Absence of a long-lived lunar paleomagnetosphere

Abstract

Determining the presence or absence of a past long-lived lunar magnetic field is crucial for understanding how the Moon's interior and surface evolved. Here, we show that Apollo impact glass associated with a young 2 million-year-old crater records a strong Earth-like magnetization, providing evidence that impacts can impart intense signals to samples recovered from the Moon and other planetary bodies. Moreover, we show that silicate crystals bearing magnetic inclusions from Apollo samples formed at ∼3.9, 3.6, 3.3, and 3.2 billion years ago are capable of recording strong core dynamo-like fields but do not. Together, these data indicate that the Moon did not have a long-lived core dynamo. As a result, the Moon was not sheltered by a sustained paleomagnetosphere, and the lunar regolith should hold buried 3 He, water, and other volatile resources acquired from solar winds and Earth's magnetosphere over some 4 billion years.

Full text

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
1 of 14
SPACE SCIENCES
Absence of a long-lived lunar paleomagnetosphere
John A. Tarduno1,2*, Rory D. Cottrell1, Kristin Lawrence3†, Richard K. Bono4, Wentao Huang1,
Catherine L. Johnson3,5, Eric G. Blackman2, Aleksey V. Smirnov6,7, Miki Nakajima1,2, Clive R. Neal8,
Tinghong Zhou1, Mauricio Ibanez-Mejia9, Hirokuni Oda10, Ben Crummins1
Determining the presence or absence of a past long-lived lunar magnetic field is crucial for understanding
how the Moon’s interior and surface evolved. Here, we show that Apollo impact glass associated with a young
2 million–year–old crater records a strong Earth-like magnetization, providing evidence that impacts can impart
intense signals to samples recovered from the Moon and other planetary bodies. Moreover, we show that silicate
crystals bearing magnetic inclusions from Apollo samples formed at ∼3.9, 3.6, 3.3, and 3.2 billion years ago are
capable of recording strong core dynamo–like fields but do not. Together, these data indicate that the Moon did
not have a long-lived core dynamo. As a result, the Moon was not sheltered by a sustained paleomagnetosphere,
and the lunar regolith should hold buried 3He, water, and other volatile resources acquired from solar winds and
Earth’s magnetosphere over some 4 billion years.
INTRODUCTION
Three outstanding questions center on the Moon’s past magnetism.
The first asks whether the lunar core could have generated a long-
lived dynamo producing a strong surface field (1–2). The second
asks whether the associated magnetosphere contributed to the pro-
tection of Earth’s atmosphere from erosion by early solar winds
(3–4). The third asks whether a lunar paleomagnetosphere blocked
ion transport from the solar wind and Earth, ultimately limiting these
as long-term sources of volatiles (5) in the lunar regolith. These
questions further stem from the surprising discovery of magnetism
in some of the lunar rocks returned from the Apollo missions.
Paleointensity estimates published in the 1970s and ’80s were
interpreted as evidence for a global lunar magnetic field between 3.9
and 3.6 billion years (Ga) ago as strong as or stronger than that of
Earth today [e.g., (1)]. The difficulties in generating such high field
strengths in the small lunar core were recognized in these works.
Further caveats about the early data stem from the nonideal nature
of magnetic carriers in lunar samples and techniques used to retrieve
paleofield strength estimates. Many lunar samples show nonideal
multidomain-like magnetic characteristics. Analysis methods have
commonly relied on the application of laboratory magnetic fields
rather than thermal treatments that duplicate the thermoremanent
magnetization (TRM) process (6), which might have imparted
primary magnetizations to lunar rocks. Moreover, many Apollo
samples are the products of impacts (glasses and breccias), and con-
cerns were raised that magnetizations could be imparted by shock
(7). Satellite data show very weak crustal magnetic fields over much
of the lunar crust [e.g., figure1 of (8)]. The low strength and deep
source depths are interpreted to record magnetizations acquired
during crustal cooling before 4.4 Ga ago (8). Relatively stronger
anomalies around part of the South Pole–Aitken basin could reflect
impact-delivered iron during this time (8). However, other magnetic
anomalies do not form a clear pattern in space or time. In particu-
lar, vast areas of mare basalts formed during the proposed high–
magnetic field epoch lack magnetic signatures (fig. S1).
The paradox of lunar magnetizations lay fallow for 25 years until
it was revisited by Lawrence etal. (9). Thellier double-heating ex-
periments that replicate the TRM process yielded results that ques-
tioned a long-lived lunar dynamo, but one of the samples analyzed
was subsequently measured by another group using nonthermal
methods, and the data were interpreted as further evidence for an
ancient dynamo (10). Moreover, the age of a strong dynamo field
was extended beyond that originally proposed (1), to older ages (10),
and then with additional nonthermal measurements to younger ages,
the latter giving rise to the concept of a “late lunar dynamo” at 3.56Ga
(11). This prompted new models of potential lunar dynamo gener-
ation (12–13), but none can successfully predict the high sustained
Earth-like field values (2). The conundrum has only deepened with
the report of a ∼1.5-Ga-old magnetization using Thellier thermal
analysis from a lunar impact breccia and its interpretation as a re-
cord of an even later core dynamo field (14).
Against this background of sample data that seem to support a
long-lived lunar dynamo [e.g., (15)], it is important to recognize
that analyses before and since the 2008 Lawrence et al. study (9)
have generated data that might indicate essentially null field values
and the absence of a lunar dynamo (see Materials and Methods).
Thus, the currently accepted concept of a long-lived lunar dynamo
extending from ∼4.2 to ∼1.5Ga relies not only on a choice of data
that appear to record strong fields but also on two key corollaries
that state the following: (i) These fields cannot be produced by any
process other than a core dynamo, and (ii) data recording null fields
are not accurate because the samples cannot record strong fields.
Recently, a Thellier analysis has been reported on a 1-Ga-old
Apollo sample and interpreted as representing the cessation of the
lunar dynamo (16). Here, we first study an Apollo glass sample
linked to a 2–million year (Ma)–old impact that, following this
timeline, is predicted to have no remanent magnetization imparted
1Department of Earth and Environmental Sciences, University of Rochester, Rochester,
NY 14627, USA. 2Department of Physics and Astronomy, University of Rochester,
Rochester, NY 14627, USA. 3Planetary Science Institute, Tucson, AZ 85719-2395,
USA. 4Geomagnetism Laboratory, University of Liverpool, Liverpool L69 3GP, UK.
5Department of Earth, Ocean and Atmospheric Sciences, University of British Columbi a,
Vancouver, BC V6T 1Z4, Canada. 6Department of Geological and Mining Engineer-
ing and Sciences, Michigan Technological University, Houghton, MI 49931, USA.
7Physics Department, Michigan Technological University, Houghton, MI 49931,
USA. 8Department of Civil and Environmental Engineering and Earth Sciences,
University of Notre Dame, Notre Dame, IN 46556, USA. 9Department of Geosciences,
University of Arizona, Tucson, AZ 85721, USA. 10Research Institute of Geology and
Geoinformation, Geological Survey of Japan, National Institute of Advanced Indus-
trial Science and Technology (AIST), Tsukuba 305-8567, Japan.
*Corresponding author. Email: john.tarduno@rochester.edu
†Work done while affiliated with PSI.
Copyright © 2021
The Authors, some
rights reserved;
exclusive licensee
American Association
for the Advancement
of Science. No claim to
original U.S. Government
Works. Distributed
under a Creative
Commons Attribution
NonCommercial
License 4.0 (CC BY-NC).
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
2 of 14
from a dynamo. We instead find a strong Earth-like magnetization
and show that its origin is related to the impact that formed the
sample. This discovery provides evidence for a mechanism that dis-
counts corollary (i) with implications more broadly for planetary
magnetizations. This leads us to an examination of corollary
(ii) through the analysis of five additional lunar samples with ages
spanning the putative lunar high-field epoch (1) and its extension to
the late lunar dynamo. We find that these samples are capable of
recording strong dynamo-like fields but instead have negligible re-
manent magnetizations, compatible with a null lunar field and
discounting corollary (ii).
These findings allow us to address the three salient questions
about the lunar magnetizations. As we will show, our new data
indicate that the Moon did not have a long-lasting core dynamo.
Thus, a sustained lunar paleomagnetosphere was not present, which
might have helped protect Earth’s atmosphere from solar winds. In-
stead, the lunar regolith should record ion transport from the solar
wind and Earth’s magnetosphere over some 4Ga.
RESULTS
We start by analyzing lunar sample 64455 (fig. S2), a ∼5 cm–
by–3cm ovoid-shaped basaltic impact melt linked to the ∼680-m-
diameter South Ray crater (17–18). The sample was collected ∼4380m
from the center of the crater and has a thick glass rim having a del-
icate, smooth exterior that almost completely covers the rock (17).
It has cosmogenic exposure ages of 2Ma [see Materials and Methods
and (19)], and the distribution of micrometeorite “zap” pits sug-
gests that it has maintained its orientation on the lunar surface since
it landed. On the basis of the physical nature of the sample, the glass
composition, the site geology, and the consistency of exposure ages
from 22 associated Apollo samples, the glass formation age is thought
to be 2 Ma, coinciding with the impact that formed South Ray crater
(see Materials and Methods). We focus our analyses on the glass
(Fig. 1A); light microscopy reveals that glass subsamples contain
spherical metallic inclusions that are <1 to 5 m in size but that are
sometimes as large as 100 m in diameter (Fig.1B) (see Materials
and Methods). Magnetic hysteresis measurements (see Materials
and Methods) show thin curves indicative of a dominance of low
coercivities typical of lunar samples (Fig.1C). A clear wasp-waisted
(20) nature to the curves may suggest the presence of ultrafine super-
paramagnetic (SP) grains. Another distinct feature is the very high
ratio of coercivity of remanence (Bcr) to coercivity (Bc) (Bcr/Bc =
16.6; table S1), different from typical terrestrial carriers but charac-
teristic of lunar samples (fig. S3). First-order reversal curve (FORC)
diagrams (see Materials and Methods) show a central peak that is
slightly asymmetric with a downward trend with increasing estimated
coercivity that may be a sign of minor interactions (Fig.1,CandD).
1 mm
200 µm
A
B
−0.4
−0.3
−0.2
−0.1
0.0
0.1
Bu (T)
0.30.20.10.0
B (T)
800 ×10
−9
6004002000−200
A m2/T2
c
C
0.10
0.05
0.00
−0.05
−0.10
Bu (T)
0.200.150.100.050.00
Bc
(T)
D
−200
200
01×(
M
9−
A m
2
)
−500 0 500
B (mT)
1 µmBSD
EFG
Fe
Fe
Fe
PNi
Ni
Ni
S
0
400
800
1200
1600
2000
2400
Counts
keV
0246
81
0
SO
NiFe
3 µm
Fig. 1. Light and electron microscope imaging and rock magnetic analyses of Apollo sample 64455. (A) Glass from Apollo sample 64455. (B) Light microscope image
of typical largest metallic spheres observed. Smaller sphere in background. (C) FORCs for subsample 64455-ss55. Saturating field, 1 T; number of FORCs, 156; field incre-
ment, 6 mT. Smoothing criteria (see Materials and Methods): Sc0 = 7, Scb = 5, Sc1 = Sb1 = 8, and h = v = 0.20. The inset shows individual magnetic hysteresis loop cor-
rected for paramagnetic slope. (D) Detail of central portion of (C). (E to G) SEM data for 64455 glass subsample, inclusion (B). (E) Backscatter detector image, 20-keV beam
strength. (F) Energy-dispersive spectroscopy (EDS) maps. (G) Spectral analyses (EDS) of spots labeled in (E). See table S2 for compositional estimates from the EDS data.
