Limitations of forcefree magnetic field extrapolations: revisiting basic assumptions
Key Words.:
Sun: corona — Sun: magnetic fields — Sun: atmosphere — Methods: data analysisAbstract
Context:Forcefree extrapolations are widely used to study the magnetic field in the solar corona based on surface measurements.
Aims:The extrapolations assume that the ratio of internal energy of the plasma to magnetic energy, the plasma is negligible. Despite the widespread use of this assumption observations, models, and theoretical considerations show that is of the order of a few percent to more than 10%, and thus not small. We investigate what consequences this has for the reliability of extrapolation results.
Methods:We use basic concepts starting with the force and the energy balance to infer relations between plasma and free magnetic energy, to study the direction of currents in the corona with respect to the magnetic field, and to estimate the errors in the free magnetic energy by neglecting effects of the plasma (). A comparison with a 3D MHD model supports our basic considerations.
Results:If plasma is of the order of the relative free energy (the ratio of the free magnetic energy to the total magnetic energy) then the pressure gradient can balance the Lorentz force. This is the case in the solar corona, and therefore the currents are not properly described. In particular the error in terms of magnetic energy by neglecting the plasma is of the order of the free magnetic energy, so that the latter can not be reliably determined by an extrapolation.
Conclusions:While a forcefree extrapolation might capture the magnetic structure and connectivity of the coronal magnetic field, the derived currents and free magnetic energy are not reliable. Thus quantitative results of extrapolations on the location and amount of heating in the corona (through current dissipation) and on the energy storage of the magnetic field (e.g. for eruptive events) are limited.
1 Introduction
Measuring the magnetic field in the outer atmosphere of the Sun is very difficult and not yet achieved on a routine basis. While ideas for advanced instrumentation exist (e.g. Peter et al., 2012a), so far we are restricted to coronagraphic observations above the limb (Lin et al., 2000, 2004) or radio data (e.g. White, 2005), both still limited in resolution and coverage of coronal features. However, magnetic field measurement at sufficient resolution would be a key to understand the heating and dynamics of the corona, because they would provide the crucial information on currents that heat the corona through Ohmic dissipation and on the free magnetic energy to be converted during eruptive events such as flares and coronal mass ejections that govern space weather.
For the chromosphere a few such measurements exist (e.g. Solanki et al., 2003; Lagg et al., 2004), but for the corona we mostly rely on extrapolations of the magnetic field from measurements mostly in the photosphere (e.g. Wiegelmann & Sakurai, 2012; Wiegelmann et al., 2014). Such extrapolations assume that the magnetic field fully dominates the plasma in the corona in terms of energy, i.e. that plasma comparing the internal (viz. thermal) energy of the plasma to the magnetic energy density is negligible, , see (1). With this assumption one can build algorithms to calculate the coronal magnetic field and deduce the currents and the free magnetic energy, even though different methods do not give consistent results (see Sect. 5.3).
However, observations, modelling, and theory do not really support that . Already in his textbook Priest (1982) writes “a magnetic field of 10 G and a density of m [is] characteristic of an active region” (his Sect. 6.4.2). Taking these numbers at face value, and employing the typical coronal temperature of 1 MK, one finds . While this magnetic field estimate is realistic (cf. Sect. 2), the density might be a factor of five to ten too high. Still for temperatures of active region loops of about 3 MK (Reale, 2014) one would then end up with a value of .
For the quiet Sun, Schrijver & van Ballegooijen (2005) provided evidence that in the upper atmosphere plasma can be even of order unity. While one might argue that these quiet Sun results are not directly transferable to active regions, our discussion in Sect. 2 will show that the quiet Sun is not just an exception. Based on published studies employing different techniques, we will conclude that is not necessarily small in the corona, but is of the order of several percent to more than 10%.
This finding, i.e. that plasma is nonnegliable, has severe consequences for forcefree extrapolations of the magnetic field. This is especially true if plasma is of the same order as the free energy of the magnetic field, which is the case as we will discuss in Sect. 3. Actually, it is not sufficient that only on average, but this has to be fulfilled throughout the whole volume under consideration. The main consequences for forcefree extrapolations of this are twofold. Firstly the currents do not have to be fieldaligned and are not constrained (Sect. 4) and secondly the free magnetic energy as derived from the extrapolation is not reliable (Sect. 5). Therefore one has to be careful not to overinterpret quantitative results on heating through current dissipation or on energy storage of the magnetic field based on forcefree extrapolations.
This paper is mainly relying on basic considerations concerning the force and energy balance. However, in the figures will compare these basic results to a 3D magnetohydrodynamics (MHD) model of the corona above an evolved active region by Bingert & Peter (2011) and find that the numerical simulation is supporting our basic arguments (for details of the model we refer to the original paper). This 3D MHD model includes an energy equation accounting for radiative losses and heat conduction. Therefore the pressure derived from that model can be trusted (cf. Peter, 2015), which is essential when calculating plasma. This trust is further supported by this type of models reproducing e.g. the width of coronal loops (Peter & Bingert, 2012), the 3D structure of loops (Bourdin et al., 2013), or the patterns of loops in an emerging active region (Chen et al., 2014).
