Документ взят из кэша поисковой машины. Адрес оригинального документа : http://danp.sinp.msu.ru/Articles_GSIB/nimb206_830_GCIB_Applications_Equipment.pdf
Дата изменения: Fri Oct 7 23:06:44 2005
Дата индексирования: Mon Oct 1 22:44:43 2012
Кодировка:
Nuclear Instruments and Methods in Physics Research B 206 (2003) 830­837 www.elsevier.com/locate/nimb

Gas cluster ion beam applications and equipment
Allen Kirkpatrick
*
Epion Corporation, 37 Manning Road, Billerica, MA 01821, USA

Abstract Interactions of gas cluster ions with solid surfaces, and prospective applications for those interactions, have been research topics of interest for several years. Substantial development efforts have been directed towards the problems associated with generating and transporting cluster ion beams having intensities sufficient to allow utilization in practical processing equipment. Unique smoothing, etching and chemical transformation effects which can be produced by cluster ion beams have been applied to a broad range of materials. Original gas cluster ion beam (GCIB) equipment which provided nanoampere level beam currents has evolved into automated milliampere level machines offering excellent process throughput, uniformity and reproducibility over large area targets. This paper will review cluster ion interactions, example applications and the present status of commercial GCIB processing equipment. с 2003 Elsevier Science B.V. All rights reserved.
PACS: 36.40.c; 41.75.Ak; 68.35.Ct; 68.55.Jk Keywords: Gas; Cluster; Ion; GCIB; Applications; Equipment

1. Introduction Over the past decade, a great deal of work has been performed to investigate fundamental interactions of energetic gas cluster ions with solid targets and to identify applications of these interactions for useful processing of materials [1­5]. It has been established that gas cluster ion beam (GCIB) processes are capable of producing surface modification effects that are distinctly different from those produced by all other known ion beam and plasma methods. Initial applications being recognized for GCIB involve atomic level process challenges where GCIB is able to offer significant advantages over all available alternatives. In con-

junction with the fundamental work and applications studies, substantial engineering efforts have been directed towards resolving problems associated with generating and transporting intense cluster ion beams. Production GCIB machines are now available and GCIB techniques are beginning to be integrated into the manufacturing processes for various advanced microdevice products from a number of industries. This paper describes a few example applications for GCIB and discusses present status of GCIB as a manufacturing technology. 2. Discussion 2.1. Generation of gas cluster ion beams Fig. 1 shows a schematic configuration that is typical of most of the GCIB equipment now being

*

Tel.: +1-978-670-1910x212; fax: +1-978-670-9119. E-mail address: akirkpatrick@epion.com (A. Kirkpatrick).

0168-583X/03/$ - see front matter с 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0168-583X(03)00858-9


A. Kirkpatrick / Nucl. Instr. and Meth. in Phys. Res. B 206 (2003) 830­837

831

Fig. 1. Schematic configuration of GCIB processor.

used for research and developmental and also for production. Depending upon the application, clusters can be produced from many different gases, including Ar, O2 , N2 and mixtures of these gases with various reactive components such as CF4 or SF6 . Neutral clusters are formed by expansion of the source gas at high pressure through a supersonic nozzle into vacuum. The directed axial stream of clusters emerging from the nozzle is allowed to pass through an aligned skimmer which blocks downstream transmission of most of the excess gas. Clusters passing through the skimmer to a second vacuum stage are ionized by electron bombardment and then accelerated to high potential, typically a few kilovolts to a few tens of kilovolts, depending upon the intended use for the beam. Magnetic filtering of the extracted ion beam to eliminate monomer ion contamination results in a beam comprised only of cluster ions with size distribution ranging from a few hundred to several thousand atoms. A neutralizer assembly injects low energy electrons into the beam so as to minimize space charge blow-up and to prevent charge build-up on nonconductive targets. Mechanical scanning is employed for uniform processing of the target substrates within a third vacuum stage. A Faraday current monitor is used for dose control. Systems typically are designed to interface with optional robotic substrate handlers and cassette-to-cassette loadlocks. Fig. 2 shows a photo-

