Epitaxial growth and magnetic properties of Mn₅Ge₃/Ge and Mn₅Ge₃C_x/Ge heterostructures for spintronic applications

In a conventional MBE, the growth of an epitaxial film can proceed via two main techniques: solid phase epitaxy (SPE) and reactive deposition epitaxy (RDE). The SPE consists of deposition or co-deposition of materials at room temperature, then followed by thermal anneals in order to activate diffusion and/or interdiffusion of species. This method involves therefore two successive processes: first, diffusion and/or interdiffusion occur and then phase nucleation takes place, which starts from the interface. In RDE, materials are deposited or co-deposited on the substrate surface, which is kept at high temperatures. Depending on the substrate temperature, different phases can be formed but the phase nucleation is the main process, which occurs at the growing surface as the growth advances. Since Mn₅Ge₃ is not the most stable phase and has a hexagonal structure similar to that of Ge(111), the SPE technique, which allows Mn₅Ge₃ films to easily adopt the substrate symmetry from the interface, naturally appears more appropriate than the RDE to form epitaxial Mn₅Ge₃ films. Indeed, our unpublished results show that when using the RDE technique, a fraction of the Mn₁₁Ge₈ phase may coexist with Mn₅Ge₃. Starting from a thin Mn film deposited on a Ge(111) substrate at room temperature, we have shown that Mn₅Ge₃ is the unique epitaxial phase that can be formed upon annealing in the temperature range of 430–600 °C [4, 5]. The Mn₁₁Ge₈ phase, which is the richest in Ge, is formed only when annealing at temperatures higher than 650 °C [13]. Below 650 °C, the coexistence of two phases, Mn₅Ge₃ and Mn₁₁Ge₈, may be observed, but only when the initial Mn film is thick enough (> 210 nm) [14] or when films are grown on an amorphous oxide substrate [15].

Figures 1(a) and (b) show typical RHEED patterns observed during epitaxial growth of Mn₅Ge₃ on Ge(111). Stating from a clean (2 × 4) reconstructed Ge(111) surface, RHEED patterns indicate that Mn₅Ge₃ films display a hexagonal symmetry similar to that of the substrate. By combining RHEED patterns with TEM backside electron diffraction analysis, it is found that the hexagonal basal (0001) plane of Mn₅Ge₃ is parallel to the (111) plane of Ge. The epitaxial relationship has following form:

$$\begin{eqnarray*} &&{\rm Mn}_5 {\rm Ge}_3(0001)\parallel {\rm Ge}(111)\\ &&{\rm with} \ [110]_{{\rm Mn}5{\rm Ge}3}\parallel [1 - 10]_{\rm Ge}. \end{eqnarray*}$$

Zoom In Zoom Out Reset image size

The Mn₅Ge₃ surface is characterized by a RHEED $(\sqrt 3 \times \sqrt 3 )R30^ \circ$ reconstruction, defined by the observation of 1 × 1 streaks along the [1–10] azimuth and additional 1/3- and 2/3-ordered streaks along the [11–2] azimuth. Interestingly, long streaks are observed in RHEED patterns for film thicknesses in the range between some nm up to about 160 nm, indicating that the film surface is highly smooth.

A high-resolution TEM image taken near the interface region of a 25 nm thick Mn₅Ge₃ film is shown in figure 2(a). The image clearly reveals atomic lattice planes of both the Ge substrate and Mn₅Ge₃ film. The film is highly ordered and the interface is atomically abrupt, no intermixing is observed. Of particular interest, with a misfit as high as 3.7%, almost no threading dislocations can be detected. XRD measurements of the corresponding sample are displayed in figure 2(b). In addition to the (hhh) reflections of the Ge(111) substrate [(111) and (333) reflections], only (002) and (004) hkl reflections of Mn₅Ge₃ are observed, thus confirming the epitaxial relationship of the film, being Mn₅Ge₃(001) ∥ Ge(111).

Zoom In Zoom Out Reset image size

A systematic study of the magnetic properties of Mn₅Ge₃ has been carried out as a function of the film thickness ranging from 5 up to 160 nm. We show in figure 3 some representative hysteresis (M–H) loops measured with in-plane and out-of-plane magnetic field. Insets represent a zoom around the positive saturation field of the out-of-plane configuration. Parallel and perpendicular configurations for thick samples lead apparently to similar magnetic reversal; this is surprising since we expect an easy magnetization axis along the c-axis, which is perpendicular to the sample plane. However, in-plane M–H curves reveal a steady change in the magnetic behavior: for samples thinner than 10 nm, the hysteresis loop exhibits a square shape that unambiguously indicates that the magnetization easy axis lies in-plane; for thicker samples, the M–H curves become increasingly canted as the saturation field increases with thickness but a hysteresis is still visible around zero field.

