Contents lists available atScienceDirect
Physics Letters B
www.elsevier.com/locate/physletb
Integrability of the evolution equations for heavy–light baryon distribution amplitudes
V.M. Braun
a,
∗, S.E. Derkachov
b,
c, A.N. Manashov
a,
daInstitutfürTheoretischePhysik,UniversitätRegensburg,D-93040Regensburg,Germany bSt.PetersburgDepartmentofSteklovMathematicalInstitute,191023St.Petersburg,Russia cSt.PetersburgStatePolytechnicUniversity,195251St.Petersburg,Russia
dDepartmentofTheoreticalPhysics,St.PetersburgUniversity,199034St.Petersburg,Russia
a r t i c l e i n f o a b s t ra c t
Articlehistory:
Received4June2014
Receivedinrevisedform22September 2014
Accepted29September2014 Availableonline2October2014 Editor:B.Grinstein
We consider evolution equations describing the scale dependence of the wave functionof a baryon containingan infinitelyheavyquark and apairoflightquarks atsmalltransverse separations,which is the QCD analogueofthe helium atom. Theevolution equations depend onthe relative helicityof the lightquarks.Forthealignedhelicities,wefindthatthe equationis completelyintegrable,that is, it hasanontrivialintegralofmotion,and obtainexactanalyticexpressionsfortheeigenfunctionsand theanomalousdimensions.Theevolutionequationforanti-alignedhelicitiescontainsanextratermthat breaksintegrabilityandcreatesa“boundstate”withtheanomalousdimensionseparatedfromtherest ofthespectrumbyafinitegap. Thecorresponding eigenfunctionisfoundusingnumericalmethods.It describesthemomentumfractiondistributionofthelightquarksin,e.g.,Λb-baryonatlargescales.
©2014TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/3.0/).FundedbySCOAP3.
1. PrecisiontestsoftheflavorsectoroftheStandardModelmay revealnewphysicsandremainhighontheagenda.Mainattention hasbeensofarfocusedonB-mesonsbutinterestisdevelopingto theheavy baryon decaysaswell. Such baryonsare produced co- piouslyattheLHCand,asmoredataarecollected,studiesofrare b-baryon decays involving flavor-changingneutral current transi- tionshavebecomequantitativeinordertomakeanimpactonthe field.Inparticular, the
Λ
b→Λ μ
+μ
− decaysarereceivingalotof attention,seee.g.Ref.[1]andreferencestherein.Theoreticaldescriptionoftheb-hadrondecaysisbasedonfac- torizationtheoremsthatmakeuseofthelargemassoftheb-quark inordertoseparate calculableeffectsofshortdistancesfromthe nonperturbativelargedistancephysics.The correspondingformal- ismis similar butmuch lessdevelopedfor baryonsascompared tomesons.ArecentdiscussionusingSCETformalismcanbefound in Ref. [2]. For the exclusive decays involving large energy re- lease in the final state, the relevant nonperturbative quantities arebaryonwavefunctionsatsmalltransverseseparations,dubbed light-conedistributionamplitudes (DA).Theirstudywasstartedin Refs. [3–5]where thecompleteclassification andrenormalization groupequations (RGE)that governthe scale-dependencearepre- sented.
*
Correspondingauthor.In this work we point out that these equations have a hid- densymmetryandcompletelyintegrableinthecasethatthelight quarkshavethesamehelicity.Inotherwords,we identifya non- trivial quantum number that distinguishes heavy baryon states with different scale dependence and obtain exact analytic solu- tion of the evolution equations. This phenomenon is similar to integrabilityofRGEsforthelightbaryons[6,7]and,inamoregen- eralcontext,tointegrabilityinhigh-energyQCD[8–10]andinthe N=4 supersymmetric Yang–Mills theory [11–13] that attracted hugeattentionasatooltochecktheAdS/CFTcorrespondence.The integrablemodelthat weencounterinthepresentcontextisdif- ferent;ithasbeendiscussedrecentlyin[17].
The similar equation for the case that the two light quarks haveopposite helicitycontainsan extratermthat breaksintegra- bilityandcreates a“bound state”withtheanomalous dimension separatedfromthe restofthespectrum bya finitegap.Thecor- responding eigenfunction is found numerically. It describes the momentum fractiondistributionofthelightquarks in,e.g.
