Figure 1 shows the XRD patterns of $${\left[GaSb/Mn\right]}_{3}$$ architectures varying the substrate temperature (Ts). Through the Rietveld refinement, it was possible to observe the formation of Mn-α and GaSb phases according to the synthesis method; however, it was also possible to identify binary phases of Mn2Sb2 and Mn3Ga, which can form at the interfaces of the Mn and GaSb layers due to diffusion between them. This diffusion can be favored by the high mobility of the species involved, especially of Ga that has a melting temperature close to room temperature7. Therefore, in GaMnSb thin films, when the substrate temperature is increased to 423 K, the high mobility of Ga toward the substrate can occur7 and the formation of binary phases, such as the Mn3Ga phase, can be favored.

On the other hand, the FTIR spectra show interference processes evidenced by the appearance of maxima and minima modulated by a sinusoidal function associated with the distribution of the layers (Fig. 2). In the case of multilayer with $$n=3$$, the interference conditions allow setting the sinusoidal modulation for a higher value of wave numbers, thus presenting only two interference fringes (Fig. 2a). When the number of layer increase (n = 6 and 12), interference fringes also increase (Fig. 2b,c) due to the periodicity and the thickness of the layers. This effect is not modified by the formation of binary phases in the interfaces due to diffusion processes, which shows the homogeneity of multilayers.

When the substrate temperature is increased to 423 K, it is possible to observe that the difference between maximum and minimum is reduced and there is a shift in the inflection points of the spectrum (Fig. 2a). This may be associated with the characteristics of the interfaces due to the increase in binary phases (Fig. 1) and the diffusion processes between the elements of the layers. Although the IR vibration modes related to the binary phases are not found, the FTIR spectra could reveal their effects on interference conditions due to the interfaces features.

Figure 3 shows HR-SEM micrographs of the $${\left[GaSb/Mn\right]}_{3}$$ multilayer. Figure 3a is a HR-SEM Secondary Electrons (SE) image whilst, Fig. 3b is Backscattering Electrons (BSE) image. The cross-section of the samples is characterized by the formation of a multilayer architecture with columnar structure (Fig. 3a) and grain size of 180.08 ± 3.24 nm on surface. In the BSE micrograph, the interface between the layers is identified (Fig. 3b), where the Mn layer corresponds to the dark gray region, and the GaSb layer to the light gray region. The morphology on the surface is governed by the high grains density, which is a prominent characteristic of the control with the deposition technique. A $${\left[GaSb/Mn\right]}_{3}$$ multilayer was obtained with a thickness of 327. 27 ± 9.01 nm and GaSb and Mn layers had an average size of 56.53 ± 3.02 nm and 50.31 ± 2.13 nm, respectively.

The diffusion of the layers determined through the binary phases identified by XRD (Fig. 1) is not evident in the HR-SEM micrographs. This may be associated with diffusion effects that occur between the layers, without generating interlayers formation due to the low substrate temperature. Nevertheless, 300 K is enough temperature for the mobility of the species during the deposition process, thus favoring the formation of a columnar microstructure (Fig. 3a). This microstructure is the result of low nucleation of adatoms, similar to the first stages of growth on the substrate27. Columnar structures emerge when the mobility of the deposited atoms is limited, therefore, their appearance occurs throughout the bulk of the material. However, it is evident that existence of the columnar structure is favored when deposition is performed at sufficiently low temperatures.

Among others, magnetic, electrical, and surface properties of thin films are affected, sometimes strongly, by the presence of columnar structures. It was reported that the magnetic anisotropy of apparently isotropic amorphous Gd–Co films may be due to its columnar structure28.

By increasing the number of periods in the structure from 3 to 12, it can be observed that the columnar microstructure is maintained (Fig. 4), and the layers formation is clearly defined. In this case, the thickness of the sample was 1.42 ± 0.04 µm and the grains size was 188.23 ± 5.64 nm on surface. The increase of the grain size was associated with the total deposition time (greater than that of the multilayer with n = 3), which generated increased mobility of species on the surface and greater nucleation27. As shown in Fig. 3, the BSE image (Fig. 4b) evidences the homogeneity of layers, where the Mn layer is represented by a dark gray region and the GaSb layer is the light gray region.

Hysteresis curves (Fig. 5) show the magnetization (M) as a function of the applied magnetic field (H) of the $${\left[GaSb/Mn\right]}_{3}$$ multilayer varying the substrate temperature at 5 K and 300 K; and the magnetization behavior when the temperature is reduced in two cases, in the presence of a magnetic field (FC) and without it (ZFC). In both cases, the hysteresis curves present a low coercive field (Hc) without saturation. This effect was attributed to the competition between the phases identified in XRD patterns. The magnetic characteristics of all phases have been taken into account. In the case of Mn-alpha its magnetic properties correspond to antiferromagnetic material29, whilst, it known that GaSb is a diamagnetic compound30, and the ferromagnetic state of Mn2Sb2 and Mn3Ga phases has been reported31,32,33,34.