Zoom In Zoom Out Reset image size

In the out-of plane configuration, at first sight, the hysteresis loops appear similar throughout the range of the studied thicknesses: little hysteresis is present and the saturation fields are higher than the ones observed in the parallel configuration. However, when looking at the variation of the perpendicular saturation field versus the film thickness, two regimes can clearly be distinguished: firstly, below thicknesses of about 20 nm, the perpendicular saturation field increases rapidly with the film thickness; secondly, above this thickness, it is independent of the film thickness and fluctuates around 10 000 ± 1000 Oe. More importantly, the general shape of the hysteresis curves changes. Above a threshold thickness, a singularity in all the out-of-plane M–H curves appears around the saturation field as demonstrated in the inset of figure 3(c) for a 25 nm thick sample. By describing the M–H curve from positive saturation to lower field values, a characteristic opening of the hysteresis loop appears around the saturation field over a narrow range of fields. As the field decreases, the two M–H branches return to similar field dependence and this singularity disappears. This feature is not present in hysteresis loops of layers thinner than 10 nm where the magnetization rises linearly with the applied field and almost reversibly. In [16], we have provided a detailed analysis of the evolution of Mn₅Ge₃ magnetic properties versus the film thickness and show that the reorientation of the magnetization from in-plane to out-of-plane occurs for a film thickness lying between 10 and 25 nm. This result is strongly supported by theoretical calculations based on an improved version of Kittel's model to retranscribe the magnetic behavior of domains in uniaxial thin films. Of particular interest, the size of magnetic domains in Mn₅Ge₃ is shown to be considerably smaller than the one in any other known magnetic system and it can be in addition tailored by the film thickness.

The enhancement of the magnetic properties of polycrystalline Mn₅Ge₃ films induced by carbon doping was first demonstrated by Gajdzik et al [17] using sputtering deposition technique. Later, Slipukhina et al [18] calculated exchange-coupling constants in Mn₅Ge₃C alloys and showed that the enhanced ferromagnetic stability in the alloy mainly results from interactions between Mn atoms mediated by carbon incorporated in octahedral voids of the hexagonal Mn₅Ge₃ cell. To incorporate as high as possible carbon atoms into interstitial sites of the Mn₅Ge₃ cell, we have implemented the SPE technique in order to promote carbon diffusion. Indeed, since the atomic radius of carbon is almost twice as small as that of Mn and Ge, it follows that in a growth process where carbon atoms can diffuse, it becomes easier for them to be incorporated in the interstitial site.

The effects of the carbon concentration on the magnetic properties of Mn₅Ge₃C_x films are depicted in figure 4. Figure 4(a) displays hysteresis loops of C-free Mn₅Ge₃ and Mn₅Ge₃C_x films with various carbon concentrations, measured by SQUID at 300 K with a magnetic field of 0.5 T applied in the film plane. At 300 K the hysteresis loops of the C-free Mn₅Ge₃ film exhibit a paramagnetic character, as being expected because C-free Mn₅Ge₃ is ferromagnetic only up to 296 K. For C-doped Mn₅Ge₃, hysteresis loops measured at 300 K clearly indicate that the materials are ferromagnetic within the whole range of the carbon concentration. It is worth noting that while the measurements at 300 K confirm that at x = 0.9 the film remains ferromagnetic, the hysteresis loops measured at 5 K (not shown here) display an oblate shape and show a great reduction of net magnetization [19, 20]. The temperature dependence of normalized magnetization of Mn₅Ge₃C_x with various carbon concentrations is presented in figure 4(b).

Zoom In Zoom Out Reset image size

For comparison, we also show a curve of a C-free Mn₅Ge₃ film of the same thickness. The figure clearly indicates that addition of carbon strongly enhances magnetization of Mn₅Ge₃ and such an enhancement continuously increases with the carbon concentration up to 0.7. The T_C, measured at the inflection point of the curve M versus T, reaches a value of ∼430 K for x = 0.6 and 0.7. We note that if we determine the T_C from the extrapolation of the M(T) data to M(T_C) = 0, the T_C of the Mn₅Ge₃C_0.6 curve gives a value up to 460 K.

