Novel Fabrication Routes of Metallic Micromembranes for In Situ Mechanical Testing

Abstract

A methodology to miniaturize mechanical tests of metal alloys based on membrane deformation was developed in this investigation. The buildup of this new path for miniaturization tests requires small amounts of material for testing. This is of particular interest for irradiated structural nuclear materials. Micrometric metallic circular membranes were fabricated starting from thin alloy foils and using two different paths. Serial fabrication of microspecimens was performed by means of successive focused ion beam (FIB) steps. On the other hand, high-throughput parallel fabrication was achieved by differential sputtering (DS) based on reactive ion etching followed by a final fine FIB polishing to flatten the membranes and straighten the mechanical response. Micro-punch tests were performed using spherical tips and the in situ load–displacement curves were recorded while monitoring the test in a scanning electron microscope. The values reached after testing of the DS membranes were more reliable than those of FIB samples, showing a large stretching section and higher values of maximum force (64 mN) and displacement (22.2 μm). The micro-punch testing methodology developed in this work combines the advantage of facilitating the interpretation of the mechanical response, by producing a bi-axial stress distribution during membrane stretching, while being amenable to high-throughput microspecimen fabrication.

1. Introduction

The development of novel routes to perform mechanical tests on advanced materials at the microscale range represents a breakthrough in the assessment of the behavior of materials only available in small dimensions, such as coatings, ion-irradiated layers, or metallic alloys activated by neutron irradiation [1]. In the latter, the avoidance of tedious infrastructures for their handling implies a decrease in the analysis costs and time. In recent years, the spreading and development of focused ion beam (FIB) to micromachine tailored structures has enabled the fabrication of micro-size specimens and has allowed the reduction in the volume of material needed to test basic mechanical properties [2,3]. In parallel, advances in mechanical instrumentation have led to the development of testing devices that can be operated inside a scanning electron microscope (SEM) chamber and are equipped with bi-directional transducers operating both in compression and traction modes [4,5]. Such in situ testing platforms allow the assessment of varied mechanical properties at the micro-scale under simultaneous observation of the deformation mode [6,7,8]. However, the adoption of micromechanical testing methods for engineering purposes still presents some challenges, either related to cumbersome and time-consuming specimen fabrication (e.g., to produce micro-tensile specimens), or the need to apply complex models to transfer the measured microscale properties into the macroscale-equivalent counterpart, as in the analysis of nanoindentation data. The ultimate goal of small-scale mechanical testing techniques applied to engineering is to derive material-constitutive equations to input models used for the safe design of components [9]. The micromechanical tests used so far have shortcomings, in terms of the complex stress distribution during loading (e.g., nanoindentation), the difficulty of capturing mechanical parameters characteristic of failure (nanoindentation, micro-pillar compression), or the inconvenient and time-consuming specimen and gripping assembly (micro-tensile tests) [10,11,12]. In addition, the results are affected by size effects (i.e., dependence of the measured properties on the extent of deformation or the size of the tested specimens), which are still not completely understood [13,14]. Therefore, the development of new methods able to increase the sample production rate and extract micromechanical information more readily transferable to macroscale behavior will foster the screening and assessment of safe structural materials.

