Report Gene

7/22/20 18:18

Topic

Webinar ID

2020 Sagan

940 3997 6754

Question Det

Question

Answer(s)

How does the magnetic field hold the granules in place?

The gas on the surface of the Sun is slightly ionized, so is really a plasma. Because some of its constituents are charged, they become trapped on magnetic field lines. This causes the gas to move more easily along field lines than across them, leading to very complex “magnetohydrodynamics”.

Did you say that the first effects discussed from convection/granulation have net effect of tens of cm/s but supergranulation has net effect of m/s? If so, how is this so if supergranulation is mostly horizontal motion? Also, why does supergranulation have any effect at all if it’s characterized by horizontal flow?

I think Dr. Cegla means horizonal with respect to the local gravity vector on the surface of the Sun. This is only tangential to our line of sight at disk center. At the limb, some of this motion is in our radial direction.

Is there some project, similar to the upcoming high res spec EPRV instruments, that intends to observe the Sun?

The next talk by Annelies Mortier will be answering this.

How are sunspots formed?

See more in Charbonneau 2010/14, Brun & Browning 2017 (https://www.springer.com/gp/livingreviews/solar-physics/news/lrsp-magnetism-dynamo-action-and-the-solar-stellar-connection/15087590) for more on dynamos and how magnetic fields are generated and how this translates to spots. Essentially spots form when you get intense/concentrated regions of magnetic field.

Is "network" the same as "faculae"?

No, the ‘network’ is lower magnetic field and weaves through most of the stellar surfaces and ‘faculae’ are higher magnetic field regions and appear mostly at the stellar limbs (and appears as magnetic bright points at disc center).

Regarding photometry and RV, is one of those quantities more dominated/influenced by spots or faculae, e.g. are spots better visible in the RV and faculae better visible in photometry, or can both effects influence photometry and RV at the same level same?

Spots can show up in both. Faculae do not show up well in optical photometry but more so in UV photometry as faculae are brighter in that band. They do affect the RV with a couple m/s.

Do we know why stars are so magnetically active? What is the physical process that we understand the least when it comes to stellar variability?

The rough model is called the alpha-omega dynamo which generates magnetic fields in rotating bodies with convecting, conducting material. More accurately, in stars like the sun the magnetic field is formed at the base of the convective zone where it meets the radiative core and there is a lot of shear.

Are effects modulated by stellar rotation not averaged out? Presumably they would spend equal time on each hemisphere. Is this only an issue if the lifetime of the plage is less than half the stellar rotation period? (or technically 3/4)

By “average out” Heather means during an exposure. If a spot causes a blueshift during your observation and a redshift only a of weeks later when you’re not observing, it does not average out. Even if you observe continuously, it will “average out” in searches for long-period planets but will still be a source of noise. For short period planets (near the rotation period) it will produce spurious signals.

What is the typical accuracy expected from asteroseismology on the stellar mass and radius ? Is the accuracy expected to be better or similar than the one provided by stellar models ?

Stellar models are used even with asteroseismic measurements. It strongly depends on the star, but in general we can get mass and radius to 3-5% without such models, and ages to a few Gyr, and these all shrink by a factor of a few with the addition of astroseismic measurements.

Did I understand correctly that RV signals induced by the magnetic cycle are not due to the change in total number of spot/faculae (and therefore different "strength"/amplitude of the signals induced by them) but solely by the suppresion of the convective blueshift?

Both effects are at play. Stars are noiser during activity maximum, and some also show less convective blueshift.

Will spots or plages have effects more prominent in one of the activity indicators more than the other. Do we know already know this?

Yes, different stellar effects can have different effects on the activity indicators. It's important to remember there is no "magic" indicator (yet) that tracks it all.

What are R modes of the sun?

Check out Lanza et al 2019; the oscillations are driven by Rossby waves and the restoring force is from the Coriolis effect, timescales around 20 days (if I remember right); Annelies may address this too because we can see them in the Sun.

How are stellar pulsations modeled out, given there are variety of modes (fundamental mode, first overtones, multimodes etc.) along with radial and non-radial ones?

It depends on their strength and period. Today, p-modes are treated either by averaging (exposing for an integer number of modes) or by modeling them (making many exposures that track the ups and down and removing them). Other modes are more subtle and usually not treated yet.

