1. Introduction
Nowadays, sustainable polices and strategies are based on transition from a linear to a circular economy model, where worldwide research has presented an upward trend since 2015, matching with the aims of the 2030 Agenda [
1]. Among the targets which must be addressed in the following years is advanced research on consequences of climate change; study efficient waste management to avoid a negative environmental impact; growth of agricultural activities; innovation in sustainable processes, goods, and services; and recycling of waste materials [
2,
3]. In this way, waste will become a valuable resource that must be reintroduced into the manufacturing processes; therefore, seeking material attributes (stable chemical properties, electrical, magnetic, and mechanical robustness, etc.) is still ongoing.
The attention for recycling principle has grown considerably, both for fundamental research and for novel applications [
4]. Recycling principle is focused on preventing either burning to recover some of the metals or else grinding wastes up and treating, and instead recovering 90% or more of valuable materials (gold (Au), silver (Ag), copper (Cu), iron (Fe), silicon (Si), graphite and others) through the zero-wastewater and zero-carbon dioxide emissions; avoiding these results in stationary storage for long periods of time, which might result in a black mass with dangerous solvents and tens of thousands of tons of waste entering into the landfills. Therefore, as electrical and electronic hardware production soars, so does interest in recycling, these being one of the largest sources of waste of clearly a lot of value. Rather than dismantling a hardware pack into cells and discharging them, they can be separated in modules and studied separately to identify their special properties [
5]. Due to the abundance and industrial usage today for development of worldwide items, recycling 3% Si steel and graphite as building blocks of emergent devices could be better than if these are, rather, long-lasting waste sources [
6].
Fe–Si steels play an important role in energy systems, where these are mainly used for manufacturing core-laminated transformers and rotating engines. Three percent Si steel-laminates have been widely used in electrical built-in home appliances, such as portable radios, ventilation systems, pump systems, lighting sources, and others, where low core losses and high permeability are required for the Fe–Si steels [
7]. For commercial 3% Si steel grade, well-referred here as Fe–Si steel, their chemical composition must be intentionally modified to improve magnetic properties by adding 0.4% silicon, 0.33% aluminum, 0.55% manganese and 0.013% phosphor in percent per weight in accordance to previous studies [
8]. These magnetic properties are strongly influenced by microstructural parameters such as grain size, residual stresses and crystallographic texture.
Graphite is one of the available intercalated compound-based materials with a weak interlayer interaction between individual graphene sheets, consisting of the mixture of clay particles, mainly silicon dioxide (SiO
2) and a minor amount of metal oxides in graphite lattice [
9,
10]. During the accommodation of clay particles it is expected that carbon atom vacancies and lattice defects in-plane and out-plane of the graphite structure occur, which make graphite a good lubricant and a lead pencil that people use in their daily life for writing, sketching, etc. For a long time, graphite, as an isolated layer, has been well studied, but its physical properties for engineering application became available in 2016 [
11]. One direction that today shows promising result is assembling graphene layers (stacked graphite) on silicon carbide (SiC) substrate by reverse abrasion technique, where the control over the coverage percentage and thickness are advantageous; here, low carrier mobility about 250 cm
2/Vs (several order of magnitude lower than in silicon), scattering mean free path l = 235 nm, and middle conductance at room temperature might allow charge carriers to move independent of their energy and space-charge-limited conduction (SCLC) inside the graphite bulk at nearly the length from their interface.
Today, analog computing is proposed and studied extensively to be capable of executing easily arithmetic functions at much lower power consumption than digital solutions. An analog scheme can allow versatile conduction modes under a wide range of conditions, which means that continuous time processing will be more beneficial than discrete time processing [
12,
13]. Thus, seeking ecofriendly alternatives for analog computing within a reduced number of building blocks is of importance. So, too is motivating research systems out of equilibrium, such as iron oxide-coupled Fe–Si steel, which might induce magnetostrictive behavior by perpendicular current injection inside the graphite layers. In this case, proximity effects by coexisting magnetostriction and disorder in lattice structure at the iron oxide surface might be responsible for functional properties in the Fe–Si steel/oxide interface when alternating current signal is injecting in a Fe–Si steel/iron oxide/graphite structure [
14,
15]. Thereby, the work presents the salient experimental observations of an iron oxide-coupled graphite/Fe–Si steel structure to understand how exchange interactions at the interface of iron oxide/Fe–Si steel can induce SCLC in graphite layers of different compositions. Stages of synthesis and parameters to build and analyze the structure will be explained in
Section 2. Results on structure formation in graphite and iron oxide, phenomenological description of the current injection response and an example for analog computing in a practical structure will be covered in
Section 3.
