Introduction
In modern semiconductor manufacturing, the transition from conventional oxide-based gates to advanced transistor architectures has necessitated revolutionary changes in front-end-of-the-line (FEOL) and middle-of-the-line (MOL) processing . Among these innovations, the replacement metal gate (RMG) scheme—often referred to as the gate-last approach—has emerged as a cornerstone for manufacturing high-performance logic devices . A pivotal process step that enables this integration scheme is poly open polish (POP) .
But what is poly open polish (Engineering Practice)? In a typical replacement metal gate flow, a temporary sacrificial polysilicon (poly-Si) gate is initially patterned to act as a placeholder . This dummy gate maintains the spatial layout of the transistor channel during high-temperature steps like source/drain dopant activation annealing , . After these high-temperature steps are complete, the dummy gate must be removed and replaced with the final high-k gate dielectric and metal work-function layers .
Before the sacrificial polysilicon can be selectively etched away, it must be exposed (Engineering Practice). Following dummy gate patterning, the device is covered with a contact etch stop layer (CESL) and a first interlayer dielectric (ILD0) layer . This dielectric overburden completely buries the gate structures (Engineering Practice). The process of removing this excess overburden to lay bare the top surface of the sacrificial gate is known as dummy gate planarization (Engineering Practice). This planarization sequence, which culminates in exposing the dummy polysilicon, is achieved through a multi-step chemical mechanical planarization process commonly known as ILD0 CMP . The final, most critical phase of this sequence—where the silicon nitride capping layers and dielectric residues are polished back precisely to reveal the clean, undamaged top of the polysilicon gate—is called the poly open polish (POP) process .
Without a highly uniform, low-defect, and precise poly open polish process, subsequent steps such as selective wet or dry etching of the dummy polysilicon would fail, leading to incomplete gate removal, metallic voids, and catastrophic yield loss . This article dives deep into the physical mechanisms, process principles, key engineering challenges, and technology evolution of this vital planarization process .
Physics & Mechanism
At its core, poly open polish is a chemical mechanical planarization (CMP) process that relies on the synergistic relationship between chemical surface modification and mechanical abrasion . The fundamental mechanical removal rate is traditionally described by Preston’s law, which states that the material removal rate is proportional to the contact pressure applied to the wafer and the relative velocity between the wafer and the polishing pad . However, at the nanoscale, Preston's law is modified by complex boundary layer lubrication, pad surface topography, and chemical reaction kinetics , .
The Chemical-Mechanical Synergistic Model
Pure mechanical polishing causes severe surface damage, such as deep fracturing and scratching, whereas pure chemical dissolution results in isotropic etching that destroys topography . The POP slurry is formulated to continuously react with the target surface, creating a thin, chemically modified passivation layer that is mechanically softer than the bulk material underneath . The abrasive particles suspended in the slurry then mechanically shear off this softened layer, exposing fresh material for subsequent chemical reaction .
During POP, three materials are simultaneously exposed to the polishing environment: silicon nitride ($Si_3N_4$, which forms gate spacers or capping layers), silicon dioxide ($SiO_2$, from the ILD0 layer), and polysilicon (the dummy gate itself) . The polishing chemistry must be designed with highly tunable selectivities to stop precisely when the polysilicon is fully cleared .
Chemical Hydrolysis of Silicon Nitride
The removal mechanism of silicon nitride in POP slurries relies heavily on hydrolysis reactions occurring at the solid-liquid interface in an aqueous environment . At elevated local temperatures and pressures generated during pad-wafer contact, water molecules react with the silicon nitride surface to form a hydrated silicate-like boundary layer , . This hydrolysis can be chemically formulated through sequential steps:
$$\text{Si}_3\text{N}_4 + \text{H}_2\text{O} \rightarrow \text{Si}_2\text{NH} + \text{Si-OH}$$
$$\text{Si}_2\text{NH} + \text{H}_2\text{O} \rightarrow \text{SiNH}_2 + \text{Si-OH}$$
This reaction sequence ultimately converts the rigid, chemically inert covalent silicon nitride network into a softened, hydrated silicon dioxide/silanol ($\text{Si-OH}$) surface layer (Engineering Practice). This modified boundary layer exhibits much lower mechanical shear strength than the bulk nitride, allowing colloidal silica or ceria abrasive particles to readily sweep it away , .