Then, the magnetization of the sample contains the contribution of magnetic characteristics of each total magnetic moment of the phases, due to the formation of the ferromagnetic phases at interfaces between GaSb semiconductor layer and Mn layer, it can be possible to obtain an environment that allows the magnetic anisotropy, and the interaction between ferromagnetic state of Mn2Sb2 and Mn3Ga phases produced by tiny crystals, and the antiferromagnetic moments configuration of the Mn-alpha. The opening of the hysteresis curve could be similar to ferromagnetic behavior (ferromagnetic-like behavior), but the ZFC–FC measurements presented in Fig. 5c revealed that the samples does not a critical temperature. The antiferromagnetic behavior of Mn-alpha was predominant when the temperature was reduced. In this way, the magnetization was decreasing as consequence of the phase transition of the paramagnetic state of Mn-alpha to its antiferromagnetic state.

Additionally, hysteresis curves are non-centrosymmetric when the substrate temperature was varied. This behavior is observed when the magnetic anisotropy between layers is greater than the interfacial exchange coupling, called exchange bias coupling35,36. This exchange bias effect can be associated with coupling between antiferromagnetic Mn and the ferromagnetic-like behavior produced through the tiny Mn2Sb2 and Mn3Ga crystals on the interface formation when the temperature was low (Fig. 5a)35,36. In this case, the sample with a Ts = 300 K has a positive exchange bias and its value corresponds to ~ 22.88 Oe at 5 K, while, this was ~ 3,30 Oe when the temperature of the measure was 300 K. In addition, M-H measurements were realized applying external magnetic field values between − 10,000 Oe to 10,000 Oe, and the samples presented a paramagnetic behavior when the H values were major than 1000 Oe, hence, these not have a saturation magnetization.

Also, the exchange bias at 5 K and its value at 300 K present a percentage difference of ~ 85, 54% for the sample $${\left[GaSb/Mn\right]}_{3}$$. This can be related to the antiferromagnetic-paramagnetic transition phase in the Mn layers when the temperature is around 100 K. In comparison, the samples with n = 3 and n = 12 present a different exchange bias when the temperature of the measure was 5 K. In the case of the M-H curve for sample $${\left[GaSb/Mn\right]}_{12}$$ shows an exchange bias value ~ 3 Oe, hence, the exchange bias was reduced when the period number $$n$$ was increased.

Figure 6 presents the current–voltage (I–V) characteristics of $${\left[GaSb/Mn\right]}_{n}$$ multilayers. The I–V curves can be arranged into two distinct regions as indicated by the range of the applied voltage. The first region, positive voltage region, is characterized by the change of high resistive state (HRS) to low resistive state (LRS) (right insets in Fig. 6). The second one, negative voltage region, is characterized by the change of LRS to HRS (left insets in Fig. 6). In consequence, the behavior of the $${\left[GaSb/Mn\right]}_{n}$$ multilayers corresponds to bipolar switching16. These curves were made for 5 cycles, which shows the stability of the resistive change present in the multilayers and the potential of this architecture for applications in non-volatile memories based on the resistive change.

The resistive switching behavior in NVM has been studied through two possible conductive mechanisms: the formation of the conductive filaments16 or the SCLC mechanism12. The conductive filament model has reported to consider the ion diffusion of oxygen or metal among contacts due to the redox process, thus allowing the formation of the filament16,37. However, this model has been recently modified to consider the redistribution of charge carriers as responsible for the formation of the conductive filament, due to the thermo-chemical stability of the crystalline structure of semiconductors (Gibbs energy)16. The SCLC mechanism is observed when the contact at the junction is ohmic19, which allowed that carriers can readily enter the interlayer or insulator and freely flow through them. Both mechanisms were observed in the I-V curves of the $${\left[GaSb/Mn\right]}_{n}$$ multilayers (see Fig. 6).

In the case of $${\left[GaSb/Mn\right]}_{n}$$ multilayers, it is possible to observe the contribution of both conductive mechanisms. The conductive filament was observed in the formation of loops (insets in Fig. 6) when changes from HRS to LRS (and vice versa) occur 16. In addition, the switching ratio (SET/RESET) of the $${\left[GaSb/Mn\right]}_{3}$$ multilayers with Ts = 300 K was 1.31, whilst, $${\left[GaSb/Mn\right]}_{12}$$ multilayers was 116.73; therefore, the multilayer with 12 periods shows greater write/erase property.

