Chapter 1 Introduction
1.3 Organization of Thesis
This work discusses the primary issues around porous low-k material integration.
Self-assembled molecularly templated mesoporous silica thin films were prepared as low-k dielectrics. Some thin-film characterization approaches are adopted to analyze the electrical, mechanical and thermal properties of the mesoporous silica thin films and metal/dielectric film stacks.
This dissertation is divided into seven chapters. The contents of each chapter are as follows.
Chapter 1 introduces the motivation and organization of this dissertation.
Chapter 2 reviews the literature. The factors that affect the dielectric constant of the materials are initially considered. Then, the general background of low-k materials, especially
mesoporous silica thin films, and their application in backend Cu dual damascene process is presented.
Chapter 3 introduces the experimental method and the analytic technology.
Chapter 4 discusses the dielectric properties and hydrophobicity of self-assembled templated mesoporous silica thin films.
Chapter 5 focuses on the microstructure and mechanical strength of mesoporous silica thin films.
Chapter 6 discusses film stresses in the mesoporous silica dielectric and its stacks. The results of the thermal stress experiment are also discussed. The coefficient of thermal expansion of the mesoporous silica thin film is derived.
Chapter 7 is concerned mainly with the thermal and electrical stability of mesoporous silica thin films. The thermal stability of metal/low-k film stacks was evaluated.
Finally, Chapter 8 summarizes all experimental results. Some unfinished work is mentioned and new issues to be tackled in future studies suggested.
Chapter 2
Literature Review
Many materials, including ceramics and polymers, are electrical insulators and candidates for IMD applications. However, the most promising dielectric is one with material characteristics that can withstand the manufacturing process. This chapter reviews the dielectric properties, material characteristics of low-k dielectrics materials. Both organic and inorganic low-k dielectrics are considered.
2.1 Electrical Performance of Low-k Dielectrics
The basic electrical phenomenon of a dielectric material is polarization: an electric dipole moment can be induced by an electric field or can be permanent. A dielectric material can exhibit three forms of polarization in an applied electric field - electronic, ionic and orientation [23, 24]. Electronic polarization arises when the cloud of bound electrons is displaced from the nucleus of an atom, resulting in an electric dipole moment. Ionic polarization refers to the distortion of the position of the nuclei by the applied field, and, therefore, a change in the bond length. Some also refer to this as atomic polarization. If a molecule has a permanent electric dipole moment, which arises from the different electronegativity or other features of bonding, then polar molecules can exhibit orientation polarization in an applied field. These three polarization phenomena reveal at the specific frequency of an ac field and determine the polarizability of materials. For instance, orientation polarization involves the motion of complete molecules, so it does not resonate at a critical frequency, as do electronic and ionic polarization. Electronic polarization dominates at high
frequency (~1015 Hz), while the other two polarization are nuclear responses and are important at lower frequencies (<1013 Hz). Currently, typical device operating frequencies are 109 to 1010 Hz, and all three polarization contribute to the dielectric constant, as discussed below. (The maximum response frequency of orientation polarization is of the order of 109 Hz). Therefore, the polarization of IMD materials should be minimized to optimize performance.
The relationship between the dielectric constant and the total polarization can be quantitatively described by the Clausius-Mossotti equation [24],
)
where εr the relative permittivity or dielectric constant of material, ε0 the permittivity of a vacuum and N the number of molecules per m3. The terms αe, αi and αo are the electronic, ionic and orientation polarizations in the molecule, respectively. According to the equation, the relative permittivity is smaller if materials contain fewer polar molecules. This fact offers an important clue to the fact that reducing the density N and the polarizabilities α can reduce the dielectric constant. Table 2.1 presents the polarizabilities of some typical chemical bonds [25]. C-F and C-C bonds have the lowest polarizability, indicating that the incorporation of fluorine atoms and reducing the number of double or triple bonds can effectively reduce the polarization of materials. The former is associated with high electronegativity, which causes tight binding of the electrons and increases free volume. The latter is associated with a drop in the mobility of the π electrons. Some fluorinated aliphatic hydrocarbon materials are designed using these rules for low-k applications. For example, polyimides, which are extensively adopted in microelectronics, have dielectric constant values in the range 2.9-3.4. Fluorine substitution can reduce the dielectric constant to a lower value of 2.6 [26].
