28nm Planar Process Flow: Integration Logic, Device Physics, and Module Dependencies

Process Map and Scope

The 28nm Planar process flow represents the final major technology generation built on classical planar metal-oxide-semiconductor field-effect transistor (MOSFET) architecture before the industry transitioned to three-dimensional fin-based structures [T1]. Understanding the 28nm fabrication process is essential because it sits at a critical inflection point: short-channel effects have become severe enough to demand sophisticated channel engineering, yet the device remains fundamentally two-dimensional, making it the most advanced node where planar integration principles fully apply [T1].

The integration objective of the 28nm Planar Flow is to construct a complete CMOS device stack—from substrate isolation through gate stack, source/drain engineering, and multi-level metallization—while maintaining electrostatic integrity under aggressively scaled channel dimensions [T1]. The planar process, invented in the early days of integrated circuit fabrication, relies on the unique ability of silicon dioxide to mask dopant diffusion and passivate junction surfaces, and this principle still underpins every module in the 28nm semiconductor process flow [T2].

At 28nm, the bulk CMOS approach faces rapidly increasing short-channel effects, including drain-induced barrier lowering (DIBL) and subthreshold slope degradation, which compel engineers to employ complex channel doping profiles, strain engineering, and advanced gate dielectric stacks [P1]. The 28nm Planar Flow process integration therefore must balance competing requirements: achieving high drive current while suppressing off-state leakage, maintaining threshold voltage stability while minimizing random dopant fluctuations, and preserving cost competitiveness while adding process complexity [T1].

Major Modules and Dependencies

Module Sequence and Ordering Logic

The 28nm Planar process flow follows a strict sequential architecture where each module creates the structural foundation for the next [T1]. The major modules, in order, are: shallow trench isolation (STI) formation, well and channel implantation, gate stack construction, extension and halo implantation, spacer formation, source/drain (S/D) epitaxy and implantation, silicidation, contact formation, and multi-level metal interconnect [T1].

The ordering is not arbitrary—it reflects deep physical dependencies (Engineering Practice). STI must precede well implantation because the isolation trenches define the active-area boundaries that mechanically and electrically confine subsequent dopant introduction [A1]. Well implantation must precede gate stack formation because the channel doping profile sets the threshold voltage that the gate stack must complement [P2]. Gate stack deposition must occur before extension implants because the gate electrode serves as a self-aligned mask for extension doping, ensuring the channel edge aligns precisely to the gate edge [T1].

STI and Active-Area Definition

The STI module is the first critical structural definition step (Engineering Practice). Trenches are etched into the silicon substrate and filled with insulating material to electrically isolate neighboring active regions [P2]. At 28nm, the high aspect ratio of these trenches makes void-free gap-fill challenging [P3]. Spin-on dielectric (SOD) materials such as perhydropolysilazane (PSZ) have been adopted because they offer superior flowability into narrow trenches compared to conventional chemical vapor deposition (CVD) oxides [P3]. The key physical principle is capillary-driven wetting: the solution must spread laterally and vertically into the trench, and this requires a highly hydrophilic liner surface to reduce the contact angle and enhance hydrogen-bond adsorption at the interface [P3].

After gap-fill and chemical-mechanical polishing (CMP), the STI surface is planarized, and the active-area pattern is fixed [P3]. This establishes the two-dimensional canvas upon which all subsequent transistor structures are built [T1]. For a deeper treatment of this module, see the 28nm Planar active-area definition process flow [T1].

Well and Channel Engineering

Well implantation creates the doping profiles that define NMOS and PMOS substrate types and set the punch-through stop (PTS) characteristics [P2]. At 28nm, maintaining a steep doping profile through subsequent thermal cycles is a central challenge [P2]. Transient enhanced diffusion (TED)—driven by silicon self-interstitials injected during ion implantation and oxidation—causes boron and phosphorus to diffuse upward toward the surface, degrading the channel profile and increasing leakage [P2].

One innovative approach to suppressing TED involves inserting a sub-monolayer of oxygen atoms during epitaxial silicon growth [P2]. This oxygen-insertion (OI) layer acts as an interstitial trap, absorbing and annihilating silicon self-interstitials, thereby blocking the interstitial-mediated diffusion pathway that boron and phosphorus rely on [P2]. The physics is rooted in point-defect equilibrium: dopant diffusion in silicon requires pairing with interstitials (e [T2].g., B–I pairs), so depleting interstitials at a specific depth fundamentally alters the diffusion driving force [P2]. More detail on this module is available in the 28nm Planar well and channel implant integration process flow [T1].

Gate Stack Construction

The gate stack module encompasses gate dielectric formation, metal gate deposition (for high-k metal gate variants), and gate electrode patterning [T1]. The gate dielectric must provide electrostatic control of the channel while minimizing gate leakage [P2]. As channel dimensions scale, the gate dielectric must thin correspondingly, but quantum-mechanical tunneling imposes a fundamental floor on usable oxide thickness [T1].

