We propose a novel explanation for the dilepton excess observed in heavy ion collisions at invariant masses below 1 GeV. We argue that the presence of local parity breaking due to a time-dependent isosinglet and/or isotriplet pseudoscalar condensate substantially modifies the dispersion relation of photons and vector mesons propagating in such a medium, changing the ρ spectral function and resulting in a potentially large excess of e + e − and µ + µ − with respect to the theoretical predictions based in a 'cocktail' of known hadronic processes. Possible signatures to prove or disprove this effect are discussed.PACS numbers: 25.75.Cj, 21.65.Jk, 12.40.Vv During the last decade several experiments in heavy ion collisions have indicated an abnormal yield of lepton pairs with invariant mass M < 1 GeV in the region of small rapidities and moderate transversal momenta [1,2] (reviewed in [3,4]). This effect is visible only for collisions that are central or semi-central, and is present for e + e − and also for µ + µ − . From a comparison to pp and p-nucleus collisions it has been established beyond doubt that such an enhancement is a nuclear medium effect [3]. experiments) the abnormal dilepton yield has not been yet explained satisfactorily by known processes in hadronic physics [3,4]. Following [2] we divide the range of dilepton invariant masses into high ( M > 3.2 GeV), low ( M < 1.2 GeV) and intermediate. The low mass region (LMR) is in turn divided into LMR I with M < 0.3 GeV and LMR II with 0.3 GeV < M < 1.2 GeV. In LMR I the enhancement could possibly be explained by modifications of meson properties in nuclear medium [6][7][8][9] as well as by protonneutron scattering [5]. But in the LMR II the ρ meson, directly via ππ fusion or indirectly through Dalitz processes, largely dominates and the in-medium effects of a dropping mass and/or broadening resonance seem unable to explain the spectacular dilepton enhancement by a 4 to 7 factor, depending on p T and centrality (see [10,11] for recent attempts).In this letter we propose a radically different explanation of this enhancement. We suggest that the effect may be a manifestation of local parity breaking (LPB) in colliding nuclei due to the generation of a pseudoscalar, isos- * Corresponding author. E-mail: andrianov@icc.ub.edu inglet or neutral isotriplet, condensate whose magnitude depends on the dynamics of the collision. It has been suggested that such a background could be due to the topological charge fluctuations leading to the so-called Chiral Magnetic Effect (CME) [12] studied by lattice QCD simulations [13] and seemingly detected in the STAR and PHENIX experiments at RHIC [14], although the issue is far from being settled. It might be also related to pseudoscalar domain walls [15]. However the fact that the observed dilepton excess is almost absent for peripheral collisions (where the CME should be more visible) and maximized in cental collisions makes us believe that it may be due to the ephemeral formation of a bona-fide thermodynamic phase where pa...