Solar Wind Structures from the Gaussianity of Magnetic Magnitude

The solar atmosphere is highly structured both spatially and temporally (Bale et al., 2019; Kasper et al., 2019). Recent studies have successfully established connections between PSP (Fox et al. (2016) and see Raouafi et al. (2023a) for a review for results of the first four years of the mission) in situ observations and solar atmospheric structures including mid-latitude coronal holes (Badman et al., 2023; Davis et al., 2023), pseudostreamers (Kasper et al., 2021), and supergranulation (Bale et al., 2021; Fargette et al., 2021; Bale et al., 2023), even though alternative explanations remain (Shi et al., 2022). Recent advances in remote sensing provide strong support for the minutes long small-scale jetting activity from magnetic reconnection (“jetlets”) as a major source of the solar wind (Raouafi et al., 2023b). In addition, EUV observations from Solar Orbiter (Müller et al., 2020) unveiled ubiquitous brightening termed “picoflare” (Chitta et al., 2023) with associated jets that last only a few tens of seconds, suggesting the solar wind source might be highly intermittent. However, magnetic footpoint mapping methods (Badman et al., 2020; Panasenco et al., 2020; Badman et al., 2023) use photospheric magnetic field observations over the whole visible disk that are refreshed at best once every six hours and lack, of course, any real temporal reliability for the far side. Therefore, such methods are hardly able to reliably contextualize and explain the boundaries of the highly structured solar wind in situ time series, except perhaps in a statistical sense.

Previous studies on the distribution of solar wind parameters primarily focused on the statistical properties of the solar wind magnetic field and plasma moments at 1 AU and beyond. For example, Whang (1977) and Padhye et al. (2001) found that the magnetic field components follow near-Gaussian distributions, and Bandyopadhyay et al. (2020) showed similar properties in Earth’s magnetosheath turbulence. For the magnetic field magnitude, Burlaga & Lazarus (2000) and Burlaga (2001) demonstrated that it generally follows a log-normal distribution. On the simulation side, Yamamoto & Kambe (1991) conducted a 3D hydrodynamic simulation and found that the velocity components follow a Gaussian distribution at small wave numbers and a long-tail distribution at larger wave numbers. It is hence empirically expected that the magnetic field components $B_{x,y,z}$ follow a Gaussian distribution, and thus the magnetic magnitude $B=\sqrt{B_{x}^{2}+B_{y}^{2}+B_{z}^{2}}$ follows a $\chi$ or Maxwell-Boltzmann distribution. However, for the solar wind measured at PSP perihelion distances, the distributions of magnetic field components generally deviate significantly from Gaussian due to the presence of large amplitude spherically polarized Alfvén waves (i.e., switchbacks, see e.g., Bale et al. (2019); Larosa et al. (2021); Drake et al. (2021); Dudok de Wit et al. (2020); Shi et al. (2022); Bale et al. (2023)). Therefore, it is not expected from any of the previous studies that the magnetic magnitude should follow a Gaussian distribution or any known distributions.

Contrary to previous attempts, using a data mining approach, this study starts by reporting an unexpected property of the solar wind magnetic field. We find that the magnetic magnitude $B$, once normalized by a power law fit with regard to the heliocentric distance, occasionally exhibits perfect Gaussian distributions. Based on this discovery, we introduce a novel time series visualization method (time-frequency representation or TFR) named Gaussianity Scalogram (GS) to visualize the spatial-scale (time-frequency) dependent Gaussianity of $B$. Applying this method to data from PSP and Ulysses, we find that the Gaussian intervals successfully map to the in situ remnants of coronal holes (Badman et al., 2023; Davis et al., 2023) and switchback patches (Bale et al., 2021; Shi et al., 2022; Bale et al., 2023). On smaller scales, Gaussian intervals map to structures that are temporally compatible with small-scale jetting activity, including ”jetlets” and ”picoflares” (Raouafi et al., 2023b; Chitta et al., 2023). The rest of the paper is structured as follows: In the next section, we describe the helio-radial power law fit and the construction of GS; in Section 3, we present some applications of GS; in Section 4, we discuss the computer simulation of Gaussianization of $B$ in MHD turbulence and the major implications; in Section 5, we conclude and summarize our results.

