Characterize Low-k and Copper Films In Situ

		At a Glance
	A new integrated metrology tool for in situ material characterization of low-k and copper film is based on a high-temperature chamber, which can simulate the high vacuum or inert gas environment of low-k and copper processes. The end result is the ability to overcome sample-to-sample and tool-to-tool variations that are inevitable in new material characterization.

With the proliferation of several types and classes of low-k materials in recent months, the screening, evaluation and integration of these new materials to next-generation devices have become daunting tasks for end users and suppliers alike. The added pressure to shorten design cycles to less than 18 months means process integration involving new materials requires more efficient methodology and toolsets than are currently available.

Evaluation of thermal-mechanical properties of new thin-film materials, like low-k dielectