MGS operated from 1997–2006 and spent most of its time in orbit at approximately 400 km altitude (Figure 2B) and at 87° inclination, crossing the equator at 2 am and 2 pm local time, the so called Mapping Orbit (MO; 1999–2006). Magnetic field data were also obtained at lower altitudes early in the mission during Aerobraking (AB) and the Science Phasing Orbit (SPO). However, other than some night–time measurements over the South polar region, all data below 400 km were obtained during the day-time (see Figure 1 in Mittelholz et al., 2018a). Thus, MGS provided repeated coverage of the magnetic field at 400 km altitude, yielding a data set that is ideal for robust characterization of crustal fields because outliers can be identified and removed, but that lacks low altitude information.

MAVEN has been in orbit for over 7 years (2014—present) providing a complementary data set to that of MGS because the MAVEN spacecraft orbit is highly eccentric (Figure 2B), and the body-fixed location and local time of periapsis evolve with time. Deep–dip campaigns have provided data at approximately 110 km altitude but over only a few locations; however, almost complete coverage is available at 150 km altitude (Figure 2C). The orbital inclination is 75°, and so data over the polar regions are not available. The MAVEN data set provides a wider range of altitude and local time coverage over any given region of the planet than that of MGS, specifically substantial data at altitudes less than 400 km. However, it does not provide multiple repeat measurements at the same altitude and local time, and so data contaminated by ionospheric fields are less easily identified.

Satellite observations have led to the generation of many crustal magnetic field models, that allow projection of data onto a constant altitude and downward-continuation to predict the surface magnetic field. MGS data have enabled global (e.g., Cain, 2003; Langlais et al., 2004; Morschhauser et al., 2014) and local (Plattner and Simons, 2015) models of the field using different techniques such as equivalent source dipoles (e.g., Langlais et al., 2004) or spherical harmonics (SH; e.g., Morschhauser et al., 2014). Lower altitude night-time MAVEN data have resulted in models with improved spatial resolution (Mittelholz et al., 2018a; Langlais et al., 2019; Gao et al., 2021). The minimum wavelength in the field visible from orbit can be approximated by the orbital altitude and so models including MAVEN data currently resolve structure globally at ∼150 km (equivalent to SH degree 142). The measured crustal field is proportional to the product of the thickness of the magnetized layer and the magnetization, and so for a given layer thickness, magnetization maps are a further outcome from modeling studies (Vervelidou et al., 2017; Langlais et al., 2019). However, in general, the thickness of the magnetized layer is unknown and most models assume that the magnetization is uniformly distributed over a 40-km-thick crustal column.

The large-scale pattern of magnetization follows the hemispheric dichotomy, suggesting that the origin of the dichotomy and crustal magnetization are linked, and raising several interconnected questions (Table 1): (H1) Was a martian dynamo active at the time the dichotomy formed? (H2) Does the hemispheric difference in magnetic fields reflect differences in the distribution and/or types of magnetic minerals and/or the properties of the dynamo field? (H3) Can the magnetic field record help elucidate the origin (exogenic versus endogenic) of the dichotomy?

Over the past couple of years, the global data set of lower altitude MAVEN observations has enabled progress on each of these questions. MGS data alone could not conclusively demonstrate the presence of magnetic fields of crustal origin over the northern lowlands; leaving open the possibility of either no magnetization or magnetization that was unresolvable by MGS. MAVEN data have allowed identification of shorter-wavelength magnetic signatures that do not show clear correlation with any other data set such as gravity or topography (Langlais et al., 2019; Mittelholz et al., 2020a). These are inferred to reflect magnetization acquired in an active dynamo period at the time of formation of the dichotomy (Mittelholz et al., 2020a). Furthermore, Mittelholz et al. (2020a) suggested that the distribution of magnetization globally, as well as the newly-detected weak northern hemisphere anomalies could be explained by different magnetic mineralogies in the two hemispheres that would result from formation of the northern lowlands crust following a giant impact.

