \ /
WF) tay ttrat M93 hot as. itety fon
x du? cane drynt} :
eae rs eae | Ry
i ls ope ocs weve eet. 2.5,. aah} (?)
Ae A 16)
2 du? Pus Cur
——— —— fe ae n(n)
+rescal tT Se SP were ta eH maaGEll yi
grt us dyn) n+l
pameeeee ee ro iy = ar Fe (z
+123. ean" tigate ‘F seat (2)
But writing for convenience
du* due
geen gS — O 17
a oe baa aa Ne
it will be found by ordinary processes that
yt 1 _— uy” +I
du™*+! = (n+ 1)u"du
_n+1
x coefficient of x in the expansion of 0*
d?u"+1 = (n + 1){nu"—1(du)? + u"d?u}
72
—
- x coefficient of 2? in the expansion of 07-!
Ss
s
—
1.2
deur *) = (n + 1) {n(n — 1)u*-?(du)® + Snu”—"dud?u + u*dut
n+1 , : :
+ x coefficient of xin the expansion of 1"?
—na—2
1.2.3
. at 1 F ,
so that the coefficient et ae in the (n+ 1)th line of
(15) will be
FO) (z)(f(z))"*"
Spd Be Aitg 3) Ab in the nth
F(z) D,(F(2))?*
oe eds Net Woks aoa in the 3d
n= Dp (De f(2))"*"
vSels ie Ne he tote ds aed ce te eae ee in the 2d
F’(z)D n 7 (2) Pt 1
Hep. oie ts ae ee. in the Ist
BY(2)DA(f(2))"*?
i i
in the expres-
consequently a whole coefficient of 1. S33 anda (aa)
sion for F(y) will be
FDNSOY + T PED (fey 4-4
apnea aoe
n. ee ee z)Dy"( f(z)" + wb
which by Leibnitz’s theorem
= DA {(/@OY"""F'(} (19)
so that the series (13) is generally true whatever be the value of 7
713
V.—VANISHING FRACTIONS.
Ir wu be a function of x of the form
f(z
f(x)
where f(x) and f(x) are any functions of w, which remain continuous
between the limits
(1)
a
ea)
zx=aandr=a+e
eee 2)
exclusive of the first of each pair of values, it sometimes happens that
for certain values of the variable x, such as
aes (3)
the functions f(x), f(7) sensibly vanish; so that, if U be the value of w
when (3) is satisfied, |
f(a) _ 0
U= ete 4
fla) ~ 0 .
and the value of U is consequently indeterminate ; the question then
arises how the value of U is to be found. Now, according to the present
theory of infinitesimals, the values of f(a), f(a) will not in general be
absolutely zero, but merely infinitely small, so that the solution of the
problem depends upon the determination of the orders to which the
functions in question respectively belong. Ifthe order of f(a) be higher
than that of f(a) the value of U will evidently be infinitely small; but
if the order of f(a) be higher than that of f(a) the value of U will be
infinitely great ; in fact, if f(a) be of the order m, and f(a) of the order
n, U will be of the order
m—n
and this will give rise to three cases,
(1) Ifm > n, (m—n) is positive, and U will be infinitely small,
(2) Ifm < n, (m—n) is negative, and U will be infinitely great,
(3) Ifm=n, (m—n) = 0, and U will be finite.
Now, since the functions f(x), f(#) remain continuous between the
limits determined by the equations (2), an approximation may be made
to the values of f(a), f(a), by assigning to x some value a little greater
or less than a contained between the above-mentioned limits. If then
L
74
- be an infinitesimal of the first order at least, the value of U will be
given nearly by one of the equations.
_ f(a+e) _ f(a—e) ,
© = Aare)" ” = Rla=e) o
Now, by Taylor’s theorem,
(ate) = f(a) +-0(a) + a eer eve |
(6)
fare) =fla) +f @+ pal O+----
and | ;
f(a — s) =) f(a) — ££" (a) 4 22 f(a) =v)
Lae m
fa-2) =f) ~ Ff) + poh" ----
Hence if these values be substituted for f(a+<), f(a+e), f(a—s),
f(a—<) respectively in (5), we shall have (remembering that by the
hypothesis
f(a) = 0 J (2) = 9) (8)
f’ (a) +f" (a) oe an
[center ee srenenenneermen (9)
F(a) +f" (a+...
or
f’ (a) — 5” (a) sel?
Uo (10)
f(a) ~2f"(a) +...
from both of which expressions there would result, if infinitesimals be
neglected in comparison with finite ones,
f’(a)
= i i
Fe) wi
If the first derived functions themselves vanished, and in consequence
of the conditions
fMa)=0 f(a) =0 (12)
the expression (11) became indeterminate, it would be necessary to
take in another term in the expressions (6) or (7), im which case
(9) or (10) combined with (8) and (12) would give
Lea
U =v
KO ve
75
And generally, if the derived functions up to the (n — 1)th vanished
together with the functions themselves, the value of U would be given
by the equation
a £07) ( a)
Pua)
It may here be remarked that the conditions (8) and (12) and
generally the conditions
fP@=0 fM(a)=0 (15)
represent (as is known by the theory of equations) that the equations
(8) have each (2 +1) equal roots, a factor corresponding to each of
which is expelled by each differentiation; so that the degree of the
factor (which in the present case will be an infinitesimal one) is con-
tinually reduced by unity, until when it is integral, it at length becomes
actually 0; in which case, if there be no other infinitesimal factors, f"*”(a),
f“*(a) have finite values, and the value of the fraction is determinate.
Sometimes, however, it happens that the index of the degree of the
infinitesimal factor is fractional, so that, although it is continually
decreased by unity, it never vanishes, but passes from positive to nega-
tive ; so that the fraction always appears under one of the indeterminate
forms
(14)
Swe (16)
and its value consequently cannot be determined by the above process.
Suppose then that the order of the infinitesimal factor is fractional and
= '" and suppose that
n
mMaprtl
n Pp 3 2 a!)
where p is a whole number. Then by p ordinary differentiations
the order of the factor will be reduced from — to 4, and it is then
required to reduce an infinitesimal of the order / to a finite number, or
n
in other words to decrease the order of the infinitesimal by the ith part of
unity; this must be done, as was seen in a former paper (on Infinitesimal
Analysis), by choosing « of an order =“. But as it was seen in the
n
L 2
76
same paper that this can scarcely be called practicable, the general
formule for
d, fx d’ f(x)
Sid and mS (18)
must be found, and afterwards 7 must be put equal to z. This, according
to the principles of infinitesimal analysis, will produce the same result.
VI.