Two of the most interesting yet overlooked features of the time series of the solar wind magnetic field magnitude $B$ are that: 1. Sometimes $B$ displays a surprisingly stable power law dependence on the heliocentric distance $R$; 2. By applying a helio-radial power law fit between $B$ and $R$, i.e. $B\propto R^{-s}$, the fit normalized magnetic magnitude $B^{*}=B(R/R_{0})^{s}$ sometimes displays a near-perfect Gaussian distribution. This is illustrated in Figure 1 (a-c), where the selected interval is highlighted with a golden bar in panel (a) and the helio-radial power law fit (fit index $s=1.86$) is shown in the inset figure. The histogram of $B$ is shown in blue in panel (b) and the normalized $B^{*}$ is shown in red. To illustrate the close proximity of the probability density function of $B^{*}$ ($PDF_{B^{*}}$) to a Gaussian distribution ($\mathcal{N}$), a standard Gaussian curve is overplotted in panel (c) (shifted with the mean value $\langle B^{*}\rangle$ and scaled with the standard deviation $\sigma_{B^{*}}$). The Jensen-Shannon Distance (JSD) of base $e$ (a statistical distance metric between probability density functions (Lin, 1991)) is calculated between $PDF_{B^{*}}$ and $\mathcal{N}$, and the value is $JSD(PDF_{B^{*}},\mathcal{N})=10^{-1.431}$, indicating considerable closeness between two distribution functions (for benchmark, see appendix). In addition, this highly Gaussian $B^{*}$ interval coincides with the radial solar wind speed profile which is visualized with radial colored lines in panel (a) and Figure 9 (c) (compiled with SPAN-ion from SWEAP suite Kasper et al. (2016)). From Nov-17 to Nov-20, the spacecraft was immersed in the high speed solar wind, indicating its coronal hole origin. The JSD produced by this process is represented as one pixel (tip of the green pyramid) in the Gaussianity Scalogram (GS) shown in panel (d3), and the scalogram for the corresponding helio-radial power law fit index $s$ is displayed in panel (d4).

Each pixel in the GS is characterized by a timestamp ($t_{mid}$) and window size ($win$), similar to wavelet scalogram. Uniquely in GS, the step size in $win$ (vertical axis) is chosen to be twice the step size in $t_{mid}$ (horizontal axis), and thus the time range covered by one pixel corresponds to the same time range covered by three pixels in the following row, and so on towards the smallest scales. Therefore, if an interval and the nested sub-intervals possess similar characteristics (e.g. relatively small JSD regardless of $t_{mid}$ and $win$ within the interval), a pyramidal structure is expected from the GS, and the base of the pyramid indicates the start and end time of the interval. One example is highlighted by the green dashed pyramid in panel (d3), where the tip of the pyramid is in fact selected a posteriori as the local minimum in the GS ($PDF_{B^{*}}$ being closest to Gaussian among the surrounding time and scales). Ample information can be inferred from the GS: 1. A semi-crossing of the heliospheric current sheet (HCS) at noon of Nov-22 is visualized as an inverted black pyramid. This is because the rapid drop of $B$ during HCS crossing can significantly destroy its Gaussianity; 2. It has been confirmed recently by Badman et al. (2023) that the solar wind can be traced back to a single mid-latitude coronal hole from Nov-17 to the end of Nov-20, and from another coronal hole for the whole day of Nov-21 (see also Bale et al. (2023); Panasenco et al. (2020); Badman et al. (2020)). The coronal holes are naturally visualized here as two white pyramids (green and red dashed lines) separated by a dark region around the mid-night of Nov-20; 3. The helio-radial power law fit index $s$ is unexpectedly stable regardless of locations and scales and systematically deviates from $R^{-2}$ ($s\simeq 1.87\pm 0.02$) (see also Bale et al. (2019)).

