On the probability distribution function of small scale interplanetary magnetic field fluctuations

In spite of a large number of papers dedicated to study MHD turbulence in the solar wind there are still some simple questions which have never been sufficiently addressed like: a)do we really know how the magnetic field vector orientation fluctuates in space? b) what is the statistics followed by the orientation of the vector itself? c) does the statistics change as the wind expands into the interplanetary space? A better understanding of these points can help us to better characterize the nature of interplanetary fluctuations and can provide useful hints to investigators who try to numerically simulate MHD turbulence. This work follows a recent paper presented by the same authors. This work follows a recent paper presented by some of the authors which shows that these fluctuations might resemble a sort of random walk governed by a Truncated Leevy Flight statistics. However, the limited statistics used in that paper did not allow final conclusions but only speculative hypotheses. In this work we aim to address the same problem using a more robust statistics which on one hand forces us not to consider velocity fluctuations but, on the other hand allows us to establish the nature of the governing statistics of magnetic fluctuations with more confidence. In addition, we show how features similar to those found in the present statistical analysis for the fast speed streams of solar wind, are qualitatively recovered in numerical simulations of the parametric instability. This might offer an alternative viewpoint for interpreting the questions raised above.


I. INTRODUCTION
The first turbulent model proposed by [23] didn't take into account that the rate of energy transfer along the turbulent cascade might not be scale-independent (Landau's objection).
This situation, in the framework of a classical Richardson's cascade, can be described by the fact that smaller and smaller eddies are less and less space-filling. In other words, turbulence would be unevenly or intermittently distributed in space. As a matter of fact, we address this phenomenon as Intermittency. Evidence of the presence of this phenomenon, when performing a statistical study on a generic fluctuating field V (x), is that the probability distribution functions (PDF hereafter) of the differences δv l (x) = |V (x+l)−V (x)| normalized to the σ of the distribution at scale l do not rescale for different scales [52]. In particular, the tails of these PDFs become more and more stretched at smaller and smaller scales. This means that the wings of the distributions become fatter and fatter. Such a behaviour implies that, at smaller scales, extreme events become statistically more probable than if they were normally distributed.
Intermittency has been found also in the solar wind fluctuations as demonstrated in the first studies performed by [13] in the outer heliosphere. In particular, this author showed an unexpected similarity between interplanetary observations on scales of 1AU and observations on scales of meters obtained for laboratory turbulence by [1]. These results suggested the universality of this phenomenon, which was independent on scale. On the other hand, [29] and [14] were the first authors to study Intermittency in the inner heliosphere. In particular, the former authors showed the different intermittent character of fast and slow wind while the latter ones showed a possible first evidence for the Kraichnan scaling ( [25]) in a magnetofluid like the solar wind. Since then, several papers ( [9,11,21,35,40,43,51]) among others followed these first approaches to the problem of understanding Intermittency in the solar wind. However, novel techniques based on the properties of wavelets introduced by [18] and firstly used within ordinary fluid dynamics by [34] and within the solar wind context by [55], [9] and [10], represented a powerful tool to finally disclose the nature of intermittent events. In this last paper, the authors showed that the intermittent event they were able to single out was a so called current-sheet-associated TD, as defined by [20]. This type of structure is associated with wind velocity gradients, a rapid magnetic field intensity change and a reversal of the maximum variance magnetic field component across the discontinuity. This discontinuity was interpreted as the interface between two adjacent flux tubes, i.e. two regions characterized by different plasma and magnetic field conditions although within the same large scale plasma region. During these studies, it was also noticed that the jumps performed by the tips of magnetic and velocity vectors were such to resemble a typical Lévy process ( [12]). This process is similar to a random walk but the statistics governing the spatial jumps is characterized by extreme behaviour. Consequently, the presence of long-range correlations makes the Gaussian statistics, which governs the Brownian motion, no longer representative of the physical process. In fact, the spatial distribution of the directions assumed by these vectors during the selected time interval was not uniform but rather