Preparing Mesoporous Low-k Films with High Mechanical Strength

ARTICLE pubs.acs.org/IECR

Preparing Mesoporous Low-k Films with High Mechanical Strength from Noncrystalline Silica Particles Hsin-Yan Lu, Chin-Lin Teng, Chien-Hao Kung, and Ben-Zu Wan* Department of Chemical Engineering, National Taiwan University, Taipei 106, Taiwan. ABSTRACT: The coating solutions composed of surfactant Tween 80 and noncrystalline silica particles with an average size of 4.6 nm were synthesized through a short hydrothermal process. They were spun onto silicon wafers to obtain mesoporous low-k films. The film properties were compared with those of the films made of MFI zeolite nanocrystals with an average size of around 40 nm. The films from noncrystalline particles showed higher surface uniformity and higher mechanical strength, and possessed k values of smaller than 2 and low leakage current densities, which all reached the needs of the future IC industry. The preparation processes and time for the noncrystalline silica particles are much simpler and shorter than those for the MFI zeolite nanocrystals. As a result, the processes developed in this research can save the production cost and time for producing low-k films in the future.

1. INTRODUCTION To solve the problem of the resistive-capacitive time delay (RC delay) caused by parasitic resistance and parasitic capacitance, introducing Cu metal wires and low dielectric constant (k value) materials into an integrated circuit (IC) to substitute the traditional used metal wires, Al, and intermetal dielectric, dense silica (k value = 3.9), is considered to be imperative. Mesoporous silica film, a combination of solid silica and mesopores filled with air, is a potential candidate for the application as an interlevel dielectric (ILD), because the k value of this film can be controlled through adjusting the pore volume. According to the report established by ITRS in 2009, the required k value should be lower than 2 for future application in the integrated circuit (IC).1 Several previous studies have produced mesoporous silica low k value (low-k) films using the surfactant-template method,2-12 which could prepare the films with k values lower than 2 by simply increasing the porosity of the film. However, the mechanical strength of the films was not strong enough to withstand the process of chemical-mechanical-polishing (CMP) during the later IC manufacturing processes.8,12 The problem is pronounced especially for the films with k values lower than 2. To meet the needs of practical applications, and meanwhile, to maintain the k value lower than 2, a strategy or new method to improve the mechanical strength is urgently needed.10-12 The requirements of mechanical strength for the industrial applications are that the elastic modulus should be higher than 10 GPa and that the hardness should be higher than 1 GPa.12 Since 2001, it has been claimed that the low-k films with puresilica-zeolite (PSZ, a microporous crystalline material) of MFItype can possess strong mechanical strength.13-16 The coating solutions containing PSZ nanocrystals were prepared through hydrothermal processes with different synthesis temperatures. A centrifuge process was applied later for removing the large crystals. Then the solutions were spun on a silicon substrate to form the films. The films with a k value of 2.1 and an elastic modulus of between 16 and 18 GPa were reported by Wang et al. r 2011 American Chemical Society

in 2001.15 The film from a two-stage hydrothermal process in the study conducted by Li et al. in 2004 was reported to possess a k value as low as 1.6 and an elastic modulus of 1.6 GPa.16 The PSZ MFI low-k films presented above seem to meet all the requirements in the future IC industry. However, Eslava et al. repeated the experiment conducted by Wang et al. and obtained the results that would be fatal for the use of this kind of low-k films in the IC industry.17,18 They observed that large voids (larger than 5 nm) were formed in the films because of the packing of PSZ MFI nanocrystals. This would cause a sealing problem that Cu metal would diffuse into the pores of low-k films in the later IC manufacturing processes, resulting in electrical degradation of electronic devices. Eslava et al. demonstrated that the surface morphology and the mechanical strength of the PSZ MFI low-k films would become worse with the increasing of the nanocrystal size. The same result was reported by Johnson et al. who synthesized zeolite low-k films from coating solutions containing nanocrystals of another type PSZ (MEL type).19 In our previous study,20 the solutions with PSZ MFI nanocrystals were prepared using a two-stage hydrothermal process (60 °C for 2 days in the first stage, and 100 °C for 36 or 48 h in the second stage) and a centrifuge process (for removing the large crystals), following the processes developed by Li et al.16 It was also found that when the films spin-coated from the solutions, the resulting surface morphologies observed through an optical microscope were very rough or even cracked, and the steady electronic and mechanical strength data could not be obtained. However, after the addition of Tween 80 (a nonionic surfactant) into the solutions, the surface morphologies of the spin-coated films from the resulting coating solutions were much improved and the electronic and mechanical strength properties could be measured. Moreover, the crystallinity and the pore volume of the Received: September 2, 2010 Revised: January 28, 2011 Accepted: February 3, 2011 Published: February 18, 2011 3265

