Flexible ε-Fe2O3-Terephthalate Thin-Film Magnets through ALD/MLD

Pliable and lightweight thin-film magnets performing at room temperature are indispensable ingredients of the next-generation flexible electronics. However, conventional inorganic magnets based on f-block metals are rigid and heavy, whereas the emerging organic/molecular magnets are inferior regarding their magnetic characteristics. Here we fuse the best features of the two worlds, by tailoring ε-Fe2O3-terephthalate superlattice thin films with inbuilt flexibility due to the thin organic layers intimately embedded within the ferrimagnetic ε-Fe2O3 matrix; these films are also sustainable as they do not contain rare heavy metals. The films are grown with sub-nanometer-scale accuracy from gaseous precursors using the atomic/molecular layer deposition (ALD/MLD) technique. Tensile tests confirm the expected increased flexibility with increasing organic content reaching a 3-fold decrease in critical bending radius (2.4 ± 0.3 mm) as compared to ε-Fe2O3 thin film (7.7 ± 0.3 mm). Most remarkably, these hybrid ε-Fe2O3-terephthalate films do not compromise the exceptional intrinsic magnetic characteristics of the ε-Fe2O3 phase, in particular the ultrahigh coercive force (∼2 kOe) even at room temperature.


INTRODUCTION
Research on flexible magnets is inspired by the strong drive to make consumer electronics thin, lightweight, and wearable; such next-generation flexible electronics should be shapeable into any arbitrary configuration depending on the intended use. 1−3 Progress in the flexible electronics has already opened the door to plethora of advanced applications such as wearable solar cells, 4 flexible transparent electrodes, 2 biocompatible electronic devices, 1,4 stretchable energy harvesters, 1,4 full color displays, 3 and flexible optoelectronic devices. 1 Because magnets are inevitable components of electronics, development of new types of thin-film magnets with inbuilt flexibility is an urgent challenge.
The pioneering works by Miller et al. 5,6 opened up research on organic/molecular magnets, 7−12 forming the bases for the currently available lightweight and flexible magnets. The organic components in these magnets provide other benefits as well, such as low-temperature processing, critical-elementfree composition, and transparency. 7,11−13 For the fabrication of flexible magnetic thin films in particular, two main strategies have been envisioned: (i) nanocomposites composing of conventional inorganic magnetic materials and a polymer substrate 14 or polymeric fillers, 15 and (ii) organic/molecular materials 10,11,16 grown using solution-based 7 or gasphase 10,11,16 deposition techniques. The multistep and often harsh solution-based reaction pathways used in the first approach are not optimal for the fabrication of conformal, homogeneous, and solvent-free magnetic thin films required in practical applications. The second approach, on the other hand, is more likely to yield high-quality homogeneous thin films, but the organic/molecular magnets based on s-or porbital spins typically suffer from weak magnetization/low coercivity field, 7,10,11,16 low magnetic transition temperature, 8,9,17 structural disorder, 18 and/or instability. 7,11,19 In applications such as magnetic storage devices, hard magnets would confer to the better stability of stored data; for this, coercive field values higher than 100 Oe are desirable, 8,12 which has not been achieved with the current organic/ molecular magnets.
Here, we present a novel approach to the flexible roomtemperature magnets; we fabricate inorganic−organic superlattice (SL) thin-film structures using the currently strongly emerging atomic/molecular layer deposition (ALD/MLD) technique, 20−25 which combines the leading ALD (atomic layer deposition) 26−28 technology of advanced inorganic thin films and its less exploited MLD (molecular layer deposition) 29,30 counterpart for purely organic films. Our choice for the inorganic component is ε-Fe 2 O 3 . This uncommon Fe(III)oxide polymorph possesses the most intriguing magnetic properties, i.e., ferrimagnetism with a Curie temperature as high as ca. 500 K and a remarkably large coercive field (even up to 20 kOe at room temperature), 31,32 and on top of that, strong magnetoelectric coupling. 33 Moreover, like iron oxides in general, it is nontoxic and biocompatible and consists of Earth-abundant elements only.