We report in figure 5(a) the variations of T_C and of magnetization at saturation (M_S) as a function of the carbon concentration. The variation of T_C with x occurs in two distinct regions: T_C first linearly increases with x up to 0.6–0.7, then falls for larger values of x. Note that this behavior is different from that previously reported for polycrystalline films [17]. The evolution of M_S, measured at 5 K, is well correlated with the T_C variation. Saturation magnetization is found to linearly decrease with x and at x = 0.6–0.7, an abrupt change in the slope is observed. We note that in transition metals and their alloys, it is generally observed that magnetization at saturation increases when increasing the Curie temperature. However, as previously mentioned, in Mn₅Ge₃C_x the ferromagnetic enhancement results from Mn_II–Mn_II interactions mediated by carbon atoms inserted into the voids of Mn octahedra of the hexagonal structure. Consequently, carbon incorporation into Mn₅Ge₃ changes the 3d states of neighboring Mn atoms and the hybridization between the C 2p and Mn_II 3d states leads to a decrease of magnetization of saturation as well as magnetic moment on Mn_II.

Zoom In Zoom Out Reset image size

These results indicate that the saturation concentration of carbon, which can be inserted into interstitial sites of the Mn₅Ge₃ lattice is around x ∼0.6–0.7. From measurements of the magnetic moment at saturation and the film thickness deduced from TEM images, an average saturated Mn moment of ∼1.9 μ_B/Mn is deduced for x = 0.7. This value is close to the one obtained in C-implanted films [21] and with theoretical calculations [18], but deviates from the values obtained for the sputtered films, where the highest moment of 1.1 μ_B/Mn was observed [17].

To understand the decrease of T_C for x ⩾ 0.7, we show in figure 5(b) a typical TEM image of a sample corresponding to x = 0.7. It is worth noting that for x ⩽ 0.6, TEM images of carbon-doped Mn₅Ge₃C_x alloys are similar to that of carbon-free Mn₅Ge₃ shown in figure 2(a) in which the film is of high crystalline quality and the interface is atomically smooth. The presence of clusters or precipitates in the grown film can be clearly seen from this image. In order to identify the nature of clusters that are formed for x ⩾ 0.7, thermodynamic calculations of the formation energy of the carbon defects in Mn₅Ge₃ have shown that the Mn₅Ge₃C_0.5 alloy is a stable ternary alloy and additional carbon atoms cannot be inserted into interstitial sites, but will rather form clusters of manganese carbides (MnC) [19]. Two main phases, Mn₇C₃ and/or Mn₅C₂, appear to be energetically favorable when the carbon concentration becomes higher than 0.5.

The thermal stability of an active material is one of the crucial parameters that need to be determined and controlled for its integration in the device fabrication process. We display in figure 6(a) the change of magnetization of a 100 nm thick Mn₅Ge₃ film after annealing at 650 °C. Before annealing, hysteresis loops of Mn₅Ge₃ exhibit ferromagnetic behavior as expected. The magnetization at saturation (M_S) is ∼1300 emu cm^-3 and the average magnetic moment per Mn atom (μ_s) is ∼3.2 μ_B. These values are close to those reported in literature for thin films [3, 4] and bulk materials [22]. After annealing at 650 °C, the magnetization at saturation (M_S) is found to decrease down to 6 emu cm^-3 and the remanent magnetization reduces from 125 to 1 emu cm^-3.

Zoom In Zoom Out Reset image size

To get a better insight into the origin of the above drastic change in magnetic properties of Mn₅Ge₃ upon annealing, we present in figure 6(b) a comparison of magnetization versus temperature before and after annealing. The as-grown sample clearly displays a ferromagnetic behavior characteristic of Mn₅Ge₃; the T_C, measured at the inflection point of the M(T) curve, is ∼296 K. After annealing at 650 °C, the M(T) curve reveals two distinct transitions: an antiferromagnetic/ferromagnetic transition at ∼150 K and a ferromagnetic/paramagnetic transition at ∼270 K. Such a magnetic signature can be unambiguously attributed to the antiferromagnetic Mn₁₁Ge₈ compound [23].