patchy. This particular behavior was indicating the presence of particular directions along which the fluctuating field vector would roughly remain aligned for a longer time. Then, a rapid and large jump would characterize the transfer from one patch to another. These large jumps were recognized to make the fluctuating field more intermittent. Moreover, the highly Alfvénic character of the selected time interval clearly showed that propagating modes and coherent structures were both contributing to the observed turbulence and strengthened the already proposed view of a turbulence made of a mixture of waves and structures ( [8,22,29,30,48,49], among others). In reality, it took a long time to reach this view of interplanetary MHD turbulence. As a matter of fact, the first spectra of solar wind fluctuations obtained from the observations of Mariner 2 in 1962, were interpreted by [17] as evidence for the presence of turbulent processes, possibly MHD turbulence as described by [25]. The velocity shear mechanism proposed by [17] would be sustained by strong velocity gradients present in the solar wind and would produce large scale Alfvén waves that would transfer their energy to smaller and smaller scales through a turbulent process. On the other hand, [4], looking at the correlation between velocity and magnetic field fluctuations observed by Mariner 5, concluded that interplanetary fluctuations were exclusively made of outward propagating Alfvén modes, mostly, of solar origin. Obviously, these two point of view were in contradiction because the absence of inward propagating modes would preclude the development of turbulent cascade of energy to form a spectrum similar to the one observed by [17]. In reality, interplanetary fluctuations are not just Alfvén waves. Only several years later, a careful data analysis performed on the observations provided by the Helios spacecraft contributed to make the idea of a possible coexistence of propagating waves and convected structures accepted (see review by [50]). Moreover, only recently, theoretical efforts ( [53,54,56] have shown the possibility that propagating modes and coherent structures might share a common origin within the general view described by the physics of complexity. Propagating modes would experience resonances which generate coherent structures which, in turn, will migrate, interact and eventually generate new modes. Moreover, [37] using a 1-D MHD numerical simulations based on a pseudo-spectral code, were able to qualitatively reproduce the radial behavior of magnetic field and velocity Intermittency observed by [11] in the inner heliosphere. In particular, they numerically simulated the propagation of a turbulent Alfvénic spectrum in a uniform background magnetic field. As a matter of fact, coherent structures were created during the spectral evolution due to the parametric instability resembling a sort of shocklets or current sheets. Obviously, the model has strong limitations since while the 1-D code allows dependence on only 1 spatial coordinate, the vectors can have all 3 cartesian components. As a consequence, these results remain at a qualitative level.
Using the idea of [32] as a starting point, [10] proposed the spaghetti-like model in which Alfvénic fluctuations propagate within a convected structure made of tangled flux-tubes, each of them being characterized by its local magnetic field and plasma. These structures represent the correlated part of the signal, unlike from the Alfvénic component which does not conserve spatio-temporal coherence. Moreover, it was found that the statistics associated to these fluctuations experienced a strong radial evolution. As a matter of fact, directional jumps, within fast solar wind, evolved from a more Gaussian-like statistics at 0.3AU towards a sort of Truncated Lévy Flight ( [28] statistics at 0.9 AU. This phenomenon suggested that the possible cause of the radial evolution was the radial progressive depletion of the Alfvénic component of the turbulent fluctuations with respect to the convected structure. However, the limited statistics due to the low resolution of plasma data did not allow a more refined analysis. In this paper we used higher resolution data, about one order of magnitude higher in frequency, to study in more detail the structure of the PDFs of directional fluctuations but we had to limit this study to the only magnetic component of the fluctuations.

II. DATA ANALYSIS
Data used in this work are 6 sec averages of magnetic field measurements performed by Helios 2 s/c during its primary mission to the sun almost three decades ago, in 1976. This is the only data set available covering the heliocentric distance range between 0.3, closest approach to the sun, and 1AU. So far, this data set has been extremely valuable to study the main physical mechanisms governing the solar wind turbulence. In particular, it has been very useful to study the radial evolution of interplanetary turbulence. As a matter of fact, this data set contains in-situ observations at different heliocentric distances of magnetic field and plasma belonging to the same corotating solar source region of high velocity wind.
