Master thesis
Higher-order corrections for squark-gaugino production at the LHC
Lasse Kiesow∗
Münster, August 12, 2017
Supervisor & first examiner: Prof. Dr. M. Klasen Second examiner: Dr. K. Kovařík Time allowed for completion: 13.02.17 bis 14.08.17
Contents
1. Introduction1
2. Supersymmetry3 2.1. Motivation...... 3 2.2. SUSY Algebra...... 4 2.3. Soft SUSY breaking...... 5 2.4. The MSSM...... 6 2.4.1. Gaugino sector...... 8 2.4.2. Squark sector...... 10
3. Virtual corrections and UV-divergences 11 3.1. Dimensional regularization...... 11 3.1.1. Passarino-Veltman reduction...... 12 3.1.2. UV divergent parts...... 17
4. Renormalization 19 4.1. Renormalization of quarks...... 20 4.2. Renormalization of squarks...... 21
5. Squark-gaugino production 25 5.1. Leading order contributions...... 25 5.1.1. Analytical results...... 26 5.1.2. Implementation in Resummino...... 29 5.1.3. Numerical results...... 31 5.2. Next-to-leading order contributions...... 36 5.2.1. Self-energy corrections...... 36 5.2.2. Implementation in Resummino...... 40 5.2.3. Vertex corrections...... 41 5.2.4. Box diagrams...... 44 5.2.5. Preliminary results...... 44
6. Summary and outlook 47
A. Feynman rules 49
B. Self-energies 51 B.1. Quark self-energy...... 51 B.2. Squark self-energy...... 54
C. Vertex corrections 57 C.1. Quark-gluon-quark vertex...... 57
iii Contents
C.2. Squark-gluon-squark vertex...... 60 C.3. Quark-squark-gaugino vertex...... 64
D. Box corrections 67 D.1. Gluon-quark-quark-squark...... 67 D.2. Gluino-squark-squark-quark...... 68 D.3. Quark-gluon-gluon-squark...... 68 D.4. Squark-gluino-gluino-quark...... 69 D.5. Quark-squark-squark-gluon...... 70 D.6. Squark-quark-quark-gluino...... 71
iv 1. Introduction
Since the discovery of the electron by J. J. Thomson in 1897, physicists have discovered countless other particles. Especially with the increasing power of particle accelerators the number of discovered particles grew so much that physicists would speak of an entire particle-zoo. They methodized this zoo by sorting the particles into composite particles (hadrons) and elementary particles. The later are further subdivided into fermions (matter particles) and bosons (particles that mediate forces between the matter particles). Over the years a theoretical model was developed that successfully describes all the so far observed particles and their interactions: the Standard Model (SM) of particle physics. Its predictions have been tested over many years and are in astonishing agreement with experiments. With the discovery of a Higgs boson at the Large Hadron Collider (LHC) in 2012 the last missing piece of the SM has been found. However, the SM cannot account for everything observed in the Universe. For example Dark Matter and neutrino masses, both known to exist due to experiments, are nowhere to be found in the SM. That is why physicists have started to consider models that go beyond the SM. One of the most famous and widely discussed ones is Supersymmetry (SUSY). It introduces a new symmetry between fermions and bosons and can solve many of the problems the SM suffers from (more on this in Chapter2). The simplest supersymmetric extension of the SM is called the Minimal Supersymmetric Standard Model (MSSM). It extends the particle content of the SM by so called superpartners which differ by spin 1/2 from their SM particle counterparts. For example the superpartners of the quarks are the squarks (scalar quarks) and the ones of the gluons are the gluinos. Searches for SUSY particles are undertaken at the LHC. Such searches are mostly based on the production of one hard jet and missing transverse energy1. For example searches based on jet production from squark and gluino decays with two associated lightest stable supersymmetric particles (LSP) ( missing energy) put constraints on the parameter space of the MSSM [2,3]. The main→ production mode in the MSSM for a jet and missing energy are squark and gluino pair production. Both are mediated by the strong interaction and have been studied in e.g. [4]. Extracting any model parameters other than the masses of the new particles from the analysis of these two processes is, however, difficult [5]. For processes which also involve the weakly interacting sector of SUSY it turns out that more information can be extracted, e.g. information about the composition of the dark matter candidate [6]. One such process is the associated production of a squark with a chargino or neutralino (together referred to as gaugino), which is also the process studied in this thesis2. The main goal of this thesis is to make precise predictions for the squark-gaugino production at the LHC. The necessary theoretical background is given in the next three chapters. First, in Chapter2 some important aspects of SUSY and the MSSM are discussed.
1Note that this could also be a sign for other theories beyond the Standard Model such as large extra dimensions [1] 2The other process is the associated production of a gluino with a chargino or neutralino.
1 1. Introduction
Chapters3 and4 cover the necessary theoretical background for the extensive next-to- leading order (NLO) calculations, i.e. regularization and renormalization. In the first part of Chapter5 the analytical results for squark-gaugino production are calculated at leading order (LO). These results are implemented in Resummino, a program first written by Jonathan Debove [7] and further developed by Marcel Rothering and David Lamprea [8]. In the second part the (virtual) NLO calculations are performed and their implementation in Resummino is described. Since the calculations of all the self-energies, vertex, and box corrections are quite lengthy they are shown in AppendicesB toD. Chapter6 gives a summary and an outlook.
2 2. Supersymmetry
One of the most successful ways to understand the laws of nature at the level of particle physics is the study of symmetries. This is shown strikingly by the Standard Model (SM) of particle physics, which can describe three of the four fundamental forces with remarkable precision using external space-time and internal gauge symmetries. There are, however, some unsatisfactory aspects in the SM like the hierarchy problem. Moreover, there are some phenomena for which the SM cannot provide any kind of explanation like dark matter. All these problems might be solved by introducing an additional (and as it happens the last possible) symmetry – supersymmetry (SUSY) – which relates fermionic and bosonic degrees of freedom. Although the idea of a symmetry between fermions and bosons was first introduced in connection with string theory, it turned out that this idea applied to field theories can solve most if not all of the problems the SM suffers from. Some of the most important and pressing problems of the SM and the solutions provided by SUSY are outlined in the next section. After that, some of the algebraic structures of a SUSY theory are introduced. The chapter ends with the discussion of the Minimal Supersymmetric Standard Model (MSSM), which is one of the most important SUSY extensions of the SM and also the model of interest for this thesis.
