Charge and CP symmetry breaking in two Higgs doublet models

The new content is real but narrow. They take their 2004 CP-conserving proof (10 parameters) and push it to the full 14-parameter potential with explicit CP violation. Two things are worth having: eqs (14) and (17), which say the depth gap between a normal minimum and any CB stationary point is (M_H±^2 / 2v²) times a positive bracket, and the observation that this gap is controlled by the mass of the scalar tied to the broken symmetry. That last point is genuine physics, not just algebra, and it is the part I would actually remember. The invariant x_i basis is the right way to set this up, and the CB stationary point's uniqueness falls out cleanly from A + BY = 0. I checked the chain: stationarity at N1 (eq 7), the form of V' in eq 8, X^T V' = 0, A + BY = 0, symmetric B. Eq (14) closes, and eq (17) is the same manipulation at N2 and also closes. The stress-test worry about a degenerate bracket does not land: the bracket vanishes only when alpha = 0 and (v'_1, v'_2) is proportional to (v₁, v₂), which is not a charge-breaking point at all. For any genuine CB stationary point the inequality is strict. Where I would push back is the step from 'V_CB > V_N1' to 'the CB point is a saddle'. That runs through B⁻¹ not being positive definite, and it leans on the earlier paper's claim that B fixes the nature of the stationary point. It is plausible and probably right, but it is the one place where the reader is trusting ref [1] rather than this text; a sentence spelling out the Hessian-to-B map at the CB point would close it. The honest gap is eq (18): the sign of V_N2 - V_N1 is undetermined, and they say so plainly. That means the CP half of the title is only settled for the CP-conserving potential (eq 19), so the title reads broader than the result. Minor: the abstract says 'cannot occur' without 'at tree level' while the introduction has the qualifier, and eq (9) is imported from ref [1], which needed an erratum, so a footnote confirming the corrected form would help. This is for 2HDM vacuum-stability people and model builders who want to stop worrying about CB vacua, not a broad-audience paper. Send it out; referees should focus on the saddle-point step and the abstract wording.

The central claim, eq (14), is a clean algebraic identity: V_CB - V_N1 = (M^2_H+- / 2 v^2) [(v'_1 v_2 - v'_2 v_1)^2 + alpha^2 v_1^2], positive whenever N1 is a minimum. I re-derived the chain (stationarity conditions eq (7), the form of V' in eq (8), X_N1^T V' = 0, A + B Y = 0, and the symmetric-B manipulation in eqs (12)-(13)) and it closes. The N2 analogue, eq (17), follows the same pattern and also closes. So the depth comparison itself is not where the concern lands. The load-bearing soft spot is the step from 'V_CB > V_N1' to 'the CB stationary point is a saddle point'. The paper argues: eq (14) gives Y^T V' > 0; eq (15) rewrites this as -V'^T B^{-1} V' > 0, i.e. V'^T B^{-1} V' < 0 for one particular vector V'; hence B^{-1}, and B, is not positive definite. That inference is valid only if V' is nonzero, which requires M^2_H+- > 0 and a nondegenerate bracket. If the bracket in eq (14) vanished (e.g. v'_1 v_2 = v'_2 v_1 and alpha v_1 = 0), eq (14) would give only V_CB = V_N1, not strict positivity, and the sign argument for B would not go through. The paper does not discuss this degenerate case, nor whether the CB and N1 stationary points could coincide in field space. This is a narrow gap, not a refutation, but it is the one place where the 'saddle point' conclusion is not fully closed by the displayed algebra. Separately, the abstract's [abstract:s0] 'charge breaking cannot occur' is unqualified, while the body is explicitly tree-level; the authors do not flag this in the abstract, and the classifier records no stated limitations. That is a presentation issue rather than an error in the derivation.

The reader's strongest_claim correctly identifies eq (14) as the load-bearing result, and I agree it is correct: the derivation from the stationarity conditions through the symmetric-B manipulation is sound, and the N2 case eq (17) mirrors it. Where I part company with the reader is the choice of weakest_assumption. The reader flags (i) the tree-level qualifier missing from the abstract and (ii) the imported charged-scalar mass matrix eq (9) from ref [1], which had an erratum. On (i): the body does say 'at the tree-level' in the introduction, so this is a wording issue in [abstract:s0], not a gap in the argument; it does not threaten the central claim. On (ii): eq (9) is a standard mass-matrix expression quoted from a published source, and the paper explicitly notes it does not depend on the number of parameters; per the review rules, a cited published result is supported, and the erratum is not evidence the quoted matrix is wrong. The genuinely unresolved step is the saddle-point inference: 'V'^T B^{-1} V' < 0 for one vector V'' establishes B is not positive definite only when V' is nonzero, and the paper does not address the degenerate bracket in eq (14). That is a real, narrow soft spot the reader did not name. It does not overturn the headline CB-depth result, so I would leave the verdict at CONDITIONAL/UNCHANGED; the concrete test above would settle whether the saddle conclusion needs a nondegeneracy caveat. The authors' own acknowledged limitation (eq (18) sign undefined, so N1 vs N2 depth unsettled for the CP-violating potential) is correctly stated in the text and is not a new finding.

