Investigation of pore and network formation in spin-on ultra low-k dielectrics by spectroscopic techniques

Positronium annihilation spectroscopy has become more and more important in microelectronics industry as one of the few methods to characterize engineered nanopores in next-generation (k < 2.4) interlayer dielectrics (ILD). With the addition of infrared spectroscopy a way is found to investigate the pore and network formation during the curing process.
1. Introduction
Porous spin-on glasses are one great candidate for the integration as ultra low-k (ULK) dielectrics in Back-End of Line (BEOL) for advanced technology nodes. They offer the possibility of a structured pore network by using “Block Polymer Templated Inorganic Oxides” (USP 6,592,764) [1]. Therefore, it is possible to adjust the physical properties of these thin films [1] as well as pore size and porosity. However, these materials are also prone to dielectric damage [2,3] during the integration into back-end of Line. In addition, these material degradations will increase with porosity [3]. It follows, that beside the damage mechanism also the pore and network formation of ULK materials need to be investigated more in detail to comprehend integration damage from the very first time of appearance. This work will evaluated the formation of spin-on glasses during curing by positron annihilation spectroscopy to observe the pore formation and by Fourier Transform Infrared spectroscopy to study the network formation.
2. Experimental
2.1 Preparation of spin-on dielectrics
For preparation of spin-on ULK material a solution from SBA Materials, Inc. was used. The ULK liquid precursor consists of an amphiphilic block copolymer with silicon alkoxide esters [1]. The final thickness is supposed to be 500 nm with an initial k-value of 2.2. The solution was spin-coated on 6-inch silicon wafers with 2000 rpm for 60 s. The spin-coated samples were soft baked for 120 s at 150 °C. The curing procedure was performed with the quartz glass oven PEO 603 from ATV technologies for different curing times at 450 °C under nitrogen atmosphere. The heat ramp of curing was chosen to be 7 °C/min.
2.2 Measurement techniques of pore and network structure
Fourier Transform Infrared (FTIR) spectroscopy was used to determine the chemical and structural changes before and after different curing times. The measurements were performed in transmission mode in the mid-range from 400 to 4000 cm-1, using the Bruker Tensor 27 spectrometer. The optical response is given as absorbance and normalized to thickness as well as treated by a baseline subtraction. Thus a comparison of the different processes can be achieved. Furthermore, a deconvolution of the FTIR peaks at the oxide region (1300 cm-1 to 950 cm-1) was accomplished with the Peak-Fit Module of ORIGIN 8.5 software. As it is described in literature the oxide region is used to be deconvoluted into the following peaks: the suboxide-, network-, cage-peak [4,5] and the Si-O-C peak [6]. The area was normalized to the total area of the Si-O-Si area.
The characterization of the nanopores were carried out at the Helmholtz-Zentrum Dresden-Rossendorf (HZDR) with positron annihilation spectroscopies (PAS). For determination of different pore components the Elbe Positron Source (EPOS) was used; a pulsed positron source with high repetition rate, high intensity and choosable energies for depth profile measurements [7,8]. Furthermore, by using the The Slow-Positron System of Rossendorf (SPONSOR) Doppler broadening spectroscopy was used to acquire also information about the atomic surrounding of the pores.
Since the possibility of diffusing out of ortho-positronium (o-Ps) from film surface through open, interconnecting pores the ULK films were covered with a 20 nm thick carbon layer by evaporating using a pre-shaped carbon rod.
3. Results
Fig. 1 shows the FTIR spectrum of the uncured and cured samples after 5 min, 30 min, 60 min and 90 min at 450 °C. The curing process at 450 °C for 90 min can be considered as complete cure. The uncured sample shows the –OH absorbance peak at 3400 cm-1 and the Si-OH peak at 914 cm-1 [9], which is characteristic for spin-on glasses in sol-gel science [10]. During the curing process the methylsilsesquioxane (MSQ) based Si-OH groups condensate and cross-linking occurs to form a 3D network [11]. Besides the Si-OH peak in the uncured sample the region from 3000 cm-1 to 2800 cm-1 is more pronounced compared to the cured samples. There are located the C-Hx vibration bands, which mostly correspond to porogen to form a porous network. The region from 900 cm-1 to 700 cm-1 is called the fingerprint region, where mostly the different Si-C vibration modes are dominate [9]. The absorbance peak at 1277 cm-1 belongs to the Si-CH3 vibration band.
