Final Report Summary - TECHNOTUBES (Technology for wafer-scale carbon nanotube applications)
The TECHNOTUBES project has developed the world's first 12'' wafer-scale growth system for carbon nanotube (CNT)s. In addition, the project contains the full development chain, starting from nanotube growth, passing through material and process development, to applications development and prototype products (X-ray imagers).
The 12'' wafer system is a scaled up version of previous 4 and 6'' systems. It incorporates a 3-zone graphite wafer heater allowing temperature uniformity, a showerhead gas distribution system, and a top-side remote plasma system. The wafers are loaded by batch loading device which allows loading of preheated wafers at up to 400 degrees Celsius.
The cost of ownership has been calculated from the actual wafer handling capacity and speed. The machine achieves 1 wafer per 30 minutes, and with the capital cost over 10 years and process costs, this gives EUR 21 per wafer or USD 0.13 / inch2 of wafer. This is a factor 7 below the original objective. 60 % of the cost is capital. If depreciation is faster, over 5 years, the costs increase to EUR 35 per wafer.
In-situ monitoring tools were developed, including a laser interferometer for monitoring the instantaneous growth rate. Various ex-situ monitors were developed, including process control monitor (PCM)s and Raman. Process uniformity across the wafer is under 5 %; it is dominated by temperature uniformity.
Work package (WP) 3 developed a range of process recipes. For vertically-aligned single wall nanotube forests, a growth rate of 20 mm / hr was achieved at a process temperature of 750 degrees Celsius, and 5.7 mm / hr at 650 degrees Celsius. For interconnects and Vias, the density of such forests was increased 30-fold, from the previous literature values of 5x1011 to 1.4x1013 cm-2. In situ X-ray photoelectron spectroscopy (XPS) was used to develop growth on metallic surfaces. A process flow was developed for growth of CNTs in high aspect ratio via structures, for use in 'through silicon vias', as needed for three-dimensional integration in future integrated circuits. The complementary side of this is the development of a chemical mechanical polishing process prior to adding the top metal contacts.
WP5 was developing high current density field emission electron guns for use in X-ray imaging systems for health-care. Field emission is needed as it is the only technology that allows rapid switching, for time resolved tomography. This technology requires extremely large emission current densities because of the low contrast of body tissues to X-rays. The project was able to achieve high currents, but not as high as the application demands, under stable operating conditions.
WP6 is to develop similar high density field emission electron guns for vacuum microelectronics. The initial application was to be in travelling wave tube (TWT) microwave amplifiers. However, an amendment changed the application to the slightly easier target of use in X-ray sources for luggage scanners (security area). A prototype for this application is very successful.
WP7 developed CNTs for high sensitivity, very low power and low cost sensors. These are based to complementary metal-oxide-semiconductor (CMOS) electronics and hot plates. The power demand targets have easily been met.
WP8 developed surface based micro-fluidic devices for chromatographic applications, this part met all its deliverables. Finally, training was carried out at a number of European workshops, specifically at the annual IWEPNM meeting.
Project context and objectives:
The principle objective of the project is to build a 12'' wafer-scale growth system for CNTs, to define a process technology for the growth, and to use this develop specific applications for CNTs on surfaces, at the various end-user partners (WPs 4-8).
In detail, the objectives are to:
1. develop an automated 300 mm wafer scale high-throughput, CNT growth equipment with throughput of at least 2 wafers per hour per deposition module;
2. develop non-destructive ex-situ and in-situ characterisation techniques for process monitoring and material qualification, such as Raman, reflectometry and PCMs (test devices);
3. the equipment will have an industrial standard of process control, reliability, quality assurance, uniformity and repeatability expected by industrial end-users;
4. define process specifications and develop recipes for nanotube growth of defined location, diameter and height, on the necessary surfaces and within growth temperature limits of 400 - 750 degrees of Celsius;
5. develop robust time-modulated (100 Hz, 0.1 % duty cycle) cold cathode electron sources of over 1 A / cm2 current density, for compact X-ray sources for computer tomography (CT) X-ray scanners;
6. build a prototype of the time-resolved X-ray source;
7. develop an optically controlled X-ray tube with imaging demonstration;
8. develop processes for CNTs for very large scale integration (VLSI) CMOS interconnects with a resistance of 0.1 Ohm / via;
9. develop nanotube-enhanced surfaces as microfluidic chemical filters and bio-electrodes for chemical analysis, for use in lab-on-chip devices;
10. develop CNT thermal interface material that exhibits 'bulk' thermal conductivity of better than 10 W / mK;
11. develop CNT surfaces for low-cost (< EUR 2 /device), smart gas sensors on CMOS with ultra-low power (1 mW), high sensitivity, selectivity and capability of fast refreshing / re-use.
Note that the objective 7 was changed in year 2 after an amendment to description of work.
Project results:
WP1: High throughput equipment (AIX)
Objectives
- develop an automated 300 mm wafer scale CNT growth equipment, covering design, engineering, control systems, evaluate maintenance, reliability and life cycle cost.
WP1 is the heart of the project and is a considerable advance on the state of the art, as it existed in 2008. At that stage, there were no large area CNT growth machines suitable for integrating CNTs into microelectronics, which demands 12'' processability. In some labs, existing 12'' plasma-enhanced chemical vapour deposition (PECVD) or reactive-ion etching (RIE) had been adapted for this purpose.
The new machine is a development of machines that previously worked at the 50 mm and 150 mm scale. It consists of three modules, the reactor, the gas flow module and the control cabinet. In addition, is the batch loader.
The reactor chamber and the shower head were made by an external machine shop. This was delayed, which delayed WP1. However, this did not delay any other part of the project, because the other partners used smaller scale growth chambers for their work.
The wafers are loaded into the reactor chamber by the batch loader. It can transfer wafers at temperatures of up to 400 degrees Celsius. High temperature loading is critical to reduce cycle time per wafer.
