403 A typical photolithograpy cycle is shown on the slide. The - - PDF document

403

Photolithography is an important process for VLSI manufacturing. As we had seen during in the introduction to the course, the key to success of VLSI technology is its ability to reduce device dimensions with a resultant increase in the performance. The cost per function also came down as more devices could be fabricated in the same chip area. This ability to fabricate smaller and smaller features is mainly attributed to advances in lithography. Lithography systems are also the most expensive tools in a modern VLSI manufacturing facility. The cost of 193 nm lithography tools that are presently used in manufacturing is in the range of 10 - 15 million USD. The cost of next generation extreme UV (EUV) tool is expected to be in the range of 100 million USD (ref: http://www.eetimes.com/electronics-news/4210901/EUV-tool-costs-hit--120- million-). From the perspective of the circuit designer, the interface between process technology and circuit designers traditionally happened at the lithography. The end product of a traditional logic circuit design cycle from a process point is a mask design which contains information on the relative locations of various structures on the chip and the interconnection between them.

A typical photolithograpy cycle is shown on the slide. The design of a chip would start with specifications, circuit design etc. Eventually the circuit designer would provide what is called a layout

• ut of all the devices and structures and the interconnects. The layout is 2 dimensional. Once a

layout is made, the correctness of the design and conformance to specifications should be verified using appropriate simulations. It should be also verified that the layout passes the design rules set

• ut by technology. We would discuss this point in more detail later. Since modern VLSI circuits are

quite complex and contains millions of transistors and other electronic devices, the task of layout design, simulation and DRC are carried out with the help of computer aided design tools. One of the interesting point here is that faster and more complex VLSI chips have resulted in higher performance of computers which in turn has facilitated the design of more complex chips. Once the layout is designed, it is transferred to a mask using an electron beam lithography process. The mask thus created is subsequently used for transferring patterns on the mask to photo resist coated on a wafer. The mask contains regions that are transparent and that are opaque. The wafer exposure system consist of a source of light, which is then collimated to obtain a parallel beam using a condenser lens. The parallel beam of light is then passed through the mask. Light will pass through the transparent regions in the mask. So the light pattern that comes out of the mask would follow the image on the mask. This light is subsequently used for exposing photo resist coated

• n the wafer surface. The photo resist is a photo sensitive material. The regions of the resist that is

exposed to light would undergo certain changes. The resist regions exposed can be removed or retained selective to the regions that are not exposed. Typically the smallest features on the design could be as small as 100nm or smaller. However in advanced lithography systems, the corresponding feature may be 4 to 5 times larger. This simplifies mask fabrication. A reduction lens is used for reducing the size of the image printed on the wafer.

404

This is an example of the application of lithography. This also illustrates the lithography process in more detail. In this specific example, a feature is etched in SiO2 grown or deposited on Si. First a photo resist is spin coated on the surface of the wafer. Subsequently a lithography mask is used to optically expose the photo resist. This results in transfer

• f the image on the mask on to the resist. Subsequently the resist is developed.

Resist development is a process in which the whole wafer is dipped in a chemical solution called a developer which would selectively etch resist that has been exposed to light or that has not been exposed to light. In this particular example, the resist is removed from the areas where it was not exposed to light. Subsequently SiO2 is etched using the resist as the mask. The etchant typically used for this process is buffered HF. Buffered HF is a solution of HF and NH4F in

• water. After etching the SiO2, the resist is removed so that the wafer can be further

processed.

405

The optical source used has a profound impact on the lithography system and the minimum feature size that can be printed with the system. We would see later that the feature size is directly related to the wavelength of light. So lithography for smaller feature sizes would require smaller and smaller wavelengths of light. However there are several issues, which can be illustrated by EUV. EUV photons would have an energy of ~ 92 eV. The bandgap of SiO2 used for glass is 9 eV !. So glass can not be used for making lenses as it would completely absorb the light. In fact air would also absorb EUV. Hence the lithography process has to be carried out in vacuum! Although higher wavelengths are not fully absorbed by SiO2, impurities in the glass would enhance the absorption. White light sources like Xenon lamps may be used for low resolution (large feature size) lithography.

406

The mask contains both transparent and opaque regions on it. Most popular material for mask is glass. However e-beam lithography does not use any mask. EUV would be absorbed by glass. In the lab we also use transparency sheets when the feature sizes are more than 20

• um. The opaque regions in such cases are ink, printed using a laser printer.

Coming back to glass, the opaque regions are usually Chrome. In fact Chrome is coated on the whole surface of the glass. A lithography process is then carried out

• n this stack to etch away Chrome. Typical Chrome thickness is in the range of 100

nm. Modern VLSI processing may use 30 – 40 masks.

407

Two types of resist are used. Positive resist develops in areas where it was exposed and negative resist develops in areas where it was not exposed. Generally positive resists give better lithographic resolution than negative resists. Modern resists contains organic compounds. The resist is subsequently used as mask for etching (both wet and reaction etching processes) and ion implantation. If the intended application is etching, the resist should be robust against the chemicals and ambient (plasma, temperature etc) used for etching. For example aqueous HF is widely used for etching of SiO2. However most of the resists would be attacked by HF. On the other hand buffered HF is not seen to etch resists. In an ion implantation process, the resist should be thick enough to block implantation into the underlying layer. Another issue is related to the fact that the resist should be removed after the etching or ion implantation. However the resist may undergo chemical and physical transformation during etching and ion implantation, which makes it harder to remove.