2 Constraints on plasma in the corona from observations, models, and theory
The plasma is a dimensionless number comparing the gas pressure and the energy density of the magnetic field ,
(1) 
with the magnetic permeability . Because the pressure essentially is the internal energy density (apart of a factor of 3/2 for a monoatomic ideal gas), compares the internal energy density to the magnetic energy density in a plasma.
The magnetic field in the corona is difficult to observe directly and consequently is not well constrained by observations. However, the technique of coronal seismology allows to get access at least to the (average) magnetic field along an oscillating coronal loop (Edwin & Roberts, 1983). The applicability of this method to the real Sun and what this average value actually means have been confirmed e.g. by De Moortel & Pascoe (2009) and Chen & Peter (2015). To get the pressure, in studies of coronal seismology the density and temperature are derived using extreme UV spectroscopy or imaging (or in the worst case they are just assumed to be at the canonical coronal values for the density of cm and the temperature of MK). Typically, in coronal seismology long loops (100 Mm to 200 Mm long) are studied, and magnetic field strengths of the order of some 5 G to 20 G are found, which is consistent with direct coronagraphic measurements using the Zeeman effect (Lin et al., 2000, 2004).
Here we just highlight three studies, basically reflecting the findings in coronal seismology (for a review see Nakariakov & Verwichte, 2005). Guo et al. (2015) find in a cool (0.65 MK) and dilute ( cm) loop system a field strength of about 8 G, corresponding to . In a hotter (1.05 MK) and denser ( cm) loop Nakariakov & Ofman (2001) find a magnetic field of about 13 G, corresponding to . For a larger sample of a dozen loops, White & Verwichte (2012) find a range of magnetic field strength, from about 3 G to 20 G. While they did not employ a density analysis, when using a typical coronal density of cm, this corresponds to values of in the range of 0.4 to 0.01. So based on these observations of coronal seismology in general plasma will range from a few percent to more than 10 %, which is not really small.
One of the major references when it comes to giving support to the assumption that should be small is the group of models by Gary (2001). Even recent reviews (e.g. Wiegelmann & Sakurai, 2012) use this as the sole justification that would be small. Most quoted is Fig. 3 of Gary (2001) that shows as a function of height for two models as extreme cases, one for a fieldline coming out of the middle of the umbra and one from a plage region. At a height of some 40 Mm, i.e. at the apex of a typical coronal loop of 120 Mm length, is 0.004 and 0.1 for the umbra and plage case, respectively. However, fieldlines from the centre of the umbra will not be associated with coronal loops, and thus the line Gary (2001) shows for the plage region might be more typical for coronal loops. Therefore, even the study by Gary (2001), usually used to argue for a very small in the corona, actually does indicate that is not so small.
A good estimate for can also be inferred from 3D MHD models. If these models include a proper description of the energy equation, in particular accounting for heat conduction and radiative losses, they will give a realistic estimate for the coronal pressure and can provide a good match to the observed real Sun (Peter, 2015). These models show a wide range of plasma in the corona, with a median value of around temperatures of about 1 MK (Peter et al., 2006, Fig. 12). In places of emerging active regions where the magnetic field is comparably strong, the 3D MHD models show small values of (based on data from Chen et al., 2015), but in general is larger. In particular, Peter et al. (2006) showed that in bright regions in the corona above an active region is not small, but can reach values even above unity! Small but nonnegligible value of are also found in the coronal part ( from 1 MK to 1.5 MK) of the 3D MHD model of of an evolved active region by Bingert & Peter (2011). This and the considerable scatter of values in the corona is illustrated in Fig. 1.
Besides the above approaches through observations and modelling, one can also estimate plasma based on basic theoretical considerations by combining results from heating processes and scaling laws. In their model to heat the upper atmosphere through Alfvén waves, van Ballegooijen et al. (2011) give a parameterisation for the volumetric heat input, , depending on the magnetic field in the loop, , and loop length . Their Eq. (63) reads . At the same time, following Rosner et al. (1978) one can find scaling relations determining the temperature and density based on and . In the form of Peter et al. (2012b) these read and . Combining these and using the coefficients quoted in the original papers one finds
(2) 
The above mentioned values of magnetic field in coronal loops as derived from coronal seismology or coronagraphic observations using the Zeeman effect range from 5 G to 20 G. Therefore with (2) this consideration supports the finding that is not small also from a theoretical perspective.
The conclusion from this overview of active region observations and models is that we can expect plasma to be somewhere in the range from a few percent to 10%. So in contrast to commonly used statements that the corona is a lowbeta plasma, we see that is mostly smaller than unity, but not negligibly small. In particular the 3D MHD models show that one has to be careful with a general statement, because depending on the structure, might range from very small values below 0.01, to values significantly above 1, even in bright patches of the corona. Such large values for plasma have been discussed before already in the context of the quiet Sun (Schrijver & van Ballegooijen, 2005), so it should not be too surprising, that also active region show nonnegligible values for .
