A method for the growth of low temperature vertically-aligned carbon nanotubes, and the subsequent fabrication of vertical interconnect electrical test structures using semiconductor fabrication is presented.
We demonstrate a method for the low temperature growth (350 °C) of vertically-aligned carbon nanotubes (CNT) bundles on electrically conductive thin-films. Due to the low growth temperature, the process allows integration with modern low-κ dielectrics and some flexible substrates. The process is compatible with standard semiconductor fabrication, and a method for the fabrication of electrical 4-point probe test structures for vertical interconnect test structures is presented. Using scanning electron microscopy the morphology of the CNT bundles is investigated, which demonstrates vertical alignment of the CNT and can be used to tune the CNT growth time. With Raman spectroscopy the crystallinity of the CNT is investigated. It was found that the CNT have many defects, due to the low growth temperature. The electrical current-voltage measurements of the test vertical interconnects displays a linear response, indicating good ohmic contact was achieved between the CNT bundle and the top and bottom metal electrodes. The obtained resistivities of the CNT bundle are among the average values in the literature, while a record-low CNT growth temperature was used.
Copper and tungsten, the metals which are currently used for the interconnects in state-of-the-art very-large-scale integration (VLSI) technology, are approaching their physical limits in terms of reliability and electrical conductivity1. While down-scaling transistors generally improves their performance, it actually increases the resistance and current density of the interconnects. This resulted in interconnects dominating the integrated circuit (IC) performance in terms of delay and power consumption2.
Carbon nanotubes (CNT) have been suggested as alternative for Cu and W metallization, especially for vertical interconnects (vias) as CNT can easily been grown vertical3. CNT have been shown to have excellent electrical reliability, allowing an up to 1,000 times higher current density than Cu4. Moreover, CNT do not suffer from surface and grain boundary scattering, which is increasing the resistivity of Cu at the nanometer scale5. Finally, CNT have been shown to be excellent thermal conductors6, which can aid in the thermal management in VLSI chips.
For successful integration of CNT in VLSI technology it is important that the growth processes for the CNT is made compatible with semiconductor fabrication. This requires the low temperature growth of CNT (< 400 °C) using materials and equipment which are considered compatible and scalable to large scale manufacturing. While many examples of CNT test vias have been demonstrated in the literature7,8,9,10,11,12,13,14, most of these use Fe as catalyst which is regarded as a contaminant in IC manufacturing15. Besides, the growth temperature used in many of these works is much higher than the upper limit of 400 °C. Preferably CNT should even be grown below 350 °C, in order to allow integration with modern low-κ dielectrics or flexible substrates.
Here we present a scalable method for growing CNT at temperatures as low as 350 °C using Co as catalyst16. This method is of interest for fabricating different electrical structures consisting of vertically aligned CNT in integrated circuits, ranging from interconnect and electrodes to super capacitors and field emission devices. The Co catalyst metal is often used in IC manufacturing for the fabrication of silicide’s17, while TiN is an often used barrier material7. Moreover, we demonstrate a process for fabricating CNT test vias while only using techniques from standard semiconductor manufacturing. With this, CNT test vias are fabricated, inspected by scanning electron microscopy (SEM) and Raman spectroscopy, and electrically characterized.
Caution: Please consult all relevant material safety data sheets (MSDS) before use. Several of the chemicals used in this fabrication process are acutely toxic and carcinogenic. Nanomaterials may have additional hazards compared to their bulk counterpart. Please use all appropriate safety practices when working with equipment, chemicals or nanomaterials, including the use of engineering controls (fume hood) and personal protective equipment (safety glasses, gloves, cleanroom clothes).
1. Alignment Marker Definition for Lithography
2. Bottom Metal and Interlayer Dielectric Deposition
3. Catalyst Deposition and CNT Growth
4. Topside Metallization
5. Measurements
The design of the measurement structure used in this work can be found in Figure 1. By employing such a structure the measurement of the CNT bundle resistance and the metal-CNT contact resistances can be determined accurately, as probe and wire resistances are circumvented. The resistance of the bundle is a measure for the quality and density of the CNT bundle. In order to determine the contact resistance bundles of different lengths should be measured.
A typical SEM image of CNT grown at 350 °C for 60 min taken from the top before metallization at 45° tilt is shown in Figure 2. Such an image is useful for checking if the growth time of the CNT is correctly set in order to obtain the same length as the thickness of the SiO2 layer. A cross-section prepared by mechanical cleaving inspected by SEM of the same wafer after metallization is shown in Figure 3. This can be used to determine the alignment of the CNT, their density (for instance be counting the number of CNT per unit length), and if a high resolution SEM is used to determine their diameter. Also the contact area between the CNT and the metal layers can be investigated.
