Yongwon Chung‡
a,
Sanggeun Lee‡a,
Chandreswar Mahataa,
Jungmok Seoa,
Seung-Min Limb,
Min-su Jeongc,
Hanearl Jungd,
Young-Chang Joob,
Young-Bae Parkc,
Hyungjun Kimd and
Taeyoon Lee*a
aNanobio Device Laboratory, School of Electrical and Electronic Engineering, Yonsei University, 50 Yonsei-ro, Seodaemun-gu, Seoul 120-749, Republic of Korea. E-mail: taeyoon.lee@yonsei.ac.kr
bNanodevice Materials Laboratory, Department of Materials Science & Engineering, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 151-744, Republic of Korea
cSchool of Materials Science and Engineering, Andong National University, 1375 Gyeongdong-ro, Andong-si, Gyeongsangbuk-do 760-749, Republic of Korea
dNanodevice Laboratory, School of Electrical and Electronic Engineering, Yonsei University, 50 Yonsei-ro, Seodaemun-gu, Seoul 120-749, Republic of Korea
First published on 4th November 2014
In this work, we have demonstrated chemically coupled (3-aminopropyl)trimethoxysilane (APTMS) and 3-mercaptopropionic acid (MPA) self-assembled monolayers (SAMs) to enhance the diffusion barrier properties against copper (Cu) as well as the adhesion properties towards SiO2 and Cu electrode. The coupled-SAM (C-SAM) can attach to both Cu and SiO2 strongly which is expected to enhance both the diffusion barrier and adhesion properties. A carbodiimide-mediated amidation process was used to link NH2 terminated APTMS to COOH terminated MPA. The resulting C-SAM shows a low root-mean-square roughness of 0.44 nm and a thickness of 2 nm. Time-dependent dielectric breakdown (TDDB) tests are used to evaluate APTMS and C-SAM for their ability to block Cu ion diffusion. The average time-to-failure (TTF) is enhanced over 4 times after the MPA attachment, and is even comparable to TaN barriers. Capacitance–voltage (C–V) measurements are also conducted to monitor Cu ion diffusion. Negligible change in the flatband voltage and C–V curve is observed during the constant voltage stress C–V measurement. Enhancement of the adhesion properties are measured using four-point bending tests and shows that the C-SAM has a 33% enhancement in the adhesion properties between SiO2 and Cu compared to APTMS. The C-SAM shows potential as an ultra-thin Cu diffusion barrier which also has good adhesion properties.
Other types of diffusion barriers have been studied as a replacement for conventional Ta/TaN barriers. Self-forming diffusion barriers from CuMn alloys7 and seedless diffusion barriers using RuMo8 have been studied as alternatives; however, it is still difficult to obtain diffusion barriers under the thickness of 2 nm. Recently, graphene has been demonstrated as a diffusion barrier which shows outstanding performance with only 0.34 nm thickness.9 Nevertheless, it still lacks the process technology to grow or transfer high quality graphene to complicated structures. Moreover, the low adhesion between the graphene–Cu10 can lead to chemical mechanical polishing (CMP) failure commercially used in the dual damascene process.11
Self-assembled monolayers (SAMs) are spontaneously formed two-dimensional structures12,13 that have been used to prevent wet chemical etching14 and oxidation in air15 due to their dense molecular packing. SAMs have also been used as a diffusion barrier for Cu diffusion and as an adhesion promoter by the selection of appropriate terminal groups that can immobilize Cu and bond with both Cu and SiO2.16,17 The use of SAMs as diffusion barriers is of interest due to their few- or sub-nanometer thickness and the simple solution based processing techniques that can be used to form SAMs on complicated nanostructures.18 It has been found that (3-aminopropyl)trimethoxysilane (APTMS) which has a silane (–Si–) head group and amine (–NH2) functional group is a promising candidate for diffusion barrier against Cu diffusion and adhesion promoter.19–21 However, the NH2 group demonstrates reduced adhesion properties with Cu compared to the thiol (–SH) group which can also lead to lower Cu immobilization.