Section 4 describes the discussion of the previous studies and hopeful findings with the theoretic analog device architecture useful as an arithmetic processor. Finally, conclusions about this research are presented in
Section 5.
2. Materials and Processes
To study physical properties in the iron oxide-coupled graphite/Fe–Si steel structure, a methodology was developed based on probing dynamic action at two interfaces in the schematic of
Figure 1. The proposed structure included one graphite layer physically assembled on silicon carbide (SiC) sandpaper as a flexible substrate and one iron oxide thin-film grown on the Fe–Si steel foil by thermal oxidation. Both layers were mechanically joined between two large-area aluminum electrodes (1 cm wide × 3 cm in length), as shown in
Figure 1a, to achieve a proximity coupling into the active region of area (0.5 cm × 0.5 cm), where the iron oxide thin-film must act as a quasi-insulating layer to drive the conducting channel of the charge carrier´s in the space-charge-limited conduction (SCLC) regime.
Figure 1b shows the assembling diagram of all the layers involved in the whole structure engineered here, where magnetic character in Fe–Si steel and iron oxide layers was initially defined by equilibrium states (random arrows showing magnetic moments), and disorder in the graphite matrix depended on misfits and vacancies occurring as a function the mechanical accommodation on the SiC surface. Two interfaces emerged from the structure: first, Fe–Si steel/iron oxide; second, iron oxide/graphite layer.
In the fabrication of the structure processed here, the following two stages were involved: (1) Lead pencil rods were used as graphite source to grow the graphite layers of three different thicknesses assembled on SiC sandpaper using the reverse abrasion technique [
11], where the sizes of these layers were 0.5 cm wide and 2 cm in length. (2) Fe–Si steel foil with 3% Si steel grade, of an average thickness of 0.18 mm, recovered from a useless low power core-laminated transformer, was used as substrate for synthesis of the iron oxide thin-film.
It is well known that effects such as influence of temperature, humidity of the atmosphere, oxidation time and controlled diffusion along grain boundaries at low temperatures (<500 °C) are expected to play a major role in the oxidation kinetic [
16,
17]; therefore, in order to achieve better adherence of the oxide on the Fe–Si steel foil, and more uniformity of its thickness in nanometer scale, to be useful as the coating layer responsible for the structure performance of
Figure 1, the following stages were accomplishing for their preparation: (a) Fe–Si steel foil was cut to a cross-section area of 1 cm × 2 cm and both upper and bottom sides of the foil were mechanically grounded using 1200 grit SiC paper; (b) the Fe–Si foil was cleaned with home-used organic solvent, known as thinner, and dried for about 2 h in ambient atmosphere; (c) a sealed laboratory furnace to avoid gas leakage was used; (d) air (21% O
2, 79% N
2) was chosen as oxidizing atmosphere; (e) after 30 min, the furnace temperature was stabilized, then Fe–Si steel foil was introduced into the quartz tube; (f) thermal oxidation was performed at 250 °C for a duration of 60 min. Finally, next to the oxidation stage, the sample was cooled out of the quartz tube, whose average thickness of the dark blue iron oxide was ~120 nm and was estimated by a profilometer (KLA Tencor, P15, Milpitas, CA, USA).
Structure formation of graphite layers and iron oxide grown on Fe–Si steel foil was investigated by Raman spectroscopy to know their associated bands; the Micro-Raman System (HORIBA Jobin-Yvon, HR800, Piscataway, NJ, USA) of excitation line λ = 632.8 nm (He-Ne laser) at 20 mW and 50× objective was used to avoid laser-induced heating.
Current signal injection response was performed and evaluated by current–voltage plots at room temperature using a digital storage oscilloscope (Tektronix, TDS1012C, Beaverton, OR, USA). In addition, a correlation study between low-field magnetic ordering and structure formation was addressed, using a methodology before reported, to evaluate hysteresis parameters such as magnetization, M, effective field strength, H
eff, loss coefficient, tan δ, equivalent resistance, R
S, and equivalent self-inductance, L
S, in magnetostrictive composites based on a series resistor-inductor-capacitor (RLC) circuit [
18].
