A magnetically shielded, charge collecting rocket probe was used on two flights in the MIddle Atmosphere Dynamics and Structure (MIDAS) Studies of Layered STructures and ICE (SOLSTICE) 2001 rocket campaign over Andøya, Norway. The probe was a graphite collection surface with a permanent magnet underneath to deflect electrons. The first MIDAS was launched 17 June 2001 into a strong, multiply layered PMSE. The probe measured negative particles inside an electron biteout within the PMSE, having a peak charge number density of −1500 charges per cubic centimeter. The second MIDAS was launched 24 June 2001 into another strong, multiply layered PMSE. The probe saw a band of positive particles centered in the lowest radar echo maximum, and a negative particle layer accompanied by a positive ion excess. The charge number densities for the positive and negative PMSE particles were several thousand charges per cubic centimeter. Unexpectedly, 2 km beneath the PMSE, the probe also found a very pronounced negative layer, which was probably an NLC. Computer simulations of incoming, negatively charged ice grains were performed using a rarefied flow field representative of the MIDAS payload at zero angle of attack. Ice grains ≤1 nm in radius were diverted by the leading shock front, indicating the smallest detectable ice particle by this probe.
If you can't find a tool you're looking for, please click the link at the top of the page to go "Back to old version". We'll be adding more features regularly and your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Recent rocket, satellite, radar, and lidar studies of the polar summer mesosphere have investigated polar mesospheric summer echoes (PMSE) and noctilucent clouds (NLC, or more generally polar mesospheric clouds, PMC). PMSE are thought to be composed of small, icy, charged particles ∼10 nm in radius with charge number densities of several 1000 charges cm−3 [Havnes et al., 2001]. Repeated measurements by Havnes et al. [1996, 2001] showed positive particles in one PMSE and negative particles inside others. Recent rocket measurements of PMSE during the DROPPS campaign measured both positively charged [Croskey et al., 2001] and negatively charged [Mitchell et al., 2001] particles in the same PMSE. Particle charging models [Rapp and Lübken, 2001] have quantified the positive ion (electron) depletions and enhancements (depletions) that can occur in a mixture of plasma and particles.
 The MIDAS SOLSTICE rocket campaign was carried out at the Andøya Rocket Range (16°E, 69°N) in Norway during June 2001. The topics of investigation were the icy particles of NLC, the meter-scale plasma structures characteristic of PMSE, and the neutral number density and temperature profiles. The MIDAS payload took simultaneous in situ measurements of positive ions, electrons, charged particles, and neutral density. The magnetically shielded probe [Horanyi et al., 2000], hereafter called the MSP, detected charged particles alongside the PIP (Positive Ion Probe) [Blix et al., 1990] and CONE (COmbined Neutral and Electron) instrument [Rapp et al., 2001]. Figure 1 shows the location of these probes on the MIDAS payload. During downleg, this orientation placed CONE in the ram and the PIP in the wake of the rocket.
 The nearby ALWIN radar measured strong, multiply layered PMSE for the two launches of the MIDAS payload: SO-MI-05 on June 17, 2001 at 00:05 UT, and SO-MI-11 on June 24, 2001 at 21:21 UT. For both launches, the solar zenith angle was approximately 85°, the local Andenes magnetometer showed no significant magnetic disturbances, and cloudy weather prevented simultaneous measurements from the ALOMAR lidar. Thus ALWIN was the sole ground-based monitor of the science conditions.
2. Experimental Methods
2.1. Probe Description
 The MSP was a graphite surface that collected charge from impacting charged particles during flight. Beneath the graphite was a permanent magnet that completely deflected electrons and partially deflected positive ions. Since the graphite patch was at the same potential as the payload skin, the observed positive ion collection was likely assisted by the electric attraction of the payload potential, estimated to be −1.5 volts in the mesopause (T. A. Blix, personal communication, 2001).
