Beruflich Dokumente
Kultur Dokumente
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
The Royal Society is collaborating with JSTOR to digitize, preserve and extend access to Proceedings: Mathematical and
Physical Sciences.
http://www.jstor.org
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
1. Introduction
The purpose of this paper is to gather togethera number of results concerningthe
heticityX#of an arbitrarysolenoidal vector fieldB(x) confinedto knotted or linked
tube-like structures, particularly as regards its topological interpretation. This
when the fieldB is confinedto two linked oriented
interpretationis straightforward
fluxtubes carryingfluxes01 and 02: provided each tube is unknottedand the field
lines withineach tube are unlinkedclosed curves circulatingonce parallel to the tube
axis, the helicityis given by
Ye = 2n0>l 02,
16
Vol. 439. A
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
H. K. Moffattand R. L. Ricca
412
fluxtube with flux0, then the helicityis related to the topology ofthe knot in a more
subtle way. On purely dimensional grounds, a result of the form
XY = h02,
((1.2)
where h is a real number determinedpartly by the topology of the knot and partly
by the twist of the fieldB within the knot tube, is to be expected. Since this twist
can be arbitrarilyprescribed(equivalent to an arbitrary'framing' of the knot), the
number h can take any value, positive or negative. Nevertheless,the fact that h is
then invariant under topological deformationof the knot tube and the fieldwithin
it, does carry some importantinformationabout the knot itself.
This type of problem appears to have been firstaddressed by Calugareanu (1959,
1961; hereafter referred to as C59 and C61). Calugareanu considered two
neighbouringclosed curves C, C* formingthe boundaries of a (possibly knotted)
ribbon of small spanwise width e, and showed that the linkingnumbern of C and C*
can be expressed in the form
n = 0" + !7 + X
(1.3)
(this equation is given at the end of p. 613 of C61), where f and Y are respectively
the writheand the normalized total torsion of C (fordefinitionof these quantities,
see (4.8) and (5.13) below), and X is an integerrepresentingthe numberof rotations
of the unit spanwise vector N on the ribbon relative to the Frenet pair (n,b) (unit
principal normal and unit binormal) in one passage round C. For the moment,we
simply note that Y and X are well definedonly if C has no point of inflexion(i.e.
no point at which the curvaturevanishes). If C is continuouslydeformedthroughan
inflexionalconfiguration(i.e. througha state that does contain a point of inflexion)
then, as will be shown in ?6 below, S is discontinuous by ? 1, but X is
simultaneouslydiscontinuousby an equal and opposite amount 411 as a resultofthe
discontinuous behaviour of the Frenet pair (n, b) in going throughthe inflexion,so
that the sum 7 ?+X does vary continuously.
The difficulty
associated with inflexionpoints was recognizedby Calugareanu (see
the long footnote on p. 8 of C59) and was to some extent resolved through
consideration of a particular example of deformation through an inflexional
configurationin C61 (pp. 616-617). Deformations of this type were described as
'degenerate' by Pohl (1968, p. 83); in fact,as pointed out by Ricca & Moffatt(1992),
any deformationwhose projection on any plane involves a type I Reidemeistermove
(see, forexample, Kauffman 1991,p. 16) must involve passage throughan inflexional
configuration.General deformations(or 'ambient isotopies') do thereforetypically
involve such passage and consideration of the associated behaviour of Y and X
cannot be avoided.
The concept of the self-linkingnumberSL of a closed curve C having no points of
vanishing curvaturewas introducedby Pohl (1968). SL is definedas half the sum of
the indices of the cross-tangentsof C (i.e. the tangents of C which intersectC in a
point distinctfromthe point of tangency). Pohl showed that SL is an integer,and
he proved that
SL = Ir +,7.
(1.4)
Under regular isotopy (i.e. continuous deformationof C not passing through any
inflexionalconfiguration),SL is invariantso that Pohl's resultprovidesan alternative
proof of the invariance under regular isotopy of # +?F, as proved in C59. Pohl's
work was extended to higher dimensions by White (1969), who again restricted
Proc. R. Soc. Lond. A (1992)
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
Heticityand theCaltugareanutnvartant
413
Tw -2
(1.5)
Tw =J' + X,
(1.6)
(a result nowhere actually stated by Fuller). Fuller then definesthe writhe f'through
IF =
(1.7)
n-Tw,
wheren is the linkingnumberof the curves C, C* bounding the ribbon,but nowhere
does he prove that this F is the same as that definedby the Gauss integral(equation
(4.8) below). This identificationis however established by Ca]ugareanu's result (1.3)
in conjunction with (1.6). The definition(1.5) of Tw provides a quantity which
evidently varies continuouslyunder all continuous deformationsof the ribbon (i.e.
under ambient isotopy).