Λ
b,at large scales andcan be called“asymptoticDA” inanalogy to the acceptedterminologyforhadronsbuiltoflightquarks.2. Consideratfirsttheleading-twistDAofabaryoncontaining an infinitelyheavy quark anda transverselypolarized “diquark”:
a pair of light quarks with aligned helicities. It can be defined as[4]
http://dx.doi.org/10.1016/j.physletb.2014.09.062
0370-2693/©2014TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/3.0/).Fundedby SCOAP3.
(
z1n)
C/
nγ
⊥μq2(
z2n)
hv
(
0)
Bj=1(
v)
= √
13
⊥μu
(
v)
fB(2)( μ )Ψ
⊥(
z1,
z2; μ ).
(1) Hereq1,2=u,
d,
sarelightquarksseparatedbyalightlikedistance, hv(
0)
is theeffectiveheavy quark field withfour-velocity v,C is the charge conjugation matrix, u(
v)
is the Diracspinor/
v u(
v)
= u(
v)
, andμ is the diquark polarization vector, vμ
μ=0. The transverseprojections are definedwithrespect tothe two auxil- iary light-like vectorsn and n¯ whichwe choose such that vμ=
(
nμ+ ¯nμ)/
2, v·n=1,n· ¯n=2:⊥μ
=
gμν⊥ν
,
gμν⊥=
gμν−
nμn
¯
ν+
nνn¯
μ/(
n· ¯
n),
(2)andsimilar for
γ
⊥μ.The Wilson linesconnecting thequark fields arenot shownforbrevity. The heavyquark field hv canitself be relatedtotheWilsonlineas[18] 0hv(
0)
h,
v=
Pexp ig 0−∞
d
α
vμAμ( α
v)
,
(3)so that the operator in Eq.(1) can be viewed as a pair of light quarks(adiquark), attachedto the Wilsonlinewith acusp con- tainingone lightlike and one timelike segment. Finally,the cou- pling fB(2) is definedasthe matrixelement ofthe corresponding localq1q2hv operator;itisinsertedfornormalization[4].Thepa- rameter
μ
is the renormalization(factorization) scale. We tacitly implyusingM S scheme.The product
⊥μu
(
v)
on the right-hand-side (r.h.s.) of Eq. (1) can be expanded in irreducible representations corresponding to physical baryon states with JP =1/
2+ and JP =3/
2+ using suitable projection operators, see [4]. These (ground) statesform the SU(
3)
F multiplets (sextets),Σ
b, Ξ
b, Ω
b andΣ
b∗, Ξ
b∗, Ω
b∗, re- spectively, whichare degenerate inthe heavy b-quark limit. The double-strangeΩ
b baryon is ofspecial interest forflavor physics asitonlydecays throughweak interaction.The DAΨ
⊥(
z1,
z2;μ )
iswrittenusuallyintermsofitsFouriertransformΨ
⊥(
z; μ ) =
∞0
d
ω
1 ∞0
d
ω
2e−i(ω1z1+ω2z2)Ψ
⊥( ω
1, ω
2; μ )
=
∞ 0ω
dω
10
due−iω(uz2+¯uz1)
Ψ
⊥( ω ,
u; μ ),
(4)where z= {z1
,
z2} and in the second line we redefineω
1=uω
,ω
2= ¯uω
withu¯ =1−u.The variablesω
1 andω
2 correspondto theenergies(uptoafactortwo)oflightquarksinthebaryonrest frame.TheDA(4)isthemostimportantnonperturbativeinputin QCDcalculationsofexclusive heavy baryon decaystothe leading poweraccuracyintheheavyquarkmass.Thescaledependenceof
Ψ
⊥( ω ,
u, μ )
or,equivalently,Ψ
⊥(
z;μ )
, isgovernedbytherenormalizationgroupequation[3,4]μ ∂
∂ μ + β( α
s) ∂
∂ α
s+
2α
s3
π H
fB(2)
( μ )Ψ
⊥(
z; μ ) =
0.
(5)Theevolution kernel H is an integral operatorwhich can be de- composedas
H =
H12+
Hh1+
Hh2−
4.