Fig. 2. Epion US50M Ultra-SmootherБ processor.

graph of a GCIB Ultra-Smoother processor for substrates up to 200 mm in diameter. GCIB tools for 300 mm semiconductor substrates are presently being designed. 2.2. Cluster ion surface interactions Fig. 3 shows a simple representation of cluster impact. During transport toward the target, an


832

A. Kirkpatrick / Nucl. Instr. and Meth. in Phys. Res. B 206 (2003) 830­837

Fig. 3. Cluster ion impact.

accelerated cluster ion typically has total energy of several keV or more shared by its hundreds or thousands of weakly bound constituent atoms. Upon impact at the target surface, because the energetic cluster has large mass and only low energy per atom, penetration of cluster atoms into the target surface is determined not by the high total energy of the cluster but rather by the low energies of the individual atoms. At the moment of impact, the high total energy carried by the cluster is transferred into the small impact volume which has dimensions of the order of tens of Angstroms. As a result, extreme temperature and pressure transients are generated within the impact volume at the moment when the atoms of the cluster and atoms of the target are intimately mixed. These transient temperature/pressure conditions contribute to very strong enhancement of chemical reactivities and to the high sputtering yields and lateral directionality of ejected target material which are well-recognized characteristics of cluster ion interactions. From inspection of the schematic which was shown in Fig. 1, it can be recognized that there are similarities between the configuration of a GCIB processor and the configuration of ion implantation machines used by the semiconductor industry.

GCIB provides a highly collimated directional ion beam, precise mechanical scanning with closed loop feedback control, and process parameters defined in terms of source species, acceleration voltages and ion doses, so as to offer the same degree of process uniformity and reproducibility that is normally associated with ion implantation. However, the interactions of cluster ions at the target surface are completely dissimilar from those of the monomer ions delivered by ion implanters. If GCIB is performed using only an inert source gas such as Ar, the cluster ion impact upon the target material produces ballistic sputtering effects which are found to be useful for atomic level smoothing and cleaning. If the cluster source gas includes species such as O2 , N2 , CF4 , SF6 or F2 , which are capable of reacting with the target, then, because of the local temperature and pressure transients produced during cluster impact, exceptionally effective chemical interactions can occur between atoms of the clusters and atoms of the target. In this mode of operation, reactive GCIB processing is by a highly directional energetic chemical beam which can produce precisely uniform, controllable and efficient, anisotropic, high-rate etching. Fig. 4 shows the results of one experiment to compare depth of etching into various materials caused by 1 б 1015 ion/cm2 doses of 20 keV cluster ions produced from pure Ar, Ar with a few percent added CF4 and Ar with a few percent added SF6 . It is interesting to note that for many of the materials of the test, rate of surface removal due to cluster ion bombardment increased by an order of magnitude or more when a small amount of halogen compound was added to the Ar source gas. 2.3. Applications for GCIB Early investigations of cluster ion impact resulted in recognition of an inherent smoothing effect which is attributed primarily to the lateral directionality of target atoms ejected from the impact site. First GCIB production tools have been designed as Ultra-SmootherБ processors to employ the smoothing effects of Ar cluster ions to cause atomic scale improvement of magnetic film layers of GMR and TMR hard disk heads for data


A. Kirkpatrick / Nucl. Instr. and Meth. in Phys. Res. B 206 (2003) 830­837

833

Fig. 4. Example etch rate comparisons.

storage. As an example, Fig. 5 shows atomic force microscope (AFM) images from a sputtered nickeliron film smoothed by GCIB using argon. It has been found that GCIB processing can enhance the surface quality of almost any material. GCIB smoothing typically requires removal of only small amounts of surface material and is most

effective for producing superior surfaces on materials which already have relatively good initial surfaces. Fig. 6 shows AFM images of a representative example, in this case a CaF2 optical surface. The initial surface of Fig. 6 is typical of many materials polished to the limits of conventional techniques such as chemical mechanical

Fig. 5. AFM images of sputtered NiFe film smoothed by GCIB.


