[

XVI. Note on the Motion of_Fluid in a Curved Pipe.

By W . R. DEAs, M.A., Imperial College of Scie~lee *
N this paper the steady motion of incompressible fluid
through a pipe of circular cross-section which is coiled in
a circle is considered. It is found necessary to approximate
by supposing that the curvature of the pipe is small, or that
R, the radius of the circle in which the pipe is coiled, is large
in comparison with a, the radius of the cross-section. Ti~e
theory is in good qualitative agreement with ext,eriments
conducted by Prof. J. Eustice ~- on the stream-line motion of
water in curved pipes. A quantitative comparisolb however.
requires a closer approximation than that of this paper ; it is
found that the theoretical results apply only if n:a/1440 R,
where n is the Reynolds' number, is small, while in all the
experinaents the value of this expression is greater than 1.
It is probably for this reason that the interesting result
established experimentally that at increased velocity of flow
the curvature of the stream-lines is increased (so that the)"
follow more nearly the line of the pipe) cannot be obtained
from the theory at its present stage ; however, an alternative
explanation appears possible, and is given in 13.
2. Fig. 1 shows the system of coordinates that has been
found convenient in considering" the motion of fluid through
a pipe of circular cross-section coiled in the form of a circle.
The surface of such a pipe is an anchor ring, of which OZ in
fig. 1 is the axis. C is the centre of the section of the pipe
by a plan~ that makes an angle 0 with a fixed axial plane.
CO, the perpendicular drawn from C upon OZ, is of the
length R~ so that R is the radius of the circle in which the
pipe is coiled. The plane through O perpendicular to OZ
will he called the "central plane" of the pip% and the circle
traced out by C its "central line." P, any point of the
section drawn, is at distance r from C, while CP makes an
angle 4y with a line through C parallel to OZ. The position
of P is then specified by the orthogonal coordinates r, ~, t~.
Th~ surface of the pipe is given by r = a , a being the radius
of any section. The components of velocity corresponding
to these coordinates are U, V, W ; U is there[ore in the
di,'eetion CP, V perpendicular to U and in the plane of the
cross-seeti-n, and W perpendicular to this phme. The
general direction of flow will be taken to b, the direction
in which 0 increases.
The motion of the fluid is supposed to be due to a fall in
prossure along the pipe. There will be a fall in pressure if
Communicated by Prof. S. Chapman, D.Se., F.R.S.

f Prec. Roy. Soc. A, vol. Ixxxv. p. 119 (1911).

.Note on the Motion of Fluid in a Cu~'ved Pipe.

209

fluid enters the pipe from a container at known pressure at

one end and flows along the pipe to a container at lower
pressure at the other end. Except near the ends, where

[Z

Fi~l

<

,R

there is certain to be some irregularity of flow, we may

expect a steady motion in which U, V, and W (but not P,
the pressure) are independent of 8.
The equations for such a motion are these :
u~U+V ~)U y2 W2 sin 4F

-3~
b
ubV+V
=_1

})

~b~

R+~sin
~U

b~ u v
P

~os~

W~
R+rsin~
[_~+ s i n ~
"~[~V+V

(1)

(2)
Phil. Ma 9. S. 7. Vol. 4. No. 20. July 1_927.

Mr. W. R. Dean : Note on the

210

and
ubW+V~W

U W s i n 4F.~ V W c o s ~

-5~ ~ b ~ + ~ ~
1

"R+~sin~

3{P~+v['(

~ +I~(bW+

Wsin~_~

W c o s ~ ~-]
+ 1 b [lbW.
(3)
~-+----f-R + r sin @].J"
The fluid is supposed incompressible, so that the equation
of continuity is
3U

Usin ~

1 3V

Vcosq~

~+~-+ R+r sin~+ ; b-~4 R + ~ . s i n @

= 0.

