Spin-on glass is a liquid-applied thin film technology used in semiconductor manufacturing to deposit layers of glass-like material onto silicon wafers. Instead of growing or sputtering a film in a vacuum chamber, a technician dispenses a liquid solution onto a spinning wafer and lets centrifugal force spread it into a uniform coating, typically between 30 and 200 nanometers thick. After spinning, the film is baked to drive off solvents and transform it into a solid silica-based layer. The technique has been a workhorse of chip fabrication for decades, valued for its simplicity, low equipment cost, and ability to smooth out uneven surface topography that other deposition methods struggle to handle.
How the Spinning Process Works
The basic procedure is deceptively simple. A small puddle of liquid precursor, usually a solution of silicon-containing molecules dissolved in an organic solvent or alcohol, is dispensed onto the center of a wafer. The wafer then spins at high speed, and the liquid spreads outward under centrifugal force while excess material flies off the edge. Solvent evaporates during and after spinning, leaving behind a thin, solid film.
Film thickness depends primarily on two things: the viscosity of the liquid and the spin speed. Higher viscosity or slower spinning gives thicker films; lower viscosity or faster spinning gives thinner ones. For spin-on glass formulations and photoresists, the relationship between thickness and spin speed follows a well-characterized pattern in which thickness decreases with the inverse square root of the rotation rate. Polyimide solutions, by contrast, can show a wider range of behavior because of differences in how their solvents evaporate, not because the fluids behave strangely during flow.1Journal of The Electrochemical Society. Dependence of Film Thickness on Speed in Spin Coating This predictability is a major advantage. Engineers can dial in a target thickness with good confidence by selecting the right combination of solution concentration and spin speed.
A typical spin-on dopant process, for example, runs at around 3,000 rpm for 30 seconds.2Solid-State Electronics. Spin-on dopant technology for cost-effective source/drain formation in silicon MOSFETs The equipment itself is a benchtop spin coater, far cheaper and simpler than the vacuum deposition systems used for chemical vapor deposition or sputtering. That accessibility is part of what makes the technology attractive for research labs and lower-volume production lines.
From Liquid to Glass
Spinning creates a film, but the film at that point is essentially a fragile gel. It needs heat to become a true glass. The thermal processing step, usually called curing or baking, serves two purposes: it drives off residual solvent and water, and it triggers chemical reactions that cross-link the silicon-oxygen network into a dense, stable solid.
Curing temperatures vary by application and material, but the range of roughly 370 to 450 °C is common for standard interlayer dielectric applications.3Journal of Materials Research. Outgassing behavior of spin-on-glass Some formulations, particularly methyl silsesquioxane-based films designed for low dielectric constant applications, require curing above 400 °C to decompose sacrificial additives and generate the desired porosity.4Electrochemical and Solid-State Letters. Low k, Porous Methyl Silsesquioxane and Spin‐On‐Glass The relationship between curing conditions and final film properties is one of the most studied aspects of the technology. Higher temperatures and longer curing times generally produce denser, more mechanically robust films, but there is a trade-off: more aggressive curing can increase the film’s dielectric constant, which defeats the purpose in applications where you want that value to be low.
This tension between mechanical strength and electrical performance sits at the heart of spin-on glass engineering. Researchers have shown that the driving force for cracking, set by the thermal expansion mismatch between the film and the silicon wafer, increases less rapidly during curing than the film’s fracture resistance does, meaning that longer cures tend to net out in favor of reliability.5Journal of The Electrochemical Society. Stress‐Corrosion Cracking of Low‐Dielectric‐Constant Spin‐On‐Glass Thin Films But engineers still have to find the sweet spot for their specific material and device stack.
The Two Main Families of Spin-On Glass
Not all spin-on glass films are the same chemistry. The two most common families are silicates and organosiloxanes (often called polysiloxanes or silsesquioxanes, depending on their structure). The distinction matters because they behave differently during processing and in the finished device.
