Moisture uptake of MSQ/PS-PB-PS hybrid films

Next, a tri-block polymer, PS-PB-PS was used as a high-temperature porogen for comparison. The moisture uptakes of MSQ/PS-PB-PS compared to MSQ/PS-P4VP films were shown in Figure 4.8 and Table 4.2. The absorption and

desorption curves were attached in Appendix B as reference.

idea mixing of MSQ/PS-PB

Figure 4.8 Moisture uptake of MSQ/PS-b-P4VP and MSQ/PS-PB-PS hybrid films

Table 4.2 Moisture uptake of MSQ/PS-PB-PS hybrid films Loading Moisture uptake

(MSQ/PS-PB-PS),

The moisture uptake in MSQ/ PS-PB-PS hybrid films were much lower, < 1.0 wt% as compared to <3 wt% MSQ/PS-P4VP hybrid films. For example, at 10%

loading, the moisture adsorption of MSQ/PS-PB-PS films (1 wt%) was lower than MSQ/PS-P4VP films (1.67 wt%). The difference primarily originated from the low moisture uptakes of pure PS-PB-PS porogen, 1.8 wt% compared to 6.7 wt % in pure

PS-P4VP due to its polar pyridine moiety. We also observed that moisture uptake in MSQ/ PS-PB-PS hybrid films did not increase with increasing loading, instead remained relatively constant. The deviation from linear model could be attributed to increased level of moisture uptake at film/Au interface in QCM case. Overall, a hydrophobic PS-PB-PS as high temperature porogen could reduce moisture adsorption of hybrid films significantly as compared to PS-P4VP poorgen.

Finally, we also examined the desorption of moisture in MSQ/PS-PB-PS hybrid films. Similar to MSQ/PS-P4VP, the adsorption/desorption curve of MSQ/PS-PB-PS hybrid film could also reverse to its initial state. Thus, use of PS-P4VP and PS-PB-PS as high temperature porogen in MSQ/porogen hybrid ILD films had no difference after degassing under vacuum.

4.3 Dielectric properties

To understand the relationship between moisture uptake and dielectric properties, various porous MSQ films with and without HMDS treatment were measured at dry and RH 100% condition, shown in Figure 4.9. A porous MSQ films possessed a low dielectric constant of 2.46 when 30% porogen was added. The experimental data showed that dielectric constants of porous MSQ at dry condition were lower than those under RH 100% condition due to moisture adsorption. The amount of k reduction in dry condition increased with increasing porogen loading, in good agreement with the increasing trend of moisture uptake. After HMDS modification, the dielectric constants of porous MSQ films only slightly decreased when surface hydroxyl groups were eliminated by HMDS. This indicated after HMDS modified the surface became more hydrophobic. In addition, the number of polar Si-OH sites was relatively small compared to the total surface area in porous MSQ films. This observation was in good agreement with our finding from moisture uptake

study using QCM. Moreover, QCM method was more sensitive in qualitatively measuring the existence Si-OH compared to dielectric measurement using CV method.

0 5 10 15 20 25 30

2.4 2.5 2.6 2.7 2.8 2.9

Dielectric constant

Porogen loading Porous MSQ at dry condition HMDS modified at dry condition Porous MSQ under RH 100%

HMDS modified under RH 100%

Figure 4.9 Dielectric constants of porous MSQ and HMDS-modified porous MSQ films at dry and RH 100 % conditions.

We then examined the dielectric properties of Solid-First^TM MSQ/porogen hybrid films cured at 250 ^oC under dry or wet conditions. The dielectric constants of MSQ/PS-P4VP hybrid films under dry and wet condition were shown in Figure 4.10.

The dielectric constants of hybrid films at RH 100% condition increased from 2.7 to 5.0 if porogen loading is increased from 0 to 30%. Their dielectric constants were much higher than those under dry condition, especially at high porogen loadings, due to high moisture uptake of hybrid film and kwater~78.