D1.5: Cost of ownership
The cost of ownership, or total running costs divided by wafer throughput has been calculated in terms of capital cost, gases and electricity costs, and maintenance. It is based on two wafers per hour, per process module. Capital was amortised over 109 years, which might be too long. Nevertheless, this comes out as EUR 21 per wafer, or USD 0.13 per inch2. This is 6 times less than our target of USD 1 per inch2. This cost puts CNT at lower cost than many processes in microelectronics.
WP2: (ETH, TU Berlin, AIX)
Objectives
- to develop in-situ tools for process monitoring;
- develop techniques and metrology for quantifying the quality of CNT in production;
- to determine the standards for CNT quality.
Results
The key instrument developed in WP2 is the laser interferometer for the in-situ measurement of growth rate. The interference of light reflected from the top surface and the lower CNT-substrate interface create a series of fringes, whose amplitude is measured by a photometer. Given the light wavelength, this allows us to deduce the height. The amplitude of the fringes decreases as the height increases. Continued data can be collected by using a modulated light beam and phase sensitive detection. Nevertheless, this method does not work for the highest forests (5 mm).
The nanotube quality and type was measured ex-situ in various ways. The low frequency radial breathing mode (RBM)s seen in the Raman spectra are used to determine the chirality of single walled nanotube (SWNT)s. TUB determined the Raman cross sections of these modes as a function of chiral angle.
Assessments of CNT purity are provided by PCMs. These are simplified electronic devices built on an electrode mesh structure. Ferritin (Fe catalyst) is used to grow sparse CNT arrays across the meshes. Electrical resistance data are taken between pairs of electrodes. The statistics can be used to monitor growth quality and reproducibility of conditions.
WP3: Process development (UCAM, IMEC, FHI, CNR)
Objectives
- develop 300 mm deposition and activation method for CNT growth catalyst in relevant test structures for the prototype applications;
- develop low temperature, back-end compatible processes for specific types of CNT in relevant test structures for the prototype applications;
- develop high growth rate processes for 5 mm high nanotube mats and surface area of 600 m2 / gm (super-growth).
Results
High growth rate: A very fast growth mode was discovered in 2004 by Hata et al. for vertically-aligned CNT forests. UCAM and AIX were able to reproduce this on their chemical vapour deposition (CVD) reactors. Using lateral images by digital photographs of the forests, the growth rates have been extracted to be up to 20 mm / hr at 750 degrees Celsius. This is 30 times the objective.
There is a large demand for growth of high area density CNT forests, for use in interconnects for VLSI integrated circuits, heat spreaders, supercapacitors, and similar uses. The typical forest as grown by Hata et al. or Zhong et al. has an area density of about 6x1011 cm-2. This is quite high, but it does not fill space so efficiently. Compared to a fully dense array in which the CNTs a spaced by the inter-layer spacing of graphite, it is only 5-6 % filling. It is possible to increase the area density, either by increasing the filling fraction, or by decreasing the diameter.
Two basic ways were invented by UCAM to increase the area density. The first was to improve the catalyst design, by processing the support layer underneath. This was used to reduce the average CNT diameter from a typical 2.4 nm towards a value of 1.1 nm. This allowed us to increase the area density to 1.4x1013 cm-2, a 30-fold increase on the previous best.
The second way was to use cyclic deposition of the catalyst, to increase the density of catalyst nanoparticles. This increases the CNT density because roughly one nanoparticle germinates one CNT. In this case, the CNT diameter stays at about 2.4 nm, but the density increases to 1.0 - 1.3x1013 cm-2, again a considerable increase in density.
The densities are obtained by measuring the weight gain after growth, measuring the CNT diameter, and number of walls, to that the weight per unit of the nanotube is known, and thus the area density can be derived.
A third method, previously invented by Toshiba, was also found to increase the density. This involved immobilising the catalyst nanoparticles to stop them sintering. This was achieved in a three-step process, where a short carburisation step did the immobilisation.
Growth on non-conducting surfaces (UCAM, FHI, CNR)
As a baseline, we first characterised the catalyst behaviour on insulating Al2O3 and SiO2 supports. Fe performs much better on Al2O3 than on SiO2 because there is a catalyst-support interaction on Al2O3 which helps bind the Fe and inhibit surface diffusion. On the other hand, XPS shows that for Co on SiO2, there is some formation of Co-Si bonds, which can help to bind Co particles, and reduce surface diffusion, at least below 700 degrees Celsius.
Many of the applications of CNTs, especially in electronics, are for nanotubes on conducting surfaces. However, the standard growth process involves deposition of catalyst onto an oxide, because the low surface energy of the oxide leads to a de-wetting of the catalyst metal, and resultant formation of the CNT when the growth gas is turned on. The problem is, most of the oxides are insulating, particularly the best catalyst support Al2O3. Therefore we expended considerable effort to design supports that were conducting. A related issue is that metallic supports can easily be oxidised due to the background oxygen in the vacuum system. To study this effect, we carried numerous in-situ growth XPS experiments to discover the actual state of the catalyst and support layer during each stage of catalyst pre-treatment and growth, using the BESSY synchrotron or systems at CNR Trieste.
Three conductive support layers were studied, Ta, TiN and CoSi2. Ta is a refractory metal widely used in microelectronics as a diffusion barrier against Cu. Its oxide Ta2O5 forms exothermically, but with a large activation energy due to the strong bonding of Ta itself. In-situ XPS and ex-situ XRD data showed two limitations to its use, with Fe as catalyst. First, Ta can form a very thin protective amorphous layer of Ta2O5 on the underlying metal. The overall layer remains conductive. But the Ta must be deposited with a certain polycrystalline structure, overwise the whole Ta layer rapidly converts into an insulating oxide. Second, if growth is carried out above approximately 600 degrees Celsius, the thin amorphous Ta oxide turns polycrystalline. The resulting grain boundaries allow the Fe to diffuse rapidly away from the surface, so growth stops.