Source depth investigations (Lewis and Simons, 2012; Gong and Wieczorek, 2021) show that magnetization depths follow the dichotomy, and are on average 9 km in the Northern hemisphere and 32 km in the southern hemisphere (Gong and Wieczorek, 2021). Gong and Wieczorek (2021) suggest that the strongly magnetized, deep layer in the South might be ancient (pre-dichotomy) units, buried by ejecta material from the large impact that formed the Borealis Basin. In their scenario, the shallow northern hemisphere material magnetization was acquired either by the newly formed post-impact crust or by later lava flows (i.e., its precise timing is still unknown, but must be syn- or post- Borealis impact). In contrast, a degree-1 dynamo origin (see Section 4.2) for the magnetic signature of the dichotomy would predict hemispherical differences in surface strengths but similar magnetization depths in the South and North. However, further work is needed, in particular to narrow source depth uncertainties and to understand the nature of the southern hemisphere magnetization (Section 4.1.4).

Magnetic anomalies associated with a crater are a record of the ambient magnetic field at the time at which the crater formed/cooled. The general absence of magnetic anomalies and inferred lack of strong magnetization associated with the interiors of the HUIA basins (craters > 500 km diameter; Robbins et al., 2013) was observed with the first MGS magnetic field data (Acuna et al., 1999). Subsequent analysis of the entire MGS mission data over craters with diameters greater than 300 km, indicated that others also show signs of temperature- and shock-induced (de)magnetization (Lillis et al., 2013; Vervelidou et al., 2017). The ages of large craters and basins with, or without, interior magnetizations have been used to infer a dynamo cessation by 4.1–4.0 Ga (Lillis et al., 2013; Vervelidou et al., 2017). Although a clear magnetic field signature, or absence thereof, associated with a crater provides valuable information, interpretation of the cratering record requires caution. Questions that persist regarding the crater record include (Table 1): (C1) How reliable is the large basin catalog used in previous studies? (C2) Does the absence of magnetic fields at satellite altitudes over crater interiors imply that they are unmagnetized, or simply magnetized below the resolution of satellite data? (C3) Do craters lacking magnetic fields over their interiors require the absence of a dynamo at the time of their formation, or are there alternative explanations? (C4) Are shock effects important to the interpretation of the crater magnetic field record?

Not many large craters exist (Figure 4A) and they are generally old, implying that identification of young magnetization is not possible. The ages of the large craters in Figure 4A are 3.91–4.18 Ga (mean of 4.06 Ga) and 4.25–4.11 Ga (mean of 4.17 Ga) according to Robbins et al. (2013) and Frey (2008), respectively. Question (C1) above arises because the latter record includes 19 basins that were inferred from their muted topographic expression (referred to as quasi-circular depressions, QCDs), and were suggested to be buried by younger units leaving no expression in visible imagery and only partially intact rims. However, the existence of what was identified as a large ancient buried crater (Frey, 2008); dashed in Figure 4A) has since been challenged (Bottke and Andrews-Hanna, 2017). Furthermore, Bottke and Andrews-Hanna (2017) suggest that between 4 (HUIA) and at most 12 basins post-date the dichotomy, substantially reducing the number of large ( > 300 km) ancient craters that can provide dynamo timing constraints. Many craters of smaller diameters occur on young and old terrains (Figure 4C), and can be investigated with higher resolution data from MAVEN (Section 3.2) and potential future missions. In addition, the presence or absence of weak fields in the lower altitude MAVEN data c. f. MGS data is an important step toward addressing question (C2) above. Related to this, although the canonical explanation for the absence of magnetic fields above major basins and large craters has been the absence of a dynamo at the time they formed (question (C3) above), recent work has proposed alternative scenarios, such as excavation of “magnetizable” crustal material or episodes of weaker dynamo activity (Mittelholz et al., 2020a; Hemingway and Driscoll, 2021).

Improved understanding of the extent to which shock or thermal effects result in demagnetization of existing magnetized layers requires improved constraints on magnetization depths. For example, a strongly magnetized layer at depths greater than those affected by the shock wave and/or reheating, could retain a pre-impact magnetization. For the large basins, where transient crater diameters range from 750–1,400 km (Melosh, 1989), it is also clear that crustal modification and/or removal would be enormous. In addition, for such large basins, estimation of shock demagnetization based on scaling arguments derived for smaller impactors is likely not appropriate (similar to the Moon (Halekas et al., 2002)). However, for smaller craters, SRM could be important and future analysis over craters sufficiently large to be magnetically resolvable, but with excavation depths that do not exceed the crustal thickness locally, is required. Because of the systematic increase in excavation depths of craters with increasing diameter, future investigations of craters with a range of diameters will allow the magnetic record of the crust at different depths to be assessed.

