Suppose an experimentalist has a sensitive device, the conventional physics of which is well under control. Now let’s assume that once in while the device is perturbed by some unanticipated “new physics” events, such as an interaction with a lump of dark matter. Suppose the device has enough sensitivity to “new physics”. When would an unanticipated “new physics” event become apparent to an unsuspecting experimentalist?
This is quite different from particle colliders, when experimentalists do hunt for unusual events. Sometimes specific signal or signature is anticipated (e.g., the Higgs), so let me emphasize that our unsuspecting fellow experimentalist does not specifically look for new physics.
The discovery of the cosmic microwave background could serve as a motivating example of paying attention to “misbehaving” data.
So when would “new physics” be noticed? An obvious answer would be: when the new physics perturbs the expected signal in a significant way.
Even this simple statement requires qualifiers. Suppose new physics provides a uniform background to the signal and the signal itself cannot be computed exactly from the first principles. For example, transition frequencies of many-electron atoms can be computed only to 3-4 significant figures while experimentalists can determine some of these frequencies to 18 significant figures. Then (unless there are symmetry arguments, e.g., parity or time-reversal violation and associated external field reversals implemented in an experiment) there is no way to dissect the new physics background from the conventional one.
This leaves us with time/space-dependent new physics signals. I.e., new physics could be noticed if it leads to some noise or drift in time/space-dependent signals. For example, a uniform-in-time drift in atomic frequencies could reveal variation in fundamental constants.
What about “new-physics” noise/spike-like events, such as the perturbation by "lumps" of dark-matter? Suppose the conventional signal is interrupted by new physics events. We could characterize such events by how long an event lasts (short/long interaction times), how frequent the events are (rare/frequent) and if the device is sensitive to the event. For simplicity we assume that the events do not overlap, i.e., the average time between the individual events is much longer than the event duration.
The event would be missed if the average time between consecutive events is larger than a typical time of continuous operation of the device (i.e., the events are rare). Unfortunately, a single bona fide event could be discarded by an experimentalist as being an outlier. Indeed, one can never guarantee that everything is fully under control, as there still may be occasional perturbations present, such as a student bumping into an optical table or misbehaving power supplies.
Essentially, rare events would be registered as such only if they are anticipated.
This argument brings us to the following conclusion: unanticipated “new physics” events would be noticed only if the sizable events are frequent on time-scale of the experiment.
There is another caveat: suppose the events are so frequent that they look like a white or a flicker noise in the signal. After all it is natural to assume Poissonian distribution of time intervals between consecutive events. Then there is a danger of “integrating out” the events.
Thereby we have to revise our statement: unanticipated “new physics” events would be noticed only if the sizable events are frequent (but not too frequent) on the time-scale of the experiment.
In practical terms, for a typical atomic physics experiment, the events should last longer than a second and there should be hundreds of them per day of operation. Even then, the experimentalist should be gutsy enough to put his/her credibility and comfort at risk and publicly report the data as being unusual. “New physics” looks for the right fellow to notice and appreciate it.
P.S. I would like to thank Dima Budker and Jeff Sherman for discussions on this topic