aainstitutetext: Department of Physics and State Key Laboratory of Nuclear Physics and Technology,
Peking University,
Beijing 100871, Chinabbinstitutetext: Collaborative Innovation Center of Quantum Matter,
Beijing 100871, Chinaccinstitutetext: Center for High Energy Physics, Peking University,
Beijing 100871, China
Massive On-shell Recursion Relations for -point Amplitudes
We construct two and three-line shifts for tree-level amplitude with massless and/or massive particles, and provide a method to construct general multi-line shifts for all masses. We choose the massless-massive BCFW shift from these shifts and examine its validity in renormalizable theories. Using such a shift, we find that amplitudes with at least one massless vector boson are constructible. This reveals the importance of gauge theory in the construction of amplitudes with massive particles. We also find that this kind of amplitudes have a cancellation related to group structure among different channels, which is essential for constructibility. Furthermore, we show that in the limit of large shift parameter , the amplitude with four massive vector bosons, which can include transverse massive vector particles, have structures proportional to the amplitude with shifted vector particles replaced by Goldstone bosons in the leading order. This is responsible for the failure of massive-massive BCFW recursion relations in the amplitudes with four massive vector bosons.
{fmffile}
graph
1 Introduction
In comparison with Feynman diagrams, on-shell recursion relations provide a more efficient approach to construct higher-point tree-level amplitudes from lower-point amplitudes. It was first motivated by Britto-Cachazo-Feng-Witten (BCFW) recursion relations Britto:2004ap; Britto:2005fq in the calculation of gluon scattering, and other versions of recursion relations were proposed to study the amplitudes in gauge theories Badger:2005zh; Risager:2005vk, gravity theories Cachazo:2005ca, supersymmetric theories Brandhuber:2008pf; Arkani-Hamed:2008owk, scalar effective field theories Cheung:2015ota and more general theories Cohen:2010mi; Cheung:2015cba. They based on the same idea, namely using complex deformation of the external momenta and calculating the residues of deformed amplitudes in the complex plane, to collect the information of factorized lower-point amplitudes.
Contrary to massless particles, the momenta of massive particles cannot be written as a direct product of two spinors. To analyze amplitudes with massive particles, the method of decomposing the massive momenta into two light-like vectors was developed Schwinn:2005pi; Schwinn:2006ca; Schwinn:2007ee; Craig:2011ws; Boels:2011zz. However, this formalism is not the most convenient for specific calculation of amplitudes, since it’s not little-group covariant. Recently, Arkani-Hamed, Huang and Huang Arkani-Hamed:2017jhn introduced a new method by regarding the massive particle as a representation of its little group. In this notation, both amplitudes and complex shifts are simplified into a little-group covariant form. There have been some efforts in constructing massive BCFW shift in the massless-massive case Aoude:2019tzn; Ballav:2020ese and massive-massive case Herderschee:2019dmc; Franken:2019wqr afterwards.
After considering little-group invariance, the on-shell constructibility of amplitudes with massive particles was investigated Franken:2019wqr. Various multi-line shifts were used to estimate the large- behavior of amplitudes with all particles massive, where is the shift parameter. Although people prefer using three and more line shifts to investigate the constructibility, two-line shifts are more convenient in the computation of amplitudes.
In this work, we take the little-group covariant spinor helicity formalism and regard different spin states from one massive external leg as a whole, so these states should be deformed by a same shift. In the case that all particles are massless, some researches ArkaniHamed:2008yf; Cheung:2008dn proved that in a gauge theory coupled to scalars and fermions, any massless amplitude with at least one gluon is two-line constructible, which is a strong conclusion. We want to examine whether the amplitude with both massless and massive particles is also two-line constructible.
The present paper is organized as follows. In section 2 we review the basic idea of recursion relations and construct all possible two and three-line shifts for both massive and massless particles. Assuming the coupling is dimensionless, we evaluate the Feynman rule in a diagrammatic way in section 3 and discuss why gauge-fixing is not enough to improve the behavior of amplitude in the large- limit. In section 4 we evaluate -point massive amplitudes with at least one massless vector boson in the large- limit. We find that these amplitudes vanish in the large- limit except for the all vector amplitudes, which can have a cancellation among different channels. In section LABEL:sec5, we give the relation between such a cancellation and the group structure of massive vectors, and explore why this cancellation fails in the amplitudes without massless particles. Finally, section LABEL:sec6 presents our conclusion and discussion. Appendix LABEL:AppA gives our conventions. Appendix LABEL:AppB explicitly calculates the large- behavior of polarization vectors in the center-of-mass frame. Appendix LABEL:AppC gives an example of evaluating the diagrammatic expressions.
