E258:

Prin ipia Motus Fluidorum.

published 1761 in
Tom.

Presented to Berlin A ademy 1752,

Novi ommentarii a ademiae s ientiarum Petropolitanae

VI ad Annum MDCCLVI et MDCCLVII, pp.

271-311, +1 page of

gures. See also E225-227.

Prin iples of the Motion of Fluids

Part One

1.

Fluid bodies dier from solids prin ipally in that in general their parti les

are not bound to ea h other, so these dierent parti les an be subje t to

very dierent motions.

A motion whi h is imparted to one uid parti le is

not so determined by the motion of other parti les that it ould not pro eed
in its own way. For solid bodies it is quite a dierent situation; if they were
inexible, their gures would undergo no hange, and the individual parts
would keep at a onstant distan e from ea h other; so that the motion of all
parts would be known, on e that of two or three points were given.

Even

then, the motions of these two or three points are not ompletely arbitrary,
sin e they must keep the same distan e from ea h other.

2.

If the solid bodies were exible, however, the motion of individual par-

ti les is less xed; be ause of exure, the distan e or the relative lo ation

of diverse parti les is subje t to hange. Even then, the manner of bending
obeys a ertain rule, whi h diverse parti les of bodies of this type must follow
in their motion, to wit that the parts that are subje t to the bending will
not tear apart, or pier e ea h other; whi h indeed will be ruled out for all
su h bodies by a ommon hara ter of impenetrability.

3.

[Fluids have an innitely large number of on eivable ows.

In uid bodies,

however, whose parti les are not joined to ea h other by any bond, the motion
of diverse parti les are mu h less restri ted, and from the motion of a number
of parti les the motion of the others annot be determined. For if the motion
of even a hundred parti les were known, it is lear that the motion whi h
the remaining parti les ould take is innitely variable.

From this it an

be on luded that the motion of ea h parti le of the uid learly does not
depend on the the motion of others, unless it were bound with them in su h
a way that it must follow with them.

4.

At the same time, it annot be that the motion of all the parti les of the

uid is bound in

no

way by any law; nor an

any

on eivable motion of a

single parti le be allowed. For sin e the parti les are impenetrable, it is lear
that no motion an take pla e where some parti les go through others, or
that they penetrate ea h other. An innite number of su h motions should
be ex luded, and only the remaining are to be onsidered, and learly the
task is to determine by whi h property these remaining possibilities an be

distinguished from the others.

5.

[Cir ums ribe the types of kinemati ows to be onsidered, from whi h one an be

pi ked out by dynami s. Before we an x on the appropriate motion when a

uid is a ted on by a for e, we must delimit those motions whi h ould take
pla e in this uid.
from those

I shall all them

impossible

possible

motions, to distinguish them

motions whi h ould not take pla e.

To this end we

must de ide the hara ter appropriate to the possible motions, separating
them from the impossible ones; when this is done we need to determine in
any situation whi h one of the possible motions a tually should be hosen.
At that point we must look at the

for es

to whi h the uid is subje ted,

and then the motion ompatible with these for es an be determined by the
prin iples of me hani s.

6.

[Restri tion to in ompressible uids. I have de ided therefore to look at the

hara ter of motions that are possible for a uid that annot be penetrated. I
shall posit moreover that the uid annot be ompressed into a smaller spa e,
and its ontinuity annot be interrupted.

I stipulate without quali ation

that, in the ourse of the motion within the uid, no empty spa e is left
by the uid, but it always maintains ontinuity in this motion.

After we

have theory suitable for uids of this nature, it will not be di ult to extend
it further to uids whose density is variable, and whi h do not ne essarily
require ontinuity.

7.

If we onsider any portion of a uid of this type, the motion by whi h

its individual parti les are moved should be so onstru ted that at ea h time
they ll the same amount of spa e. If this happens for individual parti les,
the portion as a whole is prevented from expanding into a greater amount
of spa e, or being ompressed into a smaller spa e; and it is just motion of
this type, in whi h the uid is onsidered in apable of either expansion or
ompression, that we shall take as

possible motions

. What we have said here

about an arbitrary portion of uid, is to be understood as applying to ea h

element of the uid, so that the volume of ea h element of the uid ought to
remain un hanged.

8.

With this ondition satised, we are to onsider what the motion will be

at the individual points of the uid.

For an arbitrary element of the uid,

we have to nd out the instantaneous translation of its bounding surfa es, so
as to determine the new portion of spa e in whi h it will be ontained after
a very small time period. The new portion of spa e must be equal in size to
the old portion whi h the element had o upied.

This equating of size will

fully hara terize what an be said about the motion. For if the individual
elements o upy equal spa es at ea h time, no ompression or expansion will
arise in the uid; so the motion will be ompatible with our ondition, and
we must allow it as a possible motion.

9.

[Resolution into two or three dire tions. When we take into a ount not only

the speed but also the dire tion of the motion at ea h point of the uid, it
be omes useful to resolve that motion into xed dire tions. This an be done
into two or three dire tions, the rst if the motion of individual points remain
planar; otherwise the motion should be resolved along three xed axes. Sin e
this latter ase is more di ult than the former, it is onvenient to start with
the possible motions in the rst ase, and when that is worked through we
an more easily solve the latter ase.

10.

[Two-dimensional ow.

Therefore I shall attribute to the uid ow two

su h dire tions, so that the individual parti les and their motions lie in their
plane.

Let this plane be represented by the plane of Figure 1 , and onsider any
point

of the uid, whose lo ation is referred to the orthogonal oordinates

AL = x and Ll = y.

Its motion when resolved along the same two dire tions

displays a velo ity along

namely

ln = v :

AL

lm = u,

, namely

and along the other axis

thus the a tual speed of this point is

, and its dire tion will be at an angle in lined to the axis

is

AB

uu + vv [= u2 + v2

AL

, whose tangent

= v/u.

11.

Sin e we are proposing to develop the state of the motion that applies

to ea h individual point, the velo ities

of the point

and

depend only on the lo ation

, and they are to viewed as fun tions of the oordinates

and

. We an therefore write a dierential relation

du =

Ldx + ldy,

dv = Mdx + mdy,
and sin e these are to be omplete dierentials we must have
and

dM/dy = dm/dx

the dierential

dL L
of

dL/dy = dl/dx

It is to be noted in an expression like

dL/dy

is to be taken only from the variability of

a similar way in the expression

arise if only

dl/dx

the dierential

were to vary.

1 The Appendix gives the omplete set of graphi s.

2 Euler does not use the partial derivative notation

/x.

dl

that

, and in

is su h that would

12.

[Meaning of dierentials. Care should be taken that in fra tions of the

type

dL/dy, dl/dx, dM/dy, dm/dx

the numerators

dL, dl, dM, dm

L, l, M, m

be thought to denote omplete dierentials of the fun tions

not

; rather

they always denote how mu h of those dierentials will arise from variability
of just that oordinate (and that only) that appears in the denominator, so
that the expressions always represent nite and determinate quantities.
similar meaning is to be understood for

L = du/dx, l = du/dy, M = dv/dx,

m = dv/dy; this notation was rst used by the illustrious Fontaine who has
furnished us with su h a worthy ompendium of al ulus, and I shall adhere
to it also.

13.

Thus, sin e we have

du = Ldx+ ldy

and

dv = Mdx+ mdy, we may also

infer the two velo ities at any other point an innitely small distan e from
the point

; for if su h a point is at a distan e from

and along the axis

AL

will be

AB

= dy,

velo ity along the other axis

in an innitely small time interval

is moved in the dire tion of axis

AL
AB

in the dire tion of the other axis

14.

along the axis

AL = dx,

then the velo ity of this point along the axis

= u + Ldx + ldy; and the

= v + Mdx + mdy. Therefore

by the amount

AB

will be

dt this point

= dt(u + Ldx + ldy)

by the amount

and

= dt(v + mdx + mdy).

Having noted this, let us onsider a triangular element of water

lmn

and we seek the lo ation to whi h it is transferred, by the motion intrinsi to

that element. Let the side

lm

of this triangular element be parallel to the axis

AL

, the side

ln
m

that the point

and

parallel to the axis

has oordinates

AB

; and take

lm = dx

x+ dx and y, and point n

y + dy. It is lear that the dierentials dx

and

dy

ln = dy;

and

so

has oordinates

ould be either positive

or negative, sin e we have not xed them; and also that the whole mass of
the uid an be mentally divided up into elements like this, so that what we
pres ribed for one will apply equally well to all.