The clear correspondence between the white pyramid and coronal hole encourages us to predict intervals of solar wind originating from coronal holes with GS compiled from PSP data. Among the first 14 encounters (Nov-2018 to Dec-2022), we only identified one more (for a total of 2) long intervals ($>$ 3 days) characterized by high Gaussianity in $B^{*}$. A panoramic view of these two long intervals is shown in Figure 9. The newly found interval from the inbound of E12, shown in Figure 10 and Figure 9 (d), is characterized by a 5-day long highly Gaussian $B^{*}$ time series. For illustration purpose, the green pyramid in Figure 10 (d3) is selected as the deepest local minimum in GS for $win>$ 3 days. The histogram of $B^{*}$ is remarkably concentrated (panel (b)) and aligns with Gaussian almost perfectly within 4 standard deviation (panel (c)). Similar to Figure 1 (c), the non-Gaussian part of $PDF_{B^{*}}$ has a systematic bias towards magnetic holes (weaker magnetic magnitudes, for recent studies using PSP data see e.g. Yu et al. (2021, 2022)), and the helio-radial power law fit index scalogram also shows a systematic deviation from $s=2$, similarly $s\simeq 1.87\pm 0.02$. To validate this prediction, independent results from Potential Field Source Surface (PFSS) modeling is shown in Figure 2 (see Panasenco et al. (2020) for more details), which indicates that the selected interval is indeed magnetically connected to a mid-latitude coronal hole.

To substantiate the applications of GS, here we demonstrate several examples that visualize the fractal-like Gaussian structures in the solar wind (Due to the rapid movement of PSP around perihelia, the structures in the in situ time series can be categorized into two kinds. Spatial: longitudinal structures traversed by PSP; Temporal (radial): radial structures advected by the solar wind and/or propagation of Alfvén waves) based on Taylor Hypothesis (Perez et al., 2021). From the largest scales: Ulysses, years-long polar coronal hole (McComas et al., 2003), towards the smaller scales: hour-long switchback patches (Bale et al., 2021; Fargette et al., 2021; Shi et al., 2022; Bale et al., 2023); minutes-long structures compatible with “jetlets” Raouafi et al. (2023b); and second-long structures compatible with “picoflare” Chitta et al. (2023).

Figure 3 shows the GS of the first Ulysses orbit, and the colorbar in panel (b) is enhanced compared to Figure 1 (d3) for illustration purposes. The solar latitude and wind speed profile in panel (a) indicate that the spacecraft was in southern and northern polar coronal holes in the whole year of 1994, and from 1995 to 1997 (see also McComas et al. (2003)). The two large white pyramids in the GS clearly correspond to the two polar coronal holes. Notably, the boundary observed in panel (b) results from an artificial cut-off in the helio-radial power law fit, as shown in panel (c). The cut-off value is chosen to be $R_{max}/R_{min}=1.5$, below which the spacecraft is considered stationary, and $B$ is not normalized to $B^{*}$ using helio-radial power law fit prior to the calculation of Gaussianity. However, the Gaussianity is much weaker in the polar coronal holes compared to the mid-latitude coronal holes observed by PSP at much smaller heliocentric distance, and the histograms of magnetic magnitude show much more significant fat tail towards the magnetic holes side (not shown here). This indicates that the Gaussianity of magnetic magnitudes decreases with increasing heliocentric distance, possibly due to the transition from low-$\beta$ to high-$\beta$ environment ($\beta=2\mu_{0}P/B^{2}$ is the ratio between plasma thermal pressure $P$ and magnetic pressure $B^{2}/2\mu_{0}$), and the plasma thermal pressure hence has a larger influence on the distribution of $B$. Additionally this also indicates that magnetic holes are much more preferred than spikes in the solar wind plasma.