2.1. Motivation
Grand Unification One of the unsatisfying fea- tures of the SM is that it is not possible to unify all three forces (not to mention that it does not include gravity). This can be seen in Fig. 2.1 where the dashed lines show the two-loop renor- malization group evolution of the inverse gauge couplings in the SM. As one can see in the figure these lines have three distinct intersection points and so the couplings do not unify. The MSSM on the other hand includes additional particles and therefore additional loop corrections which modify the running coupling. As the solid lines show, this modification is such that the cou- 16 plings unify at a scale MGUT 2 10 GeV. Figure 2.1. – Two-loop running of inverse The unification is not perfect, but≈ the· small dif- gauge couplings in the SM (dashed lines) and MSSM (solid lines) [9]. ference in the value of the couplings at MGUT can be explained by threshold corrections due to new particles that exist near MGUT. More details on this can be found in e.g. [9]. It
3 2. Supersymmetry should also be noted that SUSY was not formulated with the aim of unification, so this result is indeed remarkable.
Dark Matter It is known from observations – e.g. the rotation curves of spiral galaxies, the cosmic microwave background radiation and gravitational lensing – that ordinary matter constitutes only about 4% of the universe’s total energy. The rest consists of 24% dark matter and 72% dark energy (see e.g. [10] for more details) Whereas the nature of dark energy is as of yet completely unknown, there are several theories concerning dark matter. A popular one is that dark matter consists of so called WIMPs, i.e. weakly interacting massive particles. However, no suitable particle can be found in the SM. The MSSM on the other hand introduces a new superpartner particle to each particle of the SM and the lightest supersymmetric particle (LSP) – in most scenarios of the MSSM this 0 is the neutralino χ˜1 – turns out to be a suitable WIMP candidate.
The hierarchy problem This is a general problem of quantum field theories with scalar particles. In the SM the scalar particle is the Higgs boson. Its mass is subject to radiative corrections due to the fermion-loop depicted on the left of Fig. 2.2. The calculation of this self-energy leads to a divergent integral that can be regularized using the cut-off method. The radiative corrections are then proportional to the square of this cut-off parameter. Assuming that the SM is valid as long as quantum gravity effects can be neglected, the 19 cut-off parameter can be identified as the Planck scale ΛPlanck 10 GeV. ∼ During renormalization these corrections are cancelled by counterterms, but in order to arrive at the known Higgs mass of about 125 GeV a very unnatural fine-tuning of these counterterms would be needed at every order of perturbation theory. SUSY offers a more natural solution. It introduces bosonic superpartners to the fermions in the loop and thus leads to two additional diagrams shown in the middle and on the right of Fig. 2.2. The contributions from these additional diagrams exactly cancel the 2 ΛPlanck dependence and only a logarithmic dependence remains.
F ermion Sfermion Sfermion
Higgs Higgs Higgs Higgs Higgs Higgs
Figure 2.2. – Corrections to the Higgs mass: Fermionic SM correction on the left and sfermionic SUSY corrections in the middle and on the right.
2.2. SUSY Algebra
The SUSY algebra extends the Poincare algebra given by
[Pµ,Pν] = 0, (2.2.1)
[Mµν,Pσ] = i(gνσPµ gµσPν), (2.2.2) − [Mµν,Mσρ] = i(gµρMνσ + gνσMµρ gµσMνρ gνρMµσ), (2.2.3) − −
4 2.3. Soft SUSY breaking where P is the generator of space-time translations and M the generator of Lorentz transformations. Note that this algebra includes only bosonic operators. Extending the Poincare group is restricted by the Coleman-Mandula theorem [11] and the only possible way to extend it is the inclusion of fermionic operators as shown by the Haag-Lapuszanski- Solnius theorem [12]. This is exactly what SUSY does. It introduces one1 additional fermionic operator Q that can be written in terms of two Weyl components Qa, where a = 1, 2. It is also convenient to introduce dotted indices, e.g. a˙, which stand for the (0, 1/2) representation of the homogeneous Lorentz group, whereas undotted ones stand for † the (1/2, 0) representation. The generators of both representations are linked via Q¯a˙ = Qa. With the notations
σµ = (1, σi) and σ¯µ = (1, σi), (2.2.4) − i i σµν = (σµσ¯ν σνσ¯µ) and σ¯µν = (¯σµσν σ¯νσµ) (2.2.5) 2 − 2 − where σi are the usual Pauli matrices, the SUSY algebra can be compactly written as
µ Qa, Q¯ ˙ = 2(σ ) ˙ Pµ, (2.2.6) { b} ab a˙ b˙ Qa,Qb = Q¯ , Q¯ = 0, (2.2.7) { } { } 1 b [Mµν,Qa] = (σµν) Qb, (2.2.8) − 2 a a˙ [Qa,Pµ] = [Q¯ ,Pµ] = 0. (2.2.9)
2.3. Soft SUSY breaking
The SUSY transformations introduced in the last chapter change only the spin of the particles. All other quantum numbers and in particular the masses are unchanged. However, if that were the case, i.e. if SUSY were an exact symmetry, supersymmetric particles like e.g. the selectron would have been detected long ago. Rather than dismissing SUSY right away at this point due to lack of experimental evidence, one can consider broken SUSY. The symmetry breaking leads to SUSY particles with higher masses than their SM counterparts. Of course, one would like to keep the benefits of SUSY described in Section 2.1. This can be done by considering softly broken SUSY, where soft means that no additional divergences are introduced, i.e. that the theory is still renormalizable. However, since it is unknown whether SUSY is realized in nature at all, the exact breaking mechanism is also unknown. The two most popular theories are gravity- and gauge-mediated SUSY breaking [13, 14]. Both assume that a hidden sector exists which contains much heavier particles than the MSSM which somehow interact with the MSSM particles. In case of gravity-mediated SUSY breaking this interaction takes place via gravity, whereas in gauge-mediated SUSY breaking new particles are believed to exist that occur in self-energy corrections of sparticles. In the end both theories lead to soft SUSY breaking terms in the Lagrangian of the MSSM which is then often written as
L MSSM = L SUSY + L soft, (2.3.1)
1In this case one speaks of simple (N = 1) SUSY. It is also possible to introduce more than one new operator, which leads to extended (N > 1) SUSY.
