Statistical Study of ICMEs with Low Mean Carbon Charge State Plasmas Detected from 1998 to 2011

We present a statistical study of 219 ICMEs measured by both ACE and WIND from 1998 to 2011. ICME plasmas are defined as possessing cold materials if the carbon average charge states are lower than those of the preceding solar wind by three standard deviations and the carbon ionic temperature is lower than 106.05 K. A total of 69 ICMEs were identified as containing cold materials. These ICMEs tend to have speeds in the range of 300–600 km s−1, with durations between 2 and 6 hr. Cold materials tend to be present once or twice per ICME. We further identify two special types of cold materials: the ionic-cold type (IC) shows simultaneous lower average charge states of O, Mg, Si, and Fe ions than those of the preceding solar wind, while the carbon-only cold type (COC) shows a totally opposite trend in that these ions show higher average charge than in the preceding solar wind. We found that the IC has a higher proton temperature than the ICME mean value, whereas the COC has a lower proton temperature than the ICME mean value, and the COC is most often measured in magnetic cloud. A detailed examination of the IC and the COC material suggests that they are related to solar filaments. Their special mean charge indicates that the filaments are a mixture of coronal and choromospheric materials. Heating and collision processes beyond the carbon freeze-in height are crucial in generating the two distinct types.


Introduction
Coronal mass ejections (CMEs) are some of the most severe explosive phenomena in the heliosphere. They have been routinely observed by coronagraphs since the 1970s (Hundhausen et al. 1984;Schwenn et al. 2006). It is well known that halo CMEs can cause geomagnetic storms. Typically, CMEs have a three-part structure: the bright core, the dark cavity, and the bright front. The bright core is supposed to be the eruptive filament, the dark cavity is the flux rope, and the bright front is where the plasmas pile up (Hudson et al. 2006;Wimmer-Schweingruber et al. 2006;Reeves et al. 2015). When CMEs propagate into the interplanetary space, their counterparts measured in situ are called interplanetary coronal mass ejections (ICMEs; Zurbuchen & Richardson 2006). If ICMEs have enhanced magnetic field strength and smoothly rotating magnetic field, they are classified as magnetic clouds (MCs; Burlaga et al. 1981).
In the past 40 years, researchers have attempted to learn more about plasma features and the acceleration process of CMEs from in situ measurements (Cane & Richardson 2003;Richardson & Cane 2010;Howard 2015aHoward , 2015bWood et al. 2016). Many works reported the identification of the plasma pileup and flux rope structures. However, cold and dense filaments have been seldom reported from in situ measurements. Usually, it is believed that the solar filament materials come from the chromosphere (Forbes 2000;Yan et al. 2013Yan et al. , 2016. Such materials should show a lower proton temperature, higher proton density, a higher alpha to proton ratio, and lower charge state ions (Priest 1988). According to these criteria, Schwenn et al. (1980), Wurz et al. (1998), Burlaga et al. (1998), Gloeckler et al. (1999), Gopalswamy et al. (1998), Lepri & Zurbuchen (2010), Yao et al. (2010), and Song et al. (2016) identified over 10 ICME events from 1976 to 2009 that likely contained filament materials. Clearly, this number is far less than the number of observed three-part CMEs (Gopalswamy 2006;McCauley et al. 2015). Recently, Song et al. (2016) and Song et al. (2017) analyzed the average ion charge state of Fe inside the ICMEs and confirmed that some cold materials in the ICMEs may not be related to the prominence in the solar atmosphere. A natural problem that arises from these studies concerns what happens to the prominence/ filament during the initiation of the CMEs and what is the nature of the cold material in the ICMEs measured in situ. As an attempt to solve these problems, we perform a statistical analysis of all the ICMEs measured by both the WIND and ACE spacecraft from 1998 to 2011 with a focus on cold materials. By combining the ion average charge of C, O, Mg, Si, and Fe; proton and α particle moments; and solar observations; we examine the features of the cold materials in the ICMEs. The data and the criteria used to identify the cold materials are described in Section 2. The statistical results of the cold materials in the ICMEs are shown in Section 3. Constraints on possible physical processes as revealed by the presence of these cold materials are discussed in Section 4.

