3.1. X-ray Diffraction (XRD) Spectroscopy
To confirm the obtained samples from their crystal structures, the XRD patterns (
Figure 1) were analyzed for the prepared samples. The XRD pattern of Mg-MOF-74 describes sharp and robust diffraction peaks, which suggests the crystallinity and crystal structure of Mg-MOF-74 are perfect. Elaborately, there were three prominent diffraction peaks at angles 2θ of 6.8°, 11.8°, and 18.5° corresponding to the crystallographic planes of (210), (300), and (510) of Mg-MOF-74 single crystal [
16], respectively. The XRD patterns of the Mg-MOF-74@GO composites were similar to that of Mg-MOF-74. The featuring characteristic XRD signals of GO at 11.4° could not be detected in Mg-MOF-74@GO composites by this analysis. No obvious characteristic diffraction peaks of GO appeared in the XRD spectrum patterns of Mg-MOF-74@GO composites because GO was encapsulated inside Mg-MOF-74 or the content was low [
16]. Moreover, a sharp peak of GO at 11.4° indicates the characteristic (002) reflection with interlayer spacing of 7.70 Å. The interplanar spacing of Mg-MOF-74@GO(2:1), Mg-MOF-74@GO(4:1), and Mg-MOF-74@GO(5:1) composites was 7.45 Å, 7.51 Å and 7.48 Å, respectively.
The pristine Mg-MOF-74 and its composites exhibited sharp characteristic peaks at 6.8°, 11.8°, 15.3°, 16.7°, 18.5°, 19.3°, 20.5°, 21.6°, 23.8°, 24.7°, 25.7°, 27.5°, 28.3°, 29.2°, 30.0°, 30.8°, 31.6°, 33.8°, and 34.5°. The well-defined sharp peaks of the XRD patterns for all samples confirm that Mg-MOF-74 and its composites have good crystallinity. However, qualitative comparison between the Mg-MOF-74 composites and the pristine material witnesses a light narrowing and an intensity diminishing in the principal peaks at 6.9° and 11.9°, indicating that the doping implies some structural defects [
16]. In contrast with Mg-MOF-74, the first Bragg diffraction peak of Mg-MOF-74@GO(2:1) has a slight shift from 6.8° to 6.9°, and the peak intensity ratio I
300/I
210 increases from 1.0 to 1.2. These changes suggest that GO was successfully encapsulated into Mg-MOF-74.
3.2. Fourier Transform Infrared (FTIR) and Raman Spectroscopy
As a matter of fact, FTIR spectra (
Figure 2a) can effectively detect the composition of the sample. The FTIR spectrum of Mg-MOF-74 can be divided into two different regions. On the one hand, most of the activation modes above the 700 cm
−1 region belong to the organic ligands [
17]. In the spectrum of Mg-MOF-74, the peak at 1574 cm
−1 is ascribed to the vibration of the C=C bond in the benzene ring. The peak at 1417 cm
−1 is associated with –O–C–O in the carbonyl group and the peak at 1208 cm
−1 conforms to the C–O vibration. On the other hand, the characteristic vibrations in the low wavenumber region (<700 cm
−1) largely cover the ones that impact on the metal center [
16,
17]. The vibration of Mg–O accounts for the vibration located at 591 cm
−1, proving the formation of the metal organic framework. Mg-MOF-74 is composed of Mg(II) ions and DOBDC, and forms a hexagonal honeycomb structure. In this structure, each Mg(II) ion is bonded to five oxygen atoms: three with carboxylate groups and two with hydroxy groups, leaving an open metal site. This open metal site provides a beneficial sorption site for the various guest molecules. After loading GO for the Mg-MOF-74@GO(2:1) composite, further investigation sees the vibration peak of Mg–O changed from 591 cm
−1 to 587 cm
−1, and the peak of –O–C–O at 1417 cm
−1 belonging to benzene ring simultaneously shift to 1420 cm
−1. The vibration deviations of the benzene ring and the coordination bond between the metal and O prove the successful loading of GO [
17].
