**2. Experimental**

For this study, a Magnesium single crystal of high purity (better than 99.999%) was purchased from Goodfellow Cambridge Ltd. (Huntingdon, United Kingdom) to ascertain a homogenous chemistry and avoid any influences from chemical segregation or inhomogeneities on the results. To expose the crystalline faces and directions of interest for testing in this study, thin slices were cut from the pre-oriented single crystal using a slow speed diamond wire saw and shaped to resemble the geometry of half-moon samples for transmission electron microscopy (TEM). These samples were subsequently etched in a water-free solution consisting of 5 mL HNO3 and 95 mL ethanol to avoid unintended oxidation of Mg and remove material potentially deformed due to the previous sawing process. Subsequently, miniaturised compression and tensile samples were fabricated along the chemically thinned wedge as shown exemplarily in Figure 1 and insets therein using a focussed ion beam (FIB, Leo 1540 XB, Carl Zeiss, Oberkochen, Germany). The acceleration voltage of the Ga ion beam was 30 keV, and the ion beam currents were gradually decreased from starting 5 nA for rough milling down to 10 pA for final polishing of the specimen surfaces. All samples were placed on somewhat larger pedestals seen in the overview in Figure 1 to provide a clear measure of potential sample sink-in [22] during in-situ testing. Furthermore, micro-compression samples have typical aspect ratios between 4:1 and 3:1 to avoid potential plastic buckling [22], while micro-tensile specimens have commonly aspect ratios of 5:1 [23]. For details on these sample fabrication strategies, please refer to [24,25]. By fabricating the samples along the wedge, many of them can be placed in close proximity of a few microns, while at the same time a free line of sight is ensured in order to facilitate in-situ testing or other means of analysis.

**Figure 1.** Low magnification SEM overview image of a series of differently sized compression and tensile samples before testing. The insets show a representative tensile and compression specimen, respectively.

Furthermore, it is essential to fabricate the samples with a well-defined orientation in a rigorous coordinate system and avoid all geometrical deficiencies. At the same time, the specimens should be transferrable between the FIB employed for specimen fabrication, the dedicated in-situ testing scanning electron microscope (SEM, DSM982, Carl Zeiss AG, Oberkochen, Germany), as well as the TEM (JEM 2100F, Jeol Ltd., Tokyo, Japan) for additional high resolution characterisation. Thus, we developed a staggered holder design as shown in Figure 2. The actual TEM-inspired sample has a flat along its circumference to ensure rotational alignment. Thus, even after repeated placement in a TEM, rotational alignment is kept [26]. This half-disk is mounted in a holder suited to fit the Bruker Hysitron Picoindenter PI-85 SEM as well as the PI-95 TEM nanoindenter with a clamping grip. As stated before, with the rotational degree of freedom fixed, the alignment during uniaxial testing is given. Finally, this whole SEM/TEM testing compatible holder can be placed in a larger mounting device that is on the one hand compatible to the Asmec UNAT 1 in-situ SEM microindenter, while on the other hand, this holder is used for the FIB processing, as it allows for controlled rotations of the mounted sample by ±45◦ and ±90◦, as well as tilt along the long axis by 90◦. Thus, by re-mounting the specimen at various tilts, perfect rectangular and orthogonal specimens with constant sample cross-sections can be fabricated. Moreover, utilising the ±45◦ tilts allows the specimens to be rotated by 45◦ with respect to the wedge, which provides a view on two faces rather than a single one during in-situ testing, which is a particular benefit during in-situ SEM testing [4,5].

**Figure 2.** Specimen mounting and transfer concept. The actual sample follows a TEM specimen geometry, with adaptive holders to fit to different in situ TEM/SEM testing platforms as well as general focussed ion beam (FIB) fabrication and post testing SEM analysis.

The specimen sizes discussed here will be kept in the micrometre regime, spanning between 2 μm and 4 μm, respectively, taking into account that the emphasis of the present study is placed on bulk deformation mechanisms rather than effects specific to a nanoscale sample nature. While the strength values might thus be to some extent affected by the specimen size, the dimensions are above the typical dislocation spacing in well annealed crystals [27]. As such, we expect largely bulk deformation mechanisms to be examined [28].

The transferable sample mounting scheme proves a valuable asset also to examine the initial crystal orientation. Deviations from the planned crystallography could cause activation of unintended slip systems. So could the presence of unnoticed growth twins or occurring small angle boundaries [29]. While the crystallographic orientation of the specimen can be routinely accessed using electron backscatter diffraction (EBSD), with some effort even in a quasi in-situ manner [11], the orientation information from diffraction information in the TEM is much more sensitive, in particular with respect to low angle crystal defects that are challenging to resolve by standard EBSD. Furthermore, the effect of surface oxides or presence of initial dislocations can be easily assessed. Here, the specimens were routinely analysed using TEM, and an example for a 4-μm Mg tensile specimen is shown in Figure 3. Of course the sample is not fully electron transparent over the entire cross-section, but the whole length along both edges can be inspected for crystallographic defects, as exemplarily shown in Figure 3a,b for the bottom and head part of the sample, and the diffraction information in these regions is assessable, as shown for regions 3 and 4 in Figure 3c,d, to detail the crystallographic loading direction. In analogy, samples can be inspected post testing in the TEM.

**Figure 3.** Pre-testing TEM analysis of a 4 μm thick Mg tensile sample, showing the gauge section towards the bottom (**a**) and head part (**b**), respectively. The diffraction patterns in (**c**) show the expected crystallography, while the image in (**d**) indicates redeposition of a nanocrystalline MgO layer. A white arrow indicates loading direction in real and reciprocal space.

To cover a number of possible deformation mechanisms [30], we examined loading directions along Mg 2112 , Mg[0001], and Mg 2110 , respectively. An overview of the respective orientations, loading modes (compression/tension) and the Schmid factors corresponding on the one hand to dislocation plasticity on the basal, prismatic or pyramidal plane, and on the other hand to tension or contraction twinning, respectively, are given in Table 1. In addition, the deformation mode noted in the in-situ experiments as detailed later is noted. To relate this crystal orientation information to the testing geometry, in


Figure 4 the different crystal orientations are visualised in combination with the unit cell and the expected deformation mechanisms based on the actual Schmid [30].

**Figure 4.** Schematic overview of the studied crystal orientations, the related unit cell orientation with respect to the uniaxial testing volume, and the conceivable deformation mechanism(s). The introduced colour code is kept consistent throughout the manuscript.

To load the compression samples, a conical conductive diamond tip with a 16-μm diameter flat punch (Synton MDP, Nidau, Switzerland) was used, while for tensile testing, a gripping tool FIB fabricated from a heavily drawn W wire was employed [23]. Before running the actual deformation experiments, the stiffness of the respective lamella was determined for later compliance corrections. The miniaturised in-situ compression and tensile tests were conducted under displacement control at a nominally constant strain rate of 1·10−<sup>3</sup> <sup>s</sup>−1, and a corresponding video of the experiment was recorded using a frame grabber at a repetition rate of 1 Hz. This allows for a trade-off between temporal resolution upon direct observation and sufficient signal-to-noise ratio for later digital image analysis, pillar sink-in correction, and analysis of local strain, respectively.


**Table 1.** Summary of Mg crystal orientations tested, loading mode, nominal Schmid factors, and experimentally observed deformation mode.