Table 2.1 Electronic polarizability of some typical chemical bonds [25].
Although the contribution of polarization to the dielectric constant of materials is very important, the ways to reduce molecular polarizability are limited. Equation (2.1) also reveals that film density, like the chemical bonds, is important in reducing relative permittivity.
Reducing the density is considered to be more effective because it can reduce k to close to unity. It can be performed using lighter atoms and/or by incorporating more free space around the atoms. Technologically, the best way to reduce film density is to introduce pores. Treating the porous film as two component materials, enables Eq. (2.1) to be rewritten as [27],
( )
where εp and εm are the permittivity of the internal pores and the solid matrix of the film, respectively. The porosity of the film, Π, defined by Eq. (2.3), is the ratio that of the volume of the pores (Vp) to the total volume of the film (Vf). Equation (2.2) is also called the Bruggeman effective medium approximation and predicts the effect of porosity on the
dielectric constant. Figure 2.1 presents the theoretical predication of the model for SiO2 (k=4) and a material with an original k value of 2.5. The curves reveal that the dielectric constant decreases as a function of porosity slightly faster than linearly. This fact is very useful to the material designer who wishes to estimate the minimum porosity required for a desired k value.
For instance, to yield k=2.0, oxide materials require a porosity of about 50%. However, as presented in Fig. 2.1, a porosity of only ~30% is required for a k=2.5 low-k material, indicating that the lower k start value corresponds to a lower required porosity, potentially increasing the integration compatibility of the low-k materials. Additionally, the k values in Fig. 2.1 are ideal (for a dry film) and quite different from typical experimental values. In a real case, the reserve environment and the test methods affect the results. Most oxides contain many terminal OH groups and thus adsorb water. This trapped water (k≒78) or chemicals are likely to increase the dielectric constant and promote crack formation; a k value of below 2.0 is difficult to obtain even though the porosity is increased.
Figure 2.1 Bruggeman’s effective medium approximation showing dielectric constant versus porosity for oxide and a low-k material.
Relative dielectric constant
2.2 Potential Solutions for Creating Low-k Dielectrics
Many low-k materials, including organic and inorganic dielectrics, have been proposed as replacements for the traditional CVD SiO2. The dielectric constant of CVD SiO2 films is 4.0±0.2, and dielectrics with a k value of below 4.0 are defined as low-k dielectric materials.
SEMATECH has basically characterized several classes of candidate low-k materials. Table 2.2 presents the results. These candidate low-k materials are generally categorized as follows.
1. Oxides or organic polymers
2. Organic, inorganic or hybrid materials 3. CVD or spin-on methods
4. Fluorinated or nonfluorinated materials 5. Porous or nonporous materials
This section introduces the potential low-k materials. The material characterization and integration issues of them are presented for comparison.
2.2.1 Fluorosilicate Glass (FSG)
Fluorine atoms were initially incorporated into SiO2 to reduce the k value relative to that of the traditional silicate glass, SiO2, in the development of low-k materials and fluorosilicate glass, FSG. The fabrication of FSG involves chemical vapor deposition, which process extends traditional SiO2 manufacture. Therefore, this material has attracted the attention of chipmakers. Figure 2.2 presents the chemical structure of FSG. Since fluorine is very electronegative, the incorporation of fluorine atoms reduces the polarization of SiO2, decreasing the dielectric constant, as described in section 2.1. Additionally, more terminal Si-F bonds result in a looser SiO2 structure and a drop in the density of the film. Therefore, the dielectric constant of SiO2 can be reduced; the drop depends on the concentration of
Table 2.2 Classification of low-k candidates.
fluorine. Experimentally, adding 10-15 at% F to SiO2 can reduce the k value by 20% to about 3.3.