The most prominent Raman characteristics displayed by the Raman results (
Figure 2b) are the D and G bands in GO. The D band stems from the breathing vibration of the sp
2 carbon atom in the ring, whereas the G band is derived from the stretching vibration of the sp
2 atom pair in the carbon chain and the ring [
17]. A D band will appear once the degree of disorder for GO increases. The G band at 1594 cm
−1 and the D band at 1354 cm
−1 can be clearly seen, proving that GO was in a disordered state. The ratio of D band intensity to G band intensity, I
D/I
G, represents the defect degree of graphene [
17]. Here, the I
D/I
G value of GO up to 0.89 indicates that GO was oxidized and that structural defects occurred. Likewise, three Mg-MOF-74@GO composites maintained the D and G bands of GO.
Table 1 compares that the I
D/I
G values of the Mg-MOF-74@GO composites which all increased, indicating that MOF was successfully modified with different contents of GO. However, there are some differences between the Mg-MOF-74@GO composite and GO. New bands appear at 1507 cm
−1 and 1462 cm
−1 with the coexistence of Mg-MOF-74 and GO in the Mg-MOF-74@GO composites. The D+D’ band was only observed in samples with significant amounts of defects, a band observed in all graphene composites accentuates in Mg-MOF-74@GO composites. The Mg-MOF-74@GO composites show a higher graphene stacking as observed in the growth of the ratio I
2D/I
G, which is proportional to the number of graphene layers.
3.3. Thermal Analyses
Figure 3a exhibits the TGA–DTG curves of the Mg-MOF-74, GO, and Mg-MOF-74@GO composites. The thermal decomposition behavior of the Mg-MOF-74@GO composites is similar to that of the pristine Mg-MOF-74. The thermal decomposition of Mg-MOF-74 can be divided into two stages: the temperature ranges from 110 °C to 150 °C, and those from 380 °C to 420 °C. The first mass loss of about 10% occurs in the range from 110 °C to 150 °C and can be attributed to desorption of the H
2O solvent trapped in the micropores of Mg-MOF-74 [
18,
19]. The second mass loss of about 30% can be ascribed to the decomposition of the framework structure. This leads to the production of carbon-containing gases like CO, CO
2, and C
xH
y hydrocarbon mixture, as well as a small amount of H
2 [
19]. For the thermal degradation behaviors of the Mg-MOF-74@GO composites seen from the DTG curves, the degradation process displays three maxima of decomposition. The initial decomposition arises as the sample gains the 5% weight loss. The 10% weight loss (T
10) happens when the temperature of maximum decomposition rate acts as the maximum signal in the DTG curves. The temperatures of 5% (T
5) and 10% (T
10) weight losses (
Table 2) illustrate that incorporating GO into Mg-MOF-74 can enhance the decomposition temperature (T
d) of the composites. Moreover, an increasing content of GO in the Mg-MOF-74@GO composites can give rise to a continuous increase in T
d. This finding originated from the fine dispersion and hydrogen bonding interaction between Mg-MOF-74 and GO [
18,
19]. The thermal stability of the Mg-MOF-74@GO composites is higher than that of Mg-MOF-74 during the first mass loss stage of Mg-MOF-74@GO composites. This illustrates that the wrapping of GO weakens the decomposition process.
DSC analyses (
Figure 3b) were implemented to investigate the T
g of the Mg-MOF-74@GO composites. An increase in T
g could be owed to an effective attachment of Mg-MOF-74 to GO sheets, which constrains the segmental motion of the Mg-MOF-74 chains by hydrogen bonding and electrostatic attraction [
19]. The T
g value of 381.9 °C for Mg-MOF-74 is consistent with that reported in the literature [
19]. For the Mg-MOF-74@GO composites, a higher T
g is observed. Moreover, this value increased from 391.2 to 395.2 °C by loading more weight ratio between Mg-MOF-74 and GO. Such high T
g values describe the high affinity between Mg-MOF-74 and GO in account of the high compatibility and adhesion between their two phases, as previously observed. The thermograms of all the Mg-MOF-74@GO composites demonstrate a decrease in the intensity of peak corresponding to GO. The intensity of the sharp exothermic peak for GO (5.305 mW/mg) separately decreased to 0.1456 mW/mg, 0.2833 mW/mg, and 0.4343 mW/mg for the Mg-MOF-74@GO(2:1), Mg-MOF-74@GO(4:1), and Mg-MOF-74@GO(5:1) composites. In addition, the DSC of composites reveals that two endothermic peaks corresponding to Mg-MOF-74 and GO are shifted from 127.3 °C to 136.8 °C; and from 414.6 °C to 422.7 °C. The above observations prove the interaction between Mg-MOF-74 and GO.