(6) ThekernelsHhkareduetoheavy–lightquarkinteractions,Hh1f
(
z) =
10
d
α α
f(
z) − ¯ α
f( α ¯
z1,
z2)
+
ln(
iz1μ )
f(
z),
Hh2f
(
z) =
10
d
α α
f(
z) − ¯ α
f(
z1, α ¯
z2)
+
ln(
iz2μ )
f(
z).
(7)Theyare identicalto theLange–Neubert kernels[19–21].The re- mainingcontribution
H12f
(
z) =
10
d
α α
2f(
z) − ¯ α
f zα12,
z2− ¯ α
f z1,
zα21(8)
takesinto account theinteraction betweenthelight quarks; itis similartothestandardEfremov–Radyushkin–Brodsky–Lepageevo- lutionkernelforthepionDA.Hereandbelowweusethenotation
zα12
= ¯ α
z1+ α
z2, α ¯ =
1− α .
(9) Theevolutionkernels(7),(8)canbewrittenintermsofthegen- eratorsofthecollinearsubgroupofconformaltransformationsS+
=
z2∂
z+
2jz,
S0=
z∂
z+
j,
S−= − ∂
z,
(10) where j=1 istheconformal spinofthelightquark.The genera- torssatisfythestandardSL(
2)
commutationrelations[
S+,
S−] =
2S0, [
S0,
S±] = ±
S±.
(11) Onecanshowthat[23]Hh1
=
ln iμ
S(+1)− ψ (
1),
Hh2=
ln iμ
S(+2)− ψ (
1),
where S(+1), S+(2)actonthefirst,z1,andthesecond,z2,light-cone coordinate,respectively.The last kernel(8)iswritten intermsof thetwo-particleCasimiroperator S212[24]
H12
=
2ψ (
J12) − ψ (
1)
,
(12)where S212=S+S−+S0
(
S0−1)
= J12(
J12−1)
, S+=S(+1)+S(+2) etc.,andψ(
x)
is Euler’sdigammafunction.Thus,thecompleteevo- lutionkerneltakesaverycompactformH =
ln iμ
S(+1)+
ln iμ
S(+2)+
2ψ (
J12) −
4ψ (
2).
(13) The evolution equation for the DA in momentum space,Ψ
⊥(
w1,
w2;μ )
, is given by the same expression with the SL(
2)
generatorsinthemomentumspacerepresentation[22].Eigenfunc- tions of H correspond to the statesthat have autonomousscale dependenceandthe corresponding eigenvaluesdefine anomalous dimensions.3. Ourmainresultisthatthisevolutionequationcanbesolved explicitly.Tothisendweconsiderthefollowingoperators
Q
1=
iS(+1)
+
S(+2), Q
2=
S(01)S(+2)−
S(02)S(+1).
(14) It ispossibleto show that Q1 andQ2 commutewitheach other andwiththeevolutionkernelH:[Q
1, Q
2] = [Q
1, H] = [Q
2, H] =
0.
(15) Thefirsttworelationsaretrivial,thelastonecanbeverifiedusing theexplicitexpressionsforQ2 andH.If H is interpreted as a Hamiltonian of a certain quantum- mechanical model, the operators Q1 and Q2 correspond to the conservedcharges.Intheformalismofthequantuminversescat- teringmethod(QISM)thechargesQ1
,
Q2appearintheexpansionoftheelementC
(
u)
ofthemonodromymatrix,C
(
u) =
uQ
1+ Q
2.
The commutation relation [C
(
u),
H]=0, and its generalization to more degrees offreedom then follow directly from the QISM [14–16]. Note that in classical applications of integrable models one encounters Hamiltonians that commute with the sumof di- agonal elements, A(
u)
+D(
u)
, of the monodromymatrix. In our casetheHamiltoniancommuteswithC(
u)
,which correspondsto anew,nonstandardintegrablemodel.The conserved chargesQ1, Q2 are self-adjoint operators with respecttotheSL
(
2,
R)
invariantscalarproductΦ|Ψ =
1π
2C−
d2z1
C−
d2z2
Φ(
z)
∗Ψ (
z),
(16)where the integration goes over the lower complex half-plane, Imzi
<
0. The eigenfunctions of C(
u)
provide the basis of the so-calledSklyanin’srepresentationofSeparatedVariablesandare knowninexplicitform[25].Theyarelabeled,forthepresentcase, bytworealnumbers:s>
0 andx∈RsuchthatC
(
u)φ
s,x(
z1,
z2) =
s(
u−
x)φ
s,x(
z1,
z2),
(17) withφ
s,x(
z) =
s z21z22 10
d
α α
¯ α
ixexp
is
( α ¯ /
z1+ α /
z2)
=
sρ (
x)
eis/z1
z21z221F1
1
+
ix,
2,
isz−21
−
z−11,
(18)where
ρ (
x) = π
x/
sinh( π
x).