To sum up, the deviation of dielectric constant between HMDS modified and un-modified porous MSQ at dry and RH 100% films was very small, since the capacitance was strongly dependent on pore morphology and geometry. In contrast the HMDS modified and un-modified porous MSQ at dry and RH 100% films measured by QCM were more comparable and clear. Therefore, QCM was a powerful and sensitive tool for direct estimating amount of moisture adsorption in low-k films without further considered pore morphology and geometry.

0 10 20 30

2.5 3.0 3.5 4.0 4.5 5.0

Dielectric constant

Porogen loading

MSQ/PS-P4VP hybrid films at dry condition MSQ/PS-P4VP hybrid films under RH 100%

Figure 4.10 Dielectric constants of MSQ/PS-P4VP hybrid films at dry and RH 100% conditions

Model of dielectric constant of porous low-k materials

Since the amount of adsorbed H2O bad been measured by QCM, the theoretic dielectric constants of porous MSQ films can by three different possible models such as (1) series model, (2) parallel model as shown by Figure 4.11, and (3) effective medium approximation (EMA model) assuming closed, spherical pores[64].

Figure 4.11 Series model, parallel model and EMA model of capacitance [64]

The total dielectric constants of the series and parallel models are expressed by Eq. 1 and 2, respectively

To calculate moisture induced dielectric constant difference, series and parallel model were employed. The models were shown in Figure 4.11

( )

( )

Where x is the porosity, y is the volume fraction of water, KMSQ is the dielectric constant of MSQ, Kwater and Kair are dielectric constants of water and air. The QCM moisture adsorption data and porosity were substituted into these three models to calculate the dielectric constants summarized in Figure 4.12. The calculations showed

that series and parallel models deviated significantly from experimental result, while EMA model strongly correlate with the dielectric properties of the porous MSQ in this study. This implied that the capacitance was strongly dependent on pore morphology and geometry. The pore morphology of these porous MSQ and MSQ/high- temperature hybrid films will be studied in the future.

0 5 10 15 20 25 30

1.8 2.1 2.4 2.7 3.0 3.3 3.6

K

porosity

EXP EMA parallel series

porous

Figure 4.12 Calculated dielectric constants based on three models: series model, parallel models, and EMA model as function of loading vs. experimental data

4.4 Diffusion behavior

From manufacturing considerations, it was important to understand the diffusion behavior of moisture in dense MSQ, porous MSQ and the MSQ/porogen hybrids in order to design appropriate outgassing pretreatment prior to its subsequent processing steps, such as TaN barrier and etch-stop layer deposition to avoid any delamination caused by tremendous pressure from trapped moisture. In MSQ/PS-P4VP system, the moisture uptake of hybrid films could be desorbed or reversed to initial state (as illustrated in appendix A) completely within time scale (<

150 seconds), which made it easier for industry to add an additional outgassing step to eliminate trapped moisture. The outgassing time could be shortened at elevated temperature.

We attempted to characterize the diffusion coefficients and the diffusion mechanism of low-k films assuming that the diffusion followed Fickian diffusion behavior. Basically, the diffusion behavior could be modeled by following equation [75]. M∞: the equilibrium mass uptake l: film thickness

D: diffusion coefficient.

A typical fitting curve illustrated in Figure 4.13 validated our assumption of Fickian

diffusion. Diffusion coefficient was then obtained from the fitting curve based on goodness of fit. Figure 4.14 summarized the diffusion coefficients of (1) MSQ/PS-P4VP hybrid films cured at 250 ^oC, (2) porous MSQ low-k films cured at 400 ^oC, and (3) Porous MSQ low-k films with HMDS treatment. Excluding the hybrid film with 30% porogen loading (D ~1.0×10^-14 m²/sec), the rest of diffusion coefficients were between 1x10^-15 and 2x10^-15 m²/sec. Moreover, the diffusion coefficient of pure PS-P4VP porogen film cured at 250 ^oC, was extracted to be 2.4×10^-14 m²/sec, which was higher than low-k films due to its high free volume in PS-P4VP. For MSQ/hybrid films, one would expect its diffusion coefficient higher than dense MSQ film. Instead, to our surprise, diffusion coefficients of hybrid, porous and HMDS-treated films were almost the same except for hybrid film with 30%

porogen loading. This implied that the diffusion process was dominated by a rate controlling step. A thin, but dense skin-layer on the surface of low-k films hypothesized.