TiN is also a metallic diffusion barrier layer. We have found it possible to grow CNTs using Fe catalysts on TiN, if plasma pretreatment (PP) is used to increase adhesion of the Fe to the TiN. On the other hand, other groups in a different project found that TiN inhibits growth. TiN can be oxidised to TiO2 in adverse cases.
CoSi2 is a useful conductive support layer. We obtained growth on CoSi2 using Fe as catalyst. The CoSi2 resists oxidation. CoSi2 can in some cases also appear to act as a catalyst. In fact what occurs is that Co excess as separated and this is the actual catalyst. This can be used to provide a self-aligned catalyst process, by using SF6 RIE to remove Si from CoSi2.
More details of this work can be found in published papers.
CNR carried out many XPS studies to understand the various different catalyst preparations. PP increases the surface binding of the catalyst to the support oxide, which explains why this can be used to enforce the root growth mechanism, and thus enforce growth of forests, in cases where spaghetti (non-forest) growth might occur. For example on TiN.
WP4: Microelectronics (IMEC, UCAM)
This WP focused mainly on the development of processes to incorporate CNT functions in microelectronics. This includes through-silicon vias (TSVs), making the top contact to such vias, and as thermal interface materials. Much relevant process development is summarised in WP3.
The focus is on benchmarking the optimised CNT integration process with the performance of Cu contacts at a relevant technology node. In order to allow for reliable benchmarking, vertical CNT interconnects were integrated in 150 nm diameter contact holes using CMOS compatible processes on 200mm wafers.
The transfer from 300 nm contact holes to 150 nm contacts (see D43) was needed in order to allow for full wafer integration, metallisation, and electrical probing. Only in this way, the electrical properties of the CNT interconnects could be determined in a similar way as done for Cu and, hence, compared directly with Cu interconnects.
The 150 nm CNT contacts are integrated using a technology that is compatibility with the 130 nm technology node on 200 mm full wafers and can be benchmarked using automated electrical characterisation. The electrical benchmarking indicate that the resistivity of the 150 nm CNT contacts metallised with a Cu single damascene top contact are still 3 orders of magnitude higher that their Cu counterparts. Compared to other values available in the literature on CNT contacts, the CNT integration process achieved here uses more aggressive contact hole dimensions. Nevertheless, that the CNT contact resistivity is found to reduce upon using such smaller contacts with aspect ratio approximately two, which suggest a significant improvement in the CNT integration process was realised.
The structure designed, the process flow presented here for CNT integration, and the automatic probing constitute a platform that allows benchmarking different recipes and process conditions. This constitutes a significant step forward towards the realisation of a contact modules with vertical CNT interconnects of superior quality because it can speed up the learning cycle time for optimising the CNT interconnect based on learning from the electrical performance.
WP5: X-ray imager cathodes (Philips)
Objective
Philips are to develop CNT field emission electron guns for compact, time-resolved X-ray sources for X-ray tomography in medical imaging. X-ray tomography builds up a three-dimensional (3D) image of the patient from a series of projections. To image a moving 3D object such as a beating heart, the X-ray tube and its detector must be rapidly rotated or the projections must be collected by switching a distributed set of stationary sources. Thermionic sources cannot be switched fast enough, because of thermal time constants. Thus, usually a single source is rotated. However, heavy metal anodes limit rotation speeds, by structural integrity under centrifugal forces. Thus, time-resolved X-ray CT sources today operate at the technological limits. (see e.g. Philips patent WO 07088497).
Field emission cathodes allow the rapid modulation of emission current, and thus time-resolved X-ray beam. images can be generated by an array of stationary X-ray sources instead. Field emission electron sources using CNT arrays are ideal candidates for this and can generate very high frequency pulsed output, with a high peak current density. Emitters need to:
- allow for fast switching and pulsed mode operation at high frequency;
- emit high absolute electron currents of 0.2-2 A from 1 cm2 preferably at DC, but at least in pulsed mode;
- operate with high stability and reliability in UHV and poor vacuum conditions;
- be produced using a cost-effective wafer-scale production technology.
Results
These high current densities are difficult to achieve. At the beginning of the project, Philips used CNT arrays grown by their own recipe. The project aimed to increase the current density and also move to using CNT arrays grown on Aixtron based CVD systems, for scale up of supply.
The emitter arrays are 5 mm x 5 mm active array areas with uniform CNT bundle lengths of about 6-8 µm, and bundle sizes of 3 µm x 3 µm at a pitch of 15 µm. The CNTs are grown from Fe catalyst on Al, Cu, Ti, TiN stacks on Si wafers.
Overall, more than 500 emitter arrays were processed. The initial stacks had less than expected emission currents. This led to a study where diffusion in the stack was found. Changes to the stack design resulted in Mo replacing Cu and Ti layers, and eventually to Mo disks as substrates. This evaluation and optimisation is still ongoing.
The emission currents later are of order 100 mA at an applied field of 5-8 V / um for an active area of 5 mm x 5 mm. Pulsed emission currents of order 200 mA were achived at 10 Hz. While this nominally satisfies one of the deliverables D5.1 the reliability of this was insufficient for ultimate industrial use. Aging effects were looked into.
Ultimately, a reproducible current of 50 mA from 0.25 cm2 was achieved with the Mo stack. This was too low to justify the incorporation of the emitter stacks into a prototype device, as envisaged in the contract.
WP6: Vacuum microelectronics (Thales)
This WP aims to develop CNT field emission electron tubes. In the initial project, the application was TWT microwave amplifiers. Because the emission current did not reach high enough values for TWTs, the application was changed at the mid-term review to time-modulated X-ray tubes for industrial and security applications. In contrast to the X-ray body scanner being developed by Philips which requires intense X-ray pulses at moderate voltages (to enhance the contrast between soft tissues), object scanning needs smaller DC currents but higher voltages (for example luggage has higher atomic number contrast, but their imaging needs greater penetration depth), so that this last application appears well adapted to CNT field emission guns.