The issue with the ε-Fe 2 O 3 phase lies in its narrow stability window; it is nearly nonexistent in nature and challenging to artificially synthesize except in certain nanoscale samples. 32 The basis for the present work is in our recent success in developing a facile ALD process for high-quality ε-Fe 2 O 3 thin films, which are free from the other Fe 2 O 3 polymorphs, α-Fe 2 O 3 (hematite), β-Fe 2 O 3 , and γ-Fe 2 O 3 , and the magnetite Fe 3 O 4 . 34 These ε-Fe 2 O 3 films grown from FeCl 3 and H 2 O precursors in the temperature range 280−300°C are perfectly stable in ambient air (even at elevated temperatures) and against insertion of organic layers through MLD cycles. 35 Here we will demonstrate for the first time the great potential of our ALD/MLD-grown ε-Fe 2 O 3 -organic superlattices as flexible thin-film magnets. The regularly inserted organic layers enhance the mechanical properties of the films without compromising their unique magnetic properties. It should be emphasized that similarly to the parent ALD technology, the combined ALD/MLD method yields high-quality ultrathin films with atomic-level thickness control, large-area homogeneity, and conformality. These superior features derive from the way of introducing the gaseous/evaporated precursors one after another into the reactor in sequential pulses to achieve the desired surface reactions. The well-controlled surface reactions moreover make the ALD/MLD method uniquely suited to the engineering of inorganic−organic SL structures with the required atomic/molecular level accuracy for the individual layer thicknesses. 36−40

EXPERIMENTAL SECTION
All the thin-film depositions were carried out in a commercial flowtype hot-wall ALD reactor (F-120 by ASM Ltd.) using iron chloride (FeCl 3 , Merck, 95%) deionized water and terephthalic acid (TPA; Tokyo Chemical Industry CO., Ltd., > 99.0%) as precursors. 35 The two solid precursors, FeCl 3 and TPA, were placed inside the reactor in open boats and heated at 158 and 180°C, respectively, whereas the deionized water cylinder was placed outside the reactor. Nitrogen (N 2 , 99.999%) was used both as the carrier gas and the purge gas between the precursor pulses; the N 2 flow rate was kept at 300 SCCM and the reactor pressure at 3−5 mbar. The depositions were carried out at 280°C on silicon (100) (Okmetic Oy) substrates cut into 2.0 × 2.0 cm 2 pieces, washed with ethanol−water mixture and acetone, and dried prior to film deposition. The films for mechanical property studies were deposited on 50 μm thick polyimide substrates (Kapton 200HN) of 4.5 × 4.5 cm 2 with a total of five precut stripes. The polyimide substrates were washed with isopropyl alcohol and distilled water and dried before taking them for deposition. These substrates were also subjected to a 1 h wait time at 280°C prior to deposition to outgas the residual water from the polyimide.
Each superlattice (SL) deposition consisted of ALD (FeCl 3 +H 2 O) cycles for ε-Fe 2 O 3 layers and MLD (FeCl 3 +TPA) cycles for the molecular organic layers; the optimized precursor/purge pulse lengths were adopted from our previous work, i.e., 2 s FeCl 3 /4 s N 2 /1 s H 2 O/ 3 s N 2 for the ALD cycles and 4 s FeCl 3 /8 s N 2 /25 s TPA/50 s N 2 for the MLD cycles. 34,35 The pulsing sequence followed the pattern: [(FeCl 3 + H 2 O) m + (FeCl 3 + TPA) k ] n + (FeCl 3 + H 2 O) m . Here, m controls the thickness of individual ε-Fe 2 O 3 layers in the superlattice and (nk) controls the total number of organic layers within the ε-Fe 2 O 3 matrix. The total number of ALD or MLD cycles (controlling the total film thickness) is thus expressed as [n (m + k) + m]. We deposited three series of samples with different total film thicknesses, such as films with thickness <160 nm for verifying the intended SL patterns, thickness ca. 250 nm for studying the influence of the organic layers on the overall magnetic properties, and finally films with thickness >250 nm for studying the mechanical flexibility. The k value was 1 (monomolecular layer) for all the SL structures in the first two series, but for last series a film with k = 10 was additionally fabricated.
