EUV analysis of polar plumes

Three polar plumes were studied using Skylab Mg x and O vi data. The plumes lie within the boundaries of a polar coronal hole. We find that the mean temperature of the plumes is about 1.1 × 10⁶ K and that they have a small vertical temperature gradient. Densities are determined and found consistent with white light analyses. The variation of density with height in the plumes is compared with that expected for hydrostatic equilibrium. As is the case for other coronal features, polar plumes will be a source of solar wind if the magnetic field lines are open. On the basis of the derived plume model and estimates of the numbers of plumes in polar coronal holes, it appears that polar plumes contain about 15% of the mass in a typical polar hole and occupy about 10% of the volume.Skylab Solar Workshop post-doc appointee 1975–1976. The Skylab Solar Workshops are sponsored by NASA and NSF and managed by the High Altitude Observatory, National Center for Atmospheric Research.     

A Time-Evolving 3D Method Dedicated to the Reconstruction of Solar Plumes and Results Using Extreme Ultraviolet Data

An important issue in the tomographic reconstruction of the solar poles is the relatively rapid evolution of the polar plumes. We demonstrate that it is possible to take into account this temporal evolution in the reconstruction. The difficulty of this problem comes from the fact that we want a four-dimensional reconstruction (three spatial dimensions plus time) whereas we only have three-dimensional data (two-dimensional images plus time). To overcome this difficulty, we introduce a model that describes polar plumes as stationary objects whose intensity varies homogeneously with time. This assumption can be physically justified if one accepts the stability of the magnetic structure. This model leads to a bilinear inverse problem. We describe how to extend linear inversion methods to these kinds of problems. Studies of simulations show the reliability of our method. Results for SOHO/EIT data show that we can estimate the temporal evolution of polar plumes to improve the reconstruction of the solar poles from only one point of view. We expect further improvements from STEREO/EUVI data when the two probes will be separated by about 60°.     

A Chandra and XMM–Newton study of the wide-angle tail radio galaxy 3C 465

We have observed the prototypical wide-angle tail (WAT) radio galaxy 3C 465 with Chandra and XMM–Newton . X-ray emission is detected from the active nucleus and the inner radio jet, as well as a small-scale, cool component of thermal emission, a number of the individual galaxies of the host cluster (Abell 2634), and the hotter thermal emission from the cluster itself. The X-ray detection of the jet allows us to argue that synchrotron emission may be an important mechanism in other well-collimated, fast jets, including those of classical double radio sources. The bases of the radio plumes are not detected in the X-ray, which supports the model in which these plumes are physically different from the twin jets of lower-power radio galaxies. The plumes are in fact spatially coincident with deficits of X-ray emission on large scales, which argues that they contain little thermal material at the cluster temperature, although the minimum pressures throughout the source are lower than the external pressures estimated from the observed thermal emission. Our observations confirm both spatially and spectrally that a component of dense, cool gas with a short cooling time is associated with the central galaxy. However, there is no evidence for the kind of discontinuity in external properties that would be required in many models of the jet–plume transition in WATs. Although the WAT jet–plume transition appears likely to be related to the interface between this central cool component and the hotter intracluster medium, the mechanism for WAT formation remains unclear. We revisit the question of the bending of WAT plumes, and show that the plumes can be bent by plausible bulk motions of the intracluster medium, or by motion of the host galaxy with respect to the cluster, as long as the plumes are light.     