The high Td and Tg of the Mg-MOF-74@GO composites particularly exhibits their applications for the long-term stability of device operation. The XRD, FTIR spectroscopy, and DSC studies were performed on the Mg-MOF-74@GO composites to identity the possible interaction between Mg-MOF-74 and GO and to confirm the formation of Mg-MOF-74@GO composites.
3.5. Multi-Bit Memristic Behaviors of MOF-Based Memory Devices
In this paper, the Mg-MOF-74@GO composite was first utilized to fabricate a novel Ni/Mg-MOF-74@GO/ITO device (
Figure 5) that displays multi-bit memristic characteristics, whose architecture is versatile and compatible with the current parallelism demands of integration. The MOF-based film was prepared by the solution-processable method on the ITO substrate without surface modification and complex equipment.
To explore the memristic behavior of the structure Ni/Mg-MOF-74@GO/ITO, the I–V curves of Ni/Mg-MOF-74@GO(5:1)/ITO (
Figure 6a) were swept with the compliance current of I
CC = 0.1 A. Originally, the device was at a high resistance state (HRS or OFF state). As the bias voltage (applied to TE) swept negatively, the current suddenly increased at the SET voltage V
SET1 = −0.71 V, indicating that the device switched to an intermediate resistance state (IRS or ON1 state). Then, the device was kept at a low resistance state (LRS or ON2 state) when swept, until the voltage was as large as V
SET2 = −1.49 V. As the voltage swept in the opposite direction, the device returned from LRS to IRS to HRS when the bias voltage separately reached 5.14 V and 5.57 V, denoted as the RESET voltage (V
RESET1 and V
RESET2). This observation demonstrates that the device exhibits ternary memristic behaviors. Moreover, the ratio of its memristance in HRS, IRS, and LRS (R
HRS:R
IRS:R
LRS) is roughly 10
3:10
2:1. When the device suffered from one hundred of times for alternative cycle sweepings (
Figure 6b), the central value of V
SET1, V
SET2, V
RESET1, and V
RESET2 (
Figure 6c) was −0.85 V, −1.75 V, 4 V, and 5.25 V, respectively. As for the statistics from the results of repeated tests, the cycle-to-cycle distribution (
Figure 6d) of R
HRS, R
IRS, and R
LRS was relatively narrow. Moreover, the retention time of the tristable states (
Figure 6e) was over 10
4 s at an absolute readout voltage (V
READ) of 0.1 V. All these results reveal that the fabricated devices have the advantages of superior stability and reliability. For comparison, the memory device based on pristine Mg-MOF-74 was fabricated but it exhibits no memristic behaviors.
In order to better understand the underlying physical mechanism, the different loading of Mg-MOF-74 to GO (4:1 and 2:1) was further explored. The electrical characteristics of the MOF-based memory device show that the ternary memristic behavior can be modulated with the presence of GO. Accordingly, the electronic properties of Ni/Mg-MOF-74@GO(4:1)/ITO (
Figure 7a) were investigated. With the increased voltage bias from 0 to −6 V, the current separately jumps at V
SET1 = −0.57 V and V
SET2 = −0.94 V, indicating the switching from HRS to IRS to LRS. Opposite to that bias direction, two sudden decreasing currents appear at V
RESET1 = 3.45 V and V
RESET2 = 3.9 V. Under the cycle-to-cycle sweeping mode of the endurance properties (
Figure 7b), the average values of V
SET1, V
SET2, V
RESET1, and V
RESET2 (
Figure 7c) are −0.82 V, −1.70 V, 4.48 V, and 5.30 V, respectively, and R
HRS, R
IRS, and R
LRS (
Figure 7d) range from 1.15 kΩ to 9.07 kΩ; from 0.061 kΩ to 1.76 kΩ; and from 30.56 Ω to 150.83 Ω, respectively. The retention ability (
Figure 7e) was also investigated, in which the voltage bias of 0.1 V was carefully selected to read R
HRS, R
IRS, and R
LRS. According to the results, the Mg-MOF-74-based device with excellent memory stability and reliability exhibited almost no fluctuation after either 10
4 s or 100 alternative cycle sweeping.
