Yong-Jin Seo

An optimization of tungsten plug chemical mechanical polishing (CMP) using different consumables

The continuous shrinkage of integrated circuit devices provides the advantage of having more electrical functions on the smaller chips. However, it has posed severe challenges to the manufacturing of these products. In patterning steps, the depth of focus becomes small, due to the nature of patterning small features and is aggravated by the presence of step height differences in the exposure field. When the aspect ratio of holes increases, aluminum, which has been the choice of metal to fill holes, no longer effectively fills small holes. On some occasions, discontinuous aluminum exists in the holes, resulting from poor step coverage. Tungsten (W), on the other hand, can satisfactorily fill small holes by chemical vapor deposition and thus fills a need to replace aluminum as an electrical connector plug for layers separated by this plug. Chemical mechanical polish (CMP) has been used recently to planarize substrates and improve the depth of focus, as in the case of hole patterning. It is also used to prepare the metal plug effectively. CMP processing involves various parameters, such as machine configuration, slurry chemistry and formulation, polishing pad configuration, pad hardness, etc. Although the mechanism is not fully understood, these parameters are known to significantly influence the degree of planarization, contamination, defects, etc. In this paper we explore the effects of consumables used in the W-CMP process on the control of several problems, such as recessed plug, dishing and oxide erosion, etc. Based on our findings, we implement a two-step W-CMP process to improve the qualities of the polished wafers. The first step involves a proper combination of slurry and pad to polish most of the bulk tungsten with a high polish rate. The second step uses the same pad, slurry and a proprietary additive to slowly polish the remaining tungsten. In this way, we are able to obtain a workable W-CMP process.

Correlation analysis between pattern and non-pattern wafer for characterization of shallow trench isolation-chemical-mechanical polishing (STI-CMP) process

This study is about control of oxide removal amounts on the shallow trench isolation (STI) patterned wafers using removal rate and thickness of blanket (non-patterned) wafers. At first, the removal properties of plasma enhanced tetra-ethyl ortho-silicate (PETEOS) blanket wafers was investigated, and then it was compared with the removal properties and the planarization (step height) as a function of polishing time of the specific STI patterned wafers. We found that there is a relationship between the amount of oxide removal by blanket and patterned wafers. We analyzed this relationship, and the post-CMP thickness of patterned wafers could be controlled by removal rate and removal target thickness of blanket wafers. As the result of correlation analysis, we confirmed that there was the strong correlation between patterned and blanket wafers (correlation factor: 0.7109). So, we could confirm the repeatability as applying to STI CMP process from the linear formula obtained.

Design of experimental optimization for ULSI CMP process applications

As the device geometry shrinks to the deep sub-micron region, chemical mechanical polishing (CMP) planarization has become a more essential technique in advanced ULSI process. In particular, the complete global planarization of multilevel interconnections can be achieved only with the CMP process. However, it still has various problems due to the CMP equipment. In particular, among the CMP components, process variables are very important parameters in determining the removal rate and non-uniformity. In this paper, we studied the design of experiments (DOE) method in order to obtain optimized CMP equipment variables. Various process parameters, such as table and head speed, slurry flow rate and down force, have been investigated from the viewpoint of high removal rate and low non-uniformity. Through the above DOE results, we can determine the optimal CMP process parameters.

Reduction of process defects using a modified set-up for chemical mechanical polishing equipment

The chemical mechanical polishing (CMP) process is widely used for global planarization of inter-metal dielectric (IMD) and inter-layer dielectric (ILD) layers for deep sub-micron technology. However, as the IMD and ILD layers get thinner, defects such as micro-scratches can lead to severe circuit failure, which affects yield. In this paper, for the improvement of the CMP process, deionized water (DIW) pressure, purified nitrogen (PN2) gas, point of use slurry filter, and high spray bar were installed. Our experimental results show that DIW pressure and PN2 gas factors are not related to removal rate, but they are closely related to the edge hot spot on a patterned wafer. Also, the filter installation in a CMP instrument could remarkably reduce the defects after CMP processing, it is shown that the slurry filter plays an important role in the determination of pad lifetime. However, since the slurry filter cannot completely prevent defect-causing particles from entering the system, the installation of the high spray bar of deionized water with high pressure is proposed.

Motor-current-based real-time end point detection of shallow-trench-isolation chemical mechanical polishing process using high-selectivity slurry

We have studied the chemical mechanical polishing (CMP) of a shallow-trench-isolation (STI) structure in 0.18 µm semiconductor device fabrication. CMP is applied for the STI structure with and without a reverse moat pattern. The real-time end point detection (EPD) method based on the motor current (MC) signal is tested, and the factors affecting the conventional motor current detection method are investigated. The results showed that the EPD method could not be applied to STI-CMP with a reverse moat etch process due mainly to the pad conditioning effect and the characteristics of an open nitride structure. In the case of direct STI-CMP without a reverse moat etch process, global planarization with low defect levels can be achieved using a high-selectivity slurry (HSS) with a high selectivity for SiO2 and Si3N4, and the EPD method shows acceptable and reproducible results.

Optimization of post-CMP cleaning process for elimination of CMP slurry-induced metallic contaminations

We have invetigated the slurry-induced metallic contaminations of undoped and doped silicate oxide surface during the post-chemical mechanical polishing (CMP) cleaning process. The metallic contaminations by CMP slurry were evaluated in the four different oxide films, such as plasma enhanced tetra-ethyl-ortho-silicate glass (PE-TEOS), O3-boro-phospho-silicate glass (O3-BPSG), PE-BPSG, and phospho-silicate glass (PSG). Prior to entering the post-CMP cleaner, all films were polished with KOH-based slurry. The total X-ray fluorescence (TXRF) measurements showed that all oxide surfaces are heavily contaminated by potassium and calcium during polishing, which is due to the CMP slurry. The polished O3 BPSG films presented higher potassium and calcium contaminations compared to PE-TEOS film because of a mobile ion gettering ability of phosphorus. For PSG oxides, the slurry-induced mobile ion contamination increased with an increase of phosphorus content. In addition, the polishing removal rate of PSG oxides had a linear relationship with phosphorus content.

Removal characteristics of hillock on SnO2 thin film by chemical mechanical polishing process

SnO2 is one of the most suitable materials for gas sensors. The microstructure and surface morphology of SnO2 films must be controlled because the electrical and optical properties of SnO2 films depend on these characteristics. We investigated the effects of chemical mechanical polishing (CMP) on the variation of morphology of SnO2 films prepared by rf sputtering system. The commercially developed ceria-based oxide slurry, silica-based oxide slurry, and alumina-based tungsten slurry were used as CMP slurry. Nonuniformities of all slurries coincided with stability standards of less than 5%. Silica slurry had the highest removal rate among three different slurries. In addition, the particle size analysis showed that silica slurry had an abrasive with the largest average particle size of the three. Based on the atomic force microscopy analysis of thin film topographies and root mean square values, silica slurry has excellent properties that allow the application of SnO2 thin films as gas sensor materials.

Advantages of point of use (POU) slurry filter and high spray method for reduction of CMP process defects

As integrated circuit devices shrink to smaller dimensions, a chemical mechanical polishing (CMP) process was required for the global planarization of the inter-metal dielectric (IMD) layer with free-defects. However, as the IMD layer gets thinner, micro-scratches are becoming a major defect. Micro-scratches are generated by agglomerated slurry, solidified and attached slurry in the pipeline of the slurry supply system. In order to prevent the agglomerated slurry particles from slurry inflow, we installed a 0.5 µm point of use (POU) filter, which is a depth-type filter and has 80% filtering efficiency for 1.0 µm size particles. Also a high spray bar of de-ionized water with high pressure was installed in the CMP equipment. Our experimental results showed that it is impossible to completely prevent defect-causing particles through the depth-type filter. Thus, we suggest that it is necessary to optimize the slurry flow rate and to install a high spray bar of de-ionized water (DIW) with high pressure, in order to overcome the weak points of a POU depth-type filter.

Signal analysis of the end point detection method based on motor current

We first studied the factors affecting the motor current (MC) signal. It was found to be strongly affected by systematic hardware noise depending on the polishing recipe, such as pad conditioning and arm oscillation. We then studied the end point detection (EPD) for the chemical mechanical polishing (CMP) process of the shallow trench isolation (STI) structure with a reverse moat pattern. The MC signal showed a high amplitude peak in the fore part caused by the reverse moat pattern. We also found that the end point (EP) could not be detected properly and reproducibly due to the pad conditioning effect, especially when a conventional low selectivity slurry was used. Even when there was no pad conditioning effect, the EPD method could not be applied, since the measured end point was always the same due to the characteristics of the open nitride structure.

In-situ end point detection of the STI-CMP process using a high selectivity slurry

We studied the end point detection (EPD) for the direct CMP of the STI structure without the reverse moat etch process. In this case, we applied a high selectivity slurry (HSS) that improves the silicon oxide removal rate and maximizes the oxide-to-nitride selectivity. Quite acceptable and reproducible EPD results could be obtained, and the wafer-to-wafer thickness variation was significantly reduced when compared with the conventional predetermined polishing time method without EPD. Therefore, it was possible to achieve a global planarization without the complicated reverse moat process. Furthermore, the problems caused by nitride residues in the active regions could be reduced by overpolishing of SiO2, since Si3N4 was hardly removed. As a result, the STI CMP process could be dramatically simplified, the defect level reduced, and therefore the throughput, yield and stability of semiconductor device fabrication could be greatly improved.