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
3 of 14
There is no evidence for a very high coercivity signal; overall, the
individual hysteresis curves and FORCs suggest pseudosingle domain–
like grains mixed with SP particles. Scanning electron microscopy
(SEM) and energy dispersive analyses of unheated subsamples (see
Materials and Methods) reveal that magnetic particles are concen-
trated in the spheres, which have a diversity of internal structures
and distributions of Fe, Ni, and S (Fig.1,EtoG, and figs. S4 to S10).
Fe/Ni estimates (table S2), together with the magnetic coercivities
and the glass setting, indicate that body-centered martensitic and
face-centered taenite compositions and structures have been quenched
in the inclusions within the glass. Ni contents are higher than some
other lunar samples and may indicate the incorporation of impactor
material into the glass.
Thermally induced alteration is a well-known problem in
paleointensity analyses of lunar samples. However, prior concerns
have focused on chemical alteration and not on attendant changes
in magnetic structure [e.g., (6)]. Here, we use rapid, brief heating
using CO2 laser methods to limit both effects (see Materials and
Methods). Total TRM (TTRM) experiments, whereby the natural
remanence is compared to that imparted at a single temperature,
can be used to yield a first-order assessment of paleointensity (see
Materials and Methods). We select a temperature (590°C) that covers
much of the predicted unblocking temperature spectrum of potential
magnetic carriers. We first apply thermal treatment to a test speci-
men (see Materials and Methods). A comparison of magnetic
hysteresis data before and after heating shows no evidence for changes
in domain state, and we conclude that the rapid and brief heating is
insufficient to create or destroy magnetic minerals (fig. S11). TTRM
demagnetization data show the inability of one subsample to accu-
rately record the field at high unblocking temperatures, and minor
structural changes with heating are hinted at by the lack of perfect
replication of the natural remanent magnetization (NRM) and
TTRM demagnetization curves (fig. S12, A to C). Nevertheless,
these changes appear to be minor because the NRM versus TTRM
loss data over a temperature segment with a characteristic com-
ponent of magnetization from two subsamples yield replicable
paleointensity estimates of 12.2±0.7 and 11.6±4 T (Fig.2A and
fig. S12).
0
1
2
3
Temperature (oC)
Paleointensity = 12.2 + 0.7 µT
NRM lost
Total TRM
ss61
0 100 200 300 400 500 600
NRM (×10−10 Am
2)
NRM (×10−11 Am
2)
NRM (×10
−10
Am
2
)
100
1
7
5
250
360
450
550
6
50
250
123
45
TRM (×10−11 Am
2)
1
3
3
4Paleointensity = 15.6 + 2.3 µT
ss1
1
2
3
0 100
IRM (×10
−10
Am
2
)
40 mT
0
mT
100 mT
N, Up
S, Down
WE
10 mT
100
oC
320
425
Tick marks: 1 × 10-11 A m2NRM
NRM
2
Paleointensity = 23.0 µT
S, Down
WE
N, Up
15 mT
30 mT
15 mT
30 mT
S, Down
WE
N, Up
Tick marks: 2 × 10–9 A m2
Tick marks: 5 × 10–11 A m2NRM
NRM
SIRM
SIRM
40 mT
100 mT
40 mT
100 mT
ss52
ss52
A
CD
B
100
oC
N, Up
S, Down
E
W
100
oC
250
250
175
550
550
650
650
Tick marks: 5 × 10-12 A m2
1 mm
1 mm
1 mm
Fig. 2. Paleointensity analyses on Apollo 64455 glass. Subsamples measured are shown as inset in (A to C). (A) TTRM experiment. Plot of the decay of NRM and
laboratory-induced TTRM with temperature. The pink shaded region represents steps used to determine the paleointensity value. The inset shows the orthogonal vector
plot of NRM demagnetization. A three-point sliding window average was used to reduce noise in the remanence signal and determine the characteristic remanent mag-
netization (ChRM) using principal components analysis (green arrows). Red, inclination (vertical) component; blue, declination (horizontal) component. (B) Thellier-Coe
paleointensity experiment. The loss of NRM is plotted against the acquisition of a laboratory-induced TRM (circles) and (pTRM) check (triangle). Black circles/blue labels
identify data used to fit paleointensity. The inset shows orthogonal vector plot of field-off steps. Conventions as in (A). Labels in italics identify temperature range used in
paleointensity fit. (C and D) REM’ paleointensity determination. (C) The loss of NRM plotted against loss of IRM. (D) Orthogonal vector plot of AF demagnetization of NRM
(top) and 3-T saturating IRM (bottom); conventions as in (A). The slope of the line in (C) that matches the AF range, where the ChRM is defined in the NRM orthogonal
vector plot [40 to 100 mT in (D)], is related to the paleofield strength by a calibration factor (see Materials and Methods).
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
4 of 14
Some subsamples chosen for the more robust double-heating
Thellier-Coe paleointensity analyses (see Materials and Methods)
showed evidence for multiple components and/or changing direc-
tions after field-off thermal treatments. Others show evidence for
thermally induced chemical or structural changes and/or nonideal
recording behavior, but three subsamples (12%) pass partial TRM
(pTRM) checks and yield paleointensity values of 15.6±2.3, 18.1±3.1,
and 23.5±4.7 T (Fig.2B, fig. S13, and table S3). We also applied
the ratio of equivalent magnetizations (REM′) nonheating paleo-
intensity method (Fig.2,CandD, and fig. S14), applying a saturation
isothermal remanent magnetization (SIRM), and demagnetizations
using alternating fields (AFs) with smoothing to address gyroremanent
magnetization (GRM) effects; ultimately, this method relies on the
use of calibration data (see Materials and Methods). This approach
yields values ranging from ∼23 to ∼39 T, not considering uncer-
tainties in data fits or calibrations, or ∼10 to ∼89 T, consider-
ing these uncertainties (see Materials and Methods and table S4).
The mean paleointensity estimates for the 64455 glass based on
thermal and nonthermal methods are fascinating because fields this
strong have been interpreted as evidence for a dynamo in samples
billions of years old (1). The linkage of 64455 glass to South Ray
crater suggests an age that is millions, not billions, of years old, and
at this time, the Moon’s interior thermal state would have been in-
distinguishable from that of today, incapable of sustaining a core
dynamo. The lack of a dominant soft coercivity component indi-
cates that spacecraft contamination is unlikely (21). Another possi-
bility is the cooling in a local crustal magnetic field, but this is
effectively discounted by the small magnetic fields of 112 ± 5 nT
measured by Apollo 16 astronauts (22) and recent satellite measure-
ments (23) that indicate surface fields orders of magnitude smaller
than those needed to explain the 64455 paleointensity data.
A remaining possibility is a field imparted by the formation of
the South Ray crater. Impacts can result in magnetizing fields
through compression of the solar wind (24–25) or through charge
separation (26); only the latter is relevant to magnetizations near the
impact site and thus is considered here (see Materials and Methods).
Experiments and detailed simulations (26) support generation of a
field B at radius r for small impactors of radius R (0.1 ≤ R ≤ 3 km)
following
B ∼ 8.3 × 1 0 −4 (r / 50R) −2 (R / 1km ) (v / 20km s −1 ) 3.6 (1)
where v is the impactor velocity and B is in tesla. For larger impacts
(3 ≤ R ≤ 20 km), the simulations using impactor velocities of 20km s−1
support a relationship of
B ∼ 7.2 × 1 0 −5 (R / 1km ) + 2.23 × 1 0 −3 (2)
where B is evaluated at r = 50R. From numerical simulations, Crawford
(26) concluded that the magnetic anomalies of the lunar Crisium,
Nectaris, Serenitatis, Humboldtianum, and Mendel-Rydberg basins
could be accounted for by the charge separation–generated magne-
tizations associated with 20-km-radius impactors. Higher-velocity
impactors such as comets can generate even larger fields (27). How-
ever, the charge-separation mechanism is expected to depend on
numerous factors including velocity, the impactor composition,
and impact angle, such that all impacts might not necessarily generate
high fields. This is consistent with lunar observations where
magnetic anomalies are associated with some large craters and not
with others.
We model the impact using the hydrocode iSALE2D (see Materials
and Methods), the findings of which suggest that an impactor 20 to
22m in diameter can form the South Ray crater (fig. S15). Extra-
polation of Eq. 1 to this size of impactor yields fields of ∼18 to
24 T at the edge of the crater, remarkably similar to the 64455
glass paleointensity (e.g., 19.1±3.6 T, mean value derived from
Thellier analyses). This, in turn, implies that the magnetic minerals
in the 64455 glass passed through their blocking temperatures during
flight, consistent with the changing magnetization directions ob-
served from some specimens during demagnetization. We note that
prior paleomagnetic analysis of another young lunar glass, less than
a few million years old and recovered from a 3-m crater, yielded
data defining a nonzero NRM/TRM slope and a nominal field value
of 2.5 T (28). While that study did not use pTRM checks, the
results nevertheless further suggest that ionization from small
impacts can generate substantial magnetic fields.
Thus, fields generated by the impact itself are consistent with the
high paleointensity values from the 64455 glass. The young lunar
surface is a far better environment for recording and preserving
impact magnetizations as compared to Earth (see Materials and
Methods) because of the lack of a background dynamo field. Our
results indicate, more generally, that magnetizations of other plan-
etary bodies can be imparted by impacts, but our findings also have
specific and profound implications for the Moon because the 64455
Apollo impact glass specimens have paleofield strengths that are
comparable to those of the prior 4.2Ga of the nominal lunar
paleointensity record. There are more than a million known craters
on the Moon similar to or larger than the size of South Ray crater
(i.e., ≥1 km) and many thousands created by much larger impactors
that would generate orders of magnitude stronger fields (see Materials
and Methods). The Apollo lunar samples recording strong paleofields
highlighted in prior works, therefore, may record external fields
produced by impacts rather than an ancient core dynamo.
This finding provides motivation to revisit corollary (ii), which
claims that prior measurements of null lunar fields from Apollo
samples are inaccurate. Here, we apply the single silicate crystal
paleointensity technique that strives to isolate samples with better
magnetic properties than bulk samples by eliminating large multi-
domain grains [see Materials and Methods and (29)]. We select
13 crystals [e.g., Fig.3(AandG)] ∼0.5mm in size (range of 0.3 to
1.1 mm) from five basalt samples from the Apollo 17, 14, and
12 missions with ages of ∼3.9Ga (sample 14053), 3.6 Ga (sample
71055), 3.3Ga (samples 12021 and 12040), and 3.2Ga (sample 12053)
(see Materials and Methods). These crystals are plagioclase or
pyroxene; in a few cases (i.e., 14053 and 12021), plagioclase and
pyroxene could not be completely separated, and the specimen
investigated consists of both minerals. The magnetic mineralogy of
lunar bulk rock basalts is mainly native iron together with minor Ni
[<5 weight % (wt %)] or cobalt (<1wt %) (30) and a kamacite
body-centered structure. However, unlike the crystallization of typical
magnetic phenocrysts in terrestrial basalts, this native iron and
ilmenite form from the reduction of a parent ulvöspinel (30).
Clear evidence for magnetic grains that carry laboratory remanences
is seen in magnetic hysteresis data (fig. S16) of both unheated
plagioclase and pyroxene. SEM analyses (see Materials and Methods)
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
5 of 14
of the exact grains used in paleointensity analyses described below
show clear evidence for native iron particles [e.g., Fig.3(B toE)]
with a range of submicrometer grain sizes (figs. S17 to S19). The
iron grains documented in these SEM observations likely represent
the large end of a spectrum that extends to even smaller sizes.
