Physicists at CERN's LHCb collaboration reported a four-sigma anomaly in the rate of a rare B-meson decay channel — the so-called "charming penguin" diagram — in a paper posted to arXiv on April 28 and previewed at a CERN seminar Friday morning. [1] The measured branching fraction of B-to-kaon-plus-pion-plus-two-muons, in the regime where the muon pair sits below the J/psi resonance, is 0.27 of the Standard Model prediction with a discrepancy of 4.0 standard deviations. The probability of a statistical fluke producing the result is roughly one in sixteen thousand. [2]

The result lands the same week the paper's May 1 account of the National Science Board's firing and the lost-science ledger becoming a thread of its own framed a federal-science-decommissioning week; CERN is not a US institution, but the contrast between an experiment whose multi-year discovery arithmetic is its operating premise and a domestic agency whose board can be dismissed in a single executive action is the day's structural pairing.

The "charming penguin" name comes from the Feynman diagram's shape — an internal charm-quark loop that looks, to physicists in the right state of mind, like a penguin — and from the charm-quark virtual state that mediates the decay. The diagram is one of the cleanest places to look for new physics because the Standard Model prediction is theoretically rigorous; deviations point at unaccounted particles or forces in the loop, not at calculation uncertainty. The 0.27-of-prediction measurement is in the direction physicists call "destructive interference," consistent with a virtual particle reducing the decay amplitude. [3]

Four sigma is not five. The high-energy-physics community treats five-sigma as the discovery threshold; 4.0 sigma is "evidence." The community's institutional memory is also long enough to remember the 750 GeV diphoton bump that ATLAS and CMS reported in December 2015 at three to four sigma and that disappeared in the 2016 data. Sean Carroll, posting on X Friday, named the precedent: "Three more sigma to go before we get the champagne." LHCb's own analysis paper acknowledges the precedent in its statistical-treatment section. [4]

The institutional context is that the analysis used roughly one-third of the data the collaboration has already collected. The Run 3 dataset, which closes in late 2025 and 2026, is approximately 2.5 inverse femtobarns; this paper analyzes 0.9. The remaining 1.6 inverse femtobarns will be folded into a follow-up analysis with a target completion before the December 2026 winter shutdown. If the four-sigma result holds at three times the dataset, it crosses the five-sigma threshold by simple arithmetic — the significance scales as the square root of the integrated luminosity. The same arithmetic in reverse is the discipline check: if the result is a fluctuation, it will weaken as the data accumulate. [5]

The HL-LHC upgrade, scheduled for 2030 commissioning, will increase the dataset by another factor of fifteen. Whether this anomaly hardens or softens between now and then will be the cleanest single test of physics beyond the Standard Model since the Higgs discovery in 2012. The list of theoretical extensions that produce destructive interference in the charming penguin channel is short and well-defined: a new Z-prime gauge boson at the TeV scale, a leptoquark coupling muons to down-type quarks, or a charged Higgs from supersymmetric extensions. Each scenario has different cross-checks in other LHCb channels. The collaboration is now running those cross-checks in parallel. [6]

The political register sits adjacent. The May 1 paper documented the firing of the U.S. National Science Board and the federal-science-decommissioning ledger now formalized as a lost-science thread. The HL-LHC upgrade is a CERN program; U.S. participation through the Department of Energy's high-energy-physics division is funded under a budget that the FY27 budget request cuts by 11%. Brookhaven and Fermilab personnel embedded in LHCb total approximately 240 scientists. If the U.S. cut goes through as proposed, the confirmation analysis on this anomaly — three years out — will run with a smaller U.S. delegation. The discovery, if it is one, will be a discovery the U.S. helped collect data for and not the discovery the U.S. helped confirm. [7]

The CERN seminar Friday morning was attended in person by approximately 180 physicists, with another 1,400 watching the livestream. The Q&A was unusually short for a result of this magnitude — the questions clustered around the systematics treatment and the bin-by-bin agreement, both of which are LHCb's traditional strengths. The paper's lead author, Marina Artuso of Syracuse University, answered each question with the data slide already on screen. The institutional posture in the room read as the posture of a collaboration that has done this analysis correctly and is waiting for the data to either confirm it or take it back. The wait is now three times the existing dataset long. [8]

-- KENJI NAKAMURA, Tokyo