The gate electrode serves dual roles: it acts as the conductive control terminal and as a self-aligned mask for extension implantation [A1]. This self-alignment principle is central to planar process integration—it ensures that the source/drain extensions are automatically aligned to the gate edge without requiring additional lithographic registration [T1]. Further integration details are covered in the 28nm Planar gate stack integration process flow [T1].

Device Physics and Integration Logic

Electrostatic Integrity at Scaled Dimensions

The fundamental device physics challenge at 28nm is maintaining gate electrostatic control over a channel that is short enough for source and drain fields to penetrate deeply into the channel region [T1]. In a planar MOSFET, the gate exerts vertical electric field control, but the lateral fields from source and drain compete for influence over the channel potential [T1]. When the channel length becomes sufficiently short, the drain field can lower the source-side potential barrier, causing DIBL and increased off-state leakage [T3].

The integration logic of the 28nm Planar Flow addresses this through a combination of channel engineering and structural design [T1]. Halo implants—angled implantation of dopants near the channel edges—create locally higher doping concentrations that shield the channel from drain field penetration [P2]. The PTS implant, placed deeper in the channel, prevents sub-surface punch-through currents that bypass gate control [P2]. Together, these implants create a two-dimensional doping profile that counteracts short-channel effects [P2].

Strain Engineering for Carrier Mobility

At 28nm, strain engineering is integrated into the source/drain module to enhance carrier mobility [P1]. For PMOS, compressive strain is introduced through embedded silicon-germanium (SiGe) source/drain epitaxy, while for NMOS, tensile strain is applied through stress liner films or embedded silicon-carbon (SiC) structures [P2]. The physical principle is band structure modification: strain splits the degeneracy of valence band and conduction band valleys, reducing the carrier effective mass and increasing mobility [T3].

The integration logic requires that strain engineering modules be placed after gate stack formation and spacer definition, because the source/drain recess etch and epitaxial growth must not damage the gate structure [T1]. The spacers protect the gate sidewalls during these aggressive S/D engineering steps (Engineering Practice).

Doping and Fermi-Level Control

The ability to precisely control doping profiles is the foundation of all planar CMOS integration [T1]. Doping shifts the Fermi level position in the silicon band structure, determining whether the substrate behaves as n-type or p-type and setting the threshold voltage [T2]. At 28nm, the interaction between doping profiles and gate work function becomes critical: the threshold voltage is determined by the interplay of channel doping concentration, gate dielectric properties, and gate electrode work function [T2].

Ion implantation has replaced classical thermal diffusion as the dominant doping method because it independently controls dose and depth, enabling high-precision spatial selectivity at low thermal budgets [T1]. The Gaussian distribution of implanted dopants, characterized by a projected range and range straggle, governs the junction depth and concentration profile [T1]. Subsequent annealing activates the dopants—moving them into substitutional lattice sites where they become electrically active—while repairing implantation-induced lattice damage [P2].

Interface Risks and Failure Propagation

STI-Related Defect Propagation

Voids in STI gap-fill propagate consequences far downstream [P3]. If the SOD solution fails to completely fill the trench, voids remain trapped within the isolation structure [P3]. During subsequent CMP and etch-back steps, these voids can be exposed, creating deeper-than-expected STI recesses that distort the active-area topography [P3]. This topographic variation then affects gate patterning: non-uniform surface heights cause focus errors in lithography, leading to gate length variation across the wafer [T1]. The result is threshold voltage mismatch and drive current inconsistency—failure modes that are invisible at the STI module but manifest as parametric yield loss at final test [P3].

The root cause is often upstream: insufficient liner hydrophilicity or inadequate surface treatment before SOD coating leads to poor solution mobility and incomplete trench refill [P3]. This illustrates a central principle of process integration: the failure signature and its root cause are frequently separated by multiple modules (Engineering Practice).

TED-Induced Profile Degradation

Thermal processing after well implantation can degrade the carefully engineered doping profile through TED [P2]. Interstitials injected during oxidation steps drive boron and phosphorus diffusion, causing the near-surface channel concentration to increase while the deeper PTS peak broadens [P2]. The electrical consequences are severe: higher channel doping raises threshold voltage, degrades mobility through impurity scattering, and increases junction capacitance—all of which reduce drive current and switching speed [P2].

The failure propagation path is: oxidation step → interstitial injection → TED of channel dopants → profile broadening → threshold voltage shift and mobility degradation → parametric drift in device performance [P2]. This chain demonstrates why thermal budget management is a cross-module concern, not merely a property of any single annealing step (Engineering Practice).

Interface Roughening and Material Sensitivity

When germanium-containing layers are present—as in SiGe source/drain or strained channels—the interface becomes highly sensitive to plasma conditions [A2]. Aggressive oxygen plasma exposure causes excessive interfacial oxidation of germanium, forming volatile germanium oxide suboxide species that lead to volume expansion and surface protrusion ("hump" formation) [A2]. This roughening degrades the gate dielectric interface quality, increasing interface trap density and degrading carrier mobility [T1].