dx.doi.org/10.1021/ie1018436 | Ind. Eng. Chem. Res. 2011, 50, 3265–3273

Industrial & Engineering Chemistry Research films increased substantially with the increase of the hydrothermal period. The films prepared from a longer period (48 h in the second stage) possessed lower k values, very weaker mechanical strength, and higher leakage current. Therefore, to produce a film with higher mechanical strength, a shorter hydrothermal period should be used for the preparation of coating solution. However, only noncrystalline particles should be produced in the solution by using very short hydrothermal period. In this study, the attempt was made to prepare low k films from the coating solutions mainly containing noncrystalline silica particles prepared by using a very short period of hydrothermal process. The films with strong mechanical strength were expected. Moreover, because of the small size of the noncrystalline silica particles, the centrifugation process was no longer needed for the removal of large particles. In other words, if the coating solution mainly containing noncrystalline silica particles can be used for the preparation of low k films with strong mechanical strength, the short hydrothermal processes should be used in the future IC industry. Consequently, some energy and time will be saved, in comparison to those for the preparation of the coating solution containing crystalline particles (e.g., PSZ MFI nanocrystals).

2. EXPERIMENTAL SECTION 2.1. Preparation of Coating Solutions Containing Particles with Different Crystallinity and Preparation of Mesoporous Low-k Films. The coating solutions containing noncrystalline

silica particles and surfactant were synthesized through the following steps. First, TEOS (>99 wt %, Merck), EtOH (Acros Organics), TPAOH (25 wt %, Acros Organics), and DI water were mixed in a molar composition ratio of 1TEOS/5.6EtOH/ 0.36TPAOH/12.2H2O to form a precursor solution, and the mixture was stirred at 30 °C for 3 h. It was then heated to 100 °C and held for 24 h in a hydrothermal process for generating the noncrystalline silica particles. The resulting solution was mixed with surfactant Tween 80 (Polyoxyethylene(20)sorbitan monooleate, Acros Organics) with a weight ratio of Tween 80 to TEOS of 0.41 (or 0.65) to 1 to form the coating solution. Some of the coating solutions were diluted with ethanol to form new coating solutions. The coating solution was spin-coated on a precleaned 4-in. P-type (1 0 0) silicon wafer (Siltronic AG) at 2600 rpm for 30 s, using an SSP-01A spinner (King Polytechnic Engineering Co.). The precleaned wafer was first cleaned with HF (48-51 vol%, Acros organic) solution and DI water with a volume ratio of 1HF/10H2O for 5 min, then immersed in a mixture of H2O2 (35 wt %, Acros organics), NH4OH (28-30 wt %, Acros organics), and DI water, with a volume ratio of 1H2O2/1NH4OH/50H2O, for 20 min. The spin-coated film was then baked at 150 °C for 1 h and calcined at 450 °C for 5 h in air. Finally, after the film cooled down to 110 °C, it was immersed into a mixture of hexamethyldisilazane (Acros Organics) and toluene (Acros Organics) at 80 °C for 1.5 h to modify the surface to a hydrophobic state. The preparation of coating solutions containing crystalline particles and Tween 80 followed a two-stage hydrothermal process that employed in our previous study.20 The composition of precursor solution was the same as that for preparing coating solutions containing noncrystalline particles. The precursor solution was stirred at 30 °C for 3 day, and then hydrothermal synthesized through the two-stage process. The first stage was at 60 °C for 2 days, and the second stage was at 100 °C for 36 or 48 h. After the addition of Tween 80, a 24 h centrifugation process

ARTICLE

was applied to remove the large nanocrystals in the coating solutions. 2.2. Characterization. To study the pore size distribution and the crystallinity of each sample, the coating solution was dried at 60 °C for 3 h, 90 °C for 3 h, and calcined at 550 °C for 5 h to obtain powder samples. Each powder sample was characterized through nitrogen adsorption/desorption measurements at 77 K using a TriStar 3000 (Micromeritics) apparatus, to determine the pore volume and pore size distribution, and analyzed using X-ray powder diffraction (X’Pert PRO (PANalytical)) and FTIR (Spectrum 100, PerkinElmer) measurements to determine the particle crystallinity. Moreover, the particle size and particle size distribution of the coating solutions were determined with a dynamic light scattering analyzer, ZetaSizer Nano ZS (Malvern), with a laser wavelength of 633 nm. 0.1 g of the coating solutions was diluted a hundredfold with DI water. The diluted samples were placed into a transparent glass vial, which allowed the laser beam to pass, and then the vial was thermal equilibrated at 25 °C in the equipment. Finally, the measurement started and each sample was measured for 3 rounds and 12-15 times for a round within 3-5 min. The surface uniformity and the thickness of the films were obtained using a FE-SEM (LEO 1530). To measure various physical properties (dielectric constant, leakage current density, hardness, and elastic modulus), the films were analyzed with several types of equipment. The capacitance was measured using a Keithley model 82 CV meter. The frequency and the oscillation level were 1 MHz and 100 mV, respectively. The dielectric constant was calculated on the basis of the capacitance in the accumulation region of the capacitance-voltage curve, the film thickness, and the area of the electrode (which was 0.0052 cm2). The film thickness was measured using cross-sectional scanning on a FE-SEM (LEO 1530). The leakage current density of the film was determined from the current-voltage characteristics, measured using a HP4156 semiconductor parameter analyzer with an electric field of 1 MV/cm. The hardness and the elastic modulus of the film were measured using a Nano Indenter XP (MTS) system. The elastic modulus and hardness were measured by indenting the films with indention depth close to the film thicknesses through the continuous stiffness measurement (CSM) technique,21 and the Poisson ratio of the film was taken as 0.25.12 The data about the range of 1/10 thickness was taken for comparison. The average values and the standard deviations of elastic modulus and hardness were calculated, based on six measurements at six different locations on the film.

