Dielectric Films
This invention relates to porous dielectric films having a dielectric constant (k) of less than about 2.5. Over the past years there has been a constant drive to produce dielectric materials having low dielectric constants for use, particularly, in semiconductor devices to accommodate the ever decreasing dimensions of the device architecture. It is presently believed that to achieve k values of less than about 2.5 for a practical insulator, there is inevitably a degree of porosity in these materials. This porosity can present major problems for integration, particularly when vias or interconnects are formed through the dielectric layer, because, when etched, the side walls of the etched formations are at least rough and possibly permeable, if the pores interconnect at all and intersect the surface.
It is into these etched trenches and vias that copper is deposited, in typical present day architectures, and because copper would readily diffuse into the dielectric material it must be contained by a diffusion barrier. An ideal would be an insulator that had barrier characteristics, but present day solutions rely upon separate deposited layers.
Traditionally these barrier layers were deposited using physical vapour deposition techniques, but these techniques struggle to provide sufficient conformity of barrier coverage and so chemical vapour deposition (CVD) techniques are used e.g. Metal Organic CVD, Metal HaJlde CVD and Atomic Layer CVD. Whilst CVD techniques can give near 100% conformity, the precursors and reactants can penetrate the porous dielectric. This effect is shown in Figure 1 (Source IMEC at the ARMM 2001 Conference) and Figure 2 (Source Passemard et ai; "integration issues of low k and ULK materials and damascene structure" at CREMSI 2001 Conference), in both cases, it will be seen that the side wail appears "fuzzy" and this indicates that CVD precursors
have been absorbed into the porous dielectric layer giving rise to an indistinct barrier between the dielectric and the barrier layer.
EP-A-1195801 describes processes which are in fact believed to increase the porosity of the side walls and proposes sealing pores created from the side walls by providing a protective or sealing layer. It suggests that such a sealing layer can be formed by a plasma comprising oxygen and nitrogen, but gives no substantive description of the process. The addition of extra material into high aspect ratio vias is undesirable, both because it increases the aspect ratio of the via and may increase the resistance of the copper in the via. It is not clear whether or not the sealed surface maintains the local low k values as suggested in the application.
From one aspect the present invention includes a porous dielectric film having a dielectric constant (k) of less than about 2.5 and a carbon content of not less than 10% including a via or other formation etched therein characterised in the exposed surface or surfaces of the film within the via or formation is substantially non-porous.
It will be understood that this approach is in complete contrast to EP-A- 1195801 where the processing of the dielectric increases the local porosity at the surface layer and this difficulty is only overcome by adding a further sealing layer. it should also be noted that sealing porous surfaces at the top and bottom of structures is considerably easier than sealing side walls that are parallel to the flux of the incoming reactants.
In a preferred embodiment the exposed surface or surfaces is formed by a layer which is carbon depleted with respect to the buik of the fiim. Additionally or alternatively the exposed surface or surfaces is formed by a layer which is of greater density than the bulk of the material film. In a particularly preferred
embodiment the exposed surface or surfaces, may be formed by a layer which is carbon depleted with respect to the bulk of the film. Additionally or alternatively the exposed surface, or surfaces may be formed by a layer constituted substantially by Si-Si bonds and these bonds may be formed between trivalent Si molecules. Other mechanisms which are currently not understood but occur at the exposed surface or surfaces.
In each of these cases it will be understood that the layer defining the exposed surface or surfaces is formed by modifying the etched dielectric material and not by further deposition. The bulk of the film is preferably formed of an SiCOH material.
The surface forming layer or layers may be formed by nitrogen and/or hydrogen containing plasma treatment of the etched surface or surfaces, which may be at least partially coincident with another process such as resist strip.
The invention further includes a barrier layer covering the exposed surface or surfaces, in which case the barrier layer does not penetrate the exposed surface or surfaces. The barrier layer is preferably deposited by chemical vapour deposition.
From a further aspect the invention consists in a method of forming an interconnect layer in a semiconductor device including: a. depositing a low k porous dielectric film on a substrate; b. depositing a resist; c. patterning the resist to define etch apertures; d. etching vias or formations in the dielectric layer through the apertures; and e. stripping the resist characterised in that the resist is stripped with a nitrogen or a noble gas, or combination thereof, and hydrogen plasma or nitrogen or a noble gas, or a combination thereof, and
oxygen plasma and the exposed surfaces of the vias or formations are simultaneously exposed to the plasma causing densification of the surface layers which define the exposed surfaces.