The two main mechanism during the curing of spin-on ULK are the cross-linking of the Si-OH groups to form a mechanically stable thin film and the porogen removal to form a porous network. The cross-linking of Si-OH to form the Si-O-Si linkage can be seen in Fig 1. in the range of 1250 cm-1 to 960 cm-1. Already after 5 min of curing the Si-O-Si region is formed almost completely and only little changes during further curing are observed. By taking a closer look at the Si-O-Si region, a shift from 1047 cm-1 to 1052 cm-1 can be observed. Within this region four peaks are overlapping: the suboxide-, network-, cage- [4,5] and the Si-O-C peak [6]. To distinguish between those peaks a deconvolution was done, where the results are shown in Fig. 2. With increasing curing time the network peak rises whereas the cage and the suboxide peak decreases. This behavior is likely due to the cross-linking of the network material and is less pronounced after 30 min of curing. The Si-O-C peak seams not to be affect by curing time.
Fig. 3 and Fig. 4 show the results from the DBS and PALS measurements. From the DBS measurements two specific line parameters can be calculated: the S- and W-parameter. The S-parameter is a measurement of the open volume and the W-parameter is a measurement of the atomic environment of the annihilation site. The mean implantation depth for the positrons is given in nanometer scale at the upper x-axes based on the density of silica. The first 20 nm (until 1.2 keV) the S- and W-parameter corresponds to the carbon capping layer and retain unchanged for all treatments. The W-parameter for the uncured sample shows the highest values. Only for 20 nm Carbon layer the W-parameter is higher. Within the first minutes of curing, the porogen is almost removed and the W-parameter decreases. From 30 min to 90 min no change in W-parameter is visible anymore. Compared to the FTIR spectra (Fig. 1) in the region of 3000 cm-1 to 2800 cm-1 no change for all curing times are visible which shows the sensitive behavior of PAS measurements for nanopore evaluations. For the cross-linking behavior (Fig. 2) the dependence on curing time can still be observed in detail by FTIR. Therefore, it can be concluded that the porogen extraction at 450 °C with a slow heating ramp is completed within the first 30 min, whereas the cross-linking of the network takes places over the complete curing time.
The calculated pore components from PALS measurements are shown in Fig. 4. Two different lifetime components were found. The upper diagrams show the diameter of the pores and the lower diagrams show the intensities of the pore components. The first pore component has a diameter of around 0.8 nm which does not change for all curing treatments and positron energies, whereas the intensities of the component 1 decrease from the uncured to the cured state. With regard to the FTIR results (Fig. 1 and Fig. 2), this pore component intensities shows the inverse behavior of the oxide region, where the network is formed. Therefore, this component is likely due to the unreacted Si-OH groups, which decreases during the curing. The second pore component arises from the pores itself and was not observed at the uncured sample. The mean diameter is about 3.4 nm, which lowers to 3 nm near the surface of the film. This can be due to the carbon capping layer deposition and needs further considerations. Also it can be seen that the pore diameter from 5 min to 30 min still increases and reaches a final value after 30 min. In addition, the intensity of the pore component 2 changes marginal. That confirms that the porogen extraction as well as the pore formation takes place within the first 30 min.
4. Summary
In this work, the pore formation of spin-on ULK materials with an initial k-value of 2.2 was studied for a thermal curing process at 450 °C with a slow heat ramp for curing. After 5 min of curing most of the porogen is extracted and the network is formed almost completely, which can be seen by PAS and FTIR. The porogen extraction as well as the pore formation appear to be complete after 30 min, whereas the remaining time is needed to form the network.
All the investigations are running right now for faster heat ramp of curing as well as for different curing temperatures to slow down the pore formation process and get a better understanding of the processes taking place inside the material during curing.
References
[1] P. Garrou, Solid State Technology, p. 10, Nov. 2010, Accessed 22 May 2017
[2] J.L. Shohet et al., ECS Transactions, 60 (1) p. 733-738 (2014)
[3] M.R. Baklanov et al., Journal of Applied Physics 113, p. 041101, (2013)
[4] L. Zhang et al., ECS Solid State Lett. 2 (2), p. N5 (2013)
[5] A. Grill et al., J. Appl. Phys. 94 (10), p. 6697 (2003).
[6] Y. Lin et al., J. Electrochem. Soc. 153 (7), p. F144 (2010)
[7] A. Uedono et al. Applied Surface Science, Bd. 368, pp 272-276 (2016)
[8] A. Wagner et al., Defect and Diffusion Forum, Bd. 331, pp. 25-40, (2012)
[9] B. Xie et al., Microelectron. Eng, 76 (1-4) p. 52-59 (2004)
[10] P. Eaton et al., J. Appl. Polym. Sci. 82 (8), p. 1097-4628 (2001)
[11] A. Zenasni et al., J. Electrochem. Soc. 154 (1) p. G6 –G12 (2007)