Data and Criteria
We used the magnetic field, proton temperature, proton speed, and ion charge state distributions measured by the ACE spacecraft from 1998 to 2011. The magnetic field magnitude and its components in the Geocentric Solar Ecliptic (GSE) coordinate system are provided by the MFI every 16s (Smith et al. 1998). The normalized ion charge distributions of C, O, Mg, Si, and Fe are published with a cadence of one hour by SWICS (Gloeckler et al. 1998), and the plasma moments are provided with a cadence of 64s by SWEPAM(McComas et al. 1998). All the ion charge state data is provided by the ACE Science Center. To minimize the effect of instrument errors and to help better identify the MC, this work also uses WIND data, including the 3s magnetic field magnitude provided by MFI (Lepping et al. 1995), the 92s cadence proton density N p and α particle density N α provided by SWE (Ogilvie et al. 1995) and the 3s cadence proton temperature T p provided by 3DP (Lin et al. 1995). The CME images observed by LASCO and UVCS on board SOHO (Dere et al. 1997) (Aschwanden et al. 2009).
The ICME list used in this work is given by Cane & Richardson (2003); Richardson & Cane (2010), and by the ACE Science Centerhttp://www.srl.caltech.edu/ACE/ASC/ DATA/level3. Of all 319 ICME events from 1998 February to 2011 August, when the SWICS data is available, 219 events were detected by both the WIND and ACE spacecraft, together with the 24 hr of solar wind before them. For this analysis, we required data gaps in the ionic charge states to be shorter than 12 hr in the preceding solar wind and 2 hr in ICME periods, as well as shorter than 6 times the sampling cadence for other plasma parameters. The detailed ICME event list is shown in Table 6 in the Appendix. All the times of the shock and ICME leading and trailing edges are shown with respect to ACE time. The shock time is used to determine the 24 hr solar wind preceding the ICME. The mean value of all the statistics is calculated from the ICME leading edge to trailing edge. To examine whether the ICMEs are MCs, we used the criterion given by Burlaga et al. (1981). Specifically, an ICME is identified as an MC if the magnetic field magnitude is enhanced stably and the magnetic field vector rotates smoothly for a large angle. To avoid subjective judgment, we quantified the criterion as follows: the maximum magnetic field magnitude of the ICME is stronger than that of the ICME sheath, the elevation angle of the magnetic field changes by at least 50°o ver more than 8 hr, and the fluctuations of the magnetic field elevation angle θ are less than 50% of the θ max −θ min in the ICME.
In all analyzed 219 ICME events, there are 99 MCs. Since we focused on the nature of the cold materials in the ICMEs, we analyzed not only the ion charge state but also the proton temperature T p , the proton density N p , the proton velocity V p , the alpha particle density N α , and the alpha to proton density ratio, R N N p p a a  , in the ICMEs and in the preceding 24 hr of solar wind.
The carbon ions have a freezing in height between 1 and 1.2R s , which is lower than that of the O, Mg, Si, and Fe ions (Geiss et al. 1995;Kocher et al. 2017). Consequently, the charge of the carbon ions is the least affected by the heating and collisions associated with the CME initiation. We used the following two criteria to identify cold material in ICMEs.
(1) The mean carbon charge state in the carbon-cold (CC) period is at least three standard deviations lower than that of the 24 hr solar wind preceding the ICME. The standard deviation is calculated in the preceding 24 hr of solar wind.
