Acceleration of Coronal Mass Ejection Plasma in the Low Corona as Measured by the Citizen CATE Experiment

The CATE telescope was manufactured by Daystar Filters, LLC; and Daystar donated 60 telescopes to the project. The telescope is an 80 mm aperture refractor with an extremely low dispersion (ED) doublet primary lens. The focal length was 500 mm, producing an f/6.25 system. The optical tube assembly (OTA) included a 2'' rack and pinon Crayford focuser with rack and pinion assist and a locking screw. A 2'' to 125 converter and a Baader C-mount to 125 camera adapter were used to insert the CATE camera at prime focus. The OTA was mounted into the telescope mount with a dovetail foot plate. For observations during the partial phase of the eclipse, a full-aperture white light solar filter from Thousand Oaks (S4250) was used at all sites. Several of these filters were cross-calibrated with the same OTA/camera system.

The ED doublet was found to have some chromatic aberration. To minimize this, two color filters were introduced into the beam. A Yellow #12 and an IR Blocking filter (IRBF1) from High Point Scientific were purchased to limit the effective transmission between 480 and 680 nm. Both filters were 125 and were threaded into the camera adapter.

The CATE sites used German Equatorial mounts during the eclipse. In particular, a CG4 tripod mount from Celestron, Inc was used; Celestron donated 60 of these mounts to the CATE project. With the CATE OTA and camera system, only a small counterweight was needed to balance the system. On the CG4 a small Televue solar pointer was mounted at the location for the decl. drive, and the R.A. motor from the CG4 dual axis motor drive kit was installed on all the mounts to track the Sun during the eclipse. The R.A. motor was powered with 4 D-cell batteries. The R.A. motor control box allowed the CATE volunteers to make fine pointing adjustments in R.A., while a slow motion manual control was used to make pointing adjustments in decl.

In the years leading up to 2017, different cameras were examined to determine which would be an optimal choice for the CATE project. The camera FOV, pixel size and cost were considered. In 2016 the Point Grey Grasshopper USB3 5 Mpix monochrome camera model GS3-U3-51S5M-C was used at several sites in Indonesia. The detector, a Sony IMX250 CMOS array with a global reset was found to be superior to the array used in the 2015 Faroe Islands eclipse project (Penn et al. 2017). The camera was the most expensive element of the CATE instrument, and unfortunately no corporate sponsor was found to provide the camera. The CATE camera transferred data to a control laptop computer along a USB3 cable, and was also powered by the laptop using the USB3 port. Camera parameters including camera temperature were monitored, and camera communications including exposure control for the calibration images were sent along the USB3 line.

As in the 2016 Indonesian eclipse, a hardware exposure control signal was provided to the camera through the GPIO port for the observations taken during totality. A set of 8 exposures (0.4, 1.3, 4.0, 13, 40, 130, 400 and 1300 ms) was used during the totality sequences, although at many sites the whole camera FOV was saturated in the 1300 ms exposure.

In addition to providing a hardware exposure trigger, the CATE Arduino packages also contained a Global Positioning System (GPS) timing board from ITEAD Royal Tek, model number IM120417017, along with a GPS antenna from Banana Robotics, model number BR010312. The GPS board was mounted in the Arduino Box and connected to the Arduino Uno Rev 3 CPU card. The MathWorks MATLAB SolarEclipseApp software package communicated with the GPS board and data strings received from the board were relayed to the control laptop via the USB2 cable. During the calibration procedures, the Arduino passed GPS strings from this board to the CATE laptop for two purposes: (1) the GPS location of the observing site was recorded, and (2) the GPS timing information was compared with the laptop timing so that the start time of each exposure from all the CATE sites could be converted to a uniform GPS time.

The Arduino operated on power from the CATE laptop using a USB2 cable. The GPIO cable interface to the Arduino digital output pins was custom made for all the CATE sites, and the Arduino packages were installed into modified Arduino project boxes.

The CATE laptops were required to run the operation software, power the CATE camera and Arduino, and store and backup the CATE data, while operating on only battery power if necessary. At least one USB3 port was needed to communicate with the CATE camera. Enough RAM was needed to support the data collection software and enough processing power was required to provide some near real-time display capability while the camera data was being written to disk. A backlit keyboard was preferred to help users needing to type during the dark conditions at totality. The disk write access speed needed to be fast enough so as to not limit the data collection, and finally all of this needed to cost very little, since unfortunately no corporate sponsors were found to supply laptops to the project.