Two classes of volcanic plumes on Io

Comparison of Voyager 1 and Voyager 2 images of the south polar region of Io has revealed that a major volcanic eruption occured there during the period between the two spacecraft encounters. An annular deposit ～1400 km in diameter formed around the Aten Patera caldera (311°W, 48°S), the floor of which changed from orange to red-black. The characteristics of this eruption are remarkably similar to those described earlier for an eruption centered on Surt caldera (338°W, 45°N) that occured during the same period, also at high latitude, but in the north. Both volcanic centers were evidently inactive during the Voyager 1 and 2 encounters but were active sometime between the two. The geometric and colorimetric characteristics, as well as scale of the two annular deposits, are virtually identical; both resemble the surface features formed by the eruption of Pele (255°W, 18°S). These three very large plume eruptions suggest a class of eruption distinct from that of six smaller plumes observed to be continously active by both Voyagers 1 and 2. The smaller plumes, of which Prometheus is the type example, are longer-lived, deposit bright, whitish material, erupt at velocities of ～0.5 km sec^?1, and are concentrated at low latitudes in an equatorial belt around the satellite. The very large Pele-type plumes, on the other hand, are relatively short-lived, deposit darker red materials, erupt at ～1.0 km sec^?1, and (rather than restricted to a latitudinal band) are restricted in longitude from 240° to 360°W. Both direct thermal infrared temperature measurements and the implied color temperatures for quenched liquid sulfur suggest that hot spot temperatures of ～650°K are associated with the large plumes and temperatures <400°K with the small plumes. The typical eruption duration of the small plumes is at least several years; that of the large plumes appears to be of the order of days to weeks. The two classes therefore differ by more than two orders of magnitude in duration of eruption. Based on uv, visible, and infrared spectra, the small plumes seem to contain and deposit SO₂ in their annuli whereas the large plumes apparently do not. Two other plumes that occur at either end of the linear feature Loki may be intermediate or hybrid between the two classes, exhibiting attributes of both. Additionally, Loki occurs in the area of overlap in the regional distributions of the two plume classes. Two distinct volcanic systems involving different volatiles may be responsible for the two classes. We propose that the discrete temperatures associated with the two classes are a direct reflection of sulfur's peculiar variation in viscosity with temperature. Over two temperature ranges (～400 to 430°K and >650°K), sulfur is a low-viscosity fluid (orange and black, respectively); at other temperatures it is either solid or has a high viscosity. As a result, there will be two zones in Io's crust in which liquid sulfur will flow freely: a shallow zone of orange sulfur and a deeper zone of black sulfur. A low-temperature system driven by SO₂ heated to 400 to 400°K by the orange sulfur zone seems the best model for the small plumes; a system driven by sulfur heated to >650°K by hot or even molten silicates in the black sulfur zone seems the best explanation for the large plume class. The large Pele-type plumes are apparently concentrated in a region of the satellite in which a thinner sulfur-rich crust overlies the tidally heated silicate lithosphere, so the black sulfur zone may be fairly shallow in this region. The Prometheus-type plumes are possibly confined to the equatorial belt by some process that concentrates SO₂ fluid in the equatorial crust.     

On the inhomogeneous structure of the chromosphere-corona transition region above sunspot umbrae

Multiple wavelength observations of sunspot umbrae can only be expalined by an inhomogeneous, two-component model for the structure of the umbral transition region and lower corona. The ‘Wroclaw-Ondrejov sunspot model’ was a first step in this direction. This working model has now been improved using analytic expressions for the atmospheric structure in each component and fitting the free parameters to recent sunspot observations, particularly in EUV lines. The main component has a shallow transition region and a deep-set corona. The second, ‘active’ component has a vast transition region in relatively cool fine structure elements embedded in the coronal main component. The spatial filling factor of this active component amounts to 5–10% in sunspots with bright EUV plumes, but is is more than ten times smaller in sunspot without such plumes. Observations with high spatial and temporal resolutions are necessary to understand in more detail the basic physical processes.     

Moist convection and the abundance of water in the troposphere of Jupiter

A model for narrow moist convective plumes in the lower troposphere of Jupiter can reconcile the apparent depletion of water in that region of the atmosphere with a near-solar abundance of oxygen throughout the interior.     

Coronal polar plumes

The purpose of this investigation was to determine what connection exists between coronal plumes and polar surface features. To this end the properties of plumes were re-examined by making a detailed statistical analysis of photographs of three eclipses (1962, 1963, and 1965) of the last sunspot minimum. It is found that a ‘typical’ plume has a core density ≈ 10⁸ cm^?3, a half width ≈ 3.3 × 10⁴km, and a density profile with distance r from its axis characterized by $$N \approx {\text{10}}^{\text{8}} \left( {1 - \frac{r}{{3.9 \times 10^4 }}} \right)^{1.6} $$ There is some (although only weak) regularity in the projected spacing of plumes with a mean separation of ≈ 7 × 10⁴ km. The relation between plumes and various surface features is examined. Although little direct evidence can be assembled, we conclude that a direct connection exists between plumes and photospheric faculae, bright K₃ faculae, and the small-scale magnetic structure present in the chromospheric network. It is hypothesized that plumes originate at the bright cores of the rosettes which lie along the chromospheric network. The distribution of magnetic field in the corona above a surface covered with idealized chromospheric network cells is calculated. The fact that the shape and size of the magnetic flux tubes originating from the rosette agrees with that of observed plumes supports the hypothesis.     