3 Free energy in forcefree extrapolations
A potential magnetic field satisfies and and is a state of minimum energy under the constraint that the normal component of the field at the photospheric boundary is prescribed. If the magnetic field is forced away from this state, the magnetic energy available for conversion into kinetic and internal energy (i.e. acceleration and heating) is the free magnetic energy, . This is the difference between the actual magnetic energy, and the magnetic energy of the potential field (with the same boundary conditions for the normal component of ), . Here the energies are integrated over a volume, e.g. a full active region under investigation.
Because the spatial distribution of the magnetic field cannot be measured directly in the corona (with some exceptions), mainly (nonlinear) forcefree extrapolations are employed to estimate the magnetic energy, (see Sect. 4). In the following we give a by far nonexhaustive list of examples for such studies that provide information on the free energy. For these we give values for the relative free energy in percent, where
(3) 
Here the index “free” refers to free energy and not to forcefree extrapolation. This quantity is positive because for a given normal component of the magnetic field at the boundary the potential field is the solution with the minimum magnetic energy.
Forcefree extrapolations evaluate this free energy in a given volume, and of course depending on the volume one would find different values. Representing the huge variety of forcefree models (see e.g. review by Wiegelmann & Sakurai, 2012) we will quote the results from extrapolations of three different structures: a normal (nonflaring) active region, a flaring active region, and a small highly twisted part of an active region. The maximum magnetic energy in a volume for a specified magnetic field at its lower boundary is given by the energy of an open field configuration. Its maximum free magnetic energy is about 100%, and this maximum also applies to forcefree extrapolations (e.g. Amari et al., 2000). While exceptions of large free magnetic energies exist (e.g. up to 60% in a flaring active region; Thalmann & Wiegelmann, 2008), typically forcefree extrapolations all show relative free energies of some 10%.
As a sample for a normal active region we quote Thalmann et al. (2012) who find a free magnetic energy of about 7% to 14%. The domain they investigated covers about 18098 Mm in the horizontal directions, basically encompassing the whole active region. One might expect a higher level of free energy for a flaring active region just before a flare. While Sun et al. (2012) find a free energy of almost 50% two days before a Xclass flare, just a few hours before the flare the free energy was at only 20% (the authors did not specify the volume but it probably encompasses the whole active region). Thus even in a flaring active region the level of free energy is modest. This is also true for a twisted subvolume in an active region. Thalmann et al. (2014) investigated the region of strong braiding identified in the images of the Highresolution Coronal Imager (HiC) rocket flight (Cirtain et al., 2013). However, even in a small volume of 1374 Mm they found the free energy of the order of only 5%. These low values for the free energy are not simply an artefact of one particular method. Comparing major codes for forcefree extrapolations De Rosa et al. (2009) found that they all give small values in the range of 5% to 25% (full active region, 9090 Mm in the horizontal directions).
These values for the free magnetic energy in forcefree extrapolations are also found in 3D MHD simulations. In the model of an evolved active region by Bingert & Peter (2011) that we use as an example throughout this paper we find a free magnetic energy of about 10% when integrating over the whole computational domain covering the active region (here the free energy is dominated by the lower part of the computational box).
The conclusion from this overview of extrapolation models is that the (relative) free magnetic energy in the corona ranges mostly from 5% to 10% in an active region. This might be slightly higher in flaring regions, but probably not (much) higher than 20%. These values should be considered as upper limits, because they depend on the volume selected for integrating the energies. When extending the volume to include, e.g. neighbouring nonactive parts, where the free magnetic energy will be lower, the (relative) free energy of the whole volume will go down.
4 Local force balance and currents
We will first check what role the plasma plays in the force balance and what implications this has for the currents derived from a forcefree extrapolation. This is a local treatment in the sense that we investigate the forces at a given spot somewhere in the corona, e.g. at the location of a bright coronal loop.
The ideal momentum balance in a static case reads
(4) 
Here is the pressure, the density, the gravitational acceleration, the magnetic field, and . We separated the Lorentz force into the magnetic pressure gradient and the magnetic tension, i.e. terms (c) and (d).
The plasma introduced in (1) compares the gas and the magnetic pressure, i.e. basically terms (a) and (c,d) in (4): In general in a low environment the magnetic force will dominate the pressure gradient.^{1}^{1}1However, if and , and are constant in space, and and even for it might well be that the pressure gradient equals the Lorentz force. This is because in this case the forces, i.e. the gradients, depend only on and , while mainly compares and . In the stationary case in most situations a hydrostatic equilibrium will govern the variation along magnetic field lines (at least for structures with constant cross section), so the terms (a) and (b) in (4) will be of the same order, implying that the gravitational term (b) would also be small if is small.
Therefore in a low regime the gradient of the magnetic pressure (c) has to be balanced by the magnetic tension force (d) alone. This is also true when flows are present as long as these are subsonic, because then the kinetic energy is still even smaller than the internal energy. Consequently we have
(5) 
This condition applies only approximately, i.e. to the order of , where we still have to define what this means (see Sect. 4.1). Because this implies that all forces are small, solutions to this state are called forcefree.