Analysis of volcanic regions and nearby magnetic field signatures have been used to place constraints on dynamo timing (Section 4.2; e.g., Hood et al., 2010; Lillis et al., 2006, 2008; Milbury et al., 2012). However, unlike the case for impacts, surface ages of volcanic constructs or flows do not necessarily record the magnetization age because “magnetic resetting” of intrusions at depth can occur with no surface record. Further, although magnetization in volcanic areas is generally inferred to be a TRM, prolonged volcanism can also fuel hydrothermal activity and chemical reactions leading to a CRM (Harrison and Grimm, 2002; Chassefière et al., 2013). Important questions regarding the volcanic record are thus (V1) (How) can we reliably use the surface volcanic record for dating magnetization? (V2) Is the remanent magnetization (or absence thereof) in volcanic regions solely a TRM or a combination of TRM/CRM?

Recent work has started to address the first of these questions. Clearly, additional source depth constraints are needed to understand whether the bulk of the magnetization is carried by near-surface or deep units. For example, regional source depth estimates for the southern hemisphere point to much of the magnetization being carried in the deeper crust (Gong and Wieczorek, 2021); however whether some magnetization could more locally be carried at shallow depths in the vicinity of volcanoes studied previously for dynamo timing (e.g., Tyrrhenus Mons; Milbury et al., 2012) is unknown. On such local scales clear indications that at least some of the magnetization is carried by a surficial unit such as the identification of a magnetized pyroclasic flow in Mittelholz et al. (2020a), require the lowest altitude MAVEN data and future high resolution studies.

Locally, and mostly concentrated in Terra Sirenum (TS) and Terra Cimmeria (TC) (see Figure 1), crustal fields are much stronger than in other regions on Mars or terrestrial crustal fields measured from Earth orbit (Langlais et al., 2010; Thébault et al., 2010). These require large amplitude magnetizations even if tens of kms of the crust are magnetized. For example, Parker (2003) showed that a minimum magnetization of 4.76 A/m was required throughout 50 km of crust to explain the strong southern hemisphere fields. The vertically integrated magnetization is 238 kA, compared with at most ∼10 kA for freshly magnetized mid-ocean ridge basalts (10 A/m for a ∼1 km magnetized layer). Pervasive in the interpretation of martian magnetization is that the magnetized layer may be tens of km thick; however in the ocean basins most of the magnetization is carried in Layer 2A of the crust which is ∼400–600 m thick (Sandwell, 2022). The main questions here are: (S1) What gives rise to the strong magnetization? Is it a thick magnetized layer, distinct mineralogy, or a combination thereof? (S2) What are the scale lengths and origins of coherent blocks of magnetization? Are there any indications of a record of reversals, versus magnetization contrasts that reflect units with different lithologies and/or post emplacement rotations?

Although the first orbital observations over the TS/TC region were inferred to result from linear magnetization contrasts akin to those in the ocean basins (Acuna et al., 1999; Connerney et al., 1999), later modeling and addition of MAVEN data revealed a more blocky magnetic field pattern (Figure 1; Langlais et al., 2019), in which anomalies do not have obvious correlations with surface features. Bouley et al. (2020) inferred the region to consist of discrete crustal blocks, associated with thicker crust and enriched in potassium and thorium. Because the terrain is overprinted by the big basins Hellas and Argyre, they suggest that TS/TC represents some of the oldest crustal material and possibly a distinct crustal formation mechanism compared with other regions. Source depths in the region are found to be exceptionally deep, specifically larger than 40 km (Gong and Wieczorek, 2021). However, analysis of satellite compositional data also indicates surficial iron enhancements, that have been suggested to contribute to the strong magnetizations (AlHantoobi et al., 2021).

Dynamo timing raises some of the most-discussed questions regarding the martian magnetic field: (DT1) When did the martian dynamo operate? Was it continuous or intermittent? (DT2) When did it cease permanently and how quickly did this happen? (DT3) How does the dynamo timing fit with models for early Mars thermal evolution and atmospheric and surface conditions?

