A NOTE ON SINGULARITIES OF THE D EULER EQUA TION S T an v eer Mathematics Departmen t Ohio State Univ ersit y Colum bus OH ABSTRA CT In this pap er w e consider analytic initial conditions with nite energy whose complex spatial con tin uation is a sup erp osition of a smo oth bac kground o w and a singular eld Through explicit calculation in the complex plane w e sho w that under some assumptions the solution to the D Euler equation ceases to b e analytic in the real domain in nite time This researc h w as partly supp orted b y the National Aeronautics and Space Administration under NASA Con tract No NAS while the author w as in residence at the Institute for Computer Applications in Science and Engineering ICASE Additional supp ort w as pro vided b y the Departmen t of Energy DE F G ER i INTR ODUCTION One fundamen tal question in uid mec hanics is whether there exists a v elo cit y eld u x t and a corresp onding v orticit y eld x t satisfying the incompressible D Euler equations u r r u t r u r u so that for some time t these elds are singular when u x and x are initially smo oth and the energy is nite i e Z d x j u x j The answ er to this and related question of enstroph y blo w up in the limit of in nite Reynolds n um b er ha v e imp ortan t rami cations in turbulence There ha v e b een man y in v estigations in the literature to address this question Exact so lutions are kno wn for whic h a smo oth o w b ecomes singular in nite time the v elo cit y eld in all suc h cases do not v anish at in nit y and therefore the energy constrain t is violated This constrain t is imp ortan t since Childress et al sho w ed that without the energy constrain t there are explicit singular solutions of the D Euler equations whic h is imp ossible for o ws with nite energy There ha v e b een also b een a n um b er of mathematically rigorous w ork on this issue While there are no de nitiv e theorems one w a y or the other a n um b er of necessary conditions for nite time singularities ha v e b een pro v ed There ha v e also b een quite a n um b er of n umerical in v estigations as w ell Nonetheless a de nitiv e conclusion based on these is di cult to reac h Motiv ated b y ph ysical insigh t in the analogous mag netoh ydro dynamics equations argumen ts based on an assumed m ulti scale expansion w ere presen ted for the o ccurrence of nite time singularit y for o ws with b ounded v elo cit y at in nit y rather than b ounded energy Recen tly there ha v e b een in v estigations in v olving singularities in the complex spatial domain Suc h studies b ecome relev an t to the issue at hand if complex singularities are found to impinge the real ph ysical domain later in time Ca isc h simpli es the D Euler equation through a so called Mo ore appro ximation Within this appro ximation a nite time singularit y is predicted Ho w ev er the relev ance of this calculation to the actual D Euler equation is unclear since the initial data in the real domain is complex and the initial v alue problem is lo cally ill p osed unlik e the actual D Euler equations T an v eer Sp eziale recen tly lo ok ed at a sp ecial class of initial condition that are analytic and real in the real x x x domain x x x x with a complex spatial plane con tin uation i e complex x x x that includes singularities on a set of p oin ts x that are zeros of a a scalar function of x A t suc h p oin ts the analytically extended v elo cit y is nite but v orticit y in nite The initial condition is decomp osed in to a smo oth spatial o w and a singular part Under certain assumptions on existence of solutions in this extended complex domain it w as sho wn that this decomp osition remains in v arian t with time though initial singularities b ecome adv ected with the smo oth part of the v elo cit y eld call it the bac kground o w T o the exten t that these assumptions hold for an y class of initial conditions the results p osed no additional restriction on the c hoice of the bac kground o w other than that it is spatially analytic in a neigh b orho o d of complex p oin ts where the remaining part of the v elo cit y and v orticit y elds are singular Indeed for steady bac kground o ws it w as seen that the dynamics of the complex singularities resulted in su cien t conditions for instabilit y through analysis of three ordinary di eren tial equations In terestingly enough these di eren tial equations are iden tical to those found b y Lifsc hitz Hameiri and Vishik F riedlander who in v estigated instabilit y of D Euler o ws using a WKB analysis for short w a v elength disturbances The latter results ha v e also b een established rigorously Recen