3. RESULTS AND DISCUSSION 3.1. Particle Size and Particle Crystallinity Characterization. The particle crystallinity and the particle size of noncrystal-

line silica and PSZ MFI nanocrystals for synthesizing mesoporous low-k films were examined. Figure 1 shows the X-ray diffraction patterns of powder samples prepared from three kinds of coating solutions containing particles with different crystallinity. The symbols in the figure HC-MFI, LC-MFI, and NCS represent the powder samples with high crystalline PSZ MFI nanocrystals, low crystalline PSZ MFI nanocrystals, and noncrystalline silica particles, respectively. As shown in the figure, the sample, HC-MFI, with the highest crystallinity showed apparent characteristic peaks of MFI-type zeolite. The sample was prepared from the coating solution, with a 48 h long period of the second stage hydrothermal process.20 As the period was 3266

dx.doi.org/10.1021/ie1018436 |Ind. Eng. Chem. Res. 2011, 50, 3265–3273

Industrial & Engineering Chemistry Research

ARTICLE

Figure 1. X-ray differaction patterns of powder samples from coating solutions containing particles with different crystallinities. Figure 3. Particle size distributions in different coating solutions.

Figure 2. FTIR spectra from powder samples from coating solutions containing particles with different crystallinities.

reduced to 36 h, the intensity of characteristic peaks of MFI-type zeolite (sample LC-MFI) was reduced substantially, indicating a reduction of particle crystallinity. It took a very long time to prepare the coating solutions for making samples HC-MFI and LC-MFI, including a 72 h aging step, a two-stage hydrothermal process, which started from the first hydrothermal step at 60 °C for 48 h and was followed by the second stage at 100 °C for 36 or 48 h, and a centrifugation step for 24 h.20 Such a long synthesis time, 156 or 168 h, ensured the formation of the crystal structure of MFI-type zeolite. To further lower the particle crystallinity, the period of the aging step was reduced to only 3 h, and the two-stage

hydrothermal process was simplified to only one stage, which was only at 100 °C for 24 h. From this much shorter process, a coating solution containing noncrystalline silica particles was obtained. The noncrystalline species was reported to be amorphous subcolloidal silica particles, which was the elementary units for the crystal growing of MFI-type zeolite during the crystallization process.22-24 The powder sample prepared from the coating solution containing the noncrystalline silica particles did not show any characteristic peaks of MFI-type zeolite through XRD characterization, as shown in Figure 1 (sample NCS). Even through the FTIR analysis, which was a more sensitive way to detect the presence of MFI-type zeolite, a pentasil framework vibration at 550 cm-1 that was assigned to be an index of the presence of MFI-type zeolite structure was not observed,22,23 as shown in Figure 2. In contrast, the other two crystalline samples (i.e., HC-MFI, LC-MFI) showed the pentasil vibration at 550 cm-1, and the optical density ratio of band 550 cm-1 to band 450 cm-1 increased from 0.07 of sample LC-MFI to 0.38 of sample HC-MFI, indicating the increase of particle crystallinity.25,26 Figure 3 presents the particle size distribution curves of particles with different crystallinity in the coating solutions, from dynamic light scattering (DLS) measurement. The distribution curve obtained from the particles with the highest crystallinity (sample HC-MFI) was similar to that obtained from the particles with lower crystallinity (sample LC-MFI), and both of them distributed between 10 and 100 nm. The mean value of particle sizes of both crystalline samples, HC-MFI and LC-MFI, were 40.3 and 35.4 nm, respectively. Those were much larger than that, 4.6 nm, of the noncrystalline silica particles (sample NCS). The size distribution of the noncrystalline silica particles was only between 2 and 10 nm, which was much narrower than those of the crystalline particles. 3.2. Nitrogen Adsorption/Desorption Analysis of the Particles with Different Crystallinity and Different Sizes. Figure 4a and b show the adsorption/desorption isotherms and the pore size distribution curves, respectively, of all the powder samples. As shown in Figure 4a, the isotherm of sample NCS was type-E hysteresis loop, and the isotherms of samples HC-MFI and LC-MFI presented combinations of typeE and type-C hysteresis loops. Both types of loops indicated the formation of mesopores with ink-bottle geometry.27 From the desorption isotherm of sample NCS, the hysteresis loop showed 3267



ARTICLE

Figure 4. (a) Nitrogen adsorption/desorption isotherms and (b) pore size distribution curves from the powder samples prepared from coating solutions containing particles with different crystallinity. The arrows pointing toward the upper right side are adsorption isotherms, and those pointing toward to the lower left side are desorption isotherms. dV/dD in panel b represents the pore volume (V) derivative with respect to pore diameter (D).