The method may also include the deposition of a barrier layer on the densified exposed surfaces. This barrier layer may be deposited by chemical vapour deposition.
Preferably the nitrogen or noble gas dominates the hydrogen or oxygen. Thus it is preferred that the ratio of N2:H2 is about 5:1.
The substrate may be RF bias during the stripping of the photo resist. In alternative approaches the densification may take place during the etch step and it may be created by means of a non-oxidising plasr ia process, e.g. when the low k material is organic in nature.
Although the invention has been defined above, it is to be understood it includes any inventive combination of the features set out above or in the following description.
The invention may be performed in various ways and a specific embodiment will now be described, by way of example, with reference to the following drawings, in which:
Figure 3 is a transmission electron micrograph (TEM) of a cross section of a dielectric material formed with trenches to illustrate features of the invention;
Figure 4 is an enlarged view of part of Figure 3; and Figures 5a - d are electron energy loss spectroscopy (EELS) graphically represented results taken across a line A as indicated in Figure 4.
Figure 6a is a bright field TEM of a cross section of a dielectric of the invention, barrier and copper fiiied trench.
Figure 6b and 6c are bright field TEMS iiiustrating precursor diffusion of the prior art.
Figure 7 shows a bright field TEM of a 0.18 micron structure illustrating features of the invention (equivalent to figure 3, but smaller structure).
Figure 8 is an EELS scan for titanium of the structure of Figure 7 for both a) nitrogen and oxygen resist strip/treatment and b) nitrogen and hydrogen resist strip/treatment.
Figure 9 is an EELS scan for carbon of the structure of Figure 7 for both a) nitrogen and oxygen resist strip/treatment and b) nitrogen and hydrogen resist strip/treatment.
Figure 10 shows bright field TEM of a cross section of a dielectric material formed with trenches to illustrate features of the invention where a) is an overview and b), c) and d) are at higher magnification.
Figure 11 shows leakage current through structures made with the invention in 0.18 micron trenches.
Figure 12 shows leakage current through structures made with the invention in 0.25 micron trenches.
Figure 13 shows RC product through structures made with the invention. Figure 14 shows comparisons between MOCVD and sputtered (PVD) barriers upon dielectrics of the invention.
Turning to Figure 3 a test stack is illustrated to reveal a structure common in forming damascene interconnect architecture. The stack 10 is upon a base dielectric iayer 11 and consists of an etch stop layer 12, a dielectric layer 13, which has trenches 14 etched therein, a silicon carbide cap 15 on the upper surface of the dielectric layer 13 and a barrier layer 16. A silicon oxide layer 17 and a planarisation layer 18 have been added purely for the purposes of TEM sample preparation.
The dielectric layers 11 and 13 are constituted by a low k SiCOH material having more than 10% carbon, which is trade marked Orion by the Applicants.
This material is porous and has a dielectric constant k of about 2.2. The planarisation layer 18 is a material which is trade marked Flowfill by the Applicants.
The dielectric layers 11 and 13 were deposited using Trikon /xP ™ tool for example as described in WO-A-01/01472, the disclosure of which is incorporated herein by reference. This material is a cold deposition of a polymer, which is then hydrogen plasma cured.
The trenches 14 were etched, without a hard mask, in a Trikon MORI ™ helicon source plasma etch tool using CF4 CH2F2 chemistry with RF wafer bias. Following the plasma etch the photo resist, which defines the etched apertures for the trenches, was stripped in situ in the MORI ™ tool (i.e. in the same chamber) using 5:1 N2:H2 chemistry again with a helicon wave mode plasma source and applied RF wafer bias. This resist stripping also removes polymer residues. As is well known in the art there may be further wet or dry processing steps to be completed between the resist strip and the subsequent MOCVD barrier deposition of the barrier layer 16, though in this case none were used.
As these processes are known to a person skilled in the art they are not detailed here.
MOCVD titanium nitride TiN (Si) was deposited in a stand alone system using TDEAT (Tetra Dϊethyl Amino Titanium) and ammonia precursors together with helium bailast. immediately after deposition, the MOCVD film was hydrogen plasma treated and then silane soaked. There were no thermal or plasma treatments prior to deposition.