Chemical Modulation and Selectivity Additives
To control the polishing rate ratios between nitride, oxide, and polysilicon, chemical additives are introduced into the POP slurry . For instance, organophosphonic acid-based additives are frequently utilized . These molecules act as chemical bridges between the abrasive nanoparticles, the polyurethane polishing pad, and the wafer surface . By adjusting the slurry pH and additive concentrations, engineers can manipulate the electrostatic forces (zeta potential) at the interfaces . This can suppress the polishing rate of the oxide and polysilicon phases while accelerating the hydrolysis and mechanical removal of the silicon nitride capping layers, thereby achieving a highly selective, self-limiting polish that stops cleanly on the dummy gate .
Process Principles
To achieve a successful poly open polish process, engineers must balance multiple interacting process parameters . These variables directionally control the material removal rates, the cross-wafer uniformity, and the overall planarization efficiency .
Directional Parameter Interactions
- Downforce (Polishing Pressure): Increasing the downward pneumatic force applied to the wafer carrier increases the mechanical friction at the pad-wafer interface . This directionally increases the removal rates of all exposed materials (nitride, oxide, and polysilicon) but also elevates the risk of mechanical defects, such as surface micro-scratches, and exacerbates dishing in wide patterned features , .
- Platen and Carrier Rotational Speeds: Higher rotational speeds increase the relative linear velocity, which generally enhances the removal rate in accordance with Prestonian behavior . However, excessively high speeds can transition the tribological regime from boundary lubrication to hydrodynamic lubrication, where a thick fluid film of slurry lifts the wafer off the pad, reducing planarization efficiency and causing edge-roll-off profile non-uniformities .
- Slurry Flow Rate: The rate at which fresh slurry is dispensed onto the platen determines the chemical replenishment rate and assists in thermal dissipation (Engineering Practice). Low flow rates can lead to slurry starvation, resulting in localized friction spikes and temperature fluctuations that cause non-uniform polishing rates across the wafer surface .
- Slurry pH and Buffering: The chemical reactivity of the slurry is highly sensitive to pH drifts . A stable pH buffer is required to maintain the ionization state of both the abrasive particles and the target surfaces, ensuring that the electrostatic repulsion or attraction remains constant throughout the tool's lifetime .
+-----------------------+ +-------------------------+ +-------------------------+
| Increase Downforce | ---> | Higher Friction/Removal | ---> | Risk of Micro-scratches |
+-----------------------+ +-------------------------+ +-------------------------+
|
v
+-----------------------+ +-------------------------+ +-------------------------+
| High Rotational Speed | ---> | Hydrodynamic Lift-off | ---> | Reduced Planarization |
+-----------------------+ +-------------------------+ +-------------------------+
Surface Engineering for Selectivity Control
In addition to chemical tuning of the slurry, advanced process integration schemes sometimes employ surface modification techniques to alter the polishing behavior of the films . For example, prior to the POP step, the pre-metal dielectric (PMD) oxide layer can be modified via ion implantation . Introducing high-dose carbon ion implantation into the surface of the silicon dioxide layer creates a chemically resistant, carbon-doped oxide network . This modification suppresses the oxide's hydration kinetics, lowering its removal rate by up to forty percent during the subsequent POP step, thereby providing an elegant physical method to enhance selectivity without altering the slurry chemistry .
Challenges & Failure Modes
The poly open polish process operates under extremely narrow margins . Even minor process deviations can lead to yield-killing electrical failures or severe topography variation (Engineering Practice).
Dishing and Erosion
One of the most persistent failure modes in any CMP step is pattern-density-dependent non-uniformity, manifested as dishing and erosion . Dishing occurs when the softer material within a trench (such as the oxide in wide trench regions) is polished at a faster rate than the surrounding harder stopper material (such as the silicon nitride spacers) . Erosion refers to the localized thinning of both the dielectric and the gate structures in high-density pattern arrays due to localized pressure concentration .
Pre-Metal Dielectric Recess and Photolithography Misalignment
When the removal rate of the $SiO_2$ in the ILD0 layer is significantly higher than that of the silicon nitride and polysilicon, a severe PMD recess around the gate occurs . This recess creates localized step-height steps at the gate edges . During subsequent contact hole patterning steps, these severe topographical steps cause destructive interference of the exposure beam during photolithography, leading to severe photo misalignment and pattern distortion .
Mechanical Scratching
Scratch formation is highly detrimental during the final stages of POP . Micro-scratches are typically caused by mechanical anomalies, such as the agglomeration of slurry abrasive nanoparticles, pad conditioning disc wear debris, or dried slurry flakes falling back onto the polishing pad . These large, anomalous particles undergo plowing and cutting mechanisms on the softened oxide and nitride surfaces, leaving behind chatter-type or continuous line scratches . In the ILD0 CMP sequence, such scratches can act as parasitic channels that trap subsequently deposited barrier metals, leading to gate-to-contact shorts and reliability failures .
