Cosmology · Distances, clocks and cosmic measurements

Standard candles need calibration and standardization

Correct a supplied luminosity trend in a noiseless source model, then show why an unknown flux loss still biases the inferred luminosity distance.

Preparation: ratios, powers of ten and ordinary arithmetic. Physical quantities and notation are introduced as they are needed.

Back to the chapter

After reading, you can

Distinguish identical intrinsic luminosity from a conditional standardization model.
Calculate distance biases from a supplied luminosity trend and a separate attenuation case.
Explain why successful correction in a toy does not remove real calibration, scatter and selection effects.

Source class alone does not fix emitted power

A standard candle supplies distance information through a known or calibrated intrinsic luminosity. Real source populations generally need more than a class label. A standardizable candle uses measured source properties and an empirical relationship to account for predictable luminosity differences.

Type Ia supernova analyses, for example, use light-curve shape, colour and related population or environment assumptions rather than declaring every raw event equally luminous. Those relationships and their uncertainties must be calibrated and tested.

Given a calibrated bolometric luminosity L_src and received bolometric flux F, luminosity distance is D_L = √[L_src/(4πF)]. An error in intrinsic luminosity or propagation treatment changes the inferred distance.

The model below supplies an exact luminosity relationship and a perfectly known shape index. It is an arithmetic demonstration of conditional standardization, not a real supernova fit, a trained light-curve model or an empirical coefficient estimate.

Make the toy’s assumptions explicit

Choose a dimensionless shape index x between −1 and +1. Set intrinsic bolometric luminosity to L_src = 100(1 + 0.2x) power units. The values at x = −1, 0 and +1 are 80, 100 and 120 power units.

Place all three sources at the same true luminosity distance of 10 length units in the supplied flux account. With no attenuation or measurement noise, F = L_src/[4π(10 length units)²]. Flux units are the chosen power unit divided by the square of the chosen length unit.

The resulting fluxes are approximately 0.063662, 0.079577 and 0.095493 in those flux units. The brighter intrinsic source has higher received flux even though the true luminosity distance is identical.

The index x is a chosen teaching variable. It is not a particular empirical supernova shape parameter, and the coefficient 0.2 is not a measured standardization coefficient. The exact trend is supplied rather than inferred from these three points.

An incorrect identical-power assumption creates distance errors

First ignore the trend and assume every source has luminosity 100 power units. Substitution into the distance definition gives D_naive = 10/√(1 + 0.2x) length units.

At x = −1, the intrinsically fainter source is interpreted as farther away: D_naive ≈ 11.18 units. At x = 0, the inferred distance is 10. At x = +1, the intrinsically brighter source is interpreted as nearer: D_naive ≈ 9.13.

Now use the exact supplied luminosity relation and the known x for each source. D_corrected = √[100(1 + 0.2x)/(4πF)] gives 10 units in every case. The correction succeeds because the model and index are exact and the toy includes no other effects.

This result does not demonstrate zero scatter in real standard candles. It isolates one predictable luminosity trend so that the reason for standardization can be followed without confusing it with calibration or measurement error.

Noiseless hypothetical luminosity trend at true luminosity distance 10 units; all powers and fluxes use the same declared arbitrary units
Shape index x	Intrinsic luminosity	Received flux	Distance assuming luminosity 100	Distance using the supplied trend
−1	80	0.063662	11.1803	10
0	100	0.079577	10	10
+1	120	0.095493	9.1287	10

An intrinsic correction does not repair unknown attenuation

Take a separate case with x = 0, so the known intrinsic luminosity is 100 power units and the true luminosity distance is still 10 units. Now let only 0.8 of the otherwise received bolometric flux remain, with that loss omitted from the distance model.

Using the correct intrinsic luminosity but the reduced flux gives D_inferred = 10/√0.8 ≈ 11.18 units. The source appears farther away because the inference has assigned the missing flux to distance rather than to the supplied attenuation.