a rapid drop from relative pressure 0.48 to 0.4 (shown as a black arrow toward the lower left side), revealing a uniform neck diameter of the ink-bottle mesopores. However, the hysteresis loop of sample LC-MFI presented a slower drop from relative pressures 0.6 to 0.4 (shown as a red arrow toward the lower left side). Such a loop was considered to be ink-bottle mesopores with various neck diameters. A similar loop, but with a much slower drop, was observed on the isotherm from relative pressures 0.85 to 0.4 of sample HC-MFI (shown as a blue arrow toward the lower left side), indicating that the neck diameters of the ink-bottle mesopores were distributed in a much wider range. From Figure 4b, the mesopore size of sample NCS was mainly distributed between 3 and 4 nm. The mesopore sizes of samples, LC-MFI and HC-MFI, showed a wider distribution between 3 and 5 nm, and a much wider distribution between 3 and 7 nm, respectively. This was in agreement with the hysteresis phenomena occurred in the isotherms in Figure 4a. The formation of the wide pore size distributions of the crystalline samples (i.e., HC-MFI and LC-MFI), shown in Figure 4b, were considered to be a consequence of their large particle sizes and the broad size distributions. In the coating solutions containing the crystalline particles and surfactant Tween 80, the micelles formed from the surfactant were surrounded by the nanocrystals. Once the surfactant was removed, the mesopores were formed not only by removing the micelles but also by the void between the packing of the large nanocrystal particles. For sample LC-MFI, the mesopores sizes between 3 and 4 nm (the primary band in Figure 4b) was considered to be resulted from removing micelles. The other small band between 4 and 5 nm was considered to be formed between the packing of large particles. For sample HC-MFI, the primary mesopores from removing micelles was between 3.5 and 4.5 nm. The larger pores between 4.5 and 7 nm were from the particle packing, which were wider and larger than those from the particle packing of sample LC-MFI, because sample HC-MFI possessed larger nanocrystals, as shown in Figure 3. In contrast, the pore size distribution of sample NCS was the most uniform one of all the samples shown in Figure 4b. Only mesopores distributed between 3 and 4 nm from removing micelles can be observed. No larger pores can be formed because the noncrystalline silica

Figure 5. Hardness and elastic modulus of the mesoporous low-k films, exhibited with open square and red solid sphere symbols, respectively, and error bars (representing standard deviation of hardness or elastic modulus), and the total pore volume of powder samples, exhibited with open triangle symbols, prepared from coating solutions containing particles with different crystallinities.

particles were densely packed. In summary, as the particle size increases, the wider distribution of mesopores would be obtained. Furthermore, the total pore volumes of all the samples, HC-MFI, LC-MFI, and NCS, were 0.463 cm3/g, 0.397 cm3/g, and 0.358 cm3/g, respectively, which decreased with the decrease of the particle crystallinity of all the samples. 3.3. Effects of Particle Crystallinity, Total Pore Volume, and Surface Uniformity on the Mechanical Strength of Mesoporous Low-k Films. The effects of particle crystallinity on mechanical strength of the low-k films are shown in Figure 5. It can be observed that the hardness and the elastic modulus of the mesoporous low-k films decreased with the increasing of the particle crystallinity, and that of the film prepared with the noncrystalline silica was the highest (sample NCS). The hardness of sample NCS was 1.81 GPa, as shown in Figure 5, which was more than 3 times higher than 0.44 GPa of the films prepared from the particles with the high crystallinity (sample HC-MFI) and which was 30% higher than 1.39 GPa of the films with the 3268



ARTICLE

Figure 6. Surface topography images, (a) a lower resolution image with a scale bar of 1 μm meter, and (b) a higher resolution image with a scale bar of 200 nm, of the mesoporous low-k films prepared from coating solutions containing particles with different crystallinities.

low crystallinity (sample LC-MFI). The ratio of standard deviation (shown with error bars) to mean of the hardness (square symbols) of sample NCS was 6%, which was much less than 22% of that of sample LC-MFI and 34% of that of sample HC-MFI. The smaller ratio of standard deviation to mean of the hardness suggests the better uniformity of the film. Similarly, the elastic modulus of the films followed the decreasing trend with large standard deviation as the particle crystallinity increased. These indicate that the hardness, the elastic modulus and the uniformity of sample NCS were much better than those of the crystalline samples (i.e., LC-MFI and HC-MFI). Note, it took only 24 h hydrothermal process to prepare a solution containing the noncrystalline silica particles, instead of 84 or 96 h of two stages hydrothermal processes for crystalline particles. It is not necessary to apply centrifugation processes to remove the large particles, because only the small particles are in the solution, as shown in Figure 3. Therefore, a mesoporous low-k film with high mechanical strength can be successfully produced with the short hydrothermal period of 24 h. Such an efficient process can save time and energy (i.e., the cost) for producing low-k films. The particles with higher crystallinity had been expected to possess higher mechanical strength than those with lower crystallinity. However, the hardness of the resulting films in this research is in the opposite trend. The measurements of pore volumes may provide the answer for the trend. Figure 5 shows the relation of pore volume and crystallinity of the samples. It can be observed that the total pore volume of NCS (a noncrystalline sample) was 10% less than that of LC-MFI (the low crystalline

