A thesis submitted forthe degreeof Do tor of Philosophy in Physi s
General Mass S heme
for Jet Produ tion in QCD
by
Piotr Kotko
writtenunderthesupervisionof
1 Hadrons, partons and jetsin QCD 13
1.1 Fa torizationtheorems . . . 13
1.2 JetsinQCD. . . 18
1.3 Quarkmassesin QCD . . . 22
2 In lusiveDIS with heavy quarks 25 2.1 Introdu tion. . . 25
2.2 Zero-massvariableavournumbers heme . . . 26
2.3 Fixedavournumbers heme . . . 28
2.4 ACOTs heme . . . 30
3 Massive dipole subtra tion method 35 3.1 Introdu tion. . . 35
3.2 Singularbehaviouroftree-levelmatrixelements. . . 36
3.2.1 Softlimit . . . 36
3.2.2 Quasi- ollinearlimit . . . 40
3.2.2.1 Initialstateemitter ase . . . 40
3.2.2.2 Finalstateemitter . . . 43
3.3 Constru tionofdipoles. . . 44
3.4 Dipolekinemati s. . . 46
3.4.1 FinalStateEmitter-Initial StateSpe tator. . . 46
3.4.1.1 Standarddipolekinemati s . . . 48
3.4.1.2 Kinemati swithadditionalinvariant(gamma-kinemati s) 48 3.4.1.3 Softand ollinearlimitsofthedipole variables . . . 50
3.4.1.4 RelationbetweendipoleandtheSudakovvariables. . . . 51
3.4.1.5 Dipolevariable asafreeparameter . . . 52
3.4.2 InitialStateEmitter-FinalStateSpe tator. . . 53
3.4.2.1 Standardkinemati s . . . 54
3.4.2.2 Gamma-kinemati s . . . 54
3.4.2.3 Dipolevariable asafreeparameter . . . 55
3.4.2.4 Softand ollinearlimits . . . 56
3.4.2.5 RelationbetweendipoleandtheSudakovvariables. . . . 57
3.5 Dipolesplittingfun tions . . . 57
3.5.1 Finalstateemitter-Initial statespe tator . . . 57
3.5.1.1
Q
→ Qg
andQ
→ Qg
splittings . . . 573.5.1.2
g
3.5.2 Initial StateEmitter-FinalStateSpe tator. . . 61
3.5.2.1
Q
→ Qg
andQ
→ Qg
splittings . . . 613.5.2.2
g
3.5.2.3
Q
→ gQ
andQ
→ gQ
splittings . . . 633.6 Phasespa efa torization . . . 64
3.6.1 Preliminaries . . . 64
3.6.1.1 Two-parti lephasespa e . . . 65
3.6.1.2 Two-parti lephasespa eindipolevariables. . . 66
3.6.2 Finalstateemitter-initialstatespe tator. . . 67
3.6.3 Initial stateemitter -nalstatespe tator . . . 68
3.6.4 Fa torizationofthree-parti lephasespa e . . . 68
3.6.4.1 Expli itexamples . . . 71
3.7 Integrationofthedipoles . . . 71
3.7.1 Thenotation . . . 73
3.7.2 Finalstateemitter-Initialstatespe tator . . . 74
3.7.2.1
Q
→ Qg
andQ
→ Qg
splittings . . . 743.7.2.2
g
3.7.2.3
g
→ gg
splitting . . . 813.7.3 Initial StateEmitter-FinalStateSpe tator. . . 84
3.7.3.1
Q
→ Qg
andQ
→ Qg
splittings . . . 843.7.3.2
g
3.7.3.3
Q
→ gQ
andQ
→ gQ
splittings . . . 903.7.4 Complete expressionsforintegrateddipoles . . . 92
4 General Mass S heme for Jets 97 4.1 Introdu tion. . . 97
4.2 PartondensitiesinCWZ renormalizations heme . . . 97
4.2.1 Unrenormalizedpartondensities . . . 98
4.2.2 Renormalizationofpartondensities . . . 102
4.3 Quasi- ollinearsubtra tiontermsformassivedipoles . . . 103
4.3.1 Masslesstreatmentoffa torization . . . 103
4.3.2 Fullymassive ase . . . 106
4.4 Masslesslimitand onsisten y he k . . . 110
4.5 Pra ti alappli ationand
MassJet
proje t . . . 1135 Summaryand outlook 117 A Kinemati s 123 A.1 Thermodynami softheinvariants . . . 123
A.2 Expli itexpressionsforres aledvariables . . . 125
A.2.1 FinalStateEmitter-InitialStateSpe tator. . . 125
A.2.1.1
Q
→ Qg
andQ
→ Qg
splittings . . . 125A.2.1.2
g
g
→ gg
splittings . . . 126A.2.2 Initial StateEmitter-FinalStateSpe tator. . . 126
A.2.2.1
Q
→ Qg
andQ
→ Qg
splittings . . . 126A.2.2.2
g
A.2.2.3
Q
→ gQ
andQ
→ gQ
splittings . . . 127B Mathemati al supplement 131
B.1 Theintegrals . . . 131
B.2 Expansionsin
ε
. . . 136B.3 Theplusdistribution . . . 138
Quantum hromodynami s(QCD) is aswebelievethe orre ttheory ofthestrong
intera tions,withquarksandgluonsbeingitsfundamentaldegreesoffreedom. Although
there are many puzzles remaining unsolved, it is verysu essful in des ribing various
aspe tsofthemodernhighenergydata. Theoreti alpredi tionsarebasedontwomajor
issues. Themostimportantoneistheasymptoti freedom,whi hassertsthatthevalue
of the strong oupling onstant is relatively low at high energy s ales. It enables us
to use perturbation theory in al ulations on ernings attering amplitudes. However,
there are no freequarksand gluons in the nature they are allbounded in olourless
hadrons,thus theperturbative al ulationsare notthewhole story,asthehadronsare
learly of non-perturbative nature. Therefore, the se ond basi issue are fa torization
theorems, whi h allow fora separationof apro ess to a non-perturbativebound state
physi s and al ulable in QCD hard s attering amplitudes. Te hni ally the former is
des ribed in termsof variousdistribution fun tions and distribution amplitudes,whi h
sofararemostreliablytakenfromexperiments.
Phenomenologi allythemostimportantnon-perturbativeinput omprisesparton
dis-tribution fun tions (PDFs). Histori ally, they appeared in a des ription of in lusive
lepton-hadrondeepinelasti s attering(DIS)asanelementofthepartonmodel.
Nowa-daystheyareusedalsointheotherhighenergyexperimentslikeproton-proton ollisions
atLHC forinstan e. Inorder toobtainPDFsonehasto tthetheoreti ally al ulated
ross se tionwith suitable analyti alparametrizationsof PDFs to thereal data. Most
often the ele tron-proton HERA data are used in this pro edure. There are several
groupsmakinganeortin extra tingPDFs,e.g. CTEQgroup[5℄orMRSTgroup[41℄
tomentiononlythemostknown.
Onemayaskthequestion: whatisthedieren ebetweenvarioussetsofPDFs? There
areatleast afewodds. Therstaredierentfun tionalparametrizationsanddierent
statisti almethodsusedinttingthedata. Howeverthemaindieren eis onne tedto
as hemeinwhi hthea tual rossse tionis al ulated. Thisissueisinseparablyrelated
to heavyquarks and the problem of s alesin QCD. Namely, there is a di ulty with
nite-orderperturbative al ulationsifthereareafewexternalparameters(s ales)that
areverydierent. Thoses ales anbexede.g. byanexternalenergyorbythemasses
ofthequarks. A tuallyitisthe aseinreality,aswehavesixquarkavourswiththreeof
thembeingmarginallyheavierthantheothers. Moreoverthereisalsosubstantialmass
splittingsbetweentheheavyquarks. Insomesituations,thisdi ulty anbesolvedby
means of the renormalizationgroup methods, however there does notexist a uniform
perturbativeexpansionsuitableforallthes ales. Thereforeonehasto hooseaspe i
s heme. Intuitivelyit orresponds toa situation,where our measurementresolution is
toosmall to distinguishsometinydetails and to largeto see the whole thing. We an
s hemesinperturbativeQCD.