15.

To make lear how the element

dt

interval

lmn

is transferred in the small time

by its intrinsi motion, we seek the points

its angles [verti es

l m
,

and

pq

are transferred in the time

have velo ities

l
m
n
AL u u+Ldx u+ldy
AB v v+Mdx v+mdy

point:

along
along

point

will ome to

, that is:

AP AL = udt,
P p Ll
Point

will ome to

= vdt.

, that is:

AQ AM =

(u + Ldx)dt,

Qq Mm = (v + Mdx)dt.
9

and

dt

, into whi h

. Sin e we shall

But point

will be brought to

, that is:

AR AL =
Rr Ln
16.

Sin e points

time interval

l m
,

dt

, the triangle

indi ated by triangle

lmn
dt

the triangle
little time

and

(u + ldy)dt,

= (v + mdy)dt.

are brought to points

lmn

p q
,

and

in the small

is to be thought as going to the lo ation

pqr

, joined by the line segments

pq pr
,

and

qr

Sin e

was set to be innitely small, after the translation over the

it will still retain a triangular gure

Sin e the element

lmn

pqr

, that is re tilinear.

ought not to be extended into a greater area, nor to

be ompressed into a smaller one, its motion must be so omposed that the
area of triangle

17.

pqr

[Paragraphs

gen e theorem.

= 12 dxdy,

equals the area of triangle

17-20

will establish that

But the triangle

lmn
pqr

lmn

u = 0,

without the benet of the diver-

, if it is a right angle at

and the area of triangle

, has an area

must also be equal to this.

that area, we must onsider the oordinates of the points

To nd

pqr
,

, whi h are:

p
q
r
x AP=x udt AQ=x dx u Ldx dt AR=x u ldy dt
y Pp=y vdt Qq=y v Mdx dt Rr=y dy v mdy dt
+

+(

+(

10

+(

+(

Then the area of the triangle

pqr

is found from the areas of the following

trapezoids, thus :

pqr = P prR + RrqQ P pqQ.

Sin e however these trapezoids have two sides parallel and perpendi ular to
the base

18.

AQ

, their areas are easily determined.

For we have, as in geometry,

1
P R(P p + Rr),
2
1
RrqQ =
RQ(Rr + Qq),
2
1
P Q(P p + Qq).
P pqQ =
2
P prR =

Colle ting these together, we nd:

1
1
1
pqr = P Q Rr RQ P p P R Qq.
2
2
2

[Euler denes new quantities Q, R, q, r.

3 See translator's omment to this paragraph at the end of this do ument.

11

For the sake of brevity, put

AQ = AP + Q
AR = AP + R
Qq = P p + q
Rr = P p + r

so that

P Q = Q, P R = R, RQ = Q R.
Then

pqr = 12 Q(P p+r) 21 (QR)P p 21 R(P p+q) or pqr = 12 Qr 12 Rq.

19.

But from the oordinate values shown above [paragraph

17

Q = dx + Ldxdt; q = 0 + Mdxdt;
R =

0 + ldydt;

r = dy + mdydt.

After substitution, the area of the triangle be omes

1
1
pqr = dxdy(1 + Ldt)(1 + mdt) Mldxdydt2 ,
2
2
or

1
pqr = dxdy(1 + Ldt + mdt + Lmdt2 Mldt2 )
2
12

and sin e this should be equal to the area of triangle

= 21 dxdy,

lmn

, whi h is

there results this equation:

Ldt + mdt +Lmdt2 Mldt2 = 0,

L+m

20.

Sin e the terms

nite

and

m,

Lmdt

+Lmdt Mldt
Mldt

and

are vanishingly small ompared to

we shall have the equation

are dealing with a

possible

= 0.

L + m = 0.

motion, the velo ities

and

For this reason, if we

of any point

must

be su h that in their dierentials

du = Ldx + ldy,
dv = Mdx + mdy

we shall have

that

Sin e

L = du/dx

v
AB

and

and

L + m = 0.

and

, whi h are on eived as those in point

m = dv/dy,

in the dire tions of

, should be thought of as fun tions of the oordinates

du/dx + dv/dy = 0,

the ondition

the velo ities

and

AL

su h

and the riterion of possible motions onsists in

du/dx + dv/dy = 0.

Iif this ondition does not hold, the uid

motion annot take pla e.

21.

[Three-dimensional ows.

uid motion does

not

We must pro eed in the same way when the

resolve into a plane. To investigate the question taken

13

in its widest sense, we shall take the individual parti les of the uid ae ting
ea h other in any sort of motion, with the only proviso being that neither
ompression nor expansion o ur in any part.

We seek to determine from

this what sort of velo ities an o ur and give a possible motion; or, what
omes to the same thing, we want to ex lude from the list of possible motions
those whi h do not observe these onditions, so that the riterion for possible
motions an be determined.

22.

So we shall onsider any point

represent using three orthogonal axes

three oordinates of the point

of the uid, whose lo ation we shall

AL, AB, AC

. [Figure 2 below. Let the

parallel to these axes be AL = x, Ll = y

and

l = z; whi h will be gotten if from the point a perpendi ular l is dropped

to the plane determined by the two axes
then take the perpendi ular
the point

lL

AL
AB.
AL

to the axis

and

From the point

we

. In this way the lo ation of

an generally be expressed by three oordinates. This will apply

at all points of the uid.

14

23.
and

The motion of the point

parallel to the axes

the velo ity of the point

an be resolved into three dire tions

AL AB
,

be

x, y

and

AC

. So let the three dire tions of

= u, = v, o = w

velo ities an vary with the point

the three oordinates

and

z.

; and sin e these

they an be onsidered as fun tions of

Taking dierentials, we get the forms:

du = Ldx + ldy + dz
dv = Mdx + mdy + dz
dw = Ndx + ndy + dz

15

L, l, , M, m, , N, n,

and the oe ients

dinates

24.

x, y

and

will be fun tions of the oor-

z.

As these dierential forms are omplete, it follows, in the same way as

the above, that

dL/dy = dl/dx;

dL/dz = d/dx;

dl/dz = d/dy

dM/dy = dm/dx; dM/dz = d/dx; dm/dz = d/dy

dN/dy = dn/dx;

dN/dz = d/dx;

dn/dz = d/dy

ea h fra tion showing how mu h the variable in the numerator hanges for a
given hange in the oordinate in the denominator.

25.

In an innitesimal time

by the amounts
the dire tion of
all

V,

dt, the point an move in all three dire tions:

udt in the dire tion of AL, vdt in the dire tion of AB, wdt in

AC

. Sin e however the speed of the point

whi h we may

arises from the omposition of the motions in the three dire tions,

whi h are orthogonal, we shall have

traveled in the time

26.

dt

will be

V =

= V dt.

(uu + vv + ww), and the distan e

Let us now onsider any volume element of the uid, to see where it

may advan e to in an innitesimal time

dt.

Sin e it does not matter what

gure we attribute to it, as long as the whole uid mass an be divided into
su h gures, for ease of al ulation let the gure be a re tangular triangular

16

pyramid, ending at the four solid angles

, , , o so that the the oordinates

are given by the s heme:

along:

x
y
z

AL
AB
AC

x + dx

y + dy

z + dz

and sin e the base of this pyramid is

is

o = dz,

27.

the volume will be

= lmn = 21 dxdy,

and the altitude

= 16 dxdydz .

We shall now investigate where these individual verti es

be arried in the innitesimal time

dt.

, , , o

will

For ea h of these, we must onsider

the three velo ities along the three oordinate axes, for these will dier from

u, v, w

the three original velo ities

28.

Parallel to

AL

u + Ldx

u + ldy

u + dz

AB

v + Mdx

v + mdy

v + dz

AC

w + Ndx

w + ndy

w + dz

If the points

, , ,

and

a ording to the following s heme.