If correct, the result is a useful and compact statement about the vacuum structure of the general 2HDM: it removes the need for a numerical scan to exclude CB vacua once a normal minimum is known, and it gives a transparent physical interpretation, namely that the depth of a symmetry-breaking stationary point is controlled by the mass of the scalar associated with the broken symmetry. The generalization from the 10-parameter CP-conserving potential of Ref. [1] to the 14-parameter case is not entirely trivial, since the explicit CP-violating terms a4, b14, b24 and b34 enter the stationarity conditions. The paper is a letter and the result is tree-level; within that scope it is a worthwhile contribution. I note that the appearance of M^2_H± in Eq. (14) is not circular: the identity is algebraic and positivity follows from the minimum condition, as the authors argue.

Two Higgs doublets are the simplest extension of the Standard Model's scalar sector, and the theory admits vacua of several kinds: some preserve electric charge, some break it by giving the photon a mass, and some break CP. The question the paper takes up is whether the vacuum we actually live in could be unstable, decaying by tunnelling into a charge-breaking state. The authors show that whenever a charge-preserving minimum exists, every charge-breaking stationary point lies above it, with the gap fixed by the charged Higgs squared mass, so the charge-breaking point is a saddle rather than a competing vacuum. The same bookkeeping, applied to a CP-conserving potential, ties the gap between a CP-breaking and a CP-preserving stationary point to the pseudoscalar mass. What the paper does not settle is which of the two charge-preserving minima is deeper when both exist.

What carries the argument. The carrying device is a change of variables from the eight real scalar fields to the four $SU(2)_W×U(1)_Y$ invariants $x_1,x_2,x_3,x_4$, which turns the potential into $V=A^TX+½X^TBX$ and makes the stationarity conditions linear in $X$. That linearity is what lets the authors compare the potential's value at two different stationary points without ever solving for the vacuum expectation values explicitly: the gradient at one point, contracted with the field vector of the other, produces the depth difference directly. The charged-Higgs mass matrix, imported from the authors' earlier paper, supplies the positive coefficient that fixes the sign.

Core claim. The paper extends the authors' earlier proof, which covered the CP-conserving two-Higgs-doublet potential with ten independent parameters, to the fully general potential with explicit CP violation, written in the basis of four invariants $x_1=|Φ_1|^2$, $x_2=|Φ_2|^2$, $x_3=Re(Φ_1^†Φ_2)$, $x_4=Im(Φ_1^†Φ_2)$ with a vector $A$ and matrix $B$. At a charge-preserving minimum $N_1$ the stationarity conditions force the gradient $V'=A+BX_{N_1}$ into a form whose components have correlated signs, and the charged-Higgs mass matrix then gives $M^2_{H^±}=-V'_3 v^2/(2v_1v_2)$. At a charge-breaking stationary point the gradient vanishes identically, so that point is unique and satisfies $A+BY=0$. Combining the two, the authors obtain $V_{CB}-V_{N_1} = (M^2_{H^±}/2v^2)[(v'_1v_2-v'_2v_1)^2+α^2v_1^2]$, strictly positive at a minimum, and a parallel identity for the second charge-preserving minimum $N_2$ with an extra $δ^2v'^2_2$ term. The same algebra also shows $B$ cannot be positive definite when such a minimum exists, which is what makes the charge-breaking point a saddle. For the CP-conserving special case the difference between the spontaneous-CP-breaking and normal stationary points reduces to $(M^2_A/2v^2)[(v''_1v_2-v''_2v_1)^2+δ^2v_2^2]$, proportional to the pseudoscalar mass.

The premise it rests on. The whole argument is a tree-level computation, and the abstract's flat statement that charge breaking cannot occur omits that qualifier, which the introduction does state; the result also leans on the charged-Higgs mass matrix imported from the authors' earlier paper, a paper that required an erratum whose content is not described here.

Any two-Higgs-doublet parameter set that has a charge-preserving minimum is automatically safe from tree-level charge breaking, so vacuum-stability scans need only check the charge-preserving candidates.

A practical sufficient condition against charge breaking is to choose the quartic couplings so that the matrix $B$ is not positive definite.

The depth gap between a symmetry-breaking stationary point and the normal minimum is set by the mass of the scalar tied to that symmetry: the charged Higgs for charge, the pseudoscalar for CP.

For the CP-conserving potential, a normal minimum with positive $M^2_A$ always lies below the spontaneous-CP-breaking stationary point, so spontaneous CP breaking at tree level cannot be the true vacuum there.

The structural pattern suggests a general rule worth testing elsewhere: for each spontaneously breakable symmetry of a scalar sector, the depth penalty of breaking it equals the mass of the scalar that would be eaten or rotated, times a geometric factor measuring how far the two vacua are apart in field space.

Because the sign of $V_{N_2}-V_{N_1}$ is left undetermined, a numerical scan of the 14-parameter potential looking for parameter sets where both charge-preserving stationary points are genuine minima would settle whether the ambiguity is real or empty.

The tree-level character of the result invites a one-loop check: if the charged-Higgs contribution to the effective potential is large, the ordering of the two stationary points could in principle reverse, which would matter for early-universe tunnelling estimates.

The same invariant-basis bookkeeping could be applied to three-doublet or singlet-extended scalar sectors, where the number of stationary-point types grows and the linear structure of the stationarity conditions may or may not survive.

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review