Strontium titanate (STO) thin films are promising for dynamic random access memory (DRAM) capacitors, owing to their high dielectric constant (κ) of ∼150 or higher in the crystalline phase,^[1] good thermal stability, good durability, and good compatibility with the device fabrication processes.^[2,3] In DRAM applications, a flat dependence of the capacitance on the applied voltage is required, but most of the existing high-κ materials exhibit a quadratic voltage dependence , where is the capacitance at 0 V and α and β are the quadratic and linear coefficients, respectively. For most high-κ materials, α is in the 100–1000 ppm/V² range^[4–6] and can be reduced in a bilayer capacitor with materials having opposite-sign (positive and negative) values of α.^[7] STO has a negative α and can be used in combination with HfO₂, ZrO₂, Al₂O₃, and TiO₂, all of which have positive α.^[4,8,9] Another challenge is a high leakage current density of STO. Hence, Al₂O₃, with its large bandgap of 8.7 eV, is considered a good candidate, because it can reduce the leakage current to less than 10⁻⁷ A/cm².^[10] However, this occurs at the expense of the capacitance density, owing to a smaller κ value of A₂O . Currently, ZrO₂ is used in DRAM capacitors, but it is generally accepted that reduction of the current density with the Al₂O₃/ZrO₂/Al₂O₃ nano-laminate for sub-20 nm technology is unrealistic.^[11] To achieve a larger capacitance density, high-κ materials such as TiO₂, with positive α, are better candidates. A capacitance density of ∼100 fF/um² has been reported for a (0.5 nm)TiO₂/(7 nm)STO bilayer; however, the current density was above A/cm² at 1 V, owing to the smaller bandgap of TiO of ∼3 eV.^[12,13] Therefore, different types of TiO₂/STO structures must be produced and the fabrication conditions must be optimized.

In this study, a MIM structure with a TiO₂ thin layer sandwiched between two STOs (STO/TiO₂/STO) was designed and prepared using RF and DC magnetron plasma sputtering. The thickness of the TiO₂ layer was varied while those of the STO layers were kept constant. The top and bottom TiN electrodes were used to avoid the formation of interfaces between TiN and STO.^[10] The SiO₂ film was deposited to minimise the leakage current by increasing the resistivity of the substrate. All the oxide films were deposited in oxygen-rich plasma by controlling the flow of oxygen to decrease the effects of the defects generated by oxygen vacancies. The variations in the capacitance density and current density were measured to determine the electrical properties, and XRD and Raman scattering analyses were performed to examine the film morphology.

An AJA ATC Orion system equipped with five magnetrons equipped with RF (13.56 MHz) and DC power supply at the City University of Hong Kong was used for depositing TiN/STO/TiO₂/STO/TiN/SiO₂/Si thin films. P-type Si (100) 1 × 1 cm² samples were used as substrates. The samples were cleaned in an ultrasonic bath with acetone, methanol, and de-ionized water sequentially, for 15 min each. The oxide films were deposited on an RF magnetron system using 99.99% pure SiO₂, STO, and titanium (Ti) targets. Oxygen was also supplied using argon (Ar) as the sputtering gas during deposition, for minimising defects induced by oxygen deficiencies. The TiN films were deposited using DC magnetron sputtering and the optimal deposition conditions were determined by measuring the film thickness using a MCM-160 rate/thickness meter with the shutter closed. The sample holder was rotated at 50 rev/min at a fixed distance of 10 cm. The vacuum chamber was evacuated to a base pressure of Torr with the shutter open and the films were deposited at room temperature. The deposition rate was 0.3 Å/s and the deposition time was increased in steps of 3 min. The approximate thicknesses, , T ₄, and T ₅ of TiO₂, were 5.4 nm, 10.8 nm, 16.2 nm, 21.6 nm, and 27 nm, respectively. Finally, the TiN top electrodes were sputter-deposited using a mask containing mm² holes. The important deposition parameters are listed in Table 1.

Table 1.

Table 1.

Table 1. Important experimental parameters. . Thin film Ar flow/sccm O2 flow/sccm N2 flow/sccm Power/W Pressure/mTorr Deposition rate/Å·s−1 Time/min SiO2 15 3 - 200 2.3 0.9 19 STO 15 5 - 200 2.7 0.9 19 TiO2 15 5 - 200 1.8 0.3 3 TiN 5 - 15 130 1.8 0.8 40

Table 1. Important experimental parameters. .