Field emission guns are needed compared to thermionic guns to allow time modulation; thermionic sources are too slow. Compared to conventional X-ray tubes based on thermionic cathodes, X-ray tubes based on CNT cathodes allow time modulation.
TRT has developed a patented photocathode, based on P-I-N diodes and CNT emitters. The emitter runs at low voltage behind a gate electrode. Without light, the reversed biased P-I-N diodes are in the off state, and there is a low emission current. When illuminated, the diode is on, and the current delivered by the diode can be emitted by the CNTs. The key point for X-ray tubes is that the optical control allows the cathode to be floated at high potential, so that the anode, which is bulky and needs cooling, can be operated at ground potential. A great advantage of this configuration is to simply the high voltage power supply for multi-source systems.
Thus WP6 has produced a product (full development in approximately two years) starting from the CVD equipment of WP1.
WP7: Sensors (CCS and ETH)
Objectives
- to develop low power CMOS based gas sensors integrating CNT;
- to develop nanotube field effect transistor (FET) based gas sensor, and study of mechanism.
ETH developed FET gas sensors, while CCS developed low power, mat sensors using their hotplate technology. ETH developed a scalable FET sensor technology, based on the PCM technology of WP2. ETH demonstrated a prototype CNT FET gas sensor with sensitivity of below 100 ppm for NOx in humid air. In dry air a limit of 50 ppb was attained.
There was also joint work of FET sensors incorporated into the CCS CMOS drive circuits.
CCS
The majority of the CCS work is described in deliverables 7.2 7.3 7.4 and 7.5. The main advantage of CCS technology is the low power demand of their CMOS based sensor. The sensor also places the sensor elements on 'hot plates' which are metal plates with underwired heater elements
WP8: Surface based devices (DTU)
Objectives:
- develop nanotube-enhanced surfaces as microfluidic chemical filters and bioelectrodes for chemical analysis, for use in lab-on-chip devices.
Electrically insulated microfluidic channels for electroosmotic pumping (EOF) of aqueous solutions are fabricated. CNTs are too conductive to be patterned homogeneously in the whole flow channel. The potential difference between the aqueous buffer solution and the CNTs becomes too high, so that bubble formation due to electrolysis of water occurs. The bubbles in the channel terminate the electroosmotic flow, and thereby prevent the device from working properly. In the case of a continuous layer of CNTs in the fluidic channel, a maximum field strength of around 100 V / cm can be applied, which is too small for most applications.
Silicon nanograss technology was also used to create highly hierarchical, three dimensional CNT forests, in order to gain maximal control of surface tension and the active interfacial contact between the chip and cells, as well cell adhesion, proliferation and electrochemical activity.
PDMS is a biocompatible, chemically inert and electrically insulating material, and therefore was chosen as the insulation layer, protecting TiW base and exposing the CNT tips only.
WP9: Training, dissemination (UCAM)
Training was carried out at a workshop / conference holed in Kirchberg, Austria, in 3-10 March 2012, the '26th international workshop on electronic properties of novel materials' or IWEPNM 2012. This is chaired by Prof. C. Thomsen, the project leader from TU Berlin.
Potential impact:
The first 6-inch wafer scale CNT system was developed not in US or Japan but in Europe by Nano-instruments UK in 2005, a spin-off from Cambridge University. Nanoinstruments was acquired by AIXTRON in 2007. The basis of this project is to develop large wafer-scale CNT equipment, and then progress several CNT applications which require large area capabilities. This 'development-chain' ensures that there will be a future applications market for CNT equipment, which would in turn fulfil the production requirements of the applications. The close-loop interaction between research institutes, enabling equipment manufacturer, and end users ensures that this project will establish Europe's competitiveness and create employment at all levels in the value chain of the CNT-manufacturing industry.
CNTs are an enabling material for sub-22nm on-wafer interconnect technology, through-wafer interconnects, strengthening solder bumps and efficient thermal interface materials; the impacts of these microelectronics applications are to lead to higher device densities, higher levels of integration and improved reliability.
The development of wafer scale, mass production CNT equipment will expand the market for CNT deposition systems. AIXTRON expects a significant increase in the demand for systems such as that developed in this project, and as the applications that this project targets and other uses of CNTs ramp up, the expected equipment market for CNT in 2012 is expected to exceed EUR 50 million. This project will strengthen AIXTRON's position as a major equipment manufacturer and will result in the creation of jobs at AIXTRON, as well as end-user / customer sites.
Thales Electron Devices is a major supplier of electron tubes and related subsystems, which has a present market size which is EUR 1700 million, comprising microwave technologies (EUR 1000 million) and X-rays (EUR 700 million).
The development of the X-ray imager as used for security screening as in WP5 will enable the development of products in the 2 - 3 timeframe.
Philips has recently transformed itself from a firm focused on electronics components into a firm focused on three areas, one of which is health care. It shares the X-ray imager market with GE, Siemens and Toshiba. This project will continue to keep Philips at the forefront of this market. The sales figures confirm that this area is a European specialty, Siemens plus Philips is larger than any other. This will have a significant effect on employment numbers within Philips healthcare.
The environmental market is however expected to grow exponentially in the next few years, with the introduction of miniaturised gas-sensors in mobile phones, PDAs, homes or car cabins. Cambridge CMOS Sensors targets two areas with its unique low-cost, high performance technology: the environmental (hydrogen, NOx, CO) and medical sectors. With CCS technology which combines smart hot-plates with ultra-sensitive CNTs and monolithic integration of read-out and processing electronics, we expect an order of magnitude reduction in power consumption to 10 mW continuous power at 400 degrees Celsius, or 1 mW pulsed power, plus an order of magnitude reduction in system cost to well under USD 10 for an integrated gas-sensor unit achieved through miniaturisation. The low sensor cost fuels its uptake, increasing its demand and creating jobs throughout the sensor manufacturing chain. The key point is that the 1 mW power consumption allows the sensors to be incorporated into wireless and hand held devices, which the existing 200-800 mW devices do not. This allows sensors to be incorporated into cell phones, Personal digital assistant (PDA)s, so that any consumer can have his/her own local pollution monitor for use anywhere.