5 2. Supersymmetry
where L SUSY is the SUSY conserving part containing Yukawa and gauge interactions, whereas L soft contains the new SUSY violating mass and interaction terms (trlinear couplings). The mass dimension of the coupling has to be positive in order to retain a renormalizable Lagrangian. The problem that the exact breaking mechanism is unknown is eluded by adding all possible terms that conserve gauge invariance and renormalizability in L soft. Unfortunately, this leads to 105 new parameters which are in general independent. Although some of them are constrained by experiments (e.g. by minimal flavor or CP violation), there is still a huge amount of arbitrariness and the theory has no predictive power [15]. Therefore, several phenomenological models were devised which lower the number of free parameters to a manageable amount by making certain assumptions. An important example is the pMSSM which reduces the amount of free parameters to 19 by three assumptions: i) no new sources of CP violation, ii) no flavor changing neutral currents, iii) first and second generation universality. More details on this can be found in [16]. The mass of the lightest sparticle can be roughly estimated by considering the largest mass parameter msoft in L soft. The terms in L soft including msoft lead to an additional 2 contribution to the Higgs self-energy which is proportional to msoft. In order to still solve the hierarchy problem one must have msoft . 1 TeV, i.e. msoft must not be much larger than the electroweak scale because otherwise the Higgs mass would again be subject to large radiative corrections. Since the mass splitting between sparticles and particles is determined by the mass parameters in L soft one can infer that at least the lightest sparticle should have a mass of about 1 TeV or smaller. ∼
2.4. The MSSM
The Minimal Supersymmetric Standard Model is a supersymmetrized version of the SM with as little new particles as possible, i.e. each particle of the SM gets an associated superpartner differing by spin 1/2 and after the symmetry breaking also in mass. In short, the gauge bosons g, B, W1,W2,W3 get fermionic superpartners g,˜ B,˜ W˜ 1, W˜ 2, W˜ 3 (gluino, bino, winos). For the Higgs boson the introduction of a SUSY partner is a bit more involved. The Higgs boson in the SM stems from a Higgs doublet which contains four degrees of freedom. Three of these render the weak gauge bosons massive and the remaining one manifests itself as the Higgs boson h0. So simply introducing one superpartner to the Higgs boson would not yield a viable SUSY model. Rather one has to introduce a second doublet leading to 8 degrees of freedom. As in the SM, three of these give mass to the weak gauge bosons, leaving five degrees of freedom, which manifest themselves as physical particles: a light h0 and heavy H0 scalar, a pseudoscalar A0 and two charged Higgs bosons H±. Associated with each of these Higgs bosons is a fermionic superpartner, which are called higgsinos. Note that the electrically neutral B˜ and W˜ 1 have the same quantum numbers as the electrically neutral higgsinos and can therefore mix into four mass eigenstates which are 0 ˜ ˜ called neutralinos χ˜i , i = 1,..., 4. Similarly, the charged W2 and W3 mix with the charged ± higgsinos into mass eigenstates called charginos χ˜i , i = 1, 2. Neutralinos and charginos together are commonly referred to as gauginos. Since the gauginos are important for this thesis, the mixings are discussed in more detail in Section 2.4.1.
6 2.4. The MSSM
Table 2.1. – Particle content of the SM and MSSM [17, 10]. SM particles SUSY particles
Symbol Name Symbol Name u, c, t Up quarks u˜i, c˜i, t˜i, where i = L, R Up squarks d, s, b Down quarks d˜i, s˜i, ˜bi where i = L, R Down squarks e, µ, τ Charged leptons e˜i, µ˜i, τ˜i where i = L, R Charged sleptons
νe, νµ, ντ Neutrinos ν˜e, ν˜µ, ν˜τ Sneutrinos g Gluons g˜ Gluinos
± ± W W bosons ± χ˜i , where i = 1, 2 Charginos H± Charged Higgs bosons
γ Photon Z0 Z boson 0 0 h Light scalar Higgs boson χ˜i , where i = 1, 2, 3, 4 Neutralinos H0 Heavy scalar Higgs boson A0 Pseudoscalar Higgs boson
The quarks q in the SM get supersymmetric scalar particles called squarks q˜. Since quarks have two independent chiral components, qL and qR, there are two squarks q˜L and q˜R associated with each quark. Note that the subscripts L and R of the squarks do not label their chirality states (they are spin zero particles and therefore chirality is not defined) but rather refer to the chirality state of the original quark partner. Since squarks are important for this thesis they will be discussed in more detail in the next section. The charged leptons e, µ, τ also get two associated scalar superpartners each, namely e˜L,R, µ˜L,R, τ˜L,R. The neutrinos νe, νµ, ντ get only one superpartner each ν˜e, ν˜µ, ν˜τ since there are only left handed neutrinos in the SM. The particles of the SM and their SUSY partners are summarized in Table 2.1. Apart from new particles the MSSM also gives rise to a new conserved multiplicative quantum number called R-parity. It is defined as
3(B−L)+2S PR = ( 1) , (2.4.1) − where S, B, and L are the spin, baryon and lepton number, respectively. For every SM particle the R-parity is +1, whereas it is 1 for every MSSM particle. If R-parity is conserved a vertex must always contain an even− number of SUSY particles which implies that the lightest supersymmetric particle (LSP) cannot decay and thus has to be stable. As mentioned above the LSP is the lightest neutralino in most MSSM scenarios and a viable dark matter candidate.
7 2. Supersymmetry
2.4.1. Gaugino sector Charginos As mentioned above the charged higgsinos and the charged winos mix and form two charginos. This is described by the Lagrangian
chargino-mass ± L = (ψ−)T M χ˜ ψ+ + h.c., (2.4.2) MSSM −
W˜ 2 W˜ 3 ± where ψ+ = , ψ− = and the chargino mass matrix M χ˜ is given by ˜ + ˜ − H2 ! H1 !
± M2 √2 mW sin β M χ˜ = (2.4.3) √2 mW cos β µ !
± It depends on the W boson mass mW , the wino mass parameter M2, the higgsino mass parameter µ and the mixing angle β. The later is defined via tan β = v2/v1, where v2 and v1 are the two Higgs vacuum expectation values. The mass eigenstates can be found by diagonalizing the chargino mass matrix via
∗ χ˜± −1 (U) M (V ) = diag(m ± , m ± ), (2.4.4) χ1 χ2
2 where U and V are unitary matrices and m ± < m ± by convention. The diagonalization χ1 χ2 induces a change in basis from interaction to mass eigenstates which are then given by
+ + − − χi = Vijψj and χi = Uijψj . (2.4.5)
These two-component Weyl spinors are the building blocks of the four-component Dirac spinors χ+ χ˜+ = i , i = 1, 2. (2.4.6) i − T (χi ) ! The Lagrangian in terms of these Dirac spinors is
chargino-mass + + + + L = m ± χ˜ χ˜ + m ± χ˜ χ˜ , (2.4.7) MSSM χ˜1 1 1 χ˜2 2 2 where m ± are the same masses as in Eq. (2.4.4) (the tilde has been added for convenience) χ˜i and they are given by
1 2 2 2 2 2 2 2 2 2 2 (2.4.8) mχ˜± = M2 + µ + 2mW (M2 + µ + 2mW ) 4(mW sin 2β µM2) . 1,2 2 ∓ − − q Neutralinos The neutralinos arise due to mixings of the neutral bino, wino, and higgsinos. The Lagrangian describing this can be written as
1 0 L neutralino-mass = (ψ0)T M χ˜ ψ0 + h.c., (2.4.9) MSSM −2
2Note that two matrices are needed because in general M 6= M T .
