TL;DR After NASA's DART mission intersected Dimorphos in 2022, telescopes measured when the two asteroids passed in front of one another or cast shadows on each other. The shifted sequence showed that Dimorphos orbited Didymos more quickly. Combining timing with radar and pre-event images enabled tighter models of the changed orbit and shape. The published orbital periods refer to the binary pair, distinct from a later result about the pair's path around the Sun.
At the telescope, both asteroids can fit into one measurement
An observer records the brightness of Didymos and its smaller companion Dimorphos. The image obtained with one suitable ground telescope can be used as a combined light signal rather than as a pair of independently resolved pictures. When the geometry produces a mutual event, one asteroid blocks part of the other's light as seen by the telescope, or a shadow reduces the reflected light. That moment can make the signal dip. The shape of the dip and the time it occurs contain information about the configuration of the pair.
NASA's Jet Propulsion Laboratory describes optical observations of such mutual events in the analysis of the September 26, 2022 DART event. It also describes other evidence, including radar observations after the impact and images gathered during the spacecraft's approach. Its account says Dimorphos previously went around Didymos in about eleven hours and fifty-five minutes, and later work found a sustained decrease of roughly thirty-three minutes. The article below explains how event times support a period result and where additional modelling enters. NASA and JPL account of the orbit and shape study NASA science overview of Didymos and Dimorphos
The word mutual helps keep two viewing processes distinct. In an occultation, a body gets in the line of sight between an observer and some of the light. In an eclipse, a body or a shadow changes the illumination reaching or leaving another object. Under useful geometries both can affect the brightness that arrives at the ground-bound observer. The orbit determines when particular configurations repeat, while viewing and illumination geometry help determine what a telescope will see. The direct observation is a pattern in received light, not an image of an orbital diagram traced in space.
A light curve is a set of brightness measurements through time. A telescope's record can contain repeated changes due to asteroid rotation and other effects as well as the desired dips. Researchers use data reduction and models to distinguish those contributions. A reader following the scientific claim should know which part of the curve is attributed to a mutual event. Otherwise any changing brightness could be misread as a completed orbit. The DART investigators discuss the need to account for the primary asteroid's rotational brightness contribution in studying the mutual orbit. DART investigation summary in The Planetary Science Journal
First annotation. Fold a fictional sequence by the old interval
The simplest timing exercise uses comparable events, all of one type, measured under a toy geometry. Set the first event at time zero. Imagine the old period is eleven hours and fifty-five minutes. A sequence of equal repeats would then show the same event at 11 hours 55 minutes, 23 hours 50 minutes, and 35 hours 45 minutes. These numbers are rounded for the exercise. No real record is being reanalyzed. The equal intervals show what a steady repeat would look like if all four measurements were available.
Now suppose the same kind of event repeats every eleven hours and twenty-two minutes. Under the same illustrative start the calendar moves to 11 hours 22 minutes, 22 hours 44 minutes, and 34 hours 6 minutes. The first later event happens thirty-three minutes earlier than the old illustrative expectation. After two repeats it is sixty-six minutes earlier. After three repeats it is ninety-nine minutes earlier, or an hour and thirty-nine minutes. Each interval adds another copy of the difference.
This is the principle behind looking at phase. With a trial period you can align multiple observations by where they fall within the assumed cycle. The right period tends to keep comparable events near the expected phase in a simple model. A wrong period produces a cumulative displacement as more cycles pass. An astronomer can track the time difference between observed and modeled events more explicitly. If an old-period prediction falls progressively behind actual observations just after a change, a shorter interval is one explanation to evaluate.
The exercise assumes the first comparable event after the change occurs at the model's chosen zero. Real observers would account for event geometry and reference timing rather than force that equality. They can estimate an epoch, a period and other model parameters from a collection of observations where appropriate. A one-time shift that makes every event arrive a little early differs from a growing shift from later cycles. Observing several repeats gives a way to tell the effects apart if the event identities and clock references are reliable.
A single lower brightness reading does not provide either result by itself. It shows that some flux was measured at a time under particular observatory conditions. A well-sampled dip across a useful part of the event can give an estimate of when its center or other relevant feature occurred. If its beginning was missed or the seeing changed sharply, the timing uncertainty can be larger than the precision of the clock. The published timing should express what the measurement and fitting support, not how many digits the instrument records.
Second annotation. The calendar leaves a hole in the middle
Return to the fictional eleven-hour-and-twenty-two-minute sequence. Clouds obscure the event predicted at 22 hours 44 minutes. The visible times at 11 hours 22 minutes and 34 hours 6 minutes are twenty-two hours and forty-four minutes apart. Dividing by two recovers eleven hours and twenty-two minutes if there is evidence two cycles elapsed. Calling the recorded gap one orbit would produce the wrong period. Establishing the count requires the other observations, model and observation conditions.
It is also possible to confuse different kinds of event. Suppose one dip comes from the smaller body passing in front as viewed from Earth. Another comes from the other configuration in its orbit. Their separation need not be the full orbital period. The spacing between different event classes can itself depend on geometry, including a noncircular orbit and the directions to the Sun and observer. Counting every brightness minimum as the same cycle boundary can produce a tidy period for the wrong repeating structure.