sample), and the mechanical strength (i.e., hardness) of NCS was 30% higher than that of LC-MFI. This was in agreement with the general idea that the mechanical strength would decrease with the increasing of total pore volume of the film.8 However, as comparing those of the samples NCS and HC-MFI (the high crystalline sample), the total pore volume of the former was 23% less than that of the latter, but the hardness of the former was 311% much higher than that of the latter. The magnitude of the decreasing of hardness between these two samples was too much larger than that between samples NCS and LC-MFI. Even comparing with the two crystalline samples, HC-MFI and LCMFI, the total pore volume of the former was only 17% higher than that of the latter, but the hardness of the latter was 216% still much higher than that of the former. The magnitude of the increasing of the total pore volume was much less than that of the decreasing of the mechanical strength, as the particle crystallinity was increased. It suggests that the total pore volume was not the only factor to influence the mechanical strength of the mesoporous low-k films when the high crystalline particles were used to form the films. The surface uniformity characterization on the films through a scanning electron microscope may provide more information for the explanation. Figure 6 shows the surface topographies of film samples NCS, LC-MFI, and HC-MFI by using a scanning electron microscope. Two images with different resolutions were taken for each sample. The image of low crystalline sample LC-MFI in Figure 6a showed that the MFI nanocrystals were uniformly distributed on the surface. The particle sizes in the image with 3269



ARTICLE

Table 1. Electronic and Mechanical Properties of Mesoporous Low-k Films Prepared from Coating Solutions Containing Noncrystalline Silica Particles and Surfactant Tween 80 sample NCS NCS-

Tween 80 to TEOS

dilution

film thickness

total pore volume

(weight ratio)

ratioa

(nm)

(cm3/g)

k value

(10-7A/cm2)

Hb (GPa)

Ec (GPa)

0.41 0.41

0 7/3

380 310

0.358 0.375

2.26 ( 0.05 2.05 ( 0.04

0.18 1.42

1.81 ( 0.11 1.46 ( 0.19

15.48 ( 0.54 13.60 ( 0.50

0.41

5/5

201

0.372

1.93 ( 0.04

2.61

1.73 ( 0.30

17.04 ( 1.15

0.65

0

375

0.444

2.20 ( 0.05

8.85

1.46 ( 0.09

13.02 ( 0.32

leakage current

D1 NCSD2 NCS-1

Dilution ratio represents the weight ratio of the coating solution of sample NCS to ethanol. b H represents the hardness of the films. c E represents the elastic modulus of the films. a

higher resolution, shown in Figure 6b, were consistent with that obtained by DLS measurement shown in Figure 3. For the images of noncrystalline sample NCS, the surface exhibited a dark color and appeared a high uniformity. Only a few small sizes of subcolloidal silica particles embedded uniformly in the film can be observed. In contrast, the surface topography of the high crystalline sample HC-MFI was not as uniform as those of the others. Moreover, the images showed serious cracks with a width between tens of nanometer and a few micrometers on the film surface. They provide more evidence for the poor mechanical strength of those high crystalline films. The cracks were also observed and reported elsewhere.17 It is thought that the crack formation, in this research, may be a result of the lack of noncrystalline small particles, which can fill the cracks and the large voids formed by the large particle packing in the high crystalline films. Both large voids and cracks can cause the low mechanical strength of the films. Moreover, the crack formation is related to the stress development in the film that is resulted from the large difference of thermal expansion coefficients of noncrystalline particles (or silicon substrate) and high crystalline MFI nanocrystals. The volumetric thermal expansion coefficients of noncrystalline particles (amorphous silica) and silicon substrate are positive values of around 2 10-6/°C and 9 10-6/°C respectively (the values are taken as 3 times of the linear thermal expansion coefficients reported by Retajczyk and Sinha),28 and it is, however, a negative value of around -9 10-6/°C for high crystalline MFI nanocrystals.29 Such a large difference in volumetric thermal expansion coefficient is the reason for the crack formation, as shown in Figure 6 HC-MFI, some nanocrystals exhibit an intensive shrinkage to aggregate together due to the large negative value, resulting in the formation of cracks. In conclusion, because of the lack of noncrystalline silica (amorphous subcolloidal silica) filled in the void of high crystalline sample HC-MFI and because of the stress development in the films, the cracks formed, resulting in a very weak mechanical strength. On the other hand, for the sample NCS, because the film was built up only with the small noncrystalline silica particles, the voids in the particles packing should be small, the surface topography was much more uniform than the others, and no cracks were observed, thus resulting in that the mechanical strength of sample NCS was the highest. 3.4. Effect of Film Thickness on k Value of Mesoporous Films Prepared with the Noncrystalline Silica Particles. The film, sample NCS, prepared from the coating solution containing the noncrystalline particles possessed the highest mechanical strength. However, the k value of the film was 2.26, which was too high to meet the required k value of lower than 2 for the