Sometimes, iron particles are found in association with a FeTi phase
and/or troilite [e.g., Fig.3(Hto K)]. Magnetic phases other than
iron are generally subordinate, except in Apollo 71055, where the
200 µm200 µm200 µm
0
5K
10K
15K
20K
Counts
0246810
keV
O
Mg
SiAl
Ca
Fe Fe
Fe
200 nm BSD20 kV
0
4
8
Intensity (×10−12 A m2)
Step
Noise
threshhold
NRM 590 off
590 on, 20 µT
0–1.4 µT
Efficiency, 100.3%
590 tail check
590 on, 40 µT
AB C
DE F
14053
plagioclase
Plg
Py
Plg
Fe
Noise
threshhold
200 µm200 µm200 µm
GH I
JKL
0
4
8
12
16
Intensity (×10−12 A m2)
Step
NRM 590 off
590 on, 20 µT
0–1.5 µT
Efficiency, 101.6%
590 tail check
590 on, 40 µT
0
2K
4K
6K
Counts
0246810
keV
S
O
Mg
Si
Al Ca Fe
Fe
Fe
200 nm BSD20 kV
Fe
Py
FeTi
Tro
12053
pyroxene
Fig. 3. Silicate crystals and TRM experiments. (A to F) Apollo sample 14053 (∼3.9 Ga). (A) Transmission light microscopy photo. (B) Reflected light image of polished
crystal. (C) SEM backscatter image. Plg, plagioclase; Py, pyroxene. The red box shows analysis area (D and E). (D) Elemental spectra [(red dot in (E)]. (E) SEM backscatter
image of analysis area. (F) TRM experiment on grain imaged in (A to E). Intensity versus experimental step. Noise threshold is sensing limit of ultrasensitive superconducting
quantum interference device (SQUID) magnetometer. Field value shown is nominal paleointensity range (see text). Efficiency is calculated from field-on values after
applied fields of 20 and 40 T. (G to L) Apollo sample 12053 (∼3.2 Ga). (G) Transmission light microscopy photo. (H) Reflected light image of polished crystal. (I) SEM back-
scatter image. The red box shows analysis area (J and K). (J) Elemental spectra [red dot in (K)]. (K) SEM backscatter image of analysis area with phases identified. Tro, troilite.
S in spectra is interpreted as contamination from nearby troilite. (L) TRM experiment on grain imaged in (G to K); plot follows conventions in (F).
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
6 of 14
magnetic mineralogy contains more common FeCr phases and troilite.
In this case, the iron is found as small isolated particles and within
troilite grains (fig. S19). Small native iron particles such as those
that we have imaged within the silicate crystals [e.g., Fig.3E and fig.
S18D (2 and 4)] are predicted to be in the single domain (SD) or single
vortex (SV) state and, hence, reliable Thellier paleointensity recorders
(31) on billion-year time scales (6,32–34).
We note that because of the iron formation mechanism (30),
replicating the oxygen fugacity of lunar basalts during paleointensity
experiments is expected to promote the formation of new particles,
inconsistent with reliable paleofield estimation (6). Instead, the
principal requirement for paleointensity analysis using heating is
choosing a method during which the magnetic grains can be con-
sidered stable. Accordingly, we focus on kinetics and select CO2 laser
heatings in air (see Materials and Methods). Specifically, this method
has the dual advantages of best replicating the physical process of
interest, that is, the acquisition of a TRM, whereas the brief CO2
heating, 20 to 50 times shorter than conventional oven heatings, is
least likely to induce chemical or structural change of the magnetic
carriers. We again select 590°C, which is a temperature less than at
which iron might sinter (6) but high enough that a range of unblocking
temperatures are represented. We also note that 590°C is within the
unblocking temperature range of the component identified as carry-
ing the characteristic remanent magnetization (ChRM) in the few
prior studies of Apollo samples using thermal methods.
All the crystals examined had very weak magnetizations (3.9×10−12
to 1.2×10−12 A m2). However, we found that in each case, the mag-
netization after heating to 590°C did not yield consistent directions.
This suggests that any remaining magnetization is (a) at a level below
the magnetometer sensitivity or (b) that there was never a remanent
magnetization imparted, and the NRM value reflects only a spurious
viscous component. Magnetization measuring thresholds (Fig.3,FandL,
and figs. S17 to S20) support interpretation (b). Notwithstanding
this early indicator of null ambient lunar fields, we proceed to fur-
ther investigate the recording fidelity of the crystals by applying a
TRM at 590°C in a 20-T field. Twelve of the crystals acquired a
consistent magnetization. We conclude that the one sample that did
not acquire a remanence does not have recording properties able to
report high fields, but the others do. For these, we can further esti-
mate a maximum paleofield value that could be suggested by the
data, assuming interpretation (a) (see Materials and Methods).
These range from 0.6 to 2.8 T (Fig.3,FandL, and figs. S17 to S20),
but as maxima and considering the caveat of assumption (a), these
small values are indistinguishable from zero (see Materials and
Methods).
As a further test, we reheat each sample to 590°C in zero field
and then in the presence of a 40-T field. The zero-field measure-
ment, when referenced to the first zero-field measurement at 590°C
(see Materials and Methods), show extraordinarily small differenc-
es (1.22 ± 0.97%), indicating a lack of alteration and a dominance of
SD or SV grains, both consistent with our SEM results. The mea-
surement after heating in a 40-T field, when referenced to the in-
tensity measured after the application of the field at 20 T, provides
a way to more directly evaluate the ability of each sample to record
high, dynamo-like fields. Specifically, perfect recorders should yield
a twofold increase in remanent intensity. We find that of the 12 crystals
that recorded a laboratory field, the average efficiency (see Materials
and Methods) of recording this Earth-like field intensity is 92±11%
(Fig.3,FandL, and figs. S17 to S20). Thus, if a high dynamo-like
field had been present on the Moon, then these samples should
have recorded that field, but instead, they carry no appreciable
magnetization.
DISCUSSION
Our five samples, indicative of negligible fields, span in age the prior
suggested episodes of high lunar dynamo and late lunar dynamo
and are supported by results from 11 other Apollo samples (Fig.4
and table S5) within this age range, consistent with null lunar fields.
Together, these data indicate that the Moon lacked any long-lived
dynamo after ∼4Ga. We conclude that if other reported high nom-
inal paleointensity values are not related to strong magnetic interac-
tions, which can result in magnetizations that are not true paleofield
signals (35), they were likely magnetized by a combination of shock
(36) and impact fields. As noted earlier, the charge separation
process depends on several factors, including impactor composi-
tion, velocity, angle, and dust generation. Therefore, not all shocked
rocks are expected to have high imparted magnetizations. How-
ever, numerical simulations (26) indicate that impactors with radii
≤100m are adequate to explain all the nominal “high-field epoch”
values (Fig.4). We note that these high values correspond in time to
the later part of heavy bombardment (see Materials and Methods),
the earlier period being in part or wholly obscured because the crust
is near saturation levels of impacts (Fig.4).
The lack of a long-lived lunar dynamo resolves the numerous
and profound conflicts between the long-lived dynamo posit and
lunar geology, crustal magnetizations, and dynamo driving mecha-
nisms. A thermochemical driven dynamo in the first ∼100Ma of
lunar history is feasible because of rapid cooling and has some sup-
port from crustal anomalies (8), which may reflect a vestige of this
magnetization, complicated by the subsequent complex and intense
impact history of the Moon. Such an early field could have contributed
to the shielding of Earth from the solar wind (3). However, the
Apollo samples examined here indicate that for most of its history,
including times in the Paleoarchean when intense solar forcing could
have led to terrestrial water loss (4,37), the Moon lacked a core dynamo
and thus could not have provided additional magnetic shielding.
The lack of a lunar core dynamo also means that a magneto-
sphere would not have been present in the past to deflect ions (4)
that could contribute to the volatile budget of the lunar surface.
These charged particles would have two principal sources: the
solar wind and Earth’s atmosphere. The transfer from Earth’s
atmosphere would have occurred in the past as today (38) when the
Moon passed through the magnetotail of Earth’s ancient magneto-
sphere (37,39).
Saturation in solar wind volatiles by fine-grained regolith at the
surface may occur more rapidly than the billion-year time scales
considered here (40), and impact gardening can expose soils to the
surface on time scales of hundreds of millions of years (41). However,
some regions of the Moon have regolith >15m deep (41), and these
likely contain buried soils (42) that have not been recycled to the
surface since ∼1 Ga, the end of the erstwhile long-lived lunar dyna-
mo, or much earlier times. We predict that these deep lunar soils
represent a rich volatile reservoir, reflecting a ∼4-Ga-old history of
ion transport that can be explored by new missions such as Artemis.
These volatiles include 3He, water, and nitrogen, which could pro-
vide data on solar wind variability (42) and on the composition of
Earth’s early atmosphere. Overall, the absence of a long-lived dynamo
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
7 of 14
indicates that the accumulation of volatiles was not limited by the
shielding of a paleomagnetosphere, and this favors resource esti-
mates suggesting that billions of kilograms of 3He are preserved in
the lunar regolith (4,43).
MATERIALS AND METHODS
The following sections describe methods and materials for Apollo
samples measured; rock magnetic and light and electron microscope
analyses; prior paleointensity analyses; and associated debate on
methods, paleointensity measurements, and impact modeling and
its interpretation with respect to imparted magnetizations.
Apollo samples analyzed
Crystals were selected such that they lacked large opaque mineral
inclusions that could be multidomain iron. This application differs
from other single-crystal studies (44) where crystals lacking any visible
inclusions are selected. This revised selection criterion is necessary
because of the common occurrence of the opaque mineral ilmenite
in the lunar samples. Here, we seek to limit visible opaque inclu-
sions, but in most cases, these cannot be eliminated entirely. Crystals
from the following samples have been studied.
Apollo 16 sample 64455 is a basaltic impact melt (17) interpreted
to have maintained its orientation since emplacement on the lunar
surface. The ∼5 cm–by–3cm ovoid-shaped sample (fig. S2) consists
of a relatively thick black glass rim with a delicate smooth exterior
covering a basaltic melt interior. The glass is not an impact melt
splash that might refer to melt coating a fixed rock but, instead, a
coating acquired during ballistic transport of the rock and molten
melt produced by the impact (45). The young exposure age is well
constrained by several cosmogenic isotope systems. The 81Kr expo-
sure age is 2.01Ma (46); the 21Ne cosmic ray exposure age is 1.2 Ma,
and the 36Ar age is 1.8Ma (47). The distribution of microcraters and
10Be activity indicate an exposure age of 2Ma (19,48), which is the
generally accepted value for the sample. Apollo 64455 is from a suite
of 22 rocks that are thought to have originated from South Ray
crater that together yield tightly clustered exposure ages of 2 .01±0.1 Ma
(17,49). Given the geology of the collection site, the physical nature
of the sample, the consistency of the cosmogenic ages, the similarity
with other Apollo 16 impact glasses linked to South Ray crater, and
geochemical inferences for a local origin, we concur with prior authors
(45,50) who concluded that it is most probable that an impact at
2Ma formed the South Ray crater and the 64455 glass. It is possible
that future 40Ar-39Ar dating of sample 64455 might help refine its
age. However, in light of the rapid melting-and-quenching thermal
loop experienced by lunar impact glasses, it has been suggested that
insufficient Ar degassing will prevent age resetting (45,51) and, thus,
that bulk-glass 40Ar-39Ar ages will reflect those of the target material.
For instance, in the specific case of Apollo sample 64455, it has been
shown experimentally that the outermost glass layer has a liquidus
1 0.1 0.01 0.001 0.0001
Lunar magnetic and impact history
0
50
100
150
2.53.03.54.04.5
Age (Ga)
10017
10020
10049
12002
70017
71505
71567
76535
62295
15475 35021
04021
12021
14053 56
12 34
70035
71055
76535
Reported magnetic field (µT)
0
50
100
150
Impactor radius R (m)
1 km
10 km
50 km
r = 0.5 km
r = 5 km
r = 50 km
r = 500 km
r = 1000 km
830 µT
2953 µT
3673 µT
Crustal impact saturation [cum. crater frequency (N/km
2
)]
0.51.01.5
15498
15015 15465
0.01
64455
70019
51356
51206
Fig. 4. Lunar magnetic and impact history. Reported field strength measurements from select Apollo samples (table S5) shown as follows: gray filled circles, nonthermal
methods; open circles, no evidence for primary remanence, interpreted as magnetic contamination or results of magnetic interactions/phase changes during analysis;
black squares, data based on thermal analyses. All sample numbers are listed except the following: 1, 68815 (open circle); 2, 62235; 3, 72215; 4, 75055; 5, 60015 (black
square); and 6, 15016. Blue diamonds, thermal analysis values (this work). The right axis shows field impactor radius (R) capable of generating the field intensities by
magnetizations induced by charge separation. Radius (r) values shown by dashed lines are field values at r = 50R. The shaded region reflects the degree of crustal impact
saturation (see Materials and Methods).