Formanyyears,a ompletelymasslesss hemehavebeenthemoststandardintreating
DIS s attering. A tually, thes heme wasmasslessin the sense of negle tingthe mass
parametersin al ulations,but hangingthenumberof avoursin thesametime. We
shallseethedetailslater,howeverevenintuitivelyweseethatsu hanapproa hisvery
limited in a ura y. It an be satisfa tory only in ertain, relatively narrowranges of
kinemati spa e. Thereforealsotheothers hemeswereused,treatingtheheavyquarks
in amoreordered way. Inparti ular, the s hemesfor harm, and bottomquarkswere
used. Those al ulationsareverya uratein asuitablekinemati range, butagain
thoserangesarelimited. Furtherdevelopmentmusthaveledtoa omposites heme,that
isto hangethenumberofavoursononehand (likeinthemasslesss hemementioned
in the beginning of the paragraph) and to keep the masses nite on the other. The
most ommonnamefortheapproa hofthistypeisageneral-masss heme orvariable
avournumbers heme. Weshallseesu hasolutionindetailsbelowinthiswork.Sofar
thegeneral-masss hemeswereusedin in lusivepro esses,bothin extra tingthePDFs
andpredi tingexperimental out ome.
There ishoweveranotherveryimportant lassofhigh energypro esses,namelythe
produ tion of jets. Sin e one measures also the spatial distribution of the outgoing
parti lesit angivemu hmoreinformationaboutunderlyingpartondynami s. Loosely
speaking,ajeta ollimatedbun hofhadronsisaremnantofapartoneje tedfromthe
enterof ollision. Thusbyanalysingthemomentumandenergyofthejet,wegetalmost
dire ta ess tothepartonlevelsubpro ess. Itallowsformorepre ise measurementsof
somequantities, forinstan e strong oupling onstant(e.g. theanalysis performed by
ZEUS ollaboration using dijet produ tion in DIS [1℄ and by H1 ollaboration using
in lusive-jet,dijetsandtrijetanalysis[32℄). Thejetsprodu tionpro essesarealsoused
toobtainthepartondistributionfun tions(togetherwiththein lusivedata). Theoreti al
al ulationsneededtothispro edure areagains hemedependent. In aseofjets,there
ishowevermu hlesstheoreti aldevelopment on erningheavyquarks. Thereareseveral
Monte-Carlo(MC)programsusingmasslessquarks,e.g. NLOJET++[42℄,DISENT[9℄,
bothforhadron-hadronandlepton-hadron ollisions(thelastforneutral urrentwithout
Z
0
-ex hange). Forheavy quarks, there are some al ulationsfor in lusive-jetand two
jets produ tion at NLO [26℄ in a s heme with xed number of avours. It should be
remarked,thatwemeanherestri tQCD al ulations,notamodel-basedones. Forjets,
theformer aremu h moreinvolved andrequirespe ial treatments ofsingularities that
appearat NLO(andhigher)orders.
Inthiswork,weproposeasolutionintendedto llthegapin existing heavyavour
treatments. It is a general-masss heme for jets produ tion pro esses, based on some
solutionsavailableonthemarket. We on entratehereinonDISpro esseswithneutral
urrentintera tions. Furtherextensionsarepossible,asweshortlydis ussinChapter5. The developments we are presenting are essentially theoreti al. However, in order to
support the validity of our al ulationswe givesome samplenumeri al resultsusing a
dedi atedMC program. Itisapartofalargerproje tthat is urrentlyunder
develop-ment.
Thematerialisorganisedasfollows. First,inChapter1were allthebasi formalism weshallusethroughout,in ludingfa torizationtheoremsandjetstreatment. Chapter2
is devoted to existing approa hes to heavy quarks in in lusive DIS pro esses and its
problems. Notably,itintrodu esthegeneral-masssolution,whi hwelaterapplytojets.
re-ofmassivepartons,in ludingpossibleinitialstateheavyquarks. Finallywegatherallthe
pie es and onstru tthe general-masss hemeforjetsin Chapter 4. Weintrodu e tons ofsymbolsthroughoutthis work. Someofthem maylook messy,howeverthisa ounts
forthepre isetheoreti alformulationofthematerial. Inorder tofa ilitatethereading
weput someofthemin aNomen lature. Thete hni aldetailsthat arenotessentialin
themaintextarelistedin theappendi es.
A knowledgements
Inthe rst pla e, I would liketo thank Prof. Woj ie h Sªomi«ski for his patien e,
substantialhelpandmanyhoursofjointdis ussions. Thewordsofgratitudebelongalso
to Prof. Mi haª Praszaªowi z, the head of Parti le Theory Department I had a great
pleasureto beamemberasPHDstudent.
I thankmywife Iwonafor onstantsupport andfaithin me, espe iallywhen
every-thing wasgoing wrong. The sameis trueabouther family: Brunonand Basia, Marek
andElatogetherwiththeir hildren.
I own spe ial thanks to Mirek Tro iuk, my high s hool tea her and Ola ubni ka,
who reatedthes ienti atmospherethatbroughtmetothis point.
Finally, I am grateful to my parents Maria and Jerzy, my brothers Damian and
Hadrons, partons and jets in
QCD
1.1 Fa torization theorems
Although QCD has in redible amount of su esses, the theory is still not solved. For
instan e, there isa olour onnementhypothesis, stating that all observableparti les
are olour singlets. This onje ture has verystrong experimental eviden e; so far free
quark or gluon has not been found. However, su h a property has not been derived
yetfrom QCD,althoughthere areseveraltheoreti al lues,bothperturbativeand
non-perturbative. Moreover,theredoesnotexista ompletedes riptionof omposedobje ts
like hadrons in terms of the fundamental QCD degrees of freedom (i.e. quarks and
gluons). Forexample,it is known that manyfeatures of aproton anbe explainedby
assuming that it is build of three quarks
u, u, d
. Theirmasses (i.e. the parameters inQCDlagrangian;massispoorlydenedquantityforanunobservableparti le)areabout
afew
MeV
. Ontheotherhand,theprotonmassiswelldenedand anbemeasureditturnsouttobearound
1 GeV
, learlynotaboutthreetimesthemassesof onstituents.Thisisaneviden eofveryimportantnon-perturbativephenomenon,namelyspontaneous
hiral symmetrybreaking. It generatesso alled onstituentquarkmass,whi h should
beaboutonethirdoftheprotonmass. Su havalue annotbedes ribedbyperturbation
theory,thetoolwhi h atpresentisbest understoodandunder ontrol. Therearemu h
moreproblemsin des ribinghadronswithin perturbativeQCD.
All these features draw hadrons as a very ompli ated, non-perturbative obje ts.
Nevertheless, there are possibilitiesto get ertain insightinto the stru ture ofhadrons
using perturbation theory. As it wasalreadymentioned, QCDhas apropertyof being
asymptoti allyfree, i.e. at veryshort distan es theQCD oupling isveryweak, giving
some han es to useperturbation theory. This is a keyobservationleading to modern
high-energy experiments; ollisionsof parti les with higher energies anprobe smaller
spa e-time volumes. However,sin e weprobeonlyasmall partofa olourlesshadron,
we anhopetohaveperturbativedes riptiononlypartiallytherestmustbesomehow
parametrized,orobtainedbyothermethods. Thisis in fa t averyloosedes riptionof
famous fa torizationtheorems, whi h we nowshall re allin somedetails. Inour
intro-du tionweshalltrytogivemostlyne essaryresults,butwere allalsosome ompletely
We are mainly on entrated on lepton-hadrondeep inelasti pro esses throughout.
Moreover, in this se tion we limit ourselves to in lusive pro esses only. There are
es-sentiallytwopossibleapproa hestofa torization,whi hper olateatsomestages. Both
haveitsown ons andpros.
Firstone,reliesontheoperatorprodu texpansion(OPE)[54℄andishistori allythe
rstapproa hto fa torization[7℄, of ourseex eptFeynman'sparton formalism
onsid-eredbefore QCDhad beenborn. Although OPEallows forverysystemati treatment
ofalltermsthat anappear, itsappli abilityisratherlimitedtothein lusivepro esses
only.
Se ondapproa hisbasedongeneralpower ountingtheorems [40,39℄and methods
developed in [21℄. Letus re all thebasi ideas, astheyshall be importantlater, when
wedis ussmore ompli atedtopi s. Forareviewsee e.g. [17,15℄.
