Star clusters, associations and membership

Apparent proximity does not establish a shared system

Stars close together on an image can lie at different distances and have unrelated motions. A concentration on the sky is therefore a candidate grouping, not by itself proof of common origin or mutual gravitational binding.

Terminology varies, especially for young populations with structure on several scales. In this lesson, a bound cluster means a stellar group held together by its collective gravity, while an association means a dispersed unbound group with evidence of related origins or motion. Some literature uses cluster more broadly, so the definition behind a catalogue should be read.

Open and globular clusters describe familiar kinds of stellar collections with different typical appearances and histories. They are not defined here by one universal number of stars, age or density threshold. A label does not replace the measurements and dynamical model needed to characterize a particular system.

Group boundaries can be ambiguous. A compact concentration, an extended population and stars leaving a system need not fit one simple circular boundary on the sky. A useful membership analysis states which physical or observational group it is trying to identify.

Position and motion supply complementary information

Sky position identifies a projected location. Parallax constrains distance through a measurement model, while proper motion describes angular movement across the sky. Radial velocity adds the line-of-sight component of motion. Their agreement can support a membership interpretation.

The observed spread contains measurement uncertainty as well as physical structure. Correlations between measured coordinates can matter, so multiplying separate likelihoods as though every measurement were independent is an additional assumption. A joint error model can be necessary.

An unresolved companion can shift a star’s measured radial velocity away from the centre-of-mass motion of its system. Differential extinction and incomplete detection also affect which stars are available for comparison. The preceding companion lessons and earlier uncertainty lessons explain why a single hard cut can be misleading.

A membership method generally compares a group model with alternatives such as a field population of unrelated stars. The inferred result depends on the adopted distributions, errors, selection and population proportions. Similar-looking motion is evidence with a context, rather than certainty.

Reverse the conditioning with the prior retained

For an original finite probability example, partition candidates into group members C and field objects F. Let S be one stipulated joint consistency event. Supply P(S|C) = 0.9 and P(S|F) = 0.3; their likelihood ratio is 3. These are teaching assumptions, not performance measurements from a stellar catalogue.

If the member fraction before applying S is p, total probability gives P(S) = 0.9p + 0.3(1 − p). Bayes’ theorem then gives P(C|S) = 0.9p/[0.9p + 0.3(1 − p)]. Both conditional pass probabilities and the prior belong to the same supplied model.

For p = 0.8, P(S) = 0.78 and P(C|S) = 0.72/0.78 = 12/13, about 0.923077. An equivalent expected-count illustration out of 100 candidates gives 72 passing members and 6 passing field objects. These are expectations under the model, not an observed sample.

For p = 0.1, the same consistency event gives P(S) = 0.36 and P(C|S) = 0.09/0.36 = 1/4. The expected illustration now gives 9 passing members and 27 passing field objects out of 100 candidates. The evidence rule has not changed; the candidate population has.

Different scores can be constructed differently

The two results do not mean that a particular star can be assigned any desired probability by choosing a convenient prior. An actual prior or population mixture must be justified for the candidate selection and model. Changing the selection can legitimately change the relevant population proportions.

Some methods report how frequently an object passes a grouping and spatial-consistency procedure when measurements are redrawn within their errors. Other methods calculate probabilities from an explicit group-and-field mixture. A redraw frequency and a posterior probability have different constructions and should not be silently treated as identical quantities.

The synthetic calculation above uses one already specified joint event S. It neither fits a real distribution nor calibrates a published algorithm’s membership score. Its lesson is narrower: P(S|C) and P(C|S) answer different questions, and the reversal requires the candidate-population prior.

A reported high score also does not establish the binding energy of the whole group, an exact formation time or a unique chemical history. Those conclusions need additional physical evidence and an analysis designed to answer them.

Use membership to support a population comparison

Common origin can motivate comparing a group with models having a related age and composition. It does not prove exactly simultaneous formation or identical abundance for every element. Some well-studied cluster populations contain important chemical differences.

Membership uncertainty can propagate into a stellar diagram: an unrelated object can resemble a bright evolved member, while a missed member can leave an apparent gap. Selection and contamination therefore affect the apparent population as well as individual membership labels.

Useful reporting identifies the candidate selection, observed quantities, uncertainty model, group and field descriptions, score definition and remaining ambiguities. Additional spectroscopy or repeated velocity measurements can test particular interpretations without making every membership question disappear.

The next lesson uses this context to examine age estimates from population diagrams. The aim is to compare a selected, characterized population with explicit evolutionary models, keeping uncertain membership and unresolved systems in the reasoning.

Worked examples

Try the reasoning

1. Does apparent crowding in an image prove that its stars form a bound cluster?

   Show the answer and reasoning

   No.

   The stars can have different distances, motions and origins; boundedness needs additional dynamical evidence.

2. Is P(S|C) generally equal to P(C|S)?

   Show the answer and reasoning

   No.

   They condition on different events. Bayes’ theorem requires the prior population probabilities to reverse them.

3. Does the example justify multiplying separate parallax and proper-motion likelihoods as independent?

   Show the answer and reasoning

   No.

   It stipulates one joint event S and does not provide independence between coordinates.

4. Does a high membership score prove identical composition and an exactly shared age for every member?

   Show the answer and reasoning

   No.

   The score addresses a specified membership model; chemistry, formation history and gravitational binding require their own evidence.

What to carry forward

• A physical group is inferred from evidence and an explicit definition, not image proximity alone.
• Membership estimates depend on measurements, uncertainties, field populations and the construction of the score.
• Common-origin assumptions can support a population comparison while retaining chemical, age and membership uncertainty.