408

Out of these factors, resolution is controlled by the resolution of the optical system and the capabilities of the photo resist and related processes. Pitch and overlay are decided by the optical system. The other parameters are controlled by the optical system as well as resist track.

409

Critical dimension that can be printed is an important parameter. In every new technology generation, the gate length has to be reduced. Presently you can buy chip with gate lengths of about 30 nm. The pitch is the minimum period of a periodic array of minimum feature size lines that can be printed. This is an important consideration because you would like to pack more devices. So it is desirable to reduce the spacing between two adjacent lines. Overlay or alignment is a very important parameter which dictates many of the design rules. Overlay error is the displacement of the features in to consecutive lithographic steps using the same mask. In practice we may not use the same mask for two consecutive litho steps. The impact of overlay on design rules can be described using an example. The CD and pitch achieved using modern lithography processes are in the same

• rder as the wavelength of light.

410

There are three types of printing systems in use. The first two types are mostly used in university laboratories and low volume production. In a contact printing system, the optical system consist of a source of light and a collimating lens. The mask is in direct contact with the photo resist. The features sizes and spacing between lines are in the same order of the wavelength of light used, diffraction of light through the mask would be a limiting factor for lithography. Due to diffraction, light would spread laterally from the edges of the mask. Hence the image registered in the resist would be larger. However in a contact printing process, the spacing between the mask and the resist is the minimum and hence diffraction effects would be minimal. High resolution, may be obtained in contact printing. The resist is in contact with the mask. This can create problems as mask particles can stick to the mask. This makes mask cleaning mandatory before the mask can be used again if high yield has to be maintained. The feature size on the mask should be the same as what is desired on resist. In proximity printing, the mask is not in contact with the resist. This avoids the issue of mask damage during printing. However diffraction at the mask edges would reduce the resolution of the system. In both the above cases, the whole wafer is usually exposed in one shot. For overlay control a lithography step has to be aligned to the previous step or to some reference features on the wafer. In contact printing and proximity printing, alignment marks are created on the wafer and all lithography steps are usually aligned to them. The alignment marks are placed at certain locations on the wafer and not on every chip or die. Such an alignment strategy is called global alignment. Both contact and proximity printing systems typically use global alignment. The projection printing system uses one more lens. The light coming out of the mask is focused by the second lens on to the wafer surface. The second lens provides an opportunity to reduce the feature sizes during projection. Typically in projection printing systems are 4 to 5x of the feature size on the circuit designers layout. This relaxes the requirements on the mask making process. Or we can say that the project system can print smaller features. Projection systems typically print a limited number of dies at a time. So the wafer is exposed and stepped for complete exposure of the wafer surface. The mask can be much smaller than the wafer size in such a scenario. Such a mask is called a reticle. It is possible to align the reticle to the already existing features on the wafer and this provides high alignment accuracy.

411

These images are abstractions of a projection system. Since the features on the mask can be of the same size as the wavelength of light used, ray optics can not be

• applied. We have to consider the light as waves. The progressing wavefronts of

light coming out of the collimating lens can be treated as several wavelets. When such a wavefront is passed through an apperture (transparent region in a mask), the wavelets that pass through would spread side ways due to diffraction. The consequence of this is that the image of the apperture on any intercepting plane would be larger than the apperture itself.

412

We would consider the formation of images through a lens system as in a projection lithography system. The light coming out of the aperture would diverge as shown. So only a fraction of the light emerging would be collected and focused by the focusing lens of the system. The sharper features of the aperture would diffract the light more and hence they are unlikely to be collected unless the lens has a large

• diameter. The image of a circular aperture formed on the image plane would be as
• shown. The intensity of light is shown in the graph. The first minima are seen at

1.22 λf/d on either side of the primary maximum of the intensity profile.

413

The resolving power of such a system can be studied by considering two point sources (two dots on a mask for example). We should be able to resolve the two dots on the image

• plane. If the two dots are too close we would not be able to resolve them. It is assumed that

the two points can be resolved if the intensity maximum of either dots images are at the minimum of the other dot or farther. This is called Rayleigh’s condition. So the two dots can be resolved if they are spaced R apart, where R is given on the slide. n is the refractive index of the medium. Usually air is used. In such a case n = 1. Usually we are interested in non circular shapes. So 0.61 is replaced with a generic parameter which captures the geometry effect. K1 is in the range 0.6 to 0.8. Some ways to improve the resolving power of the system are the following: Use lower wavelength of light. Use a larger diameter focusing lens. Use a medium with a refractive index larger than 1. For example, 193nm immersion lithography systems in use in the field uses water as the medium, and water has n = 1.33. As per the refraction equation, n1sinθ1=n2sinθ2=NA. Where n1 is the index of refraction of the lens material and n2 is the index of refraction of the medium. So replacing air with water is going to change the n2. But θ2 would also change keeping NA constant. However a higher refractive index of enables the design of a lens with which the NA can be larger. When air is used, the highest theoretical value of NA is 1. But when water is used, the highest possible NA is the value of the refractive index of water.

414

Another issue of importance is the depth of focus. Discuss the image from the context of depth of focus.

415

We consider that the image is in focus within a distance that is defined as when the path length difference between the ray passing through the centre of the aperture and the edge of the entrance aperture is one fourth of the wavelength of light. NA has to be increased for better resolving power. But then the DOF also decreases. In immersion lithography, the DOF can also be increased. To understand this, the DOF is inversely proportional to θ. Θ decreases when water is used as the focal length of the lens would increase.

416