Considerageneri unpolarized boson-hadron utamplitude,asshown inFig.1.1A. Wedenoteprotonmomentumas
P
andbosonasq
,withq
2
=
−Q
2
. Moreover,weassume
thatthebosonvirtuality
Q
2
ismu hlargerthanallthequarkmasses(in ludingpossible
heavyquarks)andthattheBjorkenvariable
x
B
= Q
2
/P
· q
isxed. ThesituationwhereQ
2
is of the same order as the mass of a given heavy quark will be dis ussed in the
next hapter. Itturnsoutthatalltheleading ontributionstothe utamplitude anbe
hara terizedbythe utamplitudesthat havetheformshowedinFig.1.1B.Theupper blob hasallthe internal momentao-shell by order
Q
2
andthus is alled ahardpart.
Note,thatalthoughsomeoftheinternallinesare utandhen eon-shell,theyee tively
an be treated as o-shell lines by virtue of the opti al theorem. The lower part in
Fig.1.1B, the soft part, in orporates hadroni states and two partoni lines joining it withthehardpart.Thoselinesareeitherquarkorgluonlineswithvirtualitymu hlower
than
Q
2
andmomenta ollineartothehadroni momentum. Itshouldbementionedthat
theinternalblobofthesoftpart, anstillhaveUVsingularities,seebelow.
The ontributions that have stru ture des ribed above are alled twist-2, as they
orrespondinOPElanguagetoaseriesofmatrixelementsoflo aloperatorswithtwist
1
equalto2. Contributionswhi hhavemorethantwolinesjoininghardandsoftpartshave
highertwist. Re allthatsu hhighertwist ontributionsaresuppressedby
m
2
/Q
2
,where
m
2
isthemassoftheheaviestquarktakenintoa ount. Thereareseveral ompli ations
(seee.g. [15℄),howeverthegeneralpi tureisasjustdes ribed.
Note,that thetwolinesjoining bothparts annot orrespondtoaheavyquarkwith
massoftheorderof
Q
2
,duetotheassertionthat theyhavelowvirtuality omparingto
Q
2
. Thisfa t shallbeimportantlateron.
Now,we ometo morepre ise denitions ofthe softpartand its onne tionto the
restofthe pro ess. As is ommonlyknown, thesoftpart anbeparametrizedinterms
ofpartondistributionfun tions (PDF)(weshallinter hangeably allitparton density)
insideahadron. Inordertopro eedweintrodu elight- one oordinates;anyfour-ve tor
v
anbede omposedasv
µ
= v
+
n
˜
µ
+ v
−
n
µ
+ v
µ
T
,
(1.1)where
v
+
= v
· n, v
−
= v
· ˜n
(1.2)withtwolight-likeve tors
n
,n
˜
denedasn =
√
1
2
(1, 0, 0,
−1) , , ˜n =
1
√
2
(1, 0, 0, 1) .
(1.3) 1A) B) PSfragrepla ements
q
P
PSfragrepla ementsq
P
k
Figure1.1: A)CutFeynmanamplitudeforunpolarizedboson-hadronpro ess. B)
Lead-ingregions ofthe utamplitudeforlargevirtuality oftheboson. Thelines onne ting
upperandlowerpartshavelowvirtualityand anbelightquarksoragluon.
Letusnowassumethatthemomentum
k
joiningthehardandsoftpartsisparametrizedusing the light- onevariables and that suitable frame is hosen, su h that
P
+
islarge
∼
p
Q
2
. Then,sin e
k
havesmallvirtuality omparingtoQ
2
its
k
−
and
k
T
omponentsanbenegle tedinthehardpart.Then,the onne tionofthetwoparts anberealisedas
anintegraloverthe
k
+
omponent. Therestofthemomentumintegration(i.etransverse
andminus omponents)areembodiedin thedenition ofPDF, wherethey annotbe
negle ted. Its external lines ( onne ting it with the hard part) ee tively lie on the
light- one.
All these remarks leadto the followingdenitions of the parton distributions. For
thequarkdensitywehave
f
q
(B)
(x) =
1
4π
ˆ
dy
−
e
−ixP
+
y
−
P
ψ
q
y
−
n
γ
+
y
−
n, 0
ψ
q
(0)
P
(1.4) andforgluonf
(B)
g
(x) =
1
2πxP
+
ˆ
dy
−
e
−ixP
+
y
−
D
P
F
A
+µ
y
−
n
y
−
n, 0
AB
F
+
B µ
(0)
P
E
.
(1.5) Letusnowexplaintheabovenotation. First,therearequarkeldoperatorsψ
q
andthegluoneldstrengthoperator
F
µν
C
= ∂
µ
A
ν
C
− ∂
ν
A
µ
C
+ g f
CDE
A
µ
D
A
ν
E
. All theseeldsare unrenormalized,thusthePDFsdened in su hawayarethebareones asindi atedbythesupers ript
2
. Parameter
x
orrespondsto afra tion ofplus omponentofhadronmomentum that is transferred to the hard part, that is we assume
k
+
= xP
+
and isxed. Next,
y
is a spa e-time point we integrate over, with however xedy
+
= 0
;
theintegrationover
y
−
indisentangled whiletheone over
y
T
isperformed(orhidden).Finally,there is agaugelink in orderto makethedenitions gaugeinvariant. It reads
in thepresent ase
y
−
n, 0
= P exp
(
ig
ˆ
y
−
0
dz A
+
C
(z) t
C
)
,
(1.6)where thepath joining bothpoints is hosen to be astraightline. Inparti ular, when
weuse light- one gaugedened as
A
· n = 0
the gauge link is a unity operator (it is2
Sin ewefollowheremainly[15℄andotherpapersofthisauthor,weusethetermbareinthesense
useful in somegeneral onsiderations). The last remark on erning (1.4), (1.5) is that only onne teddiagramsshould betakenintoa ount.
Asalreadymentioned,thepartondistributionfun tionsdenedabove ontainUV
di-vergen es. Requiredrenormalization on ernsnotonlytheelementaryelds,butalsothe
bilo alquarkorgluonoperatorsitself. Asiswellknown,therenormalizationintrodu es
additionaldependen e onaprioriunspe iedmasss ale
µ
r
.It anbeproved that the relationbetween thebare densities andthe renormalized
oneshastheform[16,15℄
f
a
(R)
x, µ
2
r
=
X
b
ˆ
dz
z
K
ab
z
x
, α
s
µ
2
r
; ε
f
b
(B)
(z) ,
(1.7)wheretherenormalizationkernel
K
ab
isaperturbatively al ulablequantity. Note thatwehaveintrodu eddimensionalUV regulator
ε
denedasD = 4
− 2ε,
(1.8)where
D
is the spa etime dimension. The summation in (1.7) goes over all possible kindsofthelinesjoiningthesoftandhardparts (exa tsetsshallbedenedin thenexthapter). The kernel
K
ab
an be al ulatedby onsidering thesameobje tsasf
a
butwiththehadroni statesrepla edbythepartoni ones. Thuswedenethequantity
F
ab
,whi hwerefertoasadensityofparton
b
insideapartona
. Thedenitionisexa tlythesameasfor
f
b
withthe hadroni staterepla edby theon-shellstatea
. ThequantitiesF
ab
an be al ulated perturbatively in QCD with the help of spe ial Feynman rules [16, 17℄ we shall use them for massive quarks in Chapter 4.2. A tually, we have to again distinguish between the bareF
(B)
ab
and the renormalized oneF
(R)
ab
, howeverthe relation between the two remains the same as (1.7). This allows to obtainK
ab
on e spe i renormalizations hemeis hosen(seealsobelow).Sin e the bare densities
f
(B)
a
aredened bymeans of the bare elds only, they are ompletelyindependentontherenormalizations ale. Thereforeitisrelativelystraight-forwardtoderiveanevolutionequationforthedensities. Itreads
d
d log µ
r
f
a
(R)
x, µ
2
r
=
X
b
ˆ
dz
z
P
ab
z
x
, α
s
f
b
(R)
z, µ
2
r
,
(1.9)wheretheevolutionkernel
P
ab
isrelatedto therenormalizationkernelbytheformulaP
ab
z
x
, α
s
= 2α
s
∂ K
ab,
1
z
x
, α
s
∂α
s
,
(1.10)with
K
ab, n
denedbytheLaurentexpansionK
ab
(z, α
s
; ε) = δ (z
− 1) δ
ab
+
∞
X
n=1
1
ε
n
K
ab, n
(z, α
s
) .