2D and 3D Polar Plume Analysis from the Three Vantage Positions of STEREO/EUVI A, B, and SOHO/EIT

Polar plumes appear as elongated objects starting at the solar polar regions. Here we analyze these objects from a sequence of images taken simultaneously by the three spacecraft telescopes STEREO/EUVI A and B, and SOHO/EIT. We establish a method capable of automatically identifying plumes in solar EUV images close to the limb at 1.01?–?1.39?R _⊙ in order to study their temporal evolution. This plume-identification method is based on a multiscale Hough-wavelet analysis. Then two methods to determine the 3D localization and structure of the plumes are discussed: first, tomography using filtered back-projection and including the differential rotation of the Sun and, second, conventional stereoscopic triangulation. We show that tomography and stereoscopy are complementary for studying polar plumes. We also show that this systematic 2D identification and the proposed methods of 3D reconstruction are well suited to identify plumes individually and also to analyze the distribution of plumes and inter-plume regions. Finally, the results are discussed, focusing on plume position and cross-sectional area.     

Hydrodynamical and radiative transfer modeling of meteoroid impacts into Saturn's rings

In a small hypervelocity impact, superheated gas and particles glow brightly with thermal emission for a brief time interval at short wavelengths; this phenomenon is referred to as an impact flash. Over the past decade, impact flashes have been observed on the Moon and in the laboratory in both the IR and visible portions of the spectrum. These phenomena have been used to constrain impactor parameters, such as impact size, velocity and composition. With the arrival of the Cassini spacecraft at Saturn, we embarked on a study of impact flashes in Saturn's rings. We present results on the feasibility of observing impact flashes and therefore estimating the flux of meteoroids impacting Saturn's rings using Cassini's Ultraviolet Imaging Spectrograph (UVIS). Our modeling effort is two-fold. We start by simulating impacts using the CTH hydrodynamical code. Impacts involve an icy ring particle and a serpentine meteoroid, modeled with the ANEOS equation of state. The objects are centimeters to meters in diameter and collide at 30 to 50 km s⁻¹. We then use the resulting temperatures and densities of the impact plumes in a radiative transfer calculation. We calculate bound-free, free-free, electron scattering and negative ion opacities along a line-of-sight through the center of each impact plume. Our model has shown that impact flashes will not be seen with the UVIS because (1) the plumes are optically thick when their central temperatures are high, with photosphere temperatures too cool to emit observable UV flux and (2) when the plumes become optically thin, even the hottest region of the plume is too cool to observe in the UV. This corroborates the lack of UVIS impact flash detections to date. Impact flashes are not likely to be seen by other Cassini instruments because of the short lifetimes of the plumes.     

Constraints on mantle plumes on Venus: Implications for volatile history

The analysis of Venus’ gravity field and topography suggests the presence of a small number of deep mantle plumes (～9). This study predicts the number of plumes formed at the core–mantle boundary, their characteristics, and the production of partial melt from adiabatic decompression. Numerical simulations are performed using a 3D spherical code that includes large viscosity variations and internal heating. This study investigates the effect of several parameters including the core–mantle boundary temperature, the amount of internal heating, and the mantle viscosity. The smallest number of plumes is achieved when no internal heating is present. However, scaling Earth’s radiogenic heating to Venus suggests a value of ～16 TW. Cases with internal heating produce more realistic lid thickness and partial melting, but produce either too many plumes or no plumes if a high mantle temperature precludes the formation of a hot thermal boundary layer. Mantle viscosity must be reduced to at least 10²⁰ Pa s in order to include significant internal heating and still produce hot plumes. In all cases that predict melting, melting occurs throughout the upper mantle. Only cases with high core temperature (>1700 K) produce dry melting. Over time the upper mantle may have lost significant volatiles. Depending on the water content of the lower mantle, deep plumes may contribute to present-day atmospheric water via volcanic outgassing. Assuming 50 ppm water in mantle, 10 plumes with a buoyancy flux of 500 kg/s continuously erupting for 4 myr will outgas an amount of water on the order of that in the lower atmosphere. A higher level of internal heating than achieved to date, as well as relatively low mantle viscosity, may be required to achieve simulations with ～10 plumes and a thinner lid. Alternatively, if the mantle is heating up due to the stagnant lid, the effect is equivalent to having lower rates of internal heating. A temperature increase of 110 K/byr is equivalent to ?13 TW. This value along with the internal heating of 3 TW used in this study may represent the approximate heat budget of Venus’ mantle.     