Curiously, Fuller (1972), in a paper dedicated to Calugareanu (on his 70th
?+ Tw, although this
anniversary),gives White (1969) the creditforthe resultn =
result (with Tw = Y + X ) can be found clearlystated, and coupled with a tentative
discussion of the role of inflexionpoints, in C61. White's achievement was to place
this result in the wider context of differentiablemanifoldsof arbitrarydimension;
but the theoremin the form(1.3), or in the equivalent form
n =K+Tw
(1.8)
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
414
H. K. Moffattand R. L. Ricca
is given by JX = nO2 (the proofwas given by Ricca & Moffatt(1992), but is repeated
here for completeness). In ?4, the helicity is decomposed into writhe and twist
contributions,the writhe contributioninvolving the Gauss integral (4.8), which
admits interpretationin termsof the sum of signed crossingsof C averaged over all
projections. In ?5, the twist contributionis considered,part of this being associated
with torsionof C and part withwhat may be describedas 'intrinsictwist' of the field
lines in the flux tube around C. In ? 6 the generic behaviour associated with
inflexionalconfigurationsis analysed and the role of the twistparameteris discussed
in ? 7. Finally in ? 8 we summarize the conclusions. Our hope is that the alternative
proofpresentedin this paper and the associated discussion may help to demonstrate
the relevance of fluiddynamical techniques to topological problems.
= {A
(2.1)
.BdV,
where dV is the volume element d3x. Note immediatelythat X does not depend on
the gauge of A; forif A is replaced by A + V#f
, then X' is unchanged since
{B
V#dV-
n B#fdS= 0.
(2.2)
If we adopt the Coulomb gauge for A (i.e. V A = 0) and impose the further
condition A = O(1XL-3) as lxi-X 00, then A(x) is given by the Biot-Savart law:
A(x)
dVJB(x)x (x-x*)dV*
(2.3)
so that, from(2.1),
x
1= JJ[B(x)
(x x ) ddv*
-II ~ B(x*)]
~ ~~~ddV.(2.4)
Consider now the special situation in which B is zero except in two fluxfilaments
centredon two unknottedorientedclosed curves C1, C2 which may be linked (figure
1). We may suppose that the cross sections of the filamentsare small, and that they
carryfluxes601, 602* We suppose furtherthat, withineach filament,the B-lines are
themselvesunlinked curves which close on themselvesafterjust one passage round
the filament,running 'parallel' to C,, C2 respectively. In these circumstances,
X# may be evaluated directly from (2.1): integratingfirstover the cross section,
= 6(if
A dxl?+602f A dx2.
C1
Now
(2.5)
C2
A dx1 = fBndS,
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
(2.6)
415
(b)
-2;
(e)
n= 0./
--
Figure 2
Figure 3
Figure 2. Two linked orientedflux tubes, each one of which is made up of a large numberof
filamentsof small crosssection.Each pair of filaments(one fromeach tube) makes a contribution
2n601 6q2 to the total helicity,and the total helicityis 2n(Pl )2
Figure 3. Knotted fluxtube whose tube axis is a trefoilknot.
B ndS = n602,
(2.7)
where n is the (Gauss) linkingnumber of {1, C2}, i.e. the algebraic number of times
that C2 crosses D1 (allowing fordirectionof crossing).Three examples are shown in
figure1. Similarly,
A dx2=n6'01,
(2.8)
C2
X? = 2n601 602l
(2.9)
Thus V is determinedsolely by the two fluxes and the linkingnumber of the two
filaments.