application in the future IC industry. In general, the higher the total pore volume is, the lower the k value of the film will be obtained. Therefore, in order to obtain a smaller k value, a film with a higher total pore volume was prepared. Table 1 shows electronic and mechanical properties of mesoporous low-k films prepared from the coating solutions containing noncrystalline silica particles and surfactant Tween 80. A film prepared with a weight ratio of Tween 80/TEOS of 0.65 (sample NCS-1) possessed a k value of 2.20 and a total pore volume of 0.444 cm3/g, which was 27% higher than that of 0.358 cm3/g of sample NCS, prepared with a weight ratio of Tween 80/TEOS of 0.41. The k value of both samples, NCS and NCS-1, were very close to each other. However, this result was not in agreement with the general idea that a higher total pore volume would result in a lower k value, revealing that there must be the other factors to influence the k value. Besides the total pore volume, the water adsorption of the pore surface was also a serious problem resulting in a higher k value because of the high k value, 78, of water. From FTIR spectra shown in Figure 2, the sample NCS showed a peak at around 970 cm-1 that was defined as a stretching vibration of Si-OH groups, which would absorb water and resulted in a higher k value.30 Compared the peak intensity at the same band to the other samples shown in Figure 2, the peak intensity of sample NCS was stronger, indicating a higher amount of Si-OH group existed inside the film. The silanol group can be modified to be hydrophobic by HMDS, through a surface modification process that described in the Experimental Section. However, the modification might not be complete, when the film thickness was too thick. Figure 7 shows a cross section image of sample NCS and a schematic illustration of surface modification occurred inside a thick film prepared with the noncrystalline silica particles. While the modification was carrying out in the sample NCS with a thickness of 380 nm, the HMDS molecular was hard to diffuse deeply into the film because the film was packed densely with the subcolloidal silica particles, as shown in Figure 7a. As a result, some of the silanol groups located at lower or deeper parts inside the film can not be modified completely. Therefore, the k value cannot be further reduced because of the remained silanol groups, even though the total pore volume became higher (sample NCS-1 in Table 1). To further reduce the k value of mesoporous low-k films prepared with the subcolloidal silica particles, the thinner films should be prepared. As shown in Table 1, the coating solution for film sample NCS was diluted with ethanol to form two new coating solutions with two different dilution ratios. The resulting films were NCS-D1 and NCS-D2. The k value of the film, sample 3270



ARTICLE

Figure 7. (a) Cross section image of sample NCS and (b) a schematic illustration of surface modification occurred inside a thick film prepared with the noncrystalline silica particles.

Figure 8. Nitrogen adsorption/desorption isotherms and (b) pore size distribution curves from the powder samples prepared from coating solutions containing the noncrystalline silica particles with different dilution ratio.

NCS-D1, with a film thickness of 310 nm was 2.05, which was lower than 2.26 of the original film sample NCS. As the film thickness was further reduced to 201 nm, sample NCS-D2, the k value was further reduced to 1.93. From Table 1, the total pore volume of samples, NCS, NCS-D1, and NCS-D2, were very close to each other. Therefore, the reduction of k value was attributed to a more completed modification of silanol groups inside the mesoporous low-k films prepared with the subcolloidal silica particles. Figure 8a and b shows the nitrogen adsorption/desorption isotherms and the pore size distribution curves, respectively, of the powder samples, NCS-D1 and NCS-D2. As shown in Figure 8a, both the isotherms were type-E hysteresis loops, indicating the formation of mesopores with ink-bottle geometry.27 The mesopore sizes of samples NCS and NCSD1, as shown in Figures 4b and 8b, respectively, are close to each other. For the sample NCS-D2 prepared from a more diluted coating solution with a dilution ratio of 5/5, the mesopores shrunk to smaller pore sizes. It indicates that the micelles of Tween 80 became smaller in the more diluted NCS-D2 coating solution. Moreover, all the pores of the three samples were smaller than 5 nm. This can prevent the sealing problem in the films in the later IC manufacturing processes.17,18 The

mechanical strengths of all the films in Table 1 were all higher than that of the required values, elastic modulus of >10 GPa and hardness of >1 GPa, and the leakage current density of the films in Table 1 was between 0.18 and 8.85 10-7A/cm2. In conclusion, a mesoporous low-k film prepared from an ethanoldiluted coating solution, with a dilution ratio of 5/5, containing the noncrystalline silica particles and surfactant Tween 80 (Tween 80/TEOS = 0.41) possessed a k value of 1.93, a leakage current of 2.61 10-7A/cm2, an elastic modulus of 17.04 GPa, and a hardness of 1.73 GPa, which all meet the needs of the future IC industry. Moreover, the process for preparing the mesoporous low-k films with the noncrystalline silica particles was very short, so that the production cost can be saved.