(4)

These four equations reduce to equations for the corresponding motion in spherical polar coordinates if w e write
R = 0 , and to equations in cylindrical coordinates if we write

l / R = 0 and b/RbO=b/bz*.
3. We now introduce the assumption that the curvature
of the pipe is small : that is, that air is small. If the pipe
were straight a/R would vanish, and the equations could be
satisfied by
U----V---0, W=A(aS--r'), P/p=Cz,
where A and (3 are constants, and z is the distance (measured along the central line) of any section of the pipe from
a fixed section. For the slightly curved pipe we assume
U = u , V = v , W=A(a~--rS)+w, P/p=Cz+p/p,. (5)
where u, v, w, and p are all small and of order a/R.
If terms of order a'~/R2 are ignored, the equation of continuity is
bu u i bv

~+~+~-~=

o,

(6)

while the first two equations of motion become

(7)
and
A~(a~--r~)2co s@__

1 3

Writing, in conformity with (5),

b/R3t~=bFOz,we

(s)
have

If either substitutionis made, the resulting equations are not,

however, in the form in which they are usually quoted; the expressions multiplied by v in equations ~1) to (3) are actually the
components of -curl curly, where v is the velocity,but this vector
is equal to V~ v since div v--O.

211

Motion of Fluid in a Curved Pipe.

from the third equation of motion

~2
+ v [~b+z~ a1 ~ ) + 1;, ~-q,)
IA(~'-r,) + w)
+ ~[\~b +l~fA(~'-r');.]\ -g sin 4p}
+ rv ~ { A ( a ' - - ~ )

eosq~},

The terms of this equation that are not small must vanish ;
these are (--C--4vA). We therefore have
C --.-4pA, . . . . . . . .
(9)
which gives the relation between pressure g.radient and rate
of flow in a straight pipe of circular secuon*. We then
have the equation

~(p)

-~Ar.=--

"

/?'w

l ~)w 1 ?'w'~

- - ~ A ~ + ~ k~ +~ ~ +;,~-~i.
(10)

The four equations (6), (7), (8), and (10) determine the
motion.
4. From equation (10) i)p/~z must be a function of r
and ~ , so that p must be of the form A z + B, where A and B
are functions of r and ~ . If, however, this expression for/9
is substituted in (7) and (8), it appears that A must be a
constant; hence/) may be taken to be a function of r and 4/"
only. I f we now write
u = u' sin @,
v = v' cos ~ ,
w = w ' s i n % p/o=p' sin@,
where u ~, v r, w', and pr are functions of r alone, the four
fundamental equations become
dt~ !

u ~

v!

-dr "1". r. . . r.
A~(a~-.~) ~
R

0,

dp I

v/dr I

v'

(n)

u'\

=- ~+~,(~+---)'

A'(e~r'- )'_. = - - ; + ~

and

a lay'

,.

(~)

v' u'"

+--T]'~

(za)

/d'~,+l d~,_,~',~

-- 2Aeul -- -- 6vA ~t + v ~-d~r2

r dr

r~]"

* H. Lamb) ' Hydrodynamics' (4th edition)) 331.

P2

(1~)

212

Mr. W. R. Dean : Note on the

If p' is eliminated from (12) and (13), the following

equation results :

+ d

:)i,z,,'+v,

The solution of this, regarded as a differential equation for

dv'/dr + (v'--u')/r, is

dr'[dr-4- (v'-- u')/r ~. B/r-t- C," + A'~r~(3a~-r2)/6Rv,

(15)

B and C being arbitrary constants. Substituting in equation (15) the value of v' given by (11), we have
+ -d2u'
- du' = B/r + Cr +
whence
u'= D + E l f ~+ (B log r)2 + Cr2/8 + A~r4(6a 2-r~)/288Ru,
where D and E are arbitrary constants. It follows that
v ' = D - - E / r 2 + B (1 + log r)/2 + 3(_~r~/8
+ A2r 4(30a ~ - 7r~)/285Rr.
We must have
B=E=0;
otherwise the velocity of the fluid at points of the central
line (r=O) will not be finite.
It is interesting to notice that another condition is automatically satisfied. When r = O we must have
U ! ~

V/,

and this is the ease since each of these expressions is equal

to D. It appeared at first sight that the conditions
u~=v'=0,
r=0
were necessary, and it is the fact that the conditions
w p~io r= O~ r = 0
are necessary for a solution of physical significance. For
consider two points on the line OC (fig. 1) which are near
to, but on opposite sides of, C. At one point ~-=~r/2,
sin q/'=l, while at the other qy=-3~r/2, s i n ~ = - - l .
There
will therefore be a finite difference in pressure, for instance,
at these two points, however close they may be together,
unless at the centre p r = 0 . A similar consideration shows
that w~ must vanish at the centre. But the case of u' and v'
is different. Let the component in the plane of the section
of the velocity of the fluid at C be of the magnitude V~ and
in the direction of the line ~ = ~ r 0. At any point (r, ~ ) in
the immediate neighbourhood of C the velocity in the plane

Motion of Fhdd in a Curved Pipe.