2 Recursion relations for all masses
For completeness, we review the general complex shift for , a tree-level amplitude with massless particles. For each external particle, we shift their momentum vector by complex-valued vector ,
(1)
where . Now we restrict these shift vectors by three conditions,
(2)
(3)
(4)
These three conditions (2–4) respectively guarantee that (a) momentum conservation holds for shifted momenta, (b) shifted momenta are still on-shell, (c) shifted propagators are linear in . We can construct a complex function , whose residue at is the unshifted amplitude . Then Cauchy’s theorem tells us,
(5)
where the boundary term is the residue at infinity. The first term on the right-hand side can factorize into two on-shell subamplitudes when the momentum of the internal line goes on-shell,
(6)
Now we consider massive amplitudes. The three conditions (2–4) still keep the shifted momenta on-shell . To generalize formula (5) into the massive case, we add a new term which corresponds to massive on-shell propagators,
(7)
At a -pole one of massive internal particles goes on-shell. We use little-group covariant spinors Arkani-Hamed:2017jhn to describe massive particles, so the two subamplitudes and have little-group indices and should be contracted,
(8)
where the lower indices in the parenthesis means symmetrization of these indices. Here we only write the little-group indices related to the internal momentum and neglect other little-group indices. If the boundary term , the -point on-shell amplitude will be completely determined in lower-point on-shell amplitudes and this recursive formula (7) becomes an on-shell recursion relation under a valid shift.
Before we discuss whether amplitudes vanish in the large- limit, we should construct complex shifts for the composition of all masses. In this section, we will solve equations (2–4) to give all possible shift vectors in two and three-line shifts. Since massless shifts have been well studied, their generalizations in the massive case will be based on these massless shifts.
2.1 Shifting massless particles
In spinor-helicity formalism, which is briefly reviewed in appendix LABEL:AppA, we write massless amplitudes in terms of two kinds of Weyl spinor and . They are two inequivalent fundamental representations of . Both of them can be shifted, we refer to the former as holomorphic shift and the latter as anti-holomorphic shift.
Since the massive particles takes little-group representation into account, the three conditions (2–4) may not be valid in the massive case. We review some specific massless amplitude recursion relations to translate the three conditions (2–4) into massless spinor-helicity variables. Although all shift-vectors could be non-trivial , two or three-line shifts are enough to construct amplitudes in many applications. Let’s start from these few-line shifts.
1) Two-line shift
Since the shifted momentum is linear in , we can’t use holomorphic and anti-holomorphic shifts simultaneously for one particle. Otherwise, the momentum conservation condition would be violated. When we shift two external lines and , there are two choices. We choose holomorphic shift for particle and anti-holomorphic shift for particle ,
(9)
We call this a -shift. The shift-vector , so momentum conservation condition (2) is automatically satisfied. The shift vector is
We find that is responsible for the condition (3). Weyl equations are responsible for condition (4).
2) Risager-type three-line shift
In Risager-type, all the shifted external lines are holomorphic shifts. The shifted spinors are
(12)
where is an arbitrary reference spinor. Here we ignore the dimension analysis for convenience111Actually The mass dimension of , and is . In order to ensure the dimensionless , we should write the shifts as , and , where the mass dimension of constant is . Since we only want to discuss the large- behavior, we ignore such constants, which will not change the result of our following analysis..
The shift vectors are
(13)
Using Schouten identity , we easily verify condition (2) . We find that ensures condition (3), and that Weyl equations are responsible for condition (4).
3) BCFW-type three-line shift
In BCFW-type, only one shifted external line is anti-holomorphic shift, the other shifted external lines are holomorphic shifts. The shifted spinors are,
(14)
where is an arbitrary reference spinor. The shift vectors are
(15)
We use Schouten identity to verify condition (2), . We find that ensures condition (3) and Weyl equations are responsible for condition (4).