(19) Theeigenfunctionsφ
s,x(
z)
forma completesysteminthe Hilbert spacedefinedbythescalarproduct(16)φ
s,x| φ
s,x=
2π
s
δ
s−
sδ
x−
x.
(20)SincetheconservedchargesQ1andQ2commutewiththeHamil- tonianH,theysharethesamesetofeigenfunctions,
Hφ
s,x(
z) = γ (
s,
x)φ
s,x(
z).
(21) The simplestway to calculate the eigenvalues is to comparethe large-z asymptotics of the expressions on the both sides of this equation.Inthiswayoneobtainstheanomalousdimensionsγ (
s,
x; μ ) =
2 ln( μ
s/
s0) +
E(
x),
E
(
x) = ψ (
1+
ix) + ψ (
1−
ix) +
2γ
E (22) wheres0=e2−γE.GoingbacktotheRGEequation(5),weexpand theDAΨ
⊥(
z, μ )
overtheeigenfunctionsofHΨ
⊥(
z, μ ) =
∞0
ds s
∞−∞
dx
2
π η
⊥(
s,
x; μ )φ
s,x(
z).
(23)The expansion coefficients
η
⊥(
s,
x;μ )
=φ
s,x|Ψ
⊥ evolve au- tonomously,fB(2)
( μ ) η
⊥(
s,
x; μ ) =
fB(2)( μ
0) η
⊥(
s,
x; μ
0) μ
μ
0 −3β80
× μ
0ss0
38β0lnL
L34β0[E(x)− 4
π β0αs(μ0)]
,
(24)where L=
α
s( μ )/ α
s( μ
0)
,β
0=113Nc−23nf. For large scales, the coefficientsη
⊥(
s,
x;μ )
slowlydrifttowardssmallervaluesofboth parameters:s→0,thankstothefactors8 lnL/3β0,and|x|→0,tak- ingintoaccountthatψ(
1+ix)
∼ln|x|forx→ ±∞.GoingovertotheDAinmomentumspace,
Ψ
⊥( ω ,
u;μ )
,wede- finethecorrespondingeigenfunctionsasφ
s,x( ω ,
u) =
e−iω(¯uz1+uz2)
φ
s,x(
z1,
z2)
.
(25)Usingthate−ikz|z−2eis/z= −
(
1/
√ks
)
J1(
2√ks
)
[23]oneobtainsφ
s,x( ω ,
u) =
1ω √
uu¯
10
d
α
√ α α ¯ α
ixα ¯
−ix×
J1(
2√
ws
α ¯
u¯ )
J1(
2√
ws
α
u).
(26)Theeigenfunctions
φ
s,x( ω ,
u)
areorthogonalandformacomplete set:s 2
π
∞ 0ω
3dω
10
du uu
¯ φ
s,x( ω ,
u) φ
∗s,x( ω ,
u) = δ
s−
sδ
x−
x,
(27)ω
3uu¯
∞0
sds
∞−∞
dx
2
π φ
s,x( ω ,
u) φ
s∗,xω
,
u= δ
ω − ω
δ
u−
u.
(28) Making useof Bateman’s expansionfortheproductoftwo Bessel functionsweobtainaseriesrepresentation
φ
s,x( ω ,
u) =
1ω
∞ n=0in
n−1Cn3/2
(
1−
2u)
Hn(
x) √
1s
ω
J2n+3(
2√
sω ).
(29) Here Cn3/2
(
x)
aretheGegenbauerpolynomials, J2n+3(
x)
areBessel functions,andn
= (
n+
1)(
n+
2)
4
(
2n+
3) .