Project website: http://www.technotubes.eu
The 12'' wafer system is a scaled up version of previous 4 and 6'' systems. It incorporates a 3-zone graphite wafer heater allowing temperature uniformity, a showerhead gas distribution system, and a top-side remote plasma system. The wafers are loaded by batch loading device which allows loading of preheated wafers at up to 400 degrees Celsius.
The cost of ownership has been calculated from the actual wafer handling capacity and speed. The machine achieves 1 wafer per 30 minutes, and with the capital cost over 10 years and process costs, this gives EUR 21 per wafer or USD 0.13 / inch2 of wafer. This is a factor 7 below the original objective. 60 % of the cost is capital. If depreciation is faster, over 5 years, the costs increase to EUR 35 per wafer.
In-situ monitoring tools were developed, including a laser interferometer for monitoring the instantaneous growth rate. Various ex-situ monitors were developed, including process control monitor (PCM)s and Raman. Process uniformity across the wafer is under 5 %; it is dominated by temperature uniformity.
Work package (WP) 3 developed a range of process recipes. For vertically-aligned single wall nanotube forests, a growth rate of 20 mm / hr was achieved at a process temperature of 750 degrees Celsius, and 5.7 mm / hr at 650 degrees Celsius. For interconnects and Vias, the density of such forests was increased 30-fold, from the previous literature values of 5x1011 to 1.4x1013 cm-2. In situ X-ray photoelectron spectroscopy (XPS) was used to develop growth on metallic surfaces. A process flow was developed for growth of CNTs in high aspect ratio via structures, for use in 'through silicon vias', as needed for three-dimensional integration in future integrated circuits. The complementary side of this is the development of a chemical mechanical polishing process prior to adding the top metal contacts.
WP5 was developing high current density field emission electron guns for use in X-ray imaging systems for health-care. Field emission is needed as it is the only technology that allows rapid switching, for time resolved tomography. This technology requires extremely large emission current densities because of the low contrast of body tissues to X-rays. The project was able to achieve high currents, but not as high as the application demands, under stable operating conditions.
WP6 is to develop similar high density field emission electron guns for vacuum microelectronics. The initial application was to be in travelling wave tube (TWT) microwave amplifiers. However, an amendment changed the application to the slightly easier target of use in X-ray sources for luggage scanners (security area). A prototype for this application is very successful.
WP7 developed CNTs for high sensitivity, very low power and low cost sensors. These are based to complementary metal-oxide-semiconductor (CMOS) electronics and hot plates. The power demand targets have easily been met.
WP8 developed surface based micro-fluidic devices for chromatographic applications, this part met all its deliverables. Finally, training was carried out at a number of European workshops, specifically at the annual IWEPNM meeting.
Project context and objectives:
The principle objective of the project is to build a 12'' wafer-scale growth system for CNTs, to define a process technology for the growth, and to use this develop specific applications for CNTs on surfaces, at the various end-user partners (WPs 4-8).
In detail, the objectives are to:
1. develop an automated 300 mm wafer scale high-throughput, CNT growth equipment with throughput of at least 2 wafers per hour per deposition module;
2. develop non-destructive ex-situ and in-situ characterisation techniques for process monitoring and material qualification, such as Raman, reflectometry and PCMs (test devices);
3. the equipment will have an industrial standard of process control, reliability, quality assurance, uniformity and repeatability expected by industrial end-users;
4. define process specifications and develop recipes for nanotube growth of defined location, diameter and height, on the necessary surfaces and within growth temperature limits of 400 - 750 degrees of Celsius;
5. develop robust time-modulated (100 Hz, 0.1 % duty cycle) cold cathode electron sources of over 1 A / cm2 current density, for compact X-ray sources for computer tomography (CT) X-ray scanners;
6. build a prototype of the time-resolved X-ray source;
7. develop an optically controlled X-ray tube with imaging demonstration;
8. develop processes for CNTs for very large scale integration (VLSI) CMOS interconnects with a resistance of 0.1 Ohm / via;
9. develop nanotube-enhanced surfaces as microfluidic chemical filters and bio-electrodes for chemical analysis, for use in lab-on-chip devices;
10. develop CNT thermal interface material that exhibits 'bulk' thermal conductivity of better than 10 W / mK;
11. develop CNT surfaces for low-cost (< EUR 2 /device), smart gas sensors on CMOS with ultra-low power (1 mW), high sensitivity, selectivity and capability of fast refreshing / re-use.
Note that the objective 7 was changed in year 2 after an amendment to description of work.
Project results:
WP1: High throughput equipment (AIX)
Objectives
- develop an automated 300 mm wafer scale CNT growth equipment, covering design, engineering, control systems, evaluate maintenance, reliability and life cycle cost.
WP1 is the heart of the project and is a considerable advance on the state of the art, as it existed in 2008. At that stage, there were no large area CNT growth machines suitable for integrating CNTs into microelectronics, which demands 12'' processability. In some labs, existing 12'' plasma-enhanced chemical vapour deposition (PECVD) or reactive-ion etching (RIE) had been adapted for this purpose.
The new machine is a development of machines that previously worked at the 50 mm and 150 mm scale. It consists of three modules, the reactor, the gas flow module and the control cabinet. In addition, is the batch loader.
The reactor chamber and the shower head were made by an external machine shop. This was delayed, which delayed WP1. However, this did not delay any other part of the project, because the other partners used smaller scale growth chambers for their work.
The wafers are loaded into the reactor chamber by the batch loader. It can transfer wafers at temperatures of up to 400 degrees Celsius. High temperature loading is critical to reduce cycle time per wafer.
D1.5: Cost of ownership
The cost of ownership, or total running costs divided by wafer throughput has been calculated in terms of capital cost, gases and electricity costs, and maintenance. It is based on two wafers per hour, per process module. Capital was amortised over 109 years, which might be too long. Nevertheless, this comes out as EUR 21 per wafer, or USD 0.13 per inch2. This is 6 times less than our target of USD 1 per inch2. This cost puts CNT at lower cost than many processes in microelectronics.