So far all possible two and three-line shifts in the massless case are presented. Since Risager and BCFW-type shifts are the only two distinct classes of recursion relations Cheung:2015cba, the analysis of four and more-line shifts are just generalizations of three-line shifts. They are sufficient for analyzing mixed or massive recursion relations. We find that the antisymmetry of brackets and ensure condition (3), and the equation of motion for massless particles is responsible for condition (4).
2.2 Shifting mixed particles
Now let’s think about how to satisfy the three conditions when massive particles are taken into account. The massive spinor-helicity variables and are not only fundamental representations of Lorentz group , but also fundamental representations of little group . We write them as and , which are the representations of . In the massive case, the antisymmetric brackets
(16)
and massive Dirac equation
(17)
are different from the massless case. Therefore, conditions (3) and (4) wouldn’t be satisfied automatically in the massive case. For example, we can give up condition (4) for massive particles and construct a shift, where masses are no longer invariants. The validity of this kind of shifts was examined numerically for the amplitudes with graviton and scalar bosons Britto:2021pud. Here we try to satisfy all conditions.
The key point is to use some variables to contract the little-group index or Weyl-spinor index of massive spinors. For example, we can introduce an unknown variable , which is the representation . We use to contract the little-group index of , so the inner product gives the antisymmetric bracket
(18)
which ensures condition (3). Another way is to introduce and contract the Weyl-spinor index of , which has been used in ref. Franken:2019wqr.
Our strategy is the former one. First, we take massless shifts in Section 2.1 and replace the massless variables and with the massive spinor-helicity variables and for massive particles, while the spinor part of the shifts remain as massless shifts. Then we introduce some unknown variables with little-group indices (, , , etc.), whose number equals the number of massive shifted external legs. With the shifts which have been replaced multiplied by or contracted with these unknown variables, condition (3) must be satisfied. Since the above manipulation is based on massless shifts, condition (2) is still satisfied. Finally, we use the last condition (4) to determine these unknown variables.
2.2.1 Two-line shifts for mixed particles
Let’s consider two-line shifts for massive particle and massless particle . There are two ways to shift them. We can choose the -shift: the massless line is shifted holomorphically, while the massive line is shifted anti-holomorphically. We introduce one unknown , so the shifted spinors are
(19)
The shift vector is . It is orthogonal to the massless momentum because of the Weyl equation . Condition (4) leads to
(20)
The solution is . We substitute the solution into eq. (19), so the explicit form of -shift is
(21)
Another choice is the -shift: the massless line uses holomorphic shift, the massive line uses anti-holomorphic shift. For real-valued momenta, angle and square spinors are not independent. For massless particles, we have and . For the massive particle , the complex conjugation of massive spinors lowers the little indices: and . Therefore, the -shift can be implemented from the complex conjugate of -shift,
(22)
2.2.2 Three-line Risager type shifts for mixed Particles
We take massless shift (12) and choose one or two shifted particles to be massive. Since all shifts are holomorphic shifts in Risager type, each shift vector must be . Condition (3) is satisfied, because .
1) One massive and two massless
Let particle be massive particle and particles and be massless particles. We introduce one unknown , so the shifted spinors are
The shift vectors should be non-trivial, so , . The only solutions are , .
2.2.3 Three-line BCFW type shifts for mixed particles
We take massless shift (14) and choose one or two shifted particles to be massive. If particle is massive, each shift vector in this type of shifts must be . Condition (3) is satisfied, because . If particle is massless, condition (3) is also satisfied for the same reason as the case of massless three-line BCFW type shifts.
1) One massive and two massless
Since BCFW type recursion relations use holomorphic and anti-holomorphic shift, there are two kinds of compositions. The first kind is that the massive particle uses holomorphic shift. Let particle be massive particle and particles and be massless particle. We introduce one unknown , so the shifted spinors are
The second kind is that the massive particle uses anti-holomorphic shift. Let particle be massive particle and particles and be massless particles. We introduce one unknown , so the shifted spinors are
The solutions are , . The solutions become more complicated.
2.3 Shifting massive particles
The all-massive recursion relations have been worked out to study the constructibility of all-massive amplitudes in spontaneously broken gauge theories Franken:2019wqr. Now we want to reproduce these massive shifts with our method. Similarly, we introduce unknown variables for -line shifts and then solve eq. (4) for these unknown variables.