A distributed observing campaign helps because a telescope cannot observe the same object from one site at every hour and under every weather condition. Measurements from more than one site can recover events another station missed. They also introduce the need to compare time systems, measurement precision, observation geometry and reduction procedures. Evidence becomes stronger when the expected event type and timing agree across appropriately interpreted measurements. The material should retain enough details to understand why the agreement was meaningful.
Weather and visibility produce an important negative case. Imagine a telescope gives no reported dip during part of the predicted interval, but the target was below the horizon during most of the relevant period. That result provides no observation of a missing dip at the chosen phase. A clear observation covering the expected event and still showing no relevant signal would give different evidence. Well-documented gaps protect a later analysis from assigning the same meaning to both silences.
A shape measurement can also suffer from partial recording. If a dataset covers only egress, the moment when the object emerges from the configuration, then a symmetric fit to the visibly sampled part may assign the wrong center. A more complete sequence can constrain how long the event lasted and whether its shape was symmetric. The observers can report how the estimate depends on available coverage. A study of period precision benefits from knowing how timestamps were derived, since not every dip's measured center is equally secure.
Third annotation. Subtracting a background does not erase its uncertainty
Before the event, the two bodies reflect sunlight as they turn. The changing visible surfaces produce their own signal alongside the pattern the team wants. An analyst can estimate the primary body's rotational contribution by fitting observations away from mutual events or using a suitable established model. The resulting residual emphasizes the relevant dips. Errors in the estimated background can still affect the timing or depth of a weak event. Different groups can make different warranted modeling choices while agreeing on the strongest period signal.
Naidu and colleagues describe using times of mutual events in primary-subtracted light curves. They distinguish their approach from another analysis that modeled more of the available light curve. The differences matter to interpreting reported values and uncertainties. Two measurements can support a similar final period while differing in the data pieces that constrain shape, eccentricity or evolution. A good reading of the papers identifies which outputs are direct features of the observations and which outputs are inferred under each model. Naidu and colleagues, post-impact orbital and physical characterization
A toy trace illustrates why separating the contributions is valuable. At six successive moments across an apparent event the aggregate signal is trending brighter because the background is changing. A short drop relative to that trend appears on top of it. The lowest absolute brightness in the run might occur before the event despite the real dip relative to the expected background. Fitting the event relative to a justified baseline can find a center that a simple minimum read would miss. The example supplies a reasoning method, not a recipe for processing the DART data.
The analysis can include other time-dependent structure. Material around the asteroids and changing illumination after the event can affect observations. The distinction between a changing baseline and orbital event depths is one reason for comparing data over a sequence of dates. A feature that repeats with orbital phase can have a stronger interpretation than a single broad trend. Radar and imaging can constrain parts of the geometry that brightness alone does not establish. Evidence with different weaknesses is particularly valuable when a physical parameter could affect several observable features.
There is also a publication lesson about digits. The 2024 analysis reports a settled orbital period of 11.3674 hours with a reported uncertainty of 0.0004 hours. Multiplying the fractional part, 0.3674 hours, by sixty gives 22.044 minutes, corresponding to roughly 22 minutes and 2.64 seconds. The uncertainty in hours corresponds to about 1.44 seconds. JPL's public explanation expresses the estimated period as 11 hours, 22 minutes and 3 seconds. The different renderings are compatible at the precision shown. Naidu et al., 2024 period result JPL public explanation
Fourth annotation. The close-up and the radar constrain different things
JPL identifies close-range spacecraft observations during approach as pre-event evidence for the pair's spacing and dimensions. It describes later radar as sensitive to relative positions and motions. The optical event timing supplies another type of constraint over repeated observations. The model sees all those sources together. It can check whether candidate orbital and shape descriptions fit several kinds of observation rather than matching one series in isolation. JPL account of the three data sources
The distinction between measured images and inferred later shape is worth keeping. A proposed shape can change where and when one body's shadow or outline affects the brightness. Orbital geometry can also shift the times and durations. By comparing combinations with the available timing and other observations, researchers can estimate which shapes and orbits better fit the data. Their result for the later shape comes through that model. It should not be described as if a camera on DART photographed the final settled object from the same vantage it had before the event. Naidu and colleagues on post-impact shape inference
Consider a thought experiment with two different ellipse models for an unresolved body's outline. Both produce about the same depth for a particular pass. One predicts that a second event with another viewing configuration lasts longer than the observation. The other better matches timings across several configurations. That comparison adds information about shape without supplying a visual image for every observation. There can still be ambiguity about which shape or orientation change caused a given feature, especially if data do not cover enough distinct geometries.
A period estimate alone also can't explain every change in orbit. JPL's analysis describes a slightly elongated later relative orbit and a temporal change in the period itself. Its reported estimates of the evolution come from modeling across multiple observations. One number from an early sequence and another from later observations can both be reasonable for different times. The article's toy sequence intentionally uses a single period per phase so the arithmetic remains transparent. Evaluating gradual changes requires additional measurements and an explicit time-dependent model.