In this derivation, it is essential that each flux tube should by itselfhave zero
helicityand this is ensured by the above assumption that the B-lines within either
tube on its own are unlinked closed curves. In these circumstancesalso, the value of
n is given from(2.4) by integratingover the two cross sections: allowing forthe fact
with
that we may have x E Cl, x*e C2 orx E C2,x* e Cl, we findagain V =2n 6q1
062
n
I=4fr
(dx
x dx*)
xx*)
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
(2.10)
416
H. K. Moffatt
and R. L. Ricca
The restrictionto small cross sections of the two filamentsis easily lifted.In the
situation indicated by figure2 in which B is confinedto two fluxtubes, in each one
of which the B-lines are again unlinked closed curves passing once (the long way)
round the torus, we may regard each flux tube as made up of a large number of
filamentsof small cross section. Each pair of filaments(one fromeach tube) make a
contribution2n b1 6012 to the total helicity,so that summingover all such pairs, this
is now given by
2nPP2
(2.11)
(X2 +
y2)2.
(3.1)
where V = 2i2e2R3is the volume of To,and 0 (as may be easily verified)is the fluxof
Bo throughany section of the tube. The fieldlines ofBo are thus unlinkedcirclesnear
r = R. The helicityof the fieldis clearly zero.
We may now inject helicity(figure5) by 'Dehn surgery',i.e. by cuttingthe tube
at a section 0 = const., twisting the free ends through a relative angle 2rrh,and
reconneeting. We may suppose that the resulting twist is uniformlydistributed
round the tube. If h is an integerno say, then each B-line in the new flux tube is a
elosed curve in the formof a helix with axis the circler = R, and each pair of B-lines
has linkingnumber no. The helicitythus generated is given by
o{
2no0 do =no
02,
(3.2)
since we may build up the tube by incrementsdo, the incrementin helicityat each
stage being 2no0qdo, from(2.1:1).
We now propose to distortCo to the curve Ko definedabove, carryingthe tube To
Proc. R. Soc. Lond. A (1992)
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
417
const.
K
C
*/)
Figure 4
Figure 5
Figure 6. Each negative (positive)switchis equivalent to the insertionof a small loop of flux0
annihilatingfluxon one side of the crossingand creatingfluxon the otherside. In the figurethe
withit. To do this we must specifythe isotopy that acts upon the fieldBo. We picture
the fieldas embedded in an incompressiblefluidmedium which moves with velocity
u(x, t) (whereV u = 0) carryingB(x, t) withit accordingto the 'frozenfield' equation
dB/at= V x (u x B).
(3.3)
It is well known that the fluxof B throughany material surfaceis conserved under
this evolution, and that the helicity of the field is also conserved (Woltjer 1958;
Moffatt 1969). We choose a velocity field u(x, t), te (-to, to), that brings CO into
coincidencewith Ko and that carries To into a tubular neighbourhoodTo of Ko. The
flux in this tu-beis then still ( and its helicityis still nO 02.
We now convert Ko to K by appropriate crossing switches. Suppose that N+
positive switches (i.e. switches which create positive crossings) and N_ negative
switchesare needed to effectthis transformation.Each positive switchis equivalent
to the insertionof a small loop of flux P annihilatingfluxon one side of the crossing
and creating flux on the other side (figure6). The net incrementof helicityis 202.
Similarly a negative switch gives a net increment of helicity -20 2. Hence the
helicityof the fieldin the new tubular neighbourhoodT around K is
X
=No2,
N= no+2(N+-N_).
(3.4)
By this construction,the B-lines within T are still clearly closed curves, all
satellites of K, and each pair of B-lines having the same linkingnumber n, since the
crossingswitches treat all pairs in the same way. Following the argument of Ricca
& Moffatt(1992), n is in fact equal to N; for suppose we divide the flux tube into
m (> 1) sub-tubes,each sub-tube carryingfluxOm = 0/m (figure7). The helicityof a
Proc. R. Soc. Lond. A (1992)
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
418
H. K. Moffatt
and R. L. Ricca
=
X
mY+rm(m-l)
whence
(3)
(3.6)
n0P2,
so that n = N as asserted.
It is obvious now that the linkingnumber n may be given any desired value by
suitable (retrospective)choice of no:
nO =
(3.7)
n-2(N+-P_).
db/ds = -Tin,
(4.1)
where c(s) is the curvatureand T(s) the torsionof C at position s. (Note that n, b and
T would not be definedat an inflexionpoint where c = 0; problems associated with
deformationthroughinflexionalconfigurationwill be treated in ?6 below.)
