Control of heterogeneous , layered successions on shallow-level magma 1 emplacement and host rock deformation

25 Host rock deformation in active volcanic settings can signal and be used to constrain magma 26 emplacement. Yet it is difficult to evaluate the accuracy of intrusion parameters derived from 27 inversion of deformation signals because we cannot test estimates by directly accessing the 28 magma body. Physical modelling is thus critical to understanding how intrusion translates 29 into host rock deformation, particularly surface uplift and/or subsidence, because we can use 30 transparent materials or excavate models to view the actual intrusion geometry. However, 31 few physical models have investigated how a heterogeneous, layered host material impacts 32 magma emplacement, despite evidence suggesting the presence of weak layers can control 33 intrusion style and geometry. We conduct several models that simulate emplacement of a 34 felsic magma at ~6 km depth within a granular (sand) host rock; in two of our models we 35 incorporate two, thin, weak microbead layers into the layered host material. We show that 36 intrusion solely within the granular material is primarily accommodated by lateral contraction 37 (compaction and folding) of the host material, resulting in a dyke-like intrusion that erupted. 38 When the microbead layers were present, a cone sheet and saucer-shaped sill preferentially 39 formed, without erupting, accommodated by forced folding. Furthermore, we demonstrate 40 that surface deformation does not simply reflect the complexity of the intrusion geometry or 41 internal host material deformation. Overall, our results indicate that physical models should 42 further explore the role of host material heterogeneity on magma emplacement. 43 44


Introduction
Sill emplacement in the shallow subsurface may be accommodated by roof uplift (e.g.,   Material properties 126 The brittle behaviour of the upper crustal rocks was simulated by layers composed of a 127 mixture of quartz (70%) and K-feldspar (30%) sand with grain dimensions <250 m, an 128 angle of internal friction of 43°, cohesion of ~10 Pa, and a density of 1408 kg m -3 (Table 2; of ~10 Pa, and a density of 1480 kg m -3 (Table 2). For a magma analogue we used pure 138 vegetable polyglycerine-3 (PG3), a low-viscosity Newtonian fluid with a viscosity of 17 Pa s 139 and density of ~1190 kg m -3 (Table 2; Montanari et al., 2017).
140 The models were scaled according to the principles of geometric, dynamic, and kinematic 145 similarity (Hubbert, 1937; Ramberg, 1981). We used a length scaling ratio l* (where * 146 denotes the ratio between the model and natural values) of 10 -5 , such that 10 mm in the 147 models corresponds to 1 km in nature. The overburden above the top of the injection inlet, at 148 the start of each model run was 60 mm (Fig. 1B), which thus corresponds to an initial 149 emplacement depth of 6 km. Microbead layers added to models FF-07 and FF-08 were each 3 150 mm thick (Fig. 1B). Both models and nature have the same gravitational acceleration (g), 151 imposing a scale factor of g*=1. The density ratio () is ~0.5, resulting from the ratio 152 between the analogue granular material (~1400 kg m -3 ) and natural rocks (~2700 kg m -3 ) 153 (e.g., Schellart, 2000), as well as the ratio between our magma analogue (~1200 kg m -3 ) and a 154 natural granitic magma at emplacement conditions (~2400 kg m-3; e.g., Montanari et al., ). These ratios result in a stress scaling ratio g* l*) of ~5 × 10 -6 . The stress scaling ratio is related to the scaling ratios of strain rate (*), viscosity (), and velocity of

162
During the experiments, deformation was monitored through top-view photos taken at regular 163 five-minute time intervals (Fig. 1A). At the end of each experiment, the models were 164 watered, frozen, and cut with a saw to obtain cross-sections that could be used to analyse 165 internal deformation (e.g., Fig. 1C Table. 3).

209
Broad antiforms are developed at horizons G and H, which have Amax values of 3.9 mm and 210 2.3 mm, respectively ( Fig. 2A; Table 3). The surface deformation associated with the

219
For each cross-section, a series of vertical profiles 10 mm apart were arbitrarily imposed and 220 used to focus measurements (see Fig. 1C).  from 4.8-6.8 mm, and for FF-08 range from 4.6-6.6 mm (Figs 2B and C; Table 3). In all 248 models, the location of Amax varies between horizons and is rarely situated directly above the       --------00.7  -120  ----------130  -----------denotes      Layer thickness (tlayer) across the modelled forced folds is variable ( Fig. 6; Table 5).    (Figs 3 and 4B), and is thus decoupled from intrusion 354 thickness (Fig. 4F); and (iv) the weak microbead layers preferentially accommodate more 355 linear strain (extension), and their layer thickness varies across the folds (Figs 4E and 6).     (Figs 2A and 8A); this is also consistent with thinning of 412 layers A-B to D-E (Figs 5A and B). When the dyke-like intrusion reached Layer D-E and 413 formed a small sill, it appears the reduced overburden meant space could be accommodated 414 by the formation of broader, dome-shaped forced folds (Figs 2A and 8B). In contrast, we suggest the injection of the magma analogue in models FF-07 and FF-08 caused the 416 microbeads layers to immediately 'flow' laterally as porosity reduced, producing a central 417 low in layer thickness, allowing underlying sand layers directly above the inlet to bend 418 upwards ( Fig. 8C and D). The capability of the microbead layers to compress laterally is which accommodate more strain than horizons between sand-sand layers (Fig. 4E).