(30)ThefunctionsHn
(
x)
aregivenbythecontinuousHahnpolynomials uptotheprefactorρ (
x)
:Hn
(
x) =
in 10
du
u¯
u ixCn3/2
(
1−
2u)
= (
n+
1)(
n+
2)
2 in
ρ (
x)
3F2−
n,
n+
3,
1+
ix 2,
21
,
(31) e.g.ρ
−1(
x)
H0(
x) =
1, ρ
−1(
x)
H1(
x) =
3x,
ρ
−1(
x)
H2(
x) =
5x2−
1,
(32) etc. Hahn polynomials are real functions, have the symmetry Hn(
x)
=(−
1)
nHn(−
x)
,andformacompleteorthogonalsystem.In ournormalization ∞−∞
dx
2
π
Hn(
x)
Hm(
x) =
nδ
mn.
(33)Collectingeverythingweobtainthefinalresult:
Ψ
⊥( ω ,
u; μ ) = ω
2uu¯
∞−∞
dx 2
π
∞0
sds
φ
s,x( ω ,
u) η
⊥(
s,
x; μ ),
(34)wherethescaledependenceof
η
⊥(
s,
x;μ )
isgiveninEq.(24).The expansioncoefficients ofη
⊥(
s,
x;μ )
in Hahnpolynomialsare re- latedto theexpansion coefficientsofΨ
⊥( ω ,
u;μ )
inGegenbauer polynomials,η
⊥(
s,
x; μ ) =
n
in
η
n⊥(
s; μ )
Hn(
x) →
Ψ
⊥( ω ,
u; μ ) =
uu¯
n
ψ
n⊥( ω ; μ )
Cn3/2(
2u−
1),
(35) bytheBesseltransform(cf.Eq. (30)inRef.[23])ψ
n⊥( ω ; μ ) =
∞ 0ds
√
s
ω
J2n+3(
2√
s
ω ) η
⊥n(
s; μ ).
(36)Making use of the asymptotic expansion for the Bessel function onefindsthatthesmall-sbehavior
η
n⊥(
s)
∼spn translatesintothe large-ω
asymptoticsofthefunctionψ
n⊥( ω )
∼ω
−1−pn unlessthere issomecancellation,seebelow.Theexpansion coefficientsat thereference(low) scale can be calculatedfromagivenmodeloftheDAas
η
⊥(
s,
x; μ
0) =
∞ 0ω
dω
10
du
φ
s∗,x( ω ,
u) Ψ
⊥( ω ,
u; μ
0).
(37)Intheexisting studies itis usually assumedthat
Ψ
⊥( ω ,
u;μ
0)
is decreasingexponentiallyatlargeenergiesω
.Forarather general modelofthistypeΨ
⊥( ω ,
u; μ
0) = ω
2uu¯
n
cn
ω
n κne−ω/nn4 C
3/2
n
(
2u−
1)
(38)oneobtains
η
⊥(
s,
x; μ
0) =
sn
incn
(
sn
)
nHn(
x)
× Γ (
n+
4+ κ
n) Γ (
2n+
4)
1F1 n+
4+ κ
n2n
+
4−
sn
.
(39)In particular, for the simplest phenomenologically acceptable model[3–5]
Ψ
⊥( ω ,
u; μ
0) = ω
2uu¯
e−ω/004
→ η
⊥(
s,
x; μ
0) = ρ (
x)
se−s0.
(40) Exponentialdecrease ∼e−ω/n ofeach Gegenbauerharmonics in (38)amounts,fromtheviewpointoftherelationinEq.(36),tothe finetuning such that all leading powerterms inthe asymptoticsω
→ ∞drop out. This finetuning is, however,destroyed by the evolutionsothatapower-likeasymptoticsisalwaysgenerated.Toseethis, considerthesimplestmodelin(40)corresponding tothetermn=0 in(38)asanexample.Astheresultoftheevo- lution(24)allharmonicswithn
>
0 becomeexcitedη
n⊥(
s, μ ) =
cn( μ )
s( μ
0s)
−δe−s0 (41) whereδ
= −8/
3β
0lnLandcn
( μ ) ∼
dx H0
(
x)
L4/3β0E(x)Hn(
x).