tly aside from presen ting argumen ts for nite time singularit y for a class of o w with b ounded v elo cit y but not energy Bhattac harjee et al applied the complex singularit y approac h of T an v eer Sp eziale to a bac kground o w that is spatially smo oth in the complex plane but temp orally singular They presen ted argumen ts suggesting that if the the assumptions made b y T an v eer Sp eziale are v alid complex singularities can impinge the real x domain in nite time resulting in an additional collapsing length scale b esides that of the bac kground o w Ho w ev er no information on the dynamic ev olution of the singular part of v elo cit y and v orticit y is obtained F urther it is not clear if an initially analytic div ergence free v elo cit y eld in the real domain satisfying the energy constrain t can b e decomp osed in the complex x domain in to a spatially smo oth bac kground o w that has explicit solution and a singular part In this pap er largely motiv ated b y the recen t w ork of Bhattac harjee et al w e presen t analytic initial conditions u x in the un b ounded real domain satisfying the energy con strain t for whic h the analyticit y strip width of solution to is found to shrink to zero in a nite time under some clearly stated assumptions Since the assumptions ab o v e ha v e not b een v eri ed the results do not pro vide a clear answ er to the question of nite time singularit y for nite energy smo oth initial conditions nonetheless w e susp ect that the results will b e of v alue in settling this issue since the assumptions can b e c hec k ed n umerically through appropriately designed calculations INITIAL CONDITIONS Consider initial v elo cit y eld giv en b y u x u x f d x q x s where u x x x x x x s f d d d x g r where x x x x r x x x and a s g s s a s s for some small real and p ositiv e constan t a and q x h r r u x h r x W x s where g r h r r g r W x x x x x x x The initial condition corresp onds to an axi symmetri c o w with swirl Nonetheless w e do not mak e m uc h use of this symmetry since w e closely follo w the form ulation of T an v eer Sp eziale that w as dev elop ed for the general D Euler equation No w w e list b elo w some claims that are sho wn to b e true Claim i The function q x is real and analytic ev erywhere in the real x x x domain It is also analytic in the nite complex x plane except at p oin ts where g r These p oin ts are far a w a y from the real domain for small a First notice from that h i a g r a r a r a r r a r r Notice that the ab o v e is nonzero for real x for whic h r When a is small a zero in the complex x x x plane o ccurs only at p oin ts far a w a y from the real domain i e for jI m x j large atleast for some j F urther while there can b e p oles of g r h r de ned in j is clearly analytic at suc h p oin ts Th us the function h r is analytic except at p oin ts where g r The same is clearly true for h r This establishes claim i since u x and W x are ob viously analytic functions of x i e of complex v ariables x x and s x Claim ii The initial condition u x is analytic ev erywhere in the real x x x domain with singularities in the complex plane determined b y the relation g r g r and g r W e note that d x g r is analytic and nonzero ev erywhere except at p oin ts in the complex plane when the denominator in v anishes F or small a zero of the denominator in o ccurs near r while the other zeros o ccur far a w a y from the real domain d x is also zero in the complex x x x plane at p oin ts where r Raising d to a fractional p o w er w e obtain singularities of f d x at r where g r and at other complex p oin ts where g r F rom the decomp osition in and claim i ab o v e claim ii follo ws Claim iii The energy constrain t is satis ed T o sho w that constrain t is satis ed it su ces to sho w that for real x with r u x O r F rom the expression of g r in and g r in it is not di cult to sho w from that for large r h r O r h r O r F urther from W O r and therefore from it follo ws that u x O r So claim iii is established Claim iv The initial condition u x giv en b y satis es the incompressibilit y condition Indeed eac h of u x and f d q x is div ergence free s The div ergence condition can b e sho wn to b e v alid directly b y calculating the div ergence of the righ t hand side of using the expressions for f d x and q x as sho wn ab o v e D YNAMICS Giv en that the initial data is analytic in the real domain it w ould app ear reasonable to assume the follo wing whic h is listed as Assumption i The initial v alue problem with analytic initial condition has an analytic solution in the real domain for t in some in terv