Silicate-based spin-on glass is closer in composition to ordinary glass. It forms a pure silicon dioxide network and tends to have a higher dielectric constant. It is also more porous and hydrophilic, meaning it absorbs moisture readily. In one study, silicate films absorbed roughly six times more water than their polysiloxane counterparts.6Journal of Materials Research. Outgassing behavior of spin-on-glass That moisture absorption causes problems down the line, as trapped water can outgas during later high-temperature processing steps.
Organosiloxane films contain organic groups, most often methyl groups bonded to silicon atoms. These organic components lower the dielectric constant and reduce moisture uptake. A well-cured polysiloxane film shows lower stress than a silicate film, making it a better candidate for planarization, the process of leveling out surface bumps so that subsequent layers can be deposited on a flat surface. The trade-off is that organic films can produce more outgassing if they are later heated above their initial cure temperature, because the organic components begin to decompose. A high-temperature rebake above 410 °C caused roughly twice as much weight loss from polysiloxane films as from silicate films in the same study.
Why Chip Makers Use It
Spin-on glass films serve several distinct roles in semiconductor manufacturing. The most traditional is planarization. Modern chips have multiple layers of metal wiring stacked on top of each other, and each layer creates surface topography: ridges over the metal lines and valleys between them. Before the next layer can be deposited, those bumps need to be smoothed out. A spin-on glass coating naturally flows into low spots and levels the surface, much like pouring syrup onto a waffle fills the squares.
The films also serve as interlayer dielectrics, the insulating layers that separate metal wiring levels and prevent electrical shorts. For this role, a low dielectric constant is prized because it reduces capacitive coupling between adjacent wires, which in turn allows signals to travel faster and with less energy loss. Standard spin-on glass has a dielectric constant around 3, and engineered versions can push this lower. A methyl silsesquioxane film, for instance, had its dielectric constant lowered from 2.7 to 2.3 by incorporating sacrificial polymer particles that burned away during curing, leaving behind closed nanopores about 70 nanometers across.7Electrochemical and Solid-State Letters. Low k, Porous Methyl Silsesquioxane and Spin‐On‐Glass A separate methyl siloxane spin-on glass was lowered from 3.1 to 2.7 using the same porosity approach.
Beyond planarization and insulation, the films can act as protective coatings, masking layers, and even gettering layers that trap unwanted impurities.8SpringerLink / Surface Coatings International Part B. ‘Spin-on glass’ films for semiconductor technology The range of roles reflects the versatility of the basic idea: spin on a liquid, bake it, and you have a glass layer tailored by the chemistry of the precursor solution.
Spin-On Dopants
One application that sits in a related but distinct space is spin-on doping. Here, the liquid precursor is loaded with dopant atoms, typically phosphorus for n-type doping or boron for p-type doping, rather than being intended as a permanent glass layer. The coated film serves as a solid source of dopant atoms that are driven into the underlying silicon by a subsequent high-temperature anneal.
This technique has real cost advantages over ion implantation, the more common method of introducing dopants into silicon. Ion implantation requires an expensive, complex machine that accelerates charged atoms into the wafer surface. Spin-on doping replaces that with a spin coater, a hot plate, and a furnace. For forming the source and drain regions of transistors, a phosphorus-based spin-on dopant annealed between 800 and 900 °C produced sheet resistances that dropped dramatically, from about 1,260 down to 161 ohms per square, as the anneal temperature rose. Junction depth simultaneously increased from about 0.14 to 0.32 micrometers.9Solid-State Electronics. Spin-on dopant technology for cost-effective source/drain formation in silicon MOSFETs Those numbers illustrate the basic knob: higher temperature drives dopant atoms deeper and in greater quantity, lowering resistance but making the doped region thicker.