To satisfy the condition (5) two (not mutually exclusive) options exist,
(6) 
Seeking forcefree solutions of (5), research focuses on option (6a), as e.g. outlined in the introduction of the review by Wiegelmann & Sakurai (2012). For such forcefree extrapolations the possibility (6b) of small currents not aligned with the magnetic field is largely ignored.
However, because the condition (5), , has to be fulfilled only approximately to the extent that the gas can balance the magnetic forces to the order of , one can have either large currents at small angles (between and ), or one could have small currents at arbitrary angles. The latter solution is not the same as a potential field where and the Lorentz force vanishes. Instead we have seen in Sect. 2 that plasma is of the order of up to 10%, and can in some places of the bright corona even reach values above 1! (See Sect. 2 and Peter et al., 2006). Thus there can be nonnegligable Lorentz forces in the corona.
The main implication following from the two options in (6) is that they can lead to significantly different results for the currents in the system, and thus for the expected heating rates derived from an extrapolation. Likewise, if the currents are quite different between the two options, also the free magnetic energy for these two options will be quite different. Therefore selecting option (6a), , makes a choice that might provide answers in terms of currents and free magnetic energy that are far from being a unique answer.
4.1 What does “small currents” mean?
We now investigate to what extent the plasma can balance the Lorentz forces due to small currents that are not parallel to the magnetic field. This is the same as to ask when the option (6b) will apply, i.e. when the currents are small.
For this we decompose the magnetic field into a part that is a potential field (for the given boundary conditions) and a second part that we call the free magnetic field, ,
(7) 
Similar to (3), here the index “free” refers to the magnetic field corresponding to the free magnetic energy, and not to the forcefree extrapolation.
The free magnetic field corresponds to the current system that is not necessarily parallel to the magnetic field. By definition we have , and the currents are given by . Since here we consider case (6b), we assume that these currents are not (predominantly) aligned with the magnetic field.
Considering only the pressure gradient and the Lorentz force (terms a,c,d) we can write the force balance (4) as
(8) 
Dividing this by the magnetic energy density and because here we investigate the small currents not aligned with the magnetic field, we find the order of magnitude estimation
(9) 
As long as is smaller than , the free magnetic energy is . For this local analysis of the force balance we will consider the relative free energy as defined in (3) locally, i.e. we evaluate it in a small volume around the location where and are taken. For an orderofmagnitude estimation we only consider the absolute value, because locally can be negative (see also Sect. 4.2). With this it follows from (3) that . Thus from (9) we find the relation between the relative free energy and plasma in order to satisfy option (6b),
(10) 
This relation, based on a local force balance, is an orderofmagnitude estimation only. The comparison with a 3D MHD model in Sect. 4.2 shows that there is larger than , typically by a factor of about five.
From the relation in (10) it follows that as long as the relative free energy (3) is of the order of plasma (or smaller), the currents not parallel to the magnetic field are small enough, so that the plasma can balance the resulting Lorentz force. In this case, even the parallel currents are not constrained, and a forcefree extrapolation would not capture the component of the (small) currents that is parallel to the magnetic field (see Appendix A).
Only if the relative free energy would be significantly larger than plasma, the corresponding currents must be dominated by those corresponding to the nonlinear forcefree modelling. Then option (6a) would apply and the currents would have to be (roughly) parallel to the magnetic field.
However, from the discussion in Sects. 2 and 3 and the MHD model in Fig. 2 it is clear, that the relative free magnetic energy is of the order of plasma, so that the condition (10) is satisfied. From this we conclude, that (in most cases) the free energy is consistent with currents not parallel to the magnetic field. Therefore the currents are not necessarily parallel to , and consequently the currents derived from (most) forcefree extrapolations are only reliable in stating that the field is close to potential. Therefore also estimations of e.g. the localisation of plasma heating based on the currents from forcefree extrapolations are not reliable.
4.2 Free energy and plasmabeta in a 3D MHD simulation
To illustrate the relation between the (local) relative free energy and plasma as derived in (10) we show a 2D histogram of this based on the 3D MHD model of Bingert & Peter (2011) in Fig. 2. Just as with (10), here is evaluated in one grid cell of the simulation when compared to at that same grid cell. Like for the orderofmagnitude estimation above, we plot the absolute values of . The respective histograms for the positive and negative values alone look basically identical to the one for absolute values displayed in Fig. 2.
The histogram in Fig. 2 shows that typically is several (typically about five) times larger than (most entries in the histogram are to the right of the green line indicating ). Furthermore, the volume with positive values of is about a factor of 1.5 larger than the one with negative values. Consequently, when considering the whole volume is positive, as it should be. As mentioned already in Sect. 3, for this model we find when considering the whole domain.
In general the 3D MHD model supports the orderofmagnitude estimation in (10), even though it clearly shows that mostly is larger than .