8 2.4. The MSSM
0 T ˜ ˜ ˜ 0 ˜ 0 where (ψ ) = (B, W1, H1 , H2 ) and
M1 0 mZ cβsW mZ sβsW − 0 0 M2 mZ cβcW mZ sβcW M χ˜ = − , (2.4.10) mZ cβsW mZ cβcW 0 µ − − mZ sβsW mZ sβcW µ 0 − − with sβ = sin β, cβ = cos β and sW = sin θW , cW = cos θW , θW being the Weinberg angle. As the chargino mixing matrix, the neutralino mixing matrix depends on the wino and higgsino mass parameters M2 and µ but additionally also on the bino mass parameter M1. The bino and wino mass parameters unify in GUT theories at the GUT scale. This unification requires that at the electroweak scale
M2(mZ ) M1(mZ ) . (2.4.11) ≈ 2 The neutralino mass matrix can be diagonalized with one unitary 4 4 matrix N via × ∗ χ0 −1 N M N = diag(m 0 , m 0 , m 0 , m 0 ), (2.4.12) χ1 χ2 χ3 χ4 where by convention m 0 < m 0 < m 0 < m 0 . After the diagonalization the basis χ1 χ2 χ3 χ4 | | | | | | | 0| 0 changes to mass eigenstates which are given by χi = Nijψj . Note that N is commonly 0 chosen to be real and orthogonal3 although M χ can have negative eigenvalues in that case. 0 00 Physical fields can be obtained by chiral rotations χi χi which lead to positive masses 00 → mχ˜0 mχ00 > 0. From the two-component χi one can construct two four-component i ≡ i Majorana spinors 0 χ0 χ˜0 = i (2.4.13) i 00 T (χi ) ! with which the Lagrangian can be written as
4 chargino-mass 1 0 0 L = mχ˜0 χ˜ χ˜i . (2.4.14) MSSM −2 i i i=1 X Note that the mixing matrix and the masses are generally obtained only numerically as analytical expressions are – albeit possible to derive for real N – rather cumbersome.
0 ± Composition of χ˜i and χ˜j Depending on the amount of mixing the composition of neutralinos and charginos changes. The amount of mixing in turn depends on the SUSY breaking parameters. Some important cases are 0 0 if µ M1,2 mZ and M1 M2/2 then the two light neutralinos χ˜1, χ˜2 are • mostly| | bino | and| wino fields respectively,≈ whereas the two heavier ones are mostly ± higgsino fields. Similarly, the light charginos χ˜1 are mostly charged wino fields and ± the heavier charginos χ˜2 are mostly charged higgsino fields. There is also a relation between the masses
µ Mχ˜0 Mχ˜0 Mχ˜± Mχ˜± Mχ˜0 2Mχ˜0 . (2.4.15) | | ≈ 3 ≈ 4 ≈ 2 1 ≈ 2 ≈ 1 0 3This can only be done if all potential phases in M χ are ignored which is for example done in the pMSSM.
9 2. Supersymmetry
If µ M1,2 the composition swaps, i.e. the light neutralinos and charginos are • mostly| | higgsino | | fields and the heavier ones are mostly bino and wino fields. If µ M1 or µ M2 there is strong mixing between the fields. • | | ≈ | | | | ≈ | | If µ, M2 mZ and M1 M2/2 then Mχ˜0 Mχ˜± . • | | ≈ 2 ≈ 1 2.4.2. Squark sector Here only left and right handed mixing of the squarks is considered. Written in the basis
q˜L with gauge eigenstates, i.e. q˜ = , the squark mass term of the MSSM Lagrangian is q˜R!
M 2 M 2 q˜−mass = q˜† 2q˜ where 2 = q˜LL q˜LR . (2.4.16) L MSSM Mq˜ Mq˜ 2 2 q˜ Mq˜ Mq˜ ! X RL RR The entries of the squark mass matrix are
2 2 2 2 2 M = M + m cos 2β I3 eq sin θW + m (2.4.17) q˜LL q˜ Z − q 2 2 2 2 2 0 (2.4.18) Mq˜RR = Mq˜ + mZ eq cos 2 β sin θW + mq, 2 2 −2I3 M = M = mq Aq µ(tan β) , (2.4.19) q˜LR q˜RL − 2 2 where Mq˜ and Mq˜0 are soft SUSY breaking mass terms for the left- and right-handed squarks respectively, mq, eq,I3 are the quark mass, electric charge, and third isospin component, mZ is the Z-boson mass, µ the higgsino mass parameter, θW the Weinberg angle and Aq is the trilinear Higgs-squark-squark coupling. Note that the off-diagonal elements are proportional to the mass of the fermionic SM partner. Therefore, mixing effects can be neglected for the first two light generations and only mixing in the heavy third generation i.e. between ˜b and t˜ is considered. By introducing the squark mixing matrix Sq˜ the mass matrix can be diagonalized via
q˜ q˜ q˜ † 2 2 S M (S ) = diag(mq˜1 , mq˜2 ) (2.4.20)
q˜1 q˜L and the mass eigenstates then are = Sq˜ . The squared eigenvalues of M q˜ are q˜2! q˜R! given by
1 2 2 2 2 2 2 2 2 (2.4.21) mq˜1,2 = Mq˜LL + Mq˜RR (Mq˜ Mq˜ ) + 4 Mq˜ , 2" ∓ LL − RR | LR | # q where as usual 2 2 . The mixing matrix can be written as a rotation matrix with mq˜1 < mq˜2 the squark mixing angle θq˜
2 2 q˜ cos θq˜ sin θq˜ Mq˜LR S = where tan 2θq˜ = 2 2 . (2.4.22) sin θq˜ cos θq˜! Mq˜LL Mq˜RR − −
10 3. Virtual corrections and UV-divergences
In higher-order corrections infinities appear due to unconstrained loop momenta which have to be integrated over. These integrals can be divergent for small momenta which leads to IR-divergences or for large momenta which leads to UV-divergences. In order to get finite, physically meaningful results these infinities have to be tamed. For UV-divergences this is done by regularization (explicitly extracting the divergent part of the integral) and renormalization (absorbing the divergent part in so-called counterterms). There are different schemes in which this can be done. The procedure used in this thesis is dimensional regularization and dimensional reduction which will be explained in more detail in the next section.