In this work, some of the above-mentioned issues have been addressed by developing a biaxial tensile punch test based on the deformation of micromembranes fabricated by FIB and differential sputtering (DS). On the one hand, FIB milling is a well-known technique capable of working at multiple scales (nano-to-micro) with great accuracy, although being limited to serial manufacturing and, thus, time-consuming [15,16]. On the other hand, microfabrication techniques are well-established processes in the silicon and micro-electro-mechanical systems (MEMS) production industries. However, the equivalent for steel samples has not yet been developed. Indeed, the existing microfabrication processes are mainly based on dry etching or reactive ion etching (RIE) in which a volatile compound is formed by a chemical reaction between a reactive gas (e.g., fluorine, chlorine) and the substrate to etch (e.g., Si). The chemical reaction is enhanced and controlled by the ion bombardment in the case of RIE. As there is no volatile compound that can be created with Fe, Ni, and Cr, at temperatures and pressures compatible with an industrial process, in this investigation a different approach was chosen, based on the differential sputtering rates between a light element (C of the photoresist) and the compounds of the steel used in these experiments. A factor of around 10 is expected between the sputtering of C and Fe, Cr, and Ni, as calculated by the software stopping and range of ions in matter (SRIM) [17] depending on ion mass and energy. Adjusting the thickness of the mask allows for controlling the depth of the etched structures of the substrate. DS therefore allows working in parallel on macroscopic areas, simultaneously milling a large number of micromembranes. However, DS requires longer batch machining times, due to lower ion current densities obtained with a Kaufman source, as compared to focused ion beam [18,19].

This study explored the viability of using two different fabrication techniques, FIB and DS, to mill circular steel micromembranes that were subsequently mechanically tested. The mechanical response of these membranes was evaluated by means of a new micromechanical test, designated as the micro-punch test. The advantage of this approach is a simple conception of the specimen geometry, which allows a high-throughput microfabrication, and an easier transferability to equivalent tensile properties, due to the tension nature of the deformation.

2. Experimental Procedure

An AISI 316 L austenitic stainless-steel rod of 20 mm in diameter (18Cr-10Ni-3Mo in wt. %) and a T91 ferritic-martensitic steel plate with the composition shown in

Table 1. Chemical composition of T91 steel in wt. % (supplied by Industeel).

Fe	C	Mn	P	S	Si	Cu	Ni	Cr	Mo	Al	Nb	V	Ti	N
Bal.	0.102	0.401	0.019	0.0007	0.235	0.085	0.121	8.895	0.889	0.010	0.079	0.202	0.004	0.048

were used to optimize some microfabrication parameters. T91 thin foils were prepared from a hot-rolled 15 mm-thick plate, which was in the normalized and tempered metallurgical condition (solution-annealed for 15 min at 1050 °C, fast cooled in water, reheated 45 min at 770 °C, and cooled in air). T91 exhibits high corrosion resistance, steam oxidation resistance, and good creep rupture strength. Due to these properties, this ferritic-martensitic steel has been widely used in large-capacity and high-efficiency subcritical and supercritical units of new and in-service power plants worldwide [20]. T91 has also been proposed as a candidate structural material for sodium- and lead-cooled fast reactors because of its resistance to swelling compared to austenitic steels [21].

FIB micromembranes were fabricated in a thin foil of a disc 3 mm in diameter, whereas DS micromembranes were milled in a rectangular thin foil of 4 × 7 mm². The side of the T91 thin foils receiving the FIB and DS milling was prepared with a P500 grinding paper, while the other side, where the micro-punch test would be performed, was polished down with 1 µm diamond paste. The final thickness of both thin foils was less than 100 µm.

The production of the circular micromembranes with final thicknesses ranging from 8 to 33 µm was carried out by FIB (22 µm diameter membranes) and DS (120 µm diameter membranes) techniques impinging on the ground side of the thin foils. The FIB micromachining was performed with a Nova 600 FEI system (FEI, Oregon, OR, USA), which combines a field emission gun (FEG) scanning electron microscope with a gallium focused ion beam. FIB was exploited for the direct fabrication of the metallic membranes or as a final fine polishing step following DS. The conditions selected for the milling of the FIB membranes can be summarized as follows: acceleration voltage ΔV_a = 30 kV, ion current I = 9.3 nA, dwell time = 1 µs, and fabrication time Δt ≈ 5 h per 22 µm diameter membrane in a thin foil around 54 µm thick. In the case of the final polishing of membranes manufactured by DS, the conditions applied were acceleration voltage ΔV_a = 30 kV, ion current I = 21 nA, dwell time = 1 µs, and fabrication time Δt ≈ 7–8 h per 120 µm-diameter membrane in a thin foil 68–95 µm thick. FIB was always operated under a vacuum of approximately 3.5 × 10⁻⁵ mbar. The Ga ion implantation took place in the milled side of the thin foils. Considering data found in the literature, the Ga⁺ concentration in steels reaches a steady-state characterized by a depth of a few tens of nanometers [22,23]. Therefore, the effects of sputtering or the Ga⁺ implanted in the backside (testing side) of membranes with thicknesses between 8 and 33 µm are expected to be negligible. Likewise, structural changes produced by the FIB or DS fabrication processes will be localized on the side of the membrane opposite to the side where the micro-punch test is conducted.