Can we do EPRV for variable stars like cepheids?

Basically, yes, but I would not characterize this work as being extremely precise. Most work on them does not require “extreme precision” that I’m aware of. Their line profiles change during their pulsations so a lot of what we measure is not pure motion of the atmosphere,

Contrasts of faculae decrease in cooler stars and so RV signals decrease due to them. So most magnetic structures tend to be darker than (or cooler) than normal photosphere, but is there a more direct thermal effects on the spectral lines in cooler stars?

I’m not 100% I understand the question. But the contrast of both the bright faculae/plage and the dark spots relative to the ‘quiet’ photospheres both decreases as you go to cooler stars. But RVs are calcuated from thousands of lines and some lines or more or less temperature sensitive as well as more or less sensitive to magnetic fields (and other factors here too) and these will all impact the net effect of the interplay of convection and magnetic fields on the RVs

Related to my question, if faculae reduce the convective blue shift effect, how does the dominance of darker magnetic structures, which defintely reduce the convection too, in cooler stars (M dwarfs) contribute to this reduction of blue shift?

And could you please point me to any work that shows spots’ contrasts decreasing in cooler stars?. Is this mostly from observations or from simulations/models ?

Is all stellar activity, like pulsation, granulation etc. caused by interaction between convection and magnetic field?

Yes, granulation is the manifestation of convection on the stellar surface. The convection drives and excited the pressure-mode osscilations. And the magnetic field will always interplay with the convection.

Yes, it’s all tied up, but even a star with no magnetic field would have granulation and p-modes.

What is the difference ( in terms of activity induced RV signal) of having a heavy spot coverage stars versus a stellar surface less spotted but spots evolving more rapidly?

If there are so many spots that they’re almost everywhere then there’s less impact on the RVs because the convection is suppessed almost the same everywhere and the darkenss from the spots is almost everywhere. If the spots evolve more rapidly then the main difference is you’ll see the velocities change more rapildly with that, as the spot decays you should see the decay in the RVs (or photometry) as well.

IS every star active?

Every star has some level of magnetic activity, but some are so quiet that we do not notice the activity. Low-mass subgiants, for instance.

Why does the corona have a very high temperature?

It is very tenuous and it gets heated by waves.

Observing by Solar telescope, do you mean observing the stray light of the sun or observing the sun directly with filters? This might have different implications if we are looking at the integrated solar light or from any particular region of the Sun?

The EPRV solar telescopes image the whole solar disk into an integrating sphere. Exceept for differential extinction across the solar disk, they produce sun-as-a-star measurements.

How do you want to to detect RVs due to Earth in the Sun when observing from Earth. Isnt the relative RV signal following Earths orbit?

The motion of the telescope with respect to the Sun in these plots has been removed. I think Annelies was simply comparing the noise level we see in the Sun with the signal strength we would observe from the Earth if it were observing the Sun from afar.

Have these temporal offsets between RVs and stellar activity indicators also been observed in other stars?

Check Santos et al. 2014 about HD41248 (https://ui.adsabs.harvard.edu/abs/2014A%26A...566A..35S/abstract)

What is the origin of the time-shift for the correlations between RVs, Bisector etc.? Do we expect to see the same for other stars?

It’s not completely clear, but features on the approaching or receding limb have a strong RV effect if they are rotationally modulated, while these features feature in the spectra most promimently when they are at disk center.

Do all spots and faculae follow the stellar rotation period?

They are blemishes at particular locations on the stellar surface, so they have to be connected the stellar rotation period since the rotation period modulates when we do and don’t see the spots/facuale. But sometimes these appear at multiple locations so exactly how to map it to the stellar rotation period can be tricky.

What is the physics behind RV magnetic flux density correlation?

The interplay between magnetic field and convection drives almost all velocity variations originating from the stellar surface, hence the magnetic flux itself correlates strongly with those velocities

is this temporal offset in RV and indicators eg FWHM and BIS also seen in things like longitudinal mag field?

Honest answer: we do not know yet (I think)

What happens if you do RV analysis on a galaxy like Andromedae as a star ?

The integrated light of galaxies is a complex combination of light from many stars at many velocities dominated by red giant stars. I don’t think it would be very meaningful to mesaure their velocity precisely.

How would someone measure magnetic field values on a star currently?

spectropolarimetry for example. Check papers by F. Donati, O. Kochukhov, ...

