Ponente
Descripción
In general relativity, cosmic acceleration is a statement about geometry---the timelike defocusing of the comoving congruence---and the field equations source it with mass--energy combinations, not with any equation-of-state ratio. Evolving dark energy is now back at the center of cosmology, and several scenarios under active discussion---sign-switching $\Lambda_{\rm s}$CDM and AdS-to-dS transitions, which simultaneously ease the $H_0$ and $S_8$ tensions and restore $\sum m_\nu$ to physical values, as well as modified-gravity backgrounds recast in GR-like form---feature an effective dark-energy density that is negative in the past and crosses zero at some redshift $z_\dagger$. Our standard language quietly assumes this never happens: $w_{\rm de}<-1/3$ signals repulsion only while $\rho_{\rm de}>0$ and the inequality reverses on the negative branch; the phantom divide $w_{\rm de}=-1$ stops separating anything once the density changes sign; at $\rho_{\rm de}=0$ the ratio $w_{\rm de}$ does not exist, though nothing physical is singular there; and sign-preserving parametrizations such as $w_0$--$w_a$ cannot represent the crossing at all. In this talk, based on arXiv:2607.18008, we ask what defines dark energy when $w$ cannot, and answer with the two combinations the field equations actually use: the active gravitational mass density $\mathcal{M}=\rho+3p$, which drives Raychaudhuri defocusing, and the inertial mass density $\mathcal{I}=\rho+p$, whose zero---the null energy condition (NEC) boundary---replaces the phantom divide as the regular separator. Sharp results follow: any smooth sign switch is phantom-like and repulsive at the crossing, with the total NEC intact; the pole in $w_{\rm de}$ is purely kinematic, with universal residue $n(1+z_\dagger)/3$, so $w_{\rm de}=-1$ is never attained and the two sides of the divide connect only through the pole; repulsion begins strictly before the switch, at $z_{\rm rep}>z_\dagger$; an exact window in $\rho_{\rm de}'(z_\dagger)$ permits acceleration at the crossing itself; and the deceleration parameter admits exactly one or three sign-changing zeros. An analytic $\Lambda_{\rm s}$CDM profile, exponential infrared $f(T)$ gravity, and the minimal phantom brane realize the pattern. The practical moral for reconstructions: across $\rho_{\rm de}=0$ the regular targets are $(\rho_{\rm de},p_{\rm de},\mathcal{I}_{\rm de},\mathcal{M}_{\rm de})$---kinematic totals need no dark-sector split---and parametrizations should be allowed to change sign.