A Kaufman KDC40 Ion Source (Kaufman & Robinson Inc., Fort Collins, CO, USA) characterized by a beam size diameter Φ ≈ 4 cm, was used to perform the differential sputtering. The experimental conditions selected for the fabrication of the membranes were Ar gas flow = 5 sccm, ion beam voltage ΔV_b = 600 V, ion beam current I_b = 30 mA, grids acceleration voltage ΔV_a = 120 V, and fabrication time t ≈ 40 h per membrane array. Spin coating (WS-400B-6NPP/LITE equipment from Laurell Technologies Corporation, North Wales, PA, USA) and photolithography (UV-KUB 2 from Kloé, Saint-Mathieu-de-Tréviers, France) techniques were conducted to prepare the photoresist masks needed for differential sputtering.

The thicknesses of the membranes were measured by means of a 3D optical microscopy technique (Alicona InfiniteFocusSL system from Bruker Alicona, Raba, Austria) equipped with a rotation unit (Alicona 3DRotationUnit G2) and coupled to the IF-MeasureSuite Version 5.1 software (Alicona Imaging GmbH, Raba, Austria).

In order to obtain the bottom membrane profiles, a contact Alpha-Step IQ stylus profilometer (KLA-Tencor, California, CA, USA), equipped with a diamond tip with radius R = 5 µm and half-opening angle α = 30°, was utilized. Scan length, scan speed, and sample rate were typically operated in the range of 0.5–2 mm, 2–200 µm/s, and 50–1000 Hz, respectively.

The steel membranes were mechanically tested inside an SEM (Zeiss, Leo Supra 50, Jena, Germany) with an in situ micromechanical testing system (Alemnis Standard Assembly from Alemnis AG, Thun, Switzerland). For controlling the testing system, the AMICS software (Alemnis AG, Thun, Switzerland) was used. In the case of FIB membranes, a spherical diamond tip of 5 µm radius was chosen for the micro-punch tests together with a punch velocity fixed to 50 nm/s, while for the DS membranes, the tests were performed using a 10 µm radius diamond tip, with a punch velocity of 200 nm/s. The displacement of the punch was selected taking into account the average diameter of the membranes (22 µm for FIB micromembranes and 120 µm for DS membranes). All tests were carried out under vacuum (2 × 10⁻⁵ mbar) at room temperature. During the tests, the load applied versus the deflection in the membranes was recorded and used to analyze the mechanical behavior of the alloy.

3. Results and Discussion

3.1. Membrane Fabrication by FIB Milling

An array of 24 wells milled by FIB is shown in Figure 1. The diameters of the bottom-wells were measured on SEM images and the estimated average value was 22 ± 1 µm. Alignment marks were created by milling 11 wells until reaching pass-through holes. The pass-through holes were used as a reference to guide the positioning of the punch on the 13 membranes during mechanical testing. In the milled side of the thin foil, we created parallel rows in which membranes and pass-through holes are alternately arranged at a fixed distance. The membranes are located either exactly in the center between two pass-through holes, or at the distance fixed between holes and membranes if they are in the extreme of a row. Thus, this configuration allowed the punch to be placed easily on the membranes in the testing side of the thin foil (nonmilled side), where only the pass-through holes are visible.