Thus, the stationary character of this region allowed studies which could focus on the radial evolution of solar wind turbulence, providing extremely important results which have been reviewed in an excellent paper by [50]. In the present analysis we will refer to this particular high velocity stream and to the low velocity wind ahead of it observed at two different heliocentric distances, namely 0.3 and 0.9 AU. These time intervals are highlighted in Fig. 1 by vertical hatched stripes which, from the top to the bottom of the figure, go across velocity and heliocentric distance in the top panel and magnetic field intensity in the bottom panel.
Each of the four time intervals lasts 2 days and it has approximately 28800 6sec averages, taking into account data gaps. Temporal extremes, average heliocentric distance, average wind speed and field intensity are reported in Table I for each interval.
Just for sake of completeness we like to add that the same corotating stream was also observed at 0.7 AU starting around day 74 but we will omit that time interval since it would be redundant for the analysis we present here. As already said in section I, the main goal of this work is the study and characterization of directional fluctuations. To do so, we start with plotting the position of the tip of the magnetic field vector within the reference system of   AU appear to be slightly larger. Definitely, we do not observe the same radial evolution noticed for fast wind enhanced regularity of this surface, if compared to slow wind in the following, depends on the fact that, during this high velocity stream, field intensity is quite constant, as generally expected for a high velocity stream. One of the main features of this stream is that it is highly Alfvénic as already stressed several times in the available literature (see for example paper by [7]). This fact, obviously, induces large directional fluctuations on the direction of the ambient magnetic field. As a consequence, the dark spot in the figure completely covers half sphere. Obviously, only half of the sphere is covered since magnetic field polarity remains generally constant within high velocity streams while it can flip from positive to negative and vice versa more easily within slow wind. However, when we reach 0.9 AU the situation dramatically evolves. Although the tip of the vector still wanders on the surface of only half a sphere, it does not cover this surface completely but leaves out wide areas.
The distribution of the dark spots suggests that the presence of preferred spatial directions, connected by large and quick jumps which take only a few data points, begins to emerge as the heliocentric distance increases.
On the contrary, the evolution we just noticed in fast wind is much less dramatic within slow wind as shown in the two panels of Thus, following this definition, each individual |δB(t)| is given by: This time sequence, the same used for the top panel of Fig. 3 which refers to slow wind at 0.3 AU, clearly shows small amplitude and high frequency fluctuations superimposed on a sort of larger amplitude low frequency background structure. This background structure is characterized by few large and quick directional jumps. The effect of these jumps is to move the fluctuating vector from one particular average direction to another, i.e. from one dark spot to another (Fig. 3). This type of information, together with the 3D graphical representation, gives an idea of how the vector direction really fluctuates in space and time.
Moreover, most of the time the largest directional jumps are associated to the largest changes of the field intensity (bottom panel). So, these two panels suggest that during short time intervals the field can be characterized by a most probable orientation and a most probable intensity. In other words, these regions appear to be distinguishable from each other and the transition from one to another is through a large rotational jump and a change in the field intensity. Similar findings have already been reported in a previous paper ( [10]) although it was focused on a single case study and on larger scales. In that same study it was found that this kind of transition, or border, was a tangential discontinuity not in pressure balance.
The features we notice in the present study might be TDs as well although we cannot prove it since we don't have plasma data with the same time resolution of magnetic field data. If this is the case, the structure we have seen at larger scales replicates at smaller scales in a kind of self-similar manner.
Results relative to slow wind at 0.9 AU are shown in Fig lies in the fact that the amplitude of Alfvénic fluctuations is largely reduced at the Earth's orbit compared to short heliocentric distances (see the wide related literature in [50]). This qualitative study has to be substantiate with some quantitative evaluation of the relative importance of these two components contributing to the observed interplanetary turbulence.
To do so, in the following, we will discuss and compare the probability distribution functions AU in the top and bottom panels, respectively. We found that both distributions can be reasonably well fitted by a double lognormal distribution in the form reported by eq. 2.
where ξ, δ, σ > 0. The variable ξ stands for the different |δB i |/σ, one for each bin of the distribution, A i is a measure of the area under each curve, δ i is called scale parameter and represents the median, σ i is the shape parameter. Larger values of σ i push the x-location of the peak of the distribution towards lower values. Obviously, using a larger number of lognormals would provide a better fit but, the real conspicuous improvement is obtained only when we use two lognormals instead of just one.
