Four external lines require four cubic interaction vertices, since each such interaction vertex contains only one heavy field. Their other eight endpoints must form four light internal lines. A connected Feynman diagram therefore has and : the leading contribution is a one-loop box diagram, of order , with four scalar propagators. Charge-two incoming and outgoing heavy lines attach alternately around the box, as shown in panel (d) of b's figure. Each incoming heavy line supplies the two outgoing light-charge arrows at its corner, and each outgoing heavy line absorbs them. With external labels fixed there is one such box and its symmetry factor is one.
For , the external energies are also small compared with , so expand the loop integrand in all external four-momenta. Its leading term is
A Wick rotation and the Euclidean massive loop integral give
Hence, more precisely than just its requested scaling,
The mass dimension makes dimensionless, as a four-dimensional scattering amplitude should be. The leading amplitude has no extra identical-particle multiplicity.
Take with the Mandelstam variables of i. The allowed tree-level Feynman diagram is the crossed -channel exchange in panel (c) of b's figure. At one end the incoming particle and outgoing antiparticle transfer charge ; the other end absorbs that charge. The zero-charge - and -channel assignments are forbidden by the interaction vertices. Thus
In the center-of-mass frame with , and . For , both scattering amplitudes approach . The elastic two-body Lorentz-invariant phase space and incident flux agree for the two experiments. Using
for distinguishable final particles, and an extra factor when integrating the identical-particle final state over the whole sphere, gives
There is no additional division for identical incoming particles in the definition of a beam scattering cross-section. Before taking , the zero-momentum ratio is , provided the threshold is away from the heavy pole.
Near , the bare -channel scalar propagator makes fixed-order perturbation theory fail: repeated self-energy insertions contain powers of . The channel is open for , giving the leading decay width
Resumming the self-energy gives a finite Breit-Wigner resonance cross section, with denominator near the pole, rather than an infinite scattering cross-section. Also, with lies outside the assumed nonrelativistic incoming kinematics; the ratio is not a resonance prediction.
Label the incoming four-momenta and the outgoing ones . Define the Mandelstam variables , , . Two incoming particles have conserved charge , so the only tree-level Feynman diagram is an -channel exchange, shown in panel (b) of b's figure. Each end converts two charge-one lines into one charge-two line. There are no - or -channel exchanges: an incoming and an outgoing particle at such an end would transfer zero charge, for which this theory has no cubic interaction vertex.
The single Feynman diagram is already symmetric under exchanging the two outgoing particles. This does not add another copy of the scattering amplitude; the identical final-state symmetry factor enters the integrated scattering cross-section instead.
Use incoming fermion and antiparticle , outgoing fermion and antiparticle . The two tree-level Feynman diagrams are the channel, with , and the channel, with . They occupy the lower row of the figure. In the channel the fermion flow follows and ; in the channel it follows at the incoming interaction vertex and at the outgoing interaction vertex.
The two contractions again have a relative fermionic sign. With the same harmless common-phase convention as i, their sum is
The auxiliary-field propagator is the same constant in both channels. There is no channel preserving the external fermion number assignments.
Label the incoming fermion momenta and outgoing fermion momenta , with . There are two tree-level Feynman diagrams: the channel connects to and to , with exchanged momentum ; the channel connects to and to , with . Both internal lines are auxiliary-field propagators. The upper row in the figure gives these labelled connections; the outgoing labels are exchanged in its second panel to avoid a visually unnecessary crossing.
Each tree-level Feynman diagram contains two factors and one factor , giving . The exchange of the two identical outgoing fermions gives the relative fermionic sign. Up to a common external-state phase, the corresponding scattering amplitude is
There is no annihilation channel: both incoming particles carry the same fermion number, whereas a interaction vertex couples one fermion to one antiparticle when both are incoming.