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
8 of 14
of ca. 1400°C and that it must have cooled rapidly from this tem-
perature, at approximately 140 K/min, to explain the general lack of
devitrification and results from differential thermal analysis (52).
Given these constraints and using the experimental parameters for
Ar diffusion in basaltic melts (53), a characteristic temperature for
Ar diffusion can be estimated at ca. 1180°C. For a 2-mm-thick melt
layer, which is the minimum observed in sample 64455 (54), a total
loss of Ar less than 3% is predicted, with substantial age resetting
(>50%) restricted to only the outermost ca. <40 m of the glass.
Thus, only the outermost surface of the glass not substantially ablated
by micrometeorites might preserve a glass formation age. In our
paleointensity analyses, we focus on glass subsamples (Fig.1A) from
NASA Apollo sample “64455,24” (fig. S2) taken from the bottom of
64455, sheltered from micrometeorites (17).
Apollo 14 sample 14053,262 is a coarse-grained high-Al basalt,
which is unusual relative to other lunar basalts because it is reduced
during what has been interpreted to be a secondary event such as
residence in an ejecta blanket (55). Specifically, our sample is from
the outer, reduced portion of the 14053. It was proposed that 14053
represents an impact melt (56–57), but detailed trace element anal-
yses indicate that it crystallized from a primary magma (58–59). An
39Ar-40Ar plateau age of 3.94Ga (60) has been reported for 14053,
as well as a Rb-Sr age of 3.96Ga (61). On the basis of the similarity
of these ages, we follow (62) in concluding that the inferred high-
temperature reduction event occurred close to, or at, the time of the
Ar-Ar plateau age (i.e., 3.94 Ga).
Several prior works discuss magnetizations from bulk samples of
14053 (36, 63–67). The work in (36) focuses on a magnetization
thought to be held at low unblocking temperatures (<300°C), and
the authors conclude that the magnetization of their bulk samples
was possibly carried by cohenite [(Fe,Ni,Co)3C]. On the basis of a
series of hydrostatic loading experiments to explore a piezoremanent
magnetization that might mimic a shock remanent magnet ization,
these authors suggested that the NRM might be a shock remanent
magnetization acquired in a field of 40 to 60 T. In (67), a “partial
TRM” recording of a somewhat lower strength field (20-T field)
was offered as an alternative interpretation of the data.
However, the suggestion of a cohenite carrier contrasts with early
studies that highlight the relatively high iron content of 1.02wt %
and interpretations that multidomain native iron carriers were
present (62–63). The presence of an iron carrier is strongly supported
by the definition of maximum unblocking temperatures between
750° and 780°C (63). These contrasting interpretations may, at least
in part, reflect the location of different subsamples analyzed from
sample 14053; the main mass shows differences in the degree of
reduction (55).
The differences between the analyses conducted here and those
of previous studies of 14053 extend beyond the potential for speci-
men level differences in magnetic behavior. Our analyses are from
small silicate crystals versus bulk samples and thus exclude large
multidomain grains. This may explain the absence of any strong
apparent magnetization seen in our specimens. Namely, the bulk
samples may predominately record either spurious magnetiza-
tions or preferentially record shock remanent magnetization
be cause their magnetic mineral assemblages are dominated by multi-
domain grains.
Apollo 12 sample 12021,30 is a coarse-grained porphyritic pigeonite
basalt with large (up to 10 mm) pyroxene phenocrysts (68–69). It
has a 3.3-Ga age based on Rb-Sr analyses (70–71).
Apollo 12 sample 12053,283 is a porphyritic pigeonite basalt
(72–73). A whole rock Ar-Ar plateau yields an age of 3.17Ga (74).
Apollo 12 sample 12040,209 is a coarse-grained olivine basalt
(69,75); melt inclusions have been reported in silicate grains (76). It
has a Rb-Sr age of 3.3Ga (71,77).
Apollo 17 mare basalt 71055,2 is a “vesicular, fine- to medium-
grained olivine-bearing ilmenite basalt” (78). It has a Rb-Sr age of
3.6Ga (79).
Rock magnetic methods and analyses
Magnetic hysteresis data, including FORC data (80) were collected
using a Princeton Measurement Corporation Model 2900 Alternating
Gradient Force Magnetometer at the University of Rochester. FORC
data were smoothed (81–82) using FORCinel version 3.01 and
VARIFORC software.
Light and electron microscope methods and observations
Light stereomicroscopy was performed with a Nikon SMZ800 with
a trinocular head, a maximum ×630 magnification, and a Spot
Insight 4MP CCD color digital camera assembly. A Nikon Eclipse
LV100POL was also used for both transmitted and reflected light
microscopy. Glass subsamples and basalt single silicates were pre-
pared in polished acrylic mounts, carbon-coated, and examined
using a Zeiss Auriga SEM with an energy dispersive x-ray analysis
(EDAX) energy dispersive spectrometer at the University of Rochest er
Integrated Nanosystems Center. Our Apollo 64455 glass sample is
relatively uniform, and therefore, the subsamples that we selected for
SEM analyses should be representative of those used for paleointensity
analyses. Single silicate crystals from other Apollo samples are, in
comparison, more variable in composition. Thus, we conducted an
SEM examination of a specific crystal from each sample used in
paleointensity analyses after the four brief (90 s) 590°C thermal
treatments. We note that there is no textural evidence in our light
microscope or SEM observations to indicate that the Apollo 64455
Fe-Ni-S spheres were incorporated into the glass after its formation.
Hence, these spheres are primary magnetic inclusions. Our SEM
observations on the lunar single silicate grains show an occurrence
of magnetic inclusions that are similar to magnetic grains seen in
bulk lunar basalt samples (30) but with sizes that are orders of
magnitude smaller. Hence, the single silicates are more suitable than
bulk samples for recording paleointensities (29,31).
Prior paleointensity analyses: Debates over methods
and interpretations
It is commonly accepted that to obtain accurate past field records,
lunar samples carrying TRMs should be sought (83). However, what
may be underappreciated is that most available sample data used to
estimate lunar paleointensity rely on nonthermal methods that do
not directly test Thellier’s laws and, hence, the presence of a TRM
(6). Accordingly, there has been considerable concern over the use
of nonthermal methods to estimate paleointensity (9), to the point
that they have been called “the methods of last resort” (6). The con-
cern is punctuated by the multiple assumptions that must be made
estimating a TRM quantity (i.e., paleointensity) using a process dif-
ferent from that which imparted any original magnetization.
Nonthermal methods ultimately rely on an assumption that
magnetic coercivities explored by the application of laboratory AFs
equate with magnetic blocking temperatures. Ideally, this need not
be an assumption if quantitative information on magnetic domain
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
9 of 14
state distributions is available and linked uniquely to magnetic co-
ercivities and blocking temperatures, but this quantitative informa-
tion and attendant linkages are unavailable.
In the absence of a core dynamo, lunar samples might still hold
a viscous magnetization and an NRM; that is, the sample might
record a measurable magnetic moment before demagnetization.
However, a key discriminating property is that in the absence of a
core dynamo, samples should not yield a ChRM, which is a compo-
nent obtained after demagnetization of any viscous contamination.
Thus, a signature of the lack of a core dynamo is the observation of
directional instability after viscous components have been reported
after relatively low AF or thermal demagnetization treatments. Several
prior studies have reported unstable AF and/or thermal demagnet-
ization behavior of lunar samples, consistent with the lack of a field.
However, bulk lunar magnetic samples typically contain large
multidomain magnetic (MD) grains, and these are thought to be
unstable during the application of AFs, demagnetization, or the ap-
plication of anhysteretic fields [anhysteretic remanent magnetizations
(ARMs)]. Hence, separating the influence of laboratory-induced
noise from evidence for a null remanence is often not straightfor-
ward when nonthermal methods (i.e., AFs) are used.
ARM data have been used to assess the recording reliability of
lunar samples (84). In this method, an ARM is applied at different
bias fields and then used to compute paleointensities, assuming a
calibration (85). For some samples, the quality of the paleointensity
determination and its agreement with the known applied field
decreased as the magnitude of the applied field was decreased. The
applied field value when differences between the expected and ob-
tained paleointensities are greater than 100% and/or the errors in
the paleointensity exceed 100% is called a minimum paleointensity
that can be recorded by the sample using this ARM method (84).
The study outlining this approach includes a discussion of the
limitations of the equipment used for AF demagnetization and
ARM acquisition and the problems with harmonics in the signals
(84). However, while the noise introduced into measurements is
well documented, the extension of the ARM measurements to the
general conclusion that a given sample cannot record a field below
the ARM method minimum (86), or, furthermore, that such sam-
ples provide no evidence for the absence of a core dynamo (87), is
not justified as we explain below.
Within the context of a planetary body where the primary ques-
tion is the absence or presence of an internally generated magnetic
field, the lack of a characteristic magnetization is, to first order, the
evidence for the lack of a field, provided that the sample contains
magnetic grains capable of recording fields on the requisite time scales
(i.e., equal to or older than the age of the sample in question). To
demonstrate that the lack of a stable magnetization is not evidence
for the lack of a magnetic field, one would need to prove that there
are no grain sizes/domain states present that could record and re-
tain that field. Although multidomain grains are reported (but not
illustrated) in microprobe analyses reported in (84), FORC diagrams
from the same samples clearly indicate another pseudosingle domain
or single vortex component [figure S5 of (84)]. These are grains that
could retain fields on the billion year time scales relevant to the Moon.
If MD grains are present, the application of AF tends to
exace rbate experimental noise. The ARM work of (84), in which
experimental noise is present and acknowledged, has been further
extended to claim that some lunar samples can only record a field as
low as that defined by their ARM error analysis [in the case of Apollo
15016, the claimed minimum is 37 T; (86)]. Instead, these experi-
ments show only that ARM methods are poorly suited for robust
paleointensity estimates in the samples. The rock magnetic demon-
stration that magnetic grains capable of recording a TRM are present
and the lack of a ChRM together suggest that these samples passed
through their blocking temperatures or were shocked in the absence
of an ambient field. Therefore, in our summary of magnetic direc-
tions, we include a select set of samples analyzed by prior authors
(cf., table S5) where no ChRM was present.
Two prior debatable interpretations figure largely in the posit of
a long-lived dynamo. These are the oldest and youngest samples
proposed to record the field. For the youngest sample, specimens
record both a measurable paleofield and no field (14). The sample
investigated (Apollo 15498) is a complex impact breccia with basaltic
clasts, a glass matrix, fissures filled with vesicular glass, and a coat-
ing of “splash” glass a few to 6mm thick (88). There are different
interpretations of the origin of the matrix glass. In one, this results
from in situ high-pressure shock that is evident by a wide variety of
shock features (e.g., shock lamellar structures), with the lack of
unshocked clasts providing evidence for an in situ origin (89). In
another interpretation, the glass matrix is thought to have originated
as an impact melt that underwent high-temperature rapid cooling
followed by a slower cooling (90), with evidence for this process
rather than shock provided in the form of experimental analyses of
annealing characteristics (although no counter argument against the
absence of unshocked clasts is provided).