For the verification of the SL structures and the film thickness determination, X-ray reflectivity (XRR; PANalytical X'Pert PRO Alfa 1; X'Pert Reflectivity software) measurements were carried out. The targeted ε-Fe 2 O 3 crystal structure was confirmed by X-ray diffraction (XRD; PANalytical X'Pert PRO MPD Alfa 1; Cu Kα 1 radiation) measurements. The surface morphology of the sample SL films was analyzed using a scanning electron microscope (SEM, Hitachi S-4700). The sample specimen for SEM measurement was mounted on a carbon tape and analyzed at a voltage of 10 kV and a current of 15 μA. The presence of terephthalate moieties in the SL films was confirmed using Fourier transform infrared (FTIR, Bruker alpha II) and Raman (Witec Raman with a 532 nm excitation wavelength) spectroscopy analysis. In order to compensate the interference from the substrate, we subtracted the FTIR spectrum of the bare silicon substrate from the spectra of the samples.
Magnetic properties were studied using a vibrating sample magnetometer (VSM; Quantum Design PPMS). For the measurements, 3 × 4 mm 2 sample was glued with GE varnish on a quartz sample holder and set parallel to the applied magnetic field. Magnetization versus magnetic field (M−H) isotherms were collected by sweeping the magnetic field from −50 to 50 kOe. Magnetization versus temperature (M−T) curves were measured both under fieldcooled (FC) and zero-field-cooled (ZFC) conditions.
Uniaxial tensile testing of the films coated on the polyimide substrates was carried out using a tensile stage (MTI 8000−0010) equipped with a digital optical microscope (Keyence 500F) for in situ monitoring of the fragmentation process. The tensile stage was operated at a constant strain rate of 1.4 × 10 −4 s −1 by means of displacement control. Microscope images were recorded at strain intervals of 1.4 × 10 −4 for the subsequent analysis. The strain values were determined via digital image correlation by tracking the distance between pairs of points (polyimide features) on the sample surface. The gauge sections of the samples were 5 × 17 mm 2 .
The thicknesses of the films on the polyimide substrates were measured from cross-section SEM (Hitachi S4800; 1 kV) images of focused-ion-beam cuts (Lyra FIB-SEM; cutting 10 nA and polishing 1 nA and 60 pA, voltage was 30 kV. The mechanical data in the text represents values normalized to 450 nm thickness via the known h f −1/2 dependence for the COS and the saturation crack density. 41

RESULTS AND DISCUSSION
We fabricated a series of ε-Fe 2 O 3 -organic thin films using terephthalic acid (TPA; benzene-1,4-dicarboxylic acid) as the organic precursor, according to the following overall deposition process: where m stands for the number of (FeCl 3 + H 2 O) ALD cycles applied for each individual ε-Fe 2 O 3 layer, k (= 1 in most of the experiments) stands for the number of (FeCl 3 + TPA) MLD cycles applied for each individual organic layer, and n defines the number of the organic/hybrid layer blocks in the SL structure in total; the number of the ε-Fe 2 O 3 layer blocks is accordingly n + 1. In our ε-Fe 2 O 3 -TP (TP = terephthalate) SL samples n varies from 3 to 75 and m from 31 to 1000, see Table 1.