Figure 4 shows the hour-long switchback patches from a single mid-latitude coronal hole in PSP E10, which have been recently proposed to be the remnants of the supergranulations in the solar atmosphere (Bale et al., 2021; Fargette et al., 2021; Bale et al., 2023). To indicate the spacecraft movement, the Carrington longitude of PSP is plotted every one degree on the top bars of both panels (a) and (b), and the color indicates spacecraft angular velocity in the corotating frame (blue: prograde, red: retrograde, see also Figure 1 for PSP trajectory in the corotating frame at the same perihelion). The magnetic magnitude is normalized with a universal helio-radial power law fit index ($s=1.87$) and the GS is compiled with the high-resolution fluxgate magnetic data ($\sim$ 292 Hz, see Bale et al. (2016); Bowen et al. (2020)). The red dashed pyramids in panel (a) and (b) are drawn to highlight the $B^{*}$ intervals with high level of Gaussianity. The selected intervals in panel (a) show that the GS effectively captures some of the switchback patches. When these are compared with the Carrington longitude, it becomes evident that some of the structures align with the size of supergranulation, as discussed in Bale et al. (2023). However, other structures, which are smaller in angular size and likely temporal in nature, could be more accurately attributed to the ’breathing’ phenomenon of the solar wind, as explained in Berger et al. (2017); Shi et al. (2022). After the “fast radial scan” phase on Nov-18, the spacecraft began to rapidly retrograde on Nov-19 and Nov-20 (see Figure 1 (a) for the spacecraft trajectory in the corotating frame). For better comparison, an expanded view is shown in panel (b). The second and third pyramids also show decent capability of capturing the switchback patches, whereas the first pyramid seems to capture a boundary between the patches. Starting from 7:00 on Nov-20, the remaining patches consistently exhibit a high level of Gaussianity across all scales and locations, resulting in indistinct boundaries between them.

Figure 5 presents a hierarchic GS of the mid-latitude coronal hole from the inbound of PSP E12. In panels (d) and (e), focusing on the smallest scales resolvable by the JSD (approximately 1 minute, corresponding to around 20,000 data points for the shortest interval. For details on how the number of data points influences this analysis, see Appendix), we observe a surprising number of structures with distinct boundaries. In fact, these structures, typically lasting 1-10 minutes, are omnipresent in the Alfvénic solar wind for all PSP encounters. Notably, they are not limited to winds with a clear coronal hole origin, such as those in the outbound paths of E12 (for more details, see the video in the supplementary materials). These structures are typically separated (interrupted) by radial jets (i.e. individual switchbacks), and these separations are frequently accompanied by close to kinetic scale ($\lesssim$ 5 seconds) fluctuations that are bursty and short-lived in all three components of magnetic field. For further illustration, refer to the skewness scalogram video in the supplementary materials. Unlike the spatial structures shown in panels (a), (b), and (c) (as well as in Figure 1 and Figure 3), the longitude change of the spacecraft for each structure in panel (d) and (e) is less than 0.1 degree, as indicated by the crosses plotted every 0.1 Carrington longitude in the top bar. Therefore, these structures are likely temporal, i.e. advected by the solar wind. All of these features are highly compatible with the “jetlets” observed in equatorial coronal holes (Raouafi et al., 2023b), and therefore could potentially be the “building blocks” of the solar wind. In fact, even finer structures can be found with the normalized standard deviation ($\sigma_{B^{*}}/\langle B^{*}\rangle$) scalogram and skewness scalogram shown in Figure 11. For example, the small white pyramid around 8:36 in Figure 5e has two 30-seconds long substructures nested beneath in Figure 11. These seconds-long structures are intervals with smaller standard deviation compared to the surroundings, and their interruptions are temporally compatible with the “picoflare” (Chitta et al., 2023).