4. Discussion
Nowadays, digital architectures result in increased power consumption and temperature runaway, because billons of transistors operate under switching conduction to ensure that routines performed by a series of discrete steps become the needed algorithm solution. In addition, it is known that electric and thermal stresses during switching conduction of transistors cause limitations as coding faults when these are operating at higher speeds [
12,
19]. To improve reliability in digital systems, emergent schemes in recent years are generating renewed interest, whereby continuous growth of neuromorphic computing for bio-inspired data processing and hardware emulation of cellular building blocks are widely explored to perform analog-valued encoding [
13]. However, a growing number of building blocks, difficulty to configure arbitrary connections between distant blocks of devices and scaling linearly with the size are some drawbacks in analog schemes.
Using the structure and conduction properties of Fe–Si steel and graphite might help enable the analog-valued encoding with advantages in operation, such as a reduced number of devices to avoid large connections and unstressed electrical conduction. Such properties will play a dominant role in the performance when energy loss-related resistive conduction of excited charge carries, and proximity effects (local temperature changes, magnetic ordering, electrostatic interactions, etc.) by magnetostriction and disorder effects, allow for adjustable exchange interactions at the surrounding interface in the α-Fe
2O
3-coupled Fe–Si steel/graphite structure. In addition, they will allow for reconfigurable action from the α-Fe
2O
3 oxide [
15,
27], being normally antiferromagnetic as a function of the structure ordering between Morin temperature (T
M = 250 K) and Néel temperature (T
N = 950 K), in which oxygen vacancies into their band gap (~2.2 eV) will be responsible of their nonlinear electrical conduction [
28,
29].
For instance, the results of Raman spectroscopy in the previous section highlight the following findings. (a) The level of defects in the GR-1 sample is small because the peak intensity of D band is negligible and the peak intensity ratios of 2D and G bands are very similar, thus this graphite layer is not defect free. However, GR-2 and GR-3 samples reveal strong disorder by possible intercalated accommodation of clay particles in the graphite lattice; (b) the D band is larger than the G band in the GR-2 sample, which means that there are more defects in sp2 hybridized-carbon atoms than the presence of stable layers. Whereas, the 2D band is higher and D band is lower than the G band in the GR-3 sample, which means that more phonon vibrational modes are excited and a growing number of layers are stacked; (c) the difference in FWHM of the measured D band (<70 cm−1) and 2D band (72 cm−1) for the GR-2 sample and that for the GR-3 sample (130 cm−1) implies that activation of the defects built-in graphite sheets will be slightly effected by local temperature changes, dependent on the proximity of randomly distributed defects; (d) magnetic ordering shown in the Fe–Si steel/α-Fe2O3 interface was understood by FWHM of the wide band from 550 to 800 cm−1 and the peak intensity ratio of 233.67 cm−1 (A1g) and 672.76 cm−1 (Eu) bands. This specifies that α-Fe2O3 corresponds to an oxygen-enhanced oxide with weak magnetostriction, which confirm that nonlinear diffusion of the charge carries will happen when α-Fe2O3 oxide becomes perturbed by interface forces near the Fe–Si steel surface, whereby randomly distributed charge in the interior of domain walls results.
Therefore, a theoretic approach for engineering the arithmetic processor where its internal architecture may be tailored from electrical conduction properties found in graphite and Fe–Si steel, mediated by α-Fe
2O
3, must include a stripe array based on distributed resistance of graphite layers, being it directly coupled to the α-Fe
2O
3 surface, as shown in
Figure 8. Here, three graphite layers of different compositions must be assembled on silicon carbide (SiC) sheets, as in an earlier report [
5]. The resulted analog device architecture could be configured to accumulate the weighted sum models of the inputs data through load resistor within the two following principles. First, variable voltage-valued data, applied as input signal and nonlinearly perturbed by magnetostriction induced at the Fe–Si steel/α-Fe
2O
3 interface, is possible using a linear ramp of different amplitudes as the activation signal. Second, modified pulses-valued data, applied as input signal, could be combined with the magnetostrictive signal produced at the interface and replicated in the load resistor. Both scenarios might be used to extract equivalent arithmetic values of many independent input states, as presented in
Figure 8, with desired features when interconnecting load resistors in systems. For example, in resistor-capacitor (RC) circuits or complementary metal-oxide-semiconductor (CMOS) devices manufactured by silicon-based technology or emerging technologies [
19,
30].