4.3 Currents in a 3D MHD simulation
The relation of the currents to the magnetic field, i.e. the two options (6a) and (6b) can be directly investigated in a 3D MHD model. In such a model also the plasma is considered, i.e. the model fully accounts for the fact that is not small in general. In the 3D MHD model by Bingert & Peter (2011) we employ here, the free magnetic energy integrated over the whole volume is about 10%. This is comparable to the values found for typical solar active regions based on forcefree extrapolations (cf. Sect. 3) and thus this active region models should be a good representation of the real Sun in terms of the available free magnetic energy and currents.
In Fig. 3 we show the distribution of the angle between the current and the magnetic field in that 3D MHD model. While in general the angles are small, the mean angle in the coronal part of the computational domain around MK is about 15, and 25% of the coronal volume (from MK to 1.5 MK) has angles above 30. In the corona we even find angles of almost 90, i.e. currents that are almost perpendicular to the magnetic field.
This comparison to a 3D MHD models confirms that the currents in the corona are not necessarily parallel to the magnetic field as it would be required for option (6a) when assuming that is very small. Instead, when accounting for the effects of the plasma, is not necessarily small (cf. Sect. 2, Fig. 1) and the currents will not be parallel to the magnetic field, inline with option (6b).
5 Global energy considerations and accuracy
In contrast to the local treatment discussed in Sect. 4, the following considerations apply to the energy contained in a volume. In this sense this is nonlocal, or global. As outlined in Sect. 3 the volume considered by forcefree extrapolations typically ranges from a several Mm to the better part of an active region.
Based on the virial theorem (e.g. Priest, 1982, Sect 2.8.4) in a stationary state the energy balance following from the momentum balance requires that the total energy within a volume, , is solely determined by an integral of properties at its surface, ,
(11) 
As before an uppercase refers to the volume integral of the respective energy density. The exact form of is not of interest here, but it can be found e.g. in Eq. 2.66 in Priest (1982).
For the following we will consider only the magnetic and the internal (viz. thermal) energy,
(12) 
Other terms such as the kinetic or gravitational energy, or , will only strengthen the arguments given below, because they will increase the nonmagnetic contributions to the total energy. In general, we will have because the flows are mostly subsonic. In most situations will be of the order of reflecting hydrostatic equilibrium (along fieldlines).
We rewrite (11) to give
(13) 
and introduce
(14) 
Because the internal energy basically is the pressure, this quantity will be of the order of the (average) plasma (1),
(15) 
With (14) we can rewrite (13) to give
(16) 
For a (nonlinear) forcefree extrapolation, one assumes that the magnetic field has to balance itself, so other forces (and energies) are negligible and thus . Consequently, the minimum magnetic energy of the forcefree extrapolation, , is given by the surface term,
(17) 
Comparing with (16) and using (15) we find the relative error of the forcefree extrapolation (in terms of energy) with respect to the true magnetic field to be
(18) 
This implies that we can determine the magnetic energy through a forcefree extrapolation only to the order of plasma. Because forcefree extrapolations neglect the existence of the plasma, it is clear that this should be a limiting factor for the accuracy of the magnetic energy. The above estimate is a conservative estimate, because also the gravitational and kinetic energy will contribute to this uncertainty, so that the true error could easily be also a factor of 2 to 3 larger than estimated in (18).
If , which is equivalent to fully ignore the effects of the plasma in the corona, the magnetic field will be in a truly forcefree state. Then the accuracy is limited only by the accuracy of the algorithm used for the forcefree extrapolation (Sect. 5.3). However, on the real Sun plasma is not arbitrarily small. Instead, observations, theory, and numerical simulations show values for in the corona ranging from a few to 10% (see Sect. 2). Therefore we can expect the magnetic field deduced from extrapolations to be accurate only within some 10% in terms of magnetic energy.
In forcefree extrapolations the free magnetic energy is only of the order of 10% of the magnetic energy (see Sect. 3). Considering that the forcefree extrapolation can be accurate only to some 10%, too, this implies that the derived free magnetic energy is not well constrained. Basically this says that the results from a forcefree extrapolation are consistent with a potential field (in terms of energy). Thus one should not use the free energy derived from a forcefree extrapolation for a quantitative analysis.
5.1 When can we trust forcefree extrapolations?
We now turn to a quantitative measure of when a forcefree extrapolation would give results in terms of free magnetic energy that can be trusted. For this we define a free, , that in analogy to plasma compares the pressure (or internal energy) to the free energy of the magnetic field,
(19) 
With the relative free energy as defined in (3), free reads
(20) 
Finally with (18) we find the ratio of the error of the forcefree extrapolation (by neglecting the impact of the plasma) to the relative free energy to be
(21) 
The conclusion is that as long as the free as defined in (19) is of unity or even larger, , then the error of the forcefree extrapolation (because of the presence of the plasma) is at least as big as the difference of the extrapolated field from a potential field. In other words: within the errors the extrapolated field is the same as a potential field (in terms of energy).
If , then also the Lorentz force is not large compared to the other terms in the momentum equation, but will be balanced by the pressure gradient. The condition for this was in (10) that . Then it follows from (20) that proving the above statement.
