GLOBAL KNOWLEDGE ARK

Astronomy · Why stellar light changes

Flares and eruptive behaviour

Measure an excess-light integral in a defined band, follow the effect of a shared quiet baseline, and distinguish magnetic flares from other kinds of stellar brightening.

Continue in the approved chapter order. Necessary vocabulary, quantities and assumptions are explained in the lesson.

Back to the chapter

After reading, you can

  • Distinguish radiation from a flare, plasma ejection and accretion-related brightening.
  • Calculate equivalent duration and conditional band energy from an explicitly defined synthetic curve.
  • Explain how baseline estimation, cadence and recovery tests affect an event measurement.
01

Several processes can brighten a star

A sudden rise in stellar light is an observation to explain. A magnetic flare, a change in accretion and a change in intervening obscuration can all affect a light curve through different physics. Duration and shape help, but the word eruption alone does not select one mechanism.

Magnetic reconnection changes magnetic-field connectivity and can convert stored magnetic energy into particle energy, plasma heating and radiation. An observed stellar flare samples some of those radiative consequences through the instrument’s wavelength response.

A coronal mass ejection is an outward-moving structure of plasma and magnetic field. A flash of electromagnetic radiation and a mass ejection are distinct phenomena even when they occur together. NASA’s flare explanation makes this distinction for the Sun; a bright flare detected on another star is not by itself a measured ejected mass.

Different wavelengths trace different parts of an event and its evolving emitting material. One optical band does not record the full radiated energy, much less the total energy in particles, bulk motion and magnetic-field changes.

02

An area under a curve can have units of time

Choose one fixed observing band B. Let F_B(t) be its calibrated flux and Fq,B a stipulated constant quiet reference flux in the same band. Define fractional excess x(t) = [F_B(t) − Fq,B]/Fq,B. Quiet light corresponds to x = 0, not to a star emitting nothing.

Equivalent duration is ED = ∫ x(t) dt over a declared interval. The integrand is dimensionless, so ED has units of time. It is the duration for which the quiet band luminosity would radiate the same energy as the integrated band excess under the stated conversion.

Davenport and collaborators use this fractional-excess integral in their flare studies. Their empirical curves and target-dependent luminosity estimates are not reused in the synthetic calculation here. A broad-band energy conversion also needs consistent photometric response and the spectral assumptions behind that calibration.

For this teaching model only, supply a quiet band luminosity Lq,B and isotropic emission with a fixed, consistent flux-to-band-luminosity conversion. Then E_B = Lq,B × ED. This is an excess energy in band B. No correction to total, or bolometric, radiation is specified.

03

Five vertices define an exact toy curve

Construct an original model with times 0, 60, 120, 300 and 600 seconds and fractional excesses 0, 0.5, 0.3, 0.1 and 0. Define the curve to be linear between those vertices, and integrate from 0 to 600 seconds. These are a mathematical curve’s defining values, not finite-exposure measurements of a real flare.

Each segment contributes its time width multiplied by the mean of its two endpoint excesses. The four areas are 60 × 0.25 = 15 s, 60 × 0.4 = 24 s, 180 × 0.2 = 36 s and 300 × 0.05 = 15 s. Their sum is ED = 90 s.

Supply Lq,B = 2 × 10²⁴ W under the declared band and isotropic model. The excess band energy is 90 s × 2 × 10²⁴ J s⁻¹ = 1.8 × 10²⁶ J. Decimal vertices and that luminosity are stipulated inputs, not claims about an observed star.

Multiplying the peak excess 0.5 by the full 600 s would give 300 s, the area of a rectangle that this curve does not fill. The exact trapezoidal result follows because linear interpolation is part of the supplied model. With sparse real observations, an unseen peak or tail is additional uncertainty.

All five vertices of the original piecewise-linear fractional-excess model in one fixed band B. The entire integration interval is 0–600 s.
Time (s)Fractional excess x
00
600.5
1200.3
3000.1
6000

Equivalent duration is an area

Scroll sideways to see the full diagram. Its explanation is below.