3.1.1. Estimation of the FIB Wells Depth by SEM-Based Geometric Calibration

A precise measurement of the main geometrical features of the microspecimens is of great importance for a correct analysis and modeling of the micromechanical tests results. The depth-to-diameter ratio of the wells carved by FIB did not allow for an accurate direct in situ measurement of the milling depth, as the SEM microscope cannot provide a cross-sectional view of the bottom of the wells after a depth of some tens of microns. An indirect depth estimation could be performed based on sputtering rates. However, theoretical sputtering rates are affected by several experimental sources of uncertainty, such as surface roughening, inhomogeneities in the specimen thickness, slight movements of the sample during the sputtering process, and polycrystalline grain orientations locally affecting sputtering efficiency. In addition, the re-deposition of sputtered material during milling causes deviations from the expected final depth of the wells that should be taken into account in order to have a more precise depth calibration curve. An increase in re-deposition rate alongside the milled depth indicates a sub-linear relationship between total ion dose (proportional to milling time) and effective experimental depth [24], which could become significant after a few tens of microns. In this investigation, an accurate evaluation of the membrane thickness was achieved following three steps: (i) building a calibration curve according to SEM-based geometric measurements at shallow depths (limited to depths that can be visualized by SEM, which were around 30 µm) (Figure 2); (ii) fitting the obtained curve with a power-law model; and (iii) estimating the expected milled depth for larger ion doses by extrapolation, based on the fitting equation parameters.

Taking into account that the ion dose used to mill the T91 FIB membranes of this work was fixed to 173.4 nC/µm², the estimated depth of the well using the previous calibration curve would be 34 µm. Although this calibration procedure is effective at estimating the membrane thickness when the initial thickness of the thin foil is known, the result depends on the structures geometry as well as FIB milling parameters, such as beam acceleration voltage and current. Therefore, a limitation of this procedure is that a change in one of these conditions would require a new calibration procedure.

3.1.2. Characterization of the FIB Membranes by 3D Optical Microscopy

Three-dimensional optical microscopy was used to image the membranes’ array and to measure the thin foil and membranes’ thicknesses. The technique is based on the so-called focus variation technology, in which, in response to surface roughness variations, the system optics are moved vertically to keep the sample surface locally focused while shifting on the X-Y plane, recording a 3D map of the sample surface with a theoretical sub-micrometric resolution. In order to avoid reflections while using 3D optical images, the surface of the sample cannot be mirror-like polished. First, the total thickness of the thin foil was measured and compared to that evaluated with a digital probe indicator. For this purpose, a 3D rotation unit was used, holding the T91 sample with a magnet grabbed to a clamp. The optical 3D image obtained and the measurement of the mean thickness of the thin foil are shown in Figure 3. The average thickness value was calculated by averaging the distance between points located on both sides of the thin foil (in blue in the diagram).

Thus, the average thickness measured in the region where the membranes were located was equal to 54 ± 1 µm, with this value being in good agreement with the measurement obtained by the probe indicator at the center of the thin foil sample (52 µm). Then, a top scan (Figure 4) was taken from the nonpolished surface to calculate the thickness of each membrane.

The average membrane thickness was determined considering the average depth of the FIB wells measured from the top scan, which was 33 ± 1 µm (value also in accordance with the previous calibration in which the estimated depth of the wells was 34 µm) and subtracting these values from the thin foil thickness. Thus, it was possible to estimate an average membrane thickness of 21 ± 1 µm.

3.2. Membrane Fabrication by DS

Compared to FIB, DS offers the advantage of working in parallel on larger areas, simultaneously milling a large number of micromembranes at the cost of longer machining times per batch, due to lower ion current densities. For DS processing, a polymer mask, properly patterned by means of photolithographic techniques, locally protects the underlying metallic sample during the sputtering process, thanks to a highly favorable sputtering yield ratio over steel. After patterning, the remaining mask is removed by solvents or diluted acid solutions selective toward carbon-based materials, without damaging the metal.