A paleointensity of a few microteslas using a modified Thellier
technique was originally reported on this sample (91) and confirmed
in the restudy (14), which focused on glassy matrix samples, but in
the restudy, specimens within 2cm of the contact with the splash
glass lacked a ChRM. These specimens were interpreted to have been
demagnetized in a later null field (14) and that only interior samples
recorded a core dynamo. It was argued that the remanence of these
interior specimens was acquired slowly, on hour time scales, i.e.,
longer than the time scales of magnetization by impacts that were
claimed to be <∼1 s (specifically for impacts after 3.3 Ga; Supple-
mentary Materials) (14). However, the remanence acquisition and
impact field time scales are incorrect, as described below.
Irrespective of uncertainty over its mode of formation described
earlier, there is an agreement that the glass matrix cooled very rap-
idly to temperatures as low at ∼620°C. In (14), a conductive cooling
model is used to conclude an hour time scale to reach ambient lunar
conditions, but this end point is not relevant to the magnetization of
15498. Instead, the relevant time is only that to span the blocking
temperature range represented by the ChRM. In reference to the
data highlighted [figure 7 of (14)], a “high temperature” component
is defined that appears to span temperatures from high to low tem-
peratures, but this component does not correspond to the temperatures
used in the paleointensity estimates. Specifically, the component
yielding a nonzero field is isolated only after heating above 560°C. The
lowest unblocking temperature where a field is recorded is within
∼60°C of the nominal temperature change from extremely rapid to
slow cooling. Given the uncertainties in these analyses related to the
unquantified complexities of the cooling (90) and the uncertainties
in uniquely relating magnetic unblocking to ambient temperatures
in the breccia, we consider this difference to be within error, and we
thus conclude that the glass matrix from sample 15498 could have
acquired its magnetization on minute time scales. Moreover, we
note that the paleointensity isolated at lower temperatures (250° to
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
10 of 14
540°C) is essentially null (0.2±0.1 T). This is interpreted as a par-
tial shock demagnetization, at a shock level that left no evidence of
the event (14). Instead, this change from a magnetized to unmagnetized
sample could represent either the rapid decay of a transient field pro-
duced by an impact by charge separation and/or the physical trans-
port of the sample out of the range of a strong magnetizing field.
The dismissal of impact magnetizations based on a lower frequency
of large impacts after 3.3Ga in (14) is inconsistent with the potential
crater sources. Specifically, both Aristillus (diameter, 55 km) and
Autolycus crater (diameter, 39 km) have been discussed as sources
of ejecta and secondary craters near the Apollo 15 site, where sample
15498 was collected (92). These craters are very far from the collec-
tion site (>180 to 130 km), so quenching of the glass during trans-
port and magnetization is likely; impactors responsible for these
craters could have generated fields many hundreds of microteslas in
strength through charge separation (26–27), which can explain the
observed nonzero paleointensity values. We again note that the outer-
most glass of 15498 is unmagnetized and is interpreted in (14) to
record a separate, later event in a null field. Alternatively, the glass
emplacement may be related to the final emplacement of the sample
in a secondary impact insufficient to drive substantial charge separa-
tion. Thus, we conclude that the magnetization characteristics of
15498 provide no conclusive evidence for a dynamo at 1.5Ga but
instead are better explained by impact processes.
The oldest sample purported to record a lunar dynamo, the
4.2-Ga-old coarse-grained troctolite 76535 (93), has a history in-
herently related to one or more impacts because these are needed to
bring the sample to the surface from a great depth (94). The magne-
tism of this sample was first studied in (9), and it was noted that the
removal of almost 80% of the signal occurred by 540°C, with a uni-
directional signal. However, pTRM checks failed at high and low
temperatures, and at high temperatures (the experiment was ceased
at 770°C), the sample was not losing NRM/gaining pTRM in a pat-
tern consistent with a TRM. Similarities in the magnetic behavior
were noted relative to other lunar samples, specifically where NRM/
TRM characteristics at low temperature were linked to magnetic
interactions and where those at higher temperatures were due to the
formation of new iron phases. This behavior suggests that the signal
might not be an accurate recorder of any ambient lunar field.
Subsequently, paleointensities were reported for 76535 using
nonthermal techniques, where it was argued that the slow cooling
and magnetization required a core dynamo (10). Ambiguity in the
internal consistency of specimens studied motivated a second study
using similar nonthermal techniques (95). The identified high-
coercivity component is very noisy; the directions define a cloud of
points with numerous instances where magnetization increases
rather than decreases with demagnetization, only to decrease at the
next demagnetization step. This appears to reflect the acquisition
and subsequent removal of AF artifacts, but after so many of these
spurious signals are imparted to each specimen [>30, figure 5 of
(95)], there are concerns over the meaning of any derived direction.
Concomitant with this noise, we see that the method of constraining
a fit to the data such that it must pass through the origin of an
orthogonal vector plot has a very large influence on the assigned
uncertainty. Without this constraint, the nominal high coercivity
components assigned to two of the three specimens studied have
median angular dispersions so high (41° and 32°) that they would
not be acceptable in studies of terrestrial materials. The third speci-
men yields a high, marginal value (29°). The high uncertainties of
the fits prevent any conclusive determination that the specimens
record a common direction indicative of a TRM.
Beyond the low quality of the AF directional data, major unre-
solved issues revolve around equating the coercivity spectra of the
specimens with the apparent thermal unblocking defined in (9) and
whether any signal could record an impact field. Specifically, in (10)
and (95), the nominal high coercivity component is related to very
high blocking temperatures typical of kamacite and the unblocking
temperatures reported in (9). However, in (9), it was cautioned that
the characteristics might relate to iron formation by ilmenite reduc-
tion in the laboratory. Moreover, the textural evidence for slow
cooling in 76535 does not require any magnetization that it holds to
be acquired over long durations, contrary to claims in (10) and (95).
A large impact is needed to bring 76535 to the surface. If this impact
occurred at ∼4.2 Ga, then the sample could have been exposed to a
large field produced by charge separation, with its magnetic miner-
als rapidly passing through their Curie temperatures and acquiring
a magnetization in the absence of a core dynamo.
Paleointensity methods and analyses
Glass samples and single silicate crystals were mounted in 2 mm–
by–2 mm–by–2mm fused quartz boxes and set with a minimum of
sodium silicate solution for all remanence measurement. The purity
of these materials has been documented by use of a scanning super-
conducting quantum interference device (SQUID) microscope (96–97).
Paleomagnetic measurements were made using an ultrahigh-
resolution 6.3-mm-bore William S. Goree. Inc. (WSGI) three-component
DC SQUID magnetometer in the magnetically shielded room at
the University of Rochester (ambient field, <200 nT). This magne-
tometer affords an order of magnitude greater sensitivity than other
2G SQUID magnetometers.
Thermal analyses
For TTRM (98) and Thellier-Coe experiments of 64455 glass, spec-
imens were heated in air using a Firestar V20 CO2 laser (also in the
University of Rochester’s magnetically shielded room). Thermal
paleointensity techniques follow those developed for single-crystal
paleointensity analysis (29,39,44,97,99). The heating time used to
evaluate alteration using magnetic hysteresis was 3 min. Heating
times for each paleointensity step were either 90 s (subsamples
<1mm in size) or 120 s (subsamples 1 to 2mm in size). For one
TTRM experiment (subsample ss40), we used a three-point sliding
window for the orthogonal vector plots to reduce noise and identify
the temperature range of the ChRM. For Thellier data, we use the
following reliability criteria (99). A sample is deemed successful if
there is a linear relationship between the loss of NRM and the ac-
quisition of a laboratory-induced magnetization (R2 value generally
greater or equal to 0.9). Four or more points should define the best-
fit line. NRM-TRM points should be evenly distributed along the
best-fit line, and pTRM checks should fall within 15% of the origi-
nal value. The maximum angular deviation should be less than
15 degrees, and the field-off steps should not trend in the direction
of the applied field. We have relaxed these criteria somewhat for our
lunar results relative to terrestrial samples (allowed for greater
maximum angular deviation (MAD) angles and deviation of the
pTRM checks), and for two subsamples (ss31 and ss42), we used a
three-point sliding window for the orthogonal vector plots to reduce
noise and identify the ChRM temperature range.
For TRM analyses of Apollo basalt silicates at 590°C, we use the
specimen preparation, CO2 laser, and magnetometer as described
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
11 of 14
above, with all heating times at 90 s. These measurements are simi-
lar to those conducted at 565°C on terrestrial zircons (39,97) but
differ in one key way. At 565°C, terrestrial zircons have a stable
magnetization with a paleointensity that is within a factor of two of
Thellier data that uses the full unblocking spectra of the ChRM. In
contrast, the lunar silicate crystals examined here lack a ChRM that
should otherwise be represented by stable magnetic direction after
heating at 590°C (in zero field). After heating to 590°C in 20 T,
we reheat the lunar silicates in zero field to conduct an MD-tail test
[see Methods in (39)]. After heating lunar silicates in 40 T, we de-
fine the TRM efficiency as
M 590,40T
─
M 590,20T × 2 × 100
where M590,40T and M590,20T are the magnetizations imparted in
applied fields of 40 and 20 T, respectively.
Nonthermal analyses
Nonthermal paleointensity use the REM’ method, which is thought
to be best suited for samples that might show multicomponent
magnetizations (100). The slope of NRM data demagnetized by AF
was normalized by the slope of demagnetization data of the SIRM
data. The demagnetization segment chosen for paleointensity deter-
mination is that which defines the component deemed to be primary.
After measurement of the NRM, samples were AF-demagnetized up
to 300 mT. The order of AF demagnetization axes with progressively
higher peak fields was permutated (101) to counter any acquisition of a
gyroscopic remanent magnetization. Following demagnetization of
the NRM, the sample was given an SIRM in a 3-T field using an ASC
Scientific Impulse Magnetizer. The initial SIRM was measured, fol-
lowed by AF demagnetization using the same step procedure used
and axis permutations for the AF demagnetization of the NRM. A
smoothing-interpolation method (101) was applied to the NRM
and IRM demagnetization data, again to mitigate any effects of
GRM. A comparison of the loss of NRM to the loss of IRM [ratio of
equivalent magnetizations (REM)] can be used to estimate for
paleointensity of the sample. Orthogonal vector plots of NRM and
IRM demagnetization were used to determine the AF demagnetization
range of the component of magnetization most likely to be of primary
origin. Calibration compilations (100,102–103) suggest that the pa-
leofield (Bo, in tesla) is equal to the ∼3.01 × 10−3 REM for single to
multidomain magnetite and titanomagnetite. FeNi alloys and lunar
samples have been interpreted to be compatible with this trend (100).
In considering uncertainties for our nonthermal analyses, we
follow the usage in prior works that assign a factor of two uncertain-
ties to calibrations. We view this as a minimum uncertainty. For the
64455 glass, there are other calibration data that might be applied
on the basis of experiments producing small iron spheres (see table
S4). These calibrations yield a different range (low field bound of
4 versus 10 T; high field bound of 82 versus 89 T) that does not
affect the conclusions here.
Impact modeling, interpretations, and history
Modeling of South Ray crater was done using the code iSALE2D
(104–105). Our input files are included as data files S1 and S2. We
use a dunite impactor, which is also assumed in the Crawford model
(26). We choose granite as a target material because it has proper-
ties more similar to lunar anorthosite and better approximates the
target considered in Crawford (26). The impact velocity (14 km/s)
is chosen as the vertical component of an impact velocity of 20 km/s
with an impact angle of 45°. Impacts can create charge separation
(26,106–107) because they generate a combination of debris and
ionized gas (plasma). The probability of electron interaction with
impact debris is higher than that for ions, and, in turn, more elec-
trons bind to the debris, making it negatively charged. As the debris
leaves the impact site, it carries away this negative charge, leaving a
slightly positively charged plasma. The net charge increases with the
impactor mass and velocity. This charge then produces an electric
field, which, in turn, drives a current that induces a magnetic field;
experiments corroborate this effect (107). Detailed simulations
confirm (26) the mechanism outlined above and provide the scaling
for the amplified surface field.