 The rectangular 6 cm by 2 cm graphite collection surface was located in a connecting ring 40 cm from the base of the rocket, see Figure 1. The permanent magnet had an effective dipole moment of 9000 Gauss·cm3 with an effective radial depth of 4.35 cm into the rocket, determined by a least squares fit of magnetic field data measured by a computer controlled apparatus.
 During both flights the MIDAS payload rotated at ∼5 Hz. There was also precession, or coning motion, of the payload whose period was ten seconds. On downleg, this made the angle of attack smoothly oscillate between zero and 70 degrees, see Figures 2b and 3b. The PIP, situated in the wake of the rocket, was affected more by the coning motion than CONE, especially when the angle of attack was close to zero.
2.2. Probe Operation
 During flight through the mesosphere, the graphite patch collected a current on the order of nanoamps, sampled at 1085 Hz by a simple current-to-voltage amplifier. The dynamic range of the probe was ±2 nA with an accuracy of ±6 pA. The probe measured the net current at a given altitude, meaning the net sum of all positive and negative particle collections. This net current can be divided into two main components. There was a positive background from the collection of positive ions embedded in the airflow across the patch, and a particle signal from the incoming heavy charge carriers. Impact ionization effects are neglected in this investigation because the probe had no constrictive geometry to collect charged particle fragments, a concern raised by Havnes et al.  who used an electrically biased, blunt cup probe. Photoelectric effects on the graphite patch are also neglected, an assumption validated not only theoretically by the high work function and low secondary electron yield of graphite, but also experimentally by the lack of any signal in the sunlight direction.
 The positive background took the form of multiple positive peaks with every rotation of the payload, see Figures 2a and 3a. These positive “flow” peaks formed as the probe turned through the different parts of the shock front. The amplitude of the flow peaks increased with altitude due to the increasing density of positive ions, saturating the input above 100 km. The coning motion shaped the flow peaks into an oscillatory envelope, reducing the peaks as the angle of attack passed through zero.
 An additional particle current was collected when the graphite patch faced toward the ram direction. This signal was either positive or negative, depending on the relative abundances of the heavy charged particles at a given altitude. This charged particle current was in the 100 pA range and was superimposed on the larger flow modulation. Since the positive background was clearly proportional to the positive ion density, the PIP signal was used to remove the positive ion contribution from the net positive MSP data.
2.3. Extracting the Heavy Charge Carrier Residual From the Net Positive Data
2.3.1. Data Sorting Technique
 The positive background at any instant was determined not only by the positive ion density but also by the flow field around the rocket. To “freeze” the flow dynamics, raw data points within ±0.75 degrees of the ram direction were set aside, which selected one to two raw data points per rotation. This sorted data set only had variations due to the positive ions and heavy charge carriers. The ram direction was used because the most cross sectional area was presented to charged particles at that moment. This careful sorting was accomplished with a “velocity centered” coordinate system, constructed with the rocket orientation data provided by the onboard three axis magnetometer. For every data point, there existed a plane which held the rocket axis and the velocity vector, hereafter referred to as the velocity plane. The instantaneous rotation angle of the probe away from the velocity plane was calculated for all points. Small sketches of the velocity plane can be seen in Figures 2a, 3a, and 6.
 The phase of the flow peaks with respect to the velocity plane was almost completely uniform during downleg, a fact that inspired the sorting technique. This regularity of the flow peaks was not present during upleg, thus the downleg MSP data was the only one used in this technique.
2.3.2. Removing the Ion Background
 For both MIDAS flights below 95 km, the downleg MSP ram direction data had a clear resemblance to the downleg PIP data, one that was largely independent of the angle of attack, see Figures 4a and 5a. This suggests that positive ions were constantly forced into the graphite patch by the flow. Above 95 km, the ram data gradually diverged from the ion signal because the mean free path became large, reducing the forced ion collection. In the shaded regions of Figures 4a and 5a, there was not only an absence of radar echoes, but also a correlation between the ram data and the PIP signal greater than 90%. This correlation would be expected if these regions were devoid of heavy charge carriers, or “clean”, since nothing but ions contributed to the signal. Applying this concept, a simple linear transform of the ion signal should reproduce clean ram data and indicate the ion contribution elsewhere. If the altitude dependent ion contribution to the ram data, in nA, is called B(z), then
where z is the altitude in km, I(z) is the PIP signal in ions·m−3, m(z) is a scale factor in units of nA·m3, and b(z) is an offset in units of nA.