We now seek to obtain an alternative expression forthe helicityin the flux tube
T constructedaround K, directlyfromthe formula (2.1) by consideringits limiting
behaviour as the crosssection of T tends to zero. The fieldB may be decomposed into
the sum of two parts:
B-Ba + Bm,
(4.2)
where Ba is the axial fieldparallel to the tube axis and Bm is the meridionalfieldin
meridian planes perpendicularto the tube axis. When the cross section of the tube
is small, we may adopt a local cylindricalcoordinatesystem(r,0, z) and suppose that
Ba = (0O,,B(r)),
Bm = (O,B0(r),0).
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
(4.3)
419
BmdV0=.
fTAm
Aa
BaVtAa
Am
Bmd+
ad
fT
fT
Aa-BadV+2 AaBmdV
(4.6)
(using integrationby parts and the divergence theorem) the integrals in each case
being over the tube T.
A
=
Aa Ba dV. Here we may use the
Consider firstthe axial contribution
Biot-Savart expression (2.3) in the limitingform
A(x) = i fdX*(x
*)
(4.7)
2((
jj
(4.8)
(4.9)
where the angular brackets denote averaging over all directionsv of projection. This
fact is evident fromconsiderationof the diagram of figure8. The elements dx, dx*
where
will intersectin projection if and only if n is parallel to + (r + Adx -a
,dx*)
r = x-x*, 0 < A < 1 and 0 < I < 1, i.e. only if v lies within a solid angle
dzu = 2(dxxdx*)
r/4icr3
(4.10)
(the factor2 allowing forthe + possibilitiesabove). Thus when we average over all
directionsof v, take account of crossing signs and then integrate over all pairs of
elements dx, dx*, we obtain
<
=Affi+
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
(4.11)
420
H. K. Moffatt
and R. L. Ricca
r-x
<-d~~~x
Figure 8. Contribution to the solid angle by elements dx, dx* on the curve C. The v-direction
indicates a line of apparent intersection of dx, dx*.
(b)
(a)
(C)
= + 1;
x dx*)
3 ) -rI
=n?(v)+nj(v)>.
<
(4.12)
When the knot K is flattenedonto the (x,y) plane except forindentationsat the
crossings,the writhemay be easily computed fromthe formula (4.1 1), since in this
case n+(v) and n_(v) are independent of the viewing direction v except for a
vanishinglysmall solid angle of directionsnearlyparallel to the (x,y) plane. Thus the
writhein this limitingsituation is simply given by
W' - (n+ - n_),
(4.13)
and is an integer.Values of the writhefor a number of flattenedconfigurationsare
shown in figure9.
We note also that the fieldAa provides a familyof Seifertsurfacesforthe knot, as
noted by Akhmet'ev & Ruzmaikin (1992). A Seifertsurfaceis a non-self-intersecting
orientedopen surfacebounded by the knot K. Let the cross section of the knot tube
T tend to zero. Outside T, V x Aa = 0 so thereexists a scalar field(not single-valued)
such that
Aa = VVa
(4.14)
near
0
where
is
the
azimuth
used
above.
Thus
the
and,
K, V'a (2'r)-100
angle
Proc. R. Soc. Lond. A (1992)
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
Helicityand theCadlugadreanu
tnvartant
421
surfaces W,a= const. are all bounded by the knot, and since VW,ais single valued the
knot does not intersectsuch a surface at any other point. Any surface Tfa= const.
is thereforea Seifertsurface.
Ao(r)Bo(r)dV,
(5.1)
arising fromthe meridional component of the fieldBm. Note that fromthe firstof
(4.3) and the firstof (4.4), Aa = (O,A(r),O) where
Id
(rA0) = Bz(r).
r dr
(5.2)
53
=re^r =r(neosO+bsinO)}
eo =-n sin0+ b cos0,
so that, assuming 64 to be the same for all (r,0),
64 = rcos06n+rsin06b,
and so
64
(5.4)
= (r cos0) d 6n+(rsin0)id
(5.5)
b.
Since it is only the variation of 64 with arc lengths that contributesto distortionof
the field,we may suppose that at s = sl, 64(sj) = 0, i.e. 6n(sl) = 6b(sj) = 0. Then,
fromthe Frenet relations (4.1), we have
d
-6fn
dsI
dd
-6b
ds/
= -6ct+6Tb
I
I
= -6Tn
at
s = s8.