al t for some t Note If an analytic solution to do es not exist b ey ond t then w e will ha v e sho wn that the analyticit y width shrinks to zero in nite time In that case there is nothing more to demonstrate So w e will assume otherwise and let t W e no w de ne u x t hence de ned to b e the bac kground o w to b e s u x t x x x x x s t corresp onding to whic h the v orticit y r u is s s x t s t Notice that and satisfy the Euler equations and are singular at t though there are no spatial singularities ho w ev er since it is un b ounded at the energy constrain t is not satis ed b y u s W e de ne d x t to b e the solution to d u r d t s satisfying initial condition F rom the standard metho d of c haracterestics d x t t d x g x x x along tra jectories determined from d x u x t s dt Solving x t x t x t x t where x t x t cos l n t x t sin l n t x t x t cos l n t x t sin l n t x x t t F rom it follo ws that d x t g x x t x t W e no w decomp ose v elo cit y u x t and the v orticit y x t r u x t in to u x t u x t f d x t q x t s x t x t f d x t p x t s where initial condition is satis ed Then in order for the v orticit y x t and the v elo cit y u x t to satisfy it is clear on substitution that p x t and q x t m ust satisfy f f f p u r p p r q f q r p q r q r d t s s s f f f p r u r q q f rd r q f p r q s s f f r q q rd f f r q rd q p f The existence of analytic p x t and q x t in the real domain satisfying for t in the in terv al follo ws from assumption i and the decomp osition and W e no w in tro duce another assumption Assumption ii p x t and q x t are analytically con tin uable upto the set of p oin ts x t where d x t t F urther eac h of r q and r p is b ounded at d f f f R emark Since eac h of f and are con tin uous functions of d at d equations f f can b e seen to ha v e con tin uous co e cien ts at d So p x t is exp ected to b e con tin uously di eren tiable at the set of p oin ts x t atleast for early time Moreo v er if p x t t is kno wn apriori to b e just a lo cally analytic function of t then the results b elo w will hold for p x t t for all t in the in terv al regardless of the assumed b oundedness in assumption ii W e no w de ne A t rd x t t B t q x t t C t p x t t W e denote x Since d x g it follo ws that F rom and A g F urther from and it follo ws that B h It is clear that since d x t t eac h of f f f are zero at d then with assumption ii and imply A B A B C It follo ws that C A A A B Also from and it follo ws that C g h F urthermore with assumption ii it follo ws from that dC A B TC s dt where T is a second rank tensor whose elemen ts are de ned b y u s j T j k x k In our case the only nonzero elemen ts of T are T T T T T t t t F rom w e get dC SC dt where S is a second rank tensor whose elemen ts are giv en b y A s S T A j k j k j k l l A A giv en that is the usual Levi Civita tensor and the Einstein summation con v en tion on j k l rep eated indices has b een used F or t it follo ws from and that h i g A t x t x t t x t t where x t x t x t x t is determined from with x Using and it follo ws h i g A A t t Using and in the expression for S in w e nd that the comp onen ts of equation satisfy the follo wing system of ordinary di eren tial equations h i dC T C T C t x t C x t C dt t h i dC T C T C x t C t x t C dt t dC T C x t C x t C dt t Exact solutions to this is p ossible when in whic h case eac h of the comp onen ts C C a is iden tically zero initally and remains so for t Using g h a it follo ws from and that the third comp onen t of C is giv en b y a C t a t Therefore for a C t a t It follo ws from and that in suc h cases B t t It is clear from and and the de nitions that eac h of p x t t and q x t t is singular as t at the same time as complex singular p oin ts x t x t x t determined from and approac hes as t Since the total v orticit y is giv en b y it follo ws that su cien tly close to these complex singular p oin ts x t h i x t A t x x t C t t where A t is determined from with x t in this case and C t is determined from It is clear from that the singularities of and f d p do not cancel out as s t at x t In the general case when do not app ear to b e exactly solv able Nonetheless asymptotic considerations suggest that in suc h cases all comp onen ts of C t generally b ecome singular as t IMPLICA TIONS OF THE CALCULA TIONS The results in the complex plane together with the result that certain class of initially complex singularities x t impinge in nite time w ould app ear to suggest that the initial v alue problem with initial condition b ecomes singular in the real domain in nite time This w ould b e true if one can assume the follo wing Assumption