 Discrete values of m(z) and b(z) were found by performing least squares fits of I(z) to segments of the ram data. To prepare the downleg PIP data for fitting, it was smoothed to a time width of 0.2 sec, the length of one rotation. This removed the spin modulation and other periodic variations, creating a spin averaged PIP curve that could then be compared to the discrete ram direction points. The least squares fits used the shortest, continuous segments of ram direction data that retained the 90% correlation with the PIP; this length was ∼14 ram points, or a “window” of ∼1.5 km. Complete, evenly spaced windows were placed in the clean areas of Figures 4a and 5a, as many as would fit, in order to sample m(z) and b(z) as often as possible. This resulted in three curve fits for SO-MI-05 data set, but four for SO-MI-11, see Figures 4 and 5.
 Linear interpolation was used to estimate m(z) and b(z) between the clean areas since their values changed slowly and in a consistent way. Figures 4b and 5b show the values of m(z) and b(z) normalized to the lowest value and the interpolation for all altitudes. Figures 4c and 5c are plots of the resulting background B(z) on top of the sorted ram data. Finally, the heavy charge carrier residual was extracted by subtracting B(z) from the ram direction data.
2.4. Charge Number Density of the Heavy Charge Carriers
 The charge number density of the residual was deduced from the relation R=(q)(n)(Aeff)(v), where R is the heavy charged particle current in the hundreds of picoamps, q is the charge per particle, n is the heavy charged particle number density, and Aeff is the effective amount of area that collects the heavy charged particles. Aeff could be less than the full geometrical area Ageometrical= 12 cm2·sin(ø) (where ø is the angle of attack from Figures 2 and 3) but never more, since all conceivable corrections divert particles away from the patch. Since a smaller area results in a larger charge number density, the geometrical cross section was adopted to find the lower bound:
This (q)(n) is a sum over all the heavy particles (of all charges) that were present during collection.
3.1. Negative Particle Layer During SO-MI-11
 The only net negative current event is shown in more detail in Figure 6. This was interpreted as a negative particle layer because of the very low concentration of negative ions in the daytime mesosphere [Kopp and Fritzenwallner, 1998; Kopp, 2000]. This event was so brief in the sorted data it is more informative to look at the raw data. The negative current began at 80.76 km, labeled event n0 in Figure 6, and went to negative saturation for three samples. Over the next two rotations the negative layer receded, exhibiting small scale density variations. The three largest sporadic events were a wide negative peak at 80.63 km, hereafter called event n1, a pair of positive spikes at 80.76 km called p1, and single positive spike at 80.70 km, dubbed p2. At 80.4 km the probe current became net positive again, but appeared depressed relative to the positive ion profile until 79 km, see Figure 5c. The angle of attack was quite steep, starting at 0.2 degrees at n0 and only growing to 8.5 degrees by 80.4 km. The downleg positive ion data showed no change during this event, but this could be because of the geometry. The steep angle of attack kept the PIP near the center of the rocket wake, making small ion enhancements or depletions hard to detect. During upleg of SO-MI-11, when the PIP was at the leading end of the payload, there was a slight ion excess seen at p2, see Figure 6. The rocket telemetry logs showed no electrical disturbances during passage through the negative layer, nor was there an abrupt change in the payload potential, so the negative layer data were considered genuine.
3.2. PMSE Particles From SO-MI-11
 There was evidence of both positive and negative particles in this PMSE, see Figure 7b. The uncertainty of the derived charge number density was large because of the summed uncertainties of the PIP data (taken to be 25%), the MSP data, m(z), and b(z). The electron density from CONE was not available due to a technical failure, but Figure 7c shows the simultaneous positive ion density from PIP.