3.1. Dimensional regularization
The basic idea of regularization is to split the analytical expression for a correction into a finite and an infinite part. There are different schemes in which this can be done. The conceptually easiest one is the cut-off method, which introduces an upper limit Λ (cut-off) on the energy component of the loop momentum over which one integrates (or a lower limit for a IR-divergent integral). The integral will then be a function of this cut-off but the divergence is only present in the limit Λ (or Λ 0). However, this method breaks Lorentz-invariance because the cut-off modifies→ ∞ only one→ component of the four-momentum and will therefore not be used here. Instead a less intuitive method which does however not break Lorentz-invariance is used, i.e. dimensional regularization (DReg) and dimensional reduction (DRed). They can regularize UV- and IR-divergences simultaneously. They are motivated by the observation that lowering the number of dimension can render a formerly UV-divergent integral finite, e.g.
∞ ∞ ∞ 1 1 1 d3r (linearly divergent), d2r (log. div.), dr (finite). r2 r2 r2 Za Za Za Similarly, raising the number of dimensions can render a formerly IR-divergent integral finite. Therefore, the basic idea of DReg and DRed is to evaluate the loop integrals not in 4 but in D = 4 2 space-time dimensions, i.e. − d4q µ4−D dDq, (3.1.1) → Z Z where the renormalization scale µ is introduced in order to keep the mass-dimension on both sides equal. This analytic continuation to D dimensions implies that all 4-momenta
11 3. Virtual corrections and UV-divergences
and the metric tensor gµν are now D-dimensional objects, so that
gµ = 4 gµ = D, (3.1.2) µ → µ qµ = (q0, q1, q2, q3) qµ = (q0, . . . , qD−1). (3.1.3) → The Dirac-matrices are still 4 4 matrices, that satisfy the Clifford Algebra × γµ, γν = 2gµν1, (3.1.4) { } but there are now D of them so that, e.g.
µ γ γµ = D, (3.1.5) µ γ kγ/ µ = (2 D)k./ (3.1.6) − However, there is a problem with γ5 in D dimensions, which is due to the inconsistency of
µ ν ρ σ µνρσ γµ, γ5 = 0 and Tr(γ γ γ γ γ5) = 4i (3.1.7) { } when going back to D 4. More details on this can be found in e.g. [18, 19, 20, 21]. Luckily, it is however possible→ to use the so called naive dimensional reduction prescription which drops the trace relation above and keeps the anticommutation relation [21]. Although formally inconsistent it works well in practice and will therefore be used when calculating the virtual corrections in Section 5.2.3.
3.1.1. Passarino-Veltman reduction The Passarino-Veltman reduction is a very important method for NLO calculations, since it reduces all possible one-loop tensor-integrals to just four scalar-integrals. A general N-point one-loop integral is shown in Fig. 3.1 and can be written as
(2πµ)4−D q q q N D µ1 µ2 µM (3.1.8) Tµ1,µ2,...,µM (p1, p2, . . . , pN ; m1, m2, . . . , mN ) = 2 d q ··· , iπ 1 2 N Z D D ···D
p p 2 2 1 N where 1 = q m + i (3.1.9) D − 1 2 2 2 = (q + p1) m + i (3.1.10) q D − 2 N 1 . m1 − . q + p1 q + pi m m i=1 N−1 2 N P 2 2 p2 pN 1 N = q + pi m + i − D − N m3 i=1 q + p + p X (3.1.11) 1 2
These denominators i come from the propa- D gators in the loop. The loop-momenta q in the p3 p4 numerator come from fermion propagators or vertices (e.g. the triple-gluon vertex) and the Figure 3.1. – Naming and momentum con- number M of loop-momenta in the numerator ventions for a general N-loop diagram. determines the Lorentz tensor structure of the
12 3.1. Dimensional regularization integral (e.g. M = 0 scalar integral, M = 1 vector integral, M = 2 tensor integral → → → of rank 2, etc.). It is common to use a naming conventions depending on the number N of propagators T 1 = A, T 2 = B,T 3 = C,T 4 = D. (3.1.12) The scalar integrals are then given as 4−D 2 (2πµ) D 1 1 A0(m1) = 2 d q , (3.1.13) iπ 1 ≡ 1 Z D Zq D 2 2 2 1 B0(p1; m1, m2) = , (3.1.14) 1 2 Zq D D 2 2 2 2 2 2 1 C0(p1, p2, (p1 + p2) ; m1, m2, m3) = , (3.1.15) 1 2 3 Zq D D D 2 2 2 2 2 2 2 2 2 2 1 D0(p1, p2, p3, p4, (p1 + p2) , (p2 + p3) ; m1, m2, m3, m4) = , (3.1.16) 1 2 3 4 Zq D D D D where the arguments are given in LoopTools’ [22] conventions since this will be used for the numerical evaluation of these integrals in this thesis. All loop integrals appearing in NLO calculations can be expressed in terms of these four scalar integrals. For the decomposition of the vector- and tensor-integrals one uses the fact that the general integral T is by definition symmetric under exchange of Lorentz indices and thus only symmetric tensors can appear in the decomposition. Moreover, the Lorentz structure of the general integral T must be retained by the decomposition, which implies that for e.g. vector integrals a four-momentum is needed. The only four-momenta available are those of the external legs, i.e. pi (i = 1,...,N). Following the conventions of LoopTools these can N be rewritten as k1 = p1, k2 = p1 + p2, ... , kN = i=1 pi and the decomposition of the integrals needed in this thesis can be compactly written as P µ µ B = k1 B1, (3.1.17) µν µν µ ν B = g B00 + k1 k1 B11, (3.1.18) 2 µ µ µ µ C = k1 C1 + k2 C2, = ki Ci, (3.1.19) i=1 X 2 µν µν µ ν C = g C00 + ki kj Cij, (3.1.20) i,j=1 X 2 2 µνρ µν ρ νρ µ µρ ν µ ν ρ C = g ki + g ki + g ki C00i + ki kj kl Cijl, (3.1.21) i=1 i,j,l=1 X X 3 µ µ D = ki Di, (3.1.22) i=1 X 3 µν µν µ ν D = g D00 + ki kj Dij, (3.1.23) i,j=1 X 3 3 µνρ µν ρ νρ µ µρ ν µ ν ρ D = g ki + g ki + g ki D00i + ki kj kl Dijl. (3.1.24) i=1 i,j,l=1 X X Note however that the results have to be written in terms of pi and not ki for Resummino.