Isn’t the activity of the star dependant to its spectral type? Do we see activity in A type of B type stars where don't have convective layers?

Yes and yes. In my talk, I was focusing on Sun-like stars which is why I stressed so heavily the interplay of convection and magnetic fields. There is some evidence for spots on A stars (maybe B stars too, but I am not sure), and we don’t really know what is driving the magnetic field/spot generation in these stars (to my knowledge).

I am about to start my PhD. I am interested in developing models to filter out stellar noise from RV signals of an exoplanet so that we can extract true RV signals. I am curious to know that to do so, do we need to start from sun? Or could we do it for any star of our interest?

You can start with the mystery systems in the hands-on sessions for this workshop :) But yes, the Sun is an excellent starting point too!

Could you please explain the mechanism that drives magnetic fields in fully convective stars as opposed to the shell dynamo that is for the solar-type stars?

That’s a big question! I would refer you to work by Matthew Browning on M dwarf magnetic field generation.

Question for the panel discussion: 1) why would convection produce an overall blue shift, if there’s both upward and doward flow in the vertical direction; and why is the blueshift being suppressed during activity cycle? 2) If linear correlation between activity indicator and RVs may introduce extra noise, then what would be the best way to confirm the RVs we see are from the stellar variability?

The blueshifed bubbles of plasm (granules) take up moer surface area and are brighter, both of these effects make the blueshift dominate over the redshift. Concentrations of magnetic field inhibit the convective motions, this means it suppresses some of the net blueshift, and why we see changes over the activity cycle (where the magnetic field is changing). At present, there is no ‘best way’ to disentangle stellar variability and is an active source of research, there are many techniques — check out Jenn’s talk now :)

ty!

What are some typical S-values for different types of stars?

One reference for this is the “The Mount Wilson Observatory S-index of the Sun” by Ricky Egeland et al 2016, from this you can see the s-index tends to be around 0.15 - 0.2 for the Sun. Another variation on the S-index is the log R’HK, where less active is closer to -5 and more active is closer to -4.6.

Subgiant stars can have S values down towards 0.12. Cooler stars have higher values, up towards 1 for M dwarfs. log(R’HK) can be as low as -5.1 for subgiants and as high as -4.1 (or higher!) for young stars.

We ask for ~3 cm/sec measurment to find temperate planets around Sun-like stars, the Shift from the star could be as high as 200 times that. This idicat eyou might need to know the stellar variability at an accuracy of 1% to distangle the two. Is this a fair way to describe the challenge, and can such measurments even be made?

Yep, you are asking exactly the right question. And we just don't know how far we will be able to get in the next years.

It’s a fair way to compare the *magnitudes* of the two effects, but remember that it’s also possible to get pretty far with averaging over large numbers of observations or mitigating them using the strategies Jenn Burt described.

Is the Chomatic Index also working for slow rotating star observed in the visible ?

The chromatic index is a very good indicator of potential trouble - if you see it vary in sync with the RV you know that you are looking at activity, not at an exoplanet. But it is much harder to use the chromatic index to "correct" the RV data.

Can RV signals induced by activity scale up with the instrumentation we use (resolution of spectrograph or rv extraction technique) or these effects always produce similar signal strengths in data and depends solely on the physics of the process?

The RV signal is the same and is just a product of physics. The advantage of higher resolution and data reduction techniques is that you can use them to probe other spcetral signatures to try to diagnose spurious Doppler shifts from center-of-mass Doppler signals.

Question for Dr. Jennifer Burt: Based on the literature, we can see that Ca HK, H alpha, and Ca IRT indicators show different types of correlation to each other. What parameters determine the correlation between these activity indicators?

I think the main driver is the where the lines that we’re measuring are formed within the star. Stars that form at different depths and pressures are sensitive to different types of stellar variability. For example, the Ca II H&K lines are influenced by both the stellar chromosphere and photosphere, whereas the H-alpha lines seem to be more closely tied to the chromosphere

Other than the Ca IRT, are there other activity indicators in the NIR? One example would be the He 10830 triplet, but these are not as well studied.