 Coincident with the lowest radar echo maximum at 84 km was a broad band of positive particles labeled β. The positive particle density generally followed the outline of the radar echo, reaching a maximum of +1700 ± 1200 charges cm−3. Below β the positive particle density reduced to zero along with the radar echo. At 83.18 km there was a negative layer that lasted for two ram direction points, labeled α. Negative layer α peaked at −1800 ± 1200 charges cm−3 and was accompanied by an obvious enhancement in the positive ion signal, see Figure 7c.
 Above β the particle density decreased, becoming net negative at 88.32 km. The negative particles increased until event γ, above which the angle of attack fell below 20 degrees and the shock front began to divert particles, a consequence of the side mounting of the probe. The ion signal showed another small enhancement at γ, see Figure 7c. The particle signal reemerged symmetrically, peaking at −2000 ± 1300 charges cm−3 at 89 km, labeled δ, approximately 2 km beneath the mesospheric temperature minimum, see Figure 7a.
3.3. PMSE Particles From SO-MI-05
 On SO-MI-05 both the electron and positive ion data were available, see Figure 8c. The MSP showed negative particles punctuated by a positive peak of +400 ± 600 charges cm−3 at 81.7 km, labeled ε, Figure 8b. ε coincides with the lowest radar maximum similar to SO-MI-11, but the large uncertainty prevents confidence. The heavy charge carrier signal rose above the error bars in region ζ where there was a significant electron and positive ion depletion, assuming that the plasma densities immediately above and below ζ were unperturbed. The MSP saw negative particles in proportion to the electron biteout, even culminating in an identical maximum of −1500 ± 700 charges cm−3 at 83.8 km, labeled η. Immediately above η the angle of attack went through zero, reemerging into more negative particles in θ. A maximum of −3400 ± 800 charges cm−3 was reached at 90.8 km, location ι, approximately 1 km below the temperature minimum, see Figure 8a. The ion density in θ appeared depleted, but notably the electron density either looked unperturbed or showed an excess.
4.1. Computer Modeling of Incoming Particle Trajectories
 Computer simulations of incoming particles were performed to check which ice grains were capable of entering the MSP. The flow field around the rocket was required for these simulations, previously described by Horanyi et al., . In short, the simulations integrate the equations of motion for nanometer-sized spheres, taking into account viscous drag forces, magnetic forces, heating, and sublimation. At precisely zero degrees angle of attack, the flow around the MIDAS payload was azimuthally symmetric. This two dimensional fluid flow problem can be solved with Direct Simulation Monte Carlo (DSMC) methods [Gumbel, 2001]. Figure 9 shows the density field found from the DSMC simulation in multiples of the ambient mesospheric density. The simulation was performed at a neutral number density representative of 85 km in altitude. Note that a simplified outline of the rear mounted CONE instrument was used as the boundary condition rather than a blunt cylinder.
 The trajectories of spherical ice grains with radii of 100, 10, and 1 nm are shown in Figure 9. The grains were given a single negative charge. The shock completely diverted the 1 nm grains, but 10 nm and 100 nm grains pushed through the shock and hit the payload. The mass loss was completely negligible for 10 nm and 100 nm grains, while the most heavily sublimated 1 nm grain lost 40% of the starting mass. 100 nm grains passed directly over the graphite patch, but the v x B force was far too weak to turn them into the patch, even with an unreasonable charge of 300 electrons!
 These simulations reproduce the trajectory deflections that caused the heavy charge carrier residual to attenuate at zero angle of attack. They add confidence that the extraction process was accurate since angle of attack did not explicitly enter that calculation. While particles ≤1 nm are highly influenced by individual molecular collisions and could conceivably bounce into the detector at zero angle of attack, the leading shock front sweeps out of the approaching air volume. Thus the detectable ice grains for the MSP are ≥1 nm.