On the other hand, the conductive filament is constituted by the redistribution of charge carriers within the GaSb semiconductor layer due to the presence of vacancies (gallium (VGa) or antimony (VSb) vacancies), the nature of the GaSb charge carrier (p-type), and Mn ions associated with the diffusion process. This conductive filament communicates the GaSb semiconductor layers with the Mn layers inside the multilayer structure. For this reason, the loop formation is affected by the voltage bias and the number of layers. Therefore, as the number of layers increases, the changes in resistivity are more difficult and the loop size increases too. The I–V hysteresis behavior revealed a major curve aperture, due to the increment of the GaSb layer number, this can be related with an increment of SCLC regions nearly to metal–semiconductor interfaces doing that the resistive switching be biggest in comparison to samples with minor GaSb/Mn spatial periods. In consequence, voltage at resistive switching occurs was increased (see Fig. 6).

In order to evaluate the reproducibility and reliability of the multilayers, the endurance, and uniformity properties of the devices have been investigated (see Fig. 7). The resistances are read at ~ 0.5 V for the 20 cycles at room temperature. In the endurance characteristic, it is possible to observe that the HRS varied with increasing of the cycles numbers whilst LRS maintain more stable during the cycling tests. This behavior may be attributed to the stability of the conductive filaments during formation due different interface structures of the multilayers38,39.

The permanence and uniformity of the resistive state is modified due to the number of layers of the sample (see Fig. 7). When the layers of the multilayer are 3, it can be observed large and stable memory window around 102, which is convenient for the identify the high and low resistance values. In the case of multilayer with 6 layers, the uniformity of HRS decrease and the memory widows was 10 units. This memory windows keeps stable without significant degradation up to 372 s, suggesting that the multilayer GaSb/Mn have the potential application in RRAMs.

For the construction of non-volatile memories based on resistive random access memory (RRAM) technology, thin layers of metal oxides have been used with metal–insulator-metal (MIM) structures described as two-terminal devices and in which the switching between two resistive states40. As an insulating material, the use of semiconductor oxides with a thin layer structure41, nanoparticles40, and more recently, nanotubes16 has been implemented. In the case of nanoparticles, it has been shown that the density and homogeneity of the nanoparticles affect resistive changes, as well as the possible interactions between the interfaces. Whilst, nanotube structures have shown a high directionality of the charge distribution, which strongly contributes to strong changes in resistivity and a stability of the conducting filament16. In comparison, multilayer structures can contribute to the formation of distributed load trapping, allowing the control of resistive changes. These multilayers show a first approach to the study of nanostructures that contribute to the construction of non-volatile memories based on two-terminal devices.

Nevertheless, in the low voltage region, the behavior corresponds to a SCLC mechanism. Figure 8 shows log I as a function of log V of multilayers architecture and its corresponding fitting, wherein the lowest voltage (region I—Fig. 8) presents an ohmic behavior and the region between 1 and 2 V (region II—Fig. 8) evidences the SCLC mechanism. In both cases, region I and Regions II, the fitting results were R2 = 0.999 and 0.993, respectively.

When the injection of the initial charge carrier is higher than that of recombination19, the injected carriers form a space-charge region, therefore, the current flow is limited in this region (region II). This occurs before the formation of the conductive filament. After region II, the trap-filled limited conduction mechanism is located in region III (Fig. 8). If the current is limited by the drift component of injected carriers and reorientation of the charges, the SCLC density is expressed as:

$$J=\frac{9\mu \varepsilon {V}^{2}}{8{L}^{3}}$$

(1)

where $$\mu$$ is the carrier mobility, $$\varepsilon$$ is the dielectric constant, $$V$$ is the applied voltage, and $$L$$ is the sample thickness19,42.

Taking into account the magnetic and resistive behavior of multilayers, Fig. 9 shows the I–V curve measurement scheme in the presence of an external magnetic field with H = 7000 Oe parallel to the current direction.

A modification of the loops is observed in I-V curves by the magnetic field on the sample (Fig. 10), as evidenced in an increased loop area (insets in Fig. 10), thus maintaining the SET and RESET states associated with bipolar behavior. In the case that the magnetic field is applied, it is observed that the changes from HRS to LRS occur in values of approximately two orders of magnitude (from − 4 to − 6 in Log(I(A))) (Fig. 10b), while for the curves without magnetic field, HRS to LRS values of less than one order of magnitude are recorded (Fig. 6b). This behavior evidences significant a magnetic switching control on resistive properties, which makes these multilayers a promising material for Magnetic Resistive Random Access Memories (M-RRAM)41.

As in the I–V characteristic in the absence of a field, an increase in the size of the loop is observed when the external magnetic field is applied. This evidence a magnetoresistive effect on the multilayers that contributes to resistance changes and allows its modification through the application of an external magnetic field.

#Exchange #bias #coupling #bipolar #resistive #switching #room #temperature #GaSbMn #multilayers #resistive #memories #applications #Scientific #Reports

Credit