(2) The carbon ionic temperature satisfies T T 10 C c r i t 6.05 < = K. This critical temperature was derived by binning all preceding solar wind periods into six velocity bins with 50km s −1 bin widths centered around 325-575km s −1 and one bin for v>600 km s −1 . We then calculated the expected mean carbon ionic temperature according to the statistical relation between solar wind speed and carbon freeze-in temperature given in Kasper et al. (2012). For the solar wind periods with speeds exceeding 600 km s −1 , we used T C =10 5.988 K as given by Landi et al. (2014) who analyzed fast (v > 500 km s −1 ) solar wind measured by Ulysses from 1994 to 1998 (Gloeckler & Geiss 2007). We computed the weighted average of these temperatures using the number of occurrences per velocity bin as weights and found T crit =10 6.05 K. The ionic temperatures were derived from their mean charge state (as provided by the ACE Science Center) using the CHIANTI database version 8 (Dere et al. 1997;Del Zanna et al. 2015). Because ICMEs can follow each other with less than 24 hr separation, we used the 24 hr of solar wind preceding the first ICME if multiple ICMEs occur. The criterion can be expressed as Q Q C C , S W á ñ < á ñ -3 Q C,SW s and T C <10 6.05 K. In addition, this ionic temperature corresponds to a mean carbon charge of 5.1336. The symbol σ is the charge state standard deviation of the 24 hr solar wind preceding the ICME. To avoid errors and uncertainties from charge state measurements, a single data point meeting the criteria was not considered in this work. Consequently, we considered a time period containing at least two SWICS data points, which is at least one hour long in between.

Statistical Results
We identified 108 segments in 69 ICMEs that contain CC material. Of the 69 ICMEs, 38 are MCs. We also examined the mean charge states of the O, Mg, Si, and Fe ions in the time periods containing cold material. The ionic temperature of C, O, Mg, Si, and Fe ions of the 108 CC segments are shown in Table 7 in the Appendix. It is very interesting that there are two extreme cases. The first one has mean charge states of all the O, Mg, Si, and Fe ions that are lower than those in the preceding 24 hr of solar wind. The other one has mean charge states of all the O, Mg, Si, and Fe ions that are 3σ higher than those in the preceding 24 hr of solar wind. We named the first case ioniccold (IC) material and the other case carbon-only cold (COC) material. Among all the 108 CC segments, we classified 6 segments as IC and 15 segments as COC. An ICME containing IC material is shown as an example in Figure 1. From the figure, we can see that the carbon mean charge state falls below the Q 3 Q C,SW C,SW s á ñ at approximately 05:30UT on 2006 April 14. The low mean charge state lasts for approximately 7hr and is marked by the blue transparent stripe. In the low carbon charge state period, Q O á ñ , Q Mg á ñ , Q Si á ñ , and Q Fe á ñ are lower than their mean values in the preceding 24 hr of solar wind. Furthermore, the proton temperature measured by both the ACE and WIND is higher than the mean value of the ICME. Of the six IC segments, five have a higher proton temperature, similar to that shown in Figure 1.
In Table 1, the six IC segments from five ICMEs are shown with their start and end times, duration, MC or not, and the mean charge state of the C, O, Mg, Si, and Fe ions in the CC segments and in the preceding solar wind. Only one IC period was observed to be inside an MC.
One ICME event containing COC material is shown in Figure 2. From the figure, we can see that the mean carbon charge falls below the Q 3 Q C,SW C, SW s á ñ at approximately 07:00UT on 2000 November 7. The carbon low charge state lasts for approximately 2hr. During this time, the mean charge states of the O, Mg, Si, and Fe ions are all above their solar wind mean charge state plus three standard deviations of the preceding 24 hr solar wind. Meanwhile, the mean proton temperature measured by ACE and WIND is lower than that of the CC ICMEs. All of the 15 COC segments except one have a mean T p lower than that of the ICME. All 15 COC segments Figure 1. One ICME event containing the IC material. The top nine panels show the magnetic field magnitude, magnetic field components in GSE coordinates, proton speed, proton temperature, The remaining four panels display the magnetic field magnitude, α particle density, proton density, and proton temperature measured by WIND. The vertical dotted line denotes the shock in front of the ICME. The vertical dashed line denotes the leading edge of the ICME plasmas. The vertical black solid line denotes the end of the ICME. The horizontal black dotted lines denote the mean charge of ions in the 24 hr of preceding solar wind. In the fifth panel, the horizontal blue dotted line denotes the mean charge of carbon minus three standard deviations. The cold material segment is indicated by the shaded blue stripe. from 13 ICMEs are listed in Table 2. It is remarkable that 13 of the 15 COC time periods are identified inside 11 MCs.