(1.11)Forexample,inthe
MS
s hemewithN
f
avoursweobtainP
ab
(z, α
s
) = δ (z
− 1) δ
ab
+
α
s
2π
P
(1)
ab
(z) +
O α
2
s
,
(1.12) whereP
(1)
ab
arefamouslowestordersplittingfun tions. TheyreadP
(1)
(z) = C
F
1 + z
2
1
− z
+
,
(1.13)P
gg
(1)
(z) = 2C
A
"
1
1
− z
+
+
1
− z
z
− 1 + z (1 − z)
#
+ δ (1
− z)
11
6
C
A
−
2
3
N
f
T
R
,
(1.14)P
gq
(1)
(z) = T
R
[1
− 2z (1 − z)] ,
(1.15)P
qg
(1)
(z) = C
F
1 + (1
− z)
2
z
.
(1.16)Theplus distributionisdened inastandardwayas
h
+
(z) = h (z)
− δ (1 − z)
ˆ
1
0
dy h (y) .
(1.17)Note, that the support is
[0, 1]
we pay attention to this detail, sin e we shall oftenuse distributions with dierent supports (see also Appendix B.3). Sin e the splitting fun tions
P
(1)
ab
areoftenusedinthis thesiswedropthesupers riptinwhat followsP
ab
(1)
(z)
≡ P
ab
(z) .
(1.18)We turn also attention to our onvention of ordering the subs ripts. The notation
ab
orresponds to a splitting pro ess
a
→ b
, where partonb
takes the fra tionz
of theoriginalmomentum. Thephysi alinterpretation ofthefun tions
P
ab
isthensu h, thatitgivesaprobabilitydensityforsu h asplitting.
Let us now ome ba k to the fa torization. On e the renormalizationof the PDFs
and of the hard part is done, we an nally write the fa torization formula. In what
followswedroptherenormalizationindi ationinthehadroni PDFs
f
a
(R)
z, µ
2
r
≡ f
a
z, µ
2
r
.
(1.19)Thefa torizationtheoremtakesthefollowingform
dσ P, q; x
B
, Q
2
=
X
a
ˆ
1
x
B
dz
z
f
a
z, µ
2
f
, µ
2
r
dˆ
σ
a
zP, q; Q
2
, µ
2
f
, µ
2
r
+
O
m
2
Q
2
.
(1.20)Here
dσ
orrespondstoadierentialDISin lusive rossse tion,whiledˆ
σ
a
isapartoniross se tion whi h is infra-red (IR) nite. Besides UV singularities, there are also
divergen eswhi horiginateinzeromassofthegluonsandtherearetwosortsofthem: the
softsingularitiesandthe ollinearones. Theyremainevenafterrenormalization,however
thesoftandmixedsoft- ollineardivergen esare an elledbetweendierent ontributions
(we shalltake upthis issue in thenext se tion). What remains are the ollinear ones.
Thefa torizationpro edureasserts,thatthey anbein ludedinPDFsasitisessentially
anonperturbativeobje tandweshallnever al ulateitusingperturbationtheory. Su h
apro edure isattheexpenseofintrodu ingadditionalfa torizations ale
µ
f
. Aprioriitisarbitrarys aleandoneoftensetsitequaltotherenormalizations ale. Moreover,there
is ertainfreedomin hoosinga tuallysubtra tedterms. Su hapres riptiondenesthe
thekernel
K
ab
). Thus,wehavedσ
a
(R)
p, q; x, Q
2
, µ
2
r
=
X
b
ˆ
1
x
dz
z
h
S µ
2
r
, µ
2
f
F
ab
(R)
z, µ
2
r
+ S
ab
z, µ
2
r
, µ
2
f
i
dˆ
σ
b
zp, q; Q
2
, µ
2
f
, µ
2
r
+
O
m
2
Q
2
,
(1.21)where the fun tions
F
(R)
ab
are the renormalized densities of parton inside aparton dis- ussedbefore (weindi ated alsothat unsubtra ted rossse tion is renormalized). Thequantities
S
andS
ab
deneourfa torizations heme,seebelow. Theaboveequation anbesolvedorderbyorder, al ulating
dσ
(R)
a
andF
(R)
ab
toadesiredorder.Asanillustration,letus onsider ompletelymassless ase. Choosing
MS
s hemetodenePDFswegetatthelowestnontrivialorder
F
ab
mMS
x, µ
2
r
= δ (1
− x) δ
ab
+
α
s
µ
2
r
2π
−
1
ε
P
ab
(x) .
(1.22)The supers ript
mMS
expli itly indi ates that weuseMS
renormalizations heme andompletelymassless al ulation. Sin e
MS
anbealsoused inamassive ase,wefeelane essitytodistinguishbothsituationsasweshallen ountertheminonepla elateron.
Weseethat there isa ollinearpole
1/ε
intheresult, whi h an elsthesimilarpoleindσ
a
. Next,ifwe hoosethefa torizations hemetobeMS
,wehaveS
ab
z, µ
2
r
, µ
2
f
= 0,
(1.23)S µ
2
r
, µ
2
f
=
1
Γ (1
− ε)
4πµ
2
r
µ
2
f
!
ε
.
(1.24)Let us on lude this se tion by giving some summarizing remarks. First is that
hadroni PDFsareessentiallynonperturbative,andhavetobeobtainedfromexperiment,
latti e al ulationsorlow energyee tivemodels. Most reliablearethose obtainedby
global tsto data (e.g. [38℄). Moreover,PDFs ares hemedependent, andassu h are
unphysi al. Thereforeonehavetobe arefulwhenmixingPDFsobtainedbyonemethod
with al ulationsin someother s heme,asthereminder(
O (. . .)
terms) infa torizationtheorem anbe omelarge.
1.2 Jets in QCD
Inthepreviousse tionwehave onsideredthefa torizationtheoremessentiallyfor
in lu-siveDISs attering. Oneoftheelementsofthea tualproofofthefa torizationproperty
isthe an ellation ofthesoftsingularities. Inthisse tion, wetakea loserlookat this
problem. Inparti ular,wedes ribeamethodallowingforthis an ellationin asewhen
thepro essisnotfullyin lusivebut onsistin jets. Thisshallbeaverygeneral
presen-tationofthetopi anditwillevolvethroughoutthewhole dissertation. Wefollow[9℄in
thisintrodu tion.
Beforewestart,letusintrodu esomenotation. The
n
-parti leinvariantphasespa e(PS)shallbedenoted as
A) B) C) PSfragrepla ements
n
PSfragrepla ementsn + 1
PSfragrepla ementsn
Figure 1.2: Illustrative presentation of the amplitudes for
n
-jet produ tion. A) LOamplitude,B)realemission orre tions,C)virtual orre tions.
where
p
andq
arein omingmomenta. Ontherighthandsidewehaveusedmathemati- ians'notationforsets, asitwilloftenallowtomakeformulaeshorter. Thephasespa e
anbeexpressedas
dΦ
n
(p, q;
{p
i
}
n
i=1
) = (2π)
D
δ
(D)
p + q
−
n
X
i=1
p
i
!
n
Y
i=1
dΓ
i
(1.26)in termsoftheinvariantmeasuresforaparti le
i
dΓ
i
≡ dΓ (p
i
) =
d
D
p
i
δ
+
p
2
i
− m
2
i
(2π)
D−1
.
(1.27)
All thedenitions arewrittenin
D
spa e-timedimensions.A tree-levelamplitude with in omingmomenta
p
,q
andn
outgoingstates shall bedenoted as
M
n
(p, q;
{p
i
}
n
i=1
)
. Atthis stage allpossible olour orspinindi es are sup-pressed. Ifrelevant,wewilladorntheamplitudebyvarioussymbolsand/orindi es,forexample we will put ahat if we onsider external fermions to underline that wework
withamatrix. Veryoftenwewillrefertoapartoftheamplitude,forexamplewhentwo
externallegsarerepla edbyone. Thenthereminderisreferredtoasredu edamplitude.