13 3. Virtual corrections and UV-divergences
Scalar integrals All scalar integrals can be written in a form containing only one generic integral of the form D 1 In(A) = d q , (3.1.25) (q2 A + i)n Z − where, of course, n should be bigger or equal to 1. For D < 2n and A > 0 this integral converges and the result is
n D/2 Γ(n D/2) D/2−n In(A) = i( 1) π − (A i) . (3.1.26) − Γ(n) − Note, that for D 4 this result only diverges if n = 1 or n = 2 since then the argument → of the Γ function in the numerator is 1 (for n = 1) or 0 (for n = 2), which are both − poles at which the Γ-function is undefined. So only I1 and I2 (and hence A0 and B0) are divergent, whereas all other generic integrals are convergent.
Scalar Integral A0 The simplest scalar integral is defined as
4−D 2 (2πµ) D 1 A0(m ) := d q (3.1.27) iπ2 q2 m2 + i 4−D Z − (2πµ) 2 = I1(m ) (3.1.28) iπ2 4−D m2 2 2 D = m2 Γ − . (3.1.29) − 4πµ2 2 It diverges for D 4 since the Γ(x) has a pole at x = 1. The trick is to evaluate the → − integral in D = 4 2ε dimensions and expand around ε = 0. Expressed in terms of ε the integral reads − 2 −ε 2 2 m A0(m ) = m Γ(ε 1) (3.1.30) − 4πµ2 − and can be expanded using −ε m2 m2 = 1 ε ln + (ε2), (Taylor series) 4πµ2 − 4πµ2 O 1 1 Γ(ε 1) = Γ(1 + ε) = Γ(1) + γE + (ε). (Laurent series) − ε(ε 1) −ε − O − Thus up to first order in ε one finds 2 2 2 1 m A0(m ) = m γE + ln 4π ln + 1 + (ε) (3.1.31) ε − − µ2 O m2 = m2 ∆ ln + 1 + (ε) , (3.1.32) − µ2 O 1 where ∆ = ε γE +ln 4π contains the divergent part and some constants that are absorbed into the counterterms− in the MS-scheme (more on this and other schemes in the next section).
14 3.1. Dimensional regularization
2 Care has to be taken if terms DA0(m ) appear in the loop integrals. Simply letting ∝ D 4 does not do the trick; rather one has to go back to Eq. (3.1.30) and expand → 2 −ε 2 2 m DA0(m ) = (4 2ε)m Γ(ε 1) (3.1.33) − − 4πµ2 − 2 2 = 4A0(m ) 2m , (3.1.34) − 2 where A0(m ) in the last line is of course in the expanded form given in Eq. (3.1.32).
The evaluation of the other scalar integrals is more involved, since the generic integral cannot be used immediately but only after performing a Feynman parametrization. How this is done in detail can be seen in e.g. [23]. Here, only some important results are mentioned: First, note that B0 is the only other UV-divergent scalar integral (the C0 and D0 integrals are proportional to I3 and I4 respectively and therefore UV-finite). The integrals A0 and B0 can be split into a divergent and a finite part 2 2 A0(m ) = m ∆ + finite terms + (), (3.1.35) O 2 2 2 B0(p ; m , m ) = ∆ + finite terms + (). (3.1.36) 1 2 O Also note that in the special case B0(0, 0, 0) = 1/UV 1/IR = 0 an IR-divergence appears in addition to the UV-divergence. Although these divergences− formally cancel, it is useful to keep them separated in order to be able to check UV-finiteness in intermediate results.
Tensor reduction µ µ Determination of B1 In order to determine B1 one contracts both sides of B = p1 B1 with p1µ which yields
µ (p1.q) 2 p1µB = 2 2 2 2 = p1B1. (3.1.37) [q m + i][(q + p1) m + i] Zq − 1 − 2 By cleverly rewriting (p1.q) as
1 2 2 2 2 2 2 2 (p1.q) = [(q + p1) m + i] [q m + i] (p m + m ) (3.1.38) 2 − 2 − − 1 − 1 − 2 1 the integral can ben written in terms of scalar integrals o
2 1 1 1 p1B1 = 2 2 2 2 (3.1.39) 2 q m + i − (q + p1) m + i Zq − 1 Zq − 2 2 2 2 1 (p1 m2 + m1) 2 2 2 2 (3.1.40) − − [q m + i][(q + p1) m + i] Zq − 1 − 2 After shifting the momentum q q p1 in the second integral the final result reads → −
2 2 2 1 2 2 2 2 2 2 2 2 B1(p1; m1, m2) = 2 A0(m1) A0(m2) (p1 m2 + m1)B0(p1; m1, m2) . 2p1 − − − h i (3.1.41) Using Eqs. (3.1.35) and (3.1.36) B1 can be written in terms of a divergent and a finite part
2 2 2 1 B1(p ; m , m ) = ∆ + finite terms + (). (3.1.42) 1 1 2 −2 O
15 3. Virtual corrections and UV-divergences
Determination of B00 and B11 In order to determine the two coefficients, two equations µν µν µ ν are needed. Starting from the decomposition B = g B00 + p1 p1B11 one equation is obtained by contracting both sides with gµν, the other by contracting with p1µ. The contraction with the metric tensor leads to 2 µν q 2 gµνB = 2 2 2 2 = DB00 + p1B11 (3.1.43) [q m + i][(q + p1) m + i] Zq − 1 − 2 which can be broken down to scalar integrals again using a different trick, i.e. rewriting 2 2 2 2 q = (q m ) + m and thus yields the first equation for determining B00 and B11 − 1 1 2 1 2 1 DB00 + p B11 = + m 1 (q + p )2 m2 + i 1 [q2 m2 + i][(q + p )2 m2 + i] Zq 1 2 Zq 1 1 2 2 2− 2 2 2 − − = A0(m2) + m1B0(p1; m1, m2). (3.1.44) The contraction with the external momentum leads to ν µν (p1.q)q ν 2 p1µB = 2 2 2 2 = p1(B00 + p1B11). (3.1.45) [q m + i][(q + p1) m + i] Zq − 1 − 2 Rewriting (p1.q) as above does not immediately lead to scalar integrals but rather to
ν ν ν 2 1 q q p (B00 + p B11) = (3.1.46) 1 1 2 q2 m2 + i − (q + p )2 m2 + i Zq 1 Zq 1 2 − − ν 2 2 2 q (p1 m2 + m1) 2 2 2 2 , − − [q m + i][(q + p1) m + i] Zq − 1 − 2 (3.1.47) which can be evaluated using the fact that