3.2.1. Pattern Transfer through the Photoresist Masks and DS Membrane Array Fabrication

The main challenges encountered during DS microfabrication were the deviations from the expected theoretical behavior of the photoresist etching due to the insulating character of the polymeric mask. According to simulations by SRIM-2013 software [17], the sputtering yield ratio of common epoxy-resin photoresists (e.g., SU8) over steel was approximately 1/10, mainly thanks to the facilitation of ion implantation or penetration over sputtering. This implies that a 10–20 µm-thick photoresist mask should be safe enough to mill the unprotected steel regions for several tens of microns in depth. However, the impossibility to dissipate the kinetic energy of ions impinging on the insulating photoresist led to a significant increase in its local temperature. This phenomenon was responsible for a relevant increase in the photoresist effective sputtering yield due to evaporation/sublimation. Thus, a large deviation from theoretical simulations was observed, as well as an overall nonlinearity and unpredictability of the system behavior for long working times. As it can be inferred from

Table 2. Measured sputtering rates in µm/hour using a Kaufman Ion Gun on Si, 316 L, and T91 steel samples, as a function of beam voltages and ion currents corresponding to approximately 50% (middle column) and 95% (last column) of total available power.

Material	Sputtering Rate (Vb = 600 V, Ib = 30 mA)	Sputtering Rate (Vb = 1000 V, Ib = 100 mA)
Si	1.7 ± 0.2	7.4 ± 0.4
316 L	1.3 ± 0.5	6.5 ± 0.2
T91	1.9 ± 0.4	8.0 ± 0.4

, the sputtering rate was greatly increased with the increase in the beam voltage and current close to the maximum power offered by the Kaufman ion gun used for DS.

Therefore, the total working time needed to mill structures characterized by a depth of several tens of microns was drastically reduced for high powers. On the other hand, unfortunately, this amount of power could not be sustained by the photoresist mask, which was essentially destroyed after one hour of sputtering. As a consequence, the milder working conditions were selected (V_b = 600 V and I_b = 30 mA) along with longer exposure times (some tens of hours). Under these experimental parameters, the photoresist mask was able to still sustain a prolonged exposure to the ion beam, despite being subjected to a progressive pyrolysis/degradation and showing a nonlinear behavior for long working times (Figure 5).

The experimental sputtering rate ratio, measured for a test with a SU8 mask pattern on silicon wafers, was far from theoretical predictions, as highlighted in Figure 5b, and a nonlinear trend with time was shown. Depth variations (Δh) of the double-layer made of silicon and SU8 were measured by a profilometer and averaged on ten wells, relative to the average “pre-sputtering” height (h₀). A zero value of Δh would imply an equal sputtering rate of SU8 and silicon, whereas negative values translate into a photoresist sputtering rate larger than that of silicon.

The first set of DS-milled membranes in a T91 steel sample were produced by applying relatively mild sputtering parameters (V_b = 600 V and I_b = 30 mA) with a total machining time of around 40 h (Figure 6).

3.2.2. Fine-Tuned Polishing of Membrane Profiles: Removing Thickness Unevenness

Both FIB and DS membranes typically showed a second-order, parabolic-like profile, characterized by uneven thickness featuring minima points along the membrane boundary and a maximum in the central region (Figure 7).

This inhomogeneity could reach values up to 10 µm or more for milled depths of some tens of microns, thus being comparable to the membrane thickness. As a direct consequence, mechanical micro-punch tests performed in the center of the membrane, to obtain information on the mechanical stress behavior up to the breaking point, could be hindered by a premature failure of the whole membrane along its border.

This phenomenon is intrinsic to the physics of ion beam sputtering and independent of the selected sputtering technique (FIB or DS). An anisotropic distribution of re-deposited material, favoring the well center over borders, may worsen the curvature of the bottom of the wells. However, the main cause is likely to be attributed to the sputtering yield dependency on the incidence angle, which typically reaches a maximum approximately at an angle α = 75–80° from the normal to the surface [22].