4.2. Negative Layer From SO-MI-11, the Suspected NLC
 There were brilliant NLC displays seen throughout Europe on June 24, the night of SO-MI-11 (see the NLC reporting website http://www.nlcnet.co.uk/) In fact, one hour after the launch, observers in Oslo (60°N, 11°E, ∼1000 km southwest of the Andøya launch site) reported NLC at a local elevation angle of 40° which covered the sky from west to north. Those NLC were assigned a brightness rating of 4 out of 5 and possessed every categorical form (diffuse, banded, wave-like, whirl-like) recorded by the site. Hence it is probable that NLC were present above Andøya for SO-MI-11.
 The location of the negative layer was close to the observed locations of NLC and PMC. Measurements of PMC [Carbary et al., 2001] found an average height of 82.6 ± 1.3 km in the northern hemisphere. NLC lidar studies [von Cossart et al., 1999] also showed an average NLC altitude of 82.5 km, suggesting that NLC and PMC are identical phenomena simply viewed from different perspectives.
 The neutral density and temperature measurements from CONE can be used to find the water mixing ratio required for ice at the altitudes occupied by the negative layer. The ice saturation pressure from Marti and Mauersperger  shows that a water concentration of 16 (23) ppm was needed at 80.7 (80.4) km to sustain ice grains, a concentration seen (not seen) by Summers et al.  in the average 82 km – 84 km band. Thus, if the negative layer was an NLC, then it was at the lowest part of the local existence zone when observed. Stevens et al.  have seen PMC as low as 80 km, so the negative layer is plausibly close to all average altitudes. Wave motions like those seen by Gerrard et al.  could also have moved the negative layer down from the average.
 The negative layer was 2 km beneath the bottom of the PMSE, similar to earlier observations [Wälchli et al., 1993; von Zahn and Bremer, 1999; Stebel et al., 2000] showing when NLC and PMSE appear in the same volume, the NLC occurs at either the lower edge of the PMSE or a few km lower. This has been partly attributed to a growth process where smaller PMSE particles at higher altitudes coagulate and descend, making larger NLC particles at lower altitudes. Another possibility was that the ambient ionization was only sufficient to create the PMSE above 83 km [Rapp et al., 2002], well above the negative layer.
 One striking feature of Figure 6 was that negative charge was collected continuously through two rotations after n0. The steep angle of attack kept the flow (partially) azimuthally symmetric, sporadically distributing negative particles all the way around. The azimuthal symmetry was lost as the angle of attack grew and the omnidirectional negative signal ceased. Another important point is that n0 and n1 both occurred 45 degrees prior to the ram direction. The consistent forward facing angle suggests that they were larger particles that entered the MSP with little deflection compared to the omnidirectional component. Figure 9 shows that the additional particles at n0 and n1 would need to be ∼100 nm. Thus there was evidence of a particle size distribution in the negative layer: smaller particles swept in by the flow and additional, heavier ones that approached with straighter trajectories.
 The positive peaks p1 and p2 were most likely ion layers, given the serendipitous alignment of p2 and the ion excess seen on upleg. Since the horizontal distance between the upleg and downleg positions at this altitude was 20.7 km, the upleg ion data cannot be expected to match perfectly. Ion enhancements can exist in negatively charged particle layers [Rapp and Lübken, 2001], although a localized shift in particle size or ion recombination coefficient would be needed to confine the ion excess to a narrow band. It is also possible that the same mesospheric winds that collect water vapor across large distances for the growth of NLC particles [Thomas, 1991] could also gather ions into a layer by windshear.
 The charge number density of the negative layer as seen in the ram data is in Figure 7b. Ignoring the single positive data point that came from p1, the layer had a peak negative charge number density of −2600 ± 1100 charges cm−3. If an average charge of 3 electrons per NLC particle is assumed (an estimate for 50 nm grains in NLC conditions taken from Rapp and Lübken ,) this makes a particle number density of 870 cm−3. The next lowest negative peak in the ram data was −1700 charges cm−3, making a particle number density of 570 cm−3. On the average, lidar derived NLC particle number densities are somewhat lower. A range of 260–610 particles cm−3 was found by Alpers et al. , and von Cossart et al.  reported an average of 82 ± 52 particles cm−3, with one NLC event above 1000 particles cm−3. If the negative layer is assumed to be an average NLC, then the top of the negative layer had a little more negative charge than estimated by applying average particle charges to average lidar number densities. Since lidar backscatter rapidly diminishes as particles become smaller than 20 nm in radius, the upper parts of the negative layer could have contained particles smaller than the detection threshold. If the negative layer is assumed to be a strong NLC, (a reasonable assumption based on the aforementioned ground-based observations) then the upper number density is comparable to the most intense NLC event from von Cossart et al. . Overall, the tapering of the particle density within the NLC with altitude was consistent with the top of the negative layer containing more numerous, smaller particles which coagulated during descent into fewer, larger particles.