Figure 4.13 Typical fitting curve based on Fickian diffusion

0 5 10 15 20 25 30 10^-16

10^-15 10^-14

Diffusion coifcient

Porogen loading Hybrid film

Porous film

After HMDS modified

MSQ/PS-P4VP

Figure 4.14 Diffusion coefficients of various low-k films versus different porogen loadings

This hypothesis was further confirmed by SEM viewgraphs of porous low-k films with 7.9 % porosity shown in Figures 4.15 (a)-(c) for (a) as-cured, (b) after FIB ion etching for 5 second with topview, and (c) cross-sectional view of as-cured porous MSQ film. No porosity was observed for as-prepared sample illustrated in Figure 14(a). In contrast, pore structures appeared after a short ion-etch, implying the existence of a skin layer in MSQ/PS-P4VP system, which was further confirmed by a cross-sectional SEM shown in Figure 14.15 (c). The skin layer was estimated to be 10-30 nm. However, detailed analysis by XRR will be carried out in a separate study.

Based on the physical evidence of a skin-layer by FIB/SEM and diffusion coefficient data, a bilayer low-k structure with a thin, but dense skin-layer was proposed and illustrated by Figure 4.16. The skin layer was speculated to be MSQ without porosity and behave as a rate controlling step in the moisture diffusion process.

(a) as-cured; before etch (b) After etch 5 sec

(c)

Figure 4.15 FIB/SEM topview of (a) as-cured, (b) after 5-second sputter etch and (c) cross-section view SEM of a porous MSQ film with 7.9 % porosity

based on PS-P4VP porogen.

Figure 4.16 Schematic diagram of a skin layer/porous low-k matrix stack under the diffusion of moisture

In contrast, no skin layer was found in MSQ/PS-PB-PS hybrid films as shown in Figure 4.18. The possible mechanism of formation of skin layer in MSQ/PS-P4VP system was illustrated in Figure 4.17. During the spin-coating step, stable micelles were formed due to amphiphilic nature of PS-P4VP di-block copolymers, and then evaporation of solvent occurred at the top [76]. The concentration of solute species was much higher at surface, and thus solute species in sol-gel solutions tended to readily form a thin layer by condensation or cross-linking reactions. Within the matrix, the porogen would aggregate, if the inside of the gel did not undergo further condensation.

They were two reasons that there was no skin layer in MSQ/PS-PB-PS hybrid films. First, stable micelles could not form since hydrophobic PS and PB segment could not act as amphiphilic block copolymers. The other reason was the low boiling point (66 ^oC) of MSQ/PS-PB-PS solvent, tetrahydrofuran (THF). In contrast, the solvent for MSQ/PS-P4VP system, n-butanol possessed high boiling point, 117 ^oC.

During the curing process of MSQ/PS-PB-PS, fast outgassing of bulk solvent made it

difficult to form any skin layer.

Figure 4.17 The proposed mechanism for skin layer formation in MSQ/PS-P4VP hybrid films

Figure 4.18 FIB/SEM top-view of as-cured MSQ/PS-PB-PS hybrid film with 5%

porogen loading

Eliminating trapped moisture was an important pretreatment step for most of processing steps in low-k integration, several approaches such as nitrogen purge and thermal degas were proposed [77, 63]. A Solid-First^TM scheme based on MSQ/high-temperature porogen has been proposed to defer the removal of porogen until the completion of a Cu/low-k interconnect layer. However, there is still little

understanding of the adsorption behavior and outgassing behavior of Solid-First^TM low-k dielectric materials. From QCM measurement in this thesis study, the sorption/desorption curve showed that moisture adsorption in MSQ/PS-P4VP and MSQ/PS-PB-PS hybrid films could be reversed to initial state (0 wt%, desorption at

~10^-3 torr) even though their moisture uptake were so different. The absorbed moisture could be desorbed or pumped out in a short time (<200 sec), which made it easier for industry to add a short outgassing step to eliminate the trapped moisture avoiding any blistering or delamination. The outgassing time could be further shortened if ougassing pretreatment was carried out at an elevated temperature.