2.3.1 Two-line shift for massive particles
Unfortunately, we can’t construct consistent massive BCFW shift for two massive lines as simply as what we do in last subsection. We don’t have a natural choice of massless spinors to contract with Weyl-spinor index, since two external lines are both massive particles. It means that the general form of two-line shift doesn’t exist. We must choose a specific spinor or reference frame to write down a particular shift. For example, we can only shift one of the helicity states instead of both Herderschee:2019dmc; Franken:2019wqr in a special frame.
Now we still introduce two unknowns and in -shift, so the shifted spinors are
The tensor has three degrees of freedom, since its determinant is zero. Two equations are not enough to determine all degrees of freedom, so we should choose and from other information. The massive momenta of two particles can be written as different linear combinations of two null vectors and :
(43)
where and are coefficient. We choose , . We can verify condition 4,
(44)
Now the shift can be written as
(45)
2.3.2 Three-line Risage-type shifts for massive particles
We introduce three unknowns , and . Since they are all massive, there is a permutation symmetry between these shifted lines. The shifted spinors are
We can choose . However, it isn’t the final solution. After substituting , we get the final result: .
2.4 Explicit form of three-line shifts
In previous discussions, we figured out solutions for all two and three-line shifts. To simplify the expressions of these shifts, we define some new Weyl spinors,
(52)
where and correspond to massive particles. The simplified expressions of three-line shifts are shown in table 1, in which the massive spinor helicity variables are denoted in BOLD notation. Now we can write down massless three-line shifts (12) and (14) and use replacements listed in table 1 to rederive three-line shifts for all masses.
Table 1: Three-line shifts for all masses, where the little-group indices are suppressed.
Furthermore, the notation is not necessary in the expressions. We can rewrite and in terms of ,
(53)
For example, we set . The all-massive BCFW-type three-line shifts reduce to
(54)
This expression coincides with Franken:2019wqr.
3 Feynman rules in the large- limit
There is no general constructive method to give an expression of the contribution in eq. (7), so the recursion relations hold as long as . To investigate the validity of recursion relations, there are many works on the large- behavior of tree-level massless amplitudes in various shifts ArkaniHamed:2008yf; Cheung:2008dn; Cohen:2010mi; Cheung:2015cba. In refs. ArkaniHamed:2008yf; Cheung:2008dn the authors focused on the BCFW recursion relations and used background field method to show that, in a theory of spin , any massless amplitudes with at least one gluon is constructible. We want to examine whether this argument is applicable to the massive case, so the steps in their proof should be carefully reconsidered.
We label two external particles of massless amplitudes by 1 and 2. Their momenta are chosen to be deformed,
(55)
which corresponds to eq. (9). In the background field method, the large- behavior of amplitudes have a nice physical interpretation. We take particles 1 and 2 to be incoming and outgoing, so this process can be interpreted as a hard particle shooting through a soft background. In the hard limit , the -independent soft physics is treated as a classical background, while the large- behavior of amplitudes is completely determined by the hard fluctuations.
Now, let’s discuss the -dependent propagators, vertices and external legs separately. We will see differences in the case when hard fluctuations correspond to massive particles.
3.1 Hard propagators
The first problem is how massive propagators scale at large-. Both massive fermions and scalar propagators scale as the same as massless propagators, while a massive vector propagator goes as :
(56)
Notice that a massless vector propagator goes as , so we should distinguish massless and massive vectors in the following discussion. As in table 2, we use single and double wavy lines to make a distinction between massless and massive vectors propagators.
Here we introduce a diagrammatic expression to represent the numerator in propagator (56):
(57)
where the line segment and the letters attached with it, which are either indices or momentum , compose a representation of Lorentz group. The first diagram represents a symmetric tensor , in which the line segment connects two Lorentz indices. The second diagram represents two Lorentz vectors, where the line segments connect the shifted momentum to Lorentz indices.
Furthermore, Einstein summation also has a diagrammatic representation,
(58)
where two same indices are equivalent to a line segment. Therefore, the first diagram reduces to a line segment with two momenta attached to it, which represents Lorentz scalar .
3.2 External polarizations
External spinors are constructed in the little-group notation,
(59)
where the little-group indices characterize two different solutions of the Dirac equation.