The two lognormal components have a distinguishable role in the total PDF.  Similar conclusions apply to the PDFs relative to the angular fluctuations δα experienced by the vector orientation shown in Fig. 9. Obviously, this measure provides information only about directional fluctuations and is not influenced by compressive effects that may act on the vector intensity. As such, information contained in Fig. 9 are less meaningful than those discussed earlier but, we like to show this kind of Figure, and the analogous one for slow wind in Fig. 11, just for sake of completeness. Moreover, these distributions have not been normalized to their respective σs since we like to show the effective angular range of these fluctuations. Also for this fit we report the relative parameters which are shown in Tab. III. For these fluctuations the ratio A 2 /A 1 varies from 56% at 0.3 AU to 40% at 0.9 AU. The PDF is clearly peaked at larger angles (5.75 • compared to 2.25 • ) at 0.3 AU and its right tail reaches values close to 100 • . In Fig. 10 we show the vector displacement |δB| normalized to σ for slow wind at 0.3 and 0.9 AU in the same format as of Fig. 8.
The situation within slow wind shows that radial evolution is almost absent. This can be inferred from the values of the best fits parameters reported in Tab. IV which show that the contribution of the smaller lognormal can be neglected at both distances. As a matter of fact, the relative contribution of the smaller lognormal is between 0.26% at 0.3 AU and 0.13% at 0.9 AU, respectively. Moreover, the values of the parameters relative to the larger lognormal only slightly change between 0.3 and 0.9 AU suggesting that these fluctuations do not evolve much with radial distance, as expected for slow wind. In summary, remarkable is the constancy of all parameters inferred from the fit of the main lognormal, which highlights the absence of radial evolution. In this case, values of δ's are considerably smaller than those obtained for fast wind, confirming that these fluctuations are generally smaller.
As already reported for fast wind, we like to show the PDFs relative to the angular fluctuations as shown in Fig. 11 and parameters relative to the best fit which are shown in Tab. V. Also in this case, the contribution of the smaller lognormal is much smaller than within fast wind. As a matter of fact, the ratio A 2 /A 1 varies from 7.5% at 0.3 AU to 4.4% at 0.9 AU. Moreover, these PDFs are roughly peaked at the same angle (0.75 • )   Fig. 12 and Fig. 13, for fast and slow wind, respectively.
The only arbitrary imposition we applied on the fast wind sample was that to keep always the same vector polarity to resemble, as much as possible, the real situation within fast wind. As a consequence, we forced these fluctuations to remain within a solid angle of 2π aperture. These plots reproduce at some level the main features that can be observed in On the other hand, we like to show temporal sequences of these |δB(t)| relative to a fixed, arbitrary direction as we did for real data shown in the top panels of Fig. 4 to 7.
Only the top panels have to be considered since, artificial data have been built keeping the vector intensity constant. Results are quite satisfactory since we are able to reproduce the typical behavior observed within both fast and slow wind. In particular, the transition from the chaotic behavior on the left panel of Fig. 14, representing fluctuations at 0.3 AU, FIG. 12: 3D representation of vector displacements relative to artificial data generated by a random-walk whose jumps are obey to a double-lognormal whose parameters have been obtained by the best fit of real fluctuations. The top and bottom panels refer to fast wind at 0.9 and 0.3 AU, respectively and, have to be compared to analogous plots for real data shown in Fig. 2.
FIG. 13: 3D representation of vector displacements relative to artificial data generated by a random-walk whose jumps are obey to a double-lognormal whose parameters have been obtained by the best fit of real fluctuations. The top and bottom panels refer to slow wind at 0.9 and 0.3 AU, respectively and, have to be compared to analogous plots for real data shown in Fig. 3.