The multi-bit data storage performance of the as-fabricated memory device Ni/Mg-MOF-74@GO(2:1)/ITO (
Figure 8a) demonstrates that the multilevel reversible resistance level can be implemented. The device characteristics are reasonably promising for a prototype: setting from HRS to IRS to LRS occurred at −2.44 V and −3.25 V, resetting from LRS to HRS at 3.57 V, and a R
HRS:R
IRS:R
LRS ratio of 12:2:1.
Figure 8b shows that the resistance values can be maintained well during a 100-cycle scanning. Investigation of the cycle-to-cycle distribution (
Figure 8c,d), the SET and RESET voltage, as well as the mean memristance are summarized in
Table 3 and
Table 4. The mean value of V
SET1, V
SET2, and V
RESET are −1.63 V, −2.96 V, and 3.62 V, respectively, while that of R
HRS, R
IRS, and R
LRS are 0.73 kΩ, 0.078 kΩ, and 58.12 Ω, respectively. This distribution supports the multi-bit information storage performance of the Mg-MOF-74@GO memory devices. An absolute read voltage of 0.1 V was required with a retention time of 10
4 s, as shown in
Figure 8e.
As summarized in
Table 3 and
Table 4, the devices with different chemical constituents of Mg-MOF-74 and GO endow a series of ternary memory behaviors. The statistical analysis of the SET and RESET voltages was assessed, as well as the memristance R
HRS, R
IRS, and R
LRS for the Mg-MOF-74-based memory devices operated at ambient temperature. With the incremental increase in content of GO from 16.7wt% to 33.3wt%, R
HRS and R
IRS decline gradually while R
LRS is well maintained, indicating that LRS is dominated by non-metallic or metallic CPs [
9]. To verify the composition of CPs, R
LRS was examined at different temperatures (
Figure 9), which were measured under 0.1 V. R
LRS was tested under different temperatures ranging from 20 °C to 80 °C. Based on the data analyses, the temperature-independent R
LRS can be verified, which may contradict the typical characteristic of the metallic CPs.
3.6. Mechanism of Multi-Bit Memory Behaviors
It was of interest to explore the conduction mechanism of the Mg-MOF-74-based devices. Then, a double logarithmic plot of the I–V curves in the SET process (
Figure 10) was developed. The charge transport obeys the model of SCLC in HRS. Elaborately, the low-voltage region with a slope close to 1 (I ∝ V) exhibits an Ohmic conductivity characteristic. In contrast, the high-voltage region follows either a square-law dependence (I ∝ V
2) where trap states are partially filled by electrons, or the power law (I ∝ V
α, α > 2) that corresponds to the trap-filled limited conduction.
The I–V relationship in HRS (
Figure 6a,
Figure 7a and
Figure 8a) contains two different conductive regions as follows: (i) a low-voltage region where the I–V curves display a linear behavior with slope values of 1.18, 1.08, and 1.07, which conform to the Ohmic-like conduction mechanism, and (ii) a high-voltage region where slope values are 3.36, 1.47, and 1.59. For the I–V curve in the ON1 state, the plot of logI–logV is still well fitted to the lines with a slope of 2.07, 3.46, and 1.73. Such linear relations demonstrate that the conductive process in the ON1 state was dominated by SCLC [
4]. In addition, the fitting result at the ON2 state exhibits the Ohmic-like conduction behavior with a slope of 1.00, 1.07, and 1.20.
Apart from the GO-mediated memristic property, the fitting results can be further illustrated. The I–V characteristics were controlled by a different content of GO adsorbed in the Mg-MOF-74 matrix that enabled for additional routes for electron-hopping. As a result, the effect of GO tuning on the ternary memristic behaviors of the Mg-MOF-74-based devices was demonstrated.