To include all momentum-dependent contributions to the two-point function, put one external line at each interaction vertex. Let be the number of lines connecting the vertices, and their numbers of tadpole diagrams. For vertex valences ,
The integer is odd, and gives a one-particle-reducible Feynman diagram and is excluded. Thus the complete list is
The coefficient of the mixed graph includes the reflected external attachment as well as its reciprocal Feynman-diagram symmetry factor. If both external lines attach to the same interaction vertex, no external momentum enters the connecting internal lines, and that Feynman diagram cannot contribute to wave-function renormalization.
For incoming external momentum , assign the connecting lines momenta . Define
The momentum space Feynman rules give, in the order listed,
Each is a reciprocal Feynman-diagram symmetry factor, including any equivalent external attachments. There are respectively independent loop momenta.
If “contribute to wavefunction renormalization” means requiring a divergent kinetic counterterm after subtraction of all subdivergences, only the last topology qualifies. The first three still have nontrivial external momentum dependence and give finite contributions to a normalization condition; their raw divergent factors arise from mass counterterms and interaction vertex counterterms. Keeping all four here makes this distinction explicit. The ultraviolet divergence analysis in d establishes it directly.
With a single quartic interaction vertex, the two unused line ends close into one tadpole diagram. With a single sextic interaction vertex, the four unused ends pair into two tadpole diagrams. These are the only one-vertex one-particle irreducible Feynman diagrams for the two-point function.
Define the Minkowski Feynman integrals
The amputated two-point function contributions are
where are the reciprocals of the corresponding Feynman-diagram symmetry factors. Both are independent of external momentum, and hence affect the mass counterterm rather than wave-function renormalization. No numerical Feynman-diagram symmetry factors are needed here.
For , the sole tree-level Feynman diagram has a Yukawa coupling interaction vertex producing an off-shell scalar field particle of four-momentum , followed by its cubic scalar field interaction vertex. Its invariant decay amplitude is
Two successive Yukawa coupling interaction vertices would change the fermion species twice and leave a , so they cannot produce this final state. Interchanging the two identical outgoing bosons does not create another Feynman diagram; the cubic factorial is already included in the Feynman rule. The integrated decay width instead includes the identical final-state symmetry factor .
For , the sole tree-level Feynman diagram is a single flavour-changing Yukawa coupling interaction vertex, with . Its invariant decay amplitude is
The solid arrow passes from the incoming to the outgoing ; the dashed outgoing line carries .
For , there are two tree-level Feynman diagrams, both with an internal fermion. Their internal four-momenta are and . The scattering amplitude is
These are the two possible absorption/emission orderings along the fermion flow. There is no third tree-level Feynman diagram using the cubic scalar field interaction vertex, since attaching its remaining scalar field line to the external line would require a forbidden interaction vertex.
Label the external four-momenta by . The only tree-level Feynman diagram exchanges a scalar field particle between the transitions and . Its internal four-momentum is . Thus its scattering amplitude, with a consistent external-state ordering, is
It is the crossed, or -channel, exchange with these labels. A direct -channel scalar field exchange would need a flavour-preserving Yukawa coupling, which the Lagrangian density does not contain. There is no annihilation Feynman diagram for two incoming fermions.
Let label the incoming and outgoing fermion, and the incoming and outgoing antifermion. The bottom row of the figure gives the two tree-level Feynman diagrams:
Thus the Mandelstam variables in the two denominators are
The two scattering amplitude terms have a relative fermionic sign of minus when written with spinor products and . The overall phase depends on the chosen ordering of the external states. No channel joins an incoming fermion to an outgoing antifermion through a species-preserving Yukawa interaction.
With all four external fermions of species 1, both final attachments are possible. The middle row of the figure gives the two tree-level Feynman diagrams:
  • The channel has lines and , joined by a scalar propagator with .
  • The channel has lines and , joined by a scalar propagator with .
For a fixed ordering of external fermion states, the fermionic sign from exchanging the final identical fermions gives
Here are Dirac spinors. In particular, the scattering amplitude changes sign when the complete final labels and are exchanged, as required for identical fermions.