The Geometric Spreading of Coronal Plumes and Coronal Holes

The geometric spreading in plumes and in the interplume region in coronal holes is calculated, using analytic and numerical theoretical models, between 1.0 and 5.0 R. We apply a two-scale approximation that permits the rapid local spreading at the base of plumes (f_l) to be evaluated separately from the global spreading (f_g) imposed by coronal hole geometry. We show that f_l can be computed from a potential-field model and f_g can be computed from global magnetohydrodynamic simulations of coronal structure. The approximations are valid when the plasma beta is small with respect to unity and for a plume separation small with respect to a solar radius.     

Galileo observations of volcanic plumes on Io

Io's volcanic plumes erupt in a dazzling variety of sizes, shapes, colors and opacities. In general, the plumes fall into two classes, representing distinct source gas temperatures. Most of the Galileo imaging observations were of the smaller, more numerous Prometheus-type plumes that are produced when hot flows of silicate lava impinge on volatile surface ices of SO₂. Few detections were made of the giant, Pele-type plumes that vent high temperature, sulfur-rich gases from the interior of Io; this was partly because of the insensitivity of Galileo's camera to ultraviolet wavelengths. Both gas and dust spout from plumes of each class. Favorably located gas plumes were detected during eclipse, when Io was in Jupiter's shadow. Dense dust columns were imaged in daylight above several Prometheus-type eruptions, reaching heights typically less than 100 km. Comparisons between eclipse observations, sunlit images, and the record of surface changes show that these optically thick dust columns are much smaller in stature than the corresponding gas plumes but are adequate to produce the observed surface deposits. Mie scattering calculations suggest that these conspicuous dust plumes are made up of coarse grained “ash” particles with radii on the order of 100 nm, and total masses on the order of 10⁶ kg per plume. Long exposure images of Thor in sunlight show a faint outer envelope apparently populated by particles small enough to be carried along with the gas flow, perhaps formed by condensation of sulfurous “snowflakes” as suggested by the plasma instrumentation aboard Galileo as it flew through Thor's plume [Frank, L.A., Paterson, W.R., 2002. J. Geophys. Res. (Space Phys.) 107, doi:10.1029/2002JA009240. 31-1]. If so, the total mass of these fine, nearly invisible particles may be comparable to the mass of the gas, and could account for much of Io's rapid resurfacing.     

The plasma plumes of Europa and Callisto

We investigate the proposition that Europa and Callisto emit plasma plumes, i.e., a contiguous body of ionospheric plasma, extended in the direction of the corotation flow, analogous to the plume of smoke emitted in the downwind direction from a smokestack. Such plumes were seen by Voyager 1 to be emitted by Titan. We find support for this proposition in published data from Galileo Plasma Science and Plasma Wave observations taken in the corotation wakes of both moons and from magnetometer measurements reported from near the orbit of, but away from, Europa itself. This lends credence to the hypothesis that the plumes escaping from the ionospheres of Europa and Callisto are wrapped around Jupiter by corotation, survive against dispersion for a fairly long time and are convected radially by magnetospheric motions. We present simple models of plume acceleration and compare the plumes of the Europa and Callisto to the known plumes of Titan.     

On the sunspot transition region

The EUV line emission and relative line-of-sight velocity in the transition region between the chromosphere and corona of 36 sunspot regions are investigated, based on observations with the Coronal Diagnostic Spectrometer – CDS and the Solar Ultraviolet Measurements of Emitted Radiation – SUMER on the Solar and Heliospheric Observatory – SOHO. The most prominent features in the transition-region intensity maps are the sunspot plumes. In the temperature range between log T=5.2 and log T=5.6 we find that 29 of the 36 sunspots contain one or two sunspot plumes. The relative line-of-sight velocity in sunspot plumes is high and directed into the Sun in the transition region, for 19 of the sunspots the maximum velocity exceeds 25 km s^?1. The velocity increases with increasing temperature, reaches a maximum close to log T=5.5 and then decreases abruptly.