WP2: (ETH, TU Berlin, AIX)
Objectives
- to develop in-situ tools for process monitoring;
- develop techniques and metrology for quantifying the quality of CNT in production;
- to determine the standards for CNT quality.
Results
The key instrument developed in WP2 is the laser interferometer for the in-situ measurement of growth rate. The interference of light reflected from the top surface and the lower CNT-substrate interface create a series of fringes, whose amplitude is measured by a photometer. Given the light wavelength, this allows us to deduce the height. The amplitude of the fringes decreases as the height increases. Continued data can be collected by using a modulated light beam and phase sensitive detection. Nevertheless, this method does not work for the highest forests (5 mm).
The nanotube quality and type was measured ex-situ in various ways. The low frequency radial breathing mode (RBM)s seen in the Raman spectra are used to determine the chirality of single walled nanotube (SWNT)s. TUB determined the Raman cross sections of these modes as a function of chiral angle.
Assessments of CNT purity are provided by PCMs. These are simplified electronic devices built on an electrode mesh structure. Ferritin (Fe catalyst) is used to grow sparse CNT arrays across the meshes. Electrical resistance data are taken between pairs of electrodes. The statistics can be used to monitor growth quality and reproducibility of conditions.
WP3: Process development (UCAM, IMEC, FHI, CNR)
Objectives
- develop 300 mm deposition and activation method for CNT growth catalyst in relevant test structures for the prototype applications;
- develop low temperature, back-end compatible processes for specific types of CNT in relevant test structures for the prototype applications;
- develop high growth rate processes for 5 mm high nanotube mats and surface area of 600 m2 / gm (super-growth).
Results
High growth rate: A very fast growth mode was discovered in 2004 by Hata et al. for vertically-aligned CNT forests. UCAM and AIX were able to reproduce this on their chemical vapour deposition (CVD) reactors. Using lateral images by digital photographs of the forests, the growth rates have been extracted to be up to 20 mm / hr at 750 degrees Celsius. This is 30 times the objective.
There is a large demand for growth of high area density CNT forests, for use in interconnects for VLSI integrated circuits, heat spreaders, supercapacitors, and similar uses. The typical forest as grown by Hata et al. or Zhong et al. has an area density of about 6x1011 cm-2. This is quite high, but it does not fill space so efficiently. Compared to a fully dense array in which the CNTs a spaced by the inter-layer spacing of graphite, it is only 5-6 % filling. It is possible to increase the area density, either by increasing the filling fraction, or by decreasing the diameter.
Two basic ways were invented by UCAM to increase the area density. The first was to improve the catalyst design, by processing the support layer underneath. This was used to reduce the average CNT diameter from a typical 2.4 nm towards a value of 1.1 nm. This allowed us to increase the area density to 1.4x1013 cm-2, a 30-fold increase on the previous best.
The second way was to use cyclic deposition of the catalyst, to increase the density of catalyst nanoparticles. This increases the CNT density because roughly one nanoparticle germinates one CNT. In this case, the CNT diameter stays at about 2.4 nm, but the density increases to 1.0 - 1.3x1013 cm-2, again a considerable increase in density.
The densities are obtained by measuring the weight gain after growth, measuring the CNT diameter, and number of walls, to that the weight per unit of the nanotube is known, and thus the area density can be derived.
A third method, previously invented by Toshiba, was also found to increase the density. This involved immobilising the catalyst nanoparticles to stop them sintering. This was achieved in a three-step process, where a short carburisation step did the immobilisation.
Growth on non-conducting surfaces (UCAM, FHI, CNR)
As a baseline, we first characterised the catalyst behaviour on insulating Al2O3 and SiO2 supports. Fe performs much better on Al2O3 than on SiO2 because there is a catalyst-support interaction on Al2O3 which helps bind the Fe and inhibit surface diffusion. On the other hand, XPS shows that for Co on SiO2, there is some formation of Co-Si bonds, which can help to bind Co particles, and reduce surface diffusion, at least below 700 degrees Celsius.
Many of the applications of CNTs, especially in electronics, are for nanotubes on conducting surfaces. However, the standard growth process involves deposition of catalyst onto an oxide, because the low surface energy of the oxide leads to a de-wetting of the catalyst metal, and resultant formation of the CNT when the growth gas is turned on. The problem is, most of the oxides are insulating, particularly the best catalyst support Al2O3. Therefore we expended considerable effort to design supports that were conducting. A related issue is that metallic supports can easily be oxidised due to the background oxygen in the vacuum system. To study this effect, we carried numerous in-situ growth XPS experiments to discover the actual state of the catalyst and support layer during each stage of catalyst pre-treatment and growth, using the BESSY synchrotron or systems at CNR Trieste.
Three conductive support layers were studied, Ta, TiN and CoSi2. Ta is a refractory metal widely used in microelectronics as a diffusion barrier against Cu. Its oxide Ta2O5 forms exothermically, but with a large activation energy due to the strong bonding of Ta itself. In-situ XPS and ex-situ XRD data showed two limitations to its use, with Fe as catalyst. First, Ta can form a very thin protective amorphous layer of Ta2O5 on the underlying metal. The overall layer remains conductive. But the Ta must be deposited with a certain polycrystalline structure, overwise the whole Ta layer rapidly converts into an insulating oxide. Second, if growth is carried out above approximately 600 degrees Celsius, the thin amorphous Ta oxide turns polycrystalline. The resulting grain boundaries allow the Fe to diffuse rapidly away from the surface, so growth stops.
TiN is also a metallic diffusion barrier layer. We have found it possible to grow CNTs using Fe catalysts on TiN, if plasma pretreatment (PP) is used to increase adhesion of the Fe to the TiN. On the other hand, other groups in a different project found that TiN inhibits growth. TiN can be oxidised to TiO2 in adverse cases.