Label the initial species-1 and species-2 fermions by and , and the final species-1 and species-2 fermions by and . There is exactly one tree-level Feynman diagram: the -channel scalar propagator joins the continuous species-1 line to the continuous species-2 line . Its exchanged four-momentum determines , in the notation of Mandelstam variables:
The top row of the figure displays this Feynman diagram. A crossed attachment would change species at an interaction vertex and is forbidden. There is no initial antifermion for an annihilation channel.
Let . Inverse images commute with complements and countable unions, so this is a sub-sigma-algebra of . The random variable is -measurable.
The assumption says that is a measure-preserving transformation. Consequently has the same probability distribution as and is integrable. It also implies for every integrable random variable , first for indicator functions, then for simple functions, and then by positive and negative approximation.
To identify a conditional expectation, take any . The assumed identity with the bounded random variable gives
Every member of has the displayed form , so this verifies all defining conditional expectation identities. Therefore
Apply part (iii), since this conditional expectation has the same probability distribution as . We conclude
This is invariance from integral duality for a measure-preserving transformation; neither invertibility of nor a second moment of is required.
We first establish the lower truncation identity as well. The pair has the same hypotheses as , because conditional expectation is linear and negation preserves equality of probability distributions. Apply part (ii) to this pair and multiply by to get
For every real ,
Using this identity, part (ii), the lower truncation identity, and linearity of conditional expectation gives
In particular the argument does not assume that a function obtained by scalar clipping is a convex function or a concave function.
Write . The bounded random variables and have the same probability distribution, and the displayed identity identifies the latter as the conditional expectation of the former. They are square-integrable, so part (i) gives almost surely. Take and discard the countable union of exceptional null sets. At every remaining outcome, let to obtain
Thus conditional expectation preserving a distribution forces equality even when only integrability is assumed.
The function is a concave function, and is integrable since . The concave form of the conditional Jensen inequality gives
The random variables and have the same probability distribution, so . Using the law of total expectation, the nonnegative difference between the right and left sides of the inequality has zero expectation. It must vanish almost surely. Therefore
The equality comes from the distributional assumption as well as the conditional Jensen inequality.
Assume is square-integrable. Since and have the same probability distribution, is also square-integrable and . The product is integrable by the Cauchy-Schwarz inequality.
We use the defining test-function identity for conditional expectation: for bounded -measurable , . Take and let . The dominated convergence theorem, with bounds and , gives
This is the extension of the conditional expectation pull-out identity to this square-integrable multiplier. Consequently
A nonnegative random variable with zero expectation is zero almost surely. Hence
Geometrically, this is the equality case in the L2 space orthogonal projection interpretation of conditional expectation.
The bounded-time optional sampling theorem says that for an integrable discrete-time martingale and a bounded stopping time , . It requires no assumption about an eventual unbounded stopping time.
Apply this theorem with , or simply take the expectation of the stopped martingale in part (ii). For we obtain
Since ,
The monotone convergence theorem now yields . In particular almost surely, which has been established without assuming that optional stopping is valid at .
For each sample path with finite , eventually equals . Also . The dominated convergence theorem gives . Taking the limit of a sequence in the earlier identity therefore proves
For example, initial holdings give expectation one, as the first transfer must eliminate its giver. This computes the first elimination time, rather than the later time when one player owns all tokens.
After , the token product remains zero, whereas the deterministic term continues to increase. Combining this with the pre-elimination calculation in part (ii),
The stochastic process is adapted and integrable at each fixed time, and the displayed conditional expectation is nonnegative. Thus is a submartingale.
For it is neither a martingale nor a supermartingale. If , there is positive probability that the first player transfers a token to the second in each of the next rounds. Until the last such round all three remain active, so this event has probability and entails . At time the displayed conditional expectation is therefore strictly positive on an event of positive probability, ruling out both the zero drift required of a martingale and the nonpositive drift required of a supermartingale. If an initial holding is zero, and the same conclusion follows immediately. The degenerate case gives a constant martingale.

There are unlisted articles, also show them or only show them.