Der. 0
v(uvw — uu’? —vv? —ww? + Qu'v'w’) > 0 (24)
w(uvw—uu'? — vv? — ww? + 2u'v'w’) > 0
But (19) require also, in order that the conditions may be all satisfied,
that
u’?—vw < 0
y?—wu< of (25)
w’*—uv < 0
hence replacing wu, v, w, u’, v’, w’, by their values in terms of the
88
differential coefficients of ©, and adopting the notation of determinants,
the conditions finally become,
In the case of two variables
20 LO > 0
dx’ dady
: (26)
0, PO,
dydx’ dy*
In the case of three variables (2, y, z) the conditions will be
#0, FO 130 OD, LOA 0 QO QO |>0
dy’? — dydz dz*° dzdx dx?’ dady
PQ OQ d*0, QO dQ, PO,
dzdy dz | dae ce dydx dy?
and
#0 &O in &O > 0
dx? dx*’ dzxdy — dxdz
#0, @0, PD,
dydx’ dye? — dydz
#0, &O, @O,
dzdx dzdy dz
LO, LO LO, &O, > 0
dy? da”? dady’ = dadz
&O, LO, LO,
dydx’ dy?’ — dydz
PO, &O, #2,
dzdx dzdy dz
2O, OQ LO, LO, > 0
dz? da? * dady dzdz
BO, &O, LO
dydx dy? : dydz
PO, LO, PQ
dzdx dzdy dz?
89
Hence in general the final condition of a maximum or minimum will
be one of the following
(a) A@>0, 28@>0, €0>0 (27)
where
6 = A, tt. &
FF, B, D (28)
€, D, €
in which expressions @, 5, €, D, &, jf are of the form either
AQ’, BO’, CO’ (29)
where
or A, F, E
Hepa (30)
BK, DC
or ;
“| ge | en
and similarly A, B, C, D, E, F will be of the same form, and so on; or
the conditions will take the form
(8) = | A, 35 | >0
ub, @
where @, &, €@ are of the form (29).
It is easily seen that, as the order of the final condition doubles itself
every time the number of the variables is increased by unity, the order
in the case of m variables will be 2”.
If all the differentials of the function ©, up to the (2nm+1)th inclu-
sively, vanish, it was seen above that
D2er+Do,
must always retain the same sign if there exists a maximum or minimum
value of 2. In order to find the conditions relative to this case a similar
course must be pursued ; thus writing for convenience
DD, = Ox, (33)
it will be necessary to find first the conditions that
D?0n
may remain always of the same sign; this may be done by the rules
given above. The order of the resulting condition will be as was seen
above of the order
(32)
Qm—-l
N
90
where m is the number of the variables; but as Q,, is of the order 2n
with respect to the differentials, the order of the resulting condition
will be
2”"n (34)
with respect to the differentials. ‘This must be looked upon as a
quadratic, and the successive conditions deduced by degrees ; thus the
condition that
ax? + bx? + ...4 1 + ja +h
= (ax? + baP3 4+... 4) + jr +h
may not change sign is
jy? — Ak(aaP + bP +...4+ 7) < 0
which again may be written
Ak(aa’— + br? +...) 2* + Ahlx —7? > 0
and so on.
So
On certain Formule made use of in Physical Astronomy.
Most of the following formule are given in the Cambridge Mathe-
matical Journal, vol. 1v., but the demonstrations are different.
By the principles of dynamics the equations of motion of a planet or
satellite, considered as a material particle, are
de _»x By _ ig &z _7 (1)
dt? dt? dt?
where X Y Z are the resolved parts of the accelerating force parallel
to the three axes of co-ordinates respectively. We shall find it convenient
to use polar co-ordinates in future approximations ; we will therefore
transform (1). Let then
P=resolved part of the accelerating force parallel to the radius
vector.
(= resolved part of the perpendicular force in the plane of the
orbit.
S = resolved part of the perpendicular force perpendicular to the
plane of the orbit.
Also let
~
a,
b
—
TUN Ysera Re
whence, by differentiation,
dx _ dr, di dy _ dr dm dz__ dr, dn
a ff ee = — i —— yee mee 3
Camda de fdene Uo geNe ce Wont aes de (3)
Pe _)@r | odldr ,
dé dé Spr eS
dé aE dt dt =
d’z ne r dn dr Hs n
ae Osan ta tos
dem died es ade :
Now
d?l dm i =n
P=/X + mY eam Ng +h 72 (5)
~@1(@) te )+(%)t
92
as is found by differentiating the equation
P4+m+nt=!]
now from (3)
@)* 4) * GG) GG) Cee at
But if © be the angle through which the radius vector moves, we
have also
ds\? arn? dO?
=) =G)+"(@) (8)
hence
dl\?, (dm\?, (dn\?_ (d®@\?__,
G)a+G =(G) = (9)
where » represents the angular velocity of the radius vector.
But if 6 be the angle through which the radius vector moves,
measured along the orbit, and & that through which it moves, measured
perpendicular to the plane of the orbit,
Dia Dae a)
also
d_did@, dda
dt dO dt | dQ dt
dm __dmd@ , dmdQ
dt d0 dt | dQ dt (11)
ees |
dt d@ dt dQ «
so that
na (@)+@ (12)
(4) + GG) + (ay (18)
di dl dm dm dn dn
dd * @ao edo vl Co
and consequently also
hence we see that
di dm dn. dl dm dn
dé’ dé” do’ dQ? dQ dY
form the direction-cosines of a rectangular system ; viz. the radius
vector, the perpendicular to the radius vector in the plane of the orbit,
im, n;
93
and the perpendicular to the same line perpendicular to the plane of
the orbit.
Hence
ir
i Oy 15
i en On (15)
Again, if we write
HN ee Se \
Piroms “ee ao (16)
ie, : per dw _ 1d(ar?) _
Q=EX + mV +L = 20 — +r = | - (17)
Similarly, if
dl dm _ ane 4
OM eae Oe oe
be the direction-cosines of a perpendicular to the plane of the orbit, and
if » represent the angular velocity of the plane of the orbit round the
line perpendicular to the radius vector,
(F) + (3 t) + (GZ) =" 2)
al d’m @n \
hence
S=rAX + py ee bon
dndl | dudm , dvdn
dt dt dt dt ° dt al
taking account of the motion of the plane only, as is allowable by a
theorem given at the end of this paper.
(20)
Also since i he are the direction-cosines of a line perpendicular
- dx du. dy
to the plane containing the lines (/mn auv), so also are ee AE
direction-cosines of a line perpendicular to the same plane, and conse-
quently parallel to the first ; we shall for convenience measure the angles
so that the cosine of the angle between these lines is negative, hence
S=ros (21)
Again, to determine the motion of the plane of the orbit we have
dr dp dv
Boe ae ihe
dn. dp dy (22)
pew tae ae de
oe
whence
2 ee
dt | dec (23)
my—no nry—vl lw—mr —
Since if
AL+ pm + vn = 0
then (24)
(my — nu)? + (rn — lv)? + (Iu— mdr)? = 1
also writing :
| me {+ GEY+G} es
the equations for determining the motion of the planet or satellite in
its orbit are
tas Shh? dh _
ago aT x dey =
and those for the motion of the plane of the orbit
dx dp dv
he) dt dt Sr (28)
my—n nnu—lv Ip—mrn h
an (26) (27)
We will now calculate the expressions for P QS.