The value 0.8 is a chosen transmission fraction, not a fitted dust law or an observational estimate. The example shows that a perfect correction of one intrinsic trend does not automatically correct a different propagation effect.

A real discrepancy can have several possible causes. This toy does not establish an actual opacity signal, a failure of distance duality or exotic physics. It makes one conditional inference error explicit.

A synthetic noiseless intrinsic luminosity L_src = 100(1 + 0.2x) in arbitrary power units uses dimensionless supplied shape index −1 ≤ x ≤ 1 and places every source at true luminosity distance 10 length units. Bolometric flux is L_src/(4πD_L²), in the same power unit per squared length unit. The upper graph plots inferred distance against x. A dashed orange circle-marked curve assumes luminosity 100 for every source and gives 10/sqrt(1 + 0.2x), approximately 11.1803, 10 and 9.1287 at x = −1, 0 and +1. A solid blue square-marked line uses the exact supplied intrinsic trend and recovers 10 throughout. The index and coefficient are teaching choices, not an empirical supernova standardizer, a measured light curve or a fitted coefficient. A separate lower comparison fixes x = 0 and lets only transmission 0.8 of the otherwise received flux remain. Bars on one shared length scale compare true distance 10 with inferred distance 10/sqrt(0.8) ≈ 11.1803 when the luminosity is correct but that flux loss is ignored. This demonstrates one conditional inference error, not observed opacity or a failure of distance duality. Real calibration, source variation, propagation, selection, lensing, bandpass and measurement effects are not removed by the exact noiseless toy correction. Original Ark teaching model.

Calibration and residual variation remain part of the method

An overall luminosity normalization must also be calibrated to obtain an absolute distance scale. Relative light-curve information alone does not independently determine both the absolute luminosity and every distance.

Real analyses account for residual source variation, calibration correlations, colour and dust treatment, peculiar motions, lensing and population or selection effects as appropriate. Measured standardization parameters and source properties also have uncertainties and can be correlated.

A method calibrated on one sample must be checked when applied to a different population or observational setting. Otherwise a change in source properties or selection can be mistaken for a geometric distance effect.

The useful question is therefore not simply whether the objects are called standard candles. It is whether the luminosity relation, calibration, propagation treatment and uncertainty account are supported for the data and inference being made.

Worked reasoning

Keep intrinsic luminosity and transmission separate

For the x = 0 source, what luminosity distance is inferred if the true distance is 10 units, only 0.8 of the flux remains and the loss is ignored?

The intrinsic luminosity model is correct at 100 power units. The received flux is nevertheless reduced by a factor 0.8.
Distance inferred from the same luminosity is multiplied by 1/√0.8.
Thus D_inferred = 10/√0.8 ≈ 11.18 units. Correcting the intrinsic trend has not corrected the omitted transmission loss.

Approximately 11.18 units, an overestimate in this supplied attenuation example.

Try the reasoning

In the unattenuated toy, does assuming luminosity 100 for the x = −1 source overestimate or underestimate distance?
Show the answer and reasoning
It overestimates distance.
The source’s actual luminosity is 80, so its lower flux is incorrectly attributed to a greater distance.
Does exact recovery of 10 units for all three toy sources establish zero residual scatter for real supernova distances?
Show the answer and reasoning
No.
The exact noiseless relationship deliberately omits the additional variation, calibration and measurement effects present in real data.
Does relative shape information by itself independently calibrate an unknown absolute luminosity normalization?
Show the answer and reasoning
No.
An absolute luminosity or distance calibration is additional information beyond the relative standardization trend.

What to carry forward

Standardization is a conditional luminosity model, not a claim that every raw source is identical.
An omitted propagation effect can bias distance even after an intrinsic trend is corrected.
Absolute calibration, residual variation and dependence between inputs remain part of a real candle inference.

References

Sources supporting the explanations, physical distinctions and examples in this lesson.