Directed Self Assembly – record breaking small features
By Michael P.C. Watts
Directed Self Assembly (DSA) was the breakout subject at this year’s SPIE Advanced Lithography Conference. This conference is the biggest annual get together for lithography nerds, and I use it to keep up with the latest academic and industrial trends. Anyone who is anyone seems to be evaluating DSA. On a personal note, as it turns out, I did my PhD in block copolymers – the materials used in DSA – many moons ago and I will make a claim to fame at the end of this blog !
The first application for DSA in the patterning business is to enable patterned media by keeping mask writing times in the realm of practicality. In semiconductor devices, the applications are more challenging because of the pattern complexity in real devices.
DSA uses a unique property of block copolymers to frequency double or quadruple a regular pattern. Block copolymers spontaneously phase separate – “SELF ASSEMBLE” – when the two ends of one molecule are sufficiently different in chemical characteristics. The size of the phases is dependent on the molecular weight of each end, and one of the breakthroughs at the conference were reports of making features as small as 7 nm half pitch, by several teams including the University of Texas, Georgia Tech. and Seagate – more on small phases at the end of this blog.
This phase separation cleverness has been used for many years to make a family of plastics called “thermoplastic rubbers”. These are materials that have the properties of a cross-linked elastic rubber and can be melt processed by extrusion and injection molding into arbitrary shapes.
For semiconductor applications, features must be placed in a known location. The “DIRECTED” part of the self-assembly is achieved by either a physical constraint or by placing key regions on the surface that attract one of the phases. The first semiconductor application for these materials, and the subject an my earlier blog, has been frequency doubling of a spot pattern for making masks for patterned media. Every other spot is patterned by electron beam lithography, and the block copolymer fills in the gaps. The Nealy group at U. Wisconsin did the earliest work on the mask application.
There are now several teams working on production IC applications. The Wisconsin group has teamed with AZ Electronic Materials, TEL (Japan) and IMEC (Belgium). There is a French based team led by CEA LETI with a materials supplier, TEL and Mapper a multibeam e-beam supplier who are working on “complimentary” lithography. In addition there is a Stanford/ JSR/ IBM / AMAT team, and an Intel / Georgia Tech team — it’s a who’s who of advanced semiconductor lithography.
If block copolymers replicate regular patterns, how can they be used for semiconductor devices where even the most regular memory array has irregular drive transistors and decoders ? It turns out that in order to improve process control, the lithographers have been driving the designers to orient all transistors in one direction. Now the lithographers can independently pattern long lines in the X axis and then make short breaks in the lines in the Y axis. The X&Y axis patterning is becoming more popular than double patterning of two interlocking random patterns. DSA allows the long lines in the X axis to be doubled at a much lower cost than all the alternatives. There is still the challenge of cutting the lines. The idea being proposed by Intel is to use a “complementary” lithography such as EUV or e-beam to cut the lines. A team from CEA-LETI showed a test write of a doubled line structure with e-beam written cut lines. Using this approach, DSA would partially enable a further doubling or quadrupling of the X&Y double pattern. DSA enables part of further feature size reduction so it helps, but is not a complete solution on its own.
An alternative application is to use DSA to improve line width control by using block copolymers on top of poorly controlled starting features. Kurt Ronse suggested that using DSA to improve contacts printed by EUV will be an early test at IMEC (Belgium).
A less obvious strategy for creating small groups of contacts was discussed by Philip Wong from Stanford University. He proposed placing relatively large constraining regions, or walls, around groups of 1-4 contacts. The block copolymer is used to fill the space between the walls and phase separation creates the contacts in the target locations. As I noted earlier the equilibrium or natural spacing between the spots is dependent on the molecular weight. Philip Wong showed that the walls can force the spots from the equilibrium positions and can be placed in specific target locations within some limits. To make this work, the final layout must be intelligently fractured to create a mask for the “guiding walls”. This whole approach takes advantage of the fact that contacts occur in clusters within logic or SRAM cells. It will not work for DRAM where neighboring transistors are all laid out with the smallest resolvable separation. Getting the right layout rules is going to need some good modeling tools. In my opinion this approach has intriguing potential – however driving a material that relies on equilibrium – out of equilibrium might be asking for trouble with defects and line width control.
All in all, as the alternatives remain “not ready for production” and optical lithography cannot change wavelength or Numerical Aperture, the most efficient way to optically double pattern has become the only game in town. DSA looks like a serious contender to extend double patterning. The fundamental questions of defect density and line width control will start to get clearer as groups, such as IMEC and IBM, test the material in a clean room environment.
Where does a 34 year old PhD thesis fit in ? At the time, I was trying to make block copolymers that just phase separated for use as vibration damping materials, and I wanted to prove that it was the blocks that phase separated not whole molecules. I used small angle X ray scattering and found phase separation with a period of 10-15 nm, and that the period scaled with block length. At the time, I did not know about lithography, and no one would have cared as these were not lithographically useful. Even if had been lithographically useful, at the time; device features were 100 x larger than my phases so they could not have been used for devices. However, it is fun to be able to claim the first demonstration of sub 10 nm half pitch phases in a block copolymer in 1978, beating the competition by 34 years !
About the author
Mike Watts has been patterning since 1 um was the critical barrier, in other words for a longtime. I am a tall limey who is failing to develop a Texas accent here in Austin. I have a consulting shingle at www.impattern.com.
My blog “ImPattering” will focus on the latest developments in the business and technology of patterning. I am particularly interested in trying to identify how the latest commercial applications evolve.