iii If analytic solution to with initial condition exist for t in then as t one or more singularities x t whose nonzero comp onen ts are computed through and is lo cated on the same Riemann sheet as the real ph ysical domain Note If assumption iii is correct then the p oin t at whic h x t impinges will actually b e part of the ph ysical real domain and not of another Riemann sheet In the ev en t that assumption iii is v alid then the calculations in section sho w that that there m ust b e a collapsing ph ysical length scale near the origin prop ortional to t corresp onding to the complex singularit y distance from the real domain This is in addition to the length scale asso ciated with u x t in that ob viously scales as t This w as s s suggested b y Bhattac harjee et al Another p oin t to note ab out the sup erp osition is that for a xed real x di eren t from as t need not b e singular temp oral singularities in can cancel out that s of f d p as w e susp ect it will though w e are unable to demonstrate this con vincingly The argumen ts ab o v e are not applicable if the solution to cease to b e analytic in the real domain for t as can happ en if complex singularities of p created in the dynamical pro cess impinges the real domain for t Note that for small a there exists complex p oles of g r and hence of d x near r from examination of it is clear that singularities of p are exp ected to b e created at suc h p oin ts at t Whatev er the case our calculations ab o v e sho w that if assumptions i iii are v alid then the solution to ceases to b e analytic in the real domain b ey ond t It w ould b e in teresting to c hec k these n umerically through sc hemes that monitor the analyticit y width as a function of t Determination of the ph ysical nature of suc h singularities if they o ccur will also b e in teresting CONCLUSION W e ha v e considered an initial condition for the three dimensional Euler equation that is analytic in the real x domain and con tains nite energy By making three clearly stated assumptions w e nd through explicit calculations that the solution m ust lose analyticit y in nite time F urther study b oth analytical and n umerical is lik ely to shed some ligh t on the v alidit y of the assumptions If these assumptions are true n umerical calculations is also lik ely to suggest if the solution loses analyticit y b efore t or if it do es so at t in the manner sho wn explicitly in this pap er A CKNO WLEDGEMENTS W e are indebted to Amita v a Bhattac harjee for sharing a preprin t that motiv ated this pap er to a large exten t References M Lesieur T urbulenc e in Fluids Martin us Nijho J T Stuart Nonlinear Euler equations and singularities in their solution in Symp osium to Honor C C Lin editted b y D J Benney F H Sh u and C Y uan W orld Scien ti c S Childress G R Ierly E A Spiegel and W R Y oung Blo w up of unsteady t w o dimen sional Euler and Na vier Stok es solutions ha ving stagnation p oin t form J Fluid Me ch W W olibner Un theoreme sur l existence du mouv em en t plan du n uide parfait homoge ne incompressible p enden t un temps in nimen t long Math Z V I Y udo vic U S S R Comput Math Ph ys T Kato On the classical solution of the t w o dimensional non stationary Euler equa tion Arc h Rat Mec h Anal Andrew J Ma jda V orticit y T urbulence Accoustics in Fluid Flo w Siam R eview V ol P Constan tin C F e erman Direction of v orticit y and the problem of global regularit y for the Na vier Stok es equation Indian Univ Math J R Grauer T Sideris Numerical computation of three dimensional incompressible ideal uids with swirl Phys R ev L ett A Pumir E Siggia Dev elopmen t of singular solutions to the axi symmetric Euler equations Phys Fluids A M Brac het M Meneguzzi A Vincen t H P olitano and P L Sulem Numerical evidence of smo oth self similar dynamics for three dimensional ideal o ws Phys Fl A R M Kerr Evidence for a singularit y of the three dimensional incompressible Euler equation Phys Fluids A A Bhattac harjee X W ang Finite time v ortex singularit y in a mo del of D Euler o ws Phys R ev L ett R E Ca isc h Singularit y formation for complex solutions of the D incompressible Euler equations Physic a D S T an v eer C G Sp eziale Singularities of the Euler Equation and Hydro dynamic Stabilit y Phys Fl A A Lifsc hitz E Hameiri Lo cal stabilit y conditions in uid dynamics Phys Fl A S F riedlander M Vishik Instabilit y criteria for the o w of an in viscid incompressible uid Phys R ev L ett A Bhattac harjee C S Ng Xiaogang W ang Finite time v ortex singularit y and Kol mogoro v Sp ectrum in a Three dimensional Spiral Mo del with a V elo cit y Null Preprin t