It will be seen, immediately, from Figure 3 and even more clearly from Figure 4 that the interface between the barrier 16 and the trench side wall is smooth and continuous and this is in complete contrast to the prior art arrangements shown in Figures 1 and 2. Further the barrier layer itself is
smooth and continuous. The micrograph further reveals that the trench side walls adjacent to the barrier layer are less porous (denser) than the regions further away from the side walls. The denser regions are darker in the bright field TEM imagining and are marked J and K in the micrograph. As the deposited film was laterally homogeneous, then densification of the etched side walls has taken place during the trench formation (plasma etch and/or subsequent resist strip). It is believed that at least the majority of the densification has taken place during the resist strip largely because during the etching of the formations there are large amounts of polymer present on the side walls (to effect anoisotrophic etching) that is removed by the subsequent strip processes.
Further evidence for side wall densification comes from Figures 5a - c. The electron energy loss spectroscopy analysis can be used to provide information about the overall thickness and composition of the sample and also the distribution of individual elements. Spatial maps can be generated in a series of such one dimensional maps were taken in the axis identified by the line A of Figure 4 and the results are shown in Figures 5a - b.
Comparisons of the maps support the existence of a densified trench side wall. The signal plotted in 5a varies with sample thickness and sample composition/density. The concave nature of the signal from the layer 13 between the twin peaks of the barrier layer 16 show that the dielectric 13 is denser or thicker near the side walls. It is not believed that the variation is due to thickness. Figure 5b shows the titanium signal and confirms that the barrier layer 16 is tightly confined, there being no detectable titanium signal from within the film 13. Figure 5c shows that the trench side waiis are depleted of carbon whereas the oxygen profile in 5d is comparatively flat.
It is therefore concluded that the trench etch and/or resist strip process is densifying the trench side walls of the porous low-k layer, thereby providing a smooth surface to prevent penetration of barrier layer precursors or reactants. This enables the deposition of a continuous barrier thereby preventing copper penetration. Whilst the experiments have only so far been performed on the
Applicants' material it is believed that the same results would be obtained with at least some other ultra low k porous dielectrics particularly those of the SiCOH family, being hydrogenated carbon containing silicon dioxides that are porous. The carbon and hydrogen is in such film typically as C-H3 groups with C-Si bonds effectively tying in large amount of hydrogen and this hydrogen is considered the main cause of the low-k value for the matrix of the film together with the resultant porosity.
The precise mechanism for the densification is not yet known, but it is believed likely that the depletion of carbon from the densified layers enables the forming of Si-Si bonds between trivalent silicon atoms.
The reactive ion etch process of the BARC and porous low-k SiCOH material with a photo resist mask on 200mm wafers was:
Process gasses CF4 , and CH2F2 in the ratio 4.4:1 to 6.6:1
Pressure 1.5-2 millitorr Plasma power 1 ,25 KW to an inductive antenna
Wafer bias power 400 watts
Platen temperature -15°C
The reactive ion photo resist strip process on 200mm wafers, carried out in the same chamber was: Process gasses N2 and H2 in the ratio 5:1
Pressure 5 millitorr
Plasma power 2.5 KW to an inductive antenna
Wafer bias power 200 watts
Platen temperature 0°C
The etch process was with electrostatic wafer clamping and helium back side pressurisation and wafer temperature will therefore be close to the platen temperature. A low temperature is used to retain resist integrity.
For the resist strip process the wafer was undamped to allow higher wafer temperatures, thereby improving residue removal efficiency and increasing strip rate. Peak wafer temperature was indicated as 121°C (be means of industry standard thermal stickers) at 0°C platen temperature and 104°C at -15°C platen temperature.
These experiments have used nitrogen and hydrogen, however if nitrogen is not chemically active in the densification process then alternatives may be substituted such as helium, neon, argon, xenon and krypton or any other suitable sputter etch gas. Alternatively they may be added to the nitrogen and/or hydrogen gas mix.
Further work has been performed to illustrate the effectiveness of the invention. At Figure 6(a) is shown a bright field TEM image of a completed structure consisting of dielectric of the invention with an MOCVD deposited titanium nitride barrier and completed copper trench fill consisting of a sputtered copper seed layer, electroplated copper and chemical mechanical polishing step. As can be seen there has been no diffusion of metal from either the barrier or copper into the dielectric.
Further Figure 6(a) shows an amorphous layer of 5 to 8 nanometers thickness that is modified by the plasma treatment and is of higher density compared to the bulk of the porous dielectric.
In contrast at Figures 6(b) and 6(c) are shown precursor diffusion. The image at 6(b) is from W. Besling, Proc. IITC 2002 Burlingame (CA) USA, 2002
pp 288-291. The image at 6(c) is from S. Kawamura et. al. Proc IITC 2001 San Francisco, USA, pp 195-197. As can be seen bright field TEM imagery is a well known and acceptable indication of metal diffusion into a dielectric material.