, , , o

are arried in the innitesimal time

dt to

points

whose oordinates are given parallel to the three axes, the

4 This is a dierent set of meanings for these symbols.

5 Remember that Euler uses , , with two sets of meanings.

17

instantaneous translations along these axes will be: [for

AP AL = udt,
P p Ll = vdt,
p l = wdt;
[for

AQ AM = (u + Ldx)dt,
Qq Mm = (v + Mdx)dt,
q m = (w + Ndx)dt,

[for

AR AL = (u + ldy)dt,
Rr Ln = (v + mdy)dt,
r n = (w + ndy)dt,
[for

AS AL = (u + dz)dt,
Ss Ll = (v + dz)dt,
s lo = (w + dz)dt.
18

Thus we shall have as oordinates for the four points : [for

AP = x + udt,
P p = y + vdt,
p = z + wdt;

[for

AQ = x + dx + (u + Ldx)dt,
Qq = y + (v + Mdy)dt,
q = z + (w + Ndx)dt;

[for

AR = x + (u + ldy)dt,
Rr = y + dy + (v + mdy)dt,
r = z + (w + ndy)dt;
6 [The original had a misprint.

19

[for

AS = x + (u + dz)dt,
Ss = y + (v + dz)dt,
s = z + dz + (w + dz)dt.

29.

When therefore the verti es

into points

, , ,

, , , o

in the innitesimal time

of the pyramid are translated

dt, these

new points are to de-

termine a triangular pyramid su h that the volume of both be equal, namely

= 61 dxdydz .
mid

So the task omes down to determining the volume of the pyra-

.7

It is lear, however, that this pyramid is what we have left if from the volume
element

pqr

we take away the element

a prism sitting perpendi ularly

oblique se tion

30.

pqr , for the latter element is

on the triangular base

pqr ,

with the upper

ut o.

In any trun ated prism of this type, the element

pqr an be resolved

into three other volumes, whi h are:

7 The following paragraphs up through

36

are devoted to this task. If we allow use of

the determinantal formula for a parallelpiped, then the equation at the end of

u = 0.
from pqr only in

immediately, whi h implies that

8 The triangle

diers

the

20

z oordinates.

35 follows

I.

pqs

II.

prs

III.

qrs

in su h a way that we must have

1
dxdydz = pqrs + prs + qrs pqr.
6
When however a prism of this sort sits perpendi ularly on its lower base, with
three dierent altitudes, then its volume is found if the base is multiplied by
the sum of the three altitudes, divided by three.

31.

Therefore the volume of these trun ated prisms will be:

1
pqs(p + q + s),
3
1
prs(p + r + s),
prs =
3
1
qrs =
qrs(q + r + s),
3
1
pqr(p + q + r).
pqr =
3

pqs =

Sin e however

pqr = pqs + prs + qrs,

the sum of the rst three volumes,

minus the last, will be

1
1
1
1
1
dxdydz = p.qrs q.prs r.pqs + s.pqr;
6
3
3
3
3
21

or

dxdydz = 2pqr.s 2pqs.r 2prs.q 2qrs.p.

32.

It remains to as ertain the bases of these prisms. Before we do this, to

redu e al ulations we put

AQ = AP + Q;

Qq = P p + q;

q = p + ,

AR = AP + R;

Rr = P p + r;

r = p + ,

AS = AP + S;

Ss = P p + s;

s = p +

and with these substitutions, the terms ontaining

an el ea h other, and

we shall have

dxdydz = 2pqr. 2pqs. 2prs.

and the number of bases to be investigated is redu ed by one.

33.

Now the triangle

from the gure

pqr

P prqQ,

will be found, if the trapezoid

or from the ombined trapezoids

P pqQ

is ut out

P prR + RrqQ

Hen e

1
1
1
pqr = P R(P p + Rr) + RQ(Rr + Qq) P Q(P p + Qq);
2
2
2
9 This orrespondd to the abbreviations in paragraph

22

18.

but sin e

P R = R, RQ = Q R,

and

P Q = Q,

we shall have

1
1
1
1
pqr = R(P p Qq) + Q(Rr P p) = Qr Rq.
2
2
2
2
In a similar way we shall have:

1
1
1
P S(P p + Ss) + SQ(Ss + Qq) P Q(P p + Qq),
2
2
2
1
1
1
S(P p + Ss) + (Q S)(Ss + Qq) Q(P p + Qq)
pqs =
2
2
2

pqs =

when e

1
1
1
pqs = S(P p Qq) + Q(Ss P p) = Qs Sq.
2
2
2
Next,

1
1
1
P R(P p + Rr) + RS(Rr + Ss) P S(P p + Ss),
2
2
2
1
1
1
R(P p + Rr) + (S R)(Rr + Ss) S(P p + Ss)
prs =
2
2
2
prs =

when e

1
1
1
1
prs = R(P p Ss) + S(Rr P p) = Sr Rs.
2
2
2
2

34.

Substituting in these values, we shall obtain

dxdydz = (Qr Rq) + (Sq Qs) + (Rs Sr),

23

so the volume of the pyramid

will be

1
1
1
(Qr Rq) + (Sq Qs) + (Rs Sr).
6
6
6
From the values re orded in paragraph 28 above,

Q = dx + Ldxdt

q = Mdxdt

= Ndxdt

R = ldydt r = dt + mdydt = ndydt

S = dzdt

35.

s = dzdt

= dz + dzdt.

Sin e it follows that

Qr Rq = dxdy(1 + Ldt + mdt + LMdt2 MLdt2 )

Sq Qs = dxdz(dt Ldt2 + Mdt2 )
Rs Sr = dydz(dt mdt2 + ldt2 ),
therefore we nd that the volume of the pyramid

+Ldt +Lmdt

+mdt Mldt2

+dt
1
dxdydz 1 +

+Ldt2
+mdt2
ndt2
Ndt2
24

is expressed as

+Lmdt

3
Mldt

3
Lndt
3

+Mndt3

3
Nmdt

3
+Nldt

whi h should be held equal to the volume of the pyramid

After dividing by

dt,

o = 16 dxdydz

there results the equation

0 = L + m + +dt(Lm + L + m Ml N n)
+dt2 (Lm + Mn + Nl Ln Ml Nl).

36.

Disregarding the innitely small terms, we have the equation

= 0,

in whi h we have as ertained the ondition on the velo ities

to admit the uid motion as possible. Sin e

dw/dz,
values

L+m+

L = du/dx, m = dv/dy

u, v, w

and

the ondition for a possible motion, when any point with oordinate

x, y, z

has orresponding velo ities

u, v, w ,

will therefore be:

du dv dw
+
+
= 0.
dx dy
dz
By this ondition

10

no part of the uid will pass into a greater or smaller

spa e, and the ontinuity of the uid (and also the density) will be maintained
without interruption.

37.

This property of the uid, however, is to be interpreted to hold for

all parts of the uid at ea h moment in time: that is, at ea h moment the

uvw
xy
z

three velo ities

oordinates

and

for all points ought to be su h fun tions of the three

, that

du/dx + dv/dy + dw/dz = 0

will hold, and this

hara ter of those fun tions limits any proposed motion of the individual

10 Euler has a tually proved that this is a ne essary ondition.

25

points of the uid. At any other time, however, the motion of those points
ould be quite dierent, restri ted only by the requirement that the above
property still take pla e. Of ourse, the same behavior up to the present is
assumed.

38.

If however we wish to think of time also as variable, so that the motion

of a point after an elapsed time

given by the oordinates
three velo ities

u v w
t

also on the time

and

, so that

is to be dened, when the position

AL = x, Ll = y ,

and

l =

11

is

, it is lear that the

depend not only on the oordinates

x y
,

and

, so that they are fun tions of these four quantities

z
x y z
, but
,

12

du =Ldx + ldy + dz + Ldt,

dv =Mdx + mdy + dz + Mdt,
dw =Ndx + ndy + dz + N dt.
Meanwhile however, we shall always have
instant the time

is to be taken as onstant, so that

ne essary however the fun tions

11 A,

L, l,

L + m + = 0,

u v w


be ause at any

dt = 0.

Thus it is

may hange with time that at ea h

represent points is spa e, while the L, l and

below

oe ients.

represent dierential

12 I have substituted alligraphi for German letters in this translation. It is too di ult

for non-Germans to distinguish between the German N and R.

26

moment there holds the ondition

du dv dw
+
+
= 0.
dx dy
dz
Sin e this ondition assures that any portion of the uid will be arried in
the innitesimal time

dt

into an equal volume, and also likewise under the

same ondition in the following element of time, it follows that this must
happen in all following elements of time.

Part Two
39.