4. CONCLUSIONS The noncrystalline silica particles with an average size of 4.6 nm and a narrow size distribution of between 2 and 10 nm can be synthesized from a short hydrothermal process. The particles have shown amorphous in XRD characterization, and shown no crystalline framework vibration in FTIR analysis. With a much complex and longer hydrothermal process, MFI zeolite nanocrystals with an average size of around 40 nm were also 3271


Industrial & Engineering Chemistry Research synthesized. It has been found that the total pore volume and the mesopore sizes of the powder sample prepared from the noncrystalline silica particles are less and smaller than those of the crystalline powder samples. Because of the small and uniform particle size of the noncrystalline silica particles, the large voids (>5 nm) formed from the packing of the nanocrystals are not present in the sample. The mechanical strength of the films made of the nanocrystals was much weaker than that of the films made of noncrystalline particles, due to the existence of cracks and large voids in the films. In contrast, the films made of noncrystalline particles showed high surface uniformity, and there were no cracks and large voids in the films, thus resulting in a much higher mechanical strength. However, due to the presence of many silanol groups on noncrystalline particles, the particles embedded in the deeper part of the thicker films cannot be completely modified to be hydrophobic, resulting higher k values. The reduction of the film thickness has to be carried out with ethanol dilution. Therefore, the thinner mesoporous low-k film prepared from the noncrystalline particles can possess a k value of 1.93, a leakage current of 2.61 10-7A/cm2, an elastic modulus of 17.04 GPa, and a hardness of 1.73 GPa, which all meet the needs of the future IC industry. As compared to the whole preparation processes and time for the films made of crystalline particles, those for the films made of noncrystalline particles are much simpler and shorter, which can save energy (the cost) and time for producing future mesoporous low-k films.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT We would like to thank the National Science Council of the Republic of China, Taiwan, for financially supporting this research, the National Nano Device Laboratory, Hsinchu, Taiwan, for the support of the measurement of electronic and mechanical strength properties. ’ REFERENCES (1) The International Technology Roadmap for Semiconductors, please see http://www.itrs.net/Links/2009ITRS/Home2009.htm, 2009. (2) Zhao, D.; Yang, P.; Melosh, N.; Feng, J.; Chmelke, B. F.; Stucky, G. D. Continuous Mesoporous Silica Films with Highly Ordered Large Pore Structures. Adv. Mater. 1998, 10, 1380–1385. (3) Baskaran, S.; Liu, J.; Domansky, Ka.; Kohler, N.; Li, X. Low Dielectric Constant Mesoporous Silica Films Through Molecularly Templated Synthesis. Adv. Mater. 2000, 12, 291–294. (4) Yang, C. M.; Cho, A. T.; Pan, F. M.; Tsai, T. G.; Chao, K. J. Spin-on Mesoporous Silica Films with Ultralow Dielectric Constants, Ordered Pore Structures, and Hydrophobic Surfaces. Adv. Mater. 2001, 13, 1099–1102. (5) de Theije, F. K.; Balkenende, A. R.; Verheijen, M. A.; Baklanov, M. R.; Mogilnikov, K. P.; Furukawa, Y. J. Structural Characterization of Mesoporous Organosilica Films for Ultralow-k Dielectrics. Phys. Chem., B. 2003, 107, 4280–4289. (6) Ting, C.-Y.; Ouyan, D.-F.; Wan, B.-Z. Preparation of Ultralow Dielectric-Constant Porous Silica Films Using Tween 80 as a Template. J. Electrochem. Soc. 2003, 150-8, F164–F167. (7) Ting, C.-Y.; Wu, C.-A.; Wan, B.-Z.; Wu, W.-F. Low Dielectric Consant Silica Films Prepared by a Templating Method. J. Chin. Inst. Chem. Engrs. 2003, 34, 211–217.