Figure 1(a) shows the XRD patterns of the deposited films. No peaks are associated with SiO₂, indicating that the SiO₂ thin film is amorphous. After the deposition of TiN on the SiO₂/Si substrate, a mixed-structure TiN film with the (111) preferred orientation (JCPDS 38-1420)^[14] was observed. The sharp peak observed for the STO thin film on TiN/SiO₂/Si was attributed to the perovskite structure of STO.^[15]

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. (color online) (a) XRD patterns of the different stacked structures on Si (100), and (b) a FE-SEM micrograph of the sandwiched TiO2 film with a thickness of T 2.

Pure perovskites have the cubic structure with poor porosity and small specific surface area. Surfaces are more suitable for avoiding intermixture of particles from the TiO₂ and STO films owing to adsorption, diffusion, and surface reactions. TiN was chosen as the electrode material because it has been reported that TiN films prevent the diffusion of STO and TiN films^[10] and are more suitable than TaN electrodes. By depositing TiO₂ films on STO/TiN/SiO₂/Si substrates, we obtained crystalline films with the predominant anatese-TiO₂ phase, also demonstrating the presence of the rutile phase. The anatese phase has been reported to exhibit comparatively smaller κ-values, in the 30–70 range,^[16] compared with the rutile phase, for which the κ-value is ∼110. The rutile phase has been reported to form for C, potentially affecting the thermal budget of the process. It has been reported that by controlling the DC or RF magnetron sputtering conditions, the rutile phase can be obtained without heat treatment.^[17–19] In Fig. 1(b), a FE-SEM micrograph of the film with a thickness of T ₂ reveals a stacked structure of the film. Although the thickness of the TiO₂ film could not be accurately measured owing to the fact that a diamond cutter eventually damages the structure, the thickness of the STO layer was estimated to be ∼100 nm, which is close to the expected thickness determined by a thickness meter. No interface was observed at the film boundaries, and the film thickness was uniform.

Figure 2 shows the Raman scattering spectra of the structure, before and after the deposition of the last STO film on TiO₂. The spectra were obtained using a Renishaw Raman System 2000 equipped with a 514 nm Ar laser. Although the TiO₂ film is very thin (∼5.4 nm), prominent peaks associated with anatese TiO₂ are observed at 147.7 cm and 630 cm .^[20] No peaks associated with the rutile phase are observed, confirming that the dominant phase is anatese. Owing to the small thickness of the TiO₂ film, the spectrum also includes contributions of reflections from the STO films. The obtained spectrum was compared to previously reported spectra for Sr- and Ti-rich or deficient STO films, and based on this comparison the deposited films were preferentially stoichiometric. The second-order peaks at 218 cm and 303 cm⁻¹ reveal the presence of the stoichiometric STO films similar to the XRD results, while the small hump at 350 cm may arise from the Sr-rich STO.^[21,22] The small-intensity broad peak at ∼535 cm indicates the first-order Raman process. The intensity depends on the film thickness and it increased for thicknesses . Better crystallinity was observed for the thinner films than the thicker ones.^[23] Although the intensity of the first-order peak is very small, it reveals the presence of crystal defects, which may be owing to the deficiency of oxygen atoms inducing a strain in the films.^[20,21,23] This is because oxygen deficiency in TiO₂ films yields structural defects. After the deposition of the STO films on TiO₂, the intensity of the peaks associated with TiO₂ diminished, and the STO peaks became narrower. The broad spectral features in the 550–800 cm indicate that the STO film is not a single crystal.^[23]

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. (color online) Raman spectra of different structures on the TiN/SiO2/Si substrate.

