Spectroscopy – The Analysis of Spectral Line Shapes The detailed analysis of the shapes of spectral lines can give you information on: 1. Differential rotation in stars 2. The convection pattern on the surface of the star 3. The location of spots on the surface of stars 4. Stellar oscillations 5. etc, etc. The Rotation Profile The equivalent width of the line is conserved under rotational broadening !!!! To match the spectrum of a star that is rotating rapidly, take a spectrum of a slowly rotating star with the same spectral type and convolve with the rotation function Rotation in Stars Note: we only can measure v sin i, the true rotational velocity times the sine of the inclination axis of the rotating star Rotation Rates in Stars Wide range at earlier spectral types are due to two reasons: 1) some stars rotate slower; 2) We view some stars at lower inclinations (from the pole, sin i is small) As one goes to higher luminosity classes the rotation break moves to later spectral types Rotation in Stars Fast rotators The breakup velocity is the speed at which the centrifugal force exceeds the gravitational force Rotation in Stars Gravity darkening The centrifigal force from rapid rotation provides hydrostatic support to the atmosphere. Less temperature is needed to maintain hydrostatic equilibrium, thus the equatorial regions of the star have a cooler temperature than the polar regions. Von Zeipel law (1924): Teff = C g0.25 ,C is a constant Eg. B-type star Teff = 20000 K R = 3 Rסּ log g = 4.0 vsini = 300 km s–1 log g centrifugal: 3.64 Tequator / Tpole = (5634/10000)0.25 = 17327 K ΔT ≈ 2700 K Evidence for Stellar Rotation: The RossiterMcClaughlin Effect 2 1 +v 0 4 3 1 4 2 –v 3 The R-M effect occurs in eclipsing systems when the companion crosses in front of the star. This creates a distortion in the normal radial velocity of the star. This occurs at point 2 in the orbit. The Rossiter-McLaughlin Effect in an Eclipsing Binary From Holger Lehmann The Rossiter-McClaughlin Effect –v +v 0 As the companion crosses the star the observed radial velocity goes from + to – (as the planet moves towards you the star is moving away). The companion covers part of the star that is rotating towards you. You see more possitive velocities from the receeding portion of the star) you thus see a displacement to + RV. +v –v When the companion covers the receeding portion of the star, you see more negatve velocities of the star rotating towards you. You thus see a displacement to negative RV. The effect was discovered in 1924 independently by Rossiter and McClaughlin Curves show Radial Velocity after removing the binary orbital motion The Rossiter-McClaughlin Effect What can the RM effect tell you? 1. The inclination of „impact parameter“ –v –v +v +v Shorter duration and smaller amplitude The Rossiter-McClaughlin Effect What can the RM effect tell you? 2. Is the companion orbit in the same direction as the rotation of the star? –v +v +v –v Note: R-M curves are schematic, drawn by hand The Rossiter-McClaughlin Effect λ What can the RM effect tell you? 3. Are the spin axes aligned? –v Symmetric R-M distortion +v Orbit axis –v +v Asymmetric R-M distortion CoRoT-2b λ = –7.2 ± 4.5 deg HAT-P7 λ = 182 deg! HARPS data : F. Bouchy Model fit: F. Pont Lambda ~ 80 deg! Basic tools for line shape analysis: 1. The Fourier transform 2. Line bisectors Both pioneered by David Gray To derive reliable information about the line shapes requires high resolution and high signal-to-noise ratios: • R = λ/δλ ≥ 100.000 • S/N > 200-300 Fourier Transform of the Rotation Profile David Gray pioneered using the Fourier transform of spectral lines to derive information from the shapes. ∞ i(f) = 2πiλf dλ I(λ)e ∫ –∞ Where I(λ) is the intensity profile (absorption line) and frequency f is in units of cycles/Å or cycles/pixel (detector units) Because of the inverse relationship between normal and Fourier space (narrow lines translates into wide features in the Fourier domain), the Fourier transform is a sensitive measure of subtle shapes in the line profile. It is also good for measuring rotation profiles. The Instrumental Profile The observed profile is the spectral line profile of the star convolved with the instrumental profile