A laptop from Acer (Aspire E15 E5-575G-527J) was found to satisfy most of these requirements. It included a dual-core Intel i5 processor running at 3.1 GHz, operated at 64 bits, and included a solid state drive which had data write speeds that were faster than other laptops without solid state drives. The laptops were augmented with a second 8 Gb RAM card (Ballistix BLS8G4S240FSD) bringing the total laptop RAM up to 16 GB. The laptops used Windows 10 operating system.

The CATE calibration and data collection procedures were written in MATLAB, by Mathworks, Inc. Mathworks was a CATE corporate sponsor and donated software, funds and licenses for CATE volunteers; additionally, three CATE volunteer observers were Mathworks employees.

Following the 2016 Indonesian eclipse, a brainstorming session was held where the software used during the 2016 eclipse was evaluated, and ideas for a better software package were developed. Key issues were identified, including the difficulty of using the Indonesian GPS package and the focus problems at several Indonesian sites, and particular strategies were developed to minimize these problems for the 2017 eclipse.

The 2017 MATLAB software package presented each volunteer with a graphical user interface (GUI) on which several tabs could be selected. These tabs stepped sequentially through the GPS Data collection, Instrument Setup, Alignment, Focus, Calibration and Totality observations procedures. The GUI enabled multi-threading to utilize both laptop processing cores. A more complete description of each of these steps is presented in Section 4.1 below.

In addition to these main components, the equipment at each CATE site included a few miscellaneous components. During several tests in the summer in Tucson, when the telescope was taking sky flats and the camera body was exposed to direct sunlight, the camera temperature reached 76°C. Communication between the camera and the laptop was lost as the camera overheated, exceeding its 50°C maximum recommended operating temperature. A small battery-powered cooling fan was mounted on the CG4 tripod blowing air across the camera body. More than 10°C of cooling was achieved this way and allowed stable camera operation.

For 2017 eclipse observing locations lacking a power source, the CATE equipment was designed to run for 3 hr on laptop batteries, however several volunteers used external battery power sources to avoid any trouble caused by low battery power on the CATE laptop.

Finally, in order to backup the files immediately after totality, each site was equipped with two USB memory sticks with at least 16 GB capacity each. Two backup copies of the calibration and science data were made directly after totality at each site using these sticks.

A set of coordinators was identified, one from each state crossed by the 2017 path of totality. Faculty mentors from the universities participating in the 2016 eclipse formed the core of this group of state coordinators and organized efforts in Wyoming, Illinois, Kentucky and South Carolina. To this core group individuals from Oregon, Idaho, Nebraska, Missouri, and Tennessee were added to build a group of nine CATE state coordinators. These coordinators included additional university faculty, high school teachers and amateur astronomers, and they participated voluntarily in the program. Frequent meetings of the state coordinators were held with teleconferencing, and these state coordinators then met with the CATE site volunteers via phone, teleconference or in-person. The state coordinators and trainers provided input into the CATE equipment selection, software control and data analysis.

Initially a set of observing locations for the 2017 eclipse was determined using an equally spaced set of locations across the path of totality from the Pacific to Atlantic coasts. Google maps was used to identify school or park locations with easy access that were close to the centerline at the desired spacing. During a second round of analysis the local times for the eclipse contacts were used to determine if the site locations needed to be adjusted. These second-level sites were then presented as suggestions to the CATE state coordinators. The coordinators used their knowledge of the locations, local weather patterns, and the experience of their site volunteers to refine the CATE observing sites a third time. With this optimal set of locations, the CATE coordinators and site volunteers then attempted to get permission to use these sites during the day of the eclipse; some of the sites were relocated a fourth and final time depending on the ability of the volunteers to observe from the area. Using this iterative procedure, the CATE observing sites were established using local knowledge to maximize the likelihood of obtaining excellent eclipse observations.

Solar and lunar observation procedures were designed in order to give the CATE 2017 volunteers a basic understanding of equipment and an introduction to the procedure at workshops held in several states across the path of totality. The procedure served as a guide for all practice sessions and ultimately the observations during eclipse. Procedures were split into five sections: Setup, Preliminary Data, Eclipse, Data Backup and Quick look. Throughout the procedures, the amount of time allotted for each task was listed as a guideline to keep volunteers on schedule so they were prepared to take totality data at the time of the eclipse. With perfect sky conditions and no instrument problems, the procedure took approximately 2.5 hr to perform.