In this work, we demonstrate the surface chemical modification of APTMS to enhance the diffusion barrier properties to Cu and enhance the nano-adhesive properties of SAMs to both Cu and SiO2. Selection of the molecule for APTMS enhancement requires several structural aspects, including reactivity with APTMS, thermal stability, and good adhesion of the terminal group to Cu. Therefore, 3-mercaptopropionic acid (MPA) was selected as the target molecule for attachment to APTMS. With both carboxyl (COOH) and SH functional groups, MPA can form an amide bond with the NH2 of APTMS and have a strong covalent bond with Cu.22 Through carbodiimide-mediated amidation,23,24 the COOH group of MPA was chemically coupled to the NH2 group of APTMS. Cu diffusion is further prevented due to its covalent attachment of S at the top surface of the bilayer SAM. Detailed comparison were studied for MPA coupled to APTMS with reference sample of APTMS through surface morphological characterization, chemical bonding characterization, electrical reliability test, and adhesion property characterization.
![]() | ||
Fig. 1 (a) Schematics of APTMS coating process and MPA coupling process. (b) The detailed chemical reactions for the chemical coupling of MPA onto APTMS. |
Molecular topography related to the surface roughness was examined by atomic force microscopy (AFM) for the APTMS and C-SAM organic layers. Fig. 2 represents typical AFM images of the surfaces of APTMS and C-SAM coated on SiO2/Si and their cross-sectional image. The root-mean-square (RMS) roughness of the APTMS surface coating is low as RS = 0.40 nm. We have used low concentration APTMS solution to prevent multilayer or 3-dimensional assembly of the molecules due to the hydrogen bonding of the amine groups to other APTMS molecules or the substrate, which can result in a rough surface.21,25 The thickness of the SAM films measured by ellipsometry (not shown here) also indicates the film is well coated. The measured thickness of the APTMS coating is 1.4 nm, which is similar to the experimental values of monolayer APTMS in previous reports.21,26 The immersion time for APTMS coating is fixed to 30 minutes at 3 mM to obtain a high monolayer film with partial multilayer film in order to avoid sub-monolayer regions that cannot block the Cu diffusion.21 On the contrary, higher degree of multilayer APTMS is also avoided as it may weaken the adhesion property through fracture failure.21 There is only a slight increase in the roughness of RS = 0.44 nm after a MPA coupling reaction. Also, the thickness of C-SAM is 2.0 nm based on ellipsometry measurement, which shows a 0.6 nm difference from APTMS alone. This difference is equivalent to the experimental thickness of MPA.27 The cross sectional AFM image shown in Fig. 2c displays the low roughness of the films with a peak-to-valley height less than 2 nm. The results obtained from the AFM and ellipsometry measurements indicate the thin and smooth coating of APTMS and MPA coupling. AFM images obtained from larger area also show the low roughness of the SAM coating (Fig. S1†).
![]() | ||
Fig. 2 AFM images of the substrates coated with (a) APTMS and (b) C-SAM and their (c) cross-sectional image. |
The detailed surface chemistry of the SAM layers was characterized by X-ray photoelectron spectroscopy (XPS) to confirm the changes after carbodiimide-mediated amidation. Fig. 3a–c shows the XPS measurement results of the APTMS and C-SAM layer. The experimental curves are deconvoluted using Gaussian–Lorentzian peaks after Shirley background corrections. The fitted curves are assigned to red and blue which are components from APTMS and MPA molecules, respectively. The XPS spectra obtained from the APTMS film are composed of peaks originating from APTMS. Fig. 3a shows the C 1s core level spectrum which is composed of three peaks. The main peak at 284.6 eV is attributed to C–C bonding, while one smaller peak at 286.1 eV origins from C–N and C–O bonding and another small peak at 287.6 eV origins from CO bonding.21,28,29 The N 1s core level spectrum shows a large main peak at 398.8 eV due to the primary amine peak along with a small amount of hydrogen-bonded amine peak at 400.7 eV as shown in Fig. 3b.21,25 The amino groups in the APTMS molecules can interact with each other or the substrate through hydrogen bonding, which leads to a rough surface. The high primary amine percentage of 80% indicates the high quality of the APTMS film which is supported by our AFM results of the APTMS layer with low roughness.21,25 As shown in Fig. 3c, no detectable peaks were observed in the S 2p core level region due to the absence of S atoms in the APTMS molecule and substrate.