of the spectrogaph, i(λ) What is an instrumental profile (IP)?: Consider a monochromatic beam of light (delta function) Perfect spectrograph A real spectrograph If the IP of the instrument is asymmetric, then this can seriously alter the shape of the observed line profile No problem with this IP Problems for line shape measurements It is important to measure the IP of an instrument if you are making line shape measurements If D(Δλ) is the observed profile (your data) then D(Δλ) = H(Δλ)*G(Δλ)*I(Δλ) Where: D = observed data H = intrinsic spectral line G = Broadening function (rotation * macroturbulence) I = Instrumental profile * = convolution In Fourier space: d(σ) = h(σ)g(σ)i(σ) You can either include the instrumental response, I, in the modeling, or deconvolve it from the observed profile. Fourier Transform of the Rotation Profile Fourier Transform of the Rotation Profile The Fourier transform of the rotational profile has zeros which move to lower frequencies as the rotation rate increases (i.e. wider profile in wavelength coordinates means narrower profile in frequency space). Limb Darkening Limb darkening shifts the zero to higher frequency Limb Darkening The limb of the star is darker so these contribute less to the observed profile. You thus see more of regions of the star that have slower rotation rate. So the spectral line should look like a more slowly rotating star, thus the first zero of the transform should move to lower frequencies Limb Darkening Ic/Ic0 = (1 – ε) + ε cosθ Effects of Differential Rotation on Line shapes The sun differentially rotates with equatorial acceleration. The equator rotational period is about 24 days, for the pole it is about 30 days. Differential rotation can be quantified by: ω = ω0 + ω2 sin2φ + ω4sin4φ α = ω2/(ω0 + ω2) Solar case α = 0.19 + → equator rotates faster – → pole rotates faster φ is the latitude Differential rotation parameter Effects of Differential Rotation on Line shapes Equatorial acceleration → lines narrower and more ‚Vshaped‘ Polar acceleration → lines fatter and more ‚U-shaped‘ Effects of Differential Rotation on Line shapes Equatorial acceleration Polar acceleration 1. First zero moves to higher frequency 1. First zero moves to lower frequency 2. Power in first sidelobe decreases 2. Power in first sidelobe increases Noise level for low S/N Data (≈10-50) Noise level for modest S/N Data (≈ 100-150) Noise level for modest S/N Data (≈ 250-500) The measurement of differential rotation requires data taken at high spectral resolution and high signal-to-noise (S/N) ratios. Ideally one would like to measure the first and second sidelobe, but that takes data with very high signal-to-noise ratios. Often this is not possible. The best method is to use the location of the first zero The inclination of the star has an effect on the Fourier transform of the differential profile. Note: this is for the same v sin i! Differential Rotation in A stars In 1977 Gray looked for differential rotation in a sample of A-type star and found none. This is not surprising since we think that the presence of a convection zone is needed for DR and A-type stars have a radiative envelope. Differential Rotation in A stars Gray found two strange stars. γ Boo has a weak first sidelobe and no second side lobe. γ Her has no sidelobes at all. This may be the effects of stellar pulsations. Differential Rotation in F stars In 1982 Gray looked for differential rotation in a sample of F-type star and concluded that there was no differential rotation. Spot activity on F-type stars is not seen, but they do have a convection zone so DR is possible. Differential Rotation in F stars ψ Cap α=0 α = 0.25 However, in 2003 Reiners et al. found evidence for differential rotation in F-stars What about G-type stars? For the method to work you need some rotational broadening of the line profile (several km/s), otherwise differential rotation will not affect the line shapes. Solar-type stars rotate too slowly. Velocity Fields in Stars Early on it was realized that the observed shapes of spectral lines indicated a velocity broadening in the photosphere termed „turbulence“ by Rosseland. A theoretical line profile with thermal broadening alone will not reproduce the observed