4.1.1. Setup

4.1.2. GPS

4.1.3. Alignment

4.1.4. Focus

4.1.5. Calibration

4.1.6. Totality

CATE coordinators hosted twelve training workshops from April 22 through May 28. At the workshops the CATE equipment was distributed to the site volunteers. State coordinators, teaming with a CATE student, guided the site volunteers through the solar observing sequence. In some states with widely distributed volunteers, two workshops were held. TN and KY combined their workshops into one location, and a workshop was held in Tucson AZ for several groups from the southwest. Overall workshop participation was good with 130+ participants attending, representing 62 out of the 70 observation sites.

The workshops shared a set of several goals, and emphasized the safety issues involved in observing the Sun. Many of the Citizen CATE sites were in public areas where the citizen CATE teams would be the eclipse guides for the public. Teams were instructed on safe public solar viewing as well as how to develop their site-specific safety plan. Teams were encouraged to identify members who would work with the public exclusively so the other team members could concentrate on running the telescope. Teams worked through the solar procedure first indoors on a test image, and then outdoors on the Sun. Test data collected at the workshops were uploaded to a common website in the same way it would be done in summer practices and during the August eclipse.

Due to the time-critical nature of the eclipse observations, it was important that teams practice and become proficient using the equipment. The goal of the workshops was to familiarize teams with the equipment and get them to the point that they could take part in practice observation campaigns as well as practice on their own. Volunteers were then asked to begin day and night time observations and upload data of the observations for evaluation. Participating in both day and night time observations prepared teams by imaging the full disk of the Sun during conditions similar to partial eclipse phases, and imaging the moon, which provided a target of similar brightness to the solar corona.

In addition to independent practice, several network-wide training dates were established. The main purpose of these practice runs was for the teams to learn how to complete observations in a timely manner, often in less than ideal conditions. Practices also revealed issues with the procedures and allowed the CATE team to provide feedback to volunteers on the quality of their data.

Table 2 lists several parameters measured during the network-wide summer training dates. Team participation in training sessions was at or above 50% for the first three training sessions and decreased during the last two sessions. The average focus quality improved after the first two training sessions and then decreased slightly again. The polar alignment for the sites, as measured by the solar image drift while the telescope was tracking, improved by a factor of two through the summer. The camera rotation, as measured by the angle of the solar image drift during the tracking off sequence (270° is ideal), improved initially and remained good. Finally, during the last three practice sessions, the ability of volunteers to center the solar image was measured, and this also showed improvement with time.

Significant improvements to the solar eclipse app were also made at this time. Required updates to Arduino code and the app were pushed out to state coordinators who beta tested the changes and quickly provided feedback to the core team. Once the code was determined to be stable, new versions were made available to all volunteers. Since the workshops included instruction on uploading Arduino code to the microcontrollers as well as basic computer configuration and app installation, the majority of the teams were able to quickly adapt the changes into their own computers and start using new procedures.

The GPS board received timing and location information. On the "GPS" tab of the SolarEclipseApp software, the CATE volunteers selected a communications channel and viewed the GPS strings in a live display. Each site was instructed to run a GPS data acquisition routine during their pre-eclipse activities. About 1300 GPS strings were collected and saved to the file during roughly 120 s of time for each site. These text files were analyzed to determine the site latitude and longitude, and to determine the timing offsets between GPS time and each laptop time.

5.1.1. Site Location

5.1.2. Site Timing

CATE site volunteers experienced a variety of weather conditions along the path of totality. Conditions were mostly clear on the west coast, and the skies remained clear, except for a few places, along the eclipse path until eastern Nebraska. A patch of clouds was found from eastern Nebraska through western Missouri, and then mostly clear conditions prevailed again until Illinois and western Kentucky. Partly cloudy conditions challenged CATE observers along the path of totality from Kentucky to the east coast.

The last row of Table 2 shows the focus, polar alignment, camera rotation, and image alignment specifications for CATE observers on the eclipse day, August 21. In order to distinguish cloudy sites from clear sites, the image intensity of the tracking-on and tracking-off time sequences was used. (It is likely that there is useful data from the sites with lower or highly variable image intensity, but those have not yet been examined.) Stable image sequences with the CG4 mount tracking the Sun were collected from 46 sites, and good image sequences with the CG4 mount turned off (and the solar image drifting across the camera) were collected from 50 CATE sites.

The image focus values obtained on the eclipse day matched the best focus values from any practice session. The quality of the mount's polar alignment is measured by the amount of solar image drift during 100 s of tracking. To ensure minimal image smearing, a tracking drift of less than 1 pixel in 1.3 s was the goal for the polar alignment error (0.77 pixels per second). Through all of the practice sessions listed in Table 2 this goal was met, and the eclipse day drift value (0.22 pixels per second) easily met the goal. The camera angle adjustment on average was the best recorded during the summer, although five sites were rotated 180° away from the optimal orientation. Finally, the image centering achieved by the CATE volunteers on eclipse day was good.

