by Griff Resor, Resor Associates

February 22, 2010 – The SPIE Advanced Lithography Conference in San Jose, CA, provides a once-a-year opportunity to see the future of IC manufacturing. For years, this has been the conference where experts in lithography, resists, metrology, and design come to tell about their recent advances. Typically the good ideas move into IC production within two to three years.

Everyone will be watching the battle between optical lithography and EUV, which for the first time will have its own conference. The EUV conference has 127 papers — compared to 124 for the optical lithography conference, a huge advance for EUV in just one year — mostly aimed at high-volume insertion at the 22nm node by 2013. It will be interesting to learn how attendees feel about this schedule. Key challenges for EUV remain mask defects, resist resolution, and line-edge roughness (LER), and source power. As source power increases, there is a renewed concern for the lifetime of mirrors in the EUV optical train. I expect to hear credible evidence that progress on resist, sources, and lifetime will support the schedule. Progress on EUV mask defects is not so certain, so will set the pace from this week on.

The optical lithography conference has 124 papers, evenly split between the 32nm and 22 nm nodes. Computer modeling to optimize what’s left for a process window is the #1 topic (30 papers); double patterning comes in second (17 papers). For several years now it has been clear that some form of double patterning lithography (DPL) would be required to move optical lithography to the next node. Many variations on this theme will be reported this week. With overlay error budgets in the single digits of nanometers, masks and matching of scanner tools are also key topics.

A quick scan of the resist track’s 123 papers provides a sanity check on the EUV vs. optical lithography battle. Thirty-six of the resist papers describe optical and DPL materials and methods; 18 papers discuss products for EUV use. Many papers describe generic improvements in materials. It seems that material suppliers still see a larger market for optical and DPL products than for EUV products.

Three emerging technologies are the focus of the "alternative lithography" track. 26 papers describe direct writing tools, most using multi-beam e-beam. IC foundries such as TSMC are pushing for these tools as a workaround to the ever higher cost of masks for optical and EUV lithography. Nano-imprint lithography (NIL) continues to conquer new applications — structured media for high-density disk drives looks like a natural, and will be covered this week. Finally, self-assembly techniques continue to get attention, as nano-structures built in the IC industry and biotechnology converge.

After a busy week, Griff will report on his findings at this key lithography conference next week.

Griff Resor, an SST editorial advisory board member, is president of Resor Associates, ph 978-897-2200, e-mail [email protected].

by Franklin Kalk, CTO, Toppan Photomasks

September 23, 2009 –  Magnetic disk drive industry players are busy exploring new ways to pack more magnetic islands onto 2.5-in. and 3.5-in. platters. Today’s media, comprised of planar magnetic films deposited on aluminum or glass substrates, pack as much as 300Gb in a square inch, and the industry roadmap calls for terabit/in² densities in a few years. To break the terabit/in² barrier, new media architectures will be required because the magnetic islands that constitute the bits of information stored on the disk become less stable as their dimensions shrink. One option to stabilize the islands’ magnetic states requires patterning the surface of the media, leading media manufacturers into the semiconductor industry’s world of lithography and metrology.

Patterned media come in two flavors: discrete track (grooved surfaces) and bit-patterned (individually defined magnetic dots). Both types require some sort of lithography to impose the pattern on the disk, and the leading (some say only) lithography candidate is nanoimprint lithography (NIL). NIL requires a master relief pattern to be written in resist and transferred into silicon or quartz by dry etching. This master pattern is then propagated via NIL replication into multiple quartz worker templates, which can in turn be replicated into patterns on multiple disks. In principle, as many as perhaps 10 million disks can be reproduced from a single master progenitor. Several papers describing the mastering, replication and metrology challenges of patterned media were presented in a series of sessions at last week’s SPIE Conference on Photomask Technology in Monterey, CA (Sept. 14-17).

Papers from FUJIFILM, Dai Nippon Printing, Hoya, and Hitachi Global Storage Technologies explored the range of methods being developed for mastering and replicating both types of patterned media. Hard-disk masters are fabricated on 6-in. silicon or fused silica wafers. The wafer is first coated with an electron-sensitive resist, which is then exposed on a rotary stage Gaussian spot electron-beam writing system, typically with 100 keV accelerating voltage. The resist is then developed to reveal the desired relief pattern, which is transferred by dry etching into the master substrate. An interim hard mask may or may not be used, depending on the resist’s ability to withstand the substrate etch.