213

of the section will be given approximately by

u = Vc cos ( ~ - 4f0), v ------Vc sin (~-4/.o).
The component velocities (u ~sin 5k, v r cos ~ ) found above
are of this form only if u'=v', and it has b~en found that
this condition is satisfied. Moreover, it is clear that we
must have ~0----~r/2; hence the velocity at the centre of
the pipe is of" magnitude D and in the direction of the
line OC.
At the surface of the pipe, r = a , the velocity components
u and v, and hence u' and v ~, must vanish. Consequently
D + Ca~/8 + 5A~a6/288Ru =- 0
and
D + 3Ca~/8 + 23A2a6/288R~ ---- O,
whence
C -- -- A~a4/4R~,,
D -----A:a6/72Rm
The resulting expressions for the two velocity components
are
u -= A ~ sin ( a 2 - r~)2(4a2--r2)/288Rv
(17)
and
v = A

70)/28

(18)

5. 5Tow that u' and v' are known, p', upon which depends
the pressure, is determinate. Before the results of (17)
and (18)can be accepted, it must be shown that p' and w'
vanish when r = 0 . From (13)
2 ~
2 2"
d[dv'
v"
~{\
and since it is clear that the terms on the right-hand side of
this equation are finite when ~=0~ it follows that p~=0
when ~'=0. From (14) and (17),
d'w' l dw' w' 6At
A3r
dr ~ +~" dr
~.~= I%
1 4 4 ~ (a2--r~)2(la~'-r~)"
The solution is

w' =- Fir + Gr + 3Ar3/4R

A3
1152Rv ~(4a%3-- 3a4r~
+ a~r 7 __

~,9/]_0).

F must vanish, and it is then evident that w ' = 0 when r=.0.

The necessary condition that w' and p should vanish at the
central lin~ is therefore fulfilled.
The boundary conditions require w ' = 0 , r = a ; hence
wr

3Ar ~ :
A3r [4a~(a~_r~)_3cd(a~_r4 )
~ (a - - r ) + 1152Hv~

(19)

214

~ r . W. R. Dean : N o t e on the

6. The results of (.17), (18), and (19) can he expressed in

a more convenient form. Let Wo be the value of W at the
centre of the pipe ; from (5) and (19),
W0 = Aa ~, . . . . . .
(20)
and is therefore constant. AaS/v, or aWe~v, is a nondimensional constant, the value of which determines the
nature of the flow ; we write
n = Aa3/v . . . . . . .
(21)
Finally let

(22)

Then
U/W0 = na sin ~ ( 1 - - r ' 2 ) ~ ( 4 - - r ' ~ ) / 2 8 8 R ,

(23)

V/W0 = na cos ~ ( 1 - - r ' 2 " ) ( 4 - - 2 3 r '~ + 7r'4)/288R,

(24)

and
W/W0 = ( 1 - - r '~) 1

3r sin ~ 4 n~r sin 4f { 1 9 - - 2 1 r '~

4R
11520R

(25)
7. It is k n o w n * that to cause a given rate of flow
a larger pressure gradient is required in a curved pipe
than in a straight one, the difference being considerable
even when the curvature is small. From this it was
anticipated that the difference would be found to depend
on the first power of a / R , but the preceding work shows
that this is not the case. I t is clear that when W is
integrated over the whole cross-section the terms involving
s i n ~ disappear; the rate of flow is therefore ra4A/2.
The pressure is not, of course, constant over a cross-section
of the pipe ; so that there is not a constant pressure
gradient as there is in the case of flow through a straight
pipe. But it is natural to define as the mean pressure
gradient the space-rate of decrease in pressure along the
central line of the pipe. lqow, it has been pointed out
that p, the part of the pressure that varies across the
section, must vanish at the central l i n e ; the pressure at
any point of this line is therefore --4rAz, and the mean
pressure gradient 4vA. The relation between this quantity
and the rate of flow is therefore the same as if the pipe
were straight. A closer approximation t than that of this
Thisis stated by Eustice, lee. tit. p. 119 ; cfi also J. H. Grindley and
A. H. Gibson, Prec. Roy. Soc. A, vol. lxxx. p. 114 (1908).
Note added.--~cVork done since this paper was written has shown
that it is necessary to retain the terms of order a~/R 2.