Let mm‘ m’ be a free system of three mutually attracting bodies ; we
may suppose one of them to become fixed (suppose m’’) if we apply to
the system forces equal and opposite to those which act on m’. Let
r= (m’’, m), y= (m’”, m’), 6= (m, m’), (29)
Then the forces which act on m are
me along mm” ie "along mm
Te 8 &
“along mm’ a parallel to mm
2 Pe)
Then
y
ea gl ie midis f+ Be)
ey. a m ,
+ =) sin né + oa sin 2nt +5 sin Sn (70)
101
Again, by Lagrange’s theorem,
9
e~
di (liven, 2h
5 aan nt—_ (cos nt)
Cos uw =cos nt + e sin nt (cos nt) +5
9
; Gee de
=cos nt —e sin °nt ——. —— (sin ®nt)
1.2 dnt
=cos nt— 0a Et {a Snt — “cos nt}
ae ¢ mid é) cos nt + © cos 2nt + Bia cos dnt (71)
2 8 2 8
(3.) To expand the radius vector in terms of the mean anomaly.
" =1—e cos u
a
2 3
=1—ecos nt + = 5 (1—cos 2nt) —— (3 cos 3nt—3 cos ne)
mee Ub “—(¢-%) cos nt — ©
(4.) To expand the mean anomaly in terms of the true.
8
cos 2nt — ae cos 3nt (72)
18 to
dt = rd = Ra capes
h WV Ka(1—e?)
7r°d0 RS dé
as meee of a) 7 Oi pereety
=(1-$¢ >) ( (1 —2¢ cos 6+ 8¢? cos *6 — 4° cos °0)
= 1—2e cos 6 + 3? cos 20 — 4e* cos oe e? + 83e? cos 9
=] —2e cos 045 e? cos 20 —e° cos 30
, SRT er.
= 6—2e sin Oe e* sin rig sin 30 (73)
To expand the true anomaly in terms of the mean. We have above
d=nt+e(2 sin g—" e sin 20 = sin 30)
=nt+ep(@) .
=nt+ ef(nt) + e { (nt f) Yat é a’ { (nt)}s
1.2 “i ut 1.2.3 dn tO)
=nt + (2¢ — =) sin nt al e? sin 2n¢ + ey e® sin 3nt (74)
4 4 12
102
To expand the true anomaly in terms of the eccentric.
: a ; 1+ 5
In equation (66), writing for convenience i =k, and putting the
whole in an exponential form
22-1 _ | e2uv-1_y]
e26V—1 1 a 22-1 a
whence
ae (h+ Devo Be) wi pPeee.
(k+1)—-(k—lhev 1 — nervi—
let |
. — Z—HX
2u/—1 —. hae 20/—1 —
é =z é inc
whence taking logarithms
r
oe Lie
207 —-1=]
y °8 ‘a —Az
2 3
= Nog a a ee Ne
e hietins®
wee ae 2
= 2uW —14+27 —1(rsin Qu+ > sin 4u+ ...)
Ste Mee Mae
6=u+2Xrsin ute sin 2u +2 — sin 3u (75)
é
1—-/1—e
The problem of disturbed motion will be much simplified by the
following theorem.
The differential equations of the motion m expressed in rectangular
co-ordinates may be written thus,
C2 yer ay
ee ad
dy , wy _ dh 7
dt2 a ye mld dy (76)
ee ILC eee
ets 7 a)
or as they may be written
dc dR. d’y, VaR eae ae
GROW de ) “dei “pee ee ee a
< ¥ y a: ~ Sra ye (77)
103
if the co-ordinates be changed by the formule
E=ar+ by+cz, n=ar+dyt+ez, C= ax + by + c’z (78)
in which the nine direction-cosines are subject to the six equations of
condition
aa De ce I a’a” + Bb" + ce” = 0
a? + $74 ¢2=1 aa+b’b 4+ cc=0 (79)
v?4+wp?4¢2=])1 aa’ + bb’ + cc’ = 0
whence also
Pape taher ee (30)
then using the factors a, 6, ¢; a’, b’, c’; a’, b”, c’, respectively, in the
equation (77), there results the following system,
VE pap ee pe _ aR
dt? dx faa oe |
E
ney lk re ARM, sik
d@ de dy ly de (81)
n
Gre m dR _ ya _ ak
= dt? dz dy Gz ape fh
c p°
where
p” =e 2 4 n? a e (82)
But
dR _ dR dx , dR dy , dR dz
dé ~ dx dé‘ dy d&~ dz dé
aR gh ak
dR _ dR a i dy , dRdz
dy dx dn dy dn dz dy
a a]
aR, dR, AR SS)
me Be ME en Pore
dR _dRdzx , dRdy , dRd&
dt dxdg& dy df dz d
ak ya, AR
Sei, it hoes
hence (81) becomes
@— dR dn_dR dg_dR
d& dé dt? dn dP dt be
ee mo er ees Fe 84
E n C p e
But since three of the nine direction-cosines are entirely arbitt
they may always be chosen so as to destroy the third of the expressi s
in (84), so that there will be only two equations of motion remaining.
Hence if the moon be first supposed to move in the plane of the eclipti c
the results so obtained will be true, when the inclination i is taken into
account.
105
On the Calculus of Variations.
],.—FUNDAMENTAL FORMULZ.
By means of the Calculus of Differentials the properties of given
functions of any variables may be investigated, and the changes of the
former traced through values of the latter differing from one another
by insensible quantities ; or, conversely, the values or a system of values
of the variables, where certain properties of the functions (represented
by given combinations of their differential coefficients, or of the dif-
ferentials of the variables) have any proposed values, may be found.
There exists, however, a more extensive class of problems, in which
the forms of the functions are themselves unknown, and consequently
become the subject of investigation. The method by which problems
of this kind are solved is called the Calculus of Variations.
This Calculus is analogous to that of Differentials; the operations of
the one are in form exactly similar to those of the other, so that certain
corresponding points in the two may be advantageously compared in
the course of the present investigations. In order to understand this
more clearly, consider the function
OC oh 0g we) 0, (1)
in which expression
ean oUt eis) Oh) fon, (cin 5 Zytah ew) 95 1 af (Ly Yy Zy 41+) oh ot, ooh 0(2)
the symbols ©, fi, fi, fs - -. representing any functions whatever.
Now the functions treated of in the Calculus of Differentials (or, that
which is nearly equivalent, the Differential Calculus) are such that
their forms are always given, and consequently in this case
Unrest ee on 10 as Chea ae (3)
where a, 6, c, . . . are constants made variable only in certain classes
of problems, such as envelopes, variation of elements of motion, &c. ; the
object of the former class (which is most to the present purpose) being
to eliminate w, y, z, . . . by means of the equations (3) and their dif-
ferentials, the original variables x, y, z, . . . being considered constant,
and the original constants a, b,c, ... . variable in the differentiations.