After MPA coupling reaction, additional peaks which are related to thiol or amide groups have appeared. The C 1s core level spectrum of the C-SAM in Fig. 3a shows peaks for C–S at 287.4 eV and OC–N at 287.7 eV.30,31 The N 1s core level region in Fig. 3b shows a large decrease in the primary amine peak intensity along with the appearance of an amide peak at 400 eV (ref. 32) due to the carbodiimide-mediated amidation process. Significant peaks in the S 2p core level region were observed after the MPA coupling reaction as shown in Fig. 3c. The S 2p spectrum shows a doublet structure that was deconvoluted into peaks that originate from S 2p3/2 and S 2p1/2. The spectra were fitted using a peak area ratio of 2
:
1 with 1.2 eV spacing.33 The largest pair was observed at 163.5 and 164.7 eV which is due to unbound thiols (–SH),33 indicating that the most of the thiols are stable without interaction with other species. Peaks positioned at 161.3 and 162.5 eV are due to atomic sulfur atoms,34 and peaks at 167.9 and 169.1 eV corresponds to oxidized S species.35 Such defects are thought to occur during the chemical reaction of the carbodiimide-mediated amidation process. However, the oxidation of the thiol groups are not a drawback as it can enhance the diffusion barrier properties and bonding strength to Cu compared to unoxidized thiol groups.36
Time-dependent dielectric breakdown (TDDB) tests were conducted in order to evaluate the Cu diffusion barrier properties under bias temperature stress (BTS). Al/Ta/Cu stacked dots with 1 mm size were deposited on SAM-coated n++ Si with 100 nm SiO2. The leakage currents of the metal-oxide-semiconductor (MOS) capacitors were measured under constant electric field (2 MV cm−1) at elevated temperature (225 °C). Fig. 4a shows the degradation and breakdown characteristics of SiO2 during the thermal and electrical stress for each representative sample using APTMS and C-SAM. It should be noted that TDDB results from devices without diffusion barriers were not included due to their breakdown at lower electric field intensity (Fig. S2a†). The time-dependent leakage current has two distinct regions. In the initial state, the current gradually decreased for ∼500 and ∼1000 seconds for the APTMS and C-SAM, respectively. This initial leakage current is occurred by the Poole–Frenkel tunnelling via the trap sites in the SiO2 insulator, which is further suppressed by the trapping of electrons in the trap sites.37 In the second region, the current starts to increase until it suddenly reaches the compliance current value (1 × 10−5 A). The measured time until this event is defined as the time-to-failure (TTF), which is ∼2000 and ∼7700 seconds for the representative samples. This region strongly indicates the degradation of the SiO2 layer with trap creation up to the formation of a percolation path for a hard breakdown. Such change in the leakage current for TDDB measurements is a typical characteristics of failures due to traps that are newly generated via Cu ion migration,38 which modifies the band structure of the MOS allowing Fowler–Nordheim current to flow.37 Such breakdown was not observed when Al was used instead of Cu (Fig. S2b†). The longer TTF shows that the C-SAM exhibits a better Cu blocking property.
![]() | ||
Fig. 4 (a) Representative TDDB test results for APTMS and C-SAM under 225 °C and 2 MV cm−1. (b) Weibull plots and analysis for APTMS, C-SAM, 25 nm TaN, and 50 nm TaN barriers. |
Fig. 4b shows the Weibull plot and the derived parameters of the TTF for more than 15 samples with APTMS and C-SAM barriers. Here, t63.2 is defined as the time when 63.2% of the failures occurred and β is defined as the slope of the Weibull distribution obtained from the Weibull plot. TDDB results show a wide range of TTF, therefore statistical analysis is required to confirm that the C-SAM has an improved Cu blocking property and a higher resistivity to breakdown than APTMS.38 Additionally we have presented a clear comparison with as-deposited TaN barriers with a thickness of 25 and 50 nm to the SAM barrier in our work. The t63.2 of the APTMS is very low at 0.707 hours, which increases to 2.931 hours after the carbodiimide-mediated reaction to couple the MPA molecules on APTMS. The four-fold increase in t63.2 compared to APTMS shows that the barrier property of C-SAM is enhanced, and is even comparable to 50 nm TaN barrier in this experiment. The β > 1 value of the slope indicates the distribution of the TTF is in a good agreement with other diffusion barriers.39 As a result, the C-SAM diffusion barrier is very promising compared to 50 nm TaN as the thickness of the bilayer of SAMs is very low compared to TaN. The enhancement of the TTF can be attributed to the chemical interaction between S and Cu, which plays an important role in the diffusion barrier,17 and also the increment in the thickness of the C-SAM layer compared to APTMS alone.