These observations indicate that the Alfvénic solar wind is permeated with highly Gaussian magnetic magnitude intervals that are often interrupted by radial jets (switchbacks) every 1-10 minutes. In addition, the magnetic fluctuations inside the intervals often resemble the small amplitude outward propagating linear Alfvén waves. It is therefore reasonable to model the system using small amplitude Alfvénic MHD turbulence. Figure 6 shows the temporal evolution of the $JSD(PDF_{B},\mathcal{N})$ of a 3D MHD small amplitude Alfvénic turbulence simulation (Shi et al., 2023). The simulation is run with $512^{3}$ periodic box, and is initialized with unidirectional small amplitude linearly polarized Alfvén waves with isotropic wave vector spectrum (see Methods section for more details). At $t=0.00\ t_{A}$ (Alfvén crossing time $t_{A}=L/v_{A}$, where $L$ is the simulation box size), $PDF_{B}$ deviates significantly from a Gaussian distribution due to the small amplitude shear Alfvén wave initialization (fluctuations in $B$ are positive definite). The corresponding JSD is highlighted as the first red dot in the lower panel and is much larger than 0. Surprisingly, within one Alfvén crossing time at $t=0.40\ t_{A}$, the distribution of $B$ rapidly relaxes to a near-perfect Gaussian distribution, and the JSD rapidly drops towards the ground truth value (see Benchmark in Methods). As the simulation evolves, the JSD remains considerably small and thus the distribution of $B$ remains very close to Gaussian. The simulation indicates that Gaussian is the natural relaxation state for magnetic magnitude in small amplitude Alfvénic turbulence, consistent with the ubiquitous 1-10 minutes Gaussian intervals found in the solar wind. However, a 3D analytical model of a switchback with constant magnetic magnitude and fully open field lines is not yet available (see Tenerani et al. (2020); Squire & Mallet (2022); Shi et al. (2024b); Matteini et al. (2024) for recent progress on switchback modeling). Therefore, our simulation can not reproduce the realistic physical condition of the solar wind turbulence in which large amplitude spherically polarized Alfvénic fluctuations dominate.

Nevertheless, the simulation suggests that information is fully exchanged within the system, as it propagates at Alfvén speed throughout the simulation box. This allows $B$ to relax to a Gaussian distribution, which occurs within about 0.5 $t_{A}$, i.e. the time it takes for Alfvén waves to carry information from the center of the simulation box to its edges. This is indicative of a dynamic Gaussianization of $B$ in MHD turbulence, which requires MHD waves to facilitate this process due to causality in this physical system. However, for the hours and days long interval, the physical distance PSP and Ulysses trajectories covered are often degrees or tens of degrees apart in Carrington longitude (and latitude). Due to the lack of perpendicular information carrier, plasma that is degrees of Carrington longitude apart should not be considered as part of the same turbulent plasma parcel, and thus our simulation does not apply to such situation.

Therefore, alternative explanations are necessary for the observed hour-long (and longer) Gaussian structures. The simplest explanation for the Gaussian $B$ structures originating from coronal holes (mid-latitude coronal holes from E10 and E12, and polar coronal holes from Ulysses) is the pressure balance between the open coronal field lines. Close to the sun, the solar wind originating from the coronal holes is mostly magnetic dominant (plasma $\beta=2\mu_{0}P/B^{2}\ll 1$, see e.g. (Kasper et al., 2021)). To maintain pressure balance, the open field lines from the same coronal hole tend to evolve to a state in which the magnetic pressure $P_{B}=B^{2}/2\mu_{0}$ is mostly uniform for a given cross section of the magnetic flux tube. In Figure 1, the helio-radial power law normalization of $B$ essentially maps the magnetic field line density, which is effectively the magnetic flux density due to the spherical polarization of the Alfvén waves, from various radial distances and transverse locations to a single cross-section of the flux tube (for more details of spherical polarization of Alfvén waves, see the appendix and Matteini et al. (2014, 2015)). As a support of this idea, from the PSP observations of E10 and E12 (Figure 1 and Figure 10), the helio-radial power law normalization of $B$ effectively collapses the histogram of $B$ into a delta-function-like histogram of $B^{*}$. This is indicative of identical field line density within a single coronal hole due to the magnetic pressure balance. The detailed distribution of $B^{*}$ is hence the feature of the noise in magnetic magnitude within a single coronal hole, which can be considered as a one-dimensional random walk (continuous addition of small amplitude random fluctuations that can be considered as samples drawn from the same stochastic source throughout its passage from the base of the corona to the spacecraft). Therefore, the Gaussian distribution of $B^{*}$ can be easily explained as the result of the stopped random walk according to central limit theorem. Nevertheless, difficulties remain for the physical origin of the hour-long structures. They may be the manifestation of the denser field line density originating from a single supergranule based on its connection with switchback patches, but a more detailed discussion lies beyond the scope of this study.