The starting point for most hypotheses regarding timing, is the assumption that the absence of substantial crustal field anomalies above the large impact basins implies that no dynamo operated when the basins formed. The traditionally most-accepted suite of studies argued that a dynamo operated prior to basin formation, the “early dynamo” hypothesis and that shock and thermal effects during basin formation erased any magnetization present. This was supported by the early observation that most magnetic signatures are seen over (Pre-)Noachian terrain (Acuña et al., 2001), on which the major basins Hellas and Argyre are superposed. The argument was further bolstered by analyses of magnetic fields over craters larger than 300 km diameter, from which the oldest magnetized crater was found to be ∼4.1 Ga, older than the large basins (Lillis et al., 2013; Vervelidou et al., 2017). Other studies have argued for a “late dynamo”, that started after the basins were emplaced. Magnetic signatures over Noachian terrains were interpreted as magnetized intrusions postdating the surface units. Identification of magnetic signatures above young volcanoes and lava flows also supported the “late dynamo” hypothesis (Hood et al., 2010; Langlais and Purucker, 2007; Lillis et al., 2006). Figure 5 shows a compilation of literature addressing dynamo timing to emphasize the breadth of studies and suggested scenarios. Recently, low altitude MAVEN data have allowed identification of magnetic signatures over the northern hemisphere (Section 3.2) and a pyroclastic flow (Section 4.1.3) indicating an active dynamo at 4.5 Ga and at 3.7 Ga, respectively (Mittelholz et al., 2020a). One interpretation of these results is that a dynamo was active from 4.5 to at least 3.7 Ga, including during the period of basin formation. This was also previously suggested by Milbury et al. (2012) who hypothesized a dynamo shutdown around 3.6 Ga while still arguing for an early dynamo. An active dynamo during basin formation leaving small amplitude signatures not visible from orbit is plausible, given the large amount of crustal material excavated, and subsequent changes in amount and type of magnetic minerals in the basin floors (Mittelholz et al., 2020a). An alternative scenario is that of changing dynamo processes during and immediately after basin formation, specifically the possibility of decreased dynamo activity during the basin forming events (Hemingway and Driscoll, 2021; Mittelholz et al., 2020a).

Dynamo timing has been suggested to affect climate evolution on early Mars (Jakosky and Phillips, 2001), linking interior and surface/atmospheric processes. The change in ancient climate conditions from a thick atmosphere that allowed liquid water to persist (at least temporarily or intermittently) on the martian surface to a much thinner atmosphere (e.g., Chassefière et al., 2007; Wordsworth et al., 2015), could suggest a causal relation between atmospheric escape and a global dynamo field. However, this notion has been questioned (Brain et al., 2013, 2016). Pinning down the active periods of the dynamo, including temporary and permanent cessation, will allow this issue to be addressed further.

The timing results summarized above prompt the major question: (DM1) What was the driving mechanism and did it change over time? Although a thermally-driven dynamo is able to explain a very early, but limited-duration dynamo (∼500 Ma), extending the dynamo lifetime to 3.7 Ga and longer is challenging. Early in a planet’s history, the core is much hotter than the mantle driving superadiabatic heat loss from the core that enables sufficiently vigorous convection to generate a dynamo (Breuer and Spohn, 2003; Williams and Nimmo, 2004). When the heat loss is no longer superadiabatic, additional sources of buoyancy must be found to sustain a convectively-driven dynamo. Once the core temperature drops below the solidus and inner core crystallization starts, both the latent heat (additional thermal buoyancy) and the release of light elements (compositional buoyancy) become important. For the Earth, inner core solidification (i.e., bottom-up crystallization) is the dominant driver of the present-day dynamo (Gubbins, 1977; Gubbins et al., 2003; Nimmo, 2007). Recent InSight results indicate a relatively large martian core (∼1860 km), requiring a low density, and thus large concentrations of light elements (Stähler et al., 2021). While the exact amount and types of light elements in the core and mantle are currently debated, a solid inner core has not been seismically detected, and is now thought to be unlikely because it is incompatible with proposed compositional and thermal models (Stähler et al., 2021; Khan et al., 2022). The latter suggest melting temperatures that are below core temperatures inferred from thermal evolution models (e.g., Plesa et al., 2018).