CoSi2 is a useful conductive support layer. We obtained growth on CoSi2 using Fe as catalyst. The CoSi2 resists oxidation. CoSi2 can in some cases also appear to act as a catalyst. In fact what occurs is that Co excess as separated and this is the actual catalyst. This can be used to provide a self-aligned catalyst process, by using SF6 RIE to remove Si from CoSi2.
More details of this work can be found in published papers.
CNR carried out many XPS studies to understand the various different catalyst preparations. PP increases the surface binding of the catalyst to the support oxide, which explains why this can be used to enforce the root growth mechanism, and thus enforce growth of forests, in cases where spaghetti (non-forest) growth might occur. For example on TiN.
WP4: Microelectronics (IMEC, UCAM)
This WP focused mainly on the development of processes to incorporate CNT functions in microelectronics. This includes through-silicon vias (TSVs), making the top contact to such vias, and as thermal interface materials. Much relevant process development is summarised in WP3.
The focus is on benchmarking the optimised CNT integration process with the performance of Cu contacts at a relevant technology node. In order to allow for reliable benchmarking, vertical CNT interconnects were integrated in 150 nm diameter contact holes using CMOS compatible processes on 200mm wafers.
The transfer from 300 nm contact holes to 150 nm contacts (see D43) was needed in order to allow for full wafer integration, metallisation, and electrical probing. Only in this way, the electrical properties of the CNT interconnects could be determined in a similar way as done for Cu and, hence, compared directly with Cu interconnects.
The 150 nm CNT contacts are integrated using a technology that is compatibility with the 130 nm technology node on 200 mm full wafers and can be benchmarked using automated electrical characterisation. The electrical benchmarking indicate that the resistivity of the 150 nm CNT contacts metallised with a Cu single damascene top contact are still 3 orders of magnitude higher that their Cu counterparts. Compared to other values available in the literature on CNT contacts, the CNT integration process achieved here uses more aggressive contact hole dimensions. Nevertheless, that the CNT contact resistivity is found to reduce upon using such smaller contacts with aspect ratio approximately two, which suggest a significant improvement in the CNT integration process was realised.
The structure designed, the process flow presented here for CNT integration, and the automatic probing constitute a platform that allows benchmarking different recipes and process conditions. This constitutes a significant step forward towards the realisation of a contact modules with vertical CNT interconnects of superior quality because it can speed up the learning cycle time for optimising the CNT interconnect based on learning from the electrical performance.
WP5: X-ray imager cathodes (Philips)
Objective
Philips are to develop CNT field emission electron guns for compact, time-resolved X-ray sources for X-ray tomography in medical imaging. X-ray tomography builds up a three-dimensional (3D) image of the patient from a series of projections. To image a moving 3D object such as a beating heart, the X-ray tube and its detector must be rapidly rotated or the projections must be collected by switching a distributed set of stationary sources. Thermionic sources cannot be switched fast enough, because of thermal time constants. Thus, usually a single source is rotated. However, heavy metal anodes limit rotation speeds, by structural integrity under centrifugal forces. Thus, time-resolved X-ray CT sources today operate at the technological limits. (see e.g. Philips patent WO 07088497).
Field emission cathodes allow the rapid modulation of emission current, and thus time-resolved X-ray beam. images can be generated by an array of stationary X-ray sources instead. Field emission electron sources using CNT arrays are ideal candidates for this and can generate very high frequency pulsed output, with a high peak current density. Emitters need to:
- allow for fast switching and pulsed mode operation at high frequency;
- emit high absolute electron currents of 0.2-2 A from 1 cm2 preferably at DC, but at least in pulsed mode;
- operate with high stability and reliability in UHV and poor vacuum conditions;
- be produced using a cost-effective wafer-scale production technology.
Results
These high current densities are difficult to achieve. At the beginning of the project, Philips used CNT arrays grown by their own recipe. The project aimed to increase the current density and also move to using CNT arrays grown on Aixtron based CVD systems, for scale up of supply.
The emitter arrays are 5 mm x 5 mm active array areas with uniform CNT bundle lengths of about 6-8 µm, and bundle sizes of 3 µm x 3 µm at a pitch of 15 µm. The CNTs are grown from Fe catalyst on Al, Cu, Ti, TiN stacks on Si wafers.
Overall, more than 500 emitter arrays were processed. The initial stacks had less than expected emission currents. This led to a study where diffusion in the stack was found. Changes to the stack design resulted in Mo replacing Cu and Ti layers, and eventually to Mo disks as substrates. This evaluation and optimisation is still ongoing.
The emission currents later are of order 100 mA at an applied field of 5-8 V / um for an active area of 5 mm x 5 mm. Pulsed emission currents of order 200 mA were achived at 10 Hz. While this nominally satisfies one of the deliverables D5.1 the reliability of this was insufficient for ultimate industrial use. Aging effects were looked into.
Ultimately, a reproducible current of 50 mA from 0.25 cm2 was achieved with the Mo stack. This was too low to justify the incorporation of the emitter stacks into a prototype device, as envisaged in the contract.
WP6: Vacuum microelectronics (Thales)
This WP aims to develop CNT field emission electron tubes. In the initial project, the application was TWT microwave amplifiers. Because the emission current did not reach high enough values for TWTs, the application was changed at the mid-term review to time-modulated X-ray tubes for industrial and security applications. In contrast to the X-ray body scanner being developed by Philips which requires intense X-ray pulses at moderate voltages (to enhance the contrast between soft tissues), object scanning needs smaller DC currents but higher voltages (for example luggage has higher atomic number contrast, but their imaging needs greater penetration depth), so that this last application appears well adapted to CNT field emission guns.
Field emission guns are needed compared to thermionic guns to allow time modulation; thermionic sources are too slow. Compared to conventional X-ray tubes based on thermionic cathodes, X-ray tubes based on CNT cathodes allow time modulation.