(42)For the corresponding coefficients in the Gegenbauer expansion (35)oneobtainsusing(36)
ψ
n⊥( ω , μ ) =
cn( μ )
0−20
μ
0 δω
0 n+2× Γ (
n+
4− δ) Γ (
2n+
4)
1F1 n+
4− δ
2n+
4− ω
0
.
(43)The confluenthypergeometricfunction 1F1
(
a,
b|ω )
decreases asa powerofω
atω
→ ∞,cf.Eq.(62)below,unlessa−bisanonneg- ativeinteger,inwhichcasetheasymptoticbehaviorisexponential.Thus,unless
δ
=0 andn=0,weobtainψ
n⊥( ω , μ ) ∼ ( ω /
0)
−2+δ.
(44) Notethattheasymptoticbehavioristhesameforanyn.4. Next,weconsiderheavybaryonswiththelightquarkshav- ing opposite helicity. The scale dependence of the leading twist DAs does not depend on the spin ofthe light quark pair andis thesame forthe jP=0+ SU
(
3)
F triplet andall longitudinalDAs ofheavybaryonsinthe jP=1+ sextets,see[4].Fordefiniteness, considertheΛ
b-baryonDA[3,4]definedas 0uT(
z1n)
Cγ
5nd/ (
z2n)
hv
(
0)Λ(
v)
=
fΛ(1)( μ )Ψ
Λ(
z1,
z2; μ )
uΛ(
v).
(45) The evolution equation forΨ
Λ(
z1,
z2;μ )
contains an additional term corresponding to the gluon exchange between the light quarks(inFeynmangauge)H12
→
H12− δ
H12, δ
H12f(
z) =
10
d
α
¯
α0
d
β
f zα12,
zβ21(46) thatcorrespondstoH→H−1
/
J12(
J12−1)
intheSL(
2)
-invariant representation of the evolution kernel in Eq. (13). ExpandingΨ
Λ(
z1,
z2;μ )
intermsoftheeigenfunctions(18)oftheintegrable Hamiltonian(13)Ψ
Λ(
z, μ ) =
∞ 0dss
∞−∞
dx
2
π η
Λ(
s,
x; μ )φ
s,x(
z)
(47)one obtains the RGE equation for the expansion coefficients
η
Λ(
s,
x, μ )
μ ∂
∂ μ + β( α
s) ∂
∂ α
s+
2α
s3
π
2 ln
μ
ss0
+
E(
x)
×
fΛ(1)( μ ) η
Λ(
s,
x, μ )
=
2α
s3
π
f(1) Λ
( μ )
∞−∞
dxV
x,
xη
Λ s,
x, μ ,
(48)whereE
(
x)
isdefinedinEq.(22)andthekernel V(
x,
x)
isgiven bythematrixelementφ
s,x| δ
H12| φ
s,x= δ
s−
sV
x,
x.
(49)Weobtain V
x
,
x=
1 2π
∞ n=0n−1 H
∗n
(
x)
Hn(
x) (
n+
1)(
n+
2)
=
12 sinh
π (
x−
x)
x−
xxx
− π
sinhπ (
x−
x)
sinhπ
xsinhπ
x.
(50)Fig. 1.ThespectrumofeigenvaluesE(xn)≡EnofthediscretizedversionofEq.(51) x→xn=(2n+1)/200,n=0,1,. . . ,99 (bluedots)comparedtothe“unperturbed”
spectrumE(xn)(redsolidcurve).Inordernottooverloadtheplot,onlyeverysec- ondeigenvalueisshown.(Forinterpretationofthereferencestocolorinthisfigure legend,thereaderisreferredtothewebversionofthisarticle.)
Fig. 2.Thex-eveneigenfunctionsηk+(xn)=η+k(−xn)fork=0 (thegroundstate), andk=10,50,100,150 (fromlefttoright),forL=5 andN=500.Normalization isarbitrary.
Itiseasytoseethat V
(
x,
x)
∼1/(
2π
x2)
forlarge x,anddecreases exponentiallyin|x−x|.Inorder tosolve (48)one needs tofindthe eigenfunctionsof theintegralequation
E
(
x) η
E(
x) − [
Vη
E] (
x) =
Eη
E(
x).