The Ca IR Triplet is (I think) the best studied example so far, but the search for new IR activity indicators is ongoing and as instruments like HPF, CARMENES, SPIROU, etc continue operating we’re hoping to identify other promising options

Heather Cegla mentioned that the shape of the bisector is c-shaped because of the superposition of light emitted from the centers of granules (plasma moving up, hot) and the intergranular lanes (plasma moving downwards, cold). How do we get the bisector to move as showed by Jennifer Burt (in the Queloz 2001 figure)? Do we need a varying size of the granules?

The Queloz signal was due to a rotationally modulated spot, which causes a different kind of bisector changeh as it blocks light first from the blueshifted side of the star, then the redshifted side.

If you’re referring to the sort of backwards C-shape in the CCF, I think this is because the template mask is weighted by line depth so the CCF is not an exactly a composite line profile. But you can get essentially ‘reverse granulation’ in hotter/giant stars, there’s a ‘granulation boundary’ here, in Gray & Toner 1986 they show this happening for supergiant F stars (also seen in the Stellar Photosphere’s book by Gray)

How do we avoid saturation if using long exposures to mitigate RV rms?

You can bin shorter observations. It has the same effect as taking the long exposure and avoids the saturation.

Have there been studies looking at lines all across the spectral range of typical spectrographs looking for correlations between the typical activity indicators and smaller lines that might also encode activity information?

Alex Wise did some great work on this in 2018, and I’m going to mention it at the start of the panel discussion!

Regarding the analysis of the periodograms, how do you discern between a peak caused by a second/third/etc planet and one due to stellar variability?

That is indeed the question. From just one periodogram, you can’t. But you can look at whether the signal related to the identified periodicities stays stable over time, whether its period is related to the stellar rotation period,... if the same periodicity shows up in a periodogram of an indicator, it likely is activity.

In order to fit the periods harmonics, what should be our precision on the rotationnal period value ? Could differential rotation make inefficient such methods ?

Yes, differential rotation will cause problems. It is somewhat mitigated by the fact that spots tend to congregate near common latitudes.

Can you please talk more about the specifics of fitting rotation periods and harmonics with additional Keplerians.

The basic approach is to include an additional “planet” in your RV model where the period is tied to the star’s rotation period or a harmonic of the rotation period. But we know that stellar activity is often not purely periodic, and so things like GPs instead allow for quasiperiodic signals to be fit instead

Why exactly are harmonics generated? and by what?

Because of the distribution of active regions on the surface and the fact you only see half a star of course

If I apply a stellar indicator am I not in the risk of creating a signal if the RV and this indicator are opposed in phase ?

You definitely want to make sure that your addition of an activity-based component to your RV model is well motivated by some other source of information whether that’s photometry, activity indicators, etc. But the model should allow for the planets and activity signals to have different phases

As a comment, in my paper Cretignier et al.+20 we pushed the rms on alpha cen B down to 89 cm/s :)

That paper is downloaded into my “Papers to Read” folder, but clearly I need to bump it to the top of the list :)

Detrending a RV data with one of the activity indicators could add spurious signals if these activity indicators are anti-correlated with RV?

No, detrending generally only removes signals, it does not add them because you tune the detrending according to the amount of correlation.

Does the rotation period being smaller than the evolutionary timescale apply to stars of other spectral types and ages? Does it correlate with the rotation period itself?

Evolutionary timesccales are millions of years to gigayears, and rotation periods are on days to months — so the rotation periods are always smaller than the evolutionary timescales. But stellar rotation is indeed linked to the evolution of the star, e.g. stars spin-down over time, so they are connected.

Paul Robertson has a paper on exactly this topic for young M dwarfs (https://ui.adsabs.harvard.edu/abs/2020ApJ...897..125R/abstract) where they see spots that persist for hundreds of rotations - many more than is typical for Sun-like stars.

If some of these stellar activity indicators don’t correlate linearly with time, just curious how would they be used to mitigate stellar activity in RV measurements?

The likely show similar periodicity structure, so you could use a GP on both time series simultaneously, see eg Vinesh Rajpaul’s work.

If we can remove almost all stellar noises through different observing techniques and smart data analysis then why do we need to study them in details in order to get true RV of exoplanets?

Because we don’t actually know how to remove them yet! But understanding their origin and form, we can tailor our observing and modeling methods to remove them faithfully.

Is there any reason beside the huge costs for which we don’t completely move all exoplanet search in space, “outside” from all telluric emission/absorption?