The increase in sputtering yield alongside incidence angle is essentially determined by a growth in the overlap between the beam–sample interaction volume and the sample surface. Thus, under fixed experimental conditions, sputtering yields are expected to be larger close to the well borders than at the center, due to a partial overlap of the ion beam with sidewall surfaces at a locally high incidence angle (Figure 7c). This effect is likely to produce negligible differences in the bottom profile of the well for shallow depths, but considerable for depths of tens of microns.

Nevertheless, FIB capabilities provide a route to flatten the parabolic profiles. Three-dimensional customized shapes can be milled by means of properly designed greyscale bitmaps where each different pixel color will translate into a different milling nominal depth, linearly interpolated from a codified table in-between the user-assigned value corresponding to maximum (white) height and the zero-value corresponding to minimum (black) height. Large membrane diameters, as in the case of DS membranes, allow for a direct characterization of the bottom profile by means of 1D profilometry, which can be used to create a greyscale bitmap to feed a second FIB-polishing step to compensate for the uneven thickness.

The following approach was used to homogenize the thickness of the DS membranes: (i) measuring the membrane bottom profile after the main sputtering process; (ii) fitting the obtained 1D profile by means of a ellipsoidal function; (iii) building a 3D greyscale matrix by evaluating the fitted 1D function over a 2D mesh grid matching with the membrane geometry, assuming rotational symmetry of the 1D profile; (iv) converting the obtained matrix into a greyscale bitmap; and (v) loading the bitmap file into FIB software and tuning milling parameters to obtain the desired milling depth.

This FIB-polishing method was applied to the DS array of membranes to remove the uneven parabolic profiles. Comparison of the bottom profiles acquired through a contact-stylus profilometer before and after the FIB-based correction showed a significant reduction in the starting curvature (Figure 8). FIB was also used to machine the small pass-through holes acting as a reference to localize the membrane position on the sample testing side.

After FIB polishing, the dimensions of these DS membranes were also characterized by 3D-optical microscopy. This optical study revealed that the diameter was around 120 µm for all the DS membranes, whereas the thickness after the fine-tuned FIB polishing was different for each membrane, varying from 8 to 33 µm. As the FIB milling conditions applied were the same for all the microsamples, these variations could be due to a nonuniform milling during the DS fabrication of the wells and/or to the differences detected in the thickness (from 95 to 68 µm thick) along the T91 thin foil.

3.3. In Situ Micro-Punch Tests

In order to analyze the viability of performing micro-punch tests on micrometer-sized metallic specimens, a first set of micromechanical tests was performed on the FIB membranes described before. The values of the average membranes’ diameter and thickness were 22 and 21 µm, respectively, as mentioned above. Thus, the aspect ratio of these T91 samples was close to the unit, below the aspect ratio for conventional miniature small punch tests, which is typically 7 when the discs tested are 3 mm in diameter and 250 µm in thickness with a receiving die diameter of 1.75 mm. The punch selected for the test was spherical with a tip radius of 5 µm. The three membranes tested were located exactly halfway between two pass-through holes. Then, the precise movement of the in situ micromechanical testing system allowed the placing of the center of the spherical tip exactly above the center of the membrane. A scheme of the micro-punch test performed is shown in Figure 9.

The tests were conducted at room temperature in displacement control mode using a punch displacement of 50 nm/s. The force needed to push the punch through the membrane was recorded and plotted as a function of the displacement of the punch into the material. Thus, the load vs. displacement curves obtained during the micro-punch tests performed on the membranes are given in Figure 10.

These curves showed an elastic bending followed by plastic deformation until reaching a maximum force. Once reaching the maximum force, the strength decreased gradually as a consequence of the crack initiation.