Researchers have also demonstrated that spin-on dopants can be used for both n-type and p-type doping simultaneously on different sides of a wafer, using a proximity diffusion technique where dopant source wafers are placed facing both surfaces during a single rapid thermal cycle.10Materials Science in Semiconductor Processing. Proximity rapid thermal diffusion for simultaneous phosphorus and boron doping of silicon Complete CMOS transistor fabrication, with both n-channel and p-channel devices, has been achieved using spin-on dopant diffusion as the sole doping technique.11Japanese Journal of Applied Physics. An experimental study of solid source diffusion by spin on dopants and its application for minimal silicon-on-insulator CMOS fabrication For university clean rooms and small-volume manufacturing lines where purchasing an ion implanter is impractical, spin-on doping can be a genuinely enabling technology.
Reliability Problems and Poisoned Vias
Spin-on glass is not without headaches. Two mechanical stability concerns dominate. First, the cured film has a higher coefficient of thermal expansion than the silicon wafer beneath it. When the wafer cools from the curing temperature, the glass wants to shrink more than the silicon lets it, putting the film under tensile stress. Films under tension are prone to cracking. Second, silica-based films react with water in the surrounding atmosphere, and this chemical attack weakens the material over time. The combination of tensile stress and moisture sensitivity can lead to stress-corrosion cracking, where small flaws in the film propagate slowly into full cracks during storage or later processing.12Journal of The Electrochemical Society. Stress‐Corrosion Cracking of Low‐Dielectric‐Constant Spin‐On‐Glass Thin Films Thicker films are more susceptible because they store more elastic energy per unit area, giving cracks more driving force.
An even more insidious failure mode is the “poisoned via.” In a multilayer metal interconnect structure, small holes called vias are etched through the insulating layers to connect one wiring level to the next. If a via passes through a spin-on glass layer, residual moisture and gases trapped in the porous film can contaminate the metal deposited inside the via, causing a large increase in electrical resistance at that connection. Organic siloxane formulations are especially prone to this because they tend to re-absorb moisture and may not fully convert to dense glass during curing. The standard workaround is an etchback step that removes the spin-on glass from the area around via openings before metal deposition, keeping the problematic material away from the electrical contacts.13SemiFlows. Spin On Glass (SOG): Physical Principles, Curing Mechanisms, and Process Integration – Section: Challenges & Failure Modes
Etching and Removal
When spin-on glass is used as a sacrificial layer rather than a permanent one, it eventually needs to be removed. This is common in microelectromechanical systems (MEMS) fabrication, where the glass might temporarily fill a cavity that later needs to be hollow. Hydrofluoric acid (HF) is the standard etchant for dissolving silica-based materials, and it works well on spin-on glass, but the etching dynamics become tricky in confined geometries.
Inside narrow microchannels, the etch rate is initially constant because fresh HF can reach the glass surface without impediment. As the channel gets longer, though, the reaction slows because used-up acid near the etching front has to diffuse out while fresh acid diffuses in. The etch rate drops by roughly 70% when this transition from reaction-limited to diffusion-limited etching occurs. At T-shaped junctions where channels meet, the rate drops further, to about two-thirds of the straight-channel value, because the HF concentration drops sharply at the intersection.14Journal of The Electrochemical Society. HF Etching of Sacrificial Spin-on Glass in Straight and Junctioned Microchannels for MEMS Microstructure Release These slowdowns are predictable and can be designed around, but they mean that removing spin-on glass from deep or complex geometries takes considerably longer than you might expect from the initial etch rate alone. Using a dilute 5% HF concentration strikes a practical balance between acceptable speed and avoiding damage to nearby structures.
Optical Waveguides and Photonics
Beyond electronics, spin-on glass has found a niche in photonics. Optical waveguides, the tiny channels that route light signals on a chip, need layers of glass with precisely controlled refractive indices. A core layer with a slightly higher index is sandwiched between cladding layers of lower index, and light stays confined inside the core by total internal reflection, much like fiber optics in miniature.