Equivalent duration is an area Original synthetic piecewise-linear flare-excess model, not measured data or a universal flare template. In one fixed passband B, the horizontal linear axis is elapsed time from 0 to 600 seconds and the vertical linear axis is fractional excess (F_B - Fq,B) / Fq,B from 0 to 0.5, with both axes starting at zero. Fq,B is the constant quiet flux. The orange line joins exactly five vertices: 0 seconds and excess 0; 60 seconds and 0.5; 120 seconds and 0.3; 300 seconds and 0.1; 600 seconds and 0. Each vertex is marked by an orange circle. The shaded area is equivalent duration ED = integral [(F_B - Fq,B) / Fq,B] dt. The four trapezoid areas are 15 seconds on 0–60 seconds, 24 seconds on 60–120 seconds, 36 seconds on 120–300 seconds and 15 seconds on 300–600 seconds, for a total ED of 90 seconds. The vertical quantity is dimensionless, so its time integral has units of seconds. Fractional excess zero means total normalized flux F_B/Fq,B = 1, not that the star is off. Under the declared constant quiet same-band luminosity Lq,B = 2 × 10^24 watts and isotropic-emission teaching model, inferred band energy is Lq,B times ED = 1.8 × 10^26 joules. It is not bolometric energy and implies no CME, particle fluence or planetary effect. In the separate shared-baseline subcase described in the article, if the quiet baseline and quiet luminosity are overestimated by 2 percent on the same fixed 0–600 second interval, the signed excess integral becomes (90 - 0.02 × 600) / 1.02 = 76.470588235 seconds, the quiet band luminosity becomes 2.04 × 10^24 watts, inferred band energy becomes 1.56 × 10^26 joules, and its ratio to the original is 13/15. Signed residuals at the interval ends are retained in that subcase; it is not a clipped-positive event-finder rerun. The separately stipulated independent uncertainties of 3 seconds in ED and 5 percent in quiet band luminosity give local relative standard uncertainty 0.060092521 only if those inputs are independent; the formula does not apply to the shared fitted baseline. Exact trapezoids describe the invented piecewise-linear curve, not an accuracy guarantee for sparsely sampled real flares. Equivalent duration is an area Synthetic model in one fixed passband B; not observed flare data. ED = ∫ [(F_B − Fq,B) / Fq,B] dt. Fq,B is the constant quiet flux. The curve is defined to be linear between the five vertices. 0 60 120 300 450 600 0.00 0.10 0.20 0.30 0.40 0.50 Fractional excess (F_B − Fq,B) / Fq,B (dimensionless) Elapsed time / seconds Zero excess means total normalized flux = 1; the star is still emitting. The shaded area has units of seconds: dimensionless excess × time. All five defining vertices All four segment areas Time / s Excess Interval / s Area / s 0 0.0 60 0.5 120 0.3 300 0.1 600 0.0 0–60 15 60–120 24 120–300 36 300–600 15 ED = 15 + 24 + 36 + 15 = 90 s. Quiet band power = 2 × 10²⁴ W → band energy = 1.8 × 10²⁶ J. Assumes fixed calibration and isotropic emission in this teaching model. This is band energy; no bolometric conversion or particle claim follows.
Original synthetic piecewise-linear flare-excess model, not measured data or a universal flare template. In one fixed passband B, the horizontal linear axis is elapsed time from 0 to 600 seconds and the vertical linear axis is fractional excess (F_B - Fq,B) / Fq,B from 0 to 0.5, with both axes starting at zero. Fq,B is the constant quiet flux. The orange line joins exactly five vertices: 0 seconds and excess 0; 60 seconds and 0.5; 120 seconds and 0.3; 300 seconds and 0.1; 600 seconds and 0. Each vertex is marked by an orange circle. The shaded area is equivalent duration ED = integral [(F_B - Fq,B) / Fq,B] dt. The four trapezoid areas are 15 seconds on 0–60 seconds, 24 seconds on 60–120 seconds, 36 seconds on 120–300 seconds and 15 seconds on 300–600 seconds, for a total ED of 90 seconds. The vertical quantity is dimensionless, so its time integral has units of seconds. Fractional excess zero means total normalized flux F_B/Fq,B = 1, not that the star is off. Under the declared constant quiet same-band luminosity Lq,B = 2 × 10^24 watts and isotropic-emission teaching model, inferred band energy is Lq,B times ED = 1.8 × 10^26 joules. It is not bolometric energy and implies no CME, particle fluence or planetary effect. In the separate shared-baseline subcase described in the article, if the quiet baseline and quiet luminosity are overestimated by 2 percent on the same fixed 0–600 second interval, the signed excess integral becomes (90 - 0.02 × 600) / 1.02 = 76.470588235 seconds, the quiet band luminosity becomes 2.04 × 10^24 watts, inferred band energy becomes 1.56 × 10^26 joules, and its ratio to the original is 13/15. Signed residuals at the interval ends are retained in that subcase; it is not a clipped-positive event-finder rerun. The separately stipulated independent uncertainties of 3 seconds in ED and 5 percent in quiet band luminosity give local relative standard uncertainty 0.060092521 only if those inputs are independent; the formula does not apply to the shared fitted baseline. Exact trapezoids describe the invented piecewise-linear curve, not an accuracy guarantee for sparsely sampled real flares.
04

Keep a baseline change consistent

Now test a separate systematic change while holding the actual calibrated total flux F_B(t) fixed. Overestimate the original quiet flux by 2%, using Fq,B′ = 1.02Fq,B. Keep the same 0–600 s integration interval, allow signed residuals and do not re-run an event finder or clip negative endpoints.

The new fractional excess is x′ = (x − 0.02)/1.02. Hence ED′ = [90 s − 0.02 × 600 s]/1.02 ≈ 76.470588235 s. The endpoints become negative relative to the overestimated baseline; that signed residual is not negative incident radiation.

Under the same fixed conversion, the inferred quiet band luminosity changes too: Lq,B′ = 1.02 × 2 × 10²⁴ W = 2.04 × 10²⁴ W. Their product is E_B′ = 1.56 × 10²⁶ J, smaller by 2.4 × 10²⁵ J, or about 13.33%, than the original excess energy.

Treating ED and quiet luminosity as independent uncertain quantities would miss this shared-baseline dependence. Clipping the negative endpoints, changing the event interval or holding the luminosity fixed would define a different calculation. State which measurement procedure is actually being tested.

05

A detected flare can still have a biased integral

Removing a slowly varying baseline can also remove part of a long event or its faint decay. Finite exposure duration can smooth the peak. Missing observations can leave the integral incomplete. Those effects need to be assessed for the actual light curve and processing method.

Injection and recovery tests add defined synthetic events to data, process them through the same search, and measure what is recovered. The conclusion is conditional on the injected shapes, amplitudes, durations, placements and background behaviour, as well as the detection rule.

Davenport’s 2016 recovery study explicitly separates finding an event from measuring its duration or energy accurately. Its reported recovery fractions are not imported into this example. A search can recover a peak while underestimating a long, faint tail.

For a different local uncertainty exercise, suppose independently supplied estimates have u(ED) = 3 s at ED = 90 s and u(Lq,B)/Lq,B = 0.05. First-order propagation for their product gives relative standard uncertainty √[(3/90)² + 0.05²] ≈ 0.06009. This conditional calculation excludes the shared-baseline systematic just tested and is not automatically a confidence interval.

06

An eruption can draw on accretion or obscuration changes

In young stellar systems, episodic accretion can increase the release of gravitational energy as material moves inward. Audard and collaborators review accretion outbursts and the diversity of their behaviour. Such an event should not automatically be interpreted as a magnetic flare or a star undergoing terminal destruction.

Changing circumstellar dust can also change the light along our line of sight. Accretion and obscuration changes may coexist. Colours and spectra, together with the sequence across wavelengths, help test which contributions are needed.

A credible interpretation connects the event’s temporal and spectral evidence with a mechanism and its alternatives. The band-energy integral alone does not establish a coronal mass ejection, a particle fluence or a planetary environmental consequence.

Retain the quiet reference, integration limits, observing band, calibration, sampling and selection history alongside an energy estimate. They explain what was measured and which future observation could sharpen the physical inference.

Worked examples

Add areas rather than peak rectangles

What equivalent duration follows from the complete five-vertex model?

  1. The four segment areas are 15, 24, 36 and 15 s.
  2. Their sum is 90 s because the curve is defined to be linear between vertices.
  3. With the stipulated quiet band power 2 × 10²⁴ W, multiply by 90 s.

ED = 90 s and E_B = 1.8 × 10²⁶ J within the declared band and conversion assumptions.

Change both quantities that share the baseline

What energy follows when the quiet baseline is increased by 2% with total flux and limits fixed?

  1. ED′ = (90 − 0.02 × 600)/1.02 ≈ 76.470588235 s.
  2. The same baseline change gives Lq,B′ = 2.04 × 10²⁴ W.
  3. Multiply the consistent pair to obtain 1.56 × 10²⁶ J.

The inferred excess band energy becomes 1.56 × 10²⁶ J. This signed fixed-interval test is different from redefining the event.

Try the reasoning

  1. What total normalized flux F_B/Fq,B corresponds to zero fractional excess?

    Show the answer and reasoning

    1.

    Zero excess denotes the quiet reference level, while the star still emits light.

  2. Why is equivalent duration expressed in seconds rather than joules?

    Show the answer and reasoning

    It integrates a dimensionless fractional excess over time.

    A consistent band luminosity supplies the power needed to convert that time into energy.

  3. Does recovering the peak time of every injected event establish unbiased flare energies?

    Show the answer and reasoning

    No.

    Duration, tails, baseline and calibration can still bias the measured integral.

  4. Does the calculated 1.8 × 10²⁶ J include all radiated and particle energy?

    Show the answer and reasoning

    No.

    It is excess radiation in the defined band under a stipulated conversion; no bolometric or particle model is given.

What to carry forward

  • Equivalent duration measures a defined fractional-excess area relative to a quiet reference.
  • A shared baseline can change both the integral and its energy conversion.
  • Event detection, band-energy measurement and physical diagnosis require different supporting evidence.

References

Teaching resources and research supporting this lesson’s explanations and examples.

  1. Solar Flares FAQsMiles Hatfield / NASA Goddard Space Flight CenterReference guidescience.nasa.govRead source
  2. Kepler Flares II: The Temporal Morphology of White-Light Flares on GJ1243James R. A. Davenport et al., University of Washington and collaboratorsReference guidearxiv.orgRead source
  3. The Kepler Catalog of Stellar FlaresJames R. A. Davenport, Western Washington UniversityReference guidearxiv.orgRead source
  4. Episodic Accretion in Young StarsMarc Audard et al., Geneva, Konkoly, Yale and international collaboratorsReference guidearxiv.orgRead source
  5. NIST TN1297 AppendixA: Law of Propagation of UncertaintyBarry N. Taylor and Chris E. Kuyatt / NISTReference guidenist.govRead source