M-SRE was obtained by mixing PDMS pre-polymer (1:10 in volume curing agent/base elastomer) and six different FF volume ratios (S₀ = 0%; S₁ = 1.7%; S₂ = 3.5%; S₃ = 9.8%; S₄ = 15.4%; S₅ = 20.3%; and S₆ = 24.6%), corresponding to Fe₃O₄ nanoparticles’ weight concentrations of S₀ = 0%; S₁ = 0.2%; S₂ = 0.4%; S₃ = 1.1%; S₄ = 1.7%; S₅ = 2.3%; and S₆ = 2.8%. After homogeneous dispersion, the non-polymerized material was degassed to remove air bubbles and baked so as to achieve 3D polymeric cross-linking.

The structural homogeneity, biocompatibility, optical transmittance, and magnetization were studied in bulk using 2 cm × 2 cm slab samples and having a thickness of 250 μm. The M-SREs Young’s modulus has been obtained by fabricating the cantilevers shown in Figure 1A with dimensions 500 μm (width, w), 4,400 μm (length, l), and 250 μm (thickness, h) and by measuring the maximum deflection when 1 g acceleration was applied. Figure 1B shows the fabricated cantilever using S₅. The final soft microsystem was focused toward enhancing its magneto–mechanical response. To this effect, a rectangular seismic mass (with dimensions w_m, l_m, and h_m, in micrometers being of 1,600, 2,500, and 250) was included at the cantilever tip, as shown in Figure 1C (design) and Figure 1D (also fabricated using S₅).

One of the most important aspects on synthetizing SRE is to set the filler concentration upper threshold that the polymeric matrix can host, since incomplete polymerization may lead to fragile, brittle, or materials with non-homogeneous response that may leach the filler and therefore have a time-dependent actuation. For thermocurable materials – such as the family of M-SREs here presented – the curing time is indicative of how the filler affects the 3D cross-linking of the polymeric matrix. In this context, the reference was taken from the curing step for the non-doped material (S₀, with 0 wt% FF), which was of 20 min at 80°C. Identical curing step was required for samples S₁–S₃, indicative that the filler, in the concentrations used for these specific M-SRE, did not affect the formation of the 3D polymeric matrix at these concentrations. For samples S₄ and S₅, the curing time was increased to 25 min and for sample S₆ up to 35 min. Although it was possible to cure this last sample, such an increase of the curing time was highly significant, and variations on the polymer structural properties were to be expected. For this reason, this FF concentration was considered to be the upper threshold.

In parallel, slab samples with low (S₂) and high (S₆) FF concentration were used for the image analysis to calculate the average particle diameter, taking into account that the MNP included in the FF have the magnetic core and a polymeric capping. Histograms illustrating nanoparticle size distribution (D_TEM) are presented in Figures 2A,B, together with representative TEM images (insets in both figures). The deconvolution of such histograms in Gaussian peaks can also be seen in Figure 2, and more detailed information is provided in Table 1. For sample S₂ (Figure 2A), a predominant peak centered at a particle size of 8.9 ± 0.1 nm was obtained, which could be associated to non-aggregated MNP. A secondary peak was also noticeable located at a particle size of 15.5 ± 0.5 nm. With an area of only 11% as compared to the predominant peak, this second Gaussian peak is due to aggregates of two MNPs. Figure 2B shows that the histogram for sample S₆ could be deconvoluted into three main peaks, corresponding to single MNPs (8.2 ± 0.1 nm), two MNPs’ aggregate (13.2 ± 0.1 nm), and three MNPs’ aggregate (19.0 ± 0.6 nm). As compared to the predominant peak of this sample, peaks #2 and #3 had a relative area of 44 and 13%, respectively. These data suggested that the increase of MNPs caused a significant increase of small aggregates – formed by two or three MNPs – which do not hamper the overall behavior of the M-SRE since their distribution is homogeneous in the material, as it can be seen at the insets of Figure 2. From these data, it can be confirmed that the PDMS matrix can host up to four times more FF volumes than those previously presented (Pirmoradi et al., 2010; Balasoiu et al., 2011) without having large aggregates. In turn, this higher FF concentration will assure a higher actuation in response to a magnetic stimulus, as it will be discussed in the following sections.

The development of lab on chip systems and MEMS has been strongly linked to biomedical applications, becoming a major application field for these architectures. Lab on a chip systems for biomedicine should guarantee the integrity and stability of the sample under study, that being biological fluids, cells, or tissues, among others. The requirements for materials and systems applied to biomedicine are thus very strict, being biocompatibility one of the most important to be determined. Based on the latter, the biocompatibility of the proposed material was evaluated using vascular smooth muscle cells (VSMC) from rat aorta as model eukaryotic cells, considering their relevance in cardiovascular pharmacology. Figure 3 shows the cell viability for M-SRE samples (from S₀ to S₆). Viability tests were performed using the colorimetric MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay (Mosmann, 1983) After pre-attachment of the VSMC in the well, M-SRE samples with a total volume of 40 mm³ were introduced in each well, and the cell viability was measured after 24 h of incubation with the MTT.