P
106
But even in this case the form of the functions represented by (2) is
known, and it is possible to pass only from one individual to another of
the same class, and not from one class to another: Thus, for example,
in analytical geometry any surface under consideration is in the case of
envelopes, or ultimate intersections, supposed to change form according
to a certain law, the class to which it belongs remaining always the
same. As however in the Calculus of Variations the forms of the
functions must, when determined, hold good generally, 2. e. for all values
of the variables 7, y, z, . . ., it is clear that the changes of the two sets
of variables v, y, z, ... U,v,W, ... must be simultaneously taken
into account. In this case it will be convenient to introduce the new
symbol 8, and to represent the changes or (as they will be hereafter
called) thewariations of 0,1/ 42, et teil, ae SY
ox, Oy, 62, . .).
bu, Be OW, 5-0 i (4)
in the same way as their differentials were represented by
UO ss eats
du, dp, HN 4 } (5)
and similarly their successive variations by
OU, Oly Oreh ste ee
O74, "O"U, OW, 2 ne i
in the same way as their successive differentials were represented by
G32, i nl altel cs
a Gs, Goi. ‘I (7)
So also proceeding according to the rules above given, and represent-
ing the partial variations of any function ® by the symbols
OAD, 0, Ady” OD, te se (8)
in the same way as the partial differentials are represented by the
symbols
(6)
dO, One (9)
we shall have as the expression for the total variation of the same
function
Hyg. ES OMe)
80 = gt E yt “oe +... (10)
a formula analogous to the differential
Be 4 WO 7 ee (11)
a
dO
dx dy d.
107
Before proceeding further it will be well to consider a particular case,
in order to throw as much light as possible upon the nature of variations.
Consider the case of a surface whose equation is
i (aay 2) 10 (12)
Now by means of the Differential Calculus we are enabled to pass
from one point of the surface to another, and also to discover certain
geometrical properties which it possesses. For instance, the direction
of the normal at any point may be determined by any two of the
equations
cosa= 6 C08 i= ' cosy = GS (13)
where
2 TW! 2 WY 2
* (Ga AG aes aa
and by assuming arbitrarily values for two of the variables as
Be aexer ay, OL 2 7 (15)
the remaining two can be determined by the equation
EB ra¥, 2) 0,0 0r h(x, 41.7 as 0; or F(x, y, 2) (16)
Whence we can find the values of the angles «, 8, y by means of the
equations (13), im which x, y, z are to be written for 2, y,2 after the
differentiations have been performed. Conversely we may find a point
where the normal has any given direction (provided that the surface
admits of such), that is, where two of the direction-cosines have any
arbitrary values, such as those given by any two of the equations
€08 i= beCOd O =, COS. = 7 (17)
by means of the equation to the surface, and the corresponding two of
the system
1 d,F Leagan 1 d,F
CO aMmcR ON Kuma Cigale (18)
Suppose however that we have a more general case, and consider any
equation of the form
F(u, v, w) = 0 (19)
where
U=fi(%, YZ) v= f(t YZ), w= f3(4s y 2) (20)
Now if uw, v, w be determinate functions of x, y, z, whose form does
not change, the case will be precisely similar to the former one ; but if
Py2
108
the forms of u,v, w vary, we shall still proceed in a similar way, which
will give rise to the three expressions
| &F 8F 8F
Sx’ by? &z
or any other that may be required, ‘Thus taking the case before con-
sidered, the equations (13), (14), and (18) would retain the same form
as before ; but, by the principles of differentiation,
6,F on 6,F 6,u a OpLd,0) BOp tt j0 we.
ox ou oz ov Ox ow ox
oF OF OOF 0,0 0,,b 0,%
jy bu dy dv dy dw dy
Opt 20, F628 OMe OO ew Es OL
oz du oz dv 6z dw 6z
So that even in this case we have only three equations as before, by
means of which we may determine u,v, w; but since these quantities
are themselves functions of x, y, z, we shall not (as in the former case)
have determined a point, but a locus; which indeed is as it should be.
Since moreover by means of the operations indicated by the symbol
82 new quantities have been introduced, it will be allowable to assume
n relations involving them; let these be that the ratio of the partial
variations of any function with respect to any variable to the variation
of that variable are equal to the ratios of the corresponding differen-
tials; a definition which will have the double advantage of giving a
definite idea of the nature of the variations, and of leaving their absolute
magnitudes as arbitrary in this Calculus as those of the differentials were
in the Differential Calculus. The analytical expression of the above
definition is
12 40 40
dx dy dz
S07 50m hen eee (21)
ox by bz
or writing
8,0 8,0 8,0
ee OF “YY Boe Je DWV Oe 25
d,0, d,O, pea O ; Co
dx dy dz
Us lay oo ae ° a Ls (23)
so that the expression (10) will become
10.80 agi 0 ( TGA) Bosse) y
so that
a I(v+1) I(n+1)
a = — )v ris ; . > —j ra—?t
oee yj+1) Pv—j+1) P¢@+1) Cm—i4+}) [Ons]
aro = 23 (—)5, D(n+1) [~+1) OTe
(i+1) P(n—i+1) P(j+1) P—j+])
111
but
LO, =f {et (n—i)de+ (v—j)8(e+ (n—s)dr)
+ yt (n—i)dy + (vj) 8(y + (n—7)dy)
+ 2+ (n—i)dy + (v—j)d(z+ (n—i)dz) + ...}
= f{et+(v—j)Se+(n—D%(2 +(v—j)de)
+ yt (v—s)ey + (n—i)8y + (v—s)dy)
+ 2+ (v—j)d24+ (n—2)8(2+(v—s)dz) + ...}
= [52] n-1
hence also
SaO = PSO, (30)
as a particular case of this, which is in frequent use, we may here notice
mamas Yom b
and consequently
sf = fd (31)
I].—GEOMETRICAL INTERPRETATIONS.
In the case of three variables some of the above formule admit of
elegant geometrical interpretations. The three variables 2, y, z will
then be the three rectangular co-ordinates of a point on the surface
represented by the equation
Q(z, y, 2) =90 (1)
Then writing for convenience
dO OMe 1d. One
Ce asp i det
the fundamental formule give
Adz + Bdy + Cdz=0
Aéz + Boy + Coz = 0
whence
A B C
dy dz — dz Sy dz 8x — de Sz dx dy — dy dx
Then in the same manner that
dx dy dz
ce a
(where
ds* = dx? + dy? + dz’)
represent the direction-cosines of a line joining the points (2, y, 2) with
112
the consecutive point (7+dr, y4dy, z4+dz) on the surface, that is, a
tangent line ; the expressions
dx by bz e
BOR 6)
(where
és? = 62° + dy? + 62°)
will represent the direction-cosines of a line joining the point (2, y, 2)
on the first surface, with a consecutive point (v+%2, y+%y, z+%z) on the
varied one. And the formulz (3) show that a transition is made from
a point on the first surface to one on the second lying in the tangent
plane to the first at the point x, y, 2; and that the normal to the first
surface is the intersection of two planes whose normals are the two
lines whose direction-cosines are (4) and (5) respectively. It is of
course obvious that the general theory cannot give the direction of the
normal to the second surface, as it would then determine the nature
of that surface, but merely shows the law according to which one
surface is arrived at from the other. Various other assumptions might
have been made instead of (21) of Sect. I. if they had been thought
convenient, such as
dx da + dy dy + dz 6z2 =0
in which case the line of transition from one surface to another would
always have been perpendicular to the tangent line to the first surface
at the point of starting. Or again,
Ue a dy weeds
which would have given as a result that the two lines which in the
former case were perpendicular should in this one be parallel, which
would be a particular case of the form which was actually chosen.