The huge cost is a huge barrier - 2 orders of magnitude. Additionally, once you launch something into space you can (usually) no longer fix it if it breaks. Our ground based instruments can be serviced and are usually designed with operational lifetimes of a decade or more. Most space missions, even NASA flagships, only have a few (1-3) year requried lifetime. Most go longer, but that's not usually a requirement.

Huge cost is indeed the big issue, especially because you need a rather large telescope to collect enough photons. There is a pre-recorded talk by Peter Plavchan discussing the space option in much more detail.

Is it hard or almost impossible to get the LSF of the detector to actually deconvolve and then divide by a telluric model?

Good question. In principle, you can estimate the instrumental LSF of fiber-fed spectrograph using a laser light source (for example). But, even with a perfect model of the LSF, the deconvolution is problematic. I would guess that at the m/s level the approach you suggest is sufficient at some wavelengths.

About OH emission lines: Why would it be difficult/problematic to just mask them out; if they are few?

Thanks for your question. At some wavelengths, the OH emission lines are pretty sparse and can be masked out. But, in other spectral regions they are very numerous and are bright enough that the light from the lines starts to become a problem in terms of the S/N of the underlying stellar spectrum.

What is the effective range of EPRV?

In terms of wavelength? From the blue end of the visible to 2.4 microns with iSHELL are current limits. John Johnson did try UV RVs with HST archival data, and I don’t know of anyone that has done mid-infrared radial velocities for exoplanets.

Is there some examples of star showing an anticorrelated behaviour of the « RV magnetic cycle » compared to the S-index ? If yes, what is the physical explanation ?

Low-mass stars in binaries tend to be “inflated”, and this is often explained as due to the fact that they are tidally locked and more rapidly rotating and thus has more spots.

Sorry this question is not related to the talks today. I had trouble understanding the radial velocity vs phase plot. Can you please explain what is the phase represented in the x axis?

You derive a period of your signal, for example your planet. Than you phase your model and data with that period to make the sinusoidal shape better visible for the eye.

Yes, that’s the orbital phase, basically the mean anomaly M of the orbit. It allows you to see all of the observed orbits “stacked” up so the signal is clearer, similar to the way an oscilliscope works.

Would active/inactive lines have to be vetted differently depending on the spectral type or luminosity type?

The short answer is yes.

Yesterday and today, we have heard that P_decay [d]>P_rot [d] is a good rule of thumb when fitting stellar activity with quasi-periodic GPs, and I strongly agree that the evolution time-scale parameter needs to be suffienctly large to modulate a meaningul (and realistic) QP-Signal. But what about slow rotators (older stars) with rotation periods of several months? I would argue that it is possible in such cases that spots evolve and decay over less than one rotation, but the stellar rotation should still manifest itself with a quasi-periodic signal driven by the period of the rotation (as spots also dominate in RVs compared to faculae which ar elonger lived) but with a lot of phase-shifts. I would be interested in the opinion of the panel speakers about this possibility.

This is certainly a good thing to worry about! Yes, especially for inactive stars spots can come and go faster than they rotate in and out of view. Indeed, spot lifetimes on the Sun are typically < its rotation period (although stars can still sometimes have “active longitudes” where spots preferentially pop up).

Have there been studies on activity in twin binary system? Could these maybe shed light on why similar stars could produce different RV jitter?

In reference to what Jennifer mentioned when getting RVs stars with similar photometric curves

Should we observe with high-res spectropolarimetrty alongside EPRV in future? Or is it too much effort?

SPIRou is exploring this in detail for more active stars, and spectropolarimetry is definitely on the table as a possible technique for mitigating activity.

Does stellar activity strongly depend on their chemical compositions?

No, not really. The biggest effect is that very metal rich or poor stars have many more or fewer calcium ions, which changes the line depths and strengths, but it’s not a strong effect. Because these are chromospheric lines, they are also sensitive to certain effects which depend on the surface gravity of the stars. Both effects can be corrected for in activity measurements, although this has not been very well calibrated as far as I know.

Does the study of Ca II H&K work for understanding the activity of M-dwarfs? Since they are dim and the S/N is not good enough in that wavelength range.

For late M dwarfs it is indeed less useful because it is so far from the Wien peak. H-alpha is usually better, and for cooler stars the Ca II IR triplet.

I get the mag. activity cycle part. I didnt quite understand the 1 d alias getting reflected specifically at 4105 d.