Viability was determined as a percentage in relation to control samples, i.e., cells proliferation in a well without M-SRE. Between three and six replicates were done for each M-SRE for statistical validation. S₀–S₄ samples showed values similar to 100% viability. On the contrary, sample S₅ and particularly S₆ showed toxicity. In sample S₄, the % cell viability was of 99 ± 12%. This magnitude decreased until 88 ± 3% for S₅ and 23 ± 13% for S₆. Therefore, it can be concluded that S₀–S₄ samples are suitable for biological applications.

Figure 4 shows the experimentally measured magnetization as a function of the applied magnetic field, μ₀H, for slab samples S₁–S₆ at 300 K. None of them shows either coercitivity or remanence, which is indicative of their superparamagnetic behavior. Actually, this is in agreement with the superparamagnetic nature of the Fe₃O₄ nanoparticles at room temperature (Ivaneyko et al., 2012). On the same figure, the saturation magnetization (m_s) (understood as the total magnetization corresponding to an applied field of 1.2 T) for each sample is included.

According to the superparamagnetic theory (Cullity, 1974), the magnetic particle size can be calculated from the true magnetic moment of each particle (μ) using the Langevin Eq. 1, which was used for fitting the magnetization curves shown in Figure 4.

(1)m=mS[coth(μHkBT)−kBTμH], where k_B is the Boltzmann constant, T is the absolute temperature, m_S is the saturation magnetization, H is the magnetic field, and μ is the true magnetic moment, which is defined as (2)μ=mSπDL3/6 being D_L is the magnetic particle size.

The obtained D_L values were compared with the values from TEM observations, i.e., those corresponding to the main peak – i.e., without aggregates – (D_TEM). The results are summarized in Table 2. Differences between D_L and D_TEM were due to the fact that magnetic measurements did not take into account the presence of the capping MNPs layer. The particle size was recalculated including the thickness of the capping, which is a₀ = 0.83 nm for magnetite (Kaiser and Miskolcz, 1970). Introducing this correction into the Langevin function leaded to a modified magnetic particle size DL′ = D_L + a₀. Once this aspect was considered, a good agreement was obtained then between the DL′ value from Eq. 2 and the TEM measurements (Table 2). On the basis of such results, it can be confirmed that the small aggregates formed in the M-SRE – comprised of two or three MNPs – and its homogeneous distribution in the sample do not cause a significant change of its magnetic properties, even in the samples with the highest FF concentration.

Materials with Young’s modulus in the order of kilopascal–megapascal are very attractive for life science applications since they present similar stiffness as biological entities such as arteries (Rodriguez-Rodriguez et al., 2014). In that context, since the main aim of this paper is to develop soft microsystems, it becomes crucial to experimentally measure the effect of the FF on the Young’s modulus of the SRE. By measuring the maximum deflection (y_max) of the cantilevers upright and upside down, the Y_M-SRE was experimentally obtained by subtracting the initial offset of the cantilever and using Eq. 3.

(3)YM-SRE=3ρgl42ymaxh2, where ρ is the M-SRE density, g the acceleration of the gravitational field, and l and h are the length and thickness of the cantilever, respectively.

Figure 5 shows the experimental Y_M-SRE values obtained. From an initial value close to 200 kPa for samples S₀–S₂, Y_M-SRE rapidly increases with the FF concentration until it reaches 20.3% volume FF. The observed increase of Y_M-SRE is quadratic with the FF concentration for samples S₁–S₅, as predicted by Bergstrom and Boyce (1999). Conversely, S₆ does not follow such tendency. The reason for this could be the above described incomplete formation of the 3D polymeric network due to the FF oversaturation, which again seems to confirm that the concentration limit for this specific FF in PDMS is obtained with S₅ sample.

As compared to the previously published results, the values here obtained are similar to those provided by the manufacturer datasheet (Y_PDMS = 300–800 kPa), even though other researchers have shown that Y_PDMS can be in the range of the megapascal (Johnston et al., 2014). As a rule, it has been demonstrated that Y_PDMS strongly varies on both the curing agent/base elastomer ratio, temperature, baking time, and baking conditions (Fuard et al., 2008; Johnston et al., 2014). Specifically, in Brau et al. (2011) and Terwagne et al. (2014), the Young’s modulus of PDMS is shown to vary between 10 and 1,000 kPa depending of the aforementioned technological conditions. In that sense, the values obtained here are in agreement with those previously published.