Only if is significantly below unity, we can trust the magnetic energy (and thus the free energy) derived from a forcefree extrapolation. Similarly, only if is significantly below unity we can trust the currents derived from a forcefree extrapolation and thus any conclusions drawn on the distribution of (Ohmic) heating in the corona based on this (cf. Appendix A).
In general, plasma ranges from several percent to 10% (see Sect. 2). The relative free energy is the order of 10% for normal active regions and might reach about 50% in flaring active regions (see Sect. 3). With this, based on (20) the free is not significantly below unity, but typically will be of the order of unity in normal active regions and maybe in flaring active regions. Therefore the trust in the free magnetic energy derived from forcefree extrapolations should be limited.
5.2 Comparison to 3D MHD models
With our estimations in Sects. 2 and 3 that plasma and the relative free energy are of the same order, we would expect the free is of order unity. While those are only estimations, one can use a 3D MHD accounting for the magnetic field and the plasma for a closer check.
Again, we use the results from the 3D simulation of Bingert & Peter (2011) to calculate free. For this we perform a potential field extrapolation using the vertical magnetic field from the 3D model at a height of 5 Mm which is (roughly) the base of the corona in that model. Together with the actual magnetic field in the MHD model we then calculate the free energy according to (3) in the volume of each grid cell of the computation. Finally we use the plasma from the model and evaluate the free at each grid cell according to (20).
In Fig. 4 we show the resulting distribution of in the 3D MHD model. This demonstrates that in the coronal part of the computational domain, where the temperatures reach values of the order of 1 MK to 1.5 MK, free has values of the order of . Thus the numerical model confirms the estimation for at the end of Sect. 5.1 for a nonflaring active region. Consequently, according to (21) the the free magnetic energy is of the order of the error in magnetic energy if neglecting the effcets of the plasma. This supports our conclusion from Sect. 5.1 that one cannot fully trust forcefree extrapolations in terms of free magnetic energy.
5.3 Accuracy of forcefree extrapolations
The above discussion does not consider the accuracy of the method to perform the forcefree extrapolation. Few analytical solutions of highly symmetric and comparably simple magnetic configurations exist (e.g., Low & Lou, 1990; Titov & Démoulin, 1999). If compared to these, forcefree extrapolations can recover the true magnetic energy from the analytical solution only to some 10% if only the lower boundary, i.e., the surface magnetic field is specified (based on six algorithms investigated by Schrijver et al., 2006). In particular, different methods “will not necessarily converge to the correct, or even the same, solution” (Metcalf et al., 2008). A comparison of a dozen methods of extrapolation using the same data set of an observed active region magnetic field provides free magnetic energies in the range of 3% to 25% (De Rosa et al., 2009). This does not reflect a real error because the true magnetic field in the corona is not know in these cases. Still, it shows that the scatter between the different methods is considerable and that the error solely introduced by the extrapolation might be at least of the order of 20%.
It would interesting to directly compare forcefree extrapolations to 3D MHD models. For this one could use the magnetic field at the base of the corona in the MHD model to perform the extrapolation. In particular when using different MHD models, e.g. for an evolved (Bingert & Peter, 2011) or an emerging active region (Chen et al., 2014), the comparison to the magnetic field and the free energy from the extrapolation could provide a further test to the accuracy and reliability of forcefree extrapolations. Such a study however, is far beyond the scope of the present investigation.
The arguments in this paper address only the principle difficulty of forcefree extrapolations due to the fact that plasma is not vanishing in the corona. The errors introduced by the extrapolation algorithms will make things worse.
6 Possible progress
To overcome the limitations of forcefree extrapolations, some progress might be achieved by combining the measured surface magnetic field and the observed loop orientation in the minimisation process (Aschwanden et al., 2012; Malanushenko et al., 2012, 2014). This would result in solutions where the magnetic energy is not minimised (for the given vector magnetic field at the boundary), but where the magnetic field might be closer to the true field on the Sun.
A direct step forward to account for the effects of the gas pressure would be to perform magnetohydrostatic (MHS) models (e.g. Ruan et al., 2008). However such models have the serious shortcoming that they cannot account for the history and evolution of the corona. This memory of the magnetic field is accounted for by magnetofrictional models (e.g. Cheung & DeRosa, 2012), but they neglect the presence of plasma and essentially face problems similar to the forcefree extrapolations discussed here.
To account for the presence of plasma and the history of the magnetic evolution one has to employ 3D MHD models. Driven by observational data, i.e., the magnetic field at the solar surface as a function of time, such models can represent observed coronal structures, as demonstrated by Bourdin et al. (2013). While that model used only lineofsight magnetograms, the models by Chen et al. (2014, 2015) have been driven by the full vector of the magnetic field at the surface, as well as by the velocity field. Thus full 3D MHD models driven by the observed vector magnetic field are within reach. At the moment such datadriven 3D MHD models, including global coronal models (e.g. Riley et al., 2011), do not produce fully realistically looking coronae. One reason for this is that timestep limitations restrict the solar time covered by the models (Peter, 2015). With further improvements of methods and computational resources, such models might provide the progress needed for a more reliable determination of currents and free energy in the solar corona.