Next, we examined various statistics of the 108 CC segments in Figure 3. The histograms of the segments numbers per ICME, of the duration of the segment, and of the proton speed in each segment are plotted. From panel (a), we can see from the figure that most of ICMEs contain one or two cold segments. Only a few of them have more than two segments. The largest number of segments in one ICME is five. Panel (b) shows the distribution of the duration of segments. The most probable value of the duration is 2hr and most periods are shorter than 10 hr. Nevertheless, there are a few periods that are as long as 37 hr. Panel (c) shows that the bulk solar wind speed in the CC segments lies predominantly between 300 and 600km s −1 , although there are a few faster segments.
Furthermore, we plotted the ratio of O, Mg, Si, and Fe mean charge states in the CC segments and the preceding 24 hr of solar wind versus their corresponding mean carbon charge states in Figure 4. It is by construction that all the IC segments have mean charge state ratios Q Q 1 CC SW < , and the COC segments have Q Q 1 CC SW > for all investigated elements: O, Mg, Si, and Fe. The two types are located on the top part and bottom part of the distribution, respectively. The IC segments tend to be carbon colder, but the difference between IC and COC segments may be mainly due to the outlier at Q 4.48 C » .
Thus, we performed a significance test on our classification of IC and COC type in the CC class. As we introduced previously, IC type was defined by its lower carbon mean charge and lower mean charge of O, Mg, Si, and Fe than the preceding solar wind, while COC type was defined by its lower carbon mean charge and higher mean charge of O, Mg, Si, and Fe. The tests are on the carbon ionic temperature and are performed to the three types of period: (1) CC, (2) IC, and (3) COC. Note that the CC type is the mother type, and IC type and COC type are two daughter types. Our null hypothesis is that the two distributions are the same. The test results are shown in Figure 5. The p values are 0.11, 0.43, and 0.06 for the three pairs: CC and IC, CC and COC, and IC and COC, respectively. Since we chose a significance level of 10%, the pvalue less than 0.1 means the null hypothesis can be rejected at the 10% significance level. Clearly, the two daughter types are similar to their mother type (as seen by the p value of 0.11 and 0.43). However, the two daughter types, having a p value of 0.06, are very different. Indeed, as shown in Figure 5, the distributions of IC and COC cover different ionic temperature range and have little in common. These may suggest that these two types of cold material have different forming mechanisms.
Moreover, we investigated the relation between the ratios of average O, Mg, Si, and Fe average charge states to C average charge state and the bulk solar wind speed measured during CC  To further understand the nature of cold materials, we calculated the CME launch time T early and T late from the in situ measured ICME maximum and minimum velocities. The velocity information is given in the first appendix table. Considering the velocity variation during the CME initiation, we defined the CME start time range as [T early −12 hr, T late +12 hr]. According to this CME start time range, we checked the SOHO and STEREO CME lists. We identified all the CMEs having an angular width of larger than 120°as Halo CMEs because Shen et al. (2014) reported that these CMEs have greater probability of hitting Earth and thus ACE and WIND than others. If there is more than one Halo CME in the start time range, we chose the Halo CME having the largest angular width. We identified the core part of the  Table 3, no data (ND) means no solar observations from SOHO/LASCO, and no report (NR) means SOHO LASCO and UVCS do not report such phenomenon though the data are available.