Although fluorosilicate glass can be fabricated by many thin film deposition approaches, including TEOS-CVD, ECR-CVD, AP-CVD and LP-CVD, the most general approach is high-density plasma CVD (HDP-CVD). SiH4 and fluorides (such as SiF4, C2F6, CF4 or TEFS) as the gas source support simultaneous deposition and etch back, resulting in good gapfilling capacity without any void even for a width of 0.25 μm [28]. FSG has now been introduced into 0.25 and 0.18 μm IC production lines because of its chemical compatibility with IC processes. However, the thermal stability of fluorine is lower, potentially making the
Figure 2.2 Chemical structures of FSG with (a) low fluorine concentration, (b) high fluorine concentration.
dielectric unstable. The thermal treatment temperature should be under 450oC, when FSG films are used for IMD. Additionally, although the incorporation of fluorine makes the geometry of the Si-O network less polarizable, the concentration should be less than 10 at% to ensure the stability of the film, because the drop in film density provides more free space for the adsorption of moisture. This limitation in the fluorine concentration or in dielectric constant, causes the FSG films to fail at the 0.13 μm IC technology node and below.
2.2.2 Silsesquioxane (SSQ)-based Materials
Silsesquioxane-based materials comprise the inorganic Si-O bonding unit and organic substituents of low molecular weight, or hydrogen. Empirically, they have the chemical formula (R-SiO1.5)n where the substituent R may be hydrogen, alkyl, alkenyl, alkoxy or aryl.
Their name is derived from the sesqui-stoichiometry of oxygen atoms bound to a silicon atom, and they are therefore also called “T-resin” (tri-substituted). The most common structures are oligomers with a cage (T8 cube, eight silicon atoms on the vertices of the cube) or a ladder structure [29], as presented in Fig. 2.3.
Figure 2.3 Ladder (a) and cage (b-d) structures of Silsesquioxanes [29].
Organic substitution on Si causes the SSQ oligomer commonly to dissolve in an organic solvent as a precursor, before it is spin-coated on a substrate. These precursor solutions, developed by many famous suppliers of chemical materials, can yield SSQ low-k films with k<3.0, as indicated in Table 2.3. Following spin-on, the film has the characteristics of a liquid gel. It necessarily suffers baking and curing steps, which promote the formation of a
Table 2.3 Available SSQ-based low-k materials.
three-dimensional network siloxane (Si-O-Si) skeleton. During baking, the residual solvent is removed. Further curing results in polycondensation, which forms the mutually bonded silsesquioxane units [30]. Liu et al. [31], and Liou and Pretzer [32, 33] backed at 180-250oC for MSQ and 150-350oC for HSQ, following curing above 400oC; the as-cured films had dielectric constants of 2.9-3.2. Although the siloxane matrix is chemically similar to SiO2, more than 25% non-bridging Si-R bonds yield a lower density than that of SiO2. The Si-R bond has a relative lower polarizability than the Si-O bond in SiO2. The lower density and polarizability bonds of SSQ are attributed to the lower k values in the range 2.5-3.3.
The three common silsesquioxane compounds are recognized by the substituents R:
Hydrogen-silsesquioxane (HSQ, HSSQ), Methyl-silsesquioxane (MSQ, MSSQ), Phenyl-silsesquioxane (PSQ, PSSQ).
The material properties of these three silicate based materials vary greatly as indicated in Table 2.4. HSQ (H-SiO1.5) and MSQ (CH3-SiO1.5) are two low-k SSQ materials used commonly in microelectronic applications. HSQ has a k value in the range 3.0-3.2 but it involves many problems of integration. For instance, Liou and Pretzer [33] found that a network structure formed upon thermal processing at over 400oC due to the disassociation of Si-H bonds, which increases the difficulty of integration. MSQ has a dielectric constant of
Table 2.4 Material properties for SSQ-based matrix resins.
2.8, lower than that of HSQ, because the the Si-CH3 bonds are larger but less polarizable than the Si-H bonds. It is also more stable during the thermal process. The challenge raised by HSQ is degradation during O2 plasma resist stripping, such that an alternative stripping technique, such as H2 plasma treatment, is required [31].
Additional pores are introduced into the matrix resins of both HSQ and MSQ by adding surfactant or porogen, to greatly reduce the dielectric constants. Table 2.3 presents some of the results, and the section on porous materials will present details of the pore-forming methods.