Aliases show up at frequencies that are the differece between a real signal and your observing window—they are beat frequencies. A signal that gives you power only at a frequency very close to a solar day will slowly shift in phase over many years and look like a long-term trend. This means that a periodogram will show power at both the true signal period and its alias against 1 day (which for long periods is very very close to 1 day).

100

Say the frequency is roughly 1200 days, what timescale would you use on the smoother?

There is probably not one right answer; it's best to try several and see how it changes your result. I would probably try ~1 week, ~2 weeks, ~1 month.

101

Could you give as a typical reference of detrending by S-index and the discovery of exoplanet ? So we can read a real example an analysis of data.

Xavier Dumusque’s paper on alpha Cen Bb (https://www.nature.com/articles/nature11572) is a good example (even though the planet may not really exist, the analysis is good).

104

I still don’t understand how the 1-day alias can come from the instrument. As I understand these aliases is that the 1-day alias come from the sampling alone (we take a measure every night if possible). But if we were to take a measure only every two nights, we should get an alias at two days also. Where the instrument could yield a 1-day alias ?

Instruments can produce 1 day power for instance due to systematics related to the airmass of the star you observe which is changing during the night and will be the same 1 day after. This is really a function of practical observing limitations, which you indicate. We often are observing stars at roughly the same time of night for many nights in a row. This is because to optimize a survey for observing efficiency, you trNo, the ‘network’ is lower magnetic field and weaves through most of the stellar surfaces and ‘faculae’ are higher magnetic field regions and appear mostly at the stellar limbs (and appears as magnetic bright points at disc center).y to get the stars when they are high in the sky, near the meridian. That minimizes atmospheric dispersion issues and also telescope slew time. So over time we bake in this 1 day alias. If you observed every 2 days, you'd get a 2 day alias but also a 1 day alias because that is 1/2 of 2 days (and harmonics pop up in frequency analyses).

Ok but still, the systematics of your instrument will have an impact of the velocity measured, airmass conditions, lunar illumination, water vapour etc… Why would this would have an impact on a possible modelisation of the overall signal by a period of 1 day ?

Not sure to get your question here. You usually get peaks around 1-day because of 1) aliasing (sampling issue) or 2) because the signal is simply real. In particular the 1-day signal is the alias of a perharps underyling 1-year signal. So both intrumental systematics acting on either 1 day (airmass) or 1 year (there are so much that could be listed…) will put power at 1 day.

It is clear to me that the 1-day alias is due to the sampling issue (that’s how I learnt it) but I don’t understand why the instrument systematics will act on a sampling issue. How for instance, the airmass alone could create an alias at 1-day ?

Due to differentiel extinction, a change in airmass is producing a change of your spectrum spectral energy distribution. You lose a lot of blue flux at high airmass. A change in your color spectrum directly induce an effect in RV since in some ways the spectrum is changing by « itself » the weight of the stellar lines. It is one way airmass can introduce RV effect but there are others. Color correction of the spectrum are usually performed to avoid that (see Malavolta+17 for instance).

Ok thanks, I will have to let that sink in because it is a bit difficult for me to see how in the end it could add power at 1 day specifically because as Chad mentioned, you try to observe at the highest in the night so usually, you will observe the same star at the same airmass approximately over days. So my guess is that the change in RV is not expected to be this important on a day by day basis.

You can believe me that stars are absolutely not always observed at the minimum airmass… I observed at two occasions, and sometimes the schedule is so hard to respect that you try to put the observations where you can. In particular some stars have to be observed 3 times in night with 2 hours between observations to dumped out stellar oscillations. in that case, you will probe airmass between 1.5 (even lower) and 1.

Yeah I know I too made some observations on site (SOPHIE) and remote (HARPS) but usually, if the airmass is bad for a star, chances are that the same program should apply on the following nights

109

What if you don’t have an S-index (say due to not enough data)? Is there anything you can do to detrend the data/ identify signals that are more likely to be stellar activity than a planet signal?

Yeah, there’s a whole host of various indicators, one we showed during this session was also the full-width-half-maximum (FWHM) of the CCF. Jenn Burt and Annelies Mortier also addressed many in their talks, such as H-alpha etc. In DACE you can check out which ones are hosted on the interface by clicking on the drop down menu.