Silicon nitride membrane

A silicon nitride membrane is a film of silicon nitride, usually a few tens of nanometers thick, that spans a window etched through a silicon wafer, so that the film is held at its edges and free over the window. It is sold as a window for electron microscopes and X-rays, and since the first sculpted pore of 2001 it has been the usual material for solid-state nanopores.^[1] The film is an insulator that resists the potassium hydroxide used to etch the silicon behind it, it can be deposited with a thickness controlled to a nanometer, and it is stiff enough to span tens of micrometers without support.

The films are grown by low-pressure chemical vapor deposition (LPCVD), in which dichlorosilane and ammonia react on the hot surface of a silicon wafer at roughly 800 °C. Silicon nitride in the strict sense, Si₃N₄, needs three silicon atoms for four of nitrogen, and a patent for membrane substrates reports that reaching that ratio takes about three times as much ammonia as dichlorosilane.^[2] A film grown that way is under tensile stress of more than 1 GPa and fractures easily. Adding more dichlorosilane gives a silicon-rich film, written SiN_x with x below 4/3, which is amorphous and holds much less stress. The silicon-rich films are the ones used for thin free-standing membranes.^[3]

The two kinds of film are on the market and have been measured. In a study of commercial membranes from one supplier, the low-stress films carried a built-in tension of 135 to 155 MPa and the high-stress film 1,030 MPa. The same study put Young's modulus at 254 to 314 GPa and the density at 3,300 to 3,400 kg/m³.^[4] Merle and Göken measured the fracture toughness of LPCVD films 40 to 108 nm thick by bulge tests and found 6.3 ± 0.4 MPa m^1/2 at every thickness.^[5]

The films have uses well beyond nanopores. They are windows through which an electron microscope looks at a specimen, and nanometer-scale windows for X-rays, where the thinness of the film is the point.^[6] Prestressed nitride membranes also serve in photonics and optomechanics, as resonators whose vibrations have a high mechanical quality factor and as sensors, and the study of commercial membranes cited above was done for that purpose; its authors measured the membranes by their resonant frequencies, a method that does not touch the film.^[4] In micromechanics the silicon-rich film is a structural layer in its own right.^[3]

The film is deposited on both faces of a silicon wafer. On the back, a square of nitride is removed by lithography, and the wafer is put into hot potassium hydroxide, which dissolves silicon along the {111} crystal planes much more slowly than in other directions. The etch therefore cuts a pit with walls at 54.74° to the surface, and stops when it reaches the nitride on the front. The geometry fixes the size of the opening: a 50 µm window in a wafer 200 µm thick needs an opening about 333 µm wide on the back, since 50 + 2 × 200 × cot 54.74° ≈ 333. For a wafer 500 µm thick and a window of 10 µm, the opening is 717 µm; the drawing below is to scale.

Many nanopore laboratories buy the result as transmission electron microscope windows. Kwok and colleagues, who made pores by controlled dielectric breakdown, used commercial membranes of low-stress (below 250 MPa) nitride 10 nm or 30 nm thick on a silicon substrate 200 µm thick, with a window of 50 µm by 50 µm opened by the KOH etch. They note that the membranes can be cleaned in oxygen plasma for 30 s at 30 W to help wetting, though this is not required, and they filtered and degassed their solutions.^[7]

A defect shows itself only as leakage. Kwok and colleagues inferred the absence of pinholes and nano-cracks from the fact that a membrane passed less than a picoampere below ±1 V before the pore was made.^[7]

The thickness also sets the length of the pore. A pore through 20 nm of nitride is as long as about 40 bases of single-stranded DNA at 0.5 nm each, which is one reason thinner materials were tried; see Graphene nanopore.

The nitride surface oxidizes in air and in water, and at neutral pH it is negatively charged, with the consequences described in Electric double layer and Surface charge (nanopore). The surface is hydrophilic when clean and acquires organic films from air and from handling, which is why chips are cleaned with plasma or solvents just before use; see Nanopore wetting.

The bill of materials of the open MR1 Nanopore Reader lists chips with membranes 10 to 50 nm thick and pores of 2 to 10 nm. A worked example in its design notes uses a membrane 200 nm thick over a window 10 µm wide in a die 4 mm square and 500 µm thick, rotated 45° inside a cast gasket of silicone, and the opening of 717 µm on the back follows from the etch geometry above.^[8]

• Nanopore fabrication
• Access resistance
• Graphene nanopore
• Molybdenum disulfide nanopore
• Focused ion beam