[Dynami s. From those possible motions whi h have satised the above

ondition we shall now investigate the nature of that motion whi h a tually
an be sustained in the uid. That is besides the ontinuity of the uid and
its density being onstant, a ount must be taken here of the for es ae ting
the movement of individual elements of the uid. For whatever the motion
of ea h element if it is not uniform or not pointed in a [ ommon dire tion
the hange of motion ought to onform to the for es applied to this element.
No matter how the hange of motion is determined by these given for es the
above formulas must still be a onstraint on this hange of motion so new
onditions must be found, by whi h any hitherto possible motion is restri ted
to the a tual motion.

27

40.

[Planar motion. Let us also set up this investigation in two parts; and

rst we shall think of all the motion of the uid as taking pla e in the

same plane
Therefore let the position oordinates of any point
as before as
velo ities of

AL = x, Ll = y;

and now in the elapsed time

in dire tions parallel to the axes

AL

and

AB

be ause we now have to take a ount the hange in time

fun tions of

x y


and

, so that

be dened

let the two

be
and

u
v

and

will be

13

du =Ldx + ldy + Ldt,

dv =Mdx + mdy + Mdt;
and on a ount of the ondition that we found above we must have

41.
to

In the innitesimal elapsed time

 with a displa ement

a displa ement

= vdt

dt,

the distan e

 let therefore the point

in the dire tion of the axis

in the dire tion of the other axis

in rease in the velo ities

time

= udt

dt

dx

and

of the point

ought to be written

udt

dv =Mudt + mvdt + Mdt,

and

M = v/t.

28

AL

and with

to obtain the

and the distan e

du =Ludt + lvdt + Ldt,

L = ut,

be brought

whi h o ur in the innitesimal

so that

13 In modern notation,

AB;

L + m = 0.

dy

as

vdt,

from whi h the a elerative for es

14

needed to produ e these a elerations in

the orresponding dire tions will be:

along AL : 2(Lu + lv + L)
along AB : 2(Mu + mv + M)

and the for es a ting on the parti le of water

should be equal to these

expressions.

42.

Among the for es that a t on the parti les of water, we have to give

rst onsideration to gravity. Its ee t will be null, if the plane of motion is
horizontal. If however the plane of motion is at an in line, in the dire tion of
the axis

AL

, the horizontal axis being

will take a onstant value

AB

, the a elerative for e due to gravity

in the dire tion of AL. Moreover, we ought not to

negle t fri tion, be ause the motion is often there-by appre iably impeded.
Although the laws governing fri tion are not yet satisfa torily established,
never-the-less we shall perhaps not err too mu h from the mark if , in analogy
with the fri tion of solid bodies, we set the fri tion to be proportional to the
pressure of the parti les of water a ting on ea h other.

14 Euler denes the measure of for e as the a eleration needed to move a unit mass a unit
distan e in the dire tion of the for e, in a unit time. Sin e our notion of a eleration will
give one-half of a unit distan e, Euler's measure of for e will require twi e our a eleration.

29

43.

As the rst step then, we must ompute the pressure with whi h the

parti les of water are a ting on ea h other. For a parti le pressed all around
by adja ent parti les, to the extent that the pressure in some dire tion is not
balan ed, just so mu h will the motion of the parti le be ae ted.

15

That is

to say, the water at ea h point is subje t to a ertain state of ompression,

whi h will be like what o urs in still water at a ertain depth. It is onvenient
to use this depth, at whi h still water is found to be in the same state of

ompression, as an expression for the pressure at any point

Therefore, if

, then

is that still-water depth that expresses the pressure at point

will be a ertain fun tion of the oordinates

pressure at

44.

also varies with time,

Therefore we shall put

ln = mo = dy;

lmno

the area being

and

, and if the the

will also be a fun tion of time

dp = Rdx + rdy + Rdt,

re tangular element of water

and

of the water.

[Fig.

and we shall onsider a

3, whose sides are

= dxdy.

15 Euler is onsidering a material parti le extending in ea h dire tion.

30

lm = no = dx

If now the pressure at

= p + rdy
a for e

and at

is

will be

= p,

the pressure at

= p + Rdx + rdy.

= dx(p + 21 Rdx). while the

= dx(p + 21 Rdx + rdy). From these

ln
be

and

no

that is

lm

at

is pressed by

will be pressed by a for e

two for es, the element

ln

= p + Rdx,

will be

Then the side

opposite side

a resulting for e in the dire tion of

from the for es

= rdxdy.

lmno

will be have

In the same way,

dy(p + 12 rdy) and dy(p + Rdx + 12 rdy) whi h

a t on the sides

mo

, the resulting for e a ting on the element in the dire tion

lm

will

= Rdxdy.

45.

Hen e there arises an a elerative for e in the dire tion of

= R,

and an a elerative for e in the dire tion of

ln

that is

rst of these along with the for e due to gravity will then be
fri tion, we shall have the equations

R = 2Lu + 2lv + 2L

31

lm

= r.

R.
or

that is
The

Absent

R = 2Lu 2lv 2L,

and

r = 2Mu + 2mv + 2M

or

r = 2Mu 2mv 2M
whi h together give us

dp = dx 2(Lu + lv + L)dx 2(Mu + mv + M)dy + Rdt.

This dierential should be omplete, that is, integrable.

46.
of

Sin e the term

Rdt,

dx is already integrable, and as of yet we know nothing

by the very nature of omplete dierentials it is ne essary for the

above expression that

d(Lu + lv + L)
d(Mu + mv + M)
=
dy
dx

and then e, be ause

Ll +

du/dx = L, du/dy = l, dv/dx = M,

and

dv/dy = m,

vdl dL
udM
vdm dM
udL
+ lm +
+
= ML +
+ mM +
+
dy
dy
dy
dx
dx
dx

32

whi h redu es to this formula:

dL dM
(L + m)(l M) + u

dy
dx

47.

Sin e

+v

dl
dm

dy
dx

dL dM

= 0.
dy
dx

Ldx+ldy+Ldt and Mdx+mdy+Mdt are omplete dierentials16 ,

we know that

dL
dl
=
;
dy
dx

dm
dM
=
;
dx
dy

dL
dl
=
dy
dt

dM
dM
=
,
dx
dt

and

and when these expressions are substituted, we shall have the equation:

(L + m)(l M) + u

dl dM
dx

whi h learly will be satised by

+v

l=M :

latter ondition requires in turn that

dl dM
dy

that is, by

udx + vdy

17

dl dM
=0
dt

du/dy = dv/dx.

This

be a omplete dierential,

and this latter is the desired ondition des ribing those motions that are to
be allowed.

48.

18

This riterion is independent of the previous one of ontinuity and

uniform density. For this property that

udx + vdy

be a omplete dierential

would still apply, even if the moving uid were to hange its density, as in
the motion of elasti uids, or as happens in air.

16 =

du

and

dv

respe tively.

17 I have orre ted some obvious misprints on this page.

18 See translator's omment to this paragraph.

33

That is, the velo ities

v
t

and

would be su h fun tions of the oordinates

time

49.

the expression

udx + vdy

and

determine ompletely the motion of the uid.

we shall have

that at any xed

would be ompletely integrable.

We are now in a position to dene the pressure

M = l,

p,

whi h we need to

Sin e we have found that

19

dp = dx 2u(Ldx + ldy) 2v(ldx + mdy) 2Ldx 2Mdy + Rdt.

From

Ldx + ldy = du Ldt

and

ldx + mdy = dv Mdt,

dp = dx 2udu 2vdv + 2Ludt + 2Mvdt 2Ldx 2Mdy + Rdt.

If we wish to dene the pressure at ea h lo ation for a given xed time

equation to be onsidered is:

dp = dx 2udu 2vdv 2Ldx 2Mdy

and, on writing

L = du/dt

and

M = dv/dt,

we then get

dp = dx 2udu 2vdv 2

du
dv
dx 2 dy.
dt
dt

19 A tually, the argument in the next two lines does not depend on
20 That is, to get the spatial gradient of pressure.

34

M = l.

20

, the

In integrating this equation, the time

50.

is to be held onstant.

Given the hypotheses, [we shall see that this equation is integrable, if

udx + vdy

we take into a ount the riterion that

keeping the time
and

, whi h gives

onstant. Let

dS = udx + vdy

be a omplete dierential,

be its integral, whi h is a fun tion of

when

dt = 0.

If we further allow

xy
,

to be

variable, this be omes

dS = udx + vdy + Udt.

Then we shall have

51.

du/dt = dU/dx

and

dv/dt = dU/dy.