(5.6)
Now under the assumed virtual displacement64(s), the axial fieldBa (and so A0(r))
is unchanged, but the meridional fieldBm at s = sj is changed by an amount
6Bm = (Ba V)6> = Bz(r)+d
&
(5.7)
due to the variation of 6& with arc length (this is the process that in
magnetohydrodynamicsis known as 'generation of toroidal field by differential
rotation' (see Moffatt1978). Hence, at s = s,
6BI3= Bz(r) d 6)
= Bz(r)L
srnc.R.Sds
(ds 6)
J+ cosb
(dsds6]
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
(5.8)
H. K. Moffattand R. L. Ricca
422
6m=
Af(r) 6B1(r)dV
(5.10)
(rAo) r 2trdr
then from(5.10)
'[(rA0)2]0 =2xt1
(p-)'
(5.12)
m=
where
g=2=
(5.11)
T(s)ds
(5.13)
27
(5.14)
J.,
By (1.5), the total twist number for a ribbon is thus
=
-c cos
(T+ d/ds)
Ot+
(N
xN) tds CJc
CJc2
ds =
ds /27c S+
[O]C
(5.15)
dg/dt,
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
(5.16)
m=
423
(5.17)
0)
dt/ds = d2X/ds2
(6.1)
(S-SC)2t?,+
te+
(6.2)
Moreover,sinceItI= 1,
(t" t)
dst22
=0,
(6.3)
where oc=
(6.4)
and
ItI=
(6.8)
1+0(s4),
so that, near s = 0, t is indeed the unit tangent vector. Figure 11 shows this family
of curves and their projections on the three coordinate planes.
From (6.7), to leading order in Itland Isi,
at/as
2(0, t,3s),
so that
tat
n(s, t) = -as
and
(0,t,3s)
((2?92)
(6.9)
(6.10)
(6.11)
Note here that for very small t, n rotates through an angle Zt about the direction
te = (1,0,0) as s increases from -so to +so where so > Itl;and that this rotation
is clockwise (right-handed)fort < 0, and anticlockwise (left-handed)fort > 0; thus
the number of rotations of the pair (n,b) about the tangent direction t in
the anticlockwise sense increases by + 1 as t increases throughzero (at the instant
t= 0, this number is undefined).
Now the binormal is given by b = t x n, and the torsionis obtained from(4.1): for
Itiand IsIsmall,
(6.12)
T(5,t)
3t/(t2 9s2).
+
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
424
H. K. Moffattand R. L. Ricca
/ t=O
tzO
=~zol
/\t<O
Figure 11l. The twistedcubic (6.6) for -1 ? s ? 1 and forvarious values of t. The curve
A point at (i) s
containsan inflexion
when t = 0. Plane projectionsare shownbelow.
0(s,t)ds=
IT
tO20
ds=2arctan(3t?)'
(613)
+ g- +
!J0).
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
(7.1)
425
(a)
(b)
<
Figure 12. (a) Writhe,(b) torsionand (c) twist contributionsof a ribbon to the CaIugareanu
invariant.If a coiled ribbonis stretchedso that its centre-linebecomesstraight,then the initial
torsionof the centre-lineis convertedto the finaltwistof the ribbonabout its centre-line.
(t(s)
XL
C'
Figure 13. Mapping of closed curvesC to theirimages C' on the unit sphere.
h-=
const.
(7.2)
/+!+o=
The writhe YP definedby (4.8) varies continuouslyas C is deformedcontinuously;
however, if C passes through an inflexionalconfiguration,then ! jumps by + 1.
Hence, as ! jumps by ? 1, the term !F must jump by a compensatingamount + 1,
to maintain the overridinginvariance of helicity.
The equal and opposite jumps in (I ? Y) and E0 may be understood with
referenceto the simple example illustratedin figure12. If a coiled ribbonis stretched
so that its centre-linebecomes straight(at which stage dt/ds 0
O on the centre-line!),
then the initial writheof the centre-lineis convertedto the finaltwist of the ribbon
about its centre-line.This example is not genericsince it involves the appearance of
a continuumofinflexionpoints. However, it captures the essence of the nature ofthe
interchangebetween (YV+ ! ) and 50: X0 representsthe intrinsictwist of the ribbon
about its centre-line,and this in general jumps by T 1 when the centre-lineis
deformedthroughan isolated inflexionpoint.