3.1. Structural and Chemical Characteristics. All the depositions yielded visually high-quality homogeneous thin films. The expected SL structures were affirmed from the XRR (X-ray reflectivity) patterns, shown in Figure 1A for representative ε-Fe 2 O 3 -TP samples, and for a parent ε-Fe 2 O 3 film for comparison. In Figure 1A, intense regularly appearing SL peaks are clearly seen for all the films containing organic layers (n > 0) but not for the parent (n = 0) ε-Fe 2 O 3 film. Moreover, between the SL peaks, smaller oscillations can be seen, the number of which corresponds to the expected value of n (as far as can be counted before the oscillations start to overlap with each other), thus confirming the excellent controllability of our ALD/MLD process. In Table 1, we also give the film thickness values determined from the XRR data. Because the XRR technique works properly for relatively thin films only, we were not able to directly determine the thickness values for the films thicker than ca. 160 nm. For these thicker films, an approximation was calculated on the basis of the growth-per-cycle (GPC) values calculated for the thinner films (from the experimental thickness value and the number of deposition cycles applied). From XRD (X-ray diffraction) analysis, our ε-Fe 2 O 3 -TP SL films are all crystalline ( Figure 1B); XRD patterns show the same diffraction peaks (002, 013, 122, 004, 015, 204, 006, and 205) but the intensities of these peaks slowly decrease with increasing n, i.e. with a reduction of individual ε-Fe 2 O 3 layer thickness and an increase of organic layers. 34,35 The presence of the expected terephthalate moiety in the films was verified by both FTIR (Fourier transform infrared; Figure 1C) and Raman ( Figure 1D) spectroscopy analyses. In the FTIR spectra, the well-known carboxylate-group fingerprint, i.e., symmetric (ν s ) and asymmetric (ν as ) stretching bands around 1400 and 1510 cm −1 , respectively, is clearly seen for all the SL samples, with the intensity of these bands increasing with n. From the splitting between the two bands, i.e., (1510−1400) cm −1 = 110 cm −1 , it can be concluded that the TP moieties are bound to the iron atoms in a bidentate binding mode. 42,43 Also seen from the FTIR spectra is that the The film thickness values were determined by XRR for the films thinner than 160 nm; for the thicker films, the thickness [given in brackets] is an estimate based on the number of ALD/MLD cycles applied. Raman analysis complemented our understanding of the bonding scheme in our ε-Fe 2 O 3 -TP superlattices. The peaks seen in the Raman spectra ( Figure 1D) at ca. 1420 and 1606 cm −1 are due to the symmetric and asymmetric stretchings of the carboxylate group. 44 The other bands due to the terephthalate group at ca. 290, 860, and 1140 cm −1 can be assigned to the out-of-plane ring bending, the C−C stretching of the carboxylate group, and the superposition of ring breathing and in-plane bending of C−H. 44,45 The band due to in-plane bending of aromatic C−H appears at 1305 cm −1 . 45 The spectrum for our parent (n = 0) sample shows all the features expected for ε-Fe 2 O 3 . 46 With an increase in the number of organic layers, the peaks due to ε-Fe 2 O 3 decrease in intensity, whereas those from the organic moiety increase with n. We also like to mention that in our previous work, we confirmed with XPS that the films grown with the FeCl 3 + TPA process did not contain detectable amounts of elements other than iron, carbon, and oxygen. 43 This rules out the possibility of chlorine contamination in our ε-Fe 2 O 3 -TP SL films.
We also investigated the surface morphology and grain size for our ε-Fe 2 O 3 -TP SL films in comparison to the parent ε-   (Figure 2). In the parent ε-Fe 2 O 3 thin film the grains are uniform and wellshaped ( Figure 2A). The effect of organic layers on the size and distribution of grains is illustrated in Figure S1. With the increasing number of organic layers (and decreasing thickness of individual ε-Fe 2 O 3 layers), the grains first start to appear polydispersed and aggregated. With a further increase in n beyond 30, the polydispersity is again reduced, and the grains become more uniform but with a different morphology and size compared to the case in the ε-Fe 2 O 3 film. In the SL films with high organic contents, the grains are fewer and more segregated with nanogaps in between ( Figure S2).

Magnetic Property Characteristics.