ARTICLE

(8) Ting, C.-Y.; Sheu, H.-S.; Wu, W.-F.; Wan, B.-Z. Porosity Effects on Properties of Mesoporous Silica Low-k Films Prepared Using Tetraethylorthosilicate with Different Templates. J. Electrochem. Soc. 2007, 154-1, G1–G5. (9) Jung, S.-B.; Park, H.-H. Concentration-Dependent Mesostructure of Surfactant-Templated Mesoporous Silica Thin Film. Thin Solid Films 2006, 494, 320–324. (10) Cho, A.-T.; Pan, F. M.; Chao, K. J.; Chen., J. Y. The Preparation of Mesoporous Silica Ultra-low k-film using Ozone Ashing Treatment. Thin Solid Films. 2005, 483, 283–286. (11) Ha, T.-J.; Choi, S.-G.; Jung, S.-B.; Yu, B.-G.; Park, H.-H. The Improvement of Mechanical and Dielectric Properties of Ordered Mesoporous Silica Films Using TEOS-MTES Mixed Silica Precursor. Ceram. Int. 2007, 34, 947–951. (12) Tsai, C.-T.; Lu, H.-Y.; Ting, C.-Y.; Wu, W.-F.; Wan, B.-Zu Increasing Mechanical Strength of Mesoporous Silica Thin Films by Addition of Tetrapropylammonium Hydroxide and Refluxing Processes. Thin Solid Films. 2009, 517, 2039–2043. (13) Wang, Z.; Wang, H.; Hitra, A.; Yan, Y. Pure-Silica Zeolite Low-k Dielectric Thin Films. Adv. Mater. 2001, 13, 746–749. (14) Wang, Z.; Hitra, A.; Wang, H.; Yan, Y. Pure Silica Zeolite Films as Low-k Dielectrics by Spin-On of Nanoparticle Suspensions. Adv. Mater. 2001, 13, 1463–1466. (15) Li, S.; Li, Z.; Yan., Y. Ultra-Low-k Pure-Silica Zeolite MFI Films Using Cyclodextrtin as Porogen. Adv. Mater. 2003, 15, 1528–1531. (16) Li, Z.; Li, S.; Luo, H.; Yan, Y. Effects of Crystallinity in Spin-On Pure-Silica-Zeolite MFI Low-Dielectric-Constant Films. Adv. Funct. Mater. 2004, 14, 1019–1024. (17) Eslava, S.; Kirschhock, C. E. A.; Aldea, S.; Baklanov, M. R.; Iacopi, F.; Maex, K.; Martens, J. A. Characterization of Spin-on Zeolite Films Prepared from Silicalite-1 Nanoparticles Suspensions. Microporous Mesoporous Mater. 2009, 118, 458–466. (18) Eslava, S.; Baklanov, M. R.; Neimark, A. V.; Iacopi, F.; Kirschhock, C. E. A.; Maex, K.; Martens, J. A. Evidence of Large Voids in PureSilica-Zeolite Low-k Dielectrics Synthesized by Spin-on of Nanoparticle Suspensions. Adv. Mater. 2008, 20, 3110–3116. (19) Johnson, M.; Li, Z.; Wang, J.; Yan, Y. Mechanical Characterization of Zeolite Low Dielectric Constant Thin Films by Nanoindentation. Thin Solid film 2007, 515, 3164–3170. (20) Lu, H.-Y.; Teng, C.-L.; Yu, C.-W.; Liu, Y.-C.; Wan, B.-Z. Addition of Surfactant Tween 80 in Coating Solutions for Making Mesoporous Pure Silica Zeolite MFI Low-k Films. Ind. Eng. Chem. Res. 2010, 49, 6279–6286. (21) Li, X.; Bhushan, B. A Review of Nanoindentation Continuous Stiffness Measurement Technique and Its Applications. Mater. Charact. 2002, 48, 11–36. (22) Ravishankar, R.; Kirschhock, C. E. A.; Knops-Gerrit, P.-P.; Feijen, E. J. P.; Grobet, P. J.; Vanoppen, P.; De Schryver, F. C.; Miehe, G.; Fuess, H.; Schoeman, B. J.; Jacobs, P. A.; Martens, J. A. Characterization of Nanosized Material Extracted from Clear Suspensions for MFI Zeolite Synthesis. J. Phys. Chem. B. 1999, 103, 4960–4964. (23) Dokter, W. H.; van Garderen, H. F.; Beelen, T. P. M.; van Santen, R. A.; Bras, W. Homogeneous versus Heterogeneous Zeolite Nucleation. Angew. Chem., Int. Ed. Engl. 1995, 34, 73–75. (24) Li, Q.; Mihailova, B.; Creaser., D.; Sterte., J. The Nucleation Period for Crystallization of Colloidal TPA-silicalite-1 with Varying Silica Source. Microporous Mesoporous Mater. 2000, 40, 53–62. (25) Coudurier, G.; Naccache, C.; Vedrine, J. C. Uses of I. R. Spectroscopy in Identifying ZSM Zeolite Structure. J. Chem. Soc., Chem. Commun. 1982, 24, 1413–1415. (26) Shukla, D. B.; Pandya, V. P. Estimation of Crystalline Phase in ZSM-5 Zeolite by Infrared Spectroscopy. J. Chem. Technol. Biotechnol. 1989, 44, 147–154. (27) Thomas, J. M.; Thomas, W. J. Principles and Practice of Heterogeneous Catalysis; Wiley-VCH: Weinheim, Germany, 1997. (28) Retajczyk, T. F.; Sinha, A. K. Elastic Stiffness and Thermal Expansion Coefficients of Various Referactory Silicides and Silicon Nitride Films. Thin Solid Films. 1980, 70, 241–247. 3272



ARTICLE

(29) Bhange, D. S.; Ramaswamy, V. Negative Thermal Expansion in Silicalite-1 and Zirconium Silicalite-1 Having MFI Structure. Mater. Res. Bull. 2006, 41, 1392–1402. (30) Kuwahara, Y.; Kamegawa, T.; Mori, K.; Matsumura, Y.; Yamashita, H. Fabrication of Hydrophobic Zeolites Using Triethoxyfluorosilane and their Application for Photocatalytic Degradation of Acetaldehyde. Top Catal. 2009, 52, 643–648.

3273


Preparing Mesoporous Low-k Films with High Mechanical Strength

Recommend Documents