Motion of Fluid in a Curved _Pipe.

215

paper is then required to find the relation between, rate of

flow and curvature [or a given pressure gradient.
8. The motion of t~he fluid is of special simplicity in the
central plane of the pipe. At any point on OC, qF is
either 7r/2 or 3~'/2 ; in either case cos ~ , and with it V,
vanishes. At any such point the direction of the velocity
of the fluid lies in the central plane; hence a particle of
the fluid once in this plane does not leave it in the subsequent motion. The motion in one half of the pipe is
therefore quite distinct from that in the other.
The differential equation to the stream-lines in the central
plane is
dr
(R_+~')d0
UW
'
but with sufficient accuracy we can ignore r in comparison
with R and write A(a~--r 2) for W. We then have
dr

R d O - W 0 ( 1 - r '~)
:

or

na(l --r '~) (i-r':li)/72R,

dr'

d-O = n(1--r")(1--r'2/4)[72'

(26)

by substituting for U from (23) and writing s i n q / . - - + l .

This equation will give a first approximation to the streamlines, but only to those parts of them on the outside of
the central line. To get the other parts we must write
s i n ~ = - - 1 , and the sign of equation (26) must be reversed.
Thus, while on the outside of the central line r' increases
with 6, the contrary is the case on the inside ; but on both
sides the general motion of the fluid is the same, namely
a continuous movement from the inner to the outer edge
of the pipe. It is, in fact, clear from (23) that the direction
of the velocity component along OC is the same at all
points of OG. It follows from (26) that

2~

F(i +r')'(i-l~)]

0 = ~- log L(1_r,),(l+r,/2)_ ],

(27)

if 0 is measured from the point where the stream-line

crosses the central line +"=0. Evidently t~ increases
steadily with r', but since 0 tends to infinity as the value
of r' approaches 1, the stream-line never reaches the outer
edge of the pipe. The same expression for 0 as in (27),
but with the sign reversed, gives the equation to the part
of the stream-line on the inside of the central line.

216

Mr. W. R. Dean : Note on the

For given r r the value of 0 varies inversely as n, and

hence inversely as the general velocity of flow. Consequently the angular distance in which a stream-line
starting from the centre gets to within a given distance
from the outer edge of the pipe is smaller the larger the
mean velocity. The relation between 8 and r z does not,
however, involve a]R~ and is thereiore independent of the
curvature of the pipe.
The table shows the relation between 0 and / on the
assumption that g, measured in degrees, is 50 times
the logarithm to base 10 of the function of r r in equation (27). The corresponding value of n is 63"3; for a
larger value n" the values oE g in the table must be reduced
in the ratio 63"3]n I.
'l'aBL~.--Relation between r' ,~nd 0.
r'(=r/a) 0 0'1 0'2 0'3 0"4 0'5 0"6 0'7 0"S 0"85 0'9
0 (degrees)0 6'6 13'3 20"3 280 36'6 46"8 59"5 77"0 89'4 106'8
The form of the whole stream-line is shown in fig. 2,
the direction of motion being indicated by the arrow.
If the curve drawn is rotated as a whole about O through
any angle the curve so obtained is another central streamline, all these lines being of exactly the same form. In
fig. 2 a/R has been assumed, for convenience, to have the
large value 1/3; the r', 0 relation is not, as has been
seen, affected by the value of a/R, but the approximate
results above would not, of course, apply to a pipe with such
large curvature.
The figure shows plainly a steady motion of the fluid in
the central plane from the inner to the outer edge of the
pipe. Since the fluid is assmned incompressible, it is clear
enough without further analysis -that at a point near the
central plane and on the outside of the central line the
motion must be directed away from the central plane, while
it must be towards the central plane at a point near this
plane and inside the central line.
9. The differential equation to any stream-line is

dr

U =

rd~

RdO

V =-k(a~-r') ' . . . . .