TRT has developed a patented photocathode, based on P-I-N diodes and CNT emitters. The emitter runs at low voltage behind a gate electrode. Without light, the reversed biased P-I-N diodes are in the off state, and there is a low emission current. When illuminated, the diode is on, and the current delivered by the diode can be emitted by the CNTs. The key point for X-ray tubes is that the optical control allows the cathode to be floated at high potential, so that the anode, which is bulky and needs cooling, can be operated at ground potential. A great advantage of this configuration is to simply the high voltage power supply for multi-source systems.
Thus WP6 has produced a product (full development in approximately two years) starting from the CVD equipment of WP1.
WP7: Sensors (CCS and ETH)
Objectives
- to develop low power CMOS based gas sensors integrating CNT;
- to develop nanotube field effect transistor (FET) based gas sensor, and study of mechanism.
ETH developed FET gas sensors, while CCS developed low power, mat sensors using their hotplate technology. ETH developed a scalable FET sensor technology, based on the PCM technology of WP2. ETH demonstrated a prototype CNT FET gas sensor with sensitivity of below 100 ppm for NOx in humid air. In dry air a limit of 50 ppb was attained.
There was also joint work of FET sensors incorporated into the CCS CMOS drive circuits.
CCS
The majority of the CCS work is described in deliverables 7.2 7.3 7.4 and 7.5. The main advantage of CCS technology is the low power demand of their CMOS based sensor. The sensor also places the sensor elements on 'hot plates' which are metal plates with underwired heater elements
WP8: Surface based devices (DTU)
Objectives:
- develop nanotube-enhanced surfaces as microfluidic chemical filters and bioelectrodes for chemical analysis, for use in lab-on-chip devices.
Electrically insulated microfluidic channels for electroosmotic pumping (EOF) of aqueous solutions are fabricated. CNTs are too conductive to be patterned homogeneously in the whole flow channel. The potential difference between the aqueous buffer solution and the CNTs becomes too high, so that bubble formation due to electrolysis of water occurs. The bubbles in the channel terminate the electroosmotic flow, and thereby prevent the device from working properly. In the case of a continuous layer of CNTs in the fluidic channel, a maximum field strength of around 100 V / cm can be applied, which is too small for most applications.
Silicon nanograss technology was also used to create highly hierarchical, three dimensional CNT forests, in order to gain maximal control of surface tension and the active interfacial contact between the chip and cells, as well cell adhesion, proliferation and electrochemical activity.
PDMS is a biocompatible, chemically inert and electrically insulating material, and therefore was chosen as the insulation layer, protecting TiW base and exposing the CNT tips only.
WP9: Training, dissemination (UCAM)
Training was carried out at a workshop / conference holed in Kirchberg, Austria, in 3-10 March 2012, the '26th international workshop on electronic properties of novel materials' or IWEPNM 2012. This is chaired by Prof. C. Thomsen, the project leader from TU Berlin.
Potential impact:
The first 6-inch wafer scale CNT system was developed not in US or Japan but in Europe by Nano-instruments UK in 2005, a spin-off from Cambridge University. Nanoinstruments was acquired by AIXTRON in 2007. The basis of this project is to develop large wafer-scale CNT equipment, and then progress several CNT applications which require large area capabilities. This 'development-chain' ensures that there will be a future applications market for CNT equipment, which would in turn fulfil the production requirements of the applications. The close-loop interaction between research institutes, enabling equipment manufacturer, and end users ensures that this project will establish Europe's competitiveness and create employment at all levels in the value chain of the CNT-manufacturing industry.
CNTs are an enabling material for sub-22nm on-wafer interconnect technology, through-wafer interconnects, strengthening solder bumps and efficient thermal interface materials; the impacts of these microelectronics applications are to lead to higher device densities, higher levels of integration and improved reliability.
The development of wafer scale, mass production CNT equipment will expand the market for CNT deposition systems. AIXTRON expects a significant increase in the demand for systems such as that developed in this project, and as the applications that this project targets and other uses of CNTs ramp up, the expected equipment market for CNT in 2012 is expected to exceed EUR 50 million. This project will strengthen AIXTRON's position as a major equipment manufacturer and will result in the creation of jobs at AIXTRON, as well as end-user / customer sites.
Thales Electron Devices is a major supplier of electron tubes and related subsystems, which has a present market size which is EUR 1700 million, comprising microwave technologies (EUR 1000 million) and X-rays (EUR 700 million).
The development of the X-ray imager as used for security screening as in WP5 will enable the development of products in the 2 - 3 timeframe.
Philips has recently transformed itself from a firm focused on electronics components into a firm focused on three areas, one of which is health care. It shares the X-ray imager market with GE, Siemens and Toshiba. This project will continue to keep Philips at the forefront of this market. The sales figures confirm that this area is a European specialty, Siemens plus Philips is larger than any other. This will have a significant effect on employment numbers within Philips healthcare.
The environmental market is however expected to grow exponentially in the next few years, with the introduction of miniaturised gas-sensors in mobile phones, PDAs, homes or car cabins. Cambridge CMOS Sensors targets two areas with its unique low-cost, high performance technology: the environmental (hydrogen, NOx, CO) and medical sectors. With CCS technology which combines smart hot-plates with ultra-sensitive CNTs and monolithic integration of read-out and processing electronics, we expect an order of magnitude reduction in power consumption to 10 mW continuous power at 400 degrees Celsius, or 1 mW pulsed power, plus an order of magnitude reduction in system cost to well under USD 10 for an integrated gas-sensor unit achieved through miniaturisation. The low sensor cost fuels its uptake, increasing its demand and creating jobs throughout the sensor manufacturing chain. The key point is that the 1 mW power consumption allows the sensors to be incorporated into wireless and hand held devices, which the existing 200-800 mW devices do not. This allows sensors to be incorporated into cell phones, Personal digital assistant (PDA)s, so that any consumer can have his/her own local pollution monitor for use anywhere.
Project website: http://www.technotubes.eu