The I–V and C–V characteristics were determined using a Keithley 4200 SCS system at 1 MHz and 25 mV AC. Figure 3(a) shows how the I–V curves of the capacitors change as the TiO₂ film thickness changes from T ₁ to T ₅. For T ₁, the leakage current is sufficiently high, A at 1 V, and it abruptly increases above ± 2.1 V. It can be assumed that the TiO₂ film with such a small thickness compared to the bulk material acts as an interface and controls the leakage current of the capacitor. In this case, the bulk material is assumed highly conductive, owing to a large concentration of defects (oxygen vacancies or other structural defects). The abrupt increase in the current at 2.3 V suggests that trap states are generated below the conduction band. It has been proposed that if the interface plays a dominant role, then the leakage current should be attributed to the Schottky emission (SE), but if the bulk plays the dominant role, the leakage current should be attributed to the Poole–Frenkel emission (PFE).^[24] With regard to T ₁, at a small positive voltage ( V), the current increases linearly, exhibiting ohmic behavior. As the voltage increases, the electric field across the dielectric becomes sufficiently strong for injecting electrons into the near-interfacial states, subsequently sharply increasing the leakage current, as shown in the figure. These states are created owing to a positive potential generated by the presence of structural or oxygen vacancy defects. With regard to T ₂, the leakage current drops to A, but the interface still appears to play a dominant role in controlling the leakage current, because the current is still very large. There is a small increase in the current at 1.5 V, exhibiting the same behaviour as that observed for T ₁. The increase in the slope of the leakage current for V in the 0–1.5 V range, and the small increase in the current indicate that the decrease in the positive potential is owing to defects. This means that by further increasing the thickness of TiO₂, the defects are alleviated, correspondingly decreasing the positive potential. For T ₃, it is difficult to determine whether the interface or the bulk is dominant. The I–V curves exhibit the same behaviour as for T ₁ and T ₂, with a reduced leakage current of A at 1 V, in addition to the abrupt current increase at 2 V, signalling the reappearance of defects. The smaller positive potential than that for T ₁ indicates that there are fewer defects for T ₃. By increasing the thickness to T ₄, the leakage current decreases further to A, and the curve does not change as sharply as before. The almost symmetrical and smooth shape of the curve around 0 V indicates that the leakage current is dominated by the bulk and the mechanism is PFE, SE, or both. For the thickness T ₅, the current decreases to A at 1 V and the curve is asymmetrical around 0 V. The minimum of the curve at the negative voltage is owing to the migration of defects through the bulk, which creates positive fixed charges under the electrode area.^[25]

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. (color online) (a) I–V characteristics and (b) CV characteristics of devices with different TiO2 thicknesses.

Figure 3(b) shows how the capacitance density ( depends on the applied voltage, for different thicknesses. In all cases, the dependencies are almost straight lines, which is owing to the combined effect of α of STO and TiO₂. The capacitance density is given by ), where is the permittivity of the dielectric material (F/cm), A is the area of the capacitor, and T is the overall effective thickness of the insulator (STO + TiO₂ + STO); thus, is inversely proportional to the effective thickness of the insulator.^[26] Figure 3(b) is consistent with previous results, according to which decreases with increasing TiO₂ thickness.^[9,27] For attains a maximal value of ∼20 fF/um² and the curve is nearly a straight line. The smooth variations in the capacitance at 0 V and 1.8 V indicate a uniform concentration of defects under the electrode. By increasing the thickness to decreases to ∼16 fF/um², with small fluctuations owing to the motion of positively charged defects. The CV curve of T ₃ is similar to that of T ₁, showing variations at 0 V and 1.85 V, but it is not smooth, indicating a non-uniform concentration of defects under the electrode, with decreasing to 11 fF/um² for T ₃. With regard to T ₄ and T ₅, the CV curves are straight owing to the bulk effect of the capacitor, as explained before when discussing the I–V characteristics. The values are 5 fF/um² and 1 fF/um² for T ₃ and T ₄, respectively, and the leakage current is inversely related to .

The effects of a sandwich TiO₂ layer on the electrical properties of STO were investigated by varying the TiO₂ layer's thickness. XRD and Raman scattering analyses suggest that the deposited films are crystalline and the predominant phase is anastase. The capacitance density varies almost linearly with the applied voltage and the fluctuations in the CV curves decrease with increasing thickness. Maximum capacitance density is obtained for the smallest thickness (∼5.4 nm) and it decreases as the film becomes thicker. The leakage current also decreases with thickness but it is still quite high. Additional research is needed for optimising the conditions required for reducing the leakage current, while retaining a high capacitance.