Failure of Oxide CMP Stop on CESL
During the initial phase of the planarization sequence, the target is to perform an oxide CMP stop on CESL . The contact etch stop layer, typically made of high-density silicon nitride, is deposited conformally over the dummy gates and source/drain regions . It serves as an essential physical barrier that protects the delicate source/drain regions underneath . If the CMP slurry selectivity is poorly controlled, or if there is severe across-wafer thickness variation, the polishing process can break through the thin CESL, leading to severe erosion of the source/drain epitaxial structures, contact resistance variations, and complete device failure .
Technology Node Evolution
The requirements and complexity of the poly open polish process have changed dramatically as transistor architectures scaled down over successive technology nodes .
The 28nm Node: Planar Transistor Era
In the planar 28nm Planar Flow, the gate structures were two-dimensional, meaning that the topography variation before planarization was relatively moderate (Engineering Practice). The POP process primarily had to manage simple, flat, uniform line patterns, with the primary goal of exposing the polysilicon dummy gates across the wafer while keeping the gate height uniform to control short-channel effects .
The 14nm to 7nm Nodes: FinFET Architectures
With the introduction of the fin field effect transistor (FinFET) architecture at the 14nm FinFET and 7nm FinFET nodes, the pre-planarization topography became highly three-dimensional . The presence of underlying silicon fins created complex pattern-density variations and step heights that made global planarization exceptionally challenging . During POP, the slurry had to planarize the overburden across both dense fin arrays and wide isolation regions without causing spacer erosion or gate height variation . Precise control over slurry chemistry, high-selectivity stopping mechanisms, and advanced multi-zone carrier heads became mandatory to achieve the sub-nanometer thickness control required at these nodes , .
Related Processes
To understand the role of poly open polish, it must be viewed within the context of its adjacent upstream and downstream processes (Engineering Practice).
+------------------------------------+
| Upstream: |
| - Dummy Gate Patterning |
| - CESL & ILD0 Deposition |
+------------------------------------+
|
v
+------------------------------------+
| POLY OPEN POLISH (POP) |
| - Exposes Sacrificial Polysilicon |
+------------------------------------+
|
v
+------------------------------------+
| Downstream: |
| - Selective Wet/Dry Dummy Etching |
| - High-K Dielectric Deposition |
| - Work Function Metal Fill & CMP |
+------------------------------------+
Upstream Steps
Before the POP step can be executed, several preparatory layers are deposited . After the dummy gate is defined and patterned, the source/drain junctions are formed by ion implantation and rapid thermal processing . Subsequently, a thin, conformal contact etch stop layer (CESL), typically composed of silicon nitride, is deposited , . This is followed by the deposition of a thick interlayer dielectric (ILD0) layer—often silicon oxide—via flowable chemical vapor deposition (FCVD) or high-density plasma (HDP) deposition to completely fill the gaps between the gate structures , .
Downstream Steps
Once the poly open polish step successfully exposes the top of the dummy polysilicon gate, the wafer proceeds to downstream replacement gate modules . First, the exposed polysilicon is selectively removed using hot ammonium hydroxide wet etching or highly selective isotropic dry etching, leaving high-aspect-ratio gate trenches , . Next, atomic layer deposition (ALD) is used to line the trench with a thin high-k dielectric layer (such as hafnium oxide) and metal work-function setting layers . Finally, a low-resistivity gate fill metal (typically aluminum or tungsten) is deposited and polished back using metal CMP, completing the replacement metal gate flow .
Future Outlook
As the semiconductor industry marches beyond the 3nm node, conventional FinFETs are being replaced by Gate-All-Around (GAA) nanosheet and forksheet architectures . This transition imposes even more stringent requirements on the poly open polish process (Engineering Practice).
GAA Nanosheet Integration Challenges
In GAA nanosheet configurations, the dummy gate is wrapped completely around multiple vertically stacked silicon channels . The initial aspect ratio of the dummy gate is higher, and the vertical spacer structure is far more delicate . The POP process must stop with absolute, atomic-scale precision on the dummy gate without imparting any mechanical stress that could cause the ultra-thin nanosheets to collapse or distort .
To meet these extreme demands, the industry is researching highly specialized, chemical-dominant slurries that utilize chemical passivation agents to virtually eliminate mechanical force during the final polish phase . Additionally, combining molecular-level surface modification with highly selective isotropic chemical dry etching holds promise as a hybrid alternative to traditional CMP, steering the future of poly open polish toward atomic-layer-precision manufacturing .