Let us now swit h to the a tual matter of this se tion. We start with the very
s hemati des riptionof NLO al ulation for
n
-jets. Suppose for simpli ity that therearenoinitial statehadrons,e.g. ele tron-positronannihilation. A detailed formulaefor
DISshallbegivenin Se tion4.3.
ToNLOa ura y,thetotal rossse tion anbewrittenas
σ
n
= σ
LO
n
+ σ
NLO
n
.
(1.28)Theleadingorder ontributionreads (Fig.1.2A)
σ
n
LO
=
ˆ
dΦ
n
|M
n
|
2
F
n
,
(1.29)where
M
n
anddΦ
n
are explainedabove(we suppressallmomenta dependen e), whileF
n
is ertain(generalized) fun tionthat givesusanobservableweareinterestedin(i.e it may in lude step-fun tions for kinemati uts, delta fun tions for dierential rossse tion, jet algorithms et .). We shall refer to
F
n
as a jet fun tion. We des ribe itspropertiesin detaillater.
Thenext-to-leadingordertermhasin turnthefollowingform
σ
NLO
where
σ
R
represents the real orre tions, i.e the ones onne ted to the emissions of
additionalon-shellparti lesin thenalstate(Fig. 1.2B).Next,
σ
V
orrespondstoloop
orre tionsto
M
n
(Fig.1.2C).Thelast anbewrittenasσ
n
V
=
ˆ
dΦ
n
M
(loop) 2
n
F
n
.
(1.31)Thenotationissymboli here,
M
(loop) 2
n
isa tuallyaninterferen ebetweenthetreelevel amplitudeandtheone ontainingloop orre tions. Forthereal orre tionswewriteσ
n
R
=
ˆ
dΦ
n+1
|M
n+1
|
2
F
n+1
.
(1.32)As alreadystated in thepreviousse tion, higherorder al ulationsin QCDlead to
divergen es.First,thereareUVsingularities,whi hareremovedbyrenormalizationand
we donot onsider them here any more. Se ond, there are mentionedIR singularities
oming from vanishing propagators due to almost zero energy of masslessparti les or
ollinear emissions. We shall dene them pre isely in Se tion 3.2. Both kinds of sin-gularities appear in
σ
R
andσ
V
and are regularized e.g. dimensionally. However the
physi al rossse tion,whi hdoesnotdistinguishbetweenthesoftor ollinearemissions,
hasto benite. Therefore IR singularities haveto an elbetween both termsin ross
se tions (ex ept possible pure ollinear singularities onne ted with initial state
emis-sionswhi h are removed byfa torization). It is pre isely statedby meansof the KLN
theorem (Kinoshita-Lee-Nauenberg)and itsextensions, see e.g. [48, 49℄and referen es
thereintotheoriginalpapers. Inwhatfollowsweassumethatthejet rossse tionunder
onsiderationisinfra-redsafe,that isitfulls alltheassumptionsoftheKLNtheorem.
This howeverrequires to impose somerestri tions on the jet fun tions. Namely, if
oneof thenal stategluonsin
(n + 1)
-parti lephasespa e is soft(itsfour-momentumvanishes)we musthave
F
n+1
= F
n
. Similarly, iftwoofthe nalstate partonsbe omeollinear,their
F
n+1
fun tionmustalso oin idewithF
n
. Ontheotherhand,ifweenterasingularregionin
n
-parti lephasespa eF
n
mustvanish. Thoserules anbeextendedtoinitialstatepartonsandmassivepartonsaswell.
Now,sin eweknowthat IRsingularities an el,thereremainstheproblem of
te h-ni al nature, whi h however is of great importan e. Namely, both orre tions
σ
R
and
σ
V
areintegratedoverdierentphasespa eswithdierentjetfun tions. Analyti al
al- ulations arehereextremelydi ult andimpra ti al, thus oneoftenusesMonte Carlo
methods. The problem is now to an elthe singularity that appears during numeri al
integrationin
σ
R
withanalyti alsingularitiesin
σ
V
,e.g.
1/ε
poles.Histori ally the rst method was so alled phase spa e sli ing method. It an be
illustratedbysimplemathemati al example(e.g. [36℄). Suppose wehavethe following
niteexpression
I = lim
κ→0
ˆ
1
0
dx
h (x)
x
1−κ
−
1
κ
h (0)
,
(1.33)where the dependen e on
x
inh
is very ompli ated but su h that the integralexists.Thersttermin urlybra ket orrespondstoareal ontributionregularized
dimension-ally, while the se ond termis the orresponding softpole in virtual orre tion. Both
singularities an elasa tuallytherealvalueoftheintegralis
I =
ˆ
1
0
dx
h (x)
− h (0)
Supposehowever,that wewantto an elthemnumeri ally. Tothisend, wedividethe integration domain
´
1
0
. . . =
´
δ
0
. . . +
´
1
δ
. . .
, withδ
≪ 1
. Sin eh (x)
is regular enough, we anapproximateh (x)
≈ h (0)
forx
∈ [0, δ]
. Then,after simplestepswegetI
≈ h (0) log δ +
ˆ
1
δ
dx
h (x)
x
.
(1.35)Note,thatthesingularities an elledandtheintegral anbenowperformednumeri ally
withremovedregularization, i.e. weset
κ = 0
. Thereis howeveradisadvantageastheresultisapproximate.
Another method, advo ated in thiswork,is thesubtra tion method [36℄. One
on-stru tsanauxiliary rossse tion
σ
sub
=
ˆ
dΦ
n+1
M
sub
n+1
2
F
n
(1.36)whi hmimi sallthesingularitiesof
σ
R
,i.e
σ
sub
= σ
R
inthesingularregionsofPS(note,
there is
F
n
forn
partons). Besidethose pointsof phasespa e it an be anything thathavethepropertiesof a rossse tion. Onthe otherhand, itmust be hosenin su h a
way,thattheanalyti alintegrationoverone-parti lesubspa eispossible. Thatis, ifwe
writePS s hemati allyas
dΦ
n+1
= dΦ
n
⊗ dφ,
(1.37)we must be able to perform
´ dφ
M
sub
n+1
2
analyti ally. It leads then in dimensional
regularizationto poles ofthe form
1/ε
whi h an el those in virtual orre tionsdue tothe KLN theorem. Thepro edure of al ulating NLO ontribution using this method
anbesummarizedasfollows
σ
NLO
= σ
R
− σ
sub
+ σ
V
+ σ
sub
=
ˆ
dΦ
n+1
|M
n+1
|
2
F
n+1
−
M
sub
n+1
2
F
n
+
ˆ
dΦ
n
M
(loop) 2
n
+
ˆ
dφ
M
sub
n+1
2
F
n
.
(1.38)Inthese ondline,duetoIRpropertiesofthejetfun tions,we anperformtheintegration
infourdimensionsanditisnite. Inthethirdlinea an ellationofthepolestakespla e
andafterthat we anset
D = 4
.This methodhasanobviousadvantage,namelyit isexa t. Se ond,alltheintegrals
overone-parti lesubspa ehavetobemadeonlyon e andtheyareuniversal. This an
bealsogeneralizedtohigherorders,wehoweverneedmu hmoresubtra tionterms.
A parti ular hoi efor
σ
sub
isrealizedin [9, 25℄ formasslesspartons,andin [10℄for
massivequarksinthe nalstate(withsomerestri tionsdis ussed in3.1). Thisspe i hoi e is alled dipole subtra tion term. A tually, asolid part of this work is devoted
to generalizingthisapproa hto ompletely massive ase, su h that one anpra ti ally
applymassivefa torizationpro eduredes ribedin theChapter2.
Thedipole methodhas,however,alsosomedrawba ks. First,itisrelatively
ompli- ated,asweshallsee. Moreover,itisunlikelytobegeneralizedeasily tohigherorders.
Thereasonisthatitoperatesontheamplitudessquaredandthenumberofsubtra tion
whi ha tually on ernsthesubtra tionpro edureingeneral,isthatofnumeri alnature.
Namely, depending on implementation, there may be someproblems when performing
theintegrationinthese ondlineof(1.38). Thusee tively,onemaybefor edto usea supportin aformofasli ing-likemethod.