For massive vector bosons, the polarization vectors transform under the three-dimensional tensor representation of the little group,
(60)
However, this little-group covariant expression is not convenient when we are talking about amplitudes in the large- limit. Usually we don’t choose the rest frame of massive particles as a reference system, since any shift should include more than one external leg. It implies that we can choose the spin axis along the 3-momentum direction and write down the spin state for massive vector particles in terms of :
(61)
where we ignore Lorentz indices.
3.2.1 -shift
To consider two-line shifts, we take particles 1 and 2 to be massless and massive respectively. Particle 1 is always a gauge boson, while particle 2 can be any massive particle whose spin . Since the large- behavior is independent of the reference system, we can choose the center-of-mass frame of the particles for simplicity. The momenta of particle 1 and 2 and shift vectors become
(62)
where the 3-momenta are normalized. The shift vector is basically the same as the polarization vectors for the real momentum .
In pure Yang-Mills theory, the shift vector is enough to construct the polarization vectors. We add a new vector to give the longitudinal polarization, whose spatial components point opposite to the direction of spatial components of . Similarly, we construct a new null vector . The expressions of and are,
(63)
We choose -shift (21). The shifted polarization vectors of particle 1 are
(64)
which are the same as in the all-massless case. They should stay orthogonal to momentum and their product is maintained. The Ward identity is still valid for complexified amplitudes:
(65)
Therefore, we can use it to replace the negative polarization,
(66)
If particle 2 is a vector boson, we choose the axis as the spin direction, and the shifted external polarizations are
(67)
where . Since we focus on the large- behavior of amplitudes, the overall factor in the expression of can be ignored. The polarizations are modified appropriately to remain normalized to unity and orthogonal to . What’s more, the longitudinal polarization should be orthogonal to other two transverse polarizations. A more detailed discussion is performed in appendix LABEL:AppB. Using Goldstone boson equivalence theorem, the scaling behaviour of massive vector bosons can also be improved Franken:2019wqr. However, this improvement changes type of particles, so we don’t apply it in the following analysis.
If particle 2 is a fermion, we need the shifted Dirac spinors,
(68)
Although particle 1 is a gauge boson, we still use and for consistency.
3.2.2 -shift
Particles 1 and 2 have equal mass. The momenta of them become
(69)
We set the shift vector to be the same as in eq. (62). Now both and are time-like vectors, we need two vectors and to give the longitudinal polarizations,
(70)
Here we only discuss the case that both particles 1 and 2 are massive vector bosons. The shifted external legs of massive vectors are
(71)
where . In the high energy limit, , and . Therefore, eqs. (64) and (67) become the high energy limit of eq. (71).
3.3 vertices
Along the hard fluctuation, the spin of the hard particle may be changed by -dependent vertices which involve soft background fields. Since we are discussing renormalizable field theory, the -dependent vertices from derivative interactions must be linear in . In massless gauge theory, they are eliminated by choosing appropriate light-cone and gauges ArkaniHamed:2008yf; Elvang:2013cua. However, the massive vector bosons don’t have such degrees of freedom to eliminate these -dependence. In renormalizable field theory, there are two classes of vertices that cannot be eliminated by gauge fixing in the massive case: triple vector coupling () and Vector-Vector-Scalar () interaction. For simplicity, we only consider massive vector bosons that have equal mass.
Since all on-shell 3-point amplitudes in the Standard Model have been figured outChristensen:2018zcq, we can translate them into the Feynman rules to find out the vertices. Then we deform these vertices to give their -dependence explicitly.
There are two kinds of possible amplitudes (see figure 1). Since amplitude has three vector external legs, the vertex should be a order-3 Lorentz tensor , which refers to attached with three line segments in the diagrammatic expression. One kind of vertex includes two massive vectors and one massless vector. Its diagrammatic representations are
(72)
where the factor is introduced by Arkani-Hamed:2017jhn, which carries helicity. The vertex in the Standard Model belongs to this kind of vertex.
Another kind of amplitude includes three massive vectors. Its diagrammatic representation is
(73)
There is no such vertex in the Standard Model, because and bosons have different masses.
Now we give the expression for the vertex ,
(74)
where . It is easy to check this expression by dotting it into vector boson polarizations. This manipulation will give the amplitudes in eqs. (72) and (73) again. We can use shifted momenta and instead of and to deform this vertex. Diagrammatically, shifted vertex is represented as
(75)
where . In the diagram, the three lines correspond to three Lorentz indices of the vertex. Two horizontal lines represent hard fluctuations, so the momenta they carry should be shifted (e.g. and in eq. (75)).