4.3. PMSE Particles, SO-MI-11
 The positive particle layer at β is not an unprecedented measurement. Croskey et al.  saw evidence of two groups of positive particles, corresponding to sizes of 1 nm and 10 nm, within a PMSE during the DROPPS campaign of 1999. The number density of the 10 nm positive particles was seen to approach that of the 1 nm positive particles precisely within the radar echo peak, becoming nearly 3000 charges cm−3. Assuming a similar situation existed during SO-MI-11, positive 10 nm particles are likely candidates for the positive peak. Using Figure 9 as an estimate, 10 nm particles are clearly heavy enough to penetrate the shock front but 1 nm particles are going to be deflected. Layer β was 1700 charges cm−3, lower than the DROPPS result if singly positively charged particles are assumed. Negative particles, like those seen by Mitchell et al.  in the same PMSE from DROPPS, could have existed alongside the positives and reduced the intensity of the heavy charge carrier signal. Thus, even if the mixture contained 1000 cm−3 singly charged negative particles (like the 10 nm ice grains in PMSE conditions from Rapp and Lübken ,) making a true positive particle density of 2670 cm−3, this is still below the maximum value seen by Croskey et al .
 The negative layer α is reminiscent of the dust measurements of Havnes et al.  because α occurred exactly in a small inflection in the radar profile with a charge number density in the thousands of charges cm−3. Because there was an ion enhancement, the charging model of Rapp and Lübken  can be checked to see which particle sizes theoretically produce ion excesses. Without electron data the relative electron depletion is unknown, but that can be left as a free parameter. If the relative ion excess is taken to be approximately 25%, then figure 1 of Rapp and Lübken  shows that singly negatively charged particles sized 2 to 5 nm in radius are eligible candidates. Thus, the charging model does not explicitly forbid ion enhancements for the particle sizes that are detectable by the MSP. These arguments would apply to the ion excess seen at γ as well.
 The altitude of δ suggests that the temperature minimum is a source of particles. The net current nature of the MSP revealed the most numerous particles at different heights, showing mostly positive (negative) particles that were correlated with the radar echo (temperature profile) at the low (high) regions of the mesopause.
4.4. PMSE Particles From SO-MI-05
 The correlation of the negatively charged heavy particle residual and the electron reduction in ζ was encouraging. At 86.3 km, near the bottom of ζ, the heavy particle charge number density was −700 charges cm−3, while the electron density was roughly 6900 cm−3, making the sum of all negative charge approximately 7600 cm−3. At the same altitude, the positive ion density was nearly 8000 cm−3, showing that the oppositely charged particle densities were within 5% of one another. Furthermore, at η, the top of ζ, the negative heavy particles peaked at −1500 cm−3 and the electron density was 5300 cm−3, bringing the negative charge total within 6% of the 7300 cm−3 positive ion density. Thus the total positive and negative particle densities are quite close throughout ζ, indicating (quasi) charge neutrality. The small discrepancies from perfect charge neutrality could be caused by several factors. Some fraction of the negative heavy particles could have been too small to penetrate the shock front, as in Figure 9. Another explanation is that simultaneous populations of positive and negative particles were present, and not all were collected with equal efficiency by the MSP.