(28)

the.same approximation as before being made in the last

ratio. W h a t is of most interest is the variation in position
with regard to the central lino of a fluid element as it
move~ along the pipe; this variation is given by the

Motion of Fluid in a Curved -Pipe.

relation between r a n d ~ .
t w o ratios in (28) we have

F r o m the equality of the first

dr
(1 -

217

rd@

r'2) (4 - r '') sin ~ -- ( 4 - - 2 3 r " + 7 r " ) cos ~ '

or

4 - 23ri2+ 7r r4
tan 'k d ~ = r'(~ --r '~) ( 4 - - r '~) dr'.
Fig. 2.

Stream-line in the central plane.

B y integration,
sec 4F ~-- k r ' ( 1 - - r ' 2 ) ~ ( l - - r ' 2 / 4 ) ,

(29)

w h e r e k is an a r b i t r a r y constant.
The relation between r ' and ~ ~herefore depends neither
on a/R nor on n. F o r a n y value of k a closed (r', ~ ) curve

218

Mr. W. R. Dean : .Note on the

is obtained. If k is positive, see ~ is positive for all values

of r' concerned, and the curve is in the upper half of the
cross-section. For an equal and opposite value of k the
closed curve obtained is the reflexion in OC of the previous
curve: it has already been seen that the motions in the
two parts into which the pipe is divided by the central
plane are independent. A series of the curves is shown in
fig. 3. They represent what may loosely be called the
Fig. 8.

Lines showing the movement of fluid elements in the cross-section

of the pipe.
projections of the paths of the fluid elements on the crosssection of the pipe ; the direction of motion of the elements
is shown by the arrows. On this motion is, of course, to be
superposed the motion of the elements along the channel ;
the closed path of the projection corresponds to a helical
motion of the fluid element.

Motion of Fluid in a Curved Pipe.

219

From equation (24), V vanishes for all values of

when

7 d 4 - - 2 3 r " + 4 ---- 0,

the only relevant solution of this eqnation being r'=0"429.

At the points where ~1=0"429 and ~ is 0 or 7r, both
U and V vanish; these are the points denoted by crosses
in fig. 3. The two stream-lines through these points are
clearly circles in planes parallel to, and equidistant from,
the central plane. The motion of the fluid as a whole can
be regarded as made up of what are roughly scraw motions
in opposite directions about these two circular stream-lines.
The general nature of the motion suggested by the theory
is now clear.
10. The precise relation between 6 and the other variables
is of little interest, for the angular distance (0) in which the
projection of a fluid element describes a closed path such as
those of fig. 3 must tend to infinity as the minimum distance
of the path from C tends to zero. This is clear from the
motion of the fluid elements in the central plane, but may
be shown directly as follows : n P a r t of a closed path that is
at one point very near to, and above, C must practically
coincide with the whole of the upper semicircle of the
section of the pipe. 0 can be found from the differential
equation
dO =
288r'd~
n cos ~F(4--23r '~ + 7r'4)"
l~ormally the relation between r r and ~ must be known
before ~ can he found, but for points practically coincide,it
with the boundary we can write r r-- 1. Then
24
dO

--

--

sec,/~d,

and the angular distance in which a fluid element near the

boundary goes from the point where ~ = a to that where
~ = 0 is proportional to logtan -~+-2 This expression
is large if ~ is nearly ~r/2.
The only point of interest is that for given r ~ and d~,
dr? is inversely proportional to n, and therefore varies
inversely with the mean velocity. It has been shown that the
form of the closed curves of fig. 3 is independent of the
velocity of the fluid, but on the other hand thi~ angular distance
in which they are completely described by the projection of
a flu':d element is inversely proportional to the mean velocity.