Next, amplitude (see figure 2) will give a simpler vertex. Since the only one external vector boson contributes one Lorentz index, the vertex must be a Lorentz vector. Suppose the momenta of scalar bosons are and , the vertex will be
(76)
This vertex can be realized in various BSM models, such as 2HDM, MSSM and the simplest Little Higgs model Gunion:1989we; He:2017jjx. In the last case, there is a Higgs-Goldstone mixing term in the non-linear Lagrangian.
Notice that the vertex seems to be a substructure of vertex. Diagrammatically, this means
(77)
Therefore, the large- behavior of vertex is contained in vertex.
4 Large- behavior of -point amplitudes
For massless amplitudes, appropriate gauges eliminate all large- contributions from derivative interactions except in the so-called “unique diagrams". Therefore, the large- behavior of amplitudes depends on the number of hard propagators. If there are more hard boson propagators, the amplitudes will be suppressed by higher powers of . As for hard fermion propagators, the contribution can be simplified by using anti-commuted gamma matrices and . At last, the hard polarizations are dotted into the sum of all contributions.
In the massive case, we have seen that the vertices are not eliminated completely, so the diagrams with hard vector propagators cannot be ignored. Thus, the large- behavior in the massive case should be evaluated directly instead of estimated.
One may wonder if the Goldstone boson equivalence theorem Lee:1977eg; Chanowitz:1985hj is useful in the complex deformation. In the high energy limit, this theorem treats the longitudinal modes of vector bosons as Goldstone bosons to simplify calculations. However, the large- limit is not exactly the same as the high energy limit. Actually, it has been used in amplitudes under the all-line shift Cohen:2010mi, but there are also many cases where it doesn’t work. Consider the following amplitude in the large- limit with particles 1 and 2 shifted:
(78)
where is the mass of scalar bosons and is the energy of particle 3. Particle 3 is unshifted, so is no longer a negligible quantity. If we insist on applying this expansion, we should sum over the contributions from infinite terms. Thus, we won’t use the Goldstone boson equivalence theorem in our analysis.
4.1 -point Vector Boson Scattering Amplitudes
In the massive case, the large- behavior of amplitudes depends on massive vector bosons, so we can first consider vector boson scattering. Besides particle 1, we set all particles to be massive. The massive vector propagator (56) has two terms. The first term connects vertices while the second term splits the diagram in two. Therefore, the amplitudes can be split into two parts,
(79)
where and correspond to the connected and disconnected diagrammatic expressions respectively.
When a hard particle shoots through a soft background, it will interact with the classical field more than once. The soft physics is parameterized by currents , so the amplitude becomes
(80)
where the second sum is over all permutations of the labels . Here is the denominator of shifted propagators. After permutation, it becomes
(a) Figure 3: The diagram without hard propagators. The blob represents the soft background.
The first term in eq. (80) corresponds to the diagram without hard propagators (see figure 3). Since the hard particle interacts with the classical field only once, the large- behavior of the numerator should be given by a Lorentz vector . Inserting polarizations (67) and (66), it becomes
(82)
The leading diagrams are equal to zero. We used and to cancel the subleading diagrams in and . It shows that the first term in eq. (80) vanishes in the large- limit.
Figure 4: The diagrams that give the leading contributions in the large- limit. The blobs correspond to soft backgrounds.
As for the second term in eq. (80), the large- behavior is given by a Lorentz tensor . Since this term includes more than three external particles, 4-vertices should be considered. We use a tensor to represent a 4-vertex. Since there are three ways to contract four polarization vectors, the diagrammatic expression of a 4-vertex should be
(83)
where , and are arbitrary coefficients. If we use one 4-vertex instead of two 3-vertices, one propagator will decrease. Basically, the more 4-vertices diagrammatic expressions have, the lower order they are. The Lorentz tensor can be expanded as
(84)
where the superscript is the number of 4-vertices. Only the first two terms and (see figure 4) give the same contributions in . Their diagrammatic expressions are
(85)
(86)
where the script denotes the position of the 4-vertex. Now we also evaluate these connected diagrammatic expressions in specific helicity and spin states.