In the case of two variables, that is of plane curves, (3) becomes
Adz + Bdy = 0
whence
dr © by) de 62
which is equivalent to the third assumption in the case of three
variables. This result might of course have been anticipated a priori,
as in plane curves the tangent plane reduces itself to a straight line.
dx _ dy oe dy _ oy
113
IIJ.—on THE VARIATION OF A DEFINITE MULTIPLE INTEGRAL,
Consider the multiple integral
Q=f fi fi -..%...biyd (1)
where ¥ is any function of the variables
ian eerak es (2)
and their differentials of any orders
re Gere eae Meee gets Cr tye Ge ney ke! « (3)
Now since, after the integrations indicated in (1) have been per-
formed, © will be a function of x, x’, y, y’,z,z,..., and since Q is a
function of Y, which although a function of the variables (2) and (3),
and consequently also of x, x’, y, y’, Z,z’, . . ., 18 still independent of
the latter set, we may at once consider © as a function of
VR BAO od Be aa oem (4)
hence by the principles laid down in the first section
SPOT LOtee dolented (0)
BT eR pe OT Sy eR! te
SMTi ah dant i
oa ON LG
x OX 50 : eon
eee a a ea
In order to be able to exhibit this expression explicitly, we may
remark, that after the integration with respect to x, Q becomes first a
function of x’ and then a similar function of x but affected with a
negative sign, and similarly with respect to the other variables. Sup-
pose then we represent the value of any function of x, ¢(x) for instance,
when «x receives a particular value x by the notation
| Lp(a)
so that the equation
os) = | $l) 6)
always holds good. ‘This being premised, the equation may be written
in any of the following forms:
D=T df? fl. 0. .dedy— Td ft fT. dedy 1
afi fe .T..dedr— | dy ff? .. 0. Aetht t (7)
STO SRST Pi cdyde— TO ff. 0. dy de |
eoee
Q
114
It may also be remarked that
dx0yy a Od eee aad en |
= fefeafet oe ... dz dy dx t (8)
Lp ph fis ST eRe |
consequently, if the other differentiations indicated in (5) be performed
by means of the expressions (7), the final form which the value of #0
will take will be the following :
SSUES dz dy de
TSS EP delye TST SE Te dea
+ fT, To. dedsdy — fr" a 1.0... dz du by
S eee .. dy dz &z ae ae aa .. dy dx &z
It is ne to be observed that the expression for 8Y may be thus
written :
se dT
éT = Tse a ie ai od—“x +. += be
my ae , 4a
+ ty ody +-., ih gearrty od”—sy +. “a”
i APL ASL
+ ears tf he - + mary OF ete + Ob
te a
mat a =m dT =n dT
= 2,29 farily parti, Oe t+ =0 pa, ee yt ,_ 0 qr—ttiz dn'8z +
where the symbol = represents that 7, s, ¢,... receive all values requisite
to make the quantities
l—r, m—s n—t...
coincide with the various orders of differentials involved in the
function Y.
Now it is easily seen by successive integration by parts that
ft fara so )e-Idt-1 Ld" + (—)-" f'd'"Lba |
JS Maney ee Sa ve )ede Md tarby ale (— yr f'd"—Mby t ( l 1)
J Naz = §(—y'd Nd" 782 + (=) +t fd Nbz J
115
hence, using for convenience the following notation
oaex?
| P(x) = $(x’)—$(x) (12)
~t=X
the general expression for 8 may be thus written
i q “4 a [Sea — ert a abe] + Tox i a dZ dy |
i x’ y= ie ‘Chine s ol Jo—l any eS ded
x ats 2a om ) eC patly y | + Toy Zax
vy! ih so fhe dv ra |
Y, op y %=z2 "ise pid =] ea} d (Ga,)4 Oz +- Toz ae dy dx
(13
f af his ae stl a 3S gon) | |
Hse)
+> (—yrar+( 5) |e +.. |
IV.—cONDITIONS OF MAXIMA AND MINIMA.
The expression (13) of § II. is of the form
@ + fib = 0 (1)
in which the term under the sign of integration by its very nature
(since it involves unknown functions) cannot be integrated ; but this
expression must hold good whatever be the form of the unknown
function ; 1. e. if & becomes successively ®,, ©, .. ., in other words
the equations
© + (dd = 0, © + {P\dh = 0, 0 + {[Pdp=0,...
must hold good; which obviously involve the conditions
@ = 0; Pr 0; (2)
Hence, in the case of a maximum or pee aa we must have
fk oo Ae 23
cif ae ee, 6 a ca ae —r+1. ae ‘bx a3 ]+tet. 2 dz dy
x’ y==y s=m ‘Soran eee :
+f. st is —)ege- (sony ya ey | + Tey } . dz da
eC - aS pa
Sad ye az seis ma —)*d (ex )a : bz] +78: } dy de
Q 2
| oy
116
and
2 (ya GS) | |
Spade [ya (aa |
s=0
nelore Gtr
Py
the latter of whicheequations separates itself into the following system
iere Gay) sola eae
S(t =0, 20.
Suppose that besides the function © the variables are subject to —
certain conditions, such that 0 is to be a maximum or minimum for given
values of certain of the functions of the variables, as
(5)
Usviwe ce (6)
and let the given values be represented by the equations
bey oe VV at eee (7)
then whatever supposition makes © a maximum or minimum, 1. e.
O10 (8)
must also make
6U = 0, 6V =.0,.5W = 0, .. .«- (9)
But equations (3) and (4) are equivalent to the following
d0'g dO gD ih
ic Tage, Fin OF ee : |
dU dU
dx dy
dU
6 — ee = |
Yate ee 0
dV dV dV = t
de? ed ae ,.=0
(10)
dW, , dW, , dW
are Clg Sand aa a
Now by methods exactly similar to those employed in the differential
calculus will give us by means of Lagrange’s theory of indeterminate
multipliers the following equations (in which 4, »,»,... are the mul-
tipliers ),
117
dO dU (dv dW
“th tee t+ et |
ah) Wale oad le dW
Soe te Need re i Tee
dy dye’ dy "dy a (11)
dQ”. dU dV dW
Xr eo ae
dz t dz he ea | ae ‘i
or multiplying these by 4, dy, dz, . . . respectively, and adding, the
general condition becomes
6Q + OU + wsV + 7dSW4+...=0 (12)
that is
6((O2+A0+pV+yW+...)=0
the usual formula for the investigation of Isoperimetrical Problems.
It remains for us to determine the conditions which will show whether
the result obtained by the above rules be a maximum or minimum; for
this purpose it will be necessary to have recourse to the second variation
of the function ©.