Cu diffusion can degrade device performance when exceeding its critical limit. It is therefore important to employ more elaborate electrical analysis to evaluate the SAM diffusion barrier performance. Capacitance–voltage (C–V) measurements of MOS capacitors were performed at room temperature to detect the Cu ion diffusion by monitoring the C–V characteristics. Cu dots with 100 μm size were deposited on SAM-coated p Si with 10 nm SiO2. Fig. 5a–c shows the C–V curves of the fabricated devices with (a) bare SiO2, (b) APTMS-coated SiO2, and (c) C-SAM-coated SiO2. The devices were stressed at an electric field of 5 MV cm−1 at room temperature and C–V characteristics were measured during stress. Bare SiO2 samples without a diffusion barrier show a negative voltage shift in the C–V curve. The shift in C–V characteristic indicates the diffusion of ionic Cu. The constant voltage stress (CVS) results establish the effectiveness of SAMs as a diffusion barrier material under different electron injection conditions. Both APTMS and C-SAM layers effectively hinder Cu penetration into underlying SiO2 and prevent subsequent degradation (Fig. 5b and c) under externally high electron injection compared to the absence of the SAM layer (Fig. 5a). In addition to the shift of the flatband voltage, distortion of the C–V characteristics after extended CVS without a diffusion layer indicates further penetration of Cu ions. The change of the oxide capacitance indicates the diffusion of Cu into SiO2 as the change in oxide charge is a result of copper diffusion that acts as an interstitial in SiO2.
![]() | ||
Fig. 5 C–V measurements of the MOS capacitor structure with (a) bare SiO2, (b) APTMS coated SiO2, and (c) C-SAM coated SiO2. |
Adhesion properties towards SiO2 and Cu were also evaluated by using four-point bending tests. The four-point bending test is a widely used method to quantitatively measure the interfacial adhesion energy between two materials.40 Fig. 6a shows the representative load and displacement curves obtained from the four-point bending tests of bare SiO2, APTMS coated SiO2, and C-SAM coated SiO2. The samples were annealed in a vacuum chamber at 400 °C after Cu deposition to form an annealing-induced siloxane bridge to increase adhesion.20,21,41 When force is applied to the sandwiched structure, a linear relation with the displacement is expressed. As the force increases, a pre-crack propagates from the notch to the Cu/SiO2 interface and at a critical point of PC, the force decreases and an interfacial crack initiates with PPlateau.42 The force during the crack propagation is the factor that determines the adhesion of the Cu film to the oxidized silicon substrate. This force was 4.0, 7.5, and 9.1 N for bare SiO2, APTMS coated SiO2, and C-SAM coated SiO2, respectively. The interfacial adhesion energy is characterized by the critical strain energy release rate defined as the following equation:40
Fig. 6b shows the converted interface toughness before and after annealing at 400 °C with at least 3 samples. Before the annealing step, all of the samples exhibited low adhesion strength of 2.01 ± 0.33, 2.82 ± 0.39, and 2.12 ± 0.11 J m−2 for bare SiO2, APTMS-coated SiO2, and C-SAM-coated SiO2, respectively. Before the annealing step, siloxane bonds formed by the dehydration of silanol groups are strained and tend to rehydrate, and revert to silanol when the samples cool down. However, after the annealing step, irreversible dehydration between the siloxane groups and SiO2 surface occurs which is a covalent bond.21 After the annealing step, the adhesion of the SAM-coated layers increased. The bare SiO2 samples still showed a very low adhesion of 2.07 ± 0.30 J m−2, indicating that Cu and SiO2 have low interaction. APTMS shows increased interfacial adhesion energy of 7.41 ± 0.25 J m−2, which demonstrates a similar value to previous reports of APTMS on SiO2.20 In our study of C-SAM, the interfacial adhesion energy increases to 9.88 ± 0.34 J m−2, which indicates a 33% increase compared to APTMS. Since the fracture mechanism after dehydration of APTMS is due to the Cu–amine interface,20 a tail group with a stronger bonding strength with Cu is needed to enhance the adhesion properties. Thiols have a binding energy of 48 kcal mol−1 on Cu (111),22 which is stronger than that of amine–Cu at approximately 9.4 kcal mol−1.43 The obtained interfacial toughness for C-SAM is higher than as-deposited and H2 plasma-treated TaNx, and is even comparable to rapid thermal annealing-treated TaNx.44 This value is sufficient to prevent delamination and cracking during CMP commercially used in the dual damascene process.11
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4ra08134j |
‡ These authors contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2014 |