A recent modeling study (25) of the hypothesis whereby impacts
generate antipodal magnetic anomalies by compression of the solar
wind magnetic field (24) also comments on the charge separation
process. Specifically, the work in (25) cites four papers (108–111)
and states “numerous paleomagnetic investigations of impact cra-
ters on the Earth have found that impact-heated rocks record the
background field and found no evidence of an amplified or locally
generated transient field.” This statement does not properly repre-
sent the cited literature. First, in the study of the Vredefort impact
structure cited (108), the authors argue that lightning remagnetiza-
tions prevent recognition of impact magnetizations. Second, in a
review of crustal anomalies from several terrestrial impact craters,
the authors of (109) note that there are anomalous high signals but
that these might be explained by high ferromagnetic mineral con-
tents. Otherwise, the work focuses on melt rocks, specifically with
the goal of determining whether impacts could affect the geodynamo
rather than testing with paleointensity analyses whether impact fields
are recorded. However, in a discussion of the small ∼1.8-km-diameter
Lonar crater of India, the authors of (109) note that evidence of shock
remanent magnetization is “hotly debated.” They further note that
in a magnetic study of the Lonar crater (110), the third study cited
by (25), the subsequent acquisition of viscous and/or chemical mag-
netization in Earth’s field prevented recognition of a shock compo-
nent. Therefore, rather than commenting on the charge separation
magnetization process, these three papers instead explain why it is
so difficult to recognize impact magnetization on Earth. One re-
corder that might be able to record such fields is impact glasses,
magnetized on short times similar to those of Apollo 64455. The
fourth cited work (111) studied tektites from the Lonar crater, which
might record such fields. There are sampling shortcomings in the
work in that some samples appear to be weathered and/or are not
pure glass. In addition, the method chosen, NRM/SIRM (versus REM’),
is thought to provide only order-of-magnitude estimates of
paleointensity (6). Nevertheless, the authors concluded that no fields
>∼100 T were observed, that these were orders of magnitude less
than those predicted by (107), and that, therefore, the Lonar tektites
provided a counter example to locally strong impact-induced fields
resulting from the charge separation process. However, in a more
recent study (26), the electrostatic charge model was revised with
Mg+ rather than Ca+ being the dominant ion source; the higher
ionization energy ultimately results in lower field values. A 35-m-radius
impactor is called upon to form the Lonar crater (111). Charge sep-
aration (Eq. 1) predicts fields of 29 T at 50R. The recent field
intensity at the site is ∼44 T, which was probably similar during
the time of the impact, dated at ∼0.57Ma (112) during the Brunhes
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
12 of 14
chron. Thus, while some of the values reported in (111) from their
“large” samples nominally agree with the charge separation pre-
dictions of (26), the similarity of the predicted and geodynamo
fields and the inaccuracy of the paleointensity method applied pre-
vent any meaningful test. Overall, while the terrestrial environment
is challenging for examining charge separation given the back-
ground field and other crustal process that can enhance bulk mag-
netic mineral content at impact sites (e.g., hydrothermal circulation),
we hope that our results from Apollo 64455 will motivate new and
more detailed magnetic examinations of impact craters and ejecta
on Earth.
The abundance of small and large lunar craters that we refer to is
derived from (113) and (114). We rely on (115) and (116) to derive
the impact frequency shown in Fig.4.
SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/
content/full/7/32/eabi7647/DC1
REFERENCES AND NOTES
1. S. M. Cisowski, D. W. Collinson, S. K. Runcorn, A. Stephenson, M. Fuller, A review of lunar
paleointensity data and implications for the origin of lunar magnetism. J. Geophys. Res.
88, A691–A704 (1983).
2. A. J. Evans, S. M. Tikoo, J. C. Andrews-Hanna, The case against an early lunar dynamo
powered by core convection. Geophys. Res. Lett. 45, 98–107 (2018).
3. J. Green, D. Draper, S. Boardsen, C. Dong, When the Moon had a magnetosphere. Sci. Adv.
6, eabc0865 (2020).
4. J. A. Tarduno, E. G. Blackman, E. E. Mamajek, Detecting the oldest geodynamo
and attendant shielding from the solar wind: Implications for habitability. Phys. Earth
Planet. Inter. 233, 68–87 (2014).
5. B. Fegley Jr., T. D. Swindle, Lunar volatiles: Implications for lunar resource utilization, in
Resources of Near Earth Space, J. Lewis, M. S. Matthews, M. L. Guerrieri, Eds. (University of
Arizona Press, 1993), pp. 367–426.
6. D. J. Dunlop, Ö. Özdemir, Rock Magnetism: Fundamentals and Frontiers (Cambridge Univ.
Press, 2001).
7. S. M. Cisowski, C. Hale, M. Fuller, On the intensity of ancient lunar fields. Proc. Lunar Sci.
Conf. 8, 725–750 (1977).
8. M. A. Wieczorek, Strength, depth, and geometry of magnetic sources in the crust on the
Moon from localized power spectrum analysis. J. Geophys. Res. 123, 291–316 (2018).
9. K. Lawrence, C. Johnson, L. Tauxe, J. Gee, Lunar paleointensity measurements:
Implications for lunar magnetic evolution. Phys. Earth Planet. Inter. 168, 71–87 (2008).
10. I. Garrick-Bethell, B. P. Weiss, D. L. Shuster, J. Buz, Early lunar magnetism. Science 323,
356–359 (2009).
11. E. K. Shea, B. P. Weiss, W. S. Cassata, D. L. Shuster, S. M. Tikoo, J. Gattacceca, T. L. Grove,
M. D. Fuller, A long-lived lunar core dynamo. Science 335, 453–456 (2012).
12. M. Le Bars, M. A. Wieczorek, O. Karatekin, D. Cébron, M. Laneuville, An impact-driven
dynamo for the early Moon. Nature 479, 215–218 (2011).
13. C. A. Dwyer, D. J. Stevenson, F. Nimmo, A long-lived lunar dynamo driven by continuous
mechanical stirring. Nature 479, 212–214 (2011).
14. S. M. Tikoo, B. P. Weiss, D. L. Shuster, C. Suavet, H. Wang, T. L. Grove, A two-billion-year
history for the lunar dynamo. Sci. Adv. 3, e1700207 (2017).
15. M. Laneuville, M. A. Wieczorek, D. Breuer, J. Aubert, G. Morard, T. Rückriemen,
A long-lived lunar dynamo powered by core crystallization. Earth Planet. Sci. Lett. 401,
251–260 (2014).
16. S. Mighani, H. Wang, D. L. Shuster, C. S. Borlina, C. I. O. Nicols, B. P. Weiss, The end
of the lunar dynamo. Sci. Adv. 6, eaaz0883 (2020).
17. G. Ryder, M. D. Norman, Catalog of Apollo 16 rocks: Part 2 63335-66095 (Curatorial Branch
Publication 52, NASA JSC 16904, 1980).
18. A. G. Sanchez, D4. Geology of Stone Mountain, in Geology of the Apollo 16 Area, Central
Lunar Highlands, in Geological Survey Professional Paper 1048, G. E. Ulrich, C. A. Hodges,
W. R. Muehlbergerm, Eds. (U.S. Gov. Print. Office, 1981), pp. 106–126.
19. K. Nishiizumi, C. P. Kohl, J. R. Arnold, R. C. Finkel, M. W. Chaffee, J. Masarik, R. C. Reedy,
Final results of comogenic nuclides in lunar rock 64455. Lunar Planet. Sci. Conf. 26,
1055–1056 (1995).
20. L. Tauxe, T. A. T. Mullender, T. Pick, Potbellies, wasp-waists, and superparamagnetism
in magnetic hysteresis. J. Geophys. Res. 101, 571–583 (1996).
21. M. Fuller, Lunar magnetism. Rev. Geophys. Space Phys. 12, 23–70 (1974).
22. P. Dyal, C. W. Parkin, W. D. Dailey, Magnetism and the interior of the Moon. Rev. Geophys.
Space Phys. 12, 568–591 (1974).
23. H. Tsunakawa, F. Takahashi, H. Shimizu, H. Shibuya, M. Matsushima, Surface vector
mapping of magnetic anomalies over the Moon using Kaguya and Lunar Prospector
observations. J. Geophys. Res. 120, 1160–1185 (2015).
24. L. L. Hood, N. C. Richmond, P. D. Spudis, Origin of strong lunar magnetic anomalies:
Further mapping and examinations of LROC imagery in regions antipodal to young large
impact basins. J. Geophys. Res. 118, 1265–1284 (2013).
25. R. Oran, B. P. Weiss, Y. Shprits, K. Miljković, G. Tóth, Was the moon magnetized by impact
plasmas? Sci. Adv. 6, eabb1475 (2020).
26. D. A. Crawford, Simulations of magnetic fields produced by asteroid impact: Possible
implications for planetary paleomagnetism. Int. J. Impact Eng. 137, 103464 (2020).
27. M. Bruck Syal, P. H. Schultz, Cometary impact effects at the Moon: Implications for lunar
swirl formation. Icarus 257, 194–206 (2015).
28. N. Sugiura, Y. M. Wu, D. W. Strangway, G. W. Pearce, L. A. Taylor, A new magnetic
paleointensity value for a ‘young lunar glass’. Proc. Lunar Planet. Sci. Conf. 10, 2189–2197
(1979).
29. J. A. Tarduno, R. D. Cottrell, A. V. Smirnov, The paleomagnetism of single silicate crystals:
Recording the geomagnetic field during mixed polarity intervals, superchrons and inner
core growth. Rev. Geophys. 44, RG1002 (2006).
30. J. Papike, L. Taylor, S. Simon, Lunar minerals, in Lunar Source Book, G. H. Heiken,
D. T. Vaniman, B. M. French, Eds. (Cambrige Univ. Press, 1991), chap. 5, pp. 137–153.
31. A. V. Smirnov, E. V. Kulakov, M. S. Foucher, K. E. Bristol, Intrinsic paleointensity bias
and the long-term history of the geodynamo. Sci. Adv. 3, e1602306 (2017).
32. A. R. Muxworthy, W. Williams, Critical single-domain grain sizes in elongated iron
particles: Implications for meteoritic and lunar magnetism. Geophys. J. Int. 202, 578–583
(2015).
33. T. P. Almeida, A. R. Muxworthy, A. Kovács, W. Williams, P. D. Brown, R. E. Dunin-Borkowski,
Direct visualization of the thermomagnetic behavior of pseudo–single-domain
magnetite particles. Sci. Adv. 2, e1501801 (2016).
34. L. Nagy, W. Williams, L. Tauxe, A. R. Muxworthy, I. Ferreira, Thermomagnetic recording
fidelity of nanometer-sized iron and implications for planetary magnetism. Proc. Natl.
Acad. Sci. U.S.A. 116, 1984–1991 (2019).
35. T. O’Brien, J. A. Tarduno, A. Anand, A. V. Smirnov, E. G. Blackman, J. Carroll-Nellenback,
A. N. Krot, Arrival and magnetization of carbonaceous chondrites in the asteroid belt
before 4562 million years ago. Commun. Earth Environ. 1, 54 (2020).
36. J. Gattacceca, M. Boustie, L. Hood, J. P. Cuq-Lelandais, M. Fuller, N. S. Bezaeva,
T. De Resseguier, L. Berthe, Can the lunar crust be magnetized by shock: Experimental
groundtruth. Earth Planet. Sci. Lett. 299, 42–53 (2010).
37. J. A. Tarduno, R. D. Cottrell, M. K. Watkeys, A. Hofmann, P. V. Doubrovine, E. E. Mamajek,
D. Liu, D. G. Sibeck, L. P. Neukirch, Y. Usui, Geodynamo, solar wind, and magnetopause
3.4 to 3.45 billion years ago. Science 327, 1238–1240 (2010).