Thus, the above results confirm that Mn₅Ge₃, which is not a stable phase, can be stabilized on Ge(111) by epitaxy due to the similarity of its hexagonal structure compared to that of Ge(111). As defined by thermodynamics, post-thermal annealing of grown films should bring the system toward a more equilibrium state, i.e. to the formation of Mn₁₁Ge₈, which is the most stable phase in the Mn/Ge phase diagram. Since the Ge concentration in Mn₁₁Ge₈ is of ∼42% compared to ∼37.5% in Mn₅Ge₃, such a phase transformation should require a long-range diffusion of Ge from the substrate.

Regarding the thermal stability of C-doped Mn₅Ge₃, as mentioned above, doping Mn₅Ge₃ with carbon allows increasing its T_C, and such an enhancement has been explained due to Mn_II–Mn_II interactions mediated by carbon atoms [18]. Figure 7(a) shows the magnetization enhancement induced by carbon doping (black curve corresponding to Mn₅Ge₃ and green one to Mn₅Ge₃C_0.6). It can be seen that the Mn₅Ge₃C_0.6 curve exhibits a T_C up to ∼460 K, compared to 296 K of Mn₅Ge₃. Figure 7(b) shows the evolution of hysteresis loops of the Mn₅Ge₃C_0.6 film upon annealing at 750 and 850 °C. The most interesting feature is that the carbon-doped Mn₅Ge₃ layers remain ferromagnetic even after annealing at 850 °C. The hysteresis loops conserve its squareness up to 750 °C, beyond which an increase of the coercive field occurs, which can be probably attributed to the formation of point defects due to high annealing temperatures. Thus, the above results provide evidence that inserting carbon into interstitial sites of Mn₅Ge₃ allows great improvement in its thermal stability.

Zoom In Zoom Out Reset image size

Another interesting feature that can be seen in figure 7(a) is a reversible transition of T_C upon carbon doping and annealing. Doping Mn₅Ge₃ with carbon allows enhancing T_C from 296 up to 460 K, which is found to decrease down to 350 and 307 K when increasing the annealing temperature to 750 °C (red curve) and 850 °C (blue curve), respectively. Other magnetic parameters, such as M_sat and μ_s, also exhibit a similar reversible transition. For example, in Mn₅Ge₃, the measured value of μ_s is ∼3.2 μ_B, which is found to decrease down to ∼1.9 μ_B in Mn₅Ge₃C_0.6. Upon annealing, μ_s progressively increases with increasing temperature and almost reaches the initial value of ∼3.2 μ_B after annealing at 850 °C. Such results imply that carbon atoms, which have been incorporated into interstitial sites of Mn₅Ge₃, are progressively extracted during annealing.

In numerous spintronic applications, such as spin valves or giant magnetoresistance (GMR) superlattices, high-quality Ge overgrowth on top of Mn₅Ge₃ films is needed. One of the difficulties inhibiting the realization of Ge/Mn₅Ge₃ heterostructures with abrupt interfaces would be probably the Mn segregation. As we have mentioned above, since epitaxial Mn₅Ge₃ films give rise to additional RHEED streaks compared to the Ge surface, we have used RHEED to monitor, in real time and in situ, the Mn segregation process by measuring the intensity evolution of these additional streaks versus the Ge deposition time or the thickness [10]. A typical result of intensity measurements of a 2/3 streak at three different substrate temperatures, 200, 450 and 550 °C, is reported in figure 8.

Zoom In Zoom Out Reset image size

Two distinct behaviors regarding the scale of the deposition time of the Ge overlayers are clearly observed. At 200 °C, the intensity of the 2/3 streak vanishes at around 95 s, while at 450 and 550 °C it continues to a much higher deposition time and completely disappears only after a Ge deposition of more than 800 s. The corresponding Ge thickness is ∼8 nm at 200 °C and it is larger than 70 nm at 450 and 550 °C. It is worth noting that the surface segregation of an element during the growth of heterostructures or multilayers has been observed in many systems, including III–V materials [24] or Si on SiGe [25]. However, at usual growth temperatures (∼ 600 °C), the segregation length does not, in general, exceed a dozen nanometers.

To understand the different behavior of Ge overgrowth observed at low and high temperatures as described above, we have systematically performed TEM analyses of samples grown at 450 and 550 °C. Figure 9(a) displays a typical cross-sectional TEM image taken after the deposition of 60 nm of Ge at 450 °C. To see the evolution of different layers in the final structure more clearly, we show, in figure 9(b), a schema of the designed sample in which the thickness of each layer is indicated.