A small uncertainty quoted for one parameter belongs to the study and its assumptions. This matters if a summary combines a period estimate, a shape conclusion, and a model for how material evolved. Each has its own support and possible alternatives. The simple shift in repeated times can give strong evidence of a shorter mutual period. Fine details about the outline, eccentricity or later evolution need more context. Preserve which uncertainty and model each result came from when comparing statements.
The two clocks have different centers
A reader in 2026 may also encounter a NASA report that the pair's orbit around the Sun changed by a fraction of a second. That March 2026 article describes a separate analysis of the system's heliocentric motion, using measurements that constrain its course around the Sun. NASA describes that orbital period as about 770 days. The roughly thirty-three-minute effect concerns the smaller body's relative orbit about Didymos. It cannot be transferred to the much longer solar orbit by treating both as references to one asteroid clock. NASA's March 2026 account of the heliocentric result
The two results can coexist naturally because one gravitationally linked pair participates in both motions. The changes are measured and modeled using observables appropriate to each orbit. It would be unsound to compare the size of the two numerical changes without their baselines and definitions. One relation is a roughly half-day mutual period. The other is a multi-year orbit of the system. Putting the center of the orbit in the heading is a useful way to avoid confusion even before reading the reported uncertainty.
The fictional student reviewing both findings now asks two separate questions. Which events are repeating at the observatory? Which orbit appears in the stated period? These questions locate the evidence quickly. For the mutual changes, the event timing in the combined optical signal and other companion data establish the relevant rhythm. For the heliocentric result, the researchers are asking where the system is during its longer trajectory and how that trajectory changed. The same 2022 mission can prompt both investigations without making their data interchangeable.
The answer also governs what could follow scientifically. A closer view of the binary can constrain its physical properties and help test shape estimates. More suitable long-term tracking of the system's path can improve a heliocentric solution. One kind of information can refine shared physical understanding without directly measuring the other orbit's period. A good account tells a reader which new observation would bear on the particular remaining uncertainty.
Compare an early result and a settled estimate at their own times
The mission's early public result described a shortening of about thirty-three minutes. The more detailed study modeled an immediate change and a further change over the following weeks. Naidu and colleagues reported that the period adjusted by an additional 34 seconds with a stated uncertainty of 15 seconds under their model before reaching their later value. JPL's account gives the early estimate as 11 hours, 22 minutes and 37 seconds and the settled value as 11 hours, 22 minutes and 3 seconds. Their difference is thirty-four seconds at the displayed precision. Naidu et al., modeled time evolution JPL explanation of orbital evolution
That structure gives a reader a way to compare an announcement and a paper without reading an extra few digits as the entire scientific change. The early number answers whether the period had moved markedly and by roughly how much. Later work asks whether one post-event period explains all available times or whether a changing solution fits better. The latter requires more observations and more explicit assumptions. It can turn a single approximate finding into evidence about the subsequent dynamics.
The student combines the illustrative series with a later night held out for a check. A model fit only to the first few events can match them within their timing uncertainties while predicting a later event poorly. Missed cycle counts, a changed baseline, or time dependence can all contribute. The stronger evaluation asks which hypotheses reproduce the broader pattern once observations and recorded uncertainties are included. The learning goal is to check the consequences of a model on a separate time span, without assuming that a more complicated fit always describes a real physical feature.
The team also compares a case where observers have several closely spaced times and one much later usable event. An uncertain early epoch affects all later residuals similarly. A wrong assumed period leads to a growing divergence across the long gap. Missed events make the gap more useful and more dangerous at once because the count itself needs support. The measured time and an interpretation of its cycle are two facts with different origins. Keeping both visible gives another researcher a chance to see why the inferred period is justified.
One summary table might give the period, the applicable stage, and the explanation of uncertainty for each reported value. It would also say which observations it is meant to explain. This preserves both the important mission outcome and the refinements that continued study made possible. It stops the phrase after DART from collapsing every observation over many dates into one supposedly interchangeable value.
Keep the annotations that make the claim reproducible
A helpful summary of the mutual event result preserves the pre-event period as reported, the date and context of the observations, and the later inferred period under its stated method. It distinguishes photographed or measured features from fitted outcomes. The general audience can understand the result through a small timing example. The source paper and JPL brief give the formal model and the reported values more directly. Neither needs to be copied word for word to make the reasoning inspectable.
One final check compares the uncertainties for two studies using their own published conventions. A precision expressed with a particular confidence convention cannot be ranked directly against a number using a different one. Similarly, a model that uses more portions of the same underlying observations shares some of the evidence with an analysis using event times alone. The agreement remains useful in showing that the period survives a change in method. Its interpretation should account for data overlap before treating the estimates as if they came from entirely separate experiments. The studies' observation tables and methods provide the information needed to make that distinction.
When the class returns to its imagined series, the main deduction is clear. Comparable events after three repeated cycles occur ninety-nine minutes early relative to the simplified old prediction. One bad point or lost night can complicate the count. Several identified events and independent geometric observations let researchers tell persistent orbital change from gaps or changing views. A tiny darkening becomes useful evidence because it has a defined kind, a reliable time, and a place within an independently tested model of the binary.