1.3 Quark masses in QCD
Inthepreviousse tionswedidnotpayspe ialattentionto thequarkmasses. Here we
re allsomebasi fa ts onne tedwiththeir in lusionin perturbative al ulations. The
followingmaterialisessentialto thewhole work. Insomepartswerelyon[13℄.
Todayweknowsixavoursofquarkswiththefollowingmasses
3
[43℄:
m
u
= 1.7
-3.1 MeV,
m
d
= 4.1
-5.7 MeV,
m
s
≈ 100 MeV,
(1.39)m
c
≈ 1.29 GeV, m
b
≈ 4.19 GeV, m
t
≈ 172.9 GeV.
(1.40) Re allnow,thatthebasi requirementtobeinaperturbativeregime,isthatthetypi alenergys ale,say
Q
, satisesQ
≫ Λ
QCD
. Sin eΛ
QCD
≈ 200 MeV
we ansafelynegle tthemasses of
u
,d
,s
quarksin perturbative al ulations. Ifthes aleishighenough,weanalsomakesu hanapproximationwiththeotherquarks.
On the other hand, aprioriwe do notknownif there exist heavierquarks. Similar
situationusedtobebeforethedis overyofthetopquark. Thus,thequestionwasabout
therelevan eofeldtheoreti al ulation,wheresomeofthequarksarepossiblymissed.
Thesolutiontothisproblemisformulatedbymeansofso alledde ouplingtheorem[4℄.
Itstatesthat for aFeynmanamplitudewith atypi almomentum s ale
Q
we andropall thediagrams with quark mass
m
≫ Q
, doing errorO (Q/m)
. Let us now assume,thattheremainingnumberofquarkavoursis
N
f
,thusalltherenormalizedparameters(masses, ouplings et .) in su h an ee tive theory are al ulated using this number.
Ingeneral,therenormalizedparametersin theee tivetheorywith
N
f
+ 1
avoursaredierent.
Theproblemhoweverarises,whenthemassesarenotextremelydierent,asa tually
happensfor harmandbottomquarks.Forinstan e,whenthes aleis loseto
m
c
,we anmakeamistakeof theorder
m
c
/m
b
≈ 30%
(foranexamplesee e.g. [13℄). Fortunately,there is a better method than su h an un ontrolled de oupling. It redu es to thelast
in thelimitof verylarge masses. It is aspe ial renormalizations hemeexisting in the
literatureas CWZ (Collins-Wil zek-Zee)renormalizations heme[19, 45, 14℄. In order
to dene its basi slet us introdu ean a tive number of quarks
N
a
. It is anumber ofquarkslighterthanthexedexternalenergys ale(note,thatwedonothavetosetthose
masses to zero). TheCWZ s heme onsist in the subs hemes hara terizedby
N
a
. Inea h subs hemetherenormalizationisdonea ordingtothefollowingpoints:
a) thegraphswithinternallines beinga tivearerenormalizedusing
MS
b) thegraphswithat leastoneinternalheavyquarkline(ina tive) arerenormalized
byzero-momentumsubtra tion
) massesofheavyquarksareusually denedasthepolemasses
3
Asthefreequarkstatesareunobservable,thesearejustparametersobtainedin
MS
s hemeats ale about2 GeV
.Thiss hemepossessesseveralimportantproperties(seee.g. [13℄). Forustwoofthemare
themostimportant. Firstisthatitsatisesmanifestde oupling. Thatis,iftheexternal
s aleismu hsmallerthanthemassesofina tivequarks,therenormalizedparametersof
asubs hemewith
N
a
a tiveavoursarethesameasinee tivetheory withN
f
= N
a
.Hen e we anjust drop all the diagrams with ina tive quarks. The se ond important
propertyisthattheevolutionoftherenormalizedparametersinea hsubs hemeisexa tly
thesameasin
MS
withN
f
= N
a
,inparti ular theevolutionkernelsaremassless.This last property isof great importan e in thisthesis. Aswehaveseenin Se tion
1.1,theoperationaldenitionofpartondistributionfun tionsin ludesarenormalization s heme. Sin ewearegoingtotreatfa torizationwiththeheavyquarksitis onvenient
todene PDFsin CWZs heme. Then, duetothese ond property,su h PDFsundergo
thestandardDGLAPevolutionequationinea hsubs heme. Weshalldis ussitindetails
in Se tion2.4,whilein Se tion4.2we al ulatesomeofthem inthiss heme.
Thereisonemore ommentinorder. Thepurposeofintrodu ingsu has heme,isto
beabletoevolveagivenparameterthroughallappli ables aleswithoutloosinga ura y.
Itis realizedbyswit hing thes hemesatgivenswit hingpoints. Therefore,wehaveto
state amat hing onditionsat those points
4
in order to have astartingparameters in
evolution. Su h onditionswereobtainedevenupto threeloopsforthe oupling(using
ee tivetheoryformalism[12℄)anduptotwoloopsforPDFs[8℄.
In theend, letus introdu esomemorenotationwe shallusethroughout. First,we
oftenneedtodistinguishbetweenheavyandlightavours. Thuswedene
N
f
= N
q
+N
Q
,where
q
is ageneri lightquark, whileQ
orrespondsto heavy quarks. Sometimes wereferto lightpartons number,whi h issimply
N
l
= N
q
+ 1
,asgluonis alwayslight. Ifwewantto referto allthe quarkavours,but in ludinggluon, we usethesymbol
N
′
f
. Forallthedenedsymbols,weintrodu ethesets, ontaining orrespondingavoursandtheiranti-avours. Thesetsshallbedenotedbybla kboardfont,forinstan e
N
f
,N
l
et .4
Ingeneral,oneshoulddistinguishbetweentheswit hingpointandamat hingpoint. Therstisthe
In lusive DIS with heavy quarks
2.1 Introdu tion
As we haveseenin Se tion 1.3, there are ertainlysome ompli ationswhen there are heavy quarkswith masses that are neither marginally largenor negligibly small. The
problemsareevenmoreevidentin thepro esseswhi h requirefa torization. InSe tion
1.1were alledthefa torizationtheoremassumingthatthemasses anbenegle ted. In ase,when they annot, su h atreatmentis obviouslyveryina urate. Inthis hapter,
weshallanalysethisissueinmoredetailsinthe ontextofin lusiveDISs attering.
First,inthenextse tionwere allthesimplestpossiblewayofin ludingheavyquarks,
a tuallytreatingthemasmasslesspartons. Thiss heme,often alledzero-massvariable
avournumbers heme(ZM-VFNS) ismostoftenused in phenomenologi alanalysis of
DIS pro esses. However, as weshall see, it is ina urate in non-asymptoti regions of
energy s ale. That se tion is also devoted to introdu ing some notationwhi h weuse
in thisandthenext hapters. Further,inSe tion 2.3webriey des ribemorea urate treatment,howeveraimingat ompletelydierentkinemati regimethanthelatter. This
se ond solution is oftenreferred to asxed-avour numbers heme (FFNS), and takes
alltheee ts ofheavyquarksintoa ount. Theproblemishowever,that astheenergy
s alein reases,su hapredi tionbe omeslessa urate,unlesswegotohigherordersof
perturbation theory. Needless to say, su h a massive high-order al ulationsare mu h
moreinvolvedandtime- onsumingthanthemasslessones,nottomentiongeneralizations
toex lusivepro esses.
Therefore,itisdesirabletohaveas hemewhi hisappli ableatintermediateenergy
s alesand ontainsbothaboves hemesasalimiting ases. Su hsolutionswere indeed
developed[2,50, 8℄, howeverwithexpli it treatment ofin lusive pro essesonly. What
isworthemphasizing,theapproa h itedas[2℄wasprovedto allordersofperturbation
theory[15℄. Weshallbrieydes ribethisapproa h,referredtoasACOT
(Aivazis-Collins-Olness-Tung) s heme, in Se tion 2.4. It is based on CWZ renormalizations heme for partondensitiesand an beeasilygeneralizedtoanotherIR safe rossse tions.
Forashortreviewofthementionedtreatmentsofheavyquarkprodu tioninin lusive
DISseee.g. [51,52℄.