III. RESULTS OF THE NUMERICAL SIMULATIONS OF PARAMETRIC IN-STABILITY
Recently, [26,27,36,37] investigated in detail how the parametric instability could be responsible for typical features observed in the radial evolution of the Alfvénic turbulence in the solar wind high speed streams. This instability develops in a compressible plasma and, in its simplest form, involves the decay of a large amplitude Alfvén wave (generally called "pump wave", or "mother wave") in a magnetosonic fluctuation and a backscattered Alfvén wave. The wavevectors and frequencies of the fluctuations generated in this process are mutually related through well precise "resonance conditions" [41]. This mechanism can be viewed as a way for decorrelating an initially coherent state (the large amplitude mother waves should travel almost undisturbed and the observed (although slower than in slow wind) radial evolution of the Alfvénic turbulence need some mechanism to be ascribed to.
Although, it has been shown ( [19,39]) that plasma instabilities generated by velocity shears play a relevant role in the radial evolution of turbulence, another possibility is represented by parametric instability as shown by [26] and [37]. These last studies related to parametric instability focus the attention on the effects of this instability on the evolution of a large amplitude, circularly polarized, non-monochromatic Alfvén wave in a one dimensional case. The spectrum of this initial perturbation had a break at a certain wavelenght, like the spectrum of the Alfvénic fluctuations coming from the sun. They found that the Alfvénic correlation of the initial perturbation is lost during the time evolution, because of the parametric instability, leading to a production of both backscattered Alfvénic perturbations and magnetosonic waves. Finally, these perturbations evolve non linearly, producing approximately power law spectra and a reduction of the normalized cross helicity. The results found are qualitatively in good agreement with solar wind observations carried out by several authors (see review by [50]). In addition, [26] observed that the turbulent development of the instability leads to the formation of shock waves and to an intermittent behaviour of the dissipation. In particular, looking at the evolution of the flatness of velocity and magnetic field fluctuations, [37] found a good qualitative agreement of the results of the simulations with the analysis of the same quantities performed by [11]. It is then natural, in order to offer a possible different interpretation of the results shown in the previous sections, at least those concerning fast solar wind streams, to see whether turbulence induced by parametric instability has characteristics similar to those described in the solar wind in the previous sections. To accomplish this aim, we furtherly analysed the results of the numerical simulations described in [37]. The details of the numerical code can be found in [36], [26] and [27], whilst further details concerning the simulations are given in [37].
We simulate the evolution of a broad-band Alfvénic fluctuations in a compressible plasma, during their outwards propagation in the heliosphere. Similarly to the in-situ observations, the initial spectrum has a break-point. During the run of the simulation, inward propagating fluctuations start to appear and form a power-law spectrum at small k ′ s. As already pointed out by [47], this feature might suggest that parametric decay mechanism is at work in the solar wind.
The simulation domain is one-dimensional, periodic and we use cartesian geometry. The reference frame is chosen in such a way that the initial Alfvén wave is circularly polarized in the x − z plane and it propagates along the y direction. A background constant magnetic field intensity B 0 is imposed in the propagation direction of the wave: the resulting total field has, therefore, uniform intensity everywhere. The homogeneous boundary conditions limit the application of this study to the fast wind, where the background magnetic field is rather homogeneous. In our framework, the time evolution of the quantities represents the radial evolution of the fluctuations in the solar wind, while the spatial variations are the numerical counterpart of samples of the observed data at a given distance from the sun. We study the evolution of the parametric instability for 180 τ A (τ A is the Alfvén time based on the initial background radial magnetic field and density, i.e. the time needed for the wave, whose wavelength is the largest in our spectral domain, to go across the simulation box).
In the rest of the paper, we plot quantities at time t 1 = 45τ A and t 2 = 180τ A , the former corresponding to a time much before the saturation of the instability, the latter to a time longer than the saturation time, which is reached at t sat ∼ 100τ A . Practically, we consider a situation in which the instability has only weakly taken place and another in which it has already completely developed, that should be representative of the state of the solar wind closer to the sun and further away from it, respectively.