834

A. Kirkpatrick / Nucl. Instr. and Meth. in Phys. Res. B 206 (2003) 830­837

Fig. 6. AFM images of CaF2 surface smoothed by GCIB.

polishing (CMP). Although the initial surface exhibits numerical roughness of only a few Angstroms, that surface still contains a high density of shallow scratches and other defects. After smoothing by Ar cluster ions, the numerical RMS roughness or average roughness value of the surface has been changed only slightly, and in some instances might be changed almost not at all, but the true quality of the surface has clearly been very substantially improved. This can be a factor of major, and often unrealized, importance in microdevices where the device performance and/or production yield are contingent upon atomic level quality of critical surfaces. For these applications, the role of GCIB can sometimes be considered as surface homogenization rather than smoothing. GCIB smoothing is found to have particular merit for a range of photonic applications. Fig. 7 shows AFM images of a glass substrate used for narrow bandpass DWDM filters employed in telecommunications. Polishing of the glass substrate to the limits achievable by CMP leaves the glass surface with extremely small local asperities having heights of the order of tens of Angstroms. During subsequent deposition of more than one hundred optical film layers to form the filter stack, the asperities on the substrate surface become replicated and amplified through the film stack to cause scattering, significant performance degradation and substantial production yield losses. Ar-GCIB

smoothing of the glass substrate to eliminate surface asperities prior to deposition of the film layers prevents development of scattering centers, results in optimization of the bandpass behavior and increases product yields. Similar advantages can be expected for other multi-layer interference coatings and for critical surfaces on many aspheric elements, X-ray mirrors, EUV mirrors and masks, etc. GCIB-assisted deposition of optical thin films is presently also being investigated as a possible approach to producing films with lower stress, superior grain structure and fewer included defects. Until recently, GCIB equipment has not been able to offer sufficient throughput to satisfy minimum requirements of the semiconductor industry, but that situation is changing. As a result of progress in GCIB source technology, high throughput GCIB processors for semiconductors are now being designed. As this new equipment becomes available, GCIB can take on new roles in the manufacturing of semiconductor devices. GCIB will offer uniform and reproducible smoothing, cleaning and etching of virtually any material. Reactive GCIB can be performed so as to etch without causing roughening or introducing subsurface damage. Because a directional reactive cluster ion beam can cause a totally anisotropic chemical etching action and can involve formation of byproducts


A. Kirkpatrick / Nucl. Instr. and Meth. in Phys. Res. B 206 (2003) 830­837

835

Fig. 7. GCIB smoothing of glass substrate used for DWDM filters.

that are volatile at room temperature, GCIB can be readily used for cleaning or etching of the bottoms of deep well structures. As an example, Fig. 8 is an SEM image showing GCIB removal of a Cu diffusion barrier from the bottom of a contact trench without altering the barrier protection along the trench sidewall. GCIB can be used for very uniform and reproducible etching over full area substrates. Example applications for this capability include reducing thickness of the active silicon layers of silicon-oninsulator materials. Other applications are facilitated by an optional mode of operation of available GCIB equipment which allows the control system to modulate either the beam parameters or the transport speed of the substrate being processed so as to produce a controlled intentionally nonuni-

form etching process over the substrate surface. The system microprocessor recognizes instantaneous position of the cluster beam on the target substrate. The cluster beam itself is typically of the order of a centimeter in diameter at the target plane and etching rates of any particular material can be precisely determined. In conjunction with maps of the desired results, the system controller is able to perform the spatially nonuniform processing for each substrate on an individual basis. Fig. 9 shows an example result of using spatially variable GCIB etching to reduce thickness of an SiO2 layer on a 200 mm silicon wafer from 526 nm to 347 nm while simultaneously reducing nonuniformity of the oxide thickness over the entire wafer from ф87 A Example applications for locato only ф3 A A. tion specific GCIB etching include correction of

Fig. 8. GCIB Etching of the Bottom of Contact Well.