Raman spectra of Co-grown CNT at 350 °C is displayed in Figure 4. Raman spectroscopy is a powerful technique for investigating the crystallinity of the CNT18, and can for instance be used to optimize the CNT growth parameters in order to obtain the highest quality CNT. I-V measurements were performed using four point probe structures and are displayed in Figure 5. When the I-V behavior is linear it indicates ohmic contact between the CNT and the metal contacts. From the slope the electric resistance can be determined. From the resistance and the dimension of the bundles the resistivity can be calculated, which for these CNT bundles is compared to the literature in Figure 6.
Figure 1. Design of 4-point probe measurement structure used in this work. In the figure the dark yellow layer indicates the TiN, the black tubes the CNT bundles, and the metallic layer the Ti and Al stack. The sacrificial Ti layer is omitted for clarity and the oxide is semi-transparent. Probe connections for electrical 4-point probe measurements are indicated. Please click here to view a larger version of this figure.
Figure 2. Top-view SEM image of a CNT bundle. This shows a 2 µm wide CNT bundle grown in a contact opening which was etched inside the SiO2. This figure has been modified from 16, with permission from Elsevier. Please click here to view a larger version of this figure.
Figure 3. SEM cross-section of CNT via. Cross-section of a 2 µm wide and 1 µm long CNT test via prepared using mechanical cleaving after metallization. This figure has been modified from 16, with permission from Elsevier. Please click here to view a larger version of this figure.
Figure 4. Raman spectrum of a CNT bundle grown using Co at 350 °C. The names of the Raman bands are indicated. The black curve displays the raw measurement data. For all bands a Lorentzian fitting is performed (green dashed curves), except for the D’ band which is fitted by a Gaussian18. Please click here to view a larger version of this figure.
Figure 5. I-V measurements of CNT test vias with different diameters. The symbols represent the measurement data, while the solid line indicates a linear least squares fitting to the measurement data. The electrical resistances of the different vias as determined from the slope of the linear fitting are indicated. This figure has been modified from 16, with permission from Elsevier. Please click here to view a larger version of this figure.
Figure 6. Comparison of CNT bundle resistivity with values from the literature. The resistivity is calculated from the resistance and the via dimensions. It is compared with values from the literature, and CNT vias fabricated at different temperatures using the method described in this work. This figure has been modified from 16, with permission from Elsevier. Please click here to view a larger version of this figure.
Figure 1 displays a schematic overview of the structure fabricated in this work, and which was used for the 4-point probe measurements. As the potential is measured through probes carrying no current, the exact potential drop (VH-VL) over the central CNT bundle and its contacts to the metal can be measured. Bigger diameter CNT bundles are used to contact the bottom TiN layer from the contact pads, in order to reduce the total resistance for the current forcing probes and maximize the potential drop over the central CNT bundle.
As can be seen from Figure 2, the CNT were successfully grown inside the openings etched in the SiO2 with a length approximately the same as the depth of the hole (1 µm). It is crucial that the length of the CNT is roughly the same as the depth of the hole, in order to achieve conformal coating of the top metal contact. The bundles appear uniform, which also aids in conformal coating of the metal. The straightness and vertical alignment of the tubes can clearly be seen in the cross-section displayed in Figure 3. By counting, the density of the CNT bundle was estimated to be around 5×1010 tubes/cm2. Using transmission electron microscopy the average diameter of the tubes was found to be 8 nm, as was shown elsewhere16. Due to the low growth temperatures the CNT walls contain many defects making determining the number of walls difficult. The tubes appear to have a hollow core, although bamboo crossings have been observed. The cross-section also shows the bottom TiN layer, and the sacrificial Ti layer which is partly removed underneath the SiO2 during the wet etching. If openings are placed closed together the etching time of the sacrificial Ti layer may have to be optimized to minimize underetch to prevent oxide delamination. Due to the dry etching of the hole, the spacing between the SiO2 and CNT bundle is minimal, which is essential to prevent the sputtered Ti and Al from forming short circuits around the CNT bundle.
Using the Raman data the crystallinity (or quality) of the CNT can be investigated. As the different Raman bands are close to each-other deconvolution of the bands is necessary, as described elsewhere18. From the Raman data in Figure 4 it is apparent that a strong D and D’ band can be observed, which are caused by Raman scattering with defects, while the G band is related to the C-C bond. The other two bands are weak Raman features which are included for more accurate fitting.
It is known that a low growth temperature in general results in a lower CNT quality18. Usually the D over G intensity ratio (ID/G) is used to assess the quality of graphitic materials, which is 1.1 in Figure 4. As has been shown by for instance Ferrari and Robertson20, care has to be taken when using only this band ratio. With increasing quality of the CNT, first the ID/G ratio increases, till a certain amount of crystallization is reached after which the ratio decreases monotonically. Due to the very low growth temperature, the CNT in this work appear to have a crystallinity below this threshold16. In these cases the full-width at half maximum of the D band can be used to compare CNT samples fabricated at different process conditions18. It can be expected that the low quality of the CNT will significantly influence the electrical performance.