We estimated the time needed by the instability to saturate and to reproduce the spectral features observed at 0.9 AU. We found that a period of time between 6 and 7 days is necessary. This estimation, although longer, is still within the same order of magnitude of evaluation is based on the fact that between t 1 and t sat there are about 55τ A . In order to estimate τ A we compared the frequencies corresponding to the observed spectral break at 0.3 AU in Helios data with the corresponding one shown in our simulation at t = t 1 .
In Fig. 17, we show the three-dimensional plots of the magnetic field components, nor-  appear to be more structured in space and some sort of correlation with the field intensity (bottom panel) can be recognized. This behaviour is qualitatively similar to that observed in fast solar wind at 0.9AU (Fig. 7).
At the later time t 2 = 180τ A the situation changes dramatically. The tip of the magnetic field vector describes a more patchy pattern, characterized by large jumps, followed by smaller fluctuations around a single direction, and so on. This pattern is qualitatively similar to that observed in Fig. 2. In conclusion, results of parametric instability simulation seem to account for the observed transition to a sort of Lévy walk observed in the real solar wind data (see also [12]).
As a second point of agreement with the observations, we plot the vector displacements |δB|, with respect to a fixed direction, normalized to the mean magnetic field intensity as a function of the independent variable y. We compute this quantity by considering the magnetic field at a given time t and evaluating the vector differences of its components  Also in this case the qualitative similarity between the results of the simulations and the observed solar wind data is remarkable. At the time t = t 1 , corresponding to 0.3AU, the vector displacements have a quite random behaviour, although not as "noisy" as in Fig. 6, and no evident correlation between the vector displacements and the magnetic field intensity is observed. Note that the fluctuations of the magnetic field intensity are rather small at this time, due to the fact that the initial wave is circularly polarized.
On the converse, in Fig. 19, the vector displacements of the magnetic field appear to be Finally, we plotted in Fig. 20 the PDFs of vector displacements |δB λ |, defined as normalized to their standard deviations, at times t 1 and t 2 . In eq.3, t represents a specific  The similarity with Fig. 8, where the PDFs of the vector displacements |δB| are shown for the fast wind at 0.3 and 0.9AU, is evident. Also in this case, it is possible to fit effectively the curves with two lognormal distributions and the trend is similar to that observed in the solar wind data. In fact, the two lognormal distributions have comparable heights at t = t 1 , while the population relative to the long vector displacements increases its importance at subsequent times (t = t 2 ). The parameters of the fits are shown in table VI.
Another point of similarity with the analysis of real data regards the power spectrum.
As a matter of fact, as already shown by [26] and [27], the power spectrum obtained from the trace of the spectral matrix of the Alfvénic fluctuations, after the saturation of the parametric instability has been reached, shows a clear evidence of a power law inertial range similar to the one observed in Helios data at 0.9 AU.
We like to stress that a direct quantitative comparison between the real data and the simulated data is meaningless. In fact, a direct comparison would require to fix appropriate length and time scales in the simulation. Moreover, as described in [37], the model used is very simple, due to the one-dimensionality constraint, to the cartesian geometry and especially to the fact that the solar wind is actually an expanding medium. However, the results of the simulations show a qualitative behaviour very similar to that observed in real data of the fast solar wind. We conclude that the parametric instability offers a possible alternative explanation of the observed data.

IV. CONCLUSIONS
In this paper we focused on the statistics followed by interplanetary magnetic field fluctuations on a 6-sec time scale, well inside the MHD regime ( [2], as observed in solar wind turbulence between 0.3 and 0.9 AU. In particular, we aimed to understand the spatiotemporal evolution of the magnetic field vector through the study of changes experienced by both vector orientation and intensity. Several previous works, which dealt with a statistical approach to this same problem, considered different aspects connected to directional fluctuations as, for example, power associated to the fluctuations, their radial evolution, their anisotropy, the nature of the fluctuations, their generation mechanisms, and so on but, none of them, to our knowledge, has ever studied how and why the orientation of these fluctuations changes with time. There have been only few attempts to study similar problems but always limited to single case studies ( [10,33,38,44,48]. The most recent statistical approach to the same problem is represented by a paper by [12] in which these authors concluded that the temporal evolution of magnetic field and wind velocity vectors directions might follow a sort of Lévy walk. That paper, although based on larger time scales and on a weak statistics, represents the first attempt to understand the influence due to propagating modes and convected structures on the orientation of velocity and magnetic field vectors within MHD turbulence. Following this analysis and, using a more robust statistics, we found that PDFs of interplanetary magnetic field vector differences within high velocity streams can be reasonably fitted by a double log-normal distribution. In other words, vector differences, which are due to the two distinct contributions of directional uncompressive fluctuations and purely compressive fluctuations, can be separated in two distinct PDFs. Moreover, the lognormal nature of the PDFs might suggest a multiplicative process at the origin of these fluctuations, that is typical of a turbulent cascade. Furthermore, it only applies to definite positive quantities, like the vector or angular displacements we analyze in this paper. Incidentally, the multiplicative cascade notion was introduced by Kolmogorov into his statistical theory ( [23,24] of turbulence as a phenomenological framework to accomodate extreme behaviour observed in real turbulent fluids.