Proc. R. Soc. Lond. A (1992)
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
H. K. Moffattand R. L. Ricca
426
CC
rc
X(s) = [(R +
rO cos mq)
cos 0, (R +
rO cos mS)
sin 0, rosinmv],
(7.3)
(where m is an integer) is particularly interestingin this respect (figure 14). If
ro = 0, this is a circle, the principal normal n = nO pointing towards its centre. For
very small values of ro,the directionof n simply oscillates about the position no as
we move round C. As roincreasesfurtherthe amplitude ofthese oscillationsincreases,
until at a criticalvalue r,(= R/m2)inflexionpoints appear on C at the points where
cos mq =-1; forrO > r, the principal normal n makes m complete rotations around
the axis of the torus in one passage round C.
If we now place a fluxtube of cross-sectionalradius e < r, around C, and consider
a time-dependentdeformationin which ro= ro(t)decreases throughthe criticalvalue
rc, then [1] = -m, [!1] = +m in going through the critical point, i.e. torsional
helicity is instantaneously converted to twist helicity, the total helicity being of
course conserved.
The fact that 70 jumps by +1 whenever C passes through a single inflexion
suggeststhe interpretationthat it represents,in some sense, the numberof rotations
of the flux tube (or of the associated set of ribbons) about its axis in one passage
around C. This concept is, however,quite elusive, because one must specifycarefully
the frame of referencewith respect to which the flux tube rotates. There is no
difficultyin this when C is not in an inflexionalconfiguration,forthen we may use
the Frenet frame(t,n, b). Let C, C* be two neighbouringB-lines in the fluxtube (the
boundaries of a ribbon), and, as before,let N(s) be the spanwise vector fromC to C*
on this ribbon. Let X4/'be the (integer) number of rotations of N(s) about t with
respect to the Frenet frame (as definedin ?5). We shall show that in fact 90 = X.
Under arbitrarycontinuous deformationsof the fluxtube, N(s, t) is a continuous
vector functionof (s, t), and its componentswith respect to a fixed cartesian frame
ofreferenceare also continuous.However, ifC passes throughan inflexionat sc, then,
as we have seen in ? 6, the number of rotations of the Frenet pair (n,b) about the
tangent vector t in one passage round C changes by ? 1; hence the number X of
rotations of N(s, t) relative to the Frenet frame changes by + 1.
Proc. R. Soc. Lond. A (1992)
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
Helicityand theCaOlngadreann
tnvartant
427
Figure 15. A plane projectionof a knot may always be arrangedso to have no inflexionpoints.
withno inflexion
Here the figure-of-eight
knot,withtwo inflexions,is isotopedto a configuration
points.
The knot K and its associated fluxtube may always be deformedto lie nearly in
the (x, y) plane (as in ?3) so that the write #Vis (in the limit) an integer (n+-n_).
We may also always arrange that this plane projection ofK has no inflexionpoints;
forexample the figure-of-eight,
containingtwo prototypeinflexions,can be isotoped
(figure15) to a curve ofnon-vanishingcurvature.In such a 'standard' configuration,
the torsionT(s) is zero everywhereexcept in the small indentationswhere it is small;
hence in the plane limit, ! = 0. In this configurationthereforethe helicityis given
by
(7.4)
X = n2 = (n+-n)) I2+ g p2
n-(n+- n_)
1!0F n
so that
(7 5)
8. Summary
In this paper we have discussed several properties of the helicity of linked and
knotted fluxtubes. Differentcontributionsto helicityhave been analysed in terms
of the Gauss linkingnumber (?2) and in termsof the Calugareanu invariant (1.3). If
the fieldlines in a single knotted fluxtube are twisted closed curves which close on
themselves after one passage around the tube, then the helicityof the flux tube is
given by
Xy=
ABdV=n2,
(8.1)
where Pis the fluxassociated with the tube (? 3). The integern is an invariant under
frozenfielddistortionof the tube, and is identifiedwith the Calugareanu invariant
(1.3). We have demonstratedthis invariance by startingfromthe known invariance
ofhelicity.The helicityhas been decomposed into writheand twistcontributions,the
writhe contributioninvolving the Gauss integral, which admits interpretationin
terms of the sum of signed crossingsof the knot averaged over all projections (?4).