Then

U = dS/dt.

Introdu ing these expressions gives us:

dv
dU
dU
du
dx + dy =
dx +
dy
dt
dt
dx
dy

whose integral for a xed time

apparent, let us put

is learly

dU = Kdx + kdy,

so

= U.

To make this more learly

dU/dx = K

and

dU
dU
dx +
dy = Kdx + kdy = dU.
dx
dy

The integral of this equation being =

U = dS/dt,

then

dp = dx 2udu 2vdv 2dU

35

dU/dy = k.

Then

whi h upon integration yields:

p = Const. + x uu vv 2
S

being a fun tion of

52.

xy
,

and

. For

dt = 0,

dS
,
dt

its dierential is

udx + vdy.

To understand better the nature of this formula, we shall onsider the

speed at a point , whi h will be

will be:
denotes

53.

=V =

(uu + vv).

The pressure moreover

p = Const. + x uu vv 2dS/dt. The S in the last

R
S = (udx + vdy), where we view the time t as variable.

term

dS

Suppose we now wished to in lude a fri tion term that is proportional

to the pressure

in ee t while the point

traverses an innitesimal distan e

ds. The retarding for e arising from the fri tion would then be = p/f. Putting
dS/dt = U,

our dierential equation at a denite point

dp = dx

Let

in time be omes:

p
ds 2V dV 2dU.
f

be the number whose hyperboli logarithm is = 1.

21

gives

s/f

p=e
21 The orginal has

dV

instead of

es/f (dx 2V dV 2dU)

dU.

36

Integration then

or

1
p = x V V 2U es/f
f

es/f (x V V 2U)ds.

54.

For the uid motion that is a tually to be sustained, the riterion

is that

udx + vdy

be a omplete dierential at any xed time

t 22

The

ontinuity ondition requires that the density stay onstant and uniform so
that

du/dx + dv/dy

omplete dierential
be fun tions of

udy + vdx
55.

x y
,

will be

23

= 0,

when e it follows that

. Thus the velo ity omponents

and

udy + vdx

and

su h that both the expressions

is a

together must

udx + vdy

and

are omplete dierentials.

[Flows in three dimensions. Let us now start to investigate the ase where

the three velo ity omponents

AL, AB, AC,

uvw
,

of the point

are fun tions of the oordinates

dire ted along the axes

xyz
,

, and of the time

su h that

du =

Ldx + ldy + dz + Ldt

dv = Mdx + mdy + dz + Mdt

dw =

Ndx + ndy + dz + N dt;

and, a ording the the ondition given earlier, we must have

even if the time

L+m+ =0

is allowed to vary. This is the same as

du dv dw
+
+
= 0.
dx dy
dz
22 See the translator's omment for paragraph 47.
23 This expression la ks the minus sign both times it appears in this paragraph.

37

This ondition will not be used in the present part of our analysis.

56.

After an innitesimal time

traverses a distan e
in the dire tion of

= udt

AB,

dt, the

point

is brought to position

AL,

in the dire tion of the axis

and a distan e

three velo ity omponents for the point

along AL =

= wdt

24

a distan e

in the dire tion of

at postion

It

= vdt

AC.

The

will be:

u + Ludt + lvdt + wdt + Ldt,

along AB = v + Mudt + mvdt + wdt + Mdt,

along AC =

w + Nudt + nvdt + wdt + N dt.

Then the a elerations along these same dire tions will be:

along AL =

2(Lu + lv + w + L),

along AB = 2(Mu + mv + w + M),

along AC =

57.

Let us take the axis

two axes

AL

and

a elerative for e

AB

AC

2(Nu + nv + w + N ).
in the verti al dire tion, so that the other

are horizontal.

= 1

Along the axis

due to gravity. The pressure

dierential

dp = Rdx + rdy + dz,

24 The ondition will be re-introdu ed in paragraph

38

64.

AC,

there shall be an

of point

will have a

if time is held onstant. There arise from this the three a elerations

AL : R, AB : r, AC : .

These various equations are ombined in the same way that we used in paragraphs 44 and 45, and we need not repeat the argument here.

The result

will be the equations:

R=

2(Lu + lv + w + L),

r=

2(Mu + mv + w + M),

= 1
58.

2(Nu + nv + w + N ).

Sin e however the form

dp = Rdx + rdy + dz

must represent a

omplete dierential, we shall have

dR/dy = dr/dx; dR/dz = d/dx; dr/dz = d/dy.

After dierentiation and dividing by -2, we shall obtain the three equations

uLy + vly + wy + Ly

+Ll + lm + n

25

uMx + vmx + wx + Mx +ML + mM + N,

25 To in rease legibility in these two paragraphs, I follow the more modern pra ti e
indi ating partial derivatives by subs ripts. In modern terms, Euler has taken the url
to eliminate the pressure gradient. The dependent variables in the resulting dierential
equations in paragraph

59 be ome the vorti ity omponents.

39

II

III

uLz + vlz + wz + Lz
+L + l +
=

uNx + vnx + wx + Nx +NL + nM + N,

uMz + vmz + wz + Mz +M + m + =

uNy + vny + wy + Ny

59.

+Nl + nm + n.

By the property of omplete dierentials,

26

x = Mz ;

Ly = lt ;

Mx = Mt ,

lz = y ;

nx = Ny ; x = Nz ;

L z = t ;

Nx = Nt ,

Ny = nx ;

mz = y ;

Mz = t ;

Ny = nt .

Ly = lx ;

mx = My ;

Lz = x ;
Mz = x ;

y = lz ;

y = nz ;

Substitute these in the above three equations, whi h then be ome



 



dl dM
dl dM
dl dM
dl dM
+u
+v
+w
+(l M) (L + m)+nN = 0,
dt
dx
dy
dz

 
 



d dN
d dN
d dN
d dN
+u
+v
+w
+( N) (L + )+lnM = 0,
dt
dx
dy
dz

 
 



d dn
d dn
d dn
d dn
+u
+v
+w
+( n) (m + )+MNl = 0.
dt
dx
dy
dz
26 Ea h line of equalities is used in the orresponding equation of the previous paragraph.

40

60.

It is lear that these three equations are satised on setting:

27

l = M; = N; = N;

and therein lies the ondition as derived from our analysis of applied for es.

28

These an be expressed in our usual notation as

du
dy

dv
dx ;

du
dz

dw
dx ;

dv
dz

dw
dy .

These, however, are the very onditions required for the form

wdz

to be a omplete dierential. This ondition then states that the three

velo ity omponents

u, v, w

that at any xed time the form

61.

udx + vdy +

be fun tions of

x, y, z

udx + vdy + wdz

For a xed moment in time (i.e.,

dt = 0),

along with

admit an integral.

then, we have

du =Ldx + Mdy + Ndz

dv =Mdx + mdy + ndz
dw =Ndx + ndy + dz
27 See translator's omment for this paragraph.
28 quibus ontinetur riterium, quod onsideratio

41

solli itationum suppeditat.

, su h

and the values for

R, r,

will be:

R=

2(Lu + Mv + Nw + L),

r=

2(Mu + mv + nw + M),

= 1

2(Nu + nv + w + N ).

We shall have this equation for the pressure:

dp = dz 2u(Ldx + Mdy + Ndz)

2v(Mdx + mdy + ndz)
2w(Ndx + ndy + dz)
2Ldx 2Mdy 2N dz

= dz 2udu 2vdv 2wdw 2Ldx 2Mdy 2N dz.

62.

Sin e

L = du/dt; M = dv/dt; N = dw/dt,

p = C z uu vv ww 2

By the ondition found above,

an take this integral to be

Z 


dv
dw
du
dx + dy +
dz .
dt
dt
dt

udx + vdy + wdz

= S.

A variable time

will be integrable, and we

an now be allowed, and

we an take

dS = udx + vdy + wdz + Udt,

42

integration gives

with

du/dt = dU/dx; dv/dt = dU/dy; dw/dt = dU/dz.

29

Consequently,

for that point in time assumed in the above integral,

dU
dU
dU
dx +
dy +
dz = dU
dx
dy
dz

and we shall have

p = C z uu vv ww 2U,

or

30

p = C z uu vv ww 2
63.