The integration risk is that this damage occurs during pad oxide deposition—a seemingly benign step—yet its effects propagate through gate stack formation and manifest as reliability failures in the final device [P2]. The mitigation strategy involves engineering the plasma power profile during oxide deposition: a lower-power initial layer suppresses germanium oxidation, while a higher-power second layer provides the density and etch resistance needed for subsequent patterning [A2].

Gate-Induced Drain Leakage

In aggressively scaled devices, the overlap region between the gate edge and the drain experiences high electric field concentration [P2]. If the gate electrode material has a single work function, band-to-band tunneling can occur at the drain edge, producing gate-induced drain leakage (GIDL) [A1]. One structural mitigation involves using composite conductive layers with different work functions in the word-line or gate stack, which redistributes the electric field and reduces the peak field intensity at the drain edge [A1]. This again illustrates how a downstream electrical failure (GIDL) is addressed by an upstream materials choice (work function engineering in the gate stack) [A1].

How to Study the Real Flow

Understanding the 28nm Planar process flow requires more than reading module descriptions in sequence—it demands engagement with the actual step-by-step construction process, where each step's inputs, outputs, and dependencies are made explicit [T1]. The interactive flow provides a structured way to trace how a silicon wafer transforms from a blank substrate into a fully functional CMOS device [T1].

To begin studying the complete 28nm Planar Flow, you can Open AA Step 1 in the interactive flow, which starts at the active-area definition module—the first structural pattern that defines where transistors will be built [T1]. Walking through each step in order reveals the dependency chains discussed above: how STI formation creates the boundaries for well implantation, how well profiles constrain gate stack design, and how each module's thermal and chemical inputs propagate forward into the final device characteristics [T1].

When studying the flow, pay particular attention to the transitions between modules (Engineering Practice). The most instructive moments are often at module boundaries—where the output of one step becomes the input of the next, and where interface quality, alignment tolerance, and thermal history collectively determine whether the integration succeeds or fails (Engineering Practice).

Related Learning Paths

The 28nm Planar process flow is best understood as a cluster of interconnected modules rather than a monolithic sequence [T1]. Engineers seeking depth should study each major module individually while maintaining awareness of the cross-module dependencies:

• Active-area definition and STI integration: The foundational isolation module that defines the two-dimensional device layout [T1]. Studying this module reveals how lithographic patterning, etch profile control, and gap-fill chemistry interact to create the structural canvas for all subsequent processing [T1].

• Well and channel implant integration: The doping engineering module that sets threshold voltage and short-channel immunity [T1]. This module connects ion implantation physics, TED suppression strategies, and thermal budget management into a coherent integration narrative [P2].

• Gate stack integration: The module where dielectric physics, work function engineering, and self-aligned patterning converge [T1]. Understanding this module is essential for grasping how the 28nm node achieved the electrostatic control needed for planar devices at scaled dimensions [T1].

Each of these modules can be studied as a standalone integration topic, but their interdependencies—thermal budget sharing, topographic inheritance, and defect propagation—form the connective tissue that makes the 28nm Planar Flow a unified process rather than a collection of independent steps [T1].

Future Outlook

The 28nm Planar process flow also served as a bridge to alternative architectures [T1]. The ultra-thin body and buried oxide (UTBB) fully depleted silicon-on-insulator (FD-SOI) approach reuses a substantial majority of the 28nm bulk front-end process and shares an identical back-end, while achieving fully depleted channel operation through geometric thinning rather than heavy doping [P1]. This approach eliminates random dopant fluctuations as a variability source and reintroduces body-biasing as an effective threshold voltage tuning tool, because the buried oxide provides a dielectric isolation layer that enables back-gate electrostatic modulation [P1].

The innovation here is process-design co-optimization: the FD-SOI variant does not require new manufacturing equipment, making it a cost-effective alternative path for extending planar CMOS beyond the conventional 28nm bulk limit [P1]. However, challenges remain in scaling to more advanced nodes, as ultra-thin silicon and buried oxide thickness control become increasingly stringent, and analog, input/output, and electrostatic discharge intellectual property require redesign [P1].

Research into point-defect engineering, such as oxygen-insertion silicon for TED suppression, continues to offer pathways for improving doping profile control in both planar and three-dimensional architectures [P2]. Meanwhile, advances in spin-on dielectric materials and surface treatments for gap-fill applications remain relevant for any technology node that requires void-free isolation in high aspect-ratio structures [P3]. These cross-cutting innovations underscore that the 28nm Planar process flow is not merely a historical artifact—it is a living integration platform whose principles continue to inform semiconductor manufacturing at and beyond the node where planar devices reached their zenith [T1].

---

References cited: [P1] UTBB FD-SOI process/design symbiosis (2013); [P2] Oxygen-insertion silicon for TED suppression (2017); [P3] PSZ-SOD gap-fill for 28nm STI (2015); [T1] Modern Semiconductor Devices—Device Fabrication Technology (2010); [T2] Silicon VLSI Technology (2000); [T3] Physics of Semiconductor Devices (2006); [A1] Semiconductor device with U-shaped word-line dielectric (2023); [A2] Germanium hump reduction via graded plasma power (2020).