2.2.3 Low-k Polymers
Since the 1980s, a large group of researchers have studied many newly synthesized or pre-existing polymers for their use as IMDs. For example, polyimide, a traditional material used in IC packaging, has a low k value from 2.3 to 2.9. Such polymers are varied but most are inappropriate for use in actual Cu metallization. Seven main considerations limit their applicability:
1. poor thermal stability (weight loss >1.0% at under 450oC) 2. anisotropic thermal, electrical and mechanical properties 3. softness
4. poor thermal conductivity 5. moisture uptake and outgassing
6. poor adherence to other contact layers, including the metal line and the diffusion barrier
7. incompatibility with traditional technological processes developed for SiO2-based dielectrics.
Figure 2.4 presents the chemical structures of some common polymer low-k materials.
Figure 2.4 Chemical structures of (a) polyarylethers; (b) FLARETM; (c) Teflon AF; (d) divinylsiloxane-benzocyclobutene (DVS-BCB); (e) SiLK.
Notably, these materials often have symmetric monomer structures, balancing the polarizability of the polymers. Additionally, organic low-k polymers are typically synthesized as non-polar or polar polymers. Non-polar polymers comprise primarily non-polar C-C bonds.
Therefore, they have dielectric constants that are independent of frequency, such that their dielectric losses are low. Unlike non-polar polymers, polar polymers have an asymmetric charge distribution, which arises from differences among the electronegativities of atoms.
They therefore have higher dielectric loss and a dielectric constant that depends on frequency and temperature. Saturated hydrocarbons have a lower polarizability and dielectric constant than unsaturated, conjugated and aromatic hydrocarbons (Table 2.1). However, they typically suffer from thermal degradation at temperatures of 300-400oC, or even much lower temperature. Only materials with aromatic, ladder, cross-linked structures can tolerate
temperatures of 450-500oC, which are required for IC interconnects. This situation represents a dilemma but the designer must find a compromise between the dielectric constant and the thermal stability when producing reliable polymer dielectric films. Empirically, most low-k polymer films with sufficient thermal stability have dielectric constants of 2.6-2.8.
In the last five years, low-k polymers have remained attractive to some researchers because of their properties are variable [34, 35]. These variations in the properties of polymers involve the partial rearrangement of monomers [36, 37]. Two polymers, PVDF [poly(vinylidene fluoride)] and PTFE [poly(tetrafluoroethene) or Teflon], have very similar chemical structures, but widely different properties. As presented in Fig. 2.5, the difference between these polymers that the two hydrogen atoms in PVDF are replaced by two fluorine atoms in PTFE. This slight difference between bonding atoms makes PVDF soluble in a polar solvent and PTFE is insoluble in organic solvents. Additionally, replacing the hydrogen with larger side groups can generate asymmetry in the polymer chain, reducing the tendency of the chains to slide past each other and increasing the tendency to form bonds between the chains, improving rigidity and increasing the melting point. Another case of interest involves copolymer, which replaces the side of the polymer chain with two or more polymer branches.
Such branched structures may reduce the packing density of linear polymer, affecting the mechanical properties and the thermally induced glass transition temperature, Tg. (More branching corresponds to lower crystallinity).
Figure 2.5 Chemical structure of PTFE and PVDF.
The glass transition temperature is some temperature, or narrow range of temperatures below which an amorphous polymer is in a glassy state, and above which it is rubbery.
Polymer is glass or rubbery, depending on whether its application temperature is above or below its Tg (>450-500oC for low-k films). The degree of cross-linking, which affects the thermal [38] and mechanical [39] properties of low-k polymer materials, also affects their glass transition temperature. Polymer chains may chemically link to each other via the reactive groups, to form a 3D network, in a process called cross-linking, which typically occurs during curing, improving stiffness, and ensuring that the behavior prior to plastic deformation under a load is sustainable. Heavily cross-linked polymers, or thermosetting polymers, typically have a high Tg and better mechanical yield strength. They are generally brittle [40]. Epoxies, polyurethane, parylenes, polyimide siloxanes, polyimides and FPI (fluorinated polyimides), benzocyclobutene (BCB) [41], SiLKTM [42, 43], FLARETM [fluorinated poly(arylethers)] [44], PAE [poly(arylethers)], HOSPTM (hybrid organosiloxane) [44] and perfluorocyclobutane are such polymers and are preferred for use in IMD applications.