Figure 7 is TEM image of a 0.18 micron structure formed as previously described in relation to Figure 3. No metal diffusion for the barrier can be seen and this is further evidenced at Figure 8(a) and 8(b) that are EELS line scans for titanium of the structure shown at Figure 7.
Figure 8(a) illustrates the EELS scan for a nitrogen and hydrogen gas mix. At Figure 8(b) a gas mix of 200sccm of nitrogen and 10 seem of oxygen was used (a ratio of 40:1 is the best that has been established at this time ).
Oxygen is well known to remove carbon and this experiment illustrates that nitrogen can reduce the carbon removal effect of the oxygen and allow porous dielectrics to withstand a high degree the absorption of a gaseous metal precursor (though not as good as nitrogen + hydrogen). This process is in contrast to that described in EP-A-1195801 in which a nitrogen/oxygen plasma is used to form a sealing layer.
Figure 9(a) and 9(b) show EELS line scans through the structure of Figure 7 for carbon. In Figure 9(a), the EELS scan for nitrogen and oxygen shows there is greater carbon loss at the dielectric side walls than is the case for nitrogen and hydrogen as illustrated at Figure 9(b).
Figure 10 is a further illustration of an embodiment of the invention in bright field TEM images. Figure 10(a) is an overview of a structure formed as described above at Figure 3. Figure 10(d) is the result of the nitrogen and hydrogen process described in detail earlier and Figure 10(b) and 10(c) are images illustrating a nitrogen and oxygen gas mix treatment.
Figures 11 to 14 show results from electrical test structures formed with dielectric which are embodiments of the invention. The test structures were
single damascene of line width/line spacing of 0.18 and 0.25 micron trenches/spacers. The interdigital comb was 100 microns by 1600 microns in size with a 44cm perimeter. Inter-line leakage was measured at 0.5 MV/cm and interline capacitance was measured at 1MHz.
The plasma treatment/resist strip processes were as follows: Nitrogen + Hydrogen (5:1 ratio)
Nitrogen 200 seem (std. cubic centilitres per minute)
Hydrogen 40 seem
Pressure 7 millitorr
Platen temperature -15°C Electrostatic chuck with 2 Torr helium back pressure
MORI™ plasma source (inductively coupled) Plasma power 2.5Kw to inductive antenna 13.56MHz Magnet power 40/60Amps inner/outer coils
Platen power (wafer bias) 200 W 13.56MHz
Nitrogen + Oxygen (20:1 ratio)
Nitrogen 200 seem
Oxygen 10 seem
Pressure 7 millitorr
Platen temperature -15°C Electrostatic chuck with 2 Torr helium back pressure
MORI™ plasma source (inductively coupled)
Plasma power 2.5Kw to inductive antenna 13.56MHz
Magnet power 60/60Amps inner/outer coils
Platen power (wafer bias) 30 W 13.56MHz
This was the best nitrogen + oxygen process for densification and is in contrast to the sealing process of EP-A-1195801.
Note that in these subsequent experiments, the wafers were clamped electrostatically thereby lowering their temperature to close to the platen temperature. It was found that the processes were still effective at these lower wafer temperatures.
Figure 11 shows the results for leakage current (less is better) on 0.18 micron trenches for both nitrogen + hydrogen and nitrogen + oxygen gas mixtures. As can be seen the oxygen degrades the performance compared to the hydrogen. This is to be expected given the EELS result of Figures 9(a) and
9(b) showing increased carbon loss for the nitrogen + oxygen process. Further, a wet clean does not degrade the nitrogen + hydrogen treated dielectric, but does degrade slightly the nitrogen + oxygen treated dielectric, further indicating a degree of porosity. Such wet cleans are widely known and used in the industry to remove any residues after a dry resist strip process.
Figure 12 is a further illustration of the comparative effects of nitrogen and hydrogen or oxygen as for Figure 11 but on 0.25 micron structures. The results and conclusions are the same as for Figure 11.
Figure 13 shoes the RC product (less is better) from the test structures. As can be seen an industry standard wet clean does degrade the RC product slightly for both nitrogen + hydrogen and nitrogen + oxygen processes, with again better results for the nitrogen + hydrogen processes.
Figure 14 shows comparative results from 0.18 and 0.25 micron test electrical test structures with the barrier deposited by MOCVD and PVD (sputtering) means. The comparison shows that the leakage currents are low, and similar.