38. K. Terada, S. Yokota, Y. Saito, N. Kitamura, K. Asamura, M. N. Nishino, Biogenic oxygen
from Earth transported to the Moon by a wind of magnetospheric ions. Nat. Astron. 1,
0026 (2017).
39. J. A. Tarduno, R. D. Cottrell, W. J. Davis, F. Nimmo, R. K. Bono, A Hadean to Paleoarchean
geodynamo recorded by single zircon crystals. Science 349, 521–524 (2015).
40. F. Hörz, R. Grieve, G. Heiken, P. Spudis, A. Binder, Lunar surface processes, in Lunar Source
Book, G. H. Heiken, D. T. Vaniman, B. M. French, Eds. (Cambrige Univ. Press, 1991), chap 4,
pp. 61–120.
41. H. J. Melosh, Planetary Surface Processes (Cambridge Univ. Press, 2011).
42. S. A. Fagents, M. E. Rumpf, I. A. Crawford, K. H. Joy, Preservation potential of implanted
solar wind volatiles in lunar paleoregolith deposits buried by lava flows. Icarus 207,
595–604 (2010).
43. W. Fa, Y.-Q. Jin, Quantitative estimation of helium-3 spatial distribution in the lunar
regolith layer. Icarus 190, 15–23 (2007).
44. J. A. Tarduno, R. D. Cottrell, M. K. Watkeys, D. Bauch, Geomagnetic field strength 3.2
billion years ago recorded by single silicate crystals. Nature 446, 657–660 (2007).
45. R. V. Morris, T. H. See, F. Hörz, Composition of the Cayley formation at Apollo 16
as inferred from impact melt splashes. J. Geophys. Res. 17, E21–E42 (1986).
46. J. R. Arnold, C. P. Kohl, K. Nishiizumi, M. W. Caffee, R. C. Finkel, J. R. Southon,
Measurements of cosmogenic nuclides in lunar rock 64455, in 24th Lunar and Planetary
Science Conference (SAO/NASA Astrophysics Data System, 1993), 39 pp.
47. D. D. Bogard, E. K. Gibson Jr., Volatile gases in breccia 68115. Lunar Planet. Sci. Conf. 6,
63–65 (1975).
48. K. Nishiizumi, J. R. Arnold, C. P. Kohl, M. W. Chaffee, J. Masarik, R. C. Reedy, Solar cosmic
ray records in lunar rock 64455. Geochim. Cosmochim. Acta 73, 2163–2176 (2009).
49. O. Eugster, Chronology of dimict breccias and the age of South Ray crater at the Apollo
16 site. Meteorit. Planet. Sci. 34, 385–391 (1999).
50. T. H. See, F. Hörz, R. V. Morris, Apollo 16 impact-melt splashes; petrography and major-
element composition. Proc. Lunar Planet. Sci. Conf. 91, E3–E20 (1986).
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
13 of 14
51. B. A. Cohen, T. D. Swindle, D. A. Kring, Geochemistry and 40Ar-39Ar geochronology
of impact-melt clasts in feldspathic lunar meteorites: Implications for lunar
bombardment history. Meteorit. Planet. Sci. 40, 755–777 (2005).
52. D. R. Ulrich, J. Weber, Correlation of the thermal history of lunar and synthetic glass by
DTA and X-ray techniques. Lunar Planet. Sci. Conf. 4, 743 (1973).
53. M. Nowak, D. Schreen, K. Spickenbom, Argon and CO2 on the race track in silicate melts:
A tool for the development of a CO2 speciation and diffusion model. Geochim.
Cosmochim. Acta 68, 5127–5138 (2004).
54. R. A. F. Grieve, A. G. Plant, Partial melting on the lunar surface, as observed in glass coated
Apollo 16 samples. Lunar Planet. Sci. 4, 667–679 (1973).
55. L. A. Taylor, A. Patchen, R. G. Mayne, D. H. Taylor, The most reduced rock from the moon,
Apollo 14 basalt 14053: Its unique features and their origin. Am. Mineral. 89, 1617–1624
(2004).
56. G. A. Synder, L. A. Taylor, Oldest mare basalts or impact melts? The role of differential
melting of plagioclase in Apollo 14 high-Al basalts. Meteorit. Planet. Sci. 36, A194 (2001).
57. F. M. McCubbin, A. Steele, E. H. Hauri, H. Nekvasil, S. Yamashita, R. J. Hemley, Nominally
hydrous magmatism on the Moon. Proc. Natl. Acad. Sci. U.S.A. 107, 11223–11228 (2010).
58. C. R. Neal, L. A. Taylor, Petrogenesis of mare basalts: A record of lunar volcanism. Geochim.
Cosmochim. Acta 56, 2177–2211 (1992).
59. C. R. Neal, G. Y. Kramer, The petrogenesis of the Apollo 14 high-Al mare basalts. Am.
Mineral. 91, 1521–1535 (2006).
60. A. Stettler, P. Eberhardt, J. Geiss, N. Grögler, P. Maurer, Ar39-Ar40 ages and Ar37-Ar38
exposure ages of lunar rocks. Proc. Lunar Planet. Sci. Conf. 4, 1865–1888 (1973).
61. D. A. Papanastassiou, G. J. Wasserburg, Rb-Sr ages of igneous rocks from the Apollo 14
mission and the age of the Fra Mauro formation. Earth Planet. Sci. Lett. 12, 36–48 (1971).
62. J. R. Dunn, M. Fuller, On the remanent magnetism of lunar samples with special reference
to 10048,55 and 14053,48. Proc. Lunar Planet. Sci. Conf. 3, 2363–2386 (1972).
63. T. Nagata, R. M. Fisher, F. C. Schwerer, Lunar rock magnetism. Moon 4, 170–196 (1972).
64. D. W. Collinson, S. K. Runcorn, A. Stephenson, A. J. Manson, Magnetic properties of Apollo
14 rocks and fines. Proc. Lunar Planet. Sci. Conf. 3, 2343–2361 (1972).
65. M. Fuller, S. M. Cisowski, Lunar paleomagnetism, in Geomagnetism, J. A. Jacobs, Ed.
(Academic Press, 1987), vol. 2, pp. 307–456.
66. P. Rochette, J. Gattacceca, A. V. Ivanov, M. A. Nazarov, N. S. Bezaeva, Magnetic properties
of lunar materials: Meteorites, Luna and Apollo returned samples. Earth Planet. Sci. Lett.
292, 383–391 (2010).
67. C. Cournéde, J. Gattacceca, P. Rochette, Magnetic study of large Apollo samples: Possible
evidence for an ancient centered dipolar field on the Moon. Earth Planet. Sci. Lett.
331-332, 31–42 (2012).
68. D. F. Weill, R. A. Grieve, I. S. McCallum, Y. Bottinga, Mineralogy-petrology of lunar
samples. Microprobe studies of samples 12021 and 12022; viscosity of melts of selected
lunar compositions. Proc. Lunar Sci. Conf. 2, 413 (1971).
69. B. M. French, L. S. Walter, K. F. J. Heinrich, P. D. Loman, A. S. Doan, I. Adler, Composition of
major and minor minerals in five Apollo 12 crystalline rocks, NASA SP-306 (NASA, Greenbelt,
MD, 1972).
70. R. A. Cliff, C. Lee-Hu, G. W. Wetherill, Rb–Sr and U, Th–Pb measurements on Apollo 12
material. Proc. Lunar Sci. Conf. 2, 1493–1502 (1971).
71. D. A. Papanastassiou, G. J. Wasserburg, Lunar chronology and evolution from Rb-Sr
studies of Apollo 11 and 12 samples. Earth Planet. Sci. Lett. 11, 37–62 (1971).
72. M. R. Dence, J. A. V. Douglas, A. G. Plant, R. J. Traill, Mineralogy and petrology of some
Apollo 12 samples. Proc. Lunar Sci. Conf. 1, 285–299 (1971).
73. W. S. Baldridge, D. W. Beaty, S. M. R. Hill, A. L. Albee, The petrology of the Apollo 12
pigeonite basalt suite. Proc. Lunar Planet. Sci. Conf. 141–179 (1979).
74. P. Horn, T. Kirsten, E. K. Jessberger, Are there a 12 mare basalts younger than 3.1 b.y.
Unsuccessful search for a 12 mare basalts with crystallization ages below 3.1 b.y.
Meteoritics 10, 417 (1975).
75. P. E. Champness, A. C. Dunham, F. G. F. Gibb, H. N. Giles, W. S. MacKenzie, E. F. Stumpel,
J. Zussman, Mineralogy and petrology of some Apollo 12 lunar samples. Proc. Lunar Sci.
Conf. 1, 359–376 (1971).
76. R. C. Newton, A. T. Anderson, J. V. Smith, Accumulation of olivine in rock 12040 and other
basaltic fragments in the light of analysis and syntheses. Proc. Lunar Sci. Conf. 2, 575 (1971).
77. W. Compston, H. Berry, M. J. Vernon, B. W. Chappell, M. J. Kay, Rubidium-strontium
chronology and chemistry of lunar material from the Ocean of Storms. Proc. Lunar Sci.
Conf. 2, 1471–1485 (1971).
78. R. F. Dymek, A. L. Albee, A. A. Chodos, Comparative mineralogy and petrology of Apollo
17 mare basalts: Samples 70215, 71055, 74255, 75055, in Proceedings of the 6th Lunar
Science Conference (SAO/NASA Astrophysics Data System, 1975), pp. 49–77.
79. F. Tera, D. A. Papanastassiou, G. J. Wasserburg, The lunar time scale and a summary of
isotopic evidence for a terminal lunar cataclysm. Proc. Lunar Planet. Sci. Conf. 5, 792 (1974).
80. A. P. Roberts, C. R. Pike, K. L. Verosub, First-order reversal curve diagrams: A new tool
for characterizing the magnetic properties of natural samples. J. Geophys. Res. 105,
28461–28475 (2000).
81. R. J. Harrison, J. M. Feinberg, FORCinel: An improved algorithm for calculating first-order
reversal curve distributions using locally weighted regression smoothing. Geochem.
Geophys. Geosyst. 9, Q05016 (2008).
82. R. Egli, VARIFORC: An optimized protocol for calculating non-regular first-order reversal
curve (FORC) diagrams. Global Planet. Change 110, 302–320 (2013).
83. B. P. Weiss, S. M. Tikoo, The lunar dynamo. Science 346, 1246753 (2014).
84. S. M. Tikoo, B. P. Weiss, J. Buz, E. A. Lima, E. K. Shea, G. Melo, T. L. Grove, Magnetic fidelity
of lunar samples and implications for an ancient core dynamo. Earth Planet. Sci. Lett.
337-338, 93–103 (2012).
85. A. Stephenson, D. Collinson, Lunar magnetic field palaeointensities determined by
an anhysteretic remanent magnetization method. Earth Planet. Sci. Lett. 23, 220–228
(1974).
86. S. M. Tikoo, B. P. Weiss, W. S. Cassata, D. L. Shuster, J. Gattacceca, E. A. Lima, C. Suavet,
F. Nimmo, M. D. Fuller, Decline of the lunar core dynamo. Earth Planet. Sci. Lett. 404,
89–97 (2014).
87. S. M. Tikoo, B. P. Weiss, J. Buz, I. Garrick-Bethell, T. L. Grove, J. Gattacceca, Ancient lunar
dynamo: Absence of evidence is not the evidence of absence. Lunar Planet. Sci. Conf. 41,
2705 (2010).
88. C. Meyer, 15498, Lunar Sample Compendium (NASA, 2011).
89. J. M. Christie, D. T. Griggs, A. H. Heuer, G. L. Nord Jr., S. V. Radcliffe, J. S. Lally, R. M. Fisher,
Electron petrography of Apollo 14 and 15 breccias and shock produced analogs. Proc.
Lunar Planet. Sci. Conf. 1, 365–382 (1973).
90. D. R. Uhlmann, L. C. Klein, Crystallization kinetics, viscous flow and thermal histories
of lunar breccias 15286 and 15498. Proc. Lunar Planet. Sci. Conf. 2, 2529–2541 (1976).
