Introduction
The self-aligned silicide (salicide) process has been a cornerstone of silicon technology since its development in the early 1990s . The primary goal of this process is to dramatically reduce the sheet and contact resistance of the active source, drain, and gate regions of metal-oxide-semiconductor field-effect transistors (MOSFETs) , . By reacting a transition metal directly with the exposed silicon areas, a highly conductive metal silicide layer is formed without the need for additional lithography masking steps , . However, as integrated circuit complexity grew, engineers realized that certain regions of the silicon substrate must be protected from this metal-silicon reaction to maintain high electrical resistance or prevent junction leakage . This requirement led to the development of the salicide block (SAB) process , (Engineering Practice).
A salicide block (SAB) is a patterned dielectric thin film that acts as a physical and chemical barrier, preventing transition metals from contacting and reacting with the underlying silicon during subsequent thermal processing , (Engineering Practice). There are several critical applications where an SAB layer is indispensable (Engineering Practice). First, electrostatic discharge (ESD) protection resistors and input protection diodes require high internal resistance to limit current flow during high-voltage events, which would be destroyed if the silicon surface were fully silicided , (Engineering Practice). Second, precise analog resistors fabricated on active silicon or polysilicon layers rely on exact sheet resistance values that cannot be achieved if high-conductivity silicides are present , (Engineering Practice). Finally, emerging non-volatile memory architectures, such as embedded resistive random-access memory (RRAM) structures, integrate selection transistors where the drain must remain unsilicided to allow direct, low-leakage electrical contact to the bottom electrode of the memory cell . If the SAB layer is omitted or fails, the resulting unblocked metal-silicon reaction dramatically lowers the localized resistivity, leading to device failures, excessive leakage currents, or non-functional circuit blocks , , (Engineering Practice).
Physics and Chemical Mechanisms of Silicidation and Blocking
Understanding how an SAB layer functions requires a deep dive into the physical chemistry of solid-state reactions and thin-film diffusion , . When a thin transition metal film (such as nickel, cobalt, or titanium) is deposited onto clean silicon and subjected to thermal energy, solid-state diffusion occurs across the metal-silicon interface , . This reaction is driven thermodynamically by the reduction of interfacial free energy and chemical potential differences between the metal and the semiconductor .
As the temperature increases during a thermal process, different intermetallic phases nucleate and grow sequentially , . In the case of nickel silicide, the process begins at lower temperatures with metal-rich phases such as $Ni_2Si$, which possess relatively high resistivity . As diffusion continues at intermediate temperatures, the system transitions to the low-resistivity monosilicide phase ($NiSi$), which is the desired phase for high-performance contacts . If the thermal budget is increased further, silicon diffusion dominates, driving the system to form the thermodynamically stable but higher-resistivity disilicide phase ($NiSi_2$) . For titanium-based silicidation, the reaction progresses from the high-resistivity $C49$ phase to the thermodynamically stable, low-resistivity $C54$ phase .
The primary physical mechanism of the SAB layer is to serve as an amorphous, high-density diffusion barrier that completely interrupts this chemical pathway (Engineering Practice). The atomic structure of the SAB dielectric—typically silicon dioxide ($SiO_2$) or silicon nitride ($Si_3N_4$)—possesses strong covalent bonds that do not easily break down in the presence of transition metals at silicidation temperatures . Consequently, the metal atoms cannot diffuse downward into the silicon, and the silicon atoms cannot diffuse upward into the metal, effectively halting any chemical reaction , , (Engineering Practice).
From a device physics standpoint, the presence or absence of the silicide layer alters the electronic band structure at the contact interface , . According to semiconductor physics, electrons in a perfect crystal lattice move within periodic potentials, a phenomenon described by Bloch's theorem :
$$\psi_{n\mathbf{k}}(\mathbf{r}) = e^{j\mathbf{k}\cdot\mathbf{r}} u_{n\mathbf{k}}(\mathbf{r})$$
At a metal-semiconductor interface, this periodic potential is broken, giving rise to metal-induced gap states and establishing a Schottky barrier . When a transition metal directly contacts doped silicon, it forms a silicide with a specific work function, which drastically lowers the Schottky barrier height and contact resistance ($R_c$) for majority carriers , . Conversely, in regions blocked by the SAB layer, the semiconductor retains its raw interface characteristics, and the carrier concentration remains governed by the standard Fermi-Dirac distribution :
$$f(E) = \frac{1}{1+\exp\left(\frac{E-E_F}{kT}\right)}$$
Without the silicide, the interface exhibits a significantly higher sheet resistance ($R_{sh}$) and contact resistance, preserving the high-impedance state necessary for analog resistors and ESD protection circuits , , (Engineering Practice).
Process Principles and Parameter Interactions
The fabrication of a salicide block involves a sequence of dielectric deposition, lithographic patterning, high-selectivity etching, and post-etch cleaning , (Engineering Practice). The directional effects of various process parameters during these steps directly dictate the final electrical and physical outcomes of the device (Engineering Practice).
Dielectric Deposition Kinetics
The SAB layer is typically deposited using chemical vapor deposition (CVD) or atomic layer deposition (ALD) processes , , . In advanced nodes, plasma-enhanced atomic layer deposition (PEALD) is favored because it separates surface-limited chemical reactions into sequential precursor adsorption and plasma-assisted oxidation steps . This separation enables the growth of high-density, conformal films at low thermal budgets .
The physical properties of the deposited SAB film are highly sensitive to deposition parameters such as plasma power, exposure time, and temperature . When the process temperature or plasma power is decreased, the thermal activation energy of the reaction is reduced . This leads to incomplete ligand removal from the metal-organic precursors, which artificially increases the growth-per-cycle (GPC) . However, these low-temperature films are chemically porous and contain high concentrations of carbon or hydrogen impurities, which degrade their performance as diffusion barriers , . Conversely, increasing the plasma power or exposure time provides the necessary kinetic energy to densify the film, improving its barrier efficiency, but prolonged plasma exposure can induce sub-surface defects and ion-bombardment damage on the silicon substrate .
Patterning and Etching Dynamics
Once the SAB dielectric is deposited, photolithography defines the regions where the block layer must be retained , (Engineering Practice). Subsequently, a dry etching process removes the unmasked dielectric from the active transistor areas where silicide is required , (Engineering Practice). The etching chemistry must exhibit extremely high selectivity toward the SAB dielectric relative to the underlying silicon substrate and the gate spacers , (Engineering Practice). If the etch selectivity is poor, the dry etching plasma can recess the silicon source/drain regions, leading to increased parasitic spreading resistance ($R_{sp}$) and threshold voltage variations , (Engineering Practice).
Following dry etching, wet chemical cleaning is performed to remove halogenated polymers and post-etch residues . The dissolution rate of these residues in fluorinated aqueous cleaners is highly temperature-dependent and follows the classic Arrhenius equation :
$$r = k \exp\left(-\frac{E_a}{RT}\right)$$
If the wet cleaning temperature is too low, residue removal is incomplete, which prevents the deposited metal from contacting the silicon during subsequent steps and leads to open contacts (Engineering Practice). However, if the temperature is excessively high, the cleaning chemistries can corrode the remaining SAB mask or erode the isolation dielectrics , (Engineering Practice).
SAB Deposition (CVD/ALD)
│
▼
Photolithographic Masking
│
▼
Dry Etching (Open Active Areas)
│
▼
Post-Etch Clean (Wet Chemical)
│
▼
Metal Sputtering (Ni/Co/Ti)
│
▼
Rapid Thermal Anneal (RTA)
│
▼
Selective Metal Wet Etch (Unreacted)
Challenges & Failure Modes
Integrating a salicide block into high-density semiconductor fabrication flows introduces several critical physical and chemical failure modes that must be carefully managed .
Blistering and Delamination
One of the most common failure modes of advanced SAB films, particularly those deposited via ALD or PEALD, is film blistering and subsequent delamination . During the deposition of materials like aluminum oxide ($Al_2O_3$) or silicon dioxide ($SiO_2$) using water or ozone-based precursors, significant amounts of hydrogen- and water-related species can become trapped within the bulk of the film or at the interface . When the wafer is subsequently subjected to high-temperature rapid thermal annealing (RTA) to drive the silicidation reaction, these trapped volatile species outgas and diffuse toward the dielectric-silicon interface , , . The accumulated gas builds up localized pressure, leading to mechanical delamination and blistering . Once a blister forms, the metal deposited on top can seep underneath, causing unwanted localized silicide reactions and catastrophic circuit shorts .
Silicide Spiking and Pinhole Defects
If the SAB dielectric layer contains structural defects, pinholes, or micro-cracks—often caused by poor step coverage over tall gate structures—it fails as a diffusion barrier . During the RTA step, metal atoms will rapidly diffuse through these localized weak points , , (Engineering Practice). This diffusion results in localized silicide "spikes" that penetrate deep into the silicon substrate , (Engineering Practice). If these spikes reach the underlying p-n junction depletion region, they create a high-field leakage path, leading to massive junction leakage current and potential device failure , (Engineering Practice).
Phase Degradation and Agglomeration
Nickel-based silicides are highly sensitive to thermal budgets . While the monosilicide phase ($NiSi$) exhibits low resistivity and is stable at intermediate temperatures, exposure to elevated temperatures during subsequent backend dielectric depositions can cause the phase to transition into the higher-resistivity $NiSi_2$ phase . Furthermore, thin silicide films tend to undergo morphological agglomeration at high temperatures, where the continuous silicide film breaks up into isolated islands due to surface tension minimization, resulting in a dramatic increase in contact resistance , .