The form

at the point

uu + vv + ww

is seen to express the square of the speed

so the equation for the pressure be omes

p =C zVV 2

To evaluate this, rst we must seek the integral

Its dierential with only the time
will give the value of

dS/dt,

dS
.
dt
S

of the form

udx+vdy+wdz.

t being variable will be divided by dt, whi h

whi h then goes into the expression we found

for the pressure.

29 Also,

dS
.
dt

U = S/t.
dS/ds.

30 The original has

43

64.

If we now onjoin the earlier riterion whi h onstrains any possible

motion, then the three velo ity omponents

of the oordinates

xy
,

and

uvw
t
,

with [xed? time

be a omplete dierential, and also that

should be su h fun tions

, that rst

udx + vdy + wdz

du/dx + dv/dy + dw/dz = 0.

Any

motion of the uid must be subje t to these onditions, if the density is

taken not to vary. Moreover, if the form
dierential with variable time

udx + vdy + wdz + Udt is a omplete

, then the state of the pressure at any point

is expressed in terms of a depth

with

p = C z uu vv ww 2U.

This holds when the uid experien es gravity in the

BAL
65.

-dire tion, the plane

being horizontal.

[More general ase. Suppose we assign a dierent dire tion for gravity,

or we allow any variable for es to a t on individual parti les of the uid,

so that a dieren e in the value of the pressure

would now enter.

There

would still be no hange in the law governing the velo ity omponents of
ea h point of the uid.
onstituted su h that
that

The three velo ity omponents must always be so

udx + vdy + wdz

du/dx + dv/dy + dw/dz = 0. The

be a omplete dierential, and su h

three velo ity omponents

uvw
,

an

be set up in an innite number of ways to satisfy these two onditions ; and

then the uid pressure an be assigned at ea h point.

44

66.

It will be mu h more di ult, however, to determine the motion of the

uid at ea h point when variable applied for es and pressure are assigned.
For in these ases we need to nd various equations of the form

z uu vv ww 2U.

The fun tions

u v w
,

p = C

have to be dened to satisfy

not only su h equations, but also the previously spe ied laws.

This will

require the utmost analyti al skill. The sensible approa h is to inquire into
the nature of suitable fun tions whi h would onform to ea h riterion.

67.

The best pla e to start is with the integral whose dierential is

vdy + wdz

when time is held xed.

fun tion of

x y
,

, and

Let

be this integral, whi h will be a

u v w
,

point whose oordinates are

x y
z
xy
z
,

and

If the quantity

is

will provide the

The question now omes down to

and

du/dx + dv/dy + dw/dz = 0.

w = dS/dz,

whi h obtain at the urrent time at the uid

this: to dene what fun tions of

also have

t
dx dy dz

, for xed but arbitrary time

dierentiated, the oe ients of the dierentials

velo ity omponents

udx +

may be allowed for

Sin e

so that we

u = dS/dx, v = dS/dy,

this means that

ddS ddS ddS

+ 2 + 2 = 0.
dx2
dy
dz

45

and

68.

[Spe ial solutions.

Sin e it is not obvious how in general this an be

made to happen, I shall onsider ertain lasses of possibilities. Let then

S = (Ax + By + Cz)n .

Then

dS/dx = nA(Ax + By + Cz)n1

Cz)n2 ,

and

and similar forms will hold for

ddS/dx2 = n(n 1)AA(Ax + By +

ddS/dy 2

and

ddS/dz 2 .

From this it

must be that

n(n 1)(Ax + By + Cz)n2 (AA + BB + CC) = 0.

This will be satised in the rst ase when either

will obtain two suitable solutions, namely

Cz,

where the onstants

69.

If however

CC = 0.

ABC
,

n = 0 or n = 1. From these

= onstant and

S = Ax + By +

and the time an be hosen arbitrarily.

is neither 0 nor 1, then we ne essarily have

A suitable solution for

31

AA + BB +

will now be

S = (Ax + By + Cz)n

in whi h the order

order

an be any number, but the time

an enter into this

. It is lear that any ombination of su h forms an be taken for the

31 Evidently we are now onsidering omplex numbers.

46

solution

, so that:

S = + x + y + z + (Ax + By + Cz)n + (Ax + B y + C z)n

+ (A x + B y + C z)n + (A x + B y + C z)n

etc.

as long as

AA + BB + CC = 0;
A A + B B + C C = 0;
A A + B B + C C = 0 etc.

70.

Suitable formulas for S for the smaller orders, where the oordinates

x, y, z are raised to the rst, se ond, third or fourth powers, will be the
following:

I.

32

S=

II. S =
III. S =

A,
Ax + By + Cx,
Axx + Byy + Czz + 2Dxy + 2Exz + 2F yz
(A + B + C = 0),

IV. S = Ax3 + By 3 + Cz 3 + 3Dxxy + 3F xxz + 3Hyyz + 6Kxyz

A + E + G = 0; B + D + I = 0; C + F + H = 0;
32 See omment on this paragraph at the end of the do ument.

47

V.
where

S=

+Ax4 + 6Dxxyy + 4Gx3 y + 4Hxy 3 + 12Nxxyz

+By 4 + 6Exxzz + 4Ix3 z + 4Kxz 3 + 12Oxyyz
+By 4 + 6Exxzz + 4Ix3 z + 4Kxz 3 + 12Oxyyz
A + D + E = 0 G + H + P = 0,
B + D + F = 0 I + K + O = 0,
C + E + F = 0 L + M + N = 0.

71.

We an now see how to get the like formulas for any order.

First,

the same numeri al oe ients are given in the individual terms as o ur in
the law of permutation of quantities  that is, whi h arise if the trinomial

x+y+z is raised to that order power.

Next, indenite literals

multiplied into these numeri al oe ients.

, et . are

Then, without regard to these

literals, he k wherever there o ur three terms of the type

NZzz,

ABC

LZxx + MZyy +

whi h have the same ommon fa tor arising from the variables.

often as this o urs, spe ify that the sum

L+M +N

33

As

of the three literals be

set to zero. For example, for the fth power there will be had

+Ax5 + 5Dx4 y + 5Dx4 z + 10Gx3 yy + 10Gx3 zz + 20Kx3 yz + 30Nxyyzz

S=

+By 5 + 5Exy 4 + 5Ey 4z + 10Hx2 y 3 + 10Hy 3zz + 20Lxy 3 z + 30Oxxyzz

+Cz 5 + 5F xz 4 + 5F yz 4 + 10Ixxz 3 + 10Iyyz 3 + 20Mxyz 3 + 30P xxyyz,

33 That is, arising from the permutations of

48

x , y, z .

with the following onditions for the literals:

A + G + G = 0;

D + H + O = 0;

D+I +P =0

B + H + H = 0; E + G + N = 0;

E + I + P = 0;

C + I + I = 0;

K +L+M =0

F + G + N = 0; F + H + O = 0.

In a similar way, there will be 15 onditions of this type for the sixth order,
21 for the seventh, 28 for the eighth, and so on.

72.

[n =0. Now in the rst formula

will be zero, sin e the oordinates

S = A,

x y
,

and

the three velo ity omponents

do not appear at all.

This

des ribes a uid at rest. The pressure at any point, however, an be variable
with the time. For
a point

and time

A
t

is an arbitrary fun tion of time, and so the pressure at

will be

p = C 2 dA
z. This formula indi ates the state
dt

of a uid subje t to any for es whatsoever at any point in time, so long as

the for es balan e ea h other so that no motion of the uid will arise. This
will happen for example if the uid is ontained in a vase without a means
of egress, yet subje t to any sort of for es.

73.

[n=1. For the formula

S = Ax + By + Cz,

dierentiation at the point

will give three velo ity omponents:

u = A; v = B & w = C.

49

Thus, at a given time, all points of the uid will be arried with the same
motion, in the same dire tion. Then the uid as a whole will move like a solid
body, arried by a ommon but hanging motion. At a dierent point in time,
as the applied external for es hange, the motion will dier a ordingly in
both speed and dire tion. If, for point

, then the pressure will be

74.
with

35

the fun tions of time are

AB
,

and

2y dB
2z dC
.34
p = C zAABBCC 2x dA
dt
dt
dt

[n =2. The third formula

A+B+C = 0

S = Axx+Byy +Czz +2Dxy +2Exz +2F yz

will yield three velo ity omponents at the point

u = 3Ax + 2Dy + 2Ez; v = 2By + 2Dx + 2F z; w = 2Cz + 2Ex + 2F y,

w = 2Ex + 2F y 2(A + B)z.

or

In this ase, at any moment in time dierent

points in the uid will be arried in dierent motions. At the next moment in
time, moreover, the motion of ea h point an be variable in any way, be ause
the fun tions for

ABCDEF
,

an be of any sort. Even more variety an

o ur if omposite fun tions are used for

75.