91. A. R. Duncan, M. K. Sher, Y. C. Abraham, A. J. Erlank, J. P. Willis, L. H. Ahrens, Interpretation
of the compositional variability of Apollo 15 soils. Proc. Lunar Planet. Sci. Conf. 2,
2309–2320 (1975).
92. W. A. Gose, D. W. Strangway, G. W. Pearce, A determination of the intensity of the ancient
lunar magnetic field. Moon 7, 196–201 (1973).
93. R. F. Dymek, A. L. Albee, A. A. Chodos, Comparative petrology of lunar cumulate rocks
of possible primary origin: Dunite 72415, troctolite 76535, norite 78235, and anorthosite
62237. Proc. Lunar Planet. Sci. Conf. 1, 301–341 (1975).
94. I. Garrick-Bethell, K. Miljković, H. Hiesinger, C. H. van der Bogert, M. Laneuville,
D. L. Shuster, D. G. Korycansky, Troctolite 76535: A sample of the Moon’s South
Pole-Aitken basin? Icarus 338, 113430 (2020).
95. I. Garrick-Bethell, B. P. Weiss, D. L. Shuster, S. M. Tikoo, M. M. Tremblay, Further evidence
for early lunar magnetism from troctolite 76535. J. Geophys. Res. 122, 76–93 (2017).
96. H. Oda, J. Kawai, M. Miyamoto, I. Miyagi, M. Sato, A. Noguchi, Y. Yamamoto, J.-I. Fujihira,
N. Natsuhara, Y. Aramaki, T. Masuda, C. Xuan, Scanning SQUID microscope system
for geological samples: System integration and initial evaluation. Earth Planets Space 68,
179 (2016).
97. J. A. Tarduno, R. D. Cottrell, R. K. Bono, H. Oda, W. J. Davis, M. Fayek, O. van ‘t Erve,
F. Nimmo, W. Huang, E. Thern, S. Fearn, G. Mitra, A. V. Smirnov, E. G. Blackman,
Paleomagnetism indicates that primary magnetite in zircon records a strong Hadean
geodynamo. Proc Natl. Acad. Sci. U.S.A. 117, 2309–2319 (2020).
98. J. A. Tarduno, R. D. Cottrell, F. Nimmo, J. Hopkins, J. Voronov, A. Erickson, E. Blackman,
E. R. D. Scott, R. McKinley, Evidence for a dynamo in the main group pallasite parent body.
Science 338, 939–942 (2012).
99. R. D. Cottrell, J. A. Tarduno, In search of high-fidelity geomagnetic paleointensities:
A comparison of single plagioclase crystal and whole rock Thellier-Thellier analyses.
J. Geophys. Res. 105, 23579–23594 (2000).
100. J. Gattacceca, P. Rochette, Toward a robust normalized magnetic paleointensity method
applied to meteorites. Earth Planet. Sci. Lett. 227, 377–393 (2004).
101. D. R. Finn, R. S. Coe, A new protocol for three-axis static alternating field demagnetization
of rocks. Geochem. Geophys. Geosyst. 17, 1815–1822 (2016).
102. G. Kletetschka, T. Kohout, P. Wasilewski, Magnetic remanence in the Murchison
meteorite. Meteorit. Planet. Sci. 38, 399–405 (2003).
103. G. Kletetschka, M. A. Wieczorek, Fundamental relations of mineral specific magnetic
carriers for paleointensity determination. Phys. Earth Planet. Inter. 272, 44–49 (2017).
104. G. S. Collins, H. J. Melosh, B. A. Ivanov, Modeling damage and deformation in impact
simulations. Meteorit. Planet. Sci. 39, 217–231 (2004).
105. K. Wünnemann, G. S. Collins, H. J. Melosh, A strain-based porosity model for use
in hydrocode simulations of impacts and implications for transient crater growth
in porous targets. Icarus 180, 514–527 (2006).
106. D. A. Crawford, P. H. Schultz, Laboratory observations of impact-generated magnetic
fields. Nature 336, 50–52 (1988).
107. D. A. Crawford, P. H. Schultz, Electromagnetic properties of impact-generated plasma,
vapor and debris. Int. J. Impact Eng. 23, 169–180 (1999).
108. L. Carporzen, B. P. Weiss, S. A. Gilder, A. Pommier, R. J. Hart, Lightning remagnetization
of the Vredefort impact crater: No evidence for impact-generated magnetic fields.
J. Geophys. Res. 117, E01007 (2012).
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Tarduno et al., Sci. Adv. 2021; 7 : eabi7647 4 August 2021
SCIENCE ADVANCES | RESEARCH ARTICLE
14 of 14
109. S. A. Gilder, J. Pohl, M. Eitel, Magnetic signatures of terrestrial meteorite impact craters:
A summary, in Magnetic Fields in the Solar System, H. Luhr, J. Wicht, S. A. Gilder,
M. Holschneider, Eds. (Springer, 2018), pp. 357–382.
110. K. L. Louzada, B. P. Weiss, A. C. Maloof, S. T. Stewart, N. L. Swanson-Hysell, S. Adam Soule,
Paleomagnetism of Lonar impact crater, India. Earth Planet. Sci. Lett. 275, 308–319
(2008).
111. B. P. Weiss, S. Pedersen, I. Garrick-Bethell, S. T. Stewart, K. L. Louzada, A. C. Maloof,
N. L. Swanson-Hysell, Paleomagnetism of impact spherules from Lonar crater, India
and a test for impact-generated fields. Earth Planet. Sci. Lett. 298, 66–76 (2010).
112. F. Jourdan, F. Moynier, C. Koeberl, 40Ar/39Ar age of the Lonar crater and consequence
for the geochronology of planetary impacts. Geology 39, 671–674 (2011).
113. S. J. Robbins, A new global database of lunar impact craters >1-2 km: 1. Crater locations
and sizes, comparisons with published databases, and global analysis. J. Geophys. Res.
124, 871–892 (2018).
114. C. Yang, H. Zhao, L. Bruzzone, J. A. Benediktsson, Y. Liang, B. Liu, X. Zeng, R. Guan, C. Li,
Z. Ouyang, Lunar impact crater identification and age estimation with Chang’E data by
deep and transfer learning. Nat. Commun. 11, 6358 (2020).
115. G. Neukum, B. A. Ivanov, W. K. Hartmann, Cratering records in the inner solar system
in relation to the lunar reference system. Space Sci. Rev. 96, 55–86 (2001).
116. D. Stöffler, G. Ryder, Stratigraphy and isotope ages of lunar geologic units: Chronological
standard for the inner Solar System. Space Sci. Rev. 96, 9–54 (2001).
117. C. M. Fortezzo, P. D. Spudis, S. L. Harrel, Release of the digital unified global geologic
map of the Moon at 1:5,000,000- Scale, paper presented at the 51st Lunar and
Planetary Science Conference, Lunar and Planetary Institute, Houston, TX, 3 March
2020.
118. M. E. Purucker, J. B. Nicholas, Global spherical harmonic models of the internal magnetic
field of the Moon based on sequential and coestimation approaches. J. Geophys. Res. 115,
E12007 (2010).
119. R. Day, M. Fuller, V. A. Schmidt, Hysteresis properties of titanomagnetites: Grain-size
and compositional dependence. Phys. Earth Planet. Inter. 13, 260–267 (1977).
120. D. J. Dunlop, Theory and application of the Day plot (Mrs/Ms versus Hcr/Hc) 1. Theoretical
curves and tests using titanomagnetite data. J. Geophys. Res. 107, EPM 4-1–EPM 4-22
(2002).
121. P. Wasilewski, Magnetization of small iron–nickel spheres. Phys. Earth Planet. Inter. 26,
149–161 (1981).
122. R. S. Coe, The determination of paleo-intensities of the Earth’s magnetic field
with emphasis on mechanisms which could cause non-ideal behavior in Thellier’s
method. J. Geomag. Geoelec. 19, 157–179 (1967).
123. P. A. Selkin, L. Tauxe, Long-term variations in palaeointensity. Phil. Trans. R. Soc. A 358,
1065–1088 (2000).
124. C. Suavet, B. P. Weiss, W. S. Cassata, D. L. Shuster, J. Gattacceca, L. Chan, I. Garrick-Bethell,
J. W. Head, T. L. Grove, M. D. Fuller, Persistence and origin of the lunar core dynamo. Proc.
Natl. Acad. Sci. U.S.A. 110, 8453–8458 (2013).
Acknowledgments: We thank G. Kloc for assistance in sample preparation and B.L. McIntyre
on electron microscopy analyses. We greatly appreciate helpful reviews from D. Dunlop and
two anonymous reviewers. Funding: This work was supported by NSF grants EAR1656348
(to J.A.T.) and PHY-2020249 (E.G.B. and M.N.), NASA grants 80NSSC19K0510 (to J.A.T.) and
PGG-NNX13AO33G (to C.L.J. and K.L.), and a JSPS fellowship (to J.A.T.). Author contributions:
K.L. selected samples for analysis. R.D.C. and K.L. performed remanence and rock magnetic
measurements. R.K.B., W.H., T.Z., and B.C. conducted SEM analyses. Experimental data were
analyzed by these authors together with J.A.T. and A.V.S. H.O. contributed measurements on
materials. M.N. performed numerical impact modeling. E.G.B. contributed impact theory.
C.R.N. and M. I.-M. provided petrologic and age context. C.L.J. and K.L. conceived the initial
study. J.A.T. designed and supervised subsequent investigations. J.A.T. wrote the manuscript
with input from all the authors. Competing interests: The authors declare that they have no
competing interests. Data and materials availability: All data needed to evaluate the
conclusions in the paper are present in the paper and/or the Supplementary Materials.
Requests for Apollo lunar samples for study are directed to NASA (https://curator.jsc.nasa.gov/
lunar/sampreq). Data presented here are available in the EarthRef (MagIC) database (earthref.
org/MagIC/17143). Additional data related to this paper may be requested from the authors.
Submitted 29 March 2021
Accepted 16 June 2021
Published 4 August 2021
10.1126/sciadv.abi7647
Citation: J. A. Tarduno, R. D. Cottrell, K. Lawrence, R. K. Bono, W. Huang, C. L. Johnson,
E. G . Blackman, A. V. Smirnov, M. Nakajima, C. R. Neal, T. Zhou, M. Ibanez-Mejia, H. Oda,
B. Crummins, Absence of a long-lived lunar paleomagnetosphere. Sci. Adv. 7, eabi7647 (2021).
on August 4, 2021http://advances.sciencemag.org/Downloaded from

Absence of a long-lived lunar paleomagnetosphere
Crummins
Blackman, Aleksey V. Smirnov, Miki Nakajima, Clive R. Neal, Tinghong Zhou, Mauricio Ibanez-Mejia, Hirokuni Oda and Ben
John A. Tarduno, Rory D. Cottrell, Kristin Lawrence, Richard K. Bono, Wentao Huang, Catherine L. Johnson, Eric G.
DOI: 10.1126/sciadv.abi7647
(32), eabi7647.7Sci Adv
ARTICLE TOOLS http://advances.sciencemag.org/content/7/32/eabi7647
MATERIALS
SUPPLEMENTARY http://advances.sciencemag.org/content/suppl/2021/08/02/7.32.eabi7647.DC1
REFERENCES http://advances.sciencemag.org/content/7/32/eabi7647#BIBL
This article cites 109 articles, 18 of which you can access for free
PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions
Terms of ServiceUse of this article is subject to the
is a registered trademark of AAAS.Science AdvancesYork Avenue NW, Washington, DC 20005. The title
(ISSN 2375-2548) is published by the American Association for the Advancement of Science, 1200 NewScience Advances
License 4.0 (CC BY-NC).
Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial
Copyright © 2021 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of
on August 4, 2021http://advances.sciencemag.org/Downloaded from