Attention is given to the properties of oscillations with a period of 3 min in the sunspot transition region, based on observations of six sunspots. Comparing loci with the same phase we find that the 3-min oscillations affect the entire umbral transition region and part of the penumbral transition region. Above the umbra the observed relation between the oscillations in peak line intensity and line-of-sight velocity is compatible with the hypothesis that the oscillations are caused by upward-propagating acoustic waves. Information about intensity oscillations in the low corona is obtained from observations of one sunspot in the 171 Å channel with the Transition Region And Coronal Explorer – TRACE. We conclude that we observe the 3-min sunspot oscillations in the chromosphere, the transition region and the low corona. The oscillations are observable over a wider temperature range than the sunspot plumes, and show a different spatial distribution than that of the plumes.

     

Surface changes on Io during the Galileo mission

A careful survey of Galileo SSI global monitoring images revealed more than 80 apparent surface changes that took place on Io during the 5 year period of observation, ranging from giant plume deposits to subtle changes in the color or albedo of patera surfaces. Explosive volcanic activity was discovered at four previously unrecognized centers: an unnamed patera to the south of Karei that produced a Pele-sized red ring, a patera to the west of Zal that produced a small circular bright deposit, a large orange ring detected near the north pole of Io, and a small bright ring near Io's south pole. Only a handful of Io's many active volcanoes produced large scale explosive eruptions, and several of these erupted repeatedly, leaving at least 83% of Io's surface unaltered throughout the Galileo mission. Most of the hot spots detected from SSI, NIMS and ground-based thermal observations caused no noticeable surface changes greater than 10 km in extent over the five year period. Surface changes were found at every location where active plumes were identified, including Acala which was never seen in sunlight and was only detected through auroral emissions during eclipse. Two types of plumes are distinguished on the basis of the size and color of their deposits, confirming post-Voyager suggestions by McEwen and Soderblom [Icarus 55 (1983) 191]. Smaller plumes produce near-circular rings typically 150-200 km in radius that are white or yellow in color unless contaminated with silicates, and frequently coat their surroundings with frosts of fine-grained SO₂. The larger plumes are much less numerous, limited to a half dozen examples, and produce oval, orange or red, sulfur-rich rings with maximum radii in the north-south direction that are typically in the range from 500 to 550 km. Both types of plumes can be either episodic or quasi-continuous over a five year period. Repeated eruptions of the smaller SO₂-rich plumes likely contribute significantly to Io's resurfacing rate, whereas dust ejection is likely dominated by the tenuous giant plumes. Both types of plume deposits fade on time-scales of months to years through burial and alteration. Episodic seepages of SO₂ at Haemus Montes, Zal Montes, Dorian Montes, and the plateau to the north of Pillan Patera may have been triggered by activity at nearby volcanic centers.     

Observations of Low-Latitude Coronal Plumes

Using Fe ix/x 17.1 nm observations from the Extreme-Ultraviolet Imaging Telescope (EIT) on the Solar and Heliospheric Observatory (SOHO), we have identified many coronal plumes inside low-latitude coronal holes as they transited the solar limb during the late declining phase of cycle 23. These diffuse, linear features appear to be completely analogous to the familiar polar plumes. By tracking them as they rotate from the limb onto the disk (or vice versa), we confirm that EUV plumes seen against the disk appear as faint, diffuse blobs of emission surrounding a brighter core. When the EIT images are compared with near-simultaneous magnetograms from the SOHO Michelson Doppler Imager (MDI), the low-latitude, on-disk plumes are found to overlie regions of mixed polarity, where small bipoles are in contact with unipolar flux concentrations inside the coronal hole. The birth and decay of the plumes are shown to be closely related to the emergence of ephemeral regions, their dispersal in the supergranular flow field, and the cancellation of the minority-polarity flux against the dominant-polarity network elements. In addition to the faint polar and nonpolar plumes associated with ephemeral regions, we note the existence of two topologically similar coronal structures: the giant plume-like features that occur above active regions inside coronal holes, and the even larger scale “pseudostreamers” that separate coronal holes of the same polarity. In all three cases, the basic structure consists of open field lines of a given polarity overlying a photospheric region of the opposite polarity; ongoing interchange reconnection at the X-point separating the open field domains from the underlying double-arcade system appears to result in the steady evaporation of material from the closed into the open region.     