7 Conclusions
We investigated the limitations of forcefree extrapolations as a consequence of neglecting the impact of the plasma by assuming plasma to be negligibly small. We find that observations, models, and theory do not really support this assumption, and that is of the order of a few percent to 10% or even more.
In our theoretical investigation we define a free, , in (19), which is the ratio of the plasma to the (relative) free magnetic energy . If , i.e. if the free energy is not substantially larger than plasma, we draw the following conclusions:

The error of the forcefree extrapolation in terms of magnetic energy is of the same order as the difference of the true magnetic field to a potential field. This is to say that a potential extrapolation is as accurate as a forcefree extrapolation.

The free magnetic energy derived from the extrapolated field will be within the errors of the forcefree extrapolation. Therefore the free magnetic energy derived from a forcefree extrapolations cannot be trusted.

The currents are not constrained at all in a forcefree extrapolation, because they do not have to be parallel to the magnetic field.
On the Sun and for extrapolations of (nonflaring) active regions in general we provided evidence that free is of order unity, . Even in a flaring active region will not be significantly below unity. Thus the above conclusions apply to forcefree extrapolations of solar active regions.
Still, one might extract important information on the magnetic connectivity and topology from a forcefree extrapolation. However, the quantitative results on the amount of the free magnetic energy available and on the currents to locate the regions of energy deposition cannot really be trusted. Approaches going beyond current forcefree extrapolations are needed (cf. Sect. 6) to possibly overcome these limitations.
Acknowledgements.
We would like to thank Sven Bingert and Feng Chen for help in providing data for the comparison to 3D MHD models. Sincere thanks are due to Bernd Inhester and Thomas Wiegelmann for discussions on the extrapolations. We kindly acknowledge comments by Karel Schrijver, in particular his suggestion to give an outlook on possible progress. We also thank the referee for constructive comments helping to improve the paper. J.W. acknowledges funding from the People Programme (Marie Curie Actions) of the European Union’s Seventh Framework Programme (FP7/20072013) under REA grant agreement No. 623609. L.P.C. acknowledges funding by the MaxPlanck Princeton Center for Plasma Physics. RHC’s contribution was carried out in the context of Deutsche Forschungsgemeinschaft SFB 963 “Astrophysical Flow Instabilities and Turbulence” (Project A16).References
 Amari et al. (2000) Amari, T., Luciani, J. F., Mikic, Z., & Linker, J. 2000, ApJ, 529, L49
 Aschwanden et al. (2012) Aschwanden, M. J., Wuelser, J.P., Nitta, N. V., et al. 2012, ApJ, 756, 124
 Bingert & Peter (2011) Bingert, S. & Peter, H. 2011, A&A, 530, A112
 Bourdin et al. (2013) Bourdin, P.A., Bingert, S., & Peter, H. 2013, A&A, 555, A123
 Chen & Peter (2015) Chen, F. & Peter, H. 2015, A&A, 581, A137
 Chen et al. (2014) Chen, F., Peter, H., Bingert, S., & Cheung, M. C. M. 2014, A&A, 564, A12
 Chen et al. (2015) Chen, F., Peter, H., Bingert, S., & Cheung, M. C. M. 2015, Nature Physics, 11, 492
 Cheung & DeRosa (2012) Cheung, M. C. M. & DeRosa, M. L. 2012, ApJ, 757, 147
 Cirtain et al. (2013) Cirtain, J. W., Golub, L., Winebarger, A. L., et al. 2013, Nature, 493, 501
 De Moortel & Pascoe (2009) De Moortel, I. & Pascoe, D. J. 2009, ApJ, 699, L72
 De Rosa et al. (2009) De Rosa, M. L., Schrijver, C. J., Barnes, G., et al. 2009, ApJ, 696, 1780
 Edwin & Roberts (1983) Edwin, P. M. & Roberts, B. 1983, Sol. Phys., 88, 179
 Gary (2001) Gary, G. A. 2001, Sol. Phys., 203, 71
 Guo et al. (2015) Guo, Y., Erdélyi, R., Srivastava, A. K., et al. 2015, ApJ, 799, 151
 Lagg et al. (2004) Lagg, A., Woch, J., Krupp, N., & Solanki, S. K. 2004, A&A, 414, 1109
 Lin et al. (2004) Lin, H., Kuhn, J. R., & Coulter, R. 2004, ApJ, 613, L177
 Lin et al. (2000) Lin, H., Penn, M. J., & Tomczyk, S. 2000, ApJ, 541, L83
 Low & Lou (1990) Low, B. C. & Lou, Y. Q. 1990, ApJ, 352, 343
 Malanushenko et al. (2014) Malanushenko, A., Schrijver, C. J., DeRosa, M. L., & Wheatland, M. S. 2014, ApJ, 783, 102
 Malanushenko et al. (2012) Malanushenko, A., Schrijver, C. J., DeRosa, M. L., Wheatland, M. S., & Gilchrist, S. A. 2012, ApJ, 756, 153
 Metcalf et al. (2008) Metcalf, T. R., De Rosa, M. L., Schrijver, C. J., et al. 2008, Sol. Phys., 247, 269