Hemingway and Driscoll (2021) explore a range of dynamo scenarios for early Mars by varying parameters such as core sulfur mass fraction and thermal conductivity, light element partitioning or initial temperature. They find a wide range of possibilities for the evolution of the past and future martian dynamo, that given the large parameter space, can satisfy any of the scenarios proposed in Figure 5. In particular, the thermal conductivity of the core affects the duration of an early thermal dynamo. Recent experiments (Pommier, 2018; Pommier et al., 2020) suggest that the thermal conductivity of a sulfur-rich core is lower and between 5 and 30 Wm⁻¹K⁻¹ depending on the iron alloy composition in the Fe-S and Fe-S-O-Mg-Si systems. This is lower than the 30–120 Wm⁻¹K⁻¹ typically assumed in previous studies (e.g., Nimmo and Stevenson, 2000; Williams and Nimmo, 2004; Arkani-Hamed, 2012; Hemingway and Driscoll, 2021). Thermal evolution modeling (Greenwood et al., 2021) confirms that this has an important effect on the dynamo duration, with lower conductivity values allowing longer thermally-driven dynamo activity. In conclusion, the evolution of the core and possible dynamo mechanisms critically depend on the light element content, and thermal state and core thermal conductivity. The recent InSight results for core size and composition (Stähler et al., 2021; Khan et al., 2022) suggest the need to re-examine alternative crystallization scenarios, e.g., iron snow from possible top-down (near the core-mantle boundary) crystallization and subsequent remelting deeper in the core as well as the possibility that different driving mechanisms might have operated at different periods in martian history and whether future dynamo episodes are a possibility (Stewart et al., 2007; Hemingway and Driscoll, 2021). A further consideration is whether the thermal consequences of one or more large impacts could have had a transient effect on the core heat flow pattern that was sufficient to either initiate (Reese and Solomatov, 2010) or inhibit (Roberts et al., 2009) dynamo generation.

A range of questions regarding the characteristics of the martian paleodynamo exist: (DC1) What was the average surface field strength, and did it vary over time? (DC2) Was the field dominantly dipolar or did it have substantial non-dipolar contributions? If dipolar, how closely was the dipole axis aligned with the rotation axis? (DC3) Did the martian dynamo reverse polarity and, if so, how frequently and what were the characteristics of the reversals (duration, global field weakening)?

The strength of the paleodynamo, B _ancient, in which a crustal magnetization, M, was acquired, is related to M, and the thermoremanent magnetic susceptibility, χ _TRM, by B _ancient = Mμ ₀/χ _TRM, where μ ₀ is the magnetic permeability of free space. Magnetization inferred from satellite data is inversely related to the thickness of the magnetized layer; as described in Section 3.3 the latter is commonly difficult to assess. Mittelholz et al. (2020a) identified a magnetic anomaly associated with a crater in Lucus Planum; using the crater depth, combined with geological constraints, to infer the thickness of the magnetized layer, they were able to conclude that the 3.7-Ga-old lava flow could have been magnetized in an Earth-like ancient field. This is consistent with conclusions from paleomagnetic studies of meteorite ALH84001 (Weiss et al., 2008). However, the strength of the field might have varied through time and therefore such inferences are recordings of the ancient field strength at the time at which the respective magnetization was acquired.

The ancient dynamo might also have reversed polarity, often studied in planetary magnetism via paleopole locations inferred from satellite-derived magnetic field maps. Different methods and assumptions to derive paleopole locations significantly affect the results and lead to different conclusions (e.g., Milbury et al., 2012; Boutin and Arkani-Hamed, 2006; Thomas et al., 2018). Thomas et al. (2018) present an overview of published paleopole locations (their Figure 1) which show a spread across the planet. In their study, uncertainties are incorporated and admissible locations of paleopoles indicate that Mars must have reversed polarity at least once and that a substantial amount of true polar wander occurred. This is consistent with reorientation of the rotation axis in response to the emplacement of Tharsis (Zuber and Smith, 1997).

Most discussions to date concerning the ancient martian field, in particular paleopole studies, assume that it was a classic dipolar field. However, one hypothesis for the north/south dichotomy in the magnetic field is that the dynamo itself produced a hemispheric field, driven by hemispheric (spherical harmonic degree 1) variations in heat flow at the core-mantle boundary (Stanley et al., 2008). This large-scale pattern in heat flow has been invoked in endogenic models for the formation of the dichotomy (Zhong and Zuber, 2001; Roberts and Zhong, 2006). Previous (Stanley et al., 2008) and more recent (Yan et al., 2021) numerical dynamo simulations indicate that such heat flow variations can result in either a stable or reversing hemispheric dynamo. However, Dietrich and Wicht (2013) have challenged this view and argued that frequent (∼10 kyr timescales) dynamo reversals in these hemispherical models would not allow strong coherent magnetization to develop. This highlights the need for future constraints on any history of magnetic field reversals on Mars.