Chemical vapor deposition grows a solid film out of gas. Reactant gases flow over a heated wafer, react at or near its surface, and leave behind a solid layer while the volatile byproducts are pumped away. Unlike PVD, where the atoms that land are the same atoms that left a target along straight-line paths, CVD builds the film from a chemical reaction that happens right at the surface. That single difference is why CVD coats the walls and floors of deep features evenly where sputtering cannot.\n\nConformality is the property that made CVD indispensable. Because the film-forming reaction happens wherever precursor molecules can reach and stick, CVD deposits nearly the same thickness on the top, sidewalls, and bottom of a trench or via. That conformal coating is exactly what you need for gate dielectrics, spacer nitrides, tungsten contact fill, and liner films inside high-aspect-ratio structures. The trade-off is that you are now running surface chemistry, so temperature, pressure, precursor flux, and reaction byproducts all become knobs you must control.\n\nThe named CVD variants are really just different ways to supply energy and manage pressure. Atmospheric-pressure CVD (APCVD) is fast but less uniform. Low-pressure CVD (LPCVD) runs hot in a vacuum furnace and trades rate for excellent uniformity and conformality across a full boat of wafers, which is why it deposits polysilicon and silicon nitride. Plasma-enhanced CVD (PECVD) uses an RF plasma to crack the precursors, so the reaction runs at a much lower wafer temperature, protecting underlying metal layers at the cost of some film quality and hydrogen incorporation. High-density-plasma CVD (HDP-CVD) adds simultaneous sputter etching to fill aggressive gaps without voids.\n\nThermal budget is the axis that decides which variant you can use. A film deposited early in the flow, before any aluminum or copper is on the wafer, can tolerate a hot LPCVD furnace. A film deposited over completed metal interconnect cannot, because the heat would degrade the metal and diffuse junctions, so it must go down cold in a PECVD chamber. Much of the art of integration is matching each deposition step to how much heat the wafer can still take at that point.\n\nStep coverage, rate, and film quality are in constant tension. Running hotter or at lower pressure improves conformality and density but costs thermal budget; adding plasma lets you go cold but can damage the surface and leave hydrogen or stress in the film. There is no single best CVD, only the right variant for a given layer's temperature ceiling, aspect ratio, and quality requirement, which is why a modern fab runs several CVD chemistries side by side.\n\n| Variant | Energy / pressure | Wafer temp | Best for |\n|---|---|---|---|\n| APCVD | Thermal, atmospheric | Moderate | Fast oxide, older flows |\n| LPCVD | Thermal, low pressure | High (550-800C) | Polysilicon, nitride, uniform batch |\n| PECVD | RF plasma, low pressure | Low (200-400C) | Dielectrics over metal, low thermal budget |\n| HDP-CVD | Dense plasma + sputter | Moderate | Void-free gap fill in tight features |\n\n``svg\n\n``\n\nRead CVD through a surface-chemistry-and-thermal-budget lens rather than a generic coating lens. Once you see the film as the product of a gas-phase reaction happening on a hot wafer, the whole variant zoo makes sense: conformality comes free from the chemistry, and the choice between LPCVD, PECVD, and HDP-CVD is really just a negotiation between how much heat the wafer can still take and how hard the gap is to fill.