Another interesting feature of these distributions is that the two PDFs have a different weight since one of them, the one that represents the smallest |δB i | is always considerably smaller than the other one. Moreover, while the smaller PDF does evolve with heliocentric distance, decreasing its own relevance, the largest PDF seems to remain almost unaffected.
Now, if we consider, as already suggested ( [11]), MHD turbulence mainly due to propagating, uncompressive fluctuations of Alfvénic origin and to convected compressive structures, it comes natural to identify these different contributions to turbulence with the two PDFs we found. In addition, we would expect a different radial evolution since only propagating Alfvénic modes, interacting non-linearly, undergo a considerable turbulent evolution as we already know from literature ( [50]. We found that the relative contribution to the total PDF of what we identify with uncompressive fluctuations varies from 34% at 0.3 AU to 18% to 0.9 AU, in terms of relative probability. In other words what we identify with the Alfvénic contribution results to be somewhat smaller than the contribution due to the convected structures. Similar conclusions were reached by [5] who gave an estimate for the Alfvénic component around 15% of the total power associated to turbulence. Thus, we might associate our convected structures to the 2-D turbulence identified in the solar wind by [5] and [30] who model interplanetary magnetic turbulence as made of slab and quasi-2D turbulence only. However, the dominant 2-D magnetic turbulence is characterized by the fact that its wave vector results to be normal to the ambient magnetic field direction. As a consequence, we would expect to see a radial evolution even stronger than the one we observed for the slab component which has its wave vectors parallel to the ambient field. As a matter of fact, the turbulent cascade acts preferably on wave numbers perpendicular to the ambient magnetic field direction, as suggested by the three-wave resonant interaction ( [6,42]. On the contrary, the dominant component of the turbulence observed by Helios is the least affected by the radial evolution and, probably, should not be identified with the 2-D turbulence. Another possibility is that the 2-D turbulence is mixed together with the slab turbulence and represented by the smaller PDF which experiences the stronger radial evolution. If this is the case, our analysis suggests that interplanetary fluctuations are made of three rather than two components: slab, 2-D and convected structures which would support the three component model by [31]. This view is corroborated by the fact that the PDF of |δB i | within slow wind can be fitted by a single lognormal whose parameters only slightly change with heliocentric distance. As a matter of fact, this behavior has to be expected if we consider that slow wind is poor in Alfvén modes and its turbulence is already fully developed by the time we observe it at 0.3 AU ( [46]. Consequently, between.3 and .9 AU, fluctuations do not undergo the same turbulent evolution observed in the fast wind and the constancy of the PDF of |δB i | should be expected. We like to stress that interplanetary observations revealed that slow wind MHD fluctuations are intrinsically different from those observed in fast wind. In fact, [3] showed that the Elsässer ratio e − /e + , which is always around 1 within slow wind, saturates to ∼ 0.5 within fast wind at a distance of ∼ 2.5 AU.