Figure 2.4 presents the chemical structures of PAE, FLARETM, BCB and SiLKTM low-k materials. These organic polymers with a dielectric constant in the range 2.5-3.0 contain large amounts of benzene or fluorine, and, therefore, have low dielectric constants and good thermal stability. They share the problems of poor mechanical strength and poor adhesion.
Therefore, a capping oxide layer and an adhesion promoter must be added to each for integration. Table 2.5 also presents some of the polymers adopted in IMD applications.
Another focus of low-k polymer candidates is SiLKTM resin. SiLKTM is a spin-on hydrocarbon dielectric [5, 6, 42, 43, 45], which was developed by the Dow Chemical company. In 2000, IBM Microelectronics first used this dielectric material was first used in 0.13 μm BEOL. Figure 2.6 presents a four-level metal SRAM structure with a three-level
Table 2.5 Some of the polymers for low-k application.
SiLKTM dielectric [5]. Water is typically generated during curing because of dehydration in the condensation reaction in the synthesis of polymer from monomers. The SiLKTM resin is an extremely dense aromatic hydrocarbon (Fig. 2.4e) and releases no moisture as a byproduct during curing. Table 2.6 summarizes the main properties of SiLKTM dielectric. As indicated in the table, SiLKTM has a k value of 2.6 and Tg>490oC, resulting in thermal stability over 425oC.
Its intrinsic properties are so intriguing that UMC actively developed it as IMD [46]. However, UMC announced that it was ending its development of SiLKTM and in Feb. 2002, when it began to use FSG and CoralTM (carbon-doped oxide). After two years, IBM also considered abandoning SiLKTM because its CTE (which is a function of temperature) is too high at temperatures of over 150oC [47]. Although other refinements were subsequently announced, they did not satisfy the chipmakers. Nowadays, some Japanese companies like SONY and Fujitsu use a hybrid scheme as an IMD structure.
Figure 2.6 SEM cross-sectional image of Cu-SiLKTM interconnect structure [5].
Table 2.6 Intrinsic properties of SiLKTM low-k dielectric.
2.2.4 Amorphous Carbon
The use of fluorinated amorphous carbon (a-C:F) thin films as IMDs with low dielectric constants was recently reported [39, 48-59]. Sometimes, such materials are given such names as diamond-like carbon (DLC), hydrogenated DLC, fluorine-containing DLC (FDLC) [51, 52], plasma-polymerized fluorocarbon (PPFC) [53], fluorinated amorphous-carbon (a-C:F:H) [39] and amorphous carbon fluoride (a-CF) [54]. All these films have a wide range of k values from 2.0 to 3.6 and are typically formed by plasma-enhanced CVD (PECVD) using a discharge with fluoro-carbon gases (such as CF4, C2F6, C4F8 + CH4 or hydrogen mixtures).
The film comprises amorphous C-C cross-linked bonds (mixing of sp3 and sp2 bonding) and C-F bonds with a PTFE-like structure. Controlling the F/C ratio of an a-C:F film is important because it is the factor that most strongly affects the dielectric constant and the thermal stability. C-F bonds with a weak tendency to polarize in external electrical fields can reduce the dielectric constant of the a-C:F thin film, but excess C-F bonds suppress the formation of C-C cross-links, which would otherwise maintain the film’s thermal stability. Therefore, the properties of the a-C:F thin films can be adjusted easily by changing the plasma process conditions.
2.2.5 Carbon-doped Silica
Alkyl groups, like fluorine, can be incorporated into silicon dioxide to reduce its k value.
It is called carbon-doped oxide (CDO) [also known as organosilicate glass (OSG), or silicon oxicarbide (SiOCH)]. Carbon-doped oxide films can be deposited at ≥400oC by PECVD, typically using O2 and such precursors as methylsilane (1MS, H3SiCH3), dimethylsilane [2MS,
It is called carbon-doped oxide (CDO) [also known as organosilicate glass (OSG), or silicon oxicarbide (SiOCH)]. Carbon-doped oxide films can be deposited at ≥400oC by PECVD, typically using O2 and such precursors as methylsilane (1MS, H3SiCH3), dimethylsilane [2MS,