High Performance and Reliability of High Dielectric Constant (Ba,Sr)TiO 3 Thin-Film Capacitor with Cu-based Electrodes
3.3.2 Impact of Cu-based Electrodes on the Reliability of Metal Insulator Metal (Ba,Sr)TiO 3 Thin-film Capacitors
3.3.2-1 Physical Characteristic
Figures 3.11 display SEM micrographs of the surface morphology of BST films deposited on Cu (BST-3) and TaN/Cu bottom electrodes (BST-4). As presented in Fig.
3.11(a), the surface of the BST film in BST-3 sample is rough because it exhibits pinholes and hillocks, mostly of CuO, formed by oxidation growth at the BST/Cu electrode interface. The oxygen atoms from the environment easily penetrate the BST layer via diffusion paths and begin oxidizing the Cu layer. Recent reports support this finding:[31,56] as the oxide continues to grow outward and the Cu ions are transported from the metal layer, a hillock of considerable size is formed on the surface. Nevertheless, the surface of the BST-4 sample, as shown in Fig. 3.11(b), has almost no hillocks or pinholes. This study established that the BST layer on the TaN/Cu bottom electrode is impermeable to diffusing oxygen and so protects the underlying Cu layer against oxidation. The SEM observation is consistent with the cross-sectional TEM micrograph of the BST-4 sample, as shown in Fig. 3.12. The interface between the TaN and Cu layers is smoother and no reaction is observed between them.
3.3.2-2 Electrical Characteristic
Figure 3.13 plots the leakage current density as a function of electrical field up to 4.0 MV/cm. Before annealing, the leakage current density of BST-2 sample is observed to be much better than that of the BST-1 sample without a thin TaN film as a
diffusion barrier. The leakage current density of the BST-2 sample is 4.0x10-8 A/cm2 at 1 MV/cm lower than that for the BST-1 sample (8.0x10-5 A/cm2). After annealing at 400°C for 30 min in O2, the leakage current density of BST-3, not shown here, is caused by serious device failure due to Cu oxidation. However, the leakage current density of the BST-4 sample is 2.0x10-8 A/cm2 at 1 MV/cm, significantly lower than that before annealing at 400°C. Also, the breakdown field (Ebd) of the BST-4 sample is around 3.2 MV/cm (at 10-6 A/cm2), higher than those of the BST-1 and BST-2 samples, which are approximately 0.4 and 1.8 MV/cm. Numerous references have reported that post-annealing treatment in an oxygen-containing atmosphere yielded oxygen atoms and reduced the number of oxygen vacancies, ultimately enhancing the quality of BST films [11,19,20].
The mechanisms that govern the conduction of leakage current in the MIM capacitor may include Schottky emission, the Poole-Frenkel effect, electronic hopping conduction and tunneling [57-59]. Schottky emission is modeled as,
⎟⎠
where A is a constant; T represents the absolute temperature; q is the electronic charge; φ0 is the barrier height; kis the Boltzmann constant, and βs is given by
where ε0 is the permittivity of free space and ε is the high-frequency dielectric constant. Figure 3.14 plots ln(J) as a function of E1/2 in the Cu/TaN/BST/TaN/Cu structure (BST-4). The figure demonstrates that different conduction mechanisms dominate in different electric field regimes. Two linear regions are observed, and the gradient yields the corresponding effective dielectric constant in the electric field E<1.0 MV/cm. The figure reveals that the dominant conduction mechanism in the
BST-4 capacitor is Schottky emission in a low electric field, in which electrons from the cathode overcome the TaN/BST energy barrier before they are emitted. The leakage current density increases with the electric field, because when an electron enters the BST, it creates an image field that adds to or is subtracted from the barrier field, reducing the barrier height and increasing the current.
When electric field E>1.25 MV/cm, electrical conduction is governed by Poole-Frenkel emission, which is described by,
⎟⎠
Figure 3.15 plots ln(J/E) versus E1/2 in the BST-4 capacitor. The gradient in the linear region yields the corresponding effective dielectric constants in electric field E>1.25 MV/cm, which value is close to that obtained from C-V measurement. Hence, Poole-Frenkel emission is determined to dominate in a high electric field.
Poole-Frenkel emission is caused by the field-enhanced excitation of trapped electrons into the conduction band of the dielectric and its existence in capacitors establishes the presence of electron traps [60]. In electric fields of over 1.25 MV/cm, electrons in BST film traps absorb sufficient energy to be excited to the conduction band and Poole-Frenkel emission then dominates conduction.
The time to breakdown (tBD) of BST capacitors is measured by applying a voltage from 10.5 to 12.25 V, which corresponds to an electric field of 1.5 to 1.75 MV/cm in the BST-4 sample. Figure 3.16(a) plots the J –t curves of TDDB measurement. These data demonstrate the tBD is a function of the total number of carriers that pass though the films [61]. At the beginning of the stress test, the leakage current declined. The
decrease in the current in the first stage is believed to be caused by electron trapping in the dielectric films. After the majority of the electron traps were filled, the leakage current densities fell in the middle stage. However, the leakage current rapidly increased and fatal breakdown occurred after stress was applied for a longer period.
Figure 3.16(b) plots log(tBD) as a function of applied field for the BST-4 capacitor.
The sample has a longer lifetime than 10 year at 1.1 MV/cm. This result indicates the long-term intrinsic reliability of the BST capacitor in Gbit-scale DRAM applications.
3.3.2-3 Frequency Effect
Figure 3.17 plots the capacitance density and dissipation factor of the BST-2 and BTS-4 capacitors versus frequency. The results indicate that the capacitance density is maintained about 11.5 fF/μm2 at frequency from 1 kHz to 1 MHz and a high dielectric constant of 91 is obtained, which is higher than that of the Al2O3 and HfO2 MIM capacitors [62-64]. The large capacitance density at a low frequency of 1 kHz, in combination with the decline in capacitance density at a high frequency of 1 MHz, reveals that the mechanism may be defect-related because the slow traps may not have sufficient speed to follow the high-frequency signals [65]. Furthermore, a low dissipation factor under 0.03 is observed.
The linearity of capacitance against voltage is a significant parameter that depends on the material properties of the BTS film. The dependence of capacitance on voltage can compared with the voltage coefficients of capacitance (VCCs), which are given by
quadratic and linear voltage coefficients, respectively. The capacitances are measured at 1 kHz, 10 kHz, 100 kHz and 1 MHz. Figures 3.18(a) and (b) plot the obtained α and β values. For BST-4 capacitor, α falls from 357 ppm/V2 to 101 ppm/V2 for and β declines from 3637 ppm/V to 1347 ppm/V. Both α and β decrease as frequency increases, and relationship can be explained by the low time constant of traps within the BST layer. A comparison between the results for the BST MIM capacitor and those reported recently for the Ta2O5 capacitor shows overall superior VCCs that suggest that the MIM capacitor is very useful in Si RF applications [66,67]. Figure 3.19 plots the normalized capacitance of the BST-4 capacitor versus temperature at frequencies from 1 kHz to 1 MHz. The results demonstrate that the capacitance increases with the temperature. The temperature coefficients of capacitance (TCC) are 633 ppm/ºC, 433 ppm/ºC, and 274 ppm/ºC at frequencies of 1 kHz, 100 kHz and 1 MHz, respectively.