Search for the decay $B^0\to \gamma \gamma$

We report the result of a search for the rare decay $B^0\to \gamma \gamma$ in $426{\mathrm{fb}}^{-1}$ of data, corresponding to $226\times {10}^6$ $B^0{\overline{B}}^0$ pairs, collected on the ${\rm Y}(4S)$ resonance at the PEP-II asymmetric-energy $e^{+}{e}^{-}$ collider using the BABAR detector. We use a maximum likelihood fit to extract the signal yield and observe ${21}_{-12}^{+13}$ signal events with a statistical significance of $1.8\sigma$. This corresponds to a branching fraction $\mathcal{B}(B^0\to \gamma \gamma )=(1.7\pm 1.1(\mathrm{stat}.)\pm 0.2(\mathrm{syst}.))\times {10}^{-7}$. Based on this result, we set a 90% confidence level upper limit of $\mathcal{B}(B^0\to \gamma \gamma )<3.2\times {10}^{-7}$.

Figure 1

Examples of lowest order SM Feynman diagrams for $B^0\to \gamma \gamma$. The symbol $q$ represents a $u$, $c$, or $t$ quark. In some new physics scenarios, the $W$ boson may be replaced by a charged Higgs particle.

Figure 2

(a) ${\pi }^0$ likelihood ratio for $B$ candidate photons in simulated signal events (open histogram) and simulated continuum background events (shaded histogram) where the photon is required to originate from a ${\pi }^0$ decay. (b) $\eta$ likelihood ratio for $B$ candidate photons in simulated signal events (open histogram) and simulated continuum background events (shaded histogram) that are required to originate from an $\eta$ decay. Events where both $B$ candidate photon likelihood ratios are less than 0.84 are selected, as denoted by the arrows.

Figure 3

Output of the neural network on the validation samples of simulated signal (open histogram) and background (shaded histogram) events. Selected events have a NN response greater than 0.54 as denoted by the arrow.

Figure 4

Projections of the ML fit onto $m_{\mathrm{ES}}$ and $\Delta E$. (a) The projection of the $m_{\mathrm{ES}}$ component when the range of $\Delta E$ has been restricted to $-0.30\le \Delta E\le 0.13\mathrm{GeV}$. (b) The $\Delta E$ projection of the fit when the range of $m_{\mathrm{ES}}$ has been restricted to $m_{\mathrm{ES}}>5.27\mathrm{GeV}/c^2$. The points represent the on-resonance data. The solid curve represents the total PDF, the dashed curve is the continuum background component, the dot-dashed curve is the signal component, and the long-dashed curve is the peaking background component. With an expected yield of approximately one event, the peaking background component is nearly indistinguishable from the $x$ axis.

Figure 5

Likelihood curve from the maximum likelihood fit, as a function of the signal yield, after convolution with a Gaussian shape whose width is equal to the total systematic error. The shaded region corresponds to the integral of the curve up to 90% of its total area. The $y$ axis, $L/L_0$, is the ratio of the likelihood function for a given $N_{\mathrm{sig}}$ to the maximum likelihood.