 The charging model of Rapp and Lübken  can be applied to region ζ to check for obvious contradictions. Taking the relative electron depletion was somewhere between 25% and 50% (estimated by drawing an imaginary line between the upper and lower edges of ζ,) and taking the relative ion depletion to be 25%, then figure 6 of Rapp and Lübken  reveals that particles 5 to 20 nm in radius are eligible, which are all detectable particle sizes.
 Similar to SO-MI-11, the residual showed more negative particles reaching a maximum at ι, 1 km below the temperature minimum. Positive ions appear depleted in θ, indicating heavy charged particles of some kind, but CONE showed either an unperturbed electron profile or a small enhancement. The transition from the good charge neutrality seen in ζ to the apparent disagreement in θ is puzzling.
 One explanation for this disagreement is that the high electron density (>10,000 cm−3) overcame the magnetic shielding. This explanation is unlikely because the peak at ι would not have formed since electron density continued to rise. Another explanation is that the negative particles in θ were small and mobile enough (≤1 nm) to be detected by the CONE instrument as if they were electrons. However, this is even more unlikely, since the attractive +6 volt potential of the CONE instrument would not create an effective cross section for other negative particles as large as the one for electrons. Also note that the shock front would easily deflect such small particles.
 Perhaps a distribution of charged particles could explain the disagreement of the MSP and CONE in θ. As a thought experiment, suppose that dust grains in θ had a low work function because they contained sodium, like those discussed in Rapp and Lübken . Assume a double distribution of 2 nm and 10 nm grains where the 2 nm grains outnumber the 10 nm grains by a few thousand cm−3. Figure 2 of Rapp and Lübken  shows that 2 nm sodium grains will tend to be singly negatively charged, while the 10 nm grains would have one positive charge. The smaller grains would only scavenge a few thousand electrons cm−3 from the preexisting distribution because the large grains emitted electrons, offsetting the depletion. The electron profile would not appear depleted because a few thousand electrons cm−3 was only 10% of the total seen by CONE at ι. When the net current is measured by the MSP, the smaller negatively charged grains outnumber the larger positive ones and create a net negative heavy particle residual. If small negative particles were the source of the MSP/CONE disagreement, and the mesospheric temperature minimum was the source of them, then the disagreement should be greatest near the minimum and decrease with altitude, since ice grains grow by coagulation as they descend. This is exactly what was seen inside θ, a negative heavy particle density that peaked near the temperature minimum and decreased with altitude, thus this thought experiment is self-consistent.
 Finally, it must be accepted that the extraction process for the MSP data could have failed at the upper part of the mesopause. Since the extraction technique completely relies on the PIP data, overestimating the positive ion contribution would create the appearance of negative particles, and vice versa. Areas where the PIP and CONE show divergent positive ion and electron densities, like the lower half of θ, are the areas where failure could be expected. However, this suspicion would not apply to region ζ, since it is closer to the expected (quasi) charge neutrality.
 During SO-MI-11, the MSP detected a pronounced negative charge layer at the probable location of an NLC. This layer had a charge number density of −2600 charges cm−3 that was large but consistent with lidar derived number densities multiplied by model particle charges. The MSP detected positive and negative particles in the PMSE at charge number densities which were mutually consistent with Havnes et al. , Croskey et al. , and Mitchell et al. . The lower negative particle collection was simultaneous with a positive ion excess, another consistency with particle charging models. During SO-MI-05, the MSP saw negative particles in an electron biteout, as expected from charging models and earlier results from Havnes et al.  and Mitchell et al. . More negative particles were seen in regions where the electron density did not look depleted, but this could be explained by a net negative particle distribution or a failure of the extraction process at the highest altitudes. This MSP offers a technologically distinct strategy for measuring charged particles in PMSE and NLC, possessing an open geometry very different from electrically biased blunt cup probes. Improvements to the MSP to reduce positive ion collection will include stronger magnetic shielding and a positive bias (+3V) for the graphite patch, making an instrument which only sees the heaviest charge carriers in a mixed plasma/particle environment.
 The authors thank DLR for their willingness to include the MSP on the MIDAS payload on short notice. B.S., S.R., and M.H. were supported by NASA.