[Investigating the possibility of a ommon rotation. In the se ond ase, the

motion of the uid will oin ide with the uniform motion of a solid body, so
at any moment in time the dierent parts of the uid will be arried in an
equal and parallel motion.

We might suspe t that the motion of the uid

in other ases an also oin ide with the motion of a solid body, whether
rotational or of some other sort. For this to happen, the pyramid

34 The original reads: p = C z AA CC

35 The orginal has instead of u .

50

dB
dC
2x dA
dt 2y dt z dt .

must

Tab.
Fig. 2.

IV,

ne essarily be equal and similar to the pyramid

the values shown in paragraph 32,

= = dy
= o = dz

76.

QQ + qq +

= RR + rr +

= SS + ss +

p
(Q R)2 + (q r)2 + ( )2
p
= (Q S)2 + (q s)2 + ( )2
p
= (R S)2 + (r s)2 + ( )2 .

dx2 + dy 2 =

dx2 + dz 2
p
= dy 2 + dz 2

= o =
= ro

that is, taking over

36

= = dx =

= =

o;

On omparison with the three rst equations, the three last equations

redu e to these:

QR + qr + = 0;
QS + qs + = 0;
RS + rs + = 0.

If however we substitute for

Q R S q r s
,

, ,

the values assigned in

36 Again, with the same letters doubly used as unrelated labels and distan es. See also
the omments to paragraphs

75-77 at the end of this do ument.

51

paragraph 34, then the rst three equations [of the previous paragraph will give:

1 = 1 + 2Ldt; l + M = 0
1 = 1 + 2mdt; + N = 0
1 = 1 + 2dt;

from whi h we would on lude that

N =
77.

and

+n=0

L = 0, m = 0,

and

= 0, M = l,

n = .

The three velo ity omponents at ea h point

onstituted that

would therefore be so

37

du = + ldy + dz
dv = ldx + dz
dw = dx dy.

Now the se ond ondition on the uid motion demands that

and

n = .

l = M, = N

Then these all vanish, and the velo ity omponents

will be the same in all parts of the uid at any given time.

u v
,

and

It is therefore

lear that the uid motion annot oin ide with the motion of a rigid body
in this ase.

37 Some minus signs were missing in the original.

52

78.

To determine the ontribution of the for es whi h a t externally on

the uid, we ought rst to nd the for e needed to produ e any spe ied
uid motion. We found in paragraph 56 that these are equal to the three a elerative for es re orded there. If we onsider a uid element whose volume
or mass is

AL :

= dxdydz,

the motive for es required are therefore:

= 2dxdydz(Lu + lv + w + L)

= 2dxdydz (uux + vuy + wuz + ut )

AB : = 2dxdydz (Mu + mv + w + M)
AC :

= 2dxdydz (Nu + nv + w + N )

38

= 2dxdydz (uvx + vvy + wvz + vt )

= 2dxdydz (uwx + vwy + wwz + wt )

and triple integration gives the total for es that ought to be applied on the
whole mass of uid in ea h dire tion.

79.

A ording to the se ond ondition the form

be a omplete dierential, whose integral is

we are to set

= S.

dS = udx + vdy + wdz + Udt.

du/dz = dw/dx; du/dt = dU/dx

udx + vdy + wdz

With time also variable,

Then, from

du/dy = dv/dx;

et ., those three motive for es be ome:

AL := 2dxdydz

udu + vdv + wdw + dU

,
dx

38 For the sake of legibility, I have resorted to subs ripts here, writing for example
pla e of

is to

du
dx .

53

ux

in

80.

Let now

x y z 39
,

AB := 2dxdydz

udu + vdv + wdw + dU

,
dy

AC := 2dxdydz

udu + vdv + wdw + dU

.
dz

uu+vv +ww +2U = T, T

being a fun tion of the oordinates

For a xed point in time, we an write

40

dT = Kdx + kdy + dz

and the three motive for es of the element

dxdydz

are

AL : = Kdxdydz
AB : = kdxdydz
AC : = dxdydz.

Upon triple integration, these formulae extend to the whole uid mass. From
these, we obtain equivalent

41

expressions for for es, and their average dire -

tions, that may be used everywhere. But this involves a truly higher level of
di ulty, and I shall not dwell further on this topi .

39 See omments for this and the following paragraph, at the end of the do ument.
40 The typesetter misread , repla ing it with u here and with k further down.
41 Equivalent to the expressions written out in paragraph

54

78.

81.

This quantity

T = uu + vv + ww + 2U

formula for the equivalent depth

introdu ed here yields a simpler

that gives the pressure; it is

p = C z T,

as long as ea h uid parti le is a ted upon only by gravity. If however any

parti le

dire tions

is subje ted to a elerative for es whose omponents along the

AF AB
,

, and

AC

are respe tively

Q q
,

and

then a similar

al ulation gives the pressure as

p=C+

It is lear then that

(Qdx + qdy + dz) T.

Qdx + qdy + dz

must be a omplete dierential, to be

ompatible with a state of equilibrium.

The elebrated Monsieur Clairaut

has indeed shown with great larity that su h a ondition must be imposed
on the for e omponents

82.

Qq
,

and

[Appli ation to hydrostati s and hydrauli s.

At rst glan e the prin iples

of the general theory of uid motion did not seem very fruitful, yet almost
everything that is known about hydrauli s and hydrostati s is ontained in
them, so it must be allowed that these prin iples have a very broad rea h. To
see this more learly, it will be worth while to show exa tly how the known
pre epts of hydrostati s and hydrauli s follow in a lear and straight-forward
manner from the prin iples developed so far.

55

83.

Let us therefore onsider rst a uid at rest, so that

w = 0.

Sin e then

T = 2U,

the pressure at any uid parti le

p=C+

Sin e

is a fun tion of time

into the onstant

C,

Qq
,

of the axes

84.

and

and

will be

(Qdx + qdy + dz) 2U.

t, whi h we take xed,

we an fold this quantity

so that

p=C+

where

u = 0, v = 0

(Qdx + qdy + dz) ,

are the for es a ting on the uid parti le in the dire tion

AL AB
,

Sin e now

42

of the oordinates

and

AC

respe tively.

the pressure

x y
,

and

is a fun tion of the position of

, then the form

(Qdx + qdy + dz)

that is,

must be

a denite integral fun tion of these oordinates. Then it is lear, using the
same sort of argument as above, that the uid ould not be in equilibrium
unless the for es a ting on the individual uid elements are so onstituted
that

Qdx + qdy + dz

be a omplete dierential. If we set its integral

then the pressure at the point

a ting in the dire tion of

CA

= P,

will be p = C + P. If the only for e is gravity

, then

p = C z.

is spe ied, thus giving the onstant

If the pressure at one point

, then, for that one moment in time,

the pressure at all other points will be ompletely determined.

42 For a xed moment in time.

56

85.

With the passage of time, the pressure at ea h position an hange.

This will surely happen if the external for es a ting on the uid are variable,
as these are not restri ted ex ept that they remain in equilibrium and do not
produ e uid motion.

On the other hand, if these for es do not suer any

hange, then the literal

a tually does signify a true onstant independent

of time, and in that position

the same pressure

p = C +P

will always be

found.

86.

[Free surfa e. In a permanent state, the uid's boundary an be deter-

mined if the boundary is not subje ted to any for e. In a vessel, on the free

43

surfa e

where the uid is not onned by walls of the vessel, the pressure

must ne esarily be zero.

Then the equation will be

P = onst.,

and the

shape of the free surfa e is thereby expressed as a relation among the three
oordinates

xy
,

and

. At the free surfa e we may set

for any internal point

the pressure will be

are subje t only to gravity, so that

have

87.

z = C,

P = E, C = E;

p = P E.

and

If the uid elements

p = C z, then at the free surfa e we shall

from whi h we may on lude that the free surfa e is horizontal.

[Flow through narrow tubes.

Finally, on erning ows through tubes,

everything that has been teased out by various means are easily dedu ed
from these prin iples.