The optical properties of a polymeric matrix may strongly vary when an additive is included. When synthetizing the M-SRE, not only the particle size, distribution, shape, or capping of the FF but also the optical and chemical properties of the liquid carrier may potentially modify the optical transmittance of the M-SRE. This change is crucial if the idea is to use this material for defining magnetically actuable elements to be implemented in photonic lab on a chip systems (PhLoC) (Vila-Planas et al., 2011), for example, valves or pumps, where the whole spectral response of a given analyte is acquired in real time. Before such integration can be even considered, it becomes necessary to determine the effect of the FF concentration on the PDMS transmittance. Results presented in Figure 6 shows the transmittance for slab samples S₀ to S₆ in a wavelength range between 400 and 900 nm. In all cases, a propagation distance of the light in material of 250 μm has been used. As expected, the increase of the FF concentration resulted in a sharp decrease of the transmittance due to scattering centers and specific optical absorption from FF. Samples S₁–S₂ presented minimum transmittance (at a wavelength of 400 nm) of 59 and 46%, respectively. Therefore, these M-SRE can be considered suitable for developing PhLoC with magnetic actuation. Sample S₃ has a transmittance at the same wavelength of 30%, which may still be valid for PhLoC with short optical paths. Finally, samples with larger FF concentration (S₄–S₆) can be considered as opaque in the VIS–NIR range and, even though they will have a higher displacement upon magnetic actuation, they cannot be considered as valid M-SRE to implement PhLoCs or, more generally to develop elastomer-based systems where visual inspection or optical detection is required.

Once the M-SRE was characterized, the next step was to show the applicability of this material. Specifically, six different soft microsystems were fabricated – one for each FF concentration – and showed deflection in the order of hundreds of micrometer when an external magnetic field is applied. Figure 7 presents a cantilever fabricated with sample S₄ showing no actuation for weak applied magnetic fields (Figure 7A) and large displacements when larger magnetic fields are applied (Figure 7B).

Quantification of this deflection as a function of the applied magnetic field was measured, and it is shown in Figure 8. As it can be seen, deflection of the soft microsystem shows the expected increase in magnitude with μ₀H for all M-SREs under study. For a specific deflection target, higher μ₀H is required for samples with low FF concentration. As an example, for reaching a deflection of 500 μm, an applied magnetic field of 0.33 T is needed for S₁, whereas it can be decreased to 0.2 T for S₅. Finally, the mechanical sensitivity of the M-SRE with larger FF content is the highest. This means that larger deflections can be obtained but, equally important, small changes of this applied magnetic field will cause very significant variances in the resulting deflection. Therefore, for robust microsystems similar to these here presented, the optimal alternative is the S₃ material, where the deflection can accurately be controlled with the applied magnetic field. Microsystems done with S₄ and S₅ are still valid for ON/OFF actuation, such as microvalves. For quantitative actuation, they could be used by making the microsystems stiffer, i.e., increase the thickness or decrease its length.

To summarize, the different studied properties of the here discussed M-SRE in view of their application to implement magnetic soft microsystem are shown in Table 3. The selection of the given FF concentration strongly depends on the application. Low FF content materials (S₁–S₂) are suitable for photonic and biological applications (higher optical transmittance and biocompatibility) even when the required applied magnetic field so as to achieve actuation is the highest. S₄ and S₅ are opaque materials that present a significant increase of its Young’s modulus. The high FF concentration allows the largest deflection for relatively small applied magnetic fields, and therefore they are may be suitable for ON/OFF applications. A strategy if strong magnetic actuation is required in combination with photonics readout is to define the system in such a way that S₄ or S₅ M-SRE are only defined in specific regions (de Pedro et al., 2014), while the rest of the PhLoC is defined using S₀ (non-doped). In this approach, the advantages of both the types of materials could be merged in a single system, thus improving its performance. Measurements done with material S₆ seem to be indicative that the polymeric 3D cross-linking is distorted due to FF oversaturation. Finally, S₃ can be considered to be the optimal material, where biocompatibility, low Y_M-SRE, acceptable transmittance (for short optical paths) and, when used for microsystems, robust and controllable deflections, is achieved.

The negative-tone epoxy-based polymer SU-8 50 and its developer propylene glycol methyl ether acetate (PGMEA) (both from MicroChem, Corp., USA) were used following the datasheet specifications. PDMS elastomer (Sylgard Silicon Elastomer 184, Dow Corning Corp., USA) was prepared in accordance to the datasheet. FF was used as provided by the manufacturer (Liquids Research Limited, North Wales). The FF presents a uniform size distribution and low Fe₃O₄ particle agglomeration (<20 nm) dispersed into isoparafin with 10 nm as the mean particle diameter and a saturation magnetization of 400 G.

M-SRE was synthesized by mixing PDMS and FF. First, the PDMS pre-polymer was obtained in a 1:10 (v:v) curing agent/base elastomer as available from the manufacturer. Just afterward, different FF volumes were included (0; 1.7; 3.5; 9.8; 15.4; 20.3; and 24.6% in volume) and manually stirred until homogeneous dispersion is achieved. This mixture was then degassed to remove air bubbles. The resulting M-SREs were labeled as S_i (i = 0, 1, …, 6; corresponding to the increasing FF concentrations).

The slab samples, cantilevers, and soft microsystems were fabricated with soft lithography technology (Vila-Planas et al., 2011). The process started using a soda–lime wafer with a thickness of 700 μm, onto which a 250 μm SU-8 50 layer was spun. After 3-h baking at 95°C, the wafer was exposed to UV light with a dose of 350 mJ/cm² using the appropriate mask. The post exposure bake for 20 min at 95°C and the development using PGMEA finished the master fabrication. M-SRE pre-polymer obtained as above mentioned was poured onto the SU-8 master, avoiding overflow. A second degasification was done so as to remove air bubbles that may be trapped in the microstructures. Then, they were baked for completing polymerization. Just afterward, the slab samples, cantilevers, and soft microsystems, having a thickness of 250 μm, were peeled off from the master with the help of a tweezer and positioned over a glass substrate for handling easiness.

Structural homogeneity was determined using images from TEM, which were obtained by the system JEOL 2011, Jeol Ltd., USA. The distribution and size of the Fe₃O₄ nanoparticles included in the M-SRE was analyzed. From these images, 10 fields containing each of them between 20 and 100 MNP depending on the sample were analyzed by image software (ImageJ, USA).

Polydimethylsiloxane is known to be a compatible material with low or no toxicity (Wu et al., 2010), making it suitable to be used with samples of biological origin (cells, enzymes, bacteria, etc.). The study of biocompatibility of the developed M-SRE was carried out for verifying whether the inclusion of FF in the PDMS altered the toxicity of the polymeric matrix. Here, samples from S₀ to S₆ were subjected to cytotoxicity study. The cytotoxicity was determined using VSMC cultured from rat aorta in combination with the magnetic samples (introduced into the wells where the experiment was conducted) by using the colorimetric MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay (Mosmann, 1983). This technique is based on the ability of viable cells to transform the MTT salt into formazan dyes. After a 24-h VSMC incubation period, wells were washed with phosphate-buffered solution, and the MTT solution (1 mg/ml) was added and incubated for 4 h. The purple formazan generated by viable cells was solubilized with 20% sodium dodecyl sulfate in 0.02 M HCl and incubated for 10 h at 37°C. The optical density of each well was determined at 540 nm in a microplate spectrophotometer reader (BioTek^® Synergy HT). Cell viability was expressed as percentage in relation to controls (non-treated cells).

In order to determine the magnetic properties of the M-SREs, a vibrating sample magnetometer (VSM) (ADE model 10 VSM Oxford Instruments, UK) was used for obtaining the magnetization of the slab samples S₁–S₆ at 300 K. Moreover, from this study the nanoparticle size was determined and compared with the results obtained from TEM.

Cantilevers and soft microsystem were fabricated using samples S₁–S₆. These structures were magnetically characterized by placing them on a positioning stage over a permanent magnet (0.45 T nominal value). This magnet was located on a vertical stage (AHT-65 PI MICOS, Eschbach, Germany); thereby, it was possible to change the applied magnetic field (μ₀H) to the microsystem by changing the relative distance. The deflection was captured with a CCD camera (Imaging Pixelfly, 205 X2 1475, Germany) in plane, orthogonal to the microsystem.

The M-SRE Young’s modulus (Y_M-SRE) for different FF concentrations was calculated by experimentally recording the deflection of cantilevers under the action of the gravity by using a confocal microscope (PLμ 2300 Optical Imaging Profiler, Sensofar, Spain) with software (PLμ Confocal Imaging Profiler, Sensofar, Spain).

The spectral response as a function of the FF concentration was studied by coupling a broadband light source (HL-2000, Ocean Optics, USA) to a multimode fiber optic (Thorlabs, Dachau, Germany) with a core diameter of 125 μm. Light emerging from this fiber was orthogonally coupled to each slab sample (with an identical optical path of 250 μm in all cases). A second fiber optics with a core diameter of 230 μm was used to collect light emerging from the M-SRE, which was directly connected to a spectrometer (HR 4000, Ocean Optics, USA). Ten consecutive scans (each of which requiring an integration time of 100 ms) were done for statistics. A non-doped slab sample was used as reference.