The curves did not present high reproducibility from the onset of plastic deformation, probably due to the premature fracture of the specimens along the boundaries between the circular membranes and the bulk material, together with small variability in the positioning of the punch with respect to the center of the membrane. In fact, the SEM observation of the membranes during the in situ micro-punch test (Figure 11) clearly showed that the specimens were preferably broken along the boundaries between the membrane and the bulk, as it would be expected due to the parabolic bottom shape of the FIB membranes that was not removed in these samples. It is well known that the geometric discontinuities in a material, as in this particular case with a nonuniform thickness and fillet radius, cause a localized stress increase above the average stress, leading to the premature fracture along the boundaries [25]. In addition, due to the membrane aspect ratio close to the unit, a prominent indent of the spherical tip appeared at the center of the membrane. Therefore, the measured displacement of the tip into the material was a convolution of indentation and membrane deformation.

To improve the reproducibility and avoid the fracture along the boundaries of the membrane, the aspect ratio of the membranes should be increased and the parabolic side of the membrane flattened. Therefore, with the aim of avoiding the problems found during the testing of the FIB membranes, DS specimens followed by FIB polishing were tested. These specimens exhibited a higher aspect ratio (between 4–15) and, due to the last step of FIB milling, which was added to remove the parabolic bottom of the wells, the micromembranes had a more uniform thickness. In these tests, the displacement control mode was selected again, whereas the punch displacement was increased and fixed to 200 nm/s taking into account that the average DS membranes diameter was around 120 µm. The curves obtained are displayed in the load vs. displacement graphic of Figure 12. It is possible to distinguish four different regions in the curves obtained for the DS membranes. There exist a first small elastic section, a second plastic region, followed by a large membrane-stretching region until reaching a maximum strength, and a last subsequent fast decrease in the load due to the fracture of the specimen. From these results, it seems that the increase in the thickness of the membrane led to a higher maximum load achieved together with a larger displacement to fracture.

One important fact is that the in situ observation during the tests confirmed that the increment in the aspect ratio of the membranes together with a fine FIB polishing, removing the parabolic profile in the bottom of the wells, avoided the significant weakness along the boundaries that would produce an early fracture. This is clearly shown in Figure 13, where it is possible to observe the fracture of the 20 µm-thick membrane inside a small circular indent of around 25 µm in diameter created by the punch tip.

The comparison of the results obtained for FIB and DS membranes (Figure 10 and Figure 12) showed that the data achieved for DS specimens were more reliable as the curves had a large stretching section and they reached higher values of maximum force and displacement. It is possible to directly compare the values of load and displacement for FIB and DS membranes of around 20 µm thick. Thus, in the case of FIB membranes, the values achieved were 69 mN and 2.6 µm in the most favorable test, whereas for the DS membranes, the values reached were 164 mN and 22.2 µm, respectively.

4. Conclusions

In this work, two different methods to fabricate metallic micromembranes were evaluated and a novel mechanical test, designated as the micro-punch test, was developed. The main conclusions are as follows:

• Micromembranes were fabricated from T91 steel thin foils using two different milling methods: FIB and DS.
• FIB exhibited different limitations as a membrane fabrication method, mainly related to its characteristic of involving a serial production with long processing times, and to uneven re-deposition phenomena of sputtered material leading to a pronounced curvature on the bottom membrane profile.
• DS also exhibited some limitations as a membrane fabrication method, again principally related to uneven re-deposition of sputtered material leading to parabolic profiles. However, its main advantage is based on the simultaneous milling of a great number of micromembranes. Thus, this method presented a significantly higher throughput and shorter fabrication times in comparison to FIB.
• A new micro-punch test for metals was developed in this investigation. This test was performed on FIB and DS micromembranes.
• A combination of DS followed by fine FIB polishing seemed to be the best route in order to achieve a faster throughput of the overall micromachining process. The premature fracture of the membrane along its border was avoided in the DS samples by flattening the parabolic profile and increasing the membrane aspect ratio. In addition, this DS fabrication method was more reliable than only using FIB to probe membrane stretching.

This work has confirmed the feasibility of (a) the fabrication protocols implying DS followed by fine FIB polishing, and (b) the testing setup for characterizing the micromechanical behavior of metallic alloys.