dDq qµ1 qµN f(q2) = 0, for odd N. ··· Z ν 2 Here, this implies that the first integral is zero, the second yields p1A0(m2) after shifting µ µ − q q p1 and the last integral is simply B = p B1. The second equation for determining → − 1 B00 and B11 then is
2 1 2 2 2 2 2 2 2 (B00 + p B11) = A0(m (p m + m )B1(p ; m , m ) . (3.1.48) 1 2 2 − 1 − 2 1 1 1 2 h i The solutions of Eqs. (3.1.44) and (3.1.48) are
1 2 2 2 2 2 2 2 2 2 2 2 B00 = A0(m ) + 2m B0(p ; m , m ) + (p m + m )B1(p ; m , m ) 2(D 1) 2 1 1 1 2 1 − 2 1 1 1 2 − h i 1 1 2 2 2 2 2 2 2 2 2 2 2 B11 = 2 (D 2)A0(m2) 2m1B0(p1; m1, m2) D(p1 m2 + m1)B1(p1; m1, m2) . 2(D 1) p1 − − − − − h i However, as was the case with DA0, simply letting D 4 doesn’t work. Rather, one has to expand the prefactor →
1 1 1 1 2 = = 1 + + (2) 2(D 1) 6 1 2 6 3 O − − 3
16 3.1. Dimensional regularization
as well as A0,B0, and B1 according to Eqs. (3.1.35), (3.1.36) and (3.1.42) in order to get correct results
1 2 2 2 2 2 B00 = A0(m ) + 2m B0(p ; m , m ) 6 2 1 1 1 2 h 2 2 2 2 2 2 2 2 2 p1 + (p m + m )B1(p ; m , m ) + m + m , (3.1.49) 1 − 2 1 1 1 2 2 1 − 3 i 1 2 2 2 2 2 B11 = 2 2A0(m2) 2m1B0(p1; m1, m2) 6p1 − h 2 2 2 2 2 2 2 2 2 p1 4(p m + m )B1(p ; m , m ) m m + . (3.1.50) − 1 − 2 1 1 1 2 − 2 − 1 3 i As with the other coefficients it is useful to write them as the sum of a divergent and finite terms
2 2 2 1 2 2 2 B00(p ; m , m ) = 3(m m ) p ∆ + finite terms + (), (3.1.51) 1 1 2 12 2 − 1 − 1 O 2 2 2 1 h i B11(p ; m , m ) = ∆ + finite terms + (). (3.1.52) 1 1 2 3 O
3.1.2. UV divergent parts For calculating the counterterms in the MS-scheme it is useful to know the UV divergent parts of the scalar integrals and of the coefficients of the tensor integrals. Moreover, one also needs the UV divergent parts of (D 4) times these integrals (cf. [23, 24]) −
2 2 UV[A0] = m ∆, UV (D 4)A0 = 2m , (3.1.53) − − UV[B0] = ∆, UV (D 4)B0 = 2, (3.1.54) − − 1 UV[B1] = ∆, UV(D 4)B1 = 1, (3.1.55) −2 − 1 2 1 2 2 2 UV[B00] = ∆p , UV (D 4)B00 = (p 3m 3m ), (3.1.56) −12 − 6 − 1 − 2 1 2 UV[B11] = ∆, UV (D 4)B11 = . (3.1.57) 3 − − 3
17
4. Renormalization
Having explicitly identified the divergent parts of the loop-integrals using regularization, the next step is to absorb these divergences into so-called counterterms in order to get UV- finite results. This is commonly done by multiplicative renormalization, i.e. by introducing renormalized fields ΨR, couplings λR, and masses mR via 1 Ψ0 = ZΨ ΨR = 1 + δZΨ ΨR = 1 + δZΨ ΨR, (4.0.1) 2 λ0 =ZpλλR = (1 +pδZλ)λ (4.0.2)
m0 =ZmmR = (1 + δZm)mR, (4.0.3) where Zi are renormalization constants, the subscript 0 denotes bare and the subscript R renormalized quantities. For tree-level calculations one simply has Zi = 1 and therefore one expands the renormalization constants as
2 Zi = 1 + δZi + (λ ) (4.0.4) O for NLO calculations. The initial Lagrangian L 0 expressed in terms of the newly defined renormalized quantities can be split into two parts
L 0 = L 0 + L ct, (4.0.5) Ψ0→ΨR,λ0→λR,... where the first term has exactly the same structure as the initial Lagrangian but only contains renormalized quantities, whereas the second term is new and contains the so called counterterms. These are defined in such a way as to cancel the UV-divergences order by order. However, the counterterms are only defined up to a finite part. This gives rise to different renormalization schemes, which differ only w.r.t. the finite part that is absorbed by the counterterms. The simplest renormalization scheme is the minimal subtraction (MS) scheme, which uses dimensional regularization to extract the divergent parts and then cancels only the pure divergence (1/) with the counterterms. But as was seen in Eq. (3.1.32) the divergence is often accompanied by the constants γE and ln 4π, so it useful to subtract these as well. This is exactly what the modified minimal subtraction (MS) scheme does. For SUSY theories, however, these schemes do not work since dimensional regularization explicitly breaks SUSY due to the mismatch of (D 2) vector boson degrees of freedom and the 2 degrees of freedom of their superpartners.− A way out of this problem is to use a different regularization technique: dimensional reduction [25, 26]. The renormalization scheme is then also called dimensional reduction (DR) or modified dimensional reduction (DR). Here, vector bosons are kept 4 dimensional in order to preserve SUSY. This comes at the cost of mathematical inconsistency but it turns out that the technique can still be used for practical purposes [27]. In contrast to these schemes the OS-scheme is a non-minimal scheme. The main idea of this scheme is that the renormalized mass should correspond to the physical mass, which
19 4. Renormalization
means the physical mass is the real part of the propagator‘s pole. Additionally, the residue of the propagator is normalized to one. Important for this thesis is the renormalization of quarks and squarks which will be described in detail in the next two sections. These are based on [8, sec. 3.4] and references therein.