To reveal more features of the cold materials and to compare them with the filaments, we studied various plasma parameters of the cold materials. Thus, Table 3 also includes the T p , N p , N α , and R p a (N α /N p ) for both IC and COC segments compared with their mean value of each ICME. The mean values of T p , N p , N α , and R p a in CC period and those in ICME are given in Table 8 in the Appendix. The comparison results of the three classes: CC, IC, and COC, are summarized in Table 4. It could be seen that the proton temperature for the IC segments and the COC segments have almost opposite trends. Furthermore, all cold materials tend to have higher R ; p a and except IC segments, they tend to have lower proton temperatures as well.
Previously, Lepri & Zurbuchen (2010) reported the existence of cold filament materials in ICMEs. To define the cold filament materials, they used a combination of the charge states of the C, O, and Fe ions. Specifically, they defined the following: if C 2 ,3 + + represents more than 3.4% of all the measured carbon ions, O 4+ represents more than 0.9% of all the measured oxygen ions, and Fe 4+,5+,6+,7+ represents more than 12.1% of all the measured iron ions, then the Figure 3. Distribution of the 108 carbon-cold segments in terms of (a) the number of segments per ICME, (b) the segment duration, and (c) the proton speed. The vertical axis in all three panels is the number of counts of the cold segments. In the top panel, the horizontal axis denotes the number of segments per ICME event. In the middle panel, the horizontal axis denotes the segment duration. In the bottom panel, the horizontal axis denotes the proton speed of each segment. corresponding plasmas are considered predominantly cold materials. In our study, we employed a different criterion for the identification of cold materials. Since the ICME events that we used in this study significantly overlap with Lepri & Zurbuchen (2010), we can compare our results with those studied by Lepri & Zurbuchen (2010). This comparison is presented in Table 5. A total of 11 cold filaments in 10 ICMEs were identified in Lepri & Zurbuchen (2010). The serial numbers of the 11 periods in Lepri & Zurbuchen (2010) and in our study are given in Table 5 as SN(L) and SN. Out of the 11 (Lepri & Zurbuchen 2010) cold material periods, 8 meet our definition of CC materials. Note that 3 of Lepri & Zurbuchen (2010) periods contain only a single data point for the carbon mean charge meeting the requirement on cold materials, marked by S. We noticed that 1 out of the 8 CC periods contains COC materials, and no periods contain IC materials. Note that the ICMEs reported by Lepri & Zurbuchen (2010) containing SN(L) 6, SN(L) 7, and SN(L) 9 cold material periods are not in our 219 ICME list, and are marked by "-" in Table 5. This is because our selection criteria require simultaneous measurements of ICMEs by both ACE and WIND as well as measurement of the solar wind within 24 hr prior to the ICMEs. For the ICMEs in Table 5, although the ICME events identified in both studies are the same, the duration of the segments differs, as they are identified by the two different criteria, as explained above. In addition, there are CMEs and filaments related to the SN(L) 4 and SN(L) 5 cold material periods in Lepri & Zurbuchen (2010). But in our study we found a Halo CME related to SN(L) 4 and no Halo CME related to SN(L) 5. Since we defined the range of CME start time using the in situ measured speed, we only checked the CME record in the defined time range. Beside, there is NR of eruptive prominence in the CME related to SN(L) 4 in our study.
Finally, we examined the solar cycle dependence of these cold material appearances. The results are shown in Figure 7. ICMEs are often classified into MCs and nonmagnetic-cloud ICMEs (non-MC). In our study, depending on if there are cold materials present, we further classified MC and non-MC into cold material ICMEs (CC ICMEs) and non-cold material ICMEs (non-CC ICMEs). Thus, a total of four types of ICMEs are shown in Figure 7. The blue bar indicates MCs with cold  . This figure shows the significance test on the carbon ionic temperature for three types of segments: carbon-cold segments, IC segments, and COC segments. The distributions of carbon ionic temperature for all three groups are assumed to be Gaussian. The Y-axis is the carbon ionic temperature in log 10 , and the x-axis shows the three groups. The horizontal red line marks the median of each group. The blue top and bottom edges of the box mark the first and third quantiles. Outliers are marked as red pluses. The black horizontal line marks the boundary between normal data and outliers.  distribution) for each category, the uncertainty of the ratio X/Y is calculated as Y represents the yearly total number of ICMEs, and X represents the yearly total MC number and CC ICME number, for the two plots respectively. It could be seen that the fraction of CC ICMEs is less than or equal to that of MCs. Panel (c) shows yearly total   Wang & Colaninno (2014), the high CME number after mid-2010 is an artifact caused by identification of "very poor" eruptions and by LASCO image cadence. After calibration in Wang & Colaninno (2014), the SOHO/LASCO CME number shows a similar trend to sunspot number.