Nobuhito Toyama of DNP described 100 keV e-beam mastering of 6-in. diameter silicon and quartz master molds for 2.5-in. discrete track media; track pitches as low as 44nm on the quartz mold were achieved. The quartz etching process provided adequate depth uniformity and line-edge roughness, but the feature width non-uniformity is still an area of active improvement. Toyama-san described the optimization of a new resist process for the silicon mold and the method of forming unequal groove and land widths, showing 45nm track pitch with a 15nm groove width. Extremely high dose is required to write such small features with high-resolution e-beam resists — 350mC/cm² for ZEP 520A and 124mC/cm² for the new chemically amplified resist.

Hiroshi Yamashita presented details of Hoya’s recently developed discrete track media mastering process. He showed 2.5-in. quartz masters with track pitches ranging from 120nm down to 50nm. The process employs a 10nm hardmask under the ZEP 520A resist to aid in the quartz etching. So far, Hoya has been able to build 20nm grooves on a 50nm pitch. Yamashita-san announced that Hoya will begin offering commercial quartz molds for 2.5-in. discrete track media with 50nm track pitch and customer servo data in early 2010.

Noriko Yamashita of Fujifilm described the use of NIL to transfer a silicon master pattern into quartz replica templates for discrete track media. The company’s new NIL resist has excellent resistance to the quartz dry etch plasma and good pattern profile, obviating the need for a hardmask. So far, Yamashita-san and her colleagues have replicated track pitches as low as 75nm on 2.5-in. disk areas with this simplified process.

The highlight of the 16 or so NIL papers at the conference was Tom Albrecht’s (Hitachi Global Storage Technologies) presentation of work on the directed self-assembly of block copolymers to form bit patterned media. In simple terms, by adjusting chain volume, degree of polymerization, and mixture ratio, two immiscible polymers can be mixed to form regular structures. Using the classic case of polystyrene and PMMA, Albrecht showed that one can form arrays of spheres, lamellae, or cylinders of one material in another. Cylinders are of particular interest for bit patterned media — a suitable COP, coated on a patterned master to form regular arrays of cylinders at a constant radial and longitudinal pitch, provides perfect sites for magnetic dots. The method can be employed to multiply the density of a sparsely written pattern of cylinders, shortening master writing times which can stretch for more than two weeks. It can also be used to regularize the dot positions of an already-written master pattern. Albrecht showed e-beam-written dot position jitter of about 3.5nm, and  after COP overcoating, the resultant cylindrical dots had position jitter of just 1.2nm. A bonus is that the COP period can be stretched or compressed a few percentage points, affording radial-band recording identical to that used in today’s hard disk drives.

Brian Grenon of Grenon Consulting provided needed insight into the manufacturing cost of a patterned media master. Even though the names of the various process steps are similar, media mastering is much different that making a conventional IC photomask. For example, the writing time for a master pattern takes 10-20 days, longer than the entire cycle time for an advanced photomask. Similarly, 100% master pattern inspection with an electron beam tool will take more than one day. Grenon pointed out that mastering costs will be high unless the equipment can find other uses while a pattern is being written. He also rightly noted the high risk of yield loss during such long writing times.

A key success factor for patterned media will be the quality and consistency of the groove or bit pattern profiles. The metrology of these tiny features is therefore important, and was the subject of four BACUS papers. In a clever extension of existing technology, Seagate’s Justin Hwu applied analytical SEM CD measurements on quartz nanoimprint templates — first characterizing the beam and optimizing it for quartz template feature metrology, then after capturing images of features, using simulation to extract profile information without cross-sectioning the template.

Roman Sappey of KLA-Tencor and Jeff Roberts of n&k Technology both applied optical scatterometry to the measurement and characterization of patterned media groove profiles. Scatterometry requires simulation to extract profile information from transmitted and reflected spectral signals. Thus, it’s important to have as many different variables as possible: polarization, wavelength, transmitted light, and reflected light. This is no problem for the technique, which can collect all data at a rate of about one spot every three seconds. Roberts performed a sensitivity analysis of the signals as a function of items such as feature size, and found that the method can have quite strong sensitivity to the CD as the feature size decreases. Both authors concluded that optical scatterometry will be suitable for characterizing pattern profiles for at least the next four years, but NIL residual layer thickness is more challenging.

Kazuhiko Omote of Rigaku applied scatterometry principles to x-ray diffraction on groove structures. He contended that conventional optical scatterometry wavelengths are too long to achieve high accuracy on patterned media’s small features. He showed the ability to measure pitch to within 0.1nm and to reconstruct the fine details of feature profiles on magnetic substrates, including corner rounding, wall angle, CD and height. At 500sec per measurement, speed is a drawback, but as an analytical technique Rigaku has found a nice application.

This year’s presentations point the way to patterned media as an expanding topic with fertile development ground in future mask symposia.