 Nakariakov & Ofman (2001) Nakariakov, V. M. & Ofman, L. 2001, A&A, 372, L53
 Nakariakov & Verwichte (2005) Nakariakov, V. M. & Verwichte, E. 2005, Living Reviews in Solar Physics, 2, 3
 Peter (2015) Peter, H. 2015, Phil. Trans. R. Soc. A, 373, 20150055
 Peter et al. (2012a) Peter, H., Abbo, L., Andretta, V., et al. 2012a, Experimental Astronomy, 33, 271
 Peter & Bingert (2012) Peter, H. & Bingert, S. 2012, A&A, 548, A1
 Peter et al. (2012b) Peter, H., Bingert, S., & Kamio, S. 2012b, A&A, 537, A152
 Peter et al. (2006) Peter, H., Gudiksen, B. V., & Nordlund, Å. 2006, ApJ, 638, 1086
 Priest (1982) Priest, E. R. 1982, Solar Magnetohydrodynamics (D. Reidel, Dordrecht)
 Reale (2014) Reale, F. 2014, Living Reviews in Solar Physics, 11, 4
 Riley et al. (2011) Riley, P., Lionello, R., Linker, J. A., et al. 2011, Sol. Phys., 274, 361
 Rosner et al. (1978) Rosner, R., Tucker, W. H., & Vaiana, G. S. 1978, ApJ, 220, 643
 Ruan et al. (2008) Ruan, P., Wiegelmann, T., Inhester, B., et al. 2008, A&A, 481, 827
 Schrijver et al. (2006) Schrijver, C. J., Derosa, M. L., Metcalf, T. R., et al. 2006, Sol. Phys., 235, 161
 Schrijver & van Ballegooijen (2005) Schrijver, C. J. & van Ballegooijen, A. A. 2005, ApJ, 630, 552
 Solanki et al. (2003) Solanki, S. K., Lagg, A., Woch, J., et al. 2003, Nature, 425, 692
 Sun et al. (2012) Sun, X., Hoeksema, J. T., Liu, Y., et al. 2012, ApJ, 748, 77
 Thalmann et al. (2012) Thalmann, J. K., Pietarila, A., Sun, X., & Wiegelmann, T. 2012, AJ, 144, 33
 Thalmann et al. (2014) Thalmann, J. K., Tiwari, S. K., & Wiegelmann, T. 2014, ApJ, 780, 102
 Thalmann & Wiegelmann (2008) Thalmann, J. K. & Wiegelmann, T. 2008, A&A, 484, 495
 Titov & Démoulin (1999) Titov, V. S. & Démoulin, P. 1999, A&A, 351, 707
 van Ballegooijen et al. (2011) van Ballegooijen, A. A., AsgariTarghi, M., Cranmer, S. R., & DeLuca, E. E. 2011, ApJ, 736, 3
 White & Verwichte (2012) White, R. S. & Verwichte, E. 2012, A&A, 537, A49
 White (2005) White, S. M. 2005, in Chromospheric and Coronal Magnetic Fields, Vol. ESA SP–596, 10
 Wiegelmann & Sakurai (2012) Wiegelmann, T. & Sakurai, T. 2012, Living Reviews in Solar Physics, 9, 5
 Wiegelmann et al. (2014) Wiegelmann, T., Thalmann, J. K., & Solanki, S. K. 2014, A&A Rev., 22, 78
Appendix A Reliability of fieldparallel currents in a forcefree extrapolation
The following discussion will show that if plasma (1) is of the same order as the free energy of the magnetic field (3), then the currents parallel to the magnetic field in general need not be dominated by those of a forcefree extrapolation alone. Following (20) the above condition is equivalent to the free being of order unity, .
In the discussion in Sect. 4 we show that the Lorentz force due to (small) currents that are perpendicular to the magnetic field can be balanced by the plasma as long as is of the order of the free energy (10). One can decompose the currents associated with the true magnetic field into a part associated with the magnetic field from the forcefree extrapolation that is parallel to and another part that is associated with the Lorentz force being balanced by the plasma and that is not necessarily aligned with . These two parts, and would correspond to the two options in (6). The part balanced by the plasma can be further separated into a component perpendicular and a component parallel to . Thus one can write the currents associated with the true magnetic field as
(22) 
The currents of the true magnetic field, , will be divergencefree, (because the divergence of a curl is zero). Likewise, the currents of the forcefree field alone will satisfy , and consequently also . The component is determined through the pressure gradient balancing the Lorentz force , and in general .
Assuming that the only parallel currents are those of the forcefree extrapolation (i.e. ) would imply , which cannot be met in general. Therefore the assumption will not apply — in general . This argument also suggests that the parallel and perpendicular currents associated with the Lorentz force balanced by the plasma are likely to have similar magnitudes (because ).
The current introduced by the nonzero plasma therefore has components both perpendicular and parallel to the magnetic field, neither of which are captured by a forcefree extrapolation. In particular this rises doubts for quantitative studies where one uses the currents derived from an extrapolation to investigate the location and magnitude of the heating in the corona.