Finally, the existence of a stable power law dependence of $B$ with regard to heliocentric distance $R$ itself already sheds light on the physics of the solar wind originating from coronal holes. As solar activities ramp up for solar cycle 25, 4 out of the 5 recent encounters (E10, E11, E12, E14) of PSP show systematic preference for a single helio-radial power law index, which consistently deviates from $R^{-2}$. However, the $R^{-2}$ power law, expected only from the dominant radial component $B_{r}$ as a result of the Parker Spiral (conservation of magnetic flux in spherical expansion), is not strictly applicable to $B$, especially for PSP, due to the ubiquitous switchbacks. Due to the relation between $B$ and the local magnetic flux density, this is indicative of a stable expansion rate for the magnetic flux tube in the magnetic dominant wind ($\beta\ll 1$) close to the sun. Such an expansion rate is crucial for the estimation of the WKB evolution of the fluctuation quantities like the magnetic and velocity field (Hollweg, 1973; Heinemann & Olbert, 1980; Velli et al., 1991; Huang et al., 2022). It should be noted that the fit indices of $B$ coincide with the helio-radial dependence of the electron density compiled from Quasi Thermal Noise (Kruparova et al., 2023; Moncuquet et al., 2020), indicating that the deviation from $R^{-2}$ could be the evidence of active acceleration of the solar wind.

Compiled from the almost featureless magnetic magnitude time series from the solar wind, the Gaussianity Scalogram (GS) unveils a striking number of fractal-like magnetic structures spanning across over seven orders of magnitude in time. These structures include spatial structures like polar coronal holes (McComas et al., 2003), mid-latitude coronal holes (Badman et al., 2023), and switchback patches (Bale et al., 2023). They also include temporal structures compatible with “jetlets” (Raouafi et al., 2023b) and “picoflare” (Chitta et al., 2023), which are often interrupted by the radial jets (switchbacks). In addition, three-dimensional MHD simulations have shown that Gaussian is the natural relaxation state for small amplitude unidirectional Alfvénic turbulence. The minute-long structures are hence likely to be the natural products of Alfvénic MHD turbulence. Thus, it is now clear that the Alfvénic solar wind is permeated with these intermittent Gaussian $B$ structures, which are self-similarly organized from seconds to years, and are possibly the remnants of the magnetic structures on the solar surface (Uritsky & Davila, 2012; Aschwanden, 2011; Aschwanden et al., 2016; Bale et al., 2023; Raouafi et al., 2023b; Chitta et al., 2023). This paper reveals just a fraction of the rich structures uncovered by the GS from the solar wind time series. The GS proves to be a versatile tool, essential not only for deciphering the structure and dynamics of plasma and magnetic fields, one of the key objectives of the PSP mission (Fox et al., 2016), but also for revitalizing decades-old solar wind data from missions like Helios, Ulysses, and WIND. These efforts unveil new physics previously hidden within these data sets. Additionally, the GS applicability may extend beyond solar wind analysis, potentially serving other kinds of high resolution stochastic time series data.

Z.H. thanks Jiace Sun, Benjamin Chandran, Anna Tenerani, Victor Réville for valuable discussions. C.S. is supported by NSF SHINE #2229566. C.S. acknowledges the support of Extreme Science and Engineering Discovery Environment (XSEDE) EXPANSE at SDSC through allocation No. TG-AST200031, which is funded by National Science Foundation grant number ACI-1548562 (J. Towns et al., 2014). C.S. also acknowledges the support of Derecho: HPE Cray EX System (https://doi.org/10.5065/qx9a-pg09) of Computational and Information Systems Laboratory (CISL), National Center for Atmospheric Research (NCAR) (allocation No. UCLA0063). This research was funded in part by the FIELDS experiment on the Parker Solar Probe spacecraft, designed and developed under NASA contract UCB #00010350/NASA NNN06AA01C, and the NASA Parker Solar Probe Observatory Scientist grant NASA NNX15AF34G, and NASA HTMS 80NSSC20K1275. M.V. acknowledges support from ISSI via the J. Geiss fellowship.