Magnetic properties of both the parent ε-Fe 2 O 3 and the ε-Fe 2 O 3 -TP SL films were investigated using a vibrating sample magnetometer (VSM). The film surface was set parallel to the applied magnetic field (H) during the magnetization (M) measurement, and isothermal M−H curves were measured from −50 to 50 kOe at various temperatures from 10 to 300 K, see Figure 3. All the M−H curves follow a hysteresis loop typical for ferrimagnetic materials; no perfect saturation is seen up to the magnetic fields measured, though. The magnetic characteristics for the parent ε-Fe 2 O 3 film (e.g., coercivity ca. 2.1 kOe) are similar to those reported earlier. 34,47 For the superlattice ε-Fe 2 O 3 -TP films, with increasing number of organic layers (and decreasing ε-Fe 2 O 3 -layer thickness) the ferrimagnetic behavior is preserved up to n = 75 where the individual ε-Fe 2 O 3 layers are as thin as 2 nm, even though the absolute magnetization naturally decreases when the content of nonmagnetic organic layers increases. From Figure 3, both the absolute magnetization and the coercivity field are essentially identical for the two samples, n = 0 (248 nm) and n = 3 (62 nm), and even up to n = 20 (individual ε-Fe 2 O 3 -layer thickness 11 nm) the coercivity field remains essentially the same at lower temperatures. However, for higher n values an abrupt change in both magnetization and coercivity at room temperature was observed ( Figure S3).
Indeed, our ε-Fe 2 O 3 -TP SL thin films retain their roomtemperature hard-magnet characteristics (coercive field higher than 100 Oe) very well upon the addition of organic layers, see also Figure 4 where we plot the coercivity values at different temperatures. For example, the n = 20 (11 nm) sample shows a coercivity field of 260 Oe at 300 K, still clearly above the critical limit. Even for the n = 75 (2 nm) sample with extremely thin ε-Fe 2 O 3 layers the ferrimagnetism still exists below ca. 50 K, with significantly lowered coercivity values though.
Finally, we discuss an interesting detail concerning an asymmetry seen in the M-H loops. Namely, the field-cooled (FC) magnetization curves for ε-Fe 2 O 3 -TP SL structures are shifted toward the negative side ( Figure 5A), whereas for ε-Fe 2 O 3 , the loops measured under FC and ZFC (zero-fieldcooled) conditions are identical and symmetric ( Figure 5B). The difference between the positive and negative coercive field values were found to be high for all our ε-Fe 2 O 3 -TP SL samples at 10 K when the magnetization curves were measured under FC condition. For example, in case of the n = 20 (11 nm) sample, the negative and positive coercive field values with FC were −10 and 5.2 kOe, respectively; for the same sample under the ZFC conditions, the coercive field values were −5.8 and 5.7 kOe. The observed asymmetry in the FC M−H curves could be attributed to an existence of exchange bias, which usually arises from the coupling of ferri/ ferromagnetic and antiferromagnetic layers that can cause a unidirectional anisotropy in the ferri/ferromagnetic layer. 48 We tentatively assume that the presence of a paramagnetic spacer ( Figure S4) between ferrimagnetic ε-Fe 2 O 3 layers might induce an indirect antiferromagnetic coupling of alternating layers, where the conjugate π electrons of the terephthalate moieties play an important role. 49,50 The interlayer exchange interaction arising from the spin polarization through the bridging terephthalate moieties can also favor antiferromagnetic ordering. 50 3.3. Mechanical Property Characteristics. In order to corroborate the positive influence of the organic layers on the mechanical flexibility of our ferrimagnetic ε-Fe 2 O 3 -TP SL thin films, we deposited few representative SL samples and reference samples in parallel on both polyimide (for mechanical tests) and silicon (for basic characterization). For reference, we deposited both ε-Fe 2 O 3 (n = 0) and pure ironterephthalate (Fe-TP) films; for the latter film, the deposition processes consisted only of (FeCl 3 +TPA) k cycles and they thus completely lacked the ε-Fe 2 O 3 layers (m = 0). 43 For the SL samples, we deposited two samples; first a film with n = 20 and k = 1. Then with another sample, we wanted to test the effect of making each organic layer thicker, in other words, we set k = 10 instead of 1 in the deposition process, [(FeCl 3 + H 2 O) m + (FeCl 3 + TPA) k ] n + (FeCl 3 + H 2 O) m . Note that, in both our k = 1 and 10 SL samples, the individual ε-Fe 2 O 3 layer thickness expected is 15 nm; we named these samples as, n = 20-1 (15 nm) and n = 20-10 (15 nm), respectively.
To confirm the expected chemical and structural state of these samples, we characterized the ε-Fe 2 O 3 and the two SL films by FTIR, XRD ( Figure S5) and cross-sectional SEM ( Figure 6). Compared to the k = 1 case, the k = 10 SL sample exhibits stronger TP IR bands but less intense XRD peaks, as expected. Cross-sectional SEM images in Figure 6 and Figure  S6 confirm the intended layer structures for the SL films. The thickness calculated from the SEM data was 423, 454, 663, and 263 nm for the ε-Fe 2 O 3 , n = 20-1 (15 nm), n = 20-10 (15 nm) and Fe-TP films, respectively. The pore type pattern observed only in case of n = 20-10 (15 nm) suggests that the pores most likely are located in the organic layers; similar pattern was observed also for a film grown on silicon substrate ( Figure S7). The surface morphology changes such that the homogeneous grains of the size of ca. 130 nm in ε-Fe 2 O 3 start to aggregate but not grow for n = 20-1 (15 nm). For n = 20-10 (15 nm), the grains are considerably bigger (ca. 800 nm) and more isolated, having a carbon-coated appearance ( Figure S8). 51 The M−T curves measured from 100 to 400 K for the three samples deposited on polyimide substrates are displayed in Figure 7. Magnetization decreases with increasing temperature in a way typical for a ferrimagnet; the T C is apparently higher than the upper limit of our measurement, i.e., 400 K, for all the samples, in accordance with the clear hysteresis loops seen for these samples at 400 K. The absolute magnetization naturally decreases with increasing portion of organic layers but−most importantly−the coercivity field remains essentially the same at lower temperatures, i.e., ca.5 kOe at 10 K for all the samples.   The mechanical properties of ε-Fe 2 O 3 , the two types of ε-Fe 2 O 3 -TP SLs, and Fe-TP (all grown on stretchable polyimide substrates) were addressed through tensile testing; 52−54 these measurements yield the crack onset strain (COS) and the closely related critical bending radius (for a given film/ substrate bilayer) as the metrics for the stretchability and flexibility, respectively. The measurements were uniaxial tensile experiments coupled with in situ optical microscopy. Channel cracks were observed to form perpendicular to the straining axis above the crack onset strain for all the samples investigated. The visual appearance of the film surfaces is illustrated in Figure 8 for the ε-Fe 2 O 3 and Fe-TP references for various tensile-strain values corresponding to various crackdensity values.
From Figure 8 and Table 2, above the crack onset strain, the number density of cracks (along the straining axis) increases rapidly with increasing strain up to a saturation value for ε-Fe 2 O 3 , n = 20-1 (15 nm), and n = 20-10 (15 nm), while for Fe-TP the complete saturation is not reached in the studied strain range of 0−10%. The saturation crack density is directly proportional to adhesive strength, and therefore provides us with an indication of the adhesion of the films to the substrate. The first three films exhibit saturation crack density values of the same order, which indicates that the adhesion of the two SL films is governed by the ε-Fe 2 O 3 layer at the film−substrate interface. 41 In contrast, the crack density (approaching saturation) for the Fe-TP film is order-of-magnitude higher, reflecting its better adhesion to the polyimide substrate. 41 Therefore, Fe-TP could potentially serve as an interface layer to enhance adhesion of ε-Fe 2 O 3 (and the SLs) to the polyimide substrate. Most importantly, with increasing portion of organic layers the COS value of the films increases progressively from 0.33% for ε-Fe 2 O 3 to 1.07% for n = 20-10 (15 nm) ( Table 2), where the decrease in crystallinity seen for the SLs could moreover contributes to the enhanced mechanical performance. 55 It should be emphasized that the relative increase is as high as 220%.
From the COS values, we calculated the critical bending radii as R c = (h s + h f )/(2COS), where h s and h f are the thicknesses of the substrate and the film, respectively. The R c values for our ε-Fe 2 O 3 -TP SL films indicate considerably enhanced flexibility compared to the parent ε-Fe 2 O 3 film, being increased by 32% for n = 20-1 (15 nm) and by 69% for n = 20-10 (15 nm) ( Table 2).
The COS value of 0.33% for our ca. 400 nm ε-Fe 2 O 3 film extrapolates to a value of 0.68% for a 100 nm film thickness (COS ∝ h f −1/2 ). 41 This is comparable to the COS value of ca. 0.5% reported for 100 nm ALD-Al 2 O 3 films on polyimide substrates; 52 hence, ε-Fe 2 O 3 exhibits behavior typical for brittle metal oxide materials. From the other end, the COS value for our Fe-TP film (0.99%) is slightly lower than that reported for ALD/MLD-grown Al-ethylene glycol films of similar thickness (1.8%), 54 where the difference could be due to differences in crystallinity and/or the rigidity of the organic component. For Al 2 O 3 /Al-ethylene glycol superlattices/nanolaminates with large organic concentrations (ALD:MLD cycle ratios of 1:1 and 3:1), Jen et al. 52 reported critical strain values of around 0.9%. 52 Our result for the n = 20-10 (15 nm) SL film is on the same order but with a lower concentration of the organic layers.

CONCLUSION
We have demonstrated the potential of the ALD/MLD technique in the fabrication of new types of flexible inorganic−organic thin-film magnets. This technique allows for the introduction of monomolecular organic layers or thicker metal−organic layer blocks between nanoscale inorganic layers in any predesigned frequency into advanced superlattice structures. Our reproducible ALD/MLD process yielded high-quality, visually homogeneous thin films with appreciable stability under ambient conditions.
In this work, our inorganic component was the ferrimagnetic ε-Fe 2 O 3 phase, with exceptionally high coercive field even at room temperature. This was possible, as we had recently developed a facile ALD process for fabricating high-quality and stable thin films of this rare but attractive iron oxide phase. Now we have shown in this work that by introducing thin, Figure 8. Top-view optical micrographs of the surfaces of ε-Fe 2 O 3 and the Fe-TP films for various values of the uniaxial (horizontal) tensile strain (ε) and crack density. For ε-Fe 2 O 3 , 0.34% represents the crack onset strain, and at 1.41%, the crack density is already at saturation. For Fe-TP, 0.96% represents the crack onset strain, and at 5.71%, the crack density slowly increases approaching saturation. organic-rich layers, either in the form of separate monomolecular terephthalate layers or relatively thin iron-terephthalate layer blocks, it is possible to considerably improve the flexibility of the otherwise relatively rigid ε-Fe 2 O 3 thin films.
Most importantly, the enhancement in mechanical properties was achieved with reasonably low organic-to-inorganic ratios such that the functionality of the inorganic layers, here the "hard" high-coercive-field room-temperature ferrimagnetism of ε-Fe 2 O 3 , was not compromised. In other words, we were able to bring together the functional properties of the inorganic layers and the mechanical flexibility of the organic layers in one single superlattice thin-film material. Both the magnetic and mechanical properties of our novel flexible magnets were unambiguously presented through an extensive investigation of coercivity, magnetization, and tensile properties including critical bending radius, crack onset strain and saturation crack density. We are convinced that our novel mechanically flexible room-temperature magnetic thin films have high potential in next-generation applications where hard magnets in the form of flexible, lightweight, metal-sparing, and nonpoisonous thin films/coatings, possibly applied on challenging surface architectures, are desired. Moreover  The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.