Abstract

The restless race of semiconductor industry toward new electronics devices with higher performances has been possible thanks to the continuous shrinking of the elementary constituents of these devices (transistors, memory cell, etc.). Dry etching is a key process step to realize semiconductor devices. This paper will give at first an overview of the main technology requirement of dielectric etching, then the basic mechanisms of dielectric etching will be briefly explained. After that, the historical trends of gases used in dielectric etching will be illustrated. The path starting from simpler fluorocarbon gases (CF₄, CHF₃) toward more and more complex ones (C₄F₈, C₄vF₆) will be explained through the basic etching mechanism and linked to the fundamental technology requirement.

Technology Challenges in Dielectric Etching

Dry etching of dielectric materials is a key process step to realize semiconductor devices, and a continuous technology evolution has been necessary to fulfill the more and more stringent requirements for dielectric etching. The technology evolution has been focused toward two parallel and complementary directions: new plasma source developments and selection of new etching gases and additives. This paper will give an overview of the evolution of the gases used for dielectric etching in the last decade, trying to highlight the technical reasons that pushed the microelectronic industry to use the new gases and try to link all the changes in a unique path.

In this paragraph we will review the most critical aspects that challenge dielectric etching technology moving from one device generation to the following. The first aspect, common to all dry etching technologies, is the ability of photoresist to “resist” to plasma etching. While having to deal with smaller dimensions, lithographic technology needed to reduce photoresist thickness. For example, a typical photoresist thickness for the 130nm technology node is 400nm, while for the 65nm node it is 180nm. Moreover, the new generation of photoresists developed for the new lithography technologies (form i-line to 248nm and 193nm Deep UV) have intrinsically lower selectivity to etching. As a global result, in the last years the selectivity of dielectric etching versus photoresist had to dramatically increase.

Unfortunately, high selectivity is being required not only with respect to photoresist, but also with respect to Si₃N₄.The most critical structure for this aspect is called SAC (self-aligned contact), even if other structures, such as borderless contact or dual damascene requires high selectivity with respect to Si₃N₄. In the SAC process, contacts holes are etched between the gate electrodes that are protected by a Si₃N₄ layer that acts as an etch stopper. It improves the lithographic process window by permitting the use of larger contact hole size or lower overlay accuracy, because the position and size of the contact should be defined by the space between gate electrodes with the etch stopping layer (see Figure 1). Values of selectivity around 20:1 are required; it is very difficult to be able to obtain this value of selectivity without having etch stop phenomena; that is, the etching naturally “stops” and it is not able to continue the definition of the structure even if the etching time is increased.

Last but not least, the shrinking of the devices has also greatly increased the aspect ratio (AR: the ratio between the dimension to be defined versus the thickness of the stack to etch) of the structures to be defined. As an example, considering contact etching in Flash memory technology, we have an AR of 5.5:1 for 130nm node and an AR of 7:1 for 65nm node. The critical point is in the fact that the increase of ARs decreases the process window with respect to etch stop.

Basics of Dielectric Etching and Selectivity Mechanism

It is now appropriate to briefly describe the basic mechanisms of selective SiO₂ etching, so that it will be possible to build a link between these basic mechanisms and chemical properties of the etching gases. When plasmas containing fluorocarbon gases are used to etch SiO₂, three phenomena occur simultaneously: the deposition of a fluorocarbon layer, the etching of this fluorocarbon layer and the etching of the substrate.[1] At low ion bombardment energy, these processes produce a net fluorocarbon deposition.

Increasing the ion bombardment energy above a given threshold energy value, it is possible to switch to etching of the earlier deposited fluorocarbon layer; etching also occurs for the substrate materials. It is known that SiO₂ etch rates obtained etching blanket samples depending on bias power.[1] It can be noticed that for highly energetic ion bombardment SiO₂ etch rate scales with the square root of the ion energy indicate that we are in presence of a chemical sputtering mechanism. At low bias power, such as 150W, there is a deviation of etch rate from the chemical sputtering curve. The reaction enters a different regime called “suppression regime.”[1] The reason for this deviation is that the balance between fluorocarbon deposition, fluorocarbon etching and substrate etching is such that a relatively thick steady-state fluorocarbon film forms on the SiO₂ surface; this film protects the underlying SiO₂ for direct ion impact and thus chemical sputtering. This is a key point in explaining the mechanism of selectivity: in fact, similar to the SiO₂ case, there is a fluorocarbon film present during plasma etching also on Si₃N₄, photoresist and Si. A good selectivity between SiO₂ and Si₃N₄ (or Si or photoresist) is achieved when, for fixed process condition, SiO₂ is in the sputtering regime and the other layer is in the “suppression regime” or even in the “deposition regime.”

Selectivity and Fluorocarbon Gas Decomposition

At this point, a question naturally arises: What is the link between the fluorocarbon film and the fluorocarbon gases used in the plasma reactor to etch dielectric films? When plasmas containing fluorocarbon gases are ignited, the gases are (usually partially) decomposed by electron impact dissociation. Let’s consider, for example, C₄F₈: It can be decomposed in[2] F, CF, CF₂, CF₃, C2F4, C3F5, C2F5, C3F5. The ratio between the different fragments depends on the reactor design and on process condition (pressure, power, flow, etc.). From that, it is clear how difficult is to give a complete, detailed description of the reaction happening during dielectric etching, but a simplified vision can be successfully given.