Thermal and density structure of polar plumes

Normal incidence multilayer coated EUV/XUV optical systems provide a powerful technique for the study of the structure of the solar corona. Such systems permit the imaging of the full solar disk and corona with high angular resolution in narrow wavelength bands that are dominated by a single line or a line multiplet excited over a well defined range of temperatures. We have photometrically analysed, and derived temperature and density information from, images of polar plumes obtained with a multilayer Cassegrain telescope operating in the wavelength interval = 171 to 175 , which is dominated by FeIX and FeX emission. This observation was obtained in October 1987, and is the first high resolution observation of an astronomical object obtained with normal incidence multilayer optics techniques. We find that photometric data taken from this observation, applied to a simple, semi-empirical model of supersonic solar wind flow, are consistent with the idea that polar plumes are a source of the solar wind. However, we are not able to uniquely trace high speed streams to polar plumes. The temperatures that we observed are typically 1 500 000 K for both the plumes and the interplume regions, with the plume temperatures slightly higher than those of the surrounding atmosphere. Typical electron densities of the plume and interplume regions, respectively, are 5 × 10⁹ cm^–3 and 1 × 10⁸ cm^–3 at the limb of the Sun.     

Atmospheric modeling of Mars methane surface releases

We utilize the MarsWRF general circulation model (GCM) to address the behavior of gas plumes in the Martian atmosphere, with the specific goal of characterizing the source of the recently identified methane detection in the Martian atmosphere. These observations have been interpreted as the release of methane from localized surface sources with spatial and temporal variabilities. Due to the limited temporal coverage of ground-based observations, we use a GCM to simulate the development of passive atmospheric plumes over relevant timescales. The observations can be reproduced best if the release occurred just before the time of observation—no more than 1–2 sols earlier—and if this release were nearly instantaneous rather than a slow, steady emission. Furthermore, it requires a source region spanning a broad latitudinal range rather than a point emission. While the accuracy of our conclusions about this specific methane release scenario is limited by the uncertainties inherent in GCM simulations of the Martian atmosphere, our findings regarding generalized plume behavior are robust, and illustrate the potential power of numerical modeling for constraining plume source conditions.     

A model for the temperature-dependence of tidal dissipation in convective plumes on icy satellites: Implications for Europa and Enceladus

To explain the formation of surface features on Europa, Enceladus, and other satellites, many authors have postulated the spatial localization of tidal heating within convective plumes. However, the concept that enhanced tidal heating can occur within a convective plume has not been rigorously tested. Most models of this phenomenon adopt a tidal heating with a temperature-dependence derived for an incompressible, homogeneous (zero-dimensional) Maxwell material, but it is unclear whether this formulation is relevant to the heterogeneous situation of a warm plume surrounded by cold ice. To determine whether concentrated dissipation can occur in convective plumes, we develop a two-dimensional model to compute the volumetric dissipation rate for an idealized, vertically oriented, isolated convective plume obeying a Maxwellian viscoelastic compressible rheology. We apply the model to the Europa and Enceladus ice shells, and we investigate the consequences for partial melting and resurfacing processes on these bodies. We find that the tidal heating is strongly temperature dependent in a convective ice plume and could produce elevated temperatures and local partial melting in the ice shells of Europa and Enceladus. Our calculation provides the first quantitative verification of the hypothesis by Sotin et al. [Sotin, C., Head, J.W., Tobie, G., 2002. Geophys. Res. Lett. 29. 74-1] and others that the tidal dissipation rate is a strong function of temperature inside a convective plume. On Europa, such localized heating could help allow the formation of domes and chaos terrains by convection. On Enceladus, localized tidal heating in a thermal plume could explain the concentrated activity at the south pole and its associated heat transport of 2-7 GW.     

Magnetic flux separation in photospheric convection

Three-dimensional non-linear magnetoconvection in a strongly stratified compressible layer exhibits different patterns as the strength of the imposed magnetic field is reduced. There is a transition from a magnetically dominated regime, with small-scale convection in slender hexagonal cells, to a convectively dominated regime, with clusters of broad rising plumes that confine the magnetic flux to narrow lanes where fields are locally intense. Both patterns can coexist for intermediate field strengths, giving rise to flux separation: clumps of vigorously convecting plumes, from which magnetic flux has been excluded, are segregated from regions with strong fields and small-scale convection. A systematic numerical investigation of these different states shows that flux separation can occur over a significant parameter range and that there is also hysteresis. The results are related to the fine structure of magnetic fields in sunspots and in the quiet Sun.