(51) If V →0,obviously all eigenfunctions are localizedin x,η
a(
x)
∼δ(
x−a)
.Thespectrumofeigenvaluesiscontinuous, Ea=E(
a)
≥0, and double degenerate since E(a)
=E(−
a)
. In order to under- stand the effect of the “perturbation” V we consider the dis- cretizedversionofthisequation:x→xn=(
n+1/
2)
x,x=L
/
N, n = −N, . . . ,
N −1. The unperturbed eigenfunctions,η
k(
xn)
=δ
nk, correspond to local excitations at the k-th site. Discretiz- ing the integral in (48) one replaces the original eigenvalue problem (51) by the eigenvalue problem for the matrix Vnk=δ
nkE(
xn)
−xV
(
xn,
xm)
.SinceV(
x,
x)
=V(
−x,
−x)
,alleigenstates have definite parity with respect to x→ −x; the double degen- eracy is lifted and one can study x-even and x-odd eigenstates separately.Diagonalizing thismatrixnumericallywefindthatthe shiftof eigenvaluesas comparedto theunperturbed spectrumis surprisinglysmall,δ
E=E−E≤0.
003, forall eigenstates except forthelowestone,cf.Fig. 1,andthecorrespondingeigenfunctionsη
k(
xn)
remainwelllocalizedaroundthepointxk,seeFig. 2.Atthe sametimethelowestx-eveneigenstatechangesdrastically:Itbe- comes delocalized, seeFig. 2,andseparatedfrom therestof the spectrumbyafinitegap1E
=
E0−
0.
3214.
(52)Inthecontinuumlimit
(
x→0,
L→ ∞)thisphenomenoncanbe understoodascreationofaboundstateinadditiontothecontin- uum spectrumthatremainstobe largelyunperturbed.The“wave function”ofthis(lowest)statecanbeapproximatedtoagoodac- curacy(betterthat1%for|x|<
3)bythefollowingexpression:η
0(
x)
√
2E0√
2+
x2ρ (
x)
[
E0−
E(
x)] , η
0(
0) =
1.
(53) ItcanbeconvenienttoexpandthisfunctioninHahnpolynomialsη
0(
x) =
∞n=0,2,...
χ
nHn(
x),
(54)wherethefirstfewcoefficientsread
χ
00.
612, χ
2−
0.
126, χ
40.
0574,
χ
6−
0.
0338, χ
80.
0226, χ
10−
0.
0163.
(55) Thenormalizationisgivenby ∞−∞
dx
2
π η
02(
x) =
∞n=0
n
χ
n20.
0758.
(56)Coming backtotherepresentationoftheDAintheform(47)we canseparatethecontributionofthediscrete(lowest)levelas
η
Λ(
s,
x, μ ) = ξ
0(
s, μ ) χ
0−1η
0(
x) + η
Λ(
s,
x, μ ).
(57) wherethefunctionη
Λ(
s,
x, μ )
accountsforthecontributionofthe continuumspectrumandmustbeorthogonaltoη
0(
x)
,dx
η
0(
x) η
Λ(
s,
x, μ ) =
0.
(58)Goingovertothemomentumspaceweobtainforthecontribution oftheloweststate(asymptoticDA)
fΛ(1)
( μ ) Ψ
Λ(0)( ω ,
u; μ )
=
fΛ(1)( μ
0) χ
0−1ω
2uu¯ μ
μ
0 −38β0L
4
3β0[E0−β0α4sπ(μ0)]
∞−∞
dx 2
π η
0(
x)
×
∞ 0sds
φ
s,x( ω ,
u)ξ
0(
s, μ
0) μ
0ss0
3β80lnL
,
(59)cf. Eq. (24). Note that the restriction to the contribution of the discrete level impliesa certain relation betweenthe momentum fractiondistribution betweenthe twolight quarksandtheir total momentum
ω
, the remaining freedom is encodedin the “profile function”ξ
0(
s, μ
0)
atthereferencescale,which canbe arbitrary.Forthesimplestansatz
ξ
0(
s, μ
0) =
se−s0,
(60)1 Thesizeofthegapcoincideswiththegapinthespectrumofanomalousdimen- sionsofthree-light-quarkoperatorsinthelarge-Nlimit[7],indicatingthatthese problemsarerelated.Unraveling thisconnectiongoesbeyondthetasksofthislet- ter.