Munim Technologies

Research/Physics

Old LEP data rule out a boson behind the ALEPH 30 GeV excess

Measurements published in the 1990s by the LEP experiments leave no room for a new boson with isolated muons at the rate the ALEPH 30 GeV dimuon excess needs. Every model we tested, including the two built to escape LHC limits, is excluded by at least two detectors, and 44 of the 47 graded expectations written before the calculations were met.

Nov 20, 2026 · 7 min read

In 2016 a re-analysis of archived ALEPH data reported an excess of about 32 b-tagged muon pairs with a mass of 30.4 GeV in Z boson decays at CERN's LEP collider. It was a local excess of about five standard deviations, three after accounting for where else it could have appeared. If real, it would be a new particle. We asked a narrower question that can be answered with published numbers alone: could the excess be a new boson decaying to two isolated muons? Across five published models, including two designed to escape LHC limits, the answer is no. Measurements made by the four LEP experiments in the 1990s would have recorded dozens of such events and saw at most one. Every model is excluded by at least two detectors.

38-41

boson events predicted in OPAL and L3's published 20-40 GeV window, per Ko-Li-Yu model, at the models' own rate

1

event those two measurements actually saw there

12-55 GeV

scalar masses at which the Z' plus scalar model is excluded by two detectors, each alone

44 of 47

graded expectations written before the calculations that were met

The question

ALEPH's excess sits in Z decays to b quarks. The natural readings are a new boson radiated off the b quarks (vector or scalar), or a Z decaying to two new Higgs-like states. Ko, Li and Yu showed in 2016 that a boson coupling directly to b quarks at this rate is excluded by LHC data, and built two models that avoid that bound: a Z' boson that links the b quark to a heavy vector-like quark, and a Z decaying to a Z' plus a scalar. Lane and Pritchett's two-Higgs-doublet model is produced at LEP only as a Z decaying to two 30 GeV states, the same topology as our Z to h A model.

In every one of these models the two muons are usually far from the b jets. That is what two kinds of LEP1 measurement selected: the 1994-96 measurements of Z decays to a muon pair plus quarks, and the searches for a light Higgs boson radiated off a virtual Z decaying to muons. Both predate the excess by twenty years, and as far as we can find, neither had been compared with it before.

What we did

We generated each model with exact tree-level matrix elements, hadronised the events with Pythia 8, and applied each experiment's published selection as read from the scanned papers: three four-fermion selections (ALEPH, OPAL, L3) and four Higgs-search selections (ALEPH, L3, DELPHI, OPAL). Every selection detail we could not read was bracketed by a stricter variant. We predicted the number of boson events from the excess's own rate and compared it with the counts the experiments published, using throughout the lowest rate in the literature, which is Ko, Li and Yu's own fitted rate. Before each job ran, we published what we expected it to show.

What we found

For the three simple models, OPAL's four-fermion selection keeps 84-88 % of boson events, and OPAL and L3 together would have seen about 40 events between 20 and 40 GeV even at the lowest rate. They saw one. L3's Higgs search, which has no cut on the mass recoiling against the muons and found no candidate in 3.05 million Z decays, keeps 54-64 % of the events, and DELPHI's keeps 69-82 %. For Ko, Li and Yu's two escape models, OPAL's four-fermion selection keeps 82-89 % of every variant, including a Z' with a 12 GeV scalar, whose muons recoil against so little mass that three of the four Higgs searches miss it.

Boson events predicted in OPAL and L3's 20-40 GeV window at the models' own rate, against the one observed

events

Nominal selections, lowest literature rate (branching ratio 1.1e-5, Ko, Li and Yu's model-II benchmark). Probability of seeing one event or fewer: 6e-17 to 1.2e-15. Source: analysis/results/kly4f/kly4f-summary.json

Every scalar mass, not just three

The Z' plus scalar model had been computed at three scalar masses (12, 30.4 and 50 GeV). Because the Higgs searches cut on the mass recoiling against the muons, which here equals the scalar mass, a referee would ask whether some mass in between escapes. We pre-registered seven expectations for a scan at 15, 20, 25, 35, 40, 45 and 55 GeV, with a closure re-run at 30.4 GeV, and all seven passed. The test is deliberately harsh: one detector at a time, its weakest selection variant, scaled down by the worst ratio we found between our efficiencies and the experiments' own published ones. At every mass, L3's Higgs search alone (zero candidates) and OPAL's four-fermion measurement alone (one event) still predict more events than p = 0.001 allows. The margin shrinks toward the kinematic end; at 55 GeV, p = 2.2e-4 for L3 and 3.9e-4 for OPAL.

Boson events each detector alone would have seen, most pessimistic reading, by scalar mass

L3 Higgs search (0 seen; p < 0.001 needs more than 6.9)OPAL four-fermion (1 seen; p < 0.001 needs more than 9.2)
0510152020304050scalar mass (GeV)predicted events
Weakest selection variant, L3 scaled by 0.78 and OPAL by 0.62, lowest literature rate, job klyscan-v2. Hadron level, muon identification included as a common factor. Source: analysis/results/klyscan/klyscan-summary.json

A condition we found, and how it was settled

While grading, we found that the 12 GeV result rested on a detail we had not read. OPAL required at least two jets, and our code counts the two muons when it finds jets. Had OPAL removed the muons first, its selection would keep only 0.8 % of 12 GeV scalar events, because the scalar's two b quarks merge into one jet, and that case would rest on L3 alone. We published the condition before claiming anything, then had OPAL's 1996 paper read by an agent with no access to our work. The paper says that "charged tracks and electromagnetic clusters not associated to any track are formed into jets" and expects "one or two 'jets' originating from the leptons": the selection as we coded it.

Each model against two kinds of LEP1 measurement, at the lowest rate in the literature

ModelHiggs searches, L3 + DELPHI (p)Four-fermion, OPAL + L3 (p)
Vector off b9e-114.5e-12 (strictest variant)
Scalar off b3e-103.7e-13 (strictest variant)
Z to h A (covers Lane-Pritchett)2e-122.5e-14 (strictest variant)
Ko-Li-Yu II (three variants)1.2e-10 to 1.3e-101.2e-16 to 1.7e-16
Ko-Li-Yu III, 12 GeV scalar2.1e-8 (L3 alone 1.4e-9)2.0e-16 (OPAL alone, weakest variant: 2.6e-12)
Ko-Li-Yu III, 15-55 GeV (scan)L3 alone, most pessimistic: 3.8e-6 to 2.2e-4OPAL alone, most pessimistic: 8.7e-8 to 3.9e-4
p is the Poisson probability of the observed count or fewer, with no background subtracted. Sources: NOTE-higgs-searches.md (version 0.10) and the draft paper in the repository.

What it means

If the ALEPH excess is a real effect, it does not come from a boson that decays to isolated muons at the needed rate, in any model we have read. ALEPH itself reported that each muon of its excess lies close to one of the leading jets, while our models put a muon within 15 degrees of a jet in only 4-11 % of events. Exact kinematics add a second constraint: two muons within 20 degrees of the same jet can have a mass of at most 16.6 GeV at LEP1, so no new state decaying inside one jet can make a 30 GeV pair. Because ALEPH's background is its own electron-muon data, ordinary b-decay effects cancel unless they treat muons and electrons differently. What is left is a statistical fluctuation, a muon-specific detector effect, or a state whose muons fall near two jets by accident. DELPHI's open data, released in 2024, can test that directly; our pre-registered, blinded DELPHI analysis is waiting for compute.

More research