Zoom In Zoom Out Reset image size

Contrary to the designed structure, the TEM image reveals that the sample surface is terminated by a Mn₅Ge₃ layer and no trace of Ge overlayers is detectable. Indeed, the observation in this TEM image of well-defined atomic rows, all being perpendicularly aligned to the interface, can be unambiguously attributed the hexagonal (0001) plane of Mn₅Ge₃, which is parallel to the (111) plane of Ge [4, 19]. Thus, TEM analyses indicate that upon Ge deposition at 450 °C, it is not a Ge layer which is progressively formed on the surface as expected but the deposited Ge reacts with Mn to form the Mn₅Ge₃ compound. Interestingly, the Mn₅Ge₃ surface layer has a thickness of ∼20 nm, a value slightly thinner than the initial thickness of the starting Mn₅Ge₃ layer. It can also be seen that the whole Mn₅Ge₃ layer continuously floats on the surface as the Ge deposition progresses, leaving behind newly formed Ge layers. In other words, the Mn₅Ge₃ film behaves here as a surfactant, which floats upwards the growing surface, similar to the case of monolayer-thick Mn adsorbed on Ge(001) [26]. An interesting feature that can be seen from the image is the presence of Mn-rich clusters embedded inside the lower Ge layers. We notice that in standard Mn₅Ge₃/Ge heterostructures, Mn–Ge clusters are never observed in the Ge substrate after Mn₅Ge₃ growth at 450 °C [4, 5, 16] or even after post-annealing up to 650 °C [13]. The formation of such clusters can be attributed to the fact that in the case of the Ge deposition on Mn₅Ge₃ at 450 °C, the initial epitaxial Mn₅Ge₃ film, because of its metastable state, is destabilized even at its interface with the substrate, which is 25 nm from the surface. Consequently, a part of Mn is detached around the interface region and diffuses into the substrate, resulting in the formation of Mn-rich clusters. This explains why the thickness of the final floating Mn₅Ge₃ film is slightly thinner than the initial one.

To prevent out-diffusion of an element, it is common to use a diffusion barrier, and materials must be not only nonreactive but also are able to strongly adhere to adjacent materials. In electronic or memory devices, multilayers of metals, WN₂, RuTiN or RuTiO, are usually used to prevent out-diffusion of dopants (B and P) or oxidation of devices [27–29]. Such materials are, however, difficult to insert in a heterostructure where epitaxial growth is needed. Since any segregation process should involve a rapid and long-range diffusion of elements, which, in general, occurs via interstitial diffusion, our approach consists of filling in the interstitial sites of Mn₅Ge₃ prior to Ge deposition. To experiment with the filling, we choose carbon for its small atomic radius. The principle of experiments is described in figure 10(a), consisting of depositing some carbon MLs on Mn₅Ge₃ prior to Ge growth.

Zoom In Zoom Out Reset image size

The amount of adsorbed carbon should be an important parameter, and it was chosen according to the change in RHEED patterns. Upon carbon adsorption, the $\sqrt 3$ RHEED characteristic of Mn₅Ge₃ remains almost unchanged up to carbon coverage of 4 ML, beyond which a faint pattern is observed. Since we search a high filling degree of Mn₅Ge₃ interstitial sites, a carbon amount of 4 ML is then chosen. It is worth noting that carbon adsorption at room temperature or at 250 °C produces almost similar results. Figure 10(b) shows a typical structure of a sample containing 4 ML of carbon adsorbed at 250 °C.

Even if the Ge overlayers are far from perfect, the image clearly reveals that the Ge/Mn₅Ge₃ interface has become much smoother. A much smaller Mn segregation length is confirmed by RHEED analyses, which reveal that a c(2 × 4) reconstruction characteristic of a clean Ge(111) surface quickly appears only after deposition of some Ge MLs. The improvement of Mn out-diffusion is also confirmed by Auger measurements, which reveal that Mn transitions, located at 537, 584 and 631 eV, almost disappear after 3 nm thick Ge deposition while on C-free samples Mn signals persist for Ge thickness larger than 10 nm. Shown in the inset is an atomically resolved image of the interface region. Clearly, no Mn-rich clusters are present, and more importantly, well-ordered (111) planes of Ge overlayers are found to be perpendicular to the atomic rows produced by Mn arrangement along [0001] direction of the underneath Mn₅Ge₃. However, it is worth noting that when carbon adsorption on the Mn₅Ge₃ surface is carried out at temperatures ⩾ 450 °C, manganese carbides can be formed and the resulting Ge overlayer changes its orientation from (001) to (111), which has a higher surface energy [30].