The tubes are usually taken to be very narrow, or

43 In this paragraph, Euler uses the terms extrema gura, extremitas, extrema super ies, and other ombinations of these words, not using the same phrase twi e. He
lands nally at extrema super ies libera.
modern term free surfa e.

57

I have translated all these terms with the

else the ow is assumed to be uniform a ross any normal ross-se tion of the
tube.

From these assumptions arose the rule that the speed of the uid at

any position in the tube is inversely proportional to the area of the rossse tion. So let the shape of the tube be expressed by two equations among
the three oordinates

other two oordinates

tube.

88.

xy
z
y z
and

and

; so that for any value of the abs issa

an be dened. Let also

, the

any point of the

44

Let moreover the area of the ross-se tion at

xed position of the tube let the area be

dt

after an innitesimal time

time

be

t,

as will be

time will be
in terms of

d
/dt.

, so that

omponents at

Thus

will be a fun tion of

The velo ity of the uid at point

V = f f /rr.

= f f, while the speed is = ,45 whi h

+ d
.

be omes

be = rr, and at another

From the shape of the tube,

dy = dx

and

in the dire tions of

dz = dx;

AL AB
,

and

at the urrent

and

are given

when e the three velo ity

AC

are respe tively

46

f f
1

;
rr 1 + +
f f

v=
;
rr 1 + +
f f

w=
;
rr 1 + +
u=

44 The marginal note refers to Figure 2, whose re-use is a bit of a stret h. Or did Euler
lift this passage from another manus ript whi h ontained additional gures?

45 Euler used the astronomi al symbol whose teX ode is \taurus. One is bound to

have strong feelings about using su h a symbol.

substitute the makeshift

46 These are the speed

The ombination

f f

Printing di ulties have for ed me to

is the mass ux.

V times the dire tion osines.

58

Tab.

IV

Fig. 2.

uu + vv + ww = V V = f 4 /r 4 .

so that

well as the dependent variables

89.

udx + vdy + wdz

Sin e

integral to be

= S,

dS =

But
as

however

then

and

is a omplete dierential and we an take its

47

ff
ds
. For a xed time
rr

S =

of the tube.

t,

If we write this

this is a fun tion of

Sin e

, but only on the shape of the

f f ds
.
rr

To nd the pressure

as

represents the element

must onsider the quantity

time

do not depend on the time

tube, it follows that

90.

is a fun tion of

f f dx (1 + + )
f f p

dx 1 + + .
=
rr
rr
1 + +

dS =

rr

and

rr

there results:

dx 1 + +

= ds,

The term

p
U

whi h obtains at the point

arising from dierentiating

in the tube, we
when only the

R
U = dS/dt. Sin e the integral form f frrds
R f f ds
d

;
dierential will be dS/dt = U =
dt
rr

is variable  that is, su h that

does not involve the time

, the

and then, from paragraph 80,

f 4 2d
T = 4 +
r
dt

For any posited for es

Q q
,

and

f f ds
.
rr

a ting on the uid, the orresponding

47 ar -length.

59

pressure at the point

p=C+

will be:

f 4 2d

(Qdx + qdy + dz) 4

r
dt

f f ds
.
rr

This is the formula that was to be extra ted for uid motion through a tube.
Sin e we allowed any sort of for es a ting on the uid [in this derivation, it will
hold all the more so when the only for e is gravity. It is well to re all that the
for es Q, q and

had to be so onstituted that the form

Qdx + qdy + dz

be a omplete dierential, i.e., that it be integrable.

Expli it dissertatio de prin ipiis motus uidorum au tore Leon. Eulero.

60

Appendix: Table IV

61

Translator's omments

I nd that Euler's Latin is in general very pre ise. Almost all obs urities in his
meaning or ambiguities in his equations an be resolved after areful parsing
of his language. The following omments on individual paragraphs are meant
to help the reader relate Euler's dis overies to modern treatments of invis id
in ompressible uids.

This requires us to take advantage of mathemati al

on epts developed after his time.

Truesdell gives further dis ussion, extensive and illuminating, in his massive
ommentaries for the various parts of

Euleri Opera Omnia

XII.

Paragraph 17: At this stage, using the determinant formula for the area of
a triangle as it appears in high-s hool texts would immediately give the area
as

whi h






0
0
1 


1 
 dx + Ldxdt
Mdxdt
1 
2 




ldydt
dy + mdydt 1 

= 12 dxdy + 12 (L + m)dxdydt + 21 (Lm lM)dxdydtdt

on lude that

0 = L + m = u/x + v/y.

and we an then

Euler is showing how to get this

determinantal formula.

Paragraph 47: The solution given by Euler says that the url of velo ity is
zero: the ow is irrotational. The ondition that
dierential

udx + vdy

be a omplete

= d, for some potential fun tion , ombined with the ontinuity

equation, implies that the potential fun tion

62

is harmoni .

How did Euler sneak in the assumption of irrotational ow?

ontinuity equation gives

L + m = 0.

The rest of the partial dierential

equation says that the material derivative of

d
dt

u v

y x

being the vorti ity. Euler then adopts

out irrotational ow.

Note that the

lM

vanishes:

d
= 0,
dt

=0

as the solution, thus singling

Of ourse there are other solutions.

What Euler has

done is to show rigorously that irrotaional ow is a valid solution. The whole
argument glows with the ex itement of dis overy of a

Paragraph 60: The quantities

of vorti ity

method

l M, N, n

= U, ex ept for order and sign.

are the omponents

In modern ve tor notation,

we would write the equations as

D
U + [ U] = 0,
Dt
where

D/Dt is the material derivative.

Euler's solution says that the ompo-

nents of the vorti ity are all zero, i.e., the ow is irrotational. There are of
ourse other solutions, just as there were in the planar ase of paragraph 47.
Euler did ome to understand that the irrotational ase was only a spe ial
solution of the general problem. He had given it up by the time of his dida ti

Novi om-

treatise [E396 Se tio Se unda de Prin ipiis Motus Fluidorum [

mentarii a ademiae s ientiarum Petropolitanae

63

14 (1769, published 1770).

Consider now an in ompressible uid. If we let S be the symmetri part of

U,

i.e.

S=

then

o
1n
[U] + [U]T ,
2

U = S. Furthermore, the tra e of

S is 0. Thus, for an in ompress-

ible uid, the vorti ity equation be omes

D
= S,
Dt

where the symmetri matrix S has at least one positive eigenvalue.

This

raises issues about the stability of the solution. These issues are ameliorated
somewhat by the fa t that in

planar

motion the eigenve tor orresponding to

the most negative eigenvalue is parallel to vorti ity. In many other ases the
largest positive eigenvalue an still be expe ted to have little ee t. It may
be that the supreme virtuoso of analyti manipulation had gone far enough
to satisfy himself that the obvious solution

lM = N = n = 0

was

at least reasonable.

Paragraph 70: In the ondition in rule III:

ponents are the

AA, BB, CC

A + B + C = 0,

of paragraph 69.

The

the three om-

A B C
,

...

of this

urrent paragraph are now real numbers.

For the general solution of Lapla e's equation following Euler's line of reasoning, see se tion 18.3 in Whittaker and Watson's

Course of Modern Analysis

Paragraphs 75-77: As Truesdell remarks, Euler fails to a ount here for all

64

se ond order terms and so misleads himself into proving that there an be
no rotational uid motion oin iding with the motion of a solid body. Indeed,
there is a ertain sloppiness in the passage. But mistakes by the residents of
Olympus help us mortals understand how they think.

Paragraph 80: The aim of this paragraph is to establish an a leration potential

whose spatial derivatives are the motive for es written out in

paragraph 78.

Paragraph 81: For a well-dened pressure, the expression

p=C+

(Qdx + qdy + dz) T

needs to be a fun tion of position independent of the path of the integral,

and so

Qdx + qdy + dz

must be a omplete dierential.

established the ne essity, at least lo ally.

with 

Euler has thus

He nishes the senten e however

alioquin status aequilibrii, vel saltem possibilis, non daretur

  other-

wise, a state of equilibrium, or at least a possible one, would not be given.

He seems to say that any for e eld under onsideration must be ompatible
with some possible equilibrium state. He may have d'Alembert's prin iple in
mind. That prin iple, however, applies lo ally, and Euler is seeking a global
ondition.

Truesdell points out that there are winding number issues that

need a more areful analysis for resolution.

Paragraphs 88-90: this nal se tion is breath-taking.

65