Researchers have built buried channel waveguides on silicon substrates by repeatedly spin-coating sol-gel precursors and rapidly annealing each layer to densify it, building up films around 10 micrometers thick. By varying the titanium dioxide concentration in the precursor solution, the refractive index of each layer can be tuned independently. A bilayer structure with different titania concentrations in the core and cladding creates the index step needed for waveguiding, and the core can be patterned into ridge shapes by reactive ion etching before being buried under additional cladding material.15Applied Optics. Fabrication of buried channel waveguides on silicon substrates using spin-on glass Careful control of the anneal temperature keeps the layers stress-free, which prevents the birefringence and cracking that would ruin optical performance. The approach is far cheaper than the plasma-enhanced CVD processes typically used for waveguide fabrication, making it attractive for prototyping and for applications where cost matters more than the ultimate performance achievable in a full-scale foundry.
Where Spin-On Glass Fits in the Landscape
The semiconductor industry’s dominant deposition method for dielectric films is chemical vapor deposition (CVD), which grows films from gaseous precursors inside a vacuum chamber. CVD films are denser, more uniform over large wafer areas, and generally better suited to the extreme dimensional requirements of leading-edge chip fabrication at nodes below 10 nanometers. Spin-on glass does not compete head-to-head at that frontier. Its strength lies in applications where planarization matters more than raw film density, where low equipment cost is important, or where the chemistry of the deposited material needs to be tunable in ways that gas-phase methods do not easily allow.
That is why spin-on glass remains relevant despite being a mature technology. The porous low-dielectric-constant films described earlier are a good example: creating controlled porosity by burning out a sacrificial component is much easier in a solution-deposited film than in a CVD film. Likewise, spin-on doping survives because it replaces a million-dollar ion implanter with a few thousand dollars’ worth of chemicals and a spin coater. And in photonics, the ability to control refractive index by tweaking the sol-gel chemistry gives spin coating a flexibility that sputtering and CVD do not match without significantly more complexity.
The limitations are equally clear. Thickness uniformity over large wafers is harder to achieve than with CVD. Films are inherently porous compared to their vacuum-deposited counterparts, making moisture absorption and outgassing persistent concerns. And the crack sensitivity of tensile films on silicon means that spin-on glass layers must be kept thin or sandwiched between compressive layers to avoid mechanical failure. These are not fatal shortcomings so much as constraints that define the technology’s natural territory: gap fill, planarization, low-cost doping, specialty dielectrics, and applications where the liquid precursor chemistry can do something that a gas-phase process cannot.
Choosing Between Silicate and Organosiloxane Formulations
If you are selecting a spin-on glass for a specific process step, the choice between silicate and organosiloxane chemistry comes down to what you care about most. Silicate films are chemically simpler and cure into something close to pure silicon dioxide, which makes their etch behavior and electrical properties more predictable. But their tendency to absorb water makes them problematic in any flow where the film will be buried under subsequent layers and then exposed to high temperatures, because the absorbed moisture has nowhere to go and causes blistering or delamination.
Organosiloxane or silsesquioxane films, by contrast, absorb far less moisture and sit under lower residual stress after curing, both of which make them easier to integrate into a multilayer stack. Their dielectric constant is lower out of the box, and it can be pushed lower still through controlled porosity. The downside is that the organic content makes them more susceptible to outgassing if the thermal budget of later process steps exceeds the original cure temperature. A well-cured organosiloxane film can be more stable than a silicate film in practice, but “well-cured” is doing a lot of work in that sentence, and the process window is narrower.16Journal of Materials Research. Outgassing behavior of spin-on-glass
For applications where the film is permanent and buried, like an interlayer dielectric in an interconnect stack, the organosiloxane family is almost always the better choice now. For applications where the film is sacrificial and will be etched away, like gap fill before a chemical-mechanical polishing step, the simpler silicate chemistry may be preferred because its etch rate in HF is more predictable and the moisture issue is moot since the film will not survive to see later thermal processing.