One more interesting observation regards the topology showed by these fluctuations within fast and slow wind. We showed that the trajectory followed by the tip of the magnetic vector during its turbulent fluctuations follows a structured path. This path appears more clearly when the PDF of |δB i | can be fitted by a single lognormal, as in the case of slow wind regardless of heliocentric distance. However, within fast wind this structured path can be more easily observed with increasing the heliocentric distance, in concurrence with the depletion of the Alfvénic fluctuations. In other words, Alfvénic modes mask the underlying magnetic, quasi-static structure convected by the wind. The superposition of these two types of fluctuations is such that the final motion is characterized by extreme behaviour. Referring to the 3D representation used in this paper, the tip of the vector appears to be trapped within a certain solid angle for sometime but, occasionally, it escapes this limited angular region and quickly travels, in a few time steps, to finally end up in another angular region characterized by a different average orientation. These large jumps should be accounted for by the larger PDF and should be related to similar large jumps studied by [10] and interpreted as tangential discontinuities marking the border between adjacent flux tubes. On the contrary, local fluctuations, clustering around certain average directions, should have Alfvénic nature and should be identified by the smaller PDF. These results support and further corroborate the recently re-proposed spaghetti-like structure model ( [10] firstly introduced, although in the context of cosmic ray modulation, by [32] to describe interplanetary magnetic field topology.
Finally, adopting a sort of feedback procedure, we cross-checked the soundness of our fitting scheme showing that artificial data obtained from the tip of a vector that randomly walk on the surface of a sphere of constant radius, performing directional jumps which obey to a double lognormal, provides results similar, in some aspects, to those observed in interplanetary space.
However, the interplanetary observations we have do not allow to understand whether these structures come directly from the Sun or are locally generated by some mechanism.
Recent theoretical results by [37] showed that coherent structures responsible for the radial dependence of Intermittency as observed in the solar wind ( [11], might be locally created by parametric decay of Alfvén waves. These authors, showed that during the turbulent evolution, coherent structures like shocklets and/or current sheets were continuously created when the instability was active. In order to see whether a similar mechanism may account for the observed behaviour of the vector displacements and their statistics, we further analyzed in this paper the results of the simulations performed in [37]. The results of this investigation show a fairly good agreement, at least under the qualitative point of view, between the simulations and the solar wind data: either the evolution of the tip of the magnetic field vector, or the correlation between the vector displacement at a given scale with the magnetic field intensity fluctuations, or the evolution of the PDFs of the vector displacement in time, all show trends similar to those observed in the real fast solar wind data. Unfortunately, a direct quantitative comparison between the simulations and the data is difficult due to the limitations of the model. However, this mechanism, which might be active within fast wind, should be less effective within slow wind given the remarkable decoupling between magnetic field and velocity field within this type of wind ( [22]. Nevertheless, the enticing nature of the parametric instability in explaining the results comes from some well defined fact: a) it is a well defined mechanism of physical origin that induces a turbulent evolution in the plasma and not an undefined turbulence model; b) it is likely applicable to explain many general observed features of fast solar wind, like the evolution of the spectra, the decrease of the Alfvénic correlation during the propagation in the heliosphere, and so on ( [37]; c) the observed evolution of the vector displacements and of their relative PDFs can be seen as a natural consequence of the formation of shocklets and discontinuities in the wind, organized in a sort of coherent structures, that explain the long jumps observed in the magnetic field and the structures in the vector displacements at larger distances from the sun. In particular, the decrease with distance of the lognormal component of the PDFs correlated to the Alfvénic part of the turbulence, can be seen as the continuous transfer of energy between the Alfvénic and magnetosonic components of the waves during the evolution of the instability. However, a definitive conclusion about this point needs further investigations.
Another recent theoretical effort by [16] models MHD turbulence in a way that tends to the view and interpretation of the interplanetary observations we presented in this paper, that is the existence of two different components both contributing to turbulence. The theoretical model presented by these authors tells us that propagating modes and coherent, convected structures are both necessary, inseparable ingredients of MHD turbulence since they share a common origin within the general view described by the physics of complexity ( [15,53,54]. Propagating modes experience resonances which generate coherent structures which, in turn, will migrate, interact and eventually generate new modes.
These theoretical models, which favour the local generation of coherent structures, fully complement the possible solar origin of the convected component of interplanetary MHD turbulence.