Part of the twist contributionis shown to be associated with the torsionof the knot
and part with what may be described as 'intrinsictwist' of the fieldlines in the flux
tube around the knot (? 5). The genericbehaviour associated with the deformationof
the knot through a configurationwith points of inflexion (points at which the
curvature vanishes) has been analysed (?6) and the role of the twist parameter has
been discussed (? 7).
Proc. R. Soc. Lond. A (1992)
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
428
H. K. Moffattand R. L. Ricca
In derivingthe relation
X, =
n02
<
,Al') 02,
(8.2)
where the writhe Y/ is definedby equation (4.8), the normalized total torsion Y is
definedby equation (5.13) and l = X0 is the twist parameter,we have shown that
(generically) (/f ?+ ) jumps discontinuouslythrough + I as C passes through an
there is then a
inflexionalconfigurationand that by virtue of the invariance of XV9,
compensatingjump of T I in X. This behaviour has been previouslyrecognizedby
Ricca & Moffatt(1992) and is associated with the classical type I Reidemeistermove
of ambient isotopies.
The CAlughreanuinvariant is fundamental in relation to many problems that
involve continuous deformationof tube-like structures.Examples in the literature
range from the theory of dynamical systems (Uezu 1990) to the biochemistryof
excitable media (Winfree1990), fromthe quantum fieldtheoryof string-likeobjects
(Tze & Nam 1989) to studies of DNA coiling (Tsuru & Wadati 1986), fromthe theory
of propagation of spinningparticles (Jaroszewicz & Kurzepa 1991) to the general
problem of protein folding(De Santis et al. 1986). Fundamental topological aspects
of these phenomena can be successfully described in terms of the Chlugareanu
invariant. In this paper we hope that the direct derivation of the CMlugareanu
invariant fromfirstprinciplesof fluidmechanics togetherwith the discussion of the
genericbehaviour associated with inflexionalconfigurations,which are ubiquitous in
many processes of continuous deformation of field structures,provides a good
demonstration of the relevance of fluid dynamical techniques to topological
problems.
We thank Dr J. S. Langer, Directorof the InstituteforTheoreticalPhysics at UCSB (whichis
supportedby NSF GrantPHY89-04035), forhis kindhospitalityat the Institute,wherethiswork
was initiated.One of us (R. L. R.) acknowledgesthe financialsupportfromAssociazioneSviluppo
Scientificoe Tecnologicodel Piemonte (Turin,Italy).
References
In Topologicalaspectsof the
Akhmet'ev,P. & Ruzmaikin,A. 1992 Borromeanismand bordismn.
dynamicsoffluidsand plasmas (ed. H.K. Moffattetat.) pp. 249-264. Dordrecht:Kluwer.
numberof
Banchoff,T. F. & White,J. H. 1975 The behaviourofthe total twistand self-linking
a closed space curve underinversions.Math. Scand. 36, 254-262.
Berger,M. A. & Field, G. B. 1984 The topologicalpropertiesof magnetichelicity.J. Fluid JMech.
147, 133-148.
Calugareanu,G. 1959 L'int6gralde Gauss et l'analyse des noeuds tridimensionnels.
Rev. lvath.
pures appl. 4, 5-20. (C59 in text.)
et leurs invariants.
Calugareanu,G. 1961 Sur les classes d'isotopie des noeuds tridimensionnels
Math. J. 11, 588-625. (C61 in text.)
Czechoslovak
De Santis, P., Palleschi,A. & Chiavarini,S. 1986 Topological approach to proteinfolding.Gazz.
Chim.It. 116, 561-567.
Freedman, M. H. & He, Z.-X. 1991 Divergence-freefields: Energy and asymptoticcrossing
number.Ann. Math. 134, 189-229.
Fuller, F. B. 1971 The writhingnumber of a space curve. Proc. natn. Acad. Sci. U.S.A. 68,
815-819.
Fuller,F. B. 1972 How the writhingnumberof a curvedependson the curve.Rev. Roum.Math.
pures appl. 17, 1329-1334.
Fuller, F. B. 1978 Decompositionof the linkingof a closed ribbon: a problemfrommolecular
biology.Proc. natn.Acad. Sci. U.S.A. 75, 3557-3561.
Proc. R. Soc. Lond. A (1992)
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions
Helicityand theCalugyareanuinvariant
429
This content downloaded from 193.147.59.3 on Tue, 04 Aug 2015 16:58:38 UTC
All use subject to JSTOR Terms and Conditions