The conditions relative to the case in which the limits are subject
to variation presents some difficulty, we shall therefore for the present
content ourselves with determining the criteria when the variations of
the limits are not taken into account ; in this case we have the principles
already laid down,
xe vy’ oz! pal dv ey gel dv is
80 =f f f.--[2 Deal aN an tne
t—n dT 4,
Lp aeseris is d”—6z + a
(13)
so that the second variation may be written
=f fi SP nego 8)
eam ET m—s
als 2 a0 Gravee, (4 dy)?
t—n PT
2-0 gat, (
&T
ele 2s, Pty ply
r=l da’?Tt
o>. 4, Day eri diol
s=m a7t
2 ie (or tly
+..|
doz)"
de—+l§z d+ 18x
dt 6x dat Oy
|
|
dass) Od 2 OZ | (14)
|
|
118
that is of the form
JS (Aw + Bu? + Cw? + ... +2Fow+ 2Gwu+2Huv +...) (15)
Now if the element under the sign of integration remains always of
the same sign, the criteria will be found in the same way as those
relating to ordinary maxima and minima of many variables. But if this
be not the case, the element in question may be divided into certain
parts, one of which remains always of the same sign, and the remainder
are integrable; the integrals of the latter will contain the variations
dx, dy, 0z,... or their differentials as factors, which will vanish at the
limits; the integrals themselves will consequently also vanish, and the
whole will be reduced to an integral, the element of which, under
the sign of integration, remains always of the same sign. In the case
under consideration the part integrated, with respect to x, will be
generally of the form
au? + bv? +cew? +... +2fow+2Qgwu+2Qhuv+...
that with respect to y,
du +e? + cw? +... 4 2f'vw+ 2q/wu+2Qh’uv+ ..-
that with respect to 2,
awe + + we + we. FOF" w+ 2Qq'wut+Qh'uv+ ...
and soon. So that tiene the first of these with respect to sz,
the second with respect to y, the third with respect to z, and so on, and
subtracting the results from the element of the second variation of ©,
that is from (15), and writing for convenience
dada’ dal” db db’ db” de de’ dc’
= —_ — B— ——————....
Si dx dh Wy de cn dz dy hy) ode ara dx dy dz |
i Oe eee dg_ df dy” wy —7_dh_ dh’ dh” t (16)
g=P-f- FT"... 6-4. y= =
the element becomes
Ay? + Uv? + OQw?+ ... + 2 ffow + 2Ghwu+2WHuv+ ... =O (17)
which must remain always of the same sign. To determine the con-
ditions that this may be the case we will follow the course taken by
Lagrange in the Théorte des Fonctions Analytiques. Suppose it must
remain aways positive, it is evident that, for the contrary case, it will be
sufficient to take the quantities 4,%,@,..., ff, €, H, . . ., affected
with a negative sign. Since then it must never become negative, it
Ig
follows that it should have a positive minimum ; and conversely, if it
has only positive minima it can never become negative. It remains
only to find the conditions that the quantity under consideration shall
have all its minima positive. Following the method of ordinary maxima
and minima, we shall find the first and second differential coefficients of
the quantity © with respect to one variable, as uw, and equate the first to
zero, and suppose the second positive. ‘That is
Au+ FHv+Giw+ ... =0
goa ty
Substituting the value of wu, deduced from the former of these
equations in the value of ©, and writing for convenience
a-u-B, g-c-f, a=
ope (19)
B=f-A a
(18)
the expression for © becomes
O= We’? + SMw? + Jr? + ... +23Bert+ 2@@rv4+Q@Mow+ ... (20)
In the same manner as before, finding the first and second differential
coefficients of this expression with respect to one variable, as v, the
conditions become
DY’ 4+Rwt+@r+...=0 \ (21)
L>0
Similarly, the expression for © may be again transformed by means of
these equations into another of the same form, such as
0 = Aw?+ P74 dAse+ ... QXrs-+2QWsw+QWwr+ ... (22)
and the conditions again become
GAw+ Zr+Ds+... =0
23
Gt > 0 j )
and so on. Now it is easily seen that the last of these transformed
expressions, which will contain only one of the quantities wu, v,w, ... .
and which will be of the form e
will be the minimum of the quantity ©; hence the conditions that ©
shall have a positive minimum will be
A>0, H>0, G>0,... B50 (24)
and since the equations which determine u,v,w, . . . are of the first
120
degree, we may conclude that that will be the only minimum; hence
the problem is completely solved.
It may also be remarked that by the above transformations the value
of © becomes
(epee
+a(o+ ROL Grt... es
“ Gt ( w + wb i.
from which expression also it appears that the conditions (24) will -
ensure that the element (17) shall always remain positive.
121
Problems in the Calculus of Variations.
The following problems are collected from various sources.
(I.) To find the shortest line in space from one given curve to
another.
We must find
Jds = a minimum. (1)
dfds = 0 (2)
but ds? = da? + dy’? + dz* (3)
Sqs — (dou t+ dydoy + dzddz (4)
ds
. _fiea= [ode + Loy 4S be]
ay!s!
es Uf (deat gt 5 yd hs Sed) =0
where wz’ y’ 2’ are the co-ordinates of one curve, and 2” y/’ 2’ those of
the other. And since the curves are entirely independent of one
another, the first part giyes
(5)
boll dy” yw, dz” Me
ee + ay + 7% =0 (6)
dz’ dy’ dz’
Gy tt t+ aa ty + 82 = 0 (7)
J/ /} //
Now, - a are the direction-cosines of a tangent to one curve,
and. ce a aa those of a tangent to the other, therefore (6) and (7)
show that the required curve cuts both the given curves at right angles.
The second part, since tx, ty, 82 are now absolutely arbitrary, gives
dx d. dz
va eS doi Ce
d= 0 : d= 0 (8)
din" CM ey dz
a aoe at (9)
which shows that the line must be straight, its equations being
122
(II.) If it were required to draw the shortest line from one given
point to another, the first part would equal zero of itself, and the second
would show that the line must be straight.
(III.) To find the shortest line on a given surface between two given
points.
Let u=0 (1)
be the equation to the surface.
Consequently, as before, we must have |
dfds = 0 (2)
and also
OF a0 + by + bz
— f° | ara + bya gy Y 4 bea | = 0
Now since 6, sy, ¢ must vanish at the fixed limits, therefore the
first part will equal zero of itself; and in order that the integral may
vanish we must have
ao ~ ) 8 +d(“)ey +a(4 = ) bz = 0 (4)
whatever be the values of 82, 8y, 8z, so that they satisfy the condition
that the point (x, y, z) is on the given surface ; viz.
wuz + 7pey + be = 0 (5)
(3)
consequently we must es
dO etn ee oe
ds PACs WIN higs
du ier ei (6)
dx dy dz
We will now show that the osculating plane to the curve is always a
normal plane to the surface,
Let
Q = dzd’?x — dzd*z
R = dad*y — dyd*x
then, the perpendicular on the osculating plane makes angles with the
co-ordinate axes, whose cosines are P, Q, R divided respectively by
VP Qa
P = dyd*z — dzd’y
;
123
Now,
qe — dsd’a—dad's _ ds*d?a — dadsd’s
dsp ds* a ds®
= a | (da? + dy? + dz*)d?x — dx(dxd*x + dyd’y + dzd*z) \
= - | dz(dzd*x — dxd*z) — dy (dxd*y — dyd*x) \
s
1
Similarly
dy _ 1 Paes bese
aY — qo Rae Paz |
dann io
ace = at Pdy Qae |
and therefore identically
pat + Qa + RaS = 0 (9).
and also by (6)
BtoF + Bea 0 (10)
i.e. the normal to the re 1S st cea to a perpendicular to the
osculating plane of the curve, i.e. coincides with the osculating plane.