4.1. Renormalization of quarks
The mass and field counterterms for non-mixing chiral quarks can be derived from the bare Lagrangian L L R L L0 = iq¯ ∂q/ m0q¯ q + (L R), (4.1.1) 0 0 − 0 0 ↔ – where the subscript 0 denotes bare quantities, and L and R denote the left- and right- L handed parts of the bare quark field q0 – by introducing renormalized quantities m, q and qR such that
2 m0 = mZm = m(1 + δZm) + (α ), (4.1.2) O s 1 qL = ZL qL = 1 + δZL qL = 1 + δZL qL + (α2), (4.1.3) 0 q q 2 q O s q q 1 qR = ZR qR = 1 + δZR qR = 1 + δZR qR + (α2). (4.1.4) 0 q q 2 q O s q q Inserting these into the bare Lagrangian yields
L = iq¯L∂q/ L mq¯RqL + (L R) (4.1.5) − ↔ L L L 1 R∗ L R L + iRe δZ q¯ ∂q/ m δZm + δZ + δZ q¯ q + (L R), (4.1.6) q − 2 q q ↔ h i where the first line is just the original Lagrangian in terms of renormalized quantities and the second line is the one from which the counterterms can be extracted. This is done by considering i L , replacing all derivatives by ( ip), where p denotes the incoming − momentum, and then removing all the fields. In this case i L (×) (where ( ) denotes the counterterm) is given by ×
(×) L L L 1 R∗ L R L i L = Re δZ q¯ ∂q/ im δZm + δZ + δZ q¯ q + (L R) (4.1.7) − q − 2 q q ↔ L h 1 R∗ Li = Re δZ q¯∂P/ Lq im δZm + δZ + δZ qP¯ Lq + (L R), (4.1.8) − q − 2 q q ↔ h i L,R L,R 2 µ µ where q = PL,Rq, q¯ = qP¯ R,L, PL,R = PL,R, and PL,Rγ = γ PR,L was used. Replacing the derivatives by ( ip) and removing the fields yields the counterterm for the two-point Green’s function −
(×) L 1 R∗ L iΣ = i Re δZ p/ m δZm + δZ + δZ PL + (L R). (4.1.9) − q − 2 q q ↔ n h io
20 4.2. Renormalization of squarks
The quark propagator at NLO can then be written as the sum of the LO propagator, the bare self-energy contribution ( iΣ) and the counterterm ( iΣ(×)) − − i(p/ + m) i(p/ + m) i(p/ + m) i(p/ + m) i(p/ + m) + ( iΣ) + ( iΣ(×)) (4.1.10) p2 m2 p2 m2 − p2 m2 p2 m2 − p2 m2 − − − − − i(p/ + m) i(p/ + m) i(p/ + m) = + ( iΣ)ˆ (4.1.11) p2 m2 p2 m2 − p2 m2 − − − i(p/ + m) (p/ + m) = 1 + Σ(ˆ p2) , (4.1.12) p2 m2 p2 m2 − − where Σˆ denotes the renormalized self-energy. Splitting the bare self-energy into a left- and right-handed part such that iΣ = i(p/ΣL + ΣL) + (L R) and adding this to − − V S ↔ iΣ(×), the renormalized self-energy can be written as − L L L 1 R∗ L ΣS Σˆ = p/ Σ Re(δZ ) + m δZm + (δZ + δZ ) + PL + (L R) (4.1.13) V − q 2 q q m ↔ n ˆ L h i o ΣV ˆ L ΣS |L L{z } = p/Σˆ + Σˆ PL + (L | R). {z } (4.1.14) V S ↔ The OS-conditions for quarks are ˆ Re Σ p2=m2 u(p) = 0, (4.1.15) p/ + m lim Re Σˆ u(p) = 0, (4.1.16) p2→m2 p2 m2 − where the first condition fixes the renormalized mass to be the physical mass and the second condition sets the residue of the propagator to one. Using these conditions and the fact that they have to be fulfilled for the left- and right-handed parts separately, yields the OS renormalization constants for quarks
L R OS 1 L R ΣS + ΣS δZm = Re ΣV + ΣV + , (4.1.17) − 2 m p2=m2 h ˙ L i ˙ R L,OS L 2 ˙ L ˙ R ΣS + ΣS δZq = Re ΣV + m ΣV + ΣV + , (4.1.18) m p2=m2 h ˙ R ˙ L i R,OS R 2 ˙ R ˙ L ΣS + ΣS δZq = Re ΣV + m ΣV + ΣV + , (4.1.19) m p2=m2 h i ˙ d where Σ = dp2 Σ. Note that these equations simplify a lot for massless quarks. The MS counterterms can be determined from the OS counterterms by taking the UV finite part according to Eqs. (3.1.53) to (3.1.57).
4.2. Renormalization of squarks
The counterterms for squarks are derived in a similar manner. The Lagrangian in this case is 2 † µ 2 † L 0 = ∂µq˜ ∂ q˜0,i m q˜ q˜i,0 , (4.2.1) 0,i − i,0 i,0 i=1 X
21 4. Renormalization
2 where i labels the two different mass eigenstates with eigenvalues mi,0. The renormalized quantities are introduced via
1 2 q˜i,0 = Zij q˜j = (δij + δZijq˜j) + (α ), (4.2.2) 2 O s m2 = pZ2 m2 = (1 + δZ2 )m2 + (α2). (4.2.3) 0,i mi i mi i O s In the same manner as for the quarks, the counterterm for the two-point Green’s function can be derived, yielding
(×) 1 ∗ 2 1 2 2 ∗ 2 2 iΣ = i δZij + δZ p m δZij + m δZ δijδZ m . (4.2.4) − ij 2 ji − 2 i j ji − m i NLO With this counterterm the squark propagator up to NLO Pij can be written in terms of LO 2 2 the LO squark propagator P = iδij/(p m ) as ij − i NLO LO LO (×) LO P = P + P iΣkl iΣ P (4.2.5) ij ij ik − − kl lj LO LO ˆ LO = Pij + Pik iΣkl Plj . (4.2.6) − The mass counterterm in the OS scheme can be determined by requiring that the pole of this propagator is at the physical mass. For i = j this is ensured by the condition ˆ 2 Re Σii(mi ) = 0 from which the mass counterterm can be determined to be