Ion beam studies[2-4], in fact, allowed to build up the etch yield response for ion energy variation of several charged species. Again, the presence of a deposition, a suppression and a chemical sputtering regime can be observed. Etch yield response depends both on fragment composition and substrate. It has been demonstrated[3] (see Figure 2) that CF₃ + ions, showing the highest etch yield, have a threshold energy comparable to SiN and SiO₂, therefore realizing low selectivities, while CF+, having the highest difference in threshold energies, enters the suppression regime very easily for oxide, therefore having a poor window against etch stop. CF₂ + has been revealed as the one able to provide both high etch rate on oxide and high selectivity to nitride.[2,4]

Another way to interpret these results is obtained considering the C/F ratio of the different fragments. CF₃ + has the lowest C/F ratio (1/2) and the lowest selectivity; CF+ has the highest ratio (1) and the highest selectivity but a poor process window respect to etch stop; and CF₂ + has an intermediate C/F ratio (1/2) but is the best compromise between selectivity and etch stop.

Historical Trends Toward Better Selectivities

In the beginning phase of the dielectric etch technology development, the main gas used to etch oxide was CF₄, providing a C/F ratio of 1/4. As far as higher selectivities required to support technological shrink, CF₄ etching gas was coupled with another gas able to provide “selective” fragments, such as CF₂ + or CF+. The general mixture formula was CF₄/CHxF₄-x, with the coupled gas varying along the sequence

CHF3 >> CH2F2 >> CH3F (C/F = 1/3< 1/2<1).

Moving from left to the right in the sequence it is possible to achieve higher selectivity. Historically, those most often used have been CHF3 and CH2F2, as it is easier to manage with respect to the suppression regime entrance (etch stop).

Once the SAC structure was introduced, requests for selectivity became more aggressive (a selectivity SiO₂/Si₃N₄ of 20:1 is typical for this application); it was not possible to match this requirement by the mixture CF₄/CHxF₄-x without seriously losing process window amplitude against etch stop. High selectivity performances were then provided by the introduction of a fluorocarbon with a longer carbon chain: c-C₄F₈. The main reasons for its selectivity were both the low release of CF₃ fragments (due to its cyclic structure) and the high release of fragments containing C=C bonds, providing better selectivity than simple CF₂. It was proven, in fact, that at the same C/F ratio, longer chains behave better than shorter ones.[2]

Following the direction of minimizing precursors for the CF₃ + fragment and supplying longer-chained fragments, gases evolution moved toward unsaturated fluorocarbons, showing their selectivities increased with the number of unsaturated C=C bonds in the chain.[5] The evolutive sequence for these gases is

C₄F₈ >> C₅F₈ >> C₄F₆ (C/F = 1/2< 5/8< 2/3)

Again, the direction is to increase C/F ratio.

Figure 3 shows the fragmentation patterns for the gases in the last sequence, highlighting how the performances in terms of selectivity are correlated with the availability of fragments with long C-chains in the plasma. It is clear from mass spectra that selectivities achieved by C₅F₈ and C₄F₆, which are higher than the one provided by C₄F₈, do correlate with the presence in the fragmentation pattern of C_xF_y with x>3, whose presence is not revealed by the C₄F₈ spectrum.[ 8]

Furthermore, this gas family has been the key point in allowing the definition of high aspect ratio contacts, whose patterning started becoming very challenging after the 130nm node. In fact, the effect of these longer chains in the plasma is more evident when evaluated on that kind of critical structure. The model is that longerchained fragments are able to stick with resist on top, therefore protecting it from chemical sputtering, while smaller chains can enter the small structure. Theoretical calculations[6] estimate that sticking coefficient varies in the following way:

C_xF_y >> CF₃ >> CF >> CF₂

From a fragment to the following one, the change is one order of magnitude.

The fragments with lower sticking probability will deeply penetrate into the contact, while the one with the highest will stick on top (resist mask), to provide protection. Accordingly, generic C_xF_y will build up the film protecting the resist mask from chemical sputtering, while CF₂ + will be the one with the highest density at contact bottom, providing, as already observed, good selectivity performance to the nitride stop layer, keeping a good amplitude for process window against etch stop.

This explains why the availability of CF₂ + fragments and long-chained fragments is the driving force moving evolutive fluorocarbon gases toward cyclic and unsaturated structures. C₄F₆ is actually the most advanced implemented gas for dielectrics etch application, while the next step in the sequence is already under evaluation. Following the direction of an increase of unsaturated C=C bonds, the best candidate is the cyclic and aromatic C₆F₆, whose high performances in terms of selectivity has already been proven.[7]

Conclusions

An overview of gases evolution for dielectric etch application has been given, explaining why gases evolved that way along the semiconductor roadmap. In particular, the behavior of different gas families has been correlated with their molecular structures, according to the actual available conclusions of fundamental studies on dielectric etch dynamics.

In the most advanced fluorocarbon gases, the importance of having long chained structures with unsaturated C=C bonds have been shown and pointed out as the driving force for future evolutions.

References

1. M. Schaepkens and G.S. Oehrlein, J. Electrochem. Soc., 148, C221 (2001)

2. M. Sekine, Appl. Surf. Sic., 192, 270 (2002)

3. K. Yanai, K. Ishikawa, K. Karahashi, H. Tsuboi, K. Kurihara, and M. Nakamura, Proceedings of the 24th Dry Process International Symposium, The Institute of Electrical Engineers of Japan, Tokyo, 2002, p. 269

4. K. Yanai et al J. Appl. Phys, 97, 053302 (2005)

5. H. Motomura et al Thin Solid Films 390, 134-138 (2001)

6. Miyata et al J. Appl. Phys., 36 5340-5345 (1997)

7. Chatterjee et al J. Electrochem. Soc., 148, G721-G724 (2001)

8. Karwacki Jr. et al Solid State technology, November 2005