836

A. Kirkpatrick / Nucl. Instr. and Meth. in Phys. Res. B 206 (2003) 830­837

Fig. 9. Location specific GCIB etching of SiO2 film.

Fig. 10. Maximum throughput of 200 mm wafers versus process dose.

nonuniformities in critical device layers and improving operational performance distributions of SAW devices by GCIB timing prior to dicing. 2.4. GCIB manufacturing throughputs Throughout development of GCIB technology to date, a parameter of concern has always been the cluster ion beam current. Production processes generally require high throughputs in order to be

economically viable. To allow GCIB to become a practical manufacturing technique, sufficiently high beam currents must be made available. A great deal of effort over a period of several years has gone into increasing GCIB currents from the few tens of nanoamperes achieved in the earliest development apparatus at Kyoto University. First commercial GCIB production processors introduced early in 2000 provided beam currents on target of about 50 lA. By mid 2001, systems able


A. Kirkpatrick / Nucl. Instr. and Meth. in Phys. Res. B 206 (2003) 830­837

837

to deliver 200 lA onto target were available and 400 lA systems are commercially available today. Milliampere cluster ion beams are now being achieved on engineering development apparatus and development efforts are underway toward multi-milliampere beams. It is important to realize that these are now intense beams. A milliampere beam of clusters with average cluster size of 2000 atoms carries the mass equivalent of a two ampere monomer ion beam. The first GCIB production processor was designed to be used for atomic scale smoothing of the surfaces of thin film layers in hard disk head devices. At typical process ion doses of 1 б 1015 ions/ cm2 or more, these first 50 lA systems could process only a few small substrates per hour. With the subsequent progress toward ever-increasing beam currents, there has also been progress towards more efficient processes employing different beam energies and incorporating reactive species into clusters which previously had usually been comprised only of inert argon. Dose requirements for prospective applications such as microscale cleaning, damage-free etching, surface homogenization, surface oxidation and nitridation, and assisted growth of films now range from a minimum of about 1 б 1013 ions/cm2 to beyond 1 б 1015 ions/ cm2 . Fig. 10 shows estimated throughput in equivalent 200 mm diameter substrates per hour versus required ion dose for several representative levels of beam current on target. The calculation of equivalent 200 mm wafers per hour assumes 70% ion utilization efficiency as a result of losses associated with beam overscan and substrate exchange. For applications which will require process doses of 1 б 1015 ions/cm2 or less, GCIB systems able to

deliver milliampere beams will be able to satisfy throughput requirements of the semiconductor industry. Continuing advances to higher beam currents and more efficient processes will result in additional applications for GCIB.

3. Summary GCIB is emerging as a useful new processing tool for applications in data storage, photonics and semiconductors. GCIB production processors are commercially available. Earliest production applications have been limited primarily to high added value microdevices such as hard disk heads and optical filters for telecommunication. Advances in machine technology and process integration are making GCIB available for applications requiring much higher throughputs. The unique characteristics of GCIB are expected to contribute significantly to advances in the future devices of many industries.

References
[1] I. Yamada, Radiat. Eff. Def. Solids 124 (1992) 69. [2] I. Yamada, W.L. Brown, J.A. Northby, M. Sosnowski, Nucl. Instr. and Meth. B 79 (1993) 223. [3] I. Yamada, J. Matsuo, Z. Insepov, M. Akizuki, Nucl. Instr. and Meth. B 106 (1995) 165. [4] Z. Insepov, I. Yamada, Nucl. Instr. and Meth. B 112 (1996) 16. [5] N. Toyoda, J. Matsuo, I. Yamada, in: T.X. Denton, J.L. Duggan, I.L. Morgan (Eds.), Proceedings of Applications of Accelerators in Industry Conference, AIP Press, 1997, p. 483.