Moreover, a skin-layer was found in MSQ/PS-P4VP porogen hybrid films because of the amphiphilic nature of porogen and the high boiling point of n-butanol solvent. The skin layer, a dense MSQ layer behaved as a rate controlling step in the moisture diffusion process resulting in unvarying diffusion constants for porogen loading <

30% in MSQ/PS-P4VP hybrid films. The skin layer on the top of porous MSQ surface will offer the same barrier property and interfacial characteristics as the pure, dense MSQ.

Based on moisture uptake analysis using QCM and dielectric properties using CV-dot measurement, the mechanism of moisture uptake in porous MSQ films or MSQ/high-temperature porogen hybrid films could be summarized in two modes:

physical sorption and chemical sorption. In physical sorption, the surface (1) within the MSQ or porogen matrix, (2) inside the pores, and (3) at the MSQ/substrate or porogen/substrate interface, interacted with the moisture adsorbent through a long range but weak Van der Waals force. In contrast, chemical sorption led to a formation of hydrogen bonding between silanol (Si-OH) and water as illustrated in Figure 4.19.

This hydrogen bonding made the complete H2O desorption from surface very difficult

unless high temperature (200~400 ^oC) was applied to overcome the bonding energy[78]. Our moisture sorption and desorption curves of the low-k films studied in this thesis all showed reversible behavior, indicating their adsorptions 30 ^oC were all in physical sorption mode. However, in Section 4.2.3, HMDS surface treatment could reduce moisture uptake further by 11-17% through chemical reaction to eliminate the polar Si-OH sites as previously illustrated by Figure 2.27. This implied that hydrogen bonding at residual Si-OH sites formed immediately after our samples were prepared, and remained intact in the subsequent moisture uptake and desorption processes. The hydrogen bonding Si-OH sites induced multilayered adsorbents through Van der Waals force, a physical sorption mode.

Figure 4.19 (a) hydrogen-bonding between silanol and water and (b) condensation reaction of silanols

Chapter 5 Conclusions

In this thesis, the impact of high-temperature porogens, their loadings on the moisture uptake and diffusion behavior was investigated for low-k films based on Solid-First^TM approach. Specifically, di-block and tri-block copolymers such as PS-P4VP and PS-PB-PS were employed as the high temperature porogens, whose decomposition temperature was higher than 300 ^oC. Three low-k dielectric systems were comprehensively studied in this thesis, namely: (1) MSQ films cured at different temperatures, (2) MSQ/porogens hybrid films with various porogen loadings, cured at 250 ^oC, which simulated the starting ILD materials in the Solid-First^TM scheme, and (3) porous films after porogens were completely removed by burn-out at 400^oC, which simulated the final porous low-k materials after the completion of Solid-First^TM integration scheme of Cu/low-k interconnect. The moisture uptake of low-k films was investigated by using a home-built quartz crystal microbalance (QCM).

The moisture uptake of porous MSQ increased from 0.51 % to 1.77 % with raising porosity up to 29 %. The increased moisture absorption could be attributed to (1) increased surface area and (2) increased Si-OH sites on the pore surface due to increased surface area and increased polarity such as Si-OH on the pore surface due to incomplete crosslinking caused by the steric hindrance effect of porogen. High surface area could provide more sites for adsorbents [66], while high concentration of Si-OH could form hydrogen bonding with H2O and even induce multilayered adsorbents through Van der Waals force. Moreover, the deviation of moisture uptake at 30% porogen loading from linearity was presumably caused by a change of pore morphology and/or increased Si-OH at higher degree, which required more in the future.