We remark,that although this hapter is onsidered to be introdu tory, wedis uss
2.2 Zero-mass variable avour number s heme
Letusstartbydeningourobje tofinterestin this hapter. Weshallbe on entrated
heremainly onthestru turefun tions parametrizingthe rossse tionforin lusiveDIS
pro esses, notably
F
2
x
B
, Q
2
, and its dependen e on the photon virtuality
Q
2
.
Re-member,that thestru turefun tions areobtainedbymeansof asuitableproje tionof
hadroni tensor
W
µν
,denedasusualintermsofmatrixelementofele troweak urrents
sandwi hedbetweenhadronstates
W
µν
q, P ; x
B
, Q
2
=
1
4π
X
spin
X
P
x
ˆ
dΦ
1
(q, P ; P
X
)
P
j
† µ
(0)
P
X
hP
X
|j
ν
(0)
| P i ,
(2.1)wherethese ondsumgoesoverallnalstates
P
X
. Theproje tionismadeusingsuitablebasetensorsmadeoftheve tors
P
,q
andthemetri tensor. Negle tingthehadronmasswegetfor
F
2
F
2
x
B
, Q
2
=
2 x
B
D
− 2
−W
µ
µ
+ (D
− 1)
2 x
B
P
· q
W
µν
P
µ
P
ν
.
(2.2)If we repla e the hadroni state by a parton, su h a tensor is alled the partoni
tensor. We shall denote it as
w
µν
. Both tensors are related by meansof fa torization
theorem we shall give some examplesbelow. We do not give further details related
to otherstru ture fun tions and relatedissues astheyare allstandard(forthe pre ise
denitions in orporatingquarkand target masses see [3℄). Su h limited onsiderations
are ompletely enoughtoelu idate thebasi problemswith heavyquarkmasses,aswe
shallsee.
Beforewepro eed,letusre all,thatwedenoteageneri heavyquarkbysymbol
Q
.Thelightquarksaredenoted as
q
,there should beno onfusion sin ethis isonly usedinthismeaningasasubs ript.
Letusstartfurther onsiderationsbynoting,that thesimplestpossibleapproa hto
heavy quarksis when
Q
2
→ ∞
withx
B
xed, su h that all the existing heavyquark masses an be negle ted. Then, the pre ise predi tions are given by the fa torizationtheorem (1.20), whi h is exa t. All the quarks (in luding heavy quarks) are treated asmassless partons having orresponding PDFs. Su h situation is obviously not very
plausible. Inpra ti etheenergys alesdonottendtoinnity,moreovermanyinteresting
phenomenaexistatlowers ales. Se ondly,wehaveseveralheavyquarkswithlargemass
splittings,asdis ussedinSe tion1.3. Ontheotherhand,when
Q
2
ismu hsmallerthan
themassofagivenheavyquark,itmaybedroppedfrom al ulationsduetode oupling
theoremmentionedalsoinSe tion1.3.
These twomarginally dierent situations (
Q
2
≫ m
2
Q
andm
2
Q
≫ Q
2
)motivate the followingsimplests hemeoftreatingheavy quarks:a) ompletelyde ouplegivenheavyquark
Q
whenm
2
Q
> Q
2
,i.e. treatitasinnitely
heavy
b) treat
Q
asamasslesspartonwithasso iatedPDF, whenQ
2
> m
2
Q
Wehaveassumedherethat thefa torizationandrenormalizations alesareequalto
Q
.Ifthereareseveralheavyquarks,wehavethe omposites heme,withsubs hemes
har-a terizedbyana tivenumberofavours
N
a
. Thuswehaveasetofpartondistributionfun tions
f
(N
a
)
a
and ouplingsα
(N
a
)
s heme mentioned earlier, in whi h PDFs are dened. All the masses are however set
tozero. Sin eCWZsatisesmanifest de oupling,wejustdroptheina tivequarks,and
ea hsubs hemeisee tivelya
MS
s hemewithN
a
avoursand orrespondingDGLAPmasslessevolutionof PDFs. As already mentioned, the s hemes with dierent
N
a
area tually dierent renormalization s hemes, they dier by nite terms and a relation
betweens hemeswith
N
a
andN
a
+ 1
avours anbestated.Here theswit hing pointis usually hosentobe
µ
th
= m
Q
, whereQ
is(N
a
+ 1)
-thavour. It is onvenient, sin e then the heavy quark density
f
(N
a
+1)
Q
is zero at the threshold1
. Itfollowsfromtwofa ts. First isjustapre iseformoftherelationbetween
PDFs in two subs hemes [18℄. Se ond is that below
µ
th
it is suppressed by power ofΛ
QCD
/m
Q
due tode ouplingtheorem. Thuswehavethe ontinuity onditionf
(N
a
)
Q
z, µ
2
th
= f
(N
a
+1)
Q
z, µ
2
th
= 0
(2.3)Then, abovethe threshold it is evolved using DGLAP equations with
N
a
+ 1
avoursstartingfromzerovalue.
Asalreadymentionedintheintrodu tion,su has hemeis alledzero-massvariable
avournumbers heme(ZM-VFNS).Correspondingfa torizationtheoremtakestheform
W
(N
a
)
µν
q, P ; x
B
, Q
2
=
X
a∈N
a
f
(N
a
)
a
µ
2
f
⊗ ˆ
w
(N
a
)
µν
q, p
a
;
Q
2
µ
2
f
!
,
(2.4)whereweexpli itlydenotedthedependen eonthefa torizations ale(equalheretothe
renormalizations ale). Wealsointrodu edthe onvolutionsymbol,whi hsimpliesthe
notation;itisdenedhereas
f
⊗ w =
ˆ
1
x
B
dξ
ξ
f (ξ) w
x
B
ξ
.
(2.5)In (2.4)
p
a
= ξP
, nevertheless we leavep
a
as this notation is more general. As we vary the s ale, thea tivenumberof partons hanges. Su h aformula isa tually validupto orre tions oforder
O m
2
N
a
/Q
2
, where
m
N
a
would bethemass ofthe heaviest a tive quark, if we did not set it to zero. Therefore, in reality su h an approa h isunreliablefor
Q
2
aroundthemasses ofheavyquarks. Moreover,aswerea htheregion
of validity of (2.4) for one heavy quark, say harm, we simultaneously an enter the regionof inappli ability forthebeautyquark. Thus, onlyat really asymptoti regimes
thiss hemeis orre t,asweremarkedearlier.
To illustrate this approa h, onsider now a al ulation of
F
2
stru ture fun tion inthiss hemeuptoorder
α
s
. Letusassumeweworkin thes hemewithN
a
= 4
,that isbesidesgluon,
u, d
ands
quarks,whi harealwaysmassless,wehavealso harmc
N
a
=
g, u, u, d, d, s, s, c, c
.
(2.6)Then,tothis order
1
x
B
F
2
x
B
, Q
2
=
X
a∈N
a
f
a
µ
2
f
⊗
"
C
a
(0)
Q
2
µ
2
f
!
+ C
a
(1)
Q
2
µ
2
f
!#
.
(2.7) 1A) B)
Figure 2.1: A)Feynmandiagrams ontributing to stru turefun tions in ZM-VFNS up
toorder
α
1
s
;B)ThesameforFFNS.Thereisonlyboson-gluonfusionatorderα
1
s
. Thi k line orrespondstoaheavyquark.Wearea tuallyinterestedin harm ontributionto
F
2
,whi h anbepi kedupfromtheaboveequation. Setting
µ
2
f
= Q
2
(andthesameforrenormalizations ale)wehaveF
2
c
x
B
, Q
2
=
f
c
Q
2
+ f
c
Q
2
⊗
h
C
c
(0)
+ C
c
(1)
i
+ f
g
Q
2
⊗ C
(1)
g
,
(2.8)where theresultfor oe ients
C
i
in masslessMS
s hemeis well known (e.g. [28℄, fororrespondingdiagramsseeFig.2.1A)andreads
C
c
(0)
(z) = e
2
c
zδ (1
− z) ,
(2.9)C
c
(1)
(z) = e
2
c
α
s
2π
C
F
1 + z
1
− z
log
1
− z
z
−
3
4
+
1
4
(9 + 5z)
+
,
(2.10)C
g
(1)
(z) = e
2
c
α
s
2π
2P
gq
(z) log
1
− z
z
+ 8z (1
− z) − 1
.