Standard dark and flat images were taken during the pre-totality operations at each CATE site. The dark images for the totality sequences were taken using the same Arduino hardware camera triggering that was used during totality; darks were taken in a sequence of 8 exposure times, consisting of 0.4, 1.3, 4.0, 13, 40, 130, 400 and 1300 milliseconds. A sequence of 100 darks was saved to the laptop disk drive, consisting of 12 complete 8 exposure sequences and four additional frames.

A close examination of the hardware triggered exposures (the totality images and the totality darks) shows some evidence of random intensity jumps. These do not seem to be sticky bits, but seem to be deviations in the exposure time caused by the timing signals from the Arduino. In the dark frames where the signal level is about 660 counts, the jumps can be about 30 counts, and only affect one frame in the sequence of 12. This suggests these trigger errors can cause intensity errors of about 1.3%. In the totality images the signals are generally 7000 counts or higher, and about 600 images are co-added to produce an HDR frame; we estimate these trigger errors can introduce intensity errors of 0.017%.

The flat field exposures and the associated darks for the flats were triggered with a software timing trigger, and neither of these images show the random variations seen in the other calibration data.

Alignment of the CATE data is challenging. Among the totality exposures frames from one site, motion occurs as (1) the telescope does not exactly track the Sun, (2) the moon moves against the solar corona and (3) the Sun moves against the background R.A./decl. coordinate frame. The fourth alignment concern is that the different CATE sites have different image rotation. Since the CATE sites are scattered around the center line of totality, the lunar disk eclipses the solar disk in different ways, leading to a different spatial distribution of atmospheric scattered light. Finally the atmospheric conditions vary from site to site, further complicating stray light contamination.

The general procedure which was used to align this initial science data set was to compute lunar disk position to align each group of 8 sub-exposures and to produce one HDR frame. All of these HDR images from one site were interpolated to a common image center. Solar features were then tracked to determine a second set of offsets, and then all of the HDR images from one site were aligned and summed. Angular cross-correlation functions were computed using an annular region of a spatially filtered version of this HDR image to determine the rotational offsets between each site.

In this set of 6 images, two stars were seen, and were used to further refine the angular and spatial offset alignment, and to orient each image onto an R.A./decl. grid. The two alignment stars (HD 86898 and GM Leo) were also used to measure the image scale from these sites. The stars are 29762 apart, and this produces an image scale of 143 per pixels. The scale variations among these six sites is less than 001 per pixel, attesting to the excellent uniformity in the CATE optics system; solar disk measurements from all the CATE sites also point to uniformity at this level. Unfortunately the stars are not visible in the individual CATE exposures or the first-round HDR images.

After image alignment, two types of spatial filtering were applied to the CATE images to emphasize the low contrast coronal structures. First a normalized radial gradient filter (Morgan et al. 2006) was used to remove the steep radial intensity gradient from the coronal images. The radial gradient in raw camera counts from the CATE sites shows a brightness variation of a factor of 300 from the limb of the moon to the edge of the FOV. While stray light, atmospheric transmission and equipment variation between these two sites is expected to change the measured coronal intensity, the signals compare very favorably among sites that have clear skies.

The second filter was a combination of box-car averaging and then a simple edge-enhancement Sobel filter. This filtering enhanced coronal fine-scale structures to facilitate feature tracking in the time series. Sample NRGF and Sobel filtered images are shown in Figure 1.

Zoom In Zoom Out Reset image size

A time sequence of the filtered images reveals a variety of changing coronal features. The most apparent changes occur in along the SE solar limb where material is moving outward through the solar corona. This location corresponds to a CME observed by LASCO SoHO instrument. LASCO height-time plots show that the event had an outflow velocity of 254 km s⁻¹ at the observed heights between 3 and 13 RSun, with an extrapolated start time of 13:55 UT and nearly no acceleration, (−0.2 m s⁻²) (CDAW Data Center 2019).

The CATE images were divided into sub-images, and spatial cross-correlation was used to determine local image shifts between sequential image pairs. The shift values were divided by the time interval between the images and the spatial scale was used to compute a velocity; both radial and tangential velocity components were computed. Figure 2 shows a grayscale, spatially filtered image of this region of the corona with the computed velocity vectors overlaid on the grayscale. The spatial ROI showing the CME outflow was determined to lie between two prominent and stationary coronal threads. The velocities measured in the CME ROI were then averaged in radial bins of 0.1 RSun; the standard deviation of the mean was also computed for each radial bin.