Discussion
In this study, we examined 219 ICME events detected by both the ACE and WIND spacecraft from 1998 to 2011. Using a criterion based solely on the average carbon charge state, we identified CC material in 69 of these ICMEs. Our results are summarized in the following text and in Figure 8. First, 69 of the 219 studied ICMEs contain cold material as defined by the criteria discussed in Section 2. Of these ICMEs, 38 are MCs. Only 18 out of 69 CC ICMEs can be confirmed to have clear three-part CME counterparts. Of these 18 events, 4 events have been previously reported by Lepri & Zurbuchen (2010). Second, the most probable duration of the segments during which the cold material is present is 2hr, while approximately 80% of the segments last for less than 10hr. The 69 CC ICMEs typically have one or two CC segments, rarely more than that. Third, the bulk solar wind proton speed for these segments typically lies between 300 and 600km s −1 . In addition, segments containing cold materials tend to have a lower proton temperature, and higher R αp comparing to their corresponding ICMEs. Fourth, we have identified two extreme populations for the periods that contain cold materials, one is the IC type and the other is the COC type. Ionic-cold type (6/108) has simultaneously lower average charge states of C, O, Mg, Si, and Fe than that of the 24 hr of solar wind preceding to the ICMEs (at 3σ-level). In comparison, COC type (15/108) has carbon being cold, but O, Mg, Si, and Fe are 3σ hotter than the preceding solar wind. We noticed that 11 of 13 ICMEs containing COC material are MCs, and 4 of 5 ICMEs containing IC material are non-MCs. To test if the appearance of COC and IC material in MC or non-MC is statistically  significant or not, we made the following hypothesis. We assumed that the COC were randomly distributed in all 219 studied ICMEs. Considering 99 of 219 ICMEs are MCs, so the probability that 11 of 13 ICMEs containing COC material are MCs is C C C 13 11 206 88 219 99 = 0.003. The very small probability means the null hypothesis should be rejected. Besides, we made Z-test on COC appearance ratios of the MC (11/99) and non-MC (2/120) groups, assuming the appearance probability of the two groups is the same. The Z value is 2.94 larger than the critical value for 95% confidence Z 0.05 =1.645. It means that the null hypothesis should be rejected. Thus, the COC material is not randomly distributed in 219 ICMEs, and its appearance in MCs has strong statistical significance. Next, we made the same hypothesis on IC material. The probability that 4 of 5 ICMEs containing IC material are non-MCs is 0.203, and the Z value is 1.14. It means the appearance of IC material may not be random, and its appearance in non-MCs is not very significant. We supposed that this weak result of IC is caused by the small number of IC samples.
We also noted that the IC material shows a higher proton temperature, and the COC material shows a lower proton temperature. Finally, there are four ICMEs containing cold material identified by both Lepri & Zurbuchen (2010) and our study. Only one of four, SN(L)4 or SN24, is classified as a COC type. This is due to the fact that the two temporal segments are slightly different.
Considering the above statistical features of the cold materials in the ICMEs, the relevant questions would be what is the nature of the cold material and how is it formed? Note that cold materials are not unique to ICMEs. A study by Zhao et al. (2017) reported that the decrease in the ratio of decreases as a result of either high-altitude reconnection heating or high-altitude wave particle interactions. The same phenomenon was also reported by Kocher et al. (2017) in the study of ICMEs. Checking our COC materials, we found all of them having lower C C O O 5 6 6 7 + + + + . In addition, the speed of our cold materials is typically less than 500km s −1 .