220

l~/ir. W. R. Dean : Note on the

11. So far as the general nature of the motion i s concerned, the theory is in complete agreement with the
.experiments of Eustice on the stream-line motion of water
m curved pipes of circular cross-section. In the experiments the motion was made visible by the introduction
into the stream at various points of dyed water, which was
drawn out into thin coloured lines fixed relatively to the
pipe. A coloured line in the central plane of the pipe was
roughly o[ the form of the stream-line shown in fig. 2,
showing that the fluid elements near the central plane
moved steadily across from the inner to the outer edge
of' the pipe. But when such a line approached the outer
edge, it was found to break up into two coloured bands,
one of which went round the boundary of the pipe, above
the central plane, to the inner edge, while the other described
a similar path below the central plane. This is exactly the
motion that the theory would suggest. What has been called
a eoloured line in the central plane consists, of course, of fluid
elements both above the plane and below i t ; it is evident
from fig. 3 that according to the theory the coloured matter
of a line in the central plane should be divided into two
parts. Again the theory accounts for the distinction implied
by the above use of the word bands (as opposed to lines).
Two fluid elements, both on the same side of the central
plane and both near to it, describe (relatiwly to the section
of the pipe) closed paths which are close to~ether. But we
have soon that the angular distance in which such a closed
path is described tends to infinity (somewhat as a logarithm)
as the minimum distance of the path from the central line
tends to zero. Thus, if the distance from the central plane
of one of the two fluid elements is twice that of the other,
the ratio of the angular distances in which they describe
their respective closed paths is likely to be in the neighbourhood of log 2. Though originally near together, two
such elements will ultimately be far apart. The coloured
matter originally concentrated begins therefor~ to be
dispersed as soon as it approaches the outer edge of the pipe.
The motion described above continues : the bands which
reach the inner edge then move to the outer edge, remaining
near the central plane, and thence round the boundary to the
inner edge again. Fig. 5 of Eustice's paper shows in a pipe
of small radius of curvature (aiR approximately 1/7) a
co]ouved line which twice approaches the outer edge and
twice the inner edge in an angular distance of less than 27r.
The dispersion of the coloured matter must however continue,
and the theory suggests, what is found to be the fact, that

Motion of_Fluid in a Curved Pipe.

221

t h e " distinctive character of the co[our filament is gradually

lost" a.
On the other hand, experiment shows that a coloured line
at a sufficient distance from the central plane and, say, above
it, is in the form of a helix which is entirely confined to the
upper part of the pipe, and that such a line is not dispersed
into a band ; in both respects the theory agrees.
12. It is not possible to compare numerically the theory
at its present stage with the experimental work. It is
usually difficult to make a definite statement as to the
limits within which a first approximation such as this is
valid; but a rough idea of the range of validity can be
formed from the above expressions for the velocity components. It has been supposed throughout that w is small
in comparison with the velocity component W of which it
is a ~art, and therefore the term in the second bracket
in the expression for W given by equation (25) must
be nearly 1. The expressio~ r'(19--21rr~+9r'~--r '6) is a
maximum (for values of r / between 0 and 1) when r / is
approximately 0"65, its value being then about 7"6. Hence
n~a/1440 R must be small. Again, U and V have been supposed small in comparison with W. From (23)and (24)
this supposition requires hal72 R to be small, but except for
relatively small values of n the former condition is the more
stringent. But how small n~a/1440R must be for a given
order of accuracy to be attained it is difficult to say, for w is
actually 0 at points of the central line, and what is of most
importance is presumably the accuracy of the assumption so
far as flow near the centre of the pipe is concerned.
In most of the experiments the values of n'Za]1440R are,
however, greater than 1, so that the theory certainly cannot
be applied numericall]r. The least value of aiR is "005, for
a pipe of I cm. diameter coiled in a circle of radius 100 cm.
The least mean velocity of flow given for this pipe is 6"4 cm.
per see. ; the corresponding value of n (the temperature of
the water being 18 C.) is over 500, while that of n2a/].440R
is nearly unity.
13. After a good general agreement between theory and
experiment, it is surprising to find what appears to be a
complete discrepancy, even although a detailed comparison
is out of the question. [t is found experimentally that the
effect of increasing the mean velocity is that the curvature
of the ooloured lines is increased: the angular distance in
*Zoc. cit. p, 123.