For MSQ/PS-b-P4VP hybrid low-k films cured at 250 ^oC, which could be used as the starting ILD for copper damascene process, the moisture adsorption were much larger than their corresponding porous MSQ. In addition, the deviation from ideal mixing rule increased with increasing porogen loading. Several factors contributed to such high moisture uptake; namely: (1) high moisture uptake of PS-P4VP porogen (6.7 wt%), (2) increased level of the residual Si-OH groups on the pores surface resulting from incomplete crosslinking at low cure temperature at 250 ^oC due to steric effect and additional interaction by the polar pyridine moiety of porogen, and possibly (3) increased moisture uptake at MSQ/PS-P4VP-Au substrate or porous MSQ-Au substrate interface.

For a tri-block copolymer, PS-PB-SP porogen, the moisture uptake in MSQ/PS-PB-PS hybrid films were much lower, ≤ 1.0 wt% as compared to MSQ/PS-P4VP hybrid films, ≤ 3 wt%. For example, at 10% loading, the moisture adsorption of MSQ/PS-PB-PS films (1 wt%) was lower than MSQ/PS-P4VP films (1.67 wt%). The difference primarily originated from the low moisture uptakes of pure PS-PB-PS porogen, 1.8 wt% compared to 6.7 wt % in pure PS-P4VP due to its polar pyridine moiety. We also observed that moisture uptake in MSQ/ PS-PB-PS hybrid films did not increase with increasing loading, instead remained relatively constant. The deviation from linear model could be attributed to increased level of moisture uptake at film/Au interface in QCM case. Overall, a hydrophobic PS-PB-PS as high temperature porogen could reduce moisture adsorption of hybrid films significantly as compared to PS-P4VP porogen.

HMDS pre-treated porous MSQ films cured at 400 ^oC showed 11-17% reduction in moisture uptake could be attributed to the elimination of residual silanol groups, which played a minor role (< 20%) in the overall moisture uptake. Based on our studies in the thesis, such moisture adsorption was believed to be physical sorption

mode by forming a multilayer H2O adsorbent through long-range Van der Waals force with the hydrogen-bonded Si-OH--H2O, which was formed immediately after sample preparation. Elimination of such Si-OH groups required either by high-temperature annealing (> 280 ^oC) or chemical reaction such as HMDS treatment employed in this thesis. Moreover, the variation of dielectric constant between HMDS modified and un-modified porous MSQ at dry and RH 100% films was very small. In contrast the HMDS modified and un-modified porous MSQ at dry and RH 100% films measured by QCM were more comparable and clear. Therefore, QCM was a powerful and sensitive tool for direct estimating amount of moisture adsorption in low-k films.

The diffusion behavior of moisture uptake and desorption in the MSQ/porogen hybrid and porous MSQ films were also investigated in this thesis. The sorption of moisture in porous MSQ and MSQ/high-temperature porogens (PS-P4VP and PS-PB-PS) was found to be Fickian diffusion and very fast (< 200 seconds) and the absorbed moisture could be completely desorbed or pumped out in a short time (< 200 sec), even though their equilibrium moisture uptake may Such reversible characteristics indicated the moisture sorption at 30 ^oC was purely in physical sorption mode. Therefore, for IC industry, a short outgassing pre-treatment step at room temperature or elevated temperature can be easily added in the low-k integration steps to eliminate the trapped moisture avoiding any blistering or delamination.

Furthermore, the diffusion constants of the MSQ/porogen hybrid and porous MSQ films were obtained by fitting the sorption curves based on Fickian diffusion. It was found that the diffusion coefficients of porous MSQ films (400 ^oC cure) at porogen loading ≤ 20% were relatively constant ranging from 1.0x10^-15 to 2.0 x10^-15 m²/sec, while the diffusion coefficient of PS-P4VP porogen film was much higher at 2.4×10^-14 m²/sec. The unvarying diffusion coefficients in MSQ/PS-P4VP system