For the COC materials related to the filaments, the mean charge difference between the carbon ions and other heavier ions may be related to the initiation process of the CMEs. It is known that the freeze-in heights of the C, O, Mg, Si, and Fe ions are different. For example, in the slow wind, the carbon ion charge state is the first to freeze-in at approximately 1.2solar radii (R s ) (Geiss et al. 1995). Therefore, one possible process for forming COC type filament materials may be heating and collision beyond the carbon freeze-in height. During the eruption of CME, if the reconnection occurs above the carbon freeze-in height, but lower than those for other ions, then one expects that the carbon is cold, but other ions can go through ionizations and become hotter. Another possible explanation for the COC filament materials' formation is the prominence condensation in hot coronal loops. This has long been studied by Antiochos & Klimchuk (1991), Antiochos et al. (2000), and Karpen et al. (2005); it has been found that the filament material laying on the hot coronal loop is a combination of the cool condensed coronal plasmas and the evaporated chromospheric plasmas resulting from the pressure imbalance. Thus, the COC type filament materials may be a bulk of mixed ions from both the corona and the chromosphere.
It is not easy to understand the process leading to the formation of IC type filament materials. On one hand, the lower mean charges of all the C, O, Mg, Si, and Fe ions compared to that of the solar wind means that plasmas may not be sufficiently heated under 5R s (Geiss et al. 1995). On the other hand, the higher proton temperature means that the protons of the IC type filament materials tend to be heated. Such a heating may be possible by waves. It seems that in the events containing IC type filament materials the CMEs erupt without strong reconnection heating but are accompanied by preferential proton-wave resonance which efficiently heats protons. The fact that the IC type filament materials are found in non-MC ICMEs after the year 2006 and that the COC type filament materials are found in MCs before the year 2006, may suggest that these two types of filaments reflect fundamental differences between the two different solar cycles (Gopalswamy et al. 2015). Indeed, the solar activity was quite weak during the early years of the 24th solar cycle.
According to our study there are some IC and COC segments not related to the filaments. One possibility is that the core of CMEs were not observed due to the SOHO position and its instrument limitation. Another possibility is that these IC/COC segments do not correspond to filaments. As far as we know, there have been NRs of IC materials identified in the solar wind.
Our study suggests that there may be two different filament eruption processes, which have opposite effects on the ionization and proton temperatures. Although the occurrence rate of CC material in ICMEs is close to that of three-part CMEs, we cannot confirm that all CC segments are filament materials due to the limitation of solar observations before the STEREO era. Similar cold plasmas also appear in slow solar wind although with a lower occurrence rate (Zhao et al. 2017). Finally, we noted that further careful studies of IC materials may help us better understand both the CME initiation process and the heating mechanisms associated with the initiation. Such a study will benefit from the high-resolution imager on board the Solar Dynamics Observatory spacecraft (Pesnell et al. 2012) and future missions such as the Solar Orbiter (Mueller et al. 2013). This work is supported by the Open Research Program of the Key Laboratory of Geospace Environment, Chinese Academy of Sciences (CAS), and by the Fundamental Research Funds for the Central Universities. The used ionization temperature is from CHIANTI database version 8. CHIANTI is a collaborative project involving George Mason University, the University of Michigan (USA) and the University of Cambridge (UK). All the ACE and WIND data is from NASA CDA web. The ICME list is originally from the ACE Science Center. The sunspot number is from Royal Observatory of Belgium. And the CME number is from CDAW Data Center. Dr. Yao and Mr. Feng thank Dr. Shen, C.L.; Dr. Wang, C.B.; Dr. Zhao, L.; and Dr. Song, H.Q. for their helpful discussions. The authors thank the referee for helpful questions and comments.