222

Note oJ~ the .]lotio~, o/ F l u i d in a Curved P i p e .

which a coloured lille in the central plane reaches the outer

edge and divides into two bands is found to increase with
the velocity. In the limited range wherein the theory
applies, the contrary appears to be the case. In 8 it has
been seen that the angular distance in which a central
stream-line passes over some given fraction of the radius of
the pipe towards the outer edge varies inversely as the mean
velocity. A similar result has been found in 10 in the
case of any stream-line*.
The figures in Prot: Eustice's paper make his conclusion
quite clear. The most direct of the comparisons of flow in a
given pipe at different ~eloeities is that shown in fig. 3 (25) t
of his paper, as in the case there illustrated the coloured lines
that are compared start from exactly the same point of the
cross-section. This figure shows the flow in a pipe of 1 em.
diameter bent in a circle of 25 era. radius. When the mean
velocity is 5"5 cm. per see. a coloured ]i,~e starting practically from a point of the central line twice reaches the outer
edge and twice the inner edge in an angul'tr distance of little
more than vr/2 ; at twice this velocity a colom'ed line starting
from the same point reaches each edge once only in the same
angular distance. The discrepancy may he due merely to
the fact that no direct comparison between theory and experiment is possible ; but there is another way in which the
results might be reconciled. Both theory and experiment
agree in that the coloured matter of a central filament is
dispersed as soon as it approaches the outer edge of the pipe.
Now the coloured matter originally at greatest distance from
the central plane suffers the least dispersion, and it rnust be
this matter which preserves the distinctive character of the
pair of bands (into which the original single line is divided)
after they have been once round the upper or lower half of
the boundary. But the angular distance in which either
of the bands describes its path from the outer to the innel
edge is considerably affected by the distance of closest
approach to the central plane of the eoloured matter composing it ; and this distance will be in effect the maximum
distance of the eoloured matter of the original filament from
the central plane. The original cross-section of the eoloured
filament is therefore of fundamental importance in deciding
how rapidly it will subsequently cross from one edge of the
In the hope of elucidating "this matter the writer has recently

considered the flow of fluid through u sinuous ehannel~ but also in

this case the departure, in a given distance, of a stream-line from
the centralline of the channelis directly proportional to the velocity.

?Zoc. tit. p. 122.

On the Spectrum of .Neon.

223

pipe to the other ; the smaller the cross-section the more

slowly will the repeated crossings take place. The discrepancy between theory and experiment therefore disappears if it is the case that at increased velocity of flow
the cross-sections of the original filaments are sufficiently
reduced. In any case it is clear that a numerical comparison between theory and experiment, so far as a filament
originally in the central plane is concerned, is impossible
unless the cross-section of the filament before dispersion
takes place is known.

X V I I . A _Note on the Spectrum of _Neon. B y MEGttNAD

SAHA., .D.S., .F.R.S., Professor of Physics, AUahabad

University, Allahabad, India *.

H O U G H the spectral lines of ~leon have been completely
grouped into series by Paschen, the nature of the
T
series terms was not clearly understood, and a good deal of
discussion has been devoted to it.

Pasehen t discovered

A set of four terms (s~ s~ s4 s~)

of value ranging between 38040-39887 ;
a set of ten terms (/01 . . . . . . . . . p10)
of value ranging between 20958-25671'65 ;
a set of 12 terms (dl all' . . . . . . . . . d6, SxI, el" . . . . . . s " ' )
of value ranging from 11493-12419.
None of these terms, however, constitute the fundamental
level of Neon, which must have a very large value corresponding to the observed ionization potential of 21 volts. This
level was discovered by Hertz ~, by means of his vacuum
spectrograph.
~t gives rise to two lines ~.ffi=735"7 and
X~ffi743"5, separated by a frequency interval of 1428, which
is just the difference between the values of Pasvhen's s~ and
s4 terms.
From these data, and from a discussion of data on the
Zeeman effect of Neon-lines, Goudsmit has proposed the
following new designation of Paschen's terms. Goudsmit's
* Communicated by the Author.
t Pasehen, Ann. d. Phys,, eel, Ix. and lxiii.

:~ Hertz, Zs.fiir Physik, eel. xxxii, p. 933.

,~ See Gou&mit, Zs. fi Phys~7~. vo]. xxxii, and Back, eel. xxxvii.
p. 197.