(2.11)Thesplitting fun tion
P
gq
and plusdistribution weredened inSe tion1.1.Thebehaviourof thissolutionwill be expli itlydemonstratedin Se tion 2.4, where wepresentsomeplots omparingthisNLO al ulationtoothers hemes.
Although aswe havejust seen su h a s heme is very simplied, it is still most
ommonlyused in PDFsglobal tsto data (e.g. CTEQ ts [38℄and earlier). Its great
advantageis simpli ity and pra ti ality. It should bealso mentionedthat it wasvery
su essfulindes ribinglargeamountofmodernhighenergydata.
2.3 Fixed avour number s heme
Letus now present anotherapproa h,whi h is appli ablewhen
Q
2
isabout theheavy
quark mass
m
2
Q
. A tually, it is a generalization of the previous s heme, whereQ
is ina tive, but has nite mass. Thus we haveN
a
masslesspartons undergoing masslessevolutionandoneheavyavour,whi h anbeonlyprodu eddynami ally. Forexample,
atLOinDISitis theboson-gluonfusion(BGF)pro essdepi tedinFig.2.1B. Thefa torizationtheoreminthis asetakestheform
W
µν
q, P ; x
B
, Q
2
, m
2
Q
=
X
a∈N
a
f
a
µ
2
f
⊗ ˆ
w
µν
q, p
a
;
Q
2
µ
2
f
,
m
2
Q
µ
2
f
!
+
O
Λ
2
QCD
m
2
Q
!
.
(2.12)Here,thehard s aleisgiven bytheheavyquarkmassand onthe ontraryto (2.4)
N
a
doesnot hange. Thereforesu has hemeis alledxedavournumbers heme(FFNS)and was pioneered in [28, 27, 37, 29℄. We note, that here the onvolution symbol is
dened as in (2.5)but the integrationlimitsdepend onquarkmasses (weshall seethe examplebelow).
In order to dis uss someof itsproperties, letus again onsider the expli it result,
namelythe ontributionto
F
2
omingfromthe harmquark. Asalreadymentioned,thesituationwhere harm onne tsthehardandsoftpartsissuppressedby
Λ
2
QCD
/m
2
c
,thus theperturbative al ulationforF
c
2
startsatα
1
s
withBGFpro ess(Fig. 2.1)1
x
B
F
c
2
x
B
, Q
2
= f
g
Q
2
⊗ C
(1)
g
m
2
c
Q
2
,
(2.13)with oe ientgivenby(e.g. [47℄)
C
(1)
g
(z, ρ) = e
2
c
α
s
2π
(
2P
gq
(z) + 4ρz
2
(1
− 3z) − 8ρ
2
z
3
log
1 + v
1
− v
+ (4z (1
− z) (2 − ρ) − 1) z v
)
,
(2.14) where we abbreviatedρ = m
2
c
/Q
2
andv =
p
1
− 4ρz/ (1 − z)
is the velo ity of the harm quarkin thephoton-gluonCM frame. Nowthelowerlimitonthe onvolutionisz
min
= x
B
(1 + 4ρ)
.Letusdis ussnowthisresult. First,letusnotethatit ontainsthepowersof
m
2
c
/Q
2
, whi harela kinginZM-VFNStreatment(highertwists). Thereforeindeeditisreliableal ulationwhen
Q
2
isoftheorderof
m
2
c
. Now,thequestioniswhatisthebehaviourof thissolutionwhenthes aleismu hlarger. Inthis ase,wendthatC
g
(1)
(z, ρ) = e
2
c
α
s
2π
P
gq
(z) log ρ +
O (ρ) .
(2.15)Thus we see, that wehave apotentiallylarge logarithm of the heavy quarkmass and
thehards aleratio. Su hlogarithmsappearineverynextorderofperturbationtheory,
typi ally
C
a
(m)
=
m
X
k=0
c
(m)
a, k
log
k
ρ,
(2.16)what makes su h an expansion unreliable. The solution is to resum all the powersof
α
s
in front ofthe givenpowerof logarithm,i.e. to suitably rearrangetheaboveseries. Then, we a tuallyarriveat thezero-masss heme with harm beingamasslessparton.However, onehas to bear in mind that it happens at apri e of loosing ontrol of the
terms
O m
2
c
/Q
2
(a tually,ifwedonottra khighertwistterms,whi hisnoteasyand
sofarhasnot beensolved). In thenextse tionweshallpresentsomeplots omparing
thiss hemeto ZM-VFNS.
There isonemore ommentin order. One anaskwhenthis xedavourapproa h
fails, sin e logarithm is a veryslowly in reasing fun tion. In [29℄ it was argued, that
the rossse tions al ulatedinthisapproa hatNLOarestableevenforrelativelylarge
s ales,howeveronehastouseaspe ialsets ofPDFs,namelyso alleddynami alPDFs
2.4 ACOT s heme
As wealreadyanti ipated in theintrodu tion,there exist solutionswhi h ontain
ZM-VFNS and FFNS asaspe ial ases. In thefollowingse tion we des ribeone ofthem,
theso alledACOTs heme[2,15℄. Webelieveitisthebest solutionthat anbeeasily
generalized to less in lusive pro esses, in parti ular jets. This is in ontrast to other
approa hes like [50, 8℄. As we havealready remarked,it has been proved forin lusive
DIStoallordersin[15℄.
Basi assumption ofthe s hemeis that thePDFs aredened using CWZ
renormal-izations hemeand thatthemassesrelevantto a tualenergys alearekeptnite. This
resultsinthehighertwisterrorsoftheorder of
Λ
2
QCD
/Q
2
overthewhole kinemati ally allowedregionofQ
2
. Weshallseehowitworksinpra ti ebelow.
Consideragainthehadroni tensor
W
µν
andsupposeforsimpli itythatthereisonly
oneheavyquark
Q
. Thefa torizationisrealiseda tuallybytwodierenttheorems[15℄.Therstoneis essentiallythesameas(2.12),i.e. itisappli ablewhen
Q
2
.
m
2
Q
. The se ond one, is whenQ
2
&
m
2
Q
, that is both theorems have anoverlap region. Let us analysethese ond ase. Thetheoremunder onsiderationhasthefollowingformW
µν
q, P ; x
B
, Q
2
, m
2
Q
=
X
a∈N
a
f
a
µ
2
f
⊗ ˆ
w
µν
q, p
a
;
Q
2
µ
2
f
,
m
2
Q
µ
2
f
!
+
O
Λ
2
QCD
Q
2
!
.
(2.17)Super ially itisalmostthesameas(2.12),howevertherearedieren es. Firstis that in thiss heme (i.e. abovesomeswit hing point
µ
th
∼ m
Q
)theset of a tivequarksN
a
does in lude the quark
Q
. Se ond dieren e is subtle. It is onne ted with IR nitepartoni tensor. Toseethis letus al ulate itto therstorderin
α
s
. Re all, thatit isdonewithfa torization(2.17),butonthepartoni level(letussetallthes alesequalto
Q
)w
µν
a
q, p
a
; Q
2
, m
2
Q
=
X
b∈N
a
F
CWZ
ab
Q
2
m
2
Q
!
⊗ ˆ
w
µν
b
q, p
b
;
m
2
Q
Q
2
!
.
(2.18)Wedenoted thatthe partondensitiesinside aparton arerenormalizedusing
CWZ
. Tothe rst order it be omes (below wedrop all the arguments, ve tor indi es and CWZ
supers riptfortransparen y)
w
(0)
a
+ w
a
(1)
=
X
b∈N
a
F
ab
(0)
+
F
(1)
ab
⊗
w
ˆ
(0)
b
+ ˆ
w
b
(1)
+
O α
2
s
.
(2.19)Thus,thezerothorderpartoni tensorisIR safe
w
a
(0)
= ˆ
w
a
(0)
.
(2.20)Solvingfurtherthere urren ewegetforalightquark
w
q
(1)
=
F
(1)
⊗ w
(0)
q
+ ˆ
w
q
(1)
,
(2.21) forheavyquarkw
(1)
Q
=
F
(1)
⊗ w
(0)
Q
+ ˆ
w
(1)
Q
(2.22)andforagluon