Technology Node Evolution
The implementation of the salicide block has undergone significant evolutionary shifts as the industry scaled from planar transistors down to complex 3D architectures (Engineering Practice).
| Technology Node | Device Architecture | SAB Material & Deposition | Primary Integration Challenges |
|---|---|---|---|
| 28nm | Planar High-K Metal Gate | PECVD $SiO_2$ / $Si_3N_4$ | Planar defect density, selectivity of dry etch over active silicon |
| 14nm | 3D FinFET | High-Conformality PEALD $SiO_2$ | 3D fin sidewall coverage, shadowing effects, spacer erosion |
| 7nm and beyond | Advanced FinFET / GAA | Selective ALD / Ultra-thin PEALD | High aspect ratio voids, thermal budget compatibility with HKMG |
At the 28nm planar node, the SAB process was relatively straightforward . The topography was flat, allowing standard plasma-enhanced chemical vapor deposition (PECVD) to deposit a uniform layer of silicon dioxide or silicon nitride, which was then patterned using traditional optical lithography and dry etching .
However, the transition to the 14nm FinFET node and 7nm FinFET node completely redefined the integration landscape (Engineering Practice). The three-dimensional vertical fin geometry introduced severe aspect ratio challenges (Engineering Practice). Standard PECVD techniques suffer from poor step coverage, resulting in much thinner SAB films on the vertical sidewalls of the fins compared to the flat top regions (Engineering Practice). This non-uniformity led to barrier breakdown on the fin sidewalls during RTA, causing severe silicide spiking (Engineering Practice).
To overcome this, the industry transitioned to highly conformal PEALD processes , . PEALD allowed atomic-scale control of the film thickness, ensuring that the vertical fin sidewalls received the exact same barrier thickness as the horizontal surfaces , . Furthermore, the introduction of high-k metal gate (HKMG) technology limited the allowable thermal budget, as excessive temperatures could degrade the delicate gate dielectric stack . Consequently, low-temperature PEALD processes had to be optimized to deposit high-density, impurity-free SAB films at temperatures that would not compromise the HKMG integrity , .
Related Processes
The salicide block does not exist in isolation; its success depends heavily on the optimization of several adjacent process steps (Engineering Practice).
Ion Implantation
Before the SAB layer is even deposited, the silicon substrate undergoes heavy ion implantation to dope the source and drain regions , . The concentration and species of these dopants (such as arsenic, phosphorus, or boron) directly affect the subsequent silicidation kinetics . High dopant concentrations can retard the diffusion of metal atoms, requiring careful optimization of the RTA thermal budget to ensure complete silicide formation in unblocked regions without causing phase degradation in sensitive areas , .
Gate Patterning and Spacer Formation
The physical boundaries of the silicide are defined by the gate spacers, which are formed immediately after gate patterning . The SAB layer must align perfectly with these spacers . Any misalignment or poor adhesion between the SAB dielectric and the spacer material can create microscopic gaps, allowing metal atoms to seep in and form a short-circuit path between the gate and the source/drain contacts , (Engineering Practice).
Metal Deposition and Rapid Thermal Annealing
Following SAB patterning, transition metals are sputtered onto the wafer, followed by RTA to drive the solid-state reaction , . The RTA process must be finely tuned; too low of a thermal budget results in incomplete silicidation of active areas, while too high of a thermal budget can lead to silicide diffusion beneath the SAB mask, causing a phenomenon known as "silicide encroachment" , , (Engineering Practice).
Chemical Mechanical Planarization
After the unreacted metal is selectively stripped, an inter-layer dielectric is deposited and planarized using chemical mechanical planarization (CMP) to prepare the surface for contact plug formation (Engineering Practice). The SAB layer must remain mechanically stable and adhere firmly to the substrate to withstand the high shear forces exerted during the CMP process (Engineering Practice).
Future Outlook
As the semiconductor industry marches beyond the 3nm node toward nanosheet gate-all-around (GAA) transistors and complementary FET (CFET) architectures, conventional SAB lithography and etching schemes are approaching their physical limits (Engineering Practice). The extremely tight pitches of these advanced devices leave virtually no margin for overlay errors during lithography (Engineering Practice).
To address these scaling challenges, researchers are actively developing area-selective atomic layer deposition (AS-ALD) techniques , . Instead of depositing a continuous SAB film and patterning it via lithography and etching, AS-ALD leverages chemical inhibitors to selectively deposit the block dielectric only on specific material surfaces (such as the gate spacer) while leaving the active silicon regions completely bare , . This bottom-up approach completely eliminates the need for critical alignment lithography and dry etching steps, providing a damage-free, self-aligned blocking mechanism (Engineering Practice). Additionally, as novel embedded memories like RRAM continue to integrate into advanced logic platforms, the development of ultra-thin, high-density, low-temperature SAB layers will remain a vital research front to ensure high-yield, low-power system-on-chip solutions , .