Q. E. D.
(IV.) Find the curve which of all that can be drawn between two
given points contains between the evolute and the radii of curvature
at its extremities the greatest area.
Area = 3/pds
where
Ae Oe CB
Cuca dxd*y — dyd?x
hence
S(pdds + dsép) = 0
and
dés = © dB a5 Y aby
Sp = — £ (dad*by — daddy y — dyd*Sx + d?yd8x) + 70 dds
hence
if [4pd8s — © (dud*by — d2xddy — dyd?8x + dy yd) | = 0
R 2
124
or
mp pe p
wit (ae TT pa’y) abe + £ (Ady + Pde) aby
+ P dyd®8x — iw ded: ‘ty| = ='0
hence, integrating by parts,
[2 (4ar— dy) - a(f ody) |e + Fe (4ay+ P@x) + a( fa) |éy
+ fe argue — " atnty
-f [1 aCe (4de— Pa? hy a (Fay) } be
e Pp ( & ] am
f | (Leas £ Shy ae ) fe i
hence
p P (Ade —P a? v) = a(f sy) = = 2a
p P (Ady +2 Le ‘a) " a(£ dx) = 2h
or
# & (Ade—22 “Paty) — dyad (€- >) = 2a
P (Ady + Pate) e dea(?-) — 2h
hence
2
Qpds + Fa (Tady—d?yde) = adzx + bdy = pds
2
that is, taking x as the independent variable,
Tee pay &: y
a dx dx?
Tey A
+E) GB
or integrating
x+b _, dy
ae ele Li Lae
t—-xX = 2 FAL TE
ree
which is not integrable ; but if a=
b dy” as \/ b :
Aa . ee 1 ie
*@—x aes et da 2¢—x
the equation to a cycloid.
125
(V.) To find the relation between x and y, so that
S (2? +7’) Eds may be a minimum,
changing to polar co-ordinate
oe V dr? + rdé? = 0
ce jun lords + 7dds = 0
dae. eden Sie et © bdr +70 by +n 8dé
cs integrating by parts, we ee
oer rg rit | +f{ (rrr tds + i oo) br— a(rme) 90 | =0
ete the assigned limits,
nds mi pad(s =) a (1)
va af (2)
and
Ss
from either of which we obtain the equation
hdr
dll no Bo ae ch
r A pnt? __ f2 pinto £2 ea 9
which being rationalized and integrated give
m+! = h sec {n+1(0+C)}
whence we know the relation between w and y.
If m = O we have the equation to a straight line.
(VI.) To find the curve of quickest descent from one given curve to
another.
ey! ds
VQ Vh—z
eae
fla bth oS i } =0
fiz 7 (S ae + dy +2 Ze) ths goz } = 0)
therefore integrating by parts,
1 oY 5
Vh—z aa si a Ay
tf 12 C7 Viz) ‘ we 7)
da} it
d et a ee nae ee 3. g]ee} = —_ 0
"i [ ds sity i fae
126
The first part shows, as before, that the curve of quickest descent
must cut the given curves at right angles ; from the second part we get
ax * | dy ]
d (F =) an) d{“Z ) 10)
ds Vp—z ds V/h—z
da © LW Wty WL Pes,
ds /j,—z ds W/h—z
Y = 2 a constant quantity ;
consequently the required curve lies wholly in one plane. Let this be
the plane of x z.
d. ss
_ = a.Vh—z
are mney eee
da® — a*(h—Z) a dx?
dz* _ 1—a’h+a’z
“so a cycloid.
If the motion commences from the origin, instead of
h—z
we must write z
We have not written down the limits, as the method is the same as in
the other problems ; and the problem
(VII.) To find the curve of quickest descent from one given point to
another, being only a particular case of this, must be treated in the
same way.
(VIII.) A particle is in motion under the action of a central force,
find the nature of the orbit described by the principle of least action.
Jods = minimum
Now
v = C— 2fFdr
v= VW C—2/Fdr
as = /,2 dy*
r + 7 00
a u=f\/ (C-2/Far) (24) a0
127
and taking the variation with respect to @ only, as the coefficients of each
of the variations 87 86 must be put = 0, we get
iu =0 = "| \/ (0-2/ Bar) (8 +) 6
__ dr? doo
+ /C—2fFar. dé Tae st
therefore integrating by parts
dr?
(C-2/Far) (8+ a8) — de’ _ \s0
/( r)( Ve Ve me
—f{ (J (c-2sFar) (24%) — a VC—2 fFdr VC—2 fdr dy?
ie) Vee es a) + fF =0
e d{/ (c-2sFa) (14%
de?
_VC—2fFdr ne be ef
ry rr de?
+ GF
dr? are far & dr®
+ Pre se. =h
cae ae ose
Te
"6 PV C—2 fdr =h Ve +@
C~2/Fdr = h? dr?
3 Pde
Vv C—Of Far (7*
Or let roma
u
c+osP = it(w+ e
which may be applied to eee cases, and is the usual formula
(1X.) A string is stretched over a given surface; show that it will
place itself along the shortest path, if the resistance of the surface be
the only acting force.
Let
be the equation to the surface.
128
Let X, Y, Z be the resolved parts of the forces acting at the point
P(z,y, 2).
Let
T = tension at P
consequently the sum of the forces acting on PP’ parallel to the axis of
x will be
s pdx! pdx
Sf Xde+ T'S -TS = 0
and also when the element is indefinitely small the three components
will be
1 \ es
Xd+d(TS) =0
Yds+d @ 2) =0
Zds+d (‘T ze) =i)
and in the present case
R = pressure or resistance of surface.
Let
MuAh deve’. ep dine
v={Q+Q+Q hb
a |
a
.X=RZ
du
d
Y=R¥
du
z-n& |
v
du
dy dy
Ray ds+d(T. 7) = 0
du
RS w+a(T%) =
du
RG a+a(T.2) = 0 |
|
Oo
EE
129
Multiplying by ‘a a a respectively, and adding, we get
dy , dy , dz ie dz\ _g
aT+1(F a+ Gass iam
but
dx? dy? | dz
=]
ds? a ds* " ds?
dx ,dx , dy ,dy , dz dz
o. d~+ —d—=0
‘ds ds %, ds ds ds ds
dT = 0
T = constant.
du
dx dx
aes ds = — Td
du
y yy
v ds= — Td me
du
dz dz
v ds= — Td, |
dx dy dz
a et Oe
: ds i ds ds
du a Pe ise du
dz dy dz
which proves the proposition.
(X.) A particle moves on a
given surface, and is acted on by no
forces, but the resistance of the surface show that it will move along
the shortest path.
Let u = 0 be the equation to the surface,
R = resistance, and «fy the angles which it makes with the axes.
X = Reosa
v={(#)
du
Let
dx
Cosa = v
Y = Reosf Z = Reosy
du du
7 dz
Cos B= Cosy =F
130
du
dx dx
qe hy a)
du
dy py
af a ie
du
d’z dz |
qa ky =9
Now,
dz _ de ds
Ut umtls eat
dx _@xds | d’sdz
dB, \dsidi* aval ads
but since the resistance of the surface is the only acting force, and this is
directed along the normal at each point, therefore the force along the -
tangent plane is equal to zero.
d’s
ae
oc
de
ay _ «) |
eee Coa
dz 2
ae i )
whence
utes it ae Ea are the direction-cosines of the radius of curvature of
the curve described by the particle relative to the three axes, which
131
radius lies in the osculating plane, therefore by what has been proved in
preceding problems the proposition follows.