Judging from the almost complete linear behavior of the I-V characteristics in Figure 5, the contacts between the CNT and the top and bottom metal layers are ohmic. The resistance of the bundle decreases with diameter, which is to be expected as more CNT can conduct in parallel for larger bundles. The good contact between the CNT and the metals is attributed to the use of Ti19, and TiN which is more resilient against oxidation21. Besides, we found that due to the lack of any dielectric covering steps of the CNT after growth (using for instance spin-on glass), something which is often used in the literature in combination with chemical mechanical polishing (CMP)22,23, the contact resistance to the CNT is low due to embedding of the CNT tips in the top metal24.
When comparing the resistivities of the CNT bundles with literature, as is done in Figure 6, the results are among the average values in the literature. However, the growth temperature used in this work is record-low. The results of Yokoyama et al. 13 are the lowest resistivity reported in the literature, using only a 40 °C higher growth temperature. However, the equipment used for Co particle deposition in their work is likely not scalable to large volume manufacturing. Clearly the resistivity decreases with increasing growth temperature, which can be advantageous for application allowing higher growth temperatures. When comparing the resistivity of the CNT bundles with traditional interconnect metals like Cu (1.7 µΩ-cm), it is apparent that a drastic reduction of the resistivity is required. Improving the quality of the CNT and the bundle density, by optimizing the growth conditions, will be required. This has to be done without increasing the growth temperature, in order to allow integration with modern low-κ materials and flexible substrates.
We have thus demonstrated a technique for integrating low temperature CNT growth and integration into standard semiconductor fabrication. This technique has been used to fabricate CNT via test structures and has recently been applied for the fabrication of CNT super capacitors25.
The authors have nothing to disclose.
Part of the work has been performed in the project JEMSiP_3D, which is funded by the Public Authorities in France, Germany, Hungary, The Netherlands, Norway and Sweden, as well as by the ENIAC Joint Undertaking. The authors would like to thank the Dimes Technology Centre staff for processing support.
Materials | Company | Catalog Number | Comments/Description |
Si (100) wafer 4" | International Wafer Service | Resisitivity: 2-5 mΩ-cm, thickness: 525 µm | |
Ti-sputtertarget (99.995 % purity) | Praxair | ||
Al (1% Si)-sputtertarget (99.999 % purity) | Praxair | ||
Co (99.95 % purity) | Kurt J. Lesker | ||
Chemicals | Company | Catalog Number | Comments/Description |
SPR3012 positive photoresist | Dow Electronic Materials | ||
MF-322 developer | Dow Electronic Materials | ||
HNO3 (99.9 %) | KMG Ultra Pure Chemicals | ||
HNO3 (69.5%) | KMG Ultra Pure Chemicals | ||
HF 0.55% | Honeywell | ||
Tetrahydrofuran | JT Baker | ||
Acetone | Sigma-Aldrich | ||
ECI3027 positive photoresist | AZ | ||
Tetraethyl orthosilicate (TEOS) | Praxair | ||
Gasses | Company | Catalog Number | Comments/Description |
N2 (99.9990%) | Praxair | ||
O2 (99.9999%) | Praxair | ||
CF4 (99.9970%) | Praxair | ||
CL2 (99.9900%) | Praxair | ||
HBr (99.9950%) | Praxair | ||
Ar (99.9990%) | Praxair | ||
C2F6 (99.9990%) | Praxair | ||
CHF3 (99.9950%) | Praxair | ||
H2 (99.9950%) | Praxair | ||
C2H2 (99.6000%) | Praxair | ||
Equipment | Company | Catalog Number | Comments/Description |
EVG 120 coater/developer | EVG | ||
ASML PAS5500/80 waferstepper | ASML | ||
SPTS Ωmega 201 plasma etcher | SPTS | Used for Si and metal etching | |
SPTS Σigma sputter coater | SPTS | ||
Novellus Concept One PECVD | LAM | ||
Drytek 384T plasma etcher | LAM | Used for oxide etching | |
CHA Solution e-beam evaporator | CHA | ||
AIXTRON BlackMagic Pro CVD tool | AIXTRON | Carbon nanotube growth | |
Philips XL50 scanning electron microscope | FEI | ||
Tepla 300 | PVA TePla | Resist plasma stripper | |
Avenger rinser dryer | Microporcess Technologies | ||
Leitz MPV-SP reflecometer | Leitz | ||
Renishaw inVia Raman spectroscope | Renishaw | ||
Agilent 4156C parameter spectrum analyzer | Agilent | ||
Cascade Microtech probe station | Cascade Microtech |