Zoom In Zoom Out Reset image size

The average radial velocity versus radial distance is shown in Figure 3. Individual points are the radially averaged outflow speeds, the error bars are the standard deviation of the mean for each bin, and a linear fit is shown by the dotted line. The outflow speed measurements from LASCO time–height analysis of the CME leading edge are shown as the horizontal line starting at 3.1 RSun at 254 km s⁻¹ outflow speed.

Zoom In Zoom Out Reset image size

There are two fits to the CATE data which are both extrapolated to a height of 3.1 RSun. First, the dotted line shows a model with a constant acceleration of the CME material; the best fit acceleration is 15 m s⁻². The dashed line shows a constant velocity increase with increasing height intervals of 175 km s⁻¹ RSun⁻¹; this represents a changing acceleration with height. Both models assume that the radial velocity equals 0 at a height of 1.2 solar radii. Both models eventually exceed the outflow speed measured by LASCO data, and so the acceleration mechanism must stop or significantly decrease below a height of 3.1 RSun. Using the constant acceleration model, the CME material in the CATE FOV at 2.1 RSun would enter the LASCO C2 FOV at about 19UT; using the other acceleration model the material would enter the C2 FOV slightly earlier. In the C2 movie, CME material continues to stream out of the Sun until after 22UT, so the plasma observed by CATE is certainly associated with the event observed by LASCO.

This analysis of CATE data measures apparent CME plasma velocity by mapping the spatial shifts of many plasma features in the images and averaging these shifts in radial bins. It provides a more flexible method for computing the plasma acceleration than other studies, since only the first derivative is needed, and the velocity (and acceleration) can be mapped across many spatial positions and times. All previous CME kinetic studies measure the changing position of one or two CME features with time, compute the velocity by taking the first derivative of the position, and then compute the acceleration by taking the second derivative of the position.

The CATE data measures the white light corona rather than particular emission lines. Coronal plasma at all temperatures is equally sampled in this data. In the EUV work of Reva et al. (2017) which used TESIS Fe 171 and EIT He 304 images, different parts of the CME that were studied had different temperatures and thus different visibilities. Furthermore, the temperature of the CME regions may change with time, and so tracking features in a sequence of EUV images involves more assumptions than tracking features in a sequence of white light images. With those assumptions, Reva et al. (2017) track two features, the leading and central regions of a CME, and find different velocities and accelerations for those regions. The leading CME region shows a terminal outflow speed of about 400 km s⁻¹, and the central region moves outward more slowly at about 250 km s⁻¹. They attribute the observed kinetics, and the temporal changes of the kinetics to a variety of mechanisms including reconnection, mass drainage and magnetic instabilities.

Like the work of Reva et al. (2017), we measure the CME acceleration not just at the leading edge, but at many locations. The CME in this CATE data has a terminal outflow speed similar to the central region of the Reva et al. CME; they find acceleration values of between −20 and +35 m s⁻², similar to our observed accelerations of around 15 m s⁻². But unlike that work, we find that the velocity and acceleration of the plasma well behind the CME front is consistent with the leading edge velocity measured by LASCO (which is also measured with white light data). While magnetic reconnection is often discussed as an acceleration mechanism associated with the initiation and leading edge of a CME (Chen 2011), the fact that the trailing plasma observed in this CATE data shows an outflow speed similar to the CME front may provide support for other acceleration mechanisms in this CME, such as magnetic instabilities. This CME may simply exhibit different characteristics from the one studied by Reva et al. (2017), or the assumptions used in the EUV analysis to track features may lead to incorrect velocity measurements for different parts of the CME.

Observations of the solar corona were taken with EUV instruments on the day of the eclipse, for instance with the Proba2 mission and with SUVI. Examining the difference movies from Proba2 at the time of the CATE observations (http://proba2.sidc.be/swap/data/mpg/movies/2017/08/) shows only noise or no signal above a height of about 1.1 solar radii. While the raw Proba2 images (which are not aligned) do show some structures at higher locations, a full analysis of that data set is beyond the scope of this paper. Similarly, SUVI data from the eclipse has not been reduced with the standard pipeline process, and a full analysis of those measurements is again beyond the scope of this paper. Future work comparing the CATE data with these EUV measurements will likely be fruitful, and future analysis of the full CATE data set from all sites will likely reveal more flows in different regions of the solar corona.