(XI.) When a point moves freely in space or on a curve surface,
under the action of forces X, Y, Z, such that Xdx + Ydy + Zdz is a
complete differential, the path the body takes is such that
Jods is a minimum,
the integral being taken between the two points at which the motion
begins and ends.
Gene Ri ede
deh ih Miia
Cy. Te x, dE
cae a Mh as [
Ca Ry, dF
faniiag a MLS de;
Let the equation to the surface be
F(zyz) =
di dif dF
and ae dy dz a
ICs ey +(¢
dz d? dy da
ramos eats AY eh aR ait gti d's = 2{Xdx + Ydy + Zdz}
dt dt? dt dt* dt dt?
dx? 4 dy’ ee One » a yl
d? ae * ae” = 2| p(ay2) le :
if Xdz + Ydy + Zdz = d d(2yz)
dfvds = f(dsdv + véds)
CL by + bz z |
and .'. dvds = vovdt = ale
me ie or abe a iby +& dz 35,
dfvds =/ | Sfbe4 & oF eet “Yay + aby + © o82+ - az }
Te roe + Vey +b: \
= oat +5 Ley +S = between the assigned limit,
‘ es the variations of z, y, z at the limits =
. fvds is a minimum.
S 2
132
(XII.) Find the equation to the curve along which, under the action
of given forces, a body will move from one point to another in the
shortest time.
ds = vdt as. fe is to be a minimum.
Sif Bap
v
6 ds a ds.bo a
v v
Let X, Y, Z represent the accelerating forces acting on the body.
Px Py _ dz
da ON dee ae
vdv = Xdx + Ydy + Zdz
vov = Xdx + Yoy + Zéz
and dds = = © bd f- au ody + “de
oe les 2 +i aby iat Ba, cy =%
Lie atte} fab ie Ha Biya)
oes + Ydy + Zdz) = 0
¢(= =) Sty ty
a(* @) +SY=0
1 dz ds
a(; =) tae Te
1 dv dz 1d eee
W ig? dade @ eaeas ie
ldu dyin lida eee
engi, a we -—=0
v? ds ds ni v ds* cf v
lide dz) Wz a7
Betufs tao — —0
~ 2 ds ds . iit we
7 aaa (Gal) eke reer)
2 2 2 2 2 2 2
alee {> 4 ay +ot 5 Ol xt avis 2h 4% 4+ Y247
~ oe ds?) ds? ds®_ ~— ds? vw ds yt
133
vt otlv? 9 2
= — Dy*_ + X? 4+ Y? + 2?
p ~ as oe i2 + Sais:
odv"
ds?
v*
tee APR 2 ok
o
= X?4 Y2+72— {Xa +e + Ca
eye | ae FON dx x dy Z dz \
mar { : R ‘ds 2H ds an =)
putting X?+ ¥7+ 7? = R?
and since = eh £ are the direction-cosines of the resultant of the forces,
dx dy dz
ds’ ds’ ds
angle between the resultant of the forces and the tangent to the curve,
and of the element of the curve, therefore if ¢ is equal to the
4
han R? sin *
= R sing
consequently the pressure on the curve arising from the centrifugal force
is equal to that which arises from the impressed forces, then time spent
in the description of the trajectory isa minimum. Q. E. F.
(XIII.) Find the curve of a given length, which being suspended
from two given points may have its centre of gravity the lowest
possible.
Take the axis of x vertical,
consequently we have
JS (@ + &) ds a minimum
df (x + kh) ds = f {dxds + (x + k) dds = 0
ie (w+ Wy | Hon + Loy | + f'[{-a(e42%) \ 2—a(eF7aZ) ay] =0
between the given limits,
d
i ds—d(z+ he =40 d(z+h%) =0
Oe dy
eae eee laa “7—)
t+ he ats x ha.
dx __a+s
dy ob
The equation to the catenary.
(cx +h)? = (at+s)?+0?
134
If the curve was to hang from some point in a given curve to some
point in another given curve, the former part of the resulting expression
shows that the curve cuts both the given curves at right angles.
(XIV.) To find the form of a curve which a string of a given length
assumes when the enclosed area is a maximum.
Sf fydx + hds} = 0
S {ya ae + h(a + dy) | — 0
(y+ 1D) a0 4 Yay sf [ fae — 2a) | by — { ay + 2a | ae =
i =o +a i = —y—b
= (etay + (y+by
equation to a circle, of which the radius equals k.
(XV.) Find the form of the solid of evolution of given contents which
exercises the greatest attraction on a particle in its axis, law varying
1
distance
DY Oe (tli eee + hy?) de =0
vA ( TERE y
otis veal
(e+ 9°)"
a + Qh (22+y2)? = 0
the equation to the curve which generates the surface of the solid.
ag
ee po na OO ce
eo
a
135
Note on Lagrange’s Condition for Maxima and Minima of
Two Variables.
THE conditions that
z=f(%y) (1)
have a maximum or minimum value are
dz dz
—)=0 (=) a 2
) 2 dy (2)
and the discriminating condition is
22 dz d?z \?
ee Ny 5S eee ee 3 aaa 0 *
du* dy? oe) Saige as (3)
in the first of which cases the value determined by (2) will be a
minimum, in the last a maximum, and in the second neither one nor
the other.
If in (3) we substitute
d°z d*z dz
de” apo” dedy° (4)
the condition will depend upon the form of the surface
Ey — OP = (5)
where 4 is positive in the case of a minimum, negative in that of a
maximum, and zero in the case of neither a maximum nor minimum.
Now (5) is the equation to a central surface of the second order,
having its origin at the centre. To refer this to its principal axes, we
have merely to substitute in the usual cubic
(P—A)(P—A’)(P— A”) —B*(P— A)— B”(P— A’) B’2(P —A”)2BB’B’=0 (6)
the three roots of which are the coefficients of the squares of the
variables in the transformed equation ; but since
Aaa. 0 ey et
> 3 7
B=0, B/=0, B’ =! (7)
136
(6) becomes
.(P4+1)(P?—1) =0 (8)
the roots of which are —1, —1, and 1, hence the transformed equals
becomes
2+y—2=— 6 (9)
which is consequently i in the case of a
minimum, a hyperboloid of revolution of two eifbets,
maximum, . + 4 one sheet,
neutral result, a cone.
a >
my * vF =i ,. +
’ : ». we
ud Gr Phere e , of
a ;
m ’ .
, ' 7
‘ @
. \
‘
* 1 s
“~
,
‘
'
'
<
‘
'
'
‘
t
“)
1a
74%
'
7
WHYS
AA
VATS
ary > *
_— :
ree ‘ t
r -
~_ ,
oo ao at
-)
; ' :
‘ .
ee ’ }
r . ‘
‘
i
. i hI
' +
'
i
‘ —?
*
g
s
' +
.
'
.
‘
.
.
f
i
’
‘ 1
ra)
i a}