spectral line profile. This macroturbulent velocity broadening is direct evidence of convective motions in the photospheres of stars From Velocity to Spectrum N(v)dv = 1 e–(v/v0)2dv π½v0 N(v)dv is the fraction of material having velocities in the range v → v + dv and v is allowed only on stellar radii. The projection of velocities along the line of sight = β0 cos θ β = λ v0cos θ c 2 Δλ 2 Δλ 1 1 = π½β cosθ exp [–( β0cosθ ) dΔλ N(Δλ)dλ = ½ exp [ –( β ) π β 0 Δλ = λ cos θ c [ [ Note that β, the width parameter, is a function of θ, β0 is constant. At disk center N(λ) reflects N(v) directly, but away from the center the Doppler distribution becomes narrower. At the limb N(Δλ) is a delta function. Including Macroturbulence in Spectra The observed spectra (ignoring other broadening mechanisms for now) is the intensity profile convolved with the macroturbulent profile: Iν = Iν0 * Θ(Δλ) Iν0 is the unbroadened profile and Θ(Δλ) is the macroturbulent velocity distribution. What do we use for Θ? The Radial-Tangential Prescription from Gray We could just use a Gaussian distribution of radial components of the velocity field (up and down motion), but this is not realistic: Horizontal motion to lane Convection zone Rising hot material Cool, sinking intergranule lane If you included only a distribution of up and down velocities, at the limb these would not alter the line profile at the limb since the motion would not be in the radial direction. The horizontal motion would contribute at the limb Radial motion at disk center → main contrbution at disk center Tangential motion at disk center → main contribution at limb The Radial-Tangential Prescription from Gray Assume that a certain fraction of the stellar surface, AT, has tangential motion, and the rest, AR, radial motion Θ(Δλ) = ARΘR(Δλ) + ATΘT(Δλ) AR e–(Δλ/ζ cos θ) + = ½ π ζRcos θ R 2 AT –(Δλ/ζ cos θ) e π½ζTsin θ And the observed flux π/2 ∫ ΘR(Δλ)*Iνsin θ cosθ dθ + Fν = 2πAR 0 π/2 2πAT ∫ ΘT(Δλ)*Iνsin θ cosθ dθ 0 T 2 The Radial-Tangential Prescription from Gray The R-T prescription produces as slightly different velocity distribution than an isotropic Gaussian. If you want to get more sophisticated you can include temperature differences between the radial and tangential flows. The Effects of Macroturbulence Macro Relative Intensity 10 km/s 5 km/s 2.5 km/s 0 km/s Pixel shift (1 pixel = 0.015 Å) Macroturbulence versus Luminosity Class Macroturbulence increases with luminosity class (decreasing surface gravity) Amplitude Relative Flux The Effects of Macroturbulence Pixel (0.015 Å/pixel) Frequency (c/Å) There is a trade off between rotation and macroturbulent velocities. You can compensate a decrease in rotation by increasing the macroturbulent velocity. At low rotational velocities it is difficult to distinguish the two. Above the red line represents V = 3 km/s, M = 0 km/s. The blue line represents V=0 km/s, and M = 3 km/s. In wavelength space (left) the differences are barely noticeable. In Fourier space (right), the differences are larger. The Effects of Macroturbulence Pure rotation Rotation affects the location of the first zero. Macroturbulence affects the size of the first side lobe and to a lesser extent the main lobe. For slowly rotating stars one should use Fourier space for measuring accurate rotational velocities Sometimes it is very important to measure the rotational velocity accurately. HD 114762 m sin i = 11 MJup Most likely vsini is 0-1 km/s. HD 114762 is an F8 star and the mean rotation of these stars is about 5 km/s. The companion could be a more massive companion, maybe even a late M-dwarf 0100200300400500.4.6.81 A word of caution about using Fourier transforms If you want to calculate the Fourier transform of the line you have to „cut out“ the line. This is the equivalent of multiplying your data with a box function. In Fourier space this is a sinc function which gets convolved with your broadening function. This changes the FT. → need to apply taper function (bell cosine, etc.) The Funny Shape of the Lines of Vega A clue may be found in the slow projected rotational velocity of Vega, an A0 V star Recall Gravity darkening Von Zeipel law (1924): equator Rotation pole Teff = C g0.25 ,C is a constant Because of gravity darkening and centrifugal force, the equator has lower gravity and a lower temperature. For a star viewed pole on this appears at the limb. Temperature/gravity sensitive weak lines will be stronger at the equator (limb) than at the poles. The Power of Spectral Line Bisectors What is a bisector? Curvature Span Bisectors as a Measure of Granulation Hot rising cell Cool sinking lane Solar Bisector Solar bisectors take on a „C“ shape due to more flux and more area of rising part of convective cells. There is considerable variations with limb angle due to the change of depth of formation and the view angle. The line profiles themselves become shallower and wider towards the limb. Bisectors as a Measure of Granulation The measurement of an individual bisector is very noisy. One should use many lines. These can be from different line strengths as one can „collapse“ them all into one grand mean. Note: this cannot be done in hotter stars the weak lines do not mimic the shape of the top portion of the bisector. Changes in the Granulation Pattern of Dwarfs Changes in the Granulation Supergiants The Granulation and Rotation Boundary Rapid rotation, Inverse „C“ bisectors Slow rotation „C“ shaped bisectors Bisectors as a Measure of Granulation Can get good results using a 4 stream model (Dravins 1989, A&A, 228, 218). These best reproduce hydrodynamic simulations 1. Granule center (rising material) 2. Granules (rising material) 3. Neutral areas (zero velocity) 4. Intergranule lanes (cool sinking material) Each has their own fractional areas An, velocity Vn, and Temperature Tn Constraints: 1. A1 + A2 + A3 + A4 = 1 2. V3 = 0 3. Mass conservation: A1×V1 + A2 ×V2 = A4×V4 Downflow = upflow Best way, is to use numerical hydrodynamic simulations Bisectors as a Measure of Granulation Examples of 4 component fits for stars from Dravins (1989) Rotation amplifies the Bisector span (Gray 1986): Using Bisectors to Study Variability The Effects of Stellar Pulsations Variations of Bisectors with Pulsations The 51 Peg Controversy Gray & Hatzes Gray reported bisector variations of 51 Peg with the same period as the planet. Gray & Hatzes modeled these with nonradial pulsations A beautiful paper that was completely wrong. Hatzes et al. More and better bisector data for 51 Peg showed that the Gray measurements were probably wrong. 51 Peg has a planet! Bisector Variations due to Spots Spot Pattern Changes in Radial Velocity due to changing shapes Note: this has the same shape as the R-M effect, as it should, the spots can be considered to be a companion (star or planet) that blocks flux from the main star Star Patches Bisectors Bisector span Star Patches ΔT = 300 K Compared to ΔT = 2000 K for sunspots Spots vs. Planets HD 166435 Radial Velocity This was reported to be another short period planet with a period of 4 days until… Spots vs. Planets Radial Velocity Ca II Brightness Color The star was found to vary in Ca II, brightness, and color with the same period as the presumed planet. This is a spotted star Correlation of bisector span with radial velocity for HD 166435. Looking at bisector variations has become a standard way of confirming planets. The spirit of David Gray continues… Disk Integration Mechanics θ Cell i,j 1. Divide the star into an x,y grid 2. At each cell calculate the limb angle θ 3. Take the appropriate limb angle intrinsic line profile from model atmospheres, or just apply limb darkening law to a line profile or even a Gaussian profile (the poor person‘s way) 4. Calculate the radial velocity using the desired vsini. Include differential rotation if desired. Doppler shift your line profile 5. Use a random number generator to calculate the radial and tangential value of the macro-turbulent velocity with maximum value ξ. Apply additional Doppler shift due to the turbulent velocity 6. If there is a spot, you can scale the flux. If there are pulsations you can add velocity field of star. 7. Can add convective velocities/fluxes 8. Take area of cell and multiply it by the projected area (cos θ) 9. Go to next i,j cell 10. Add all profiles from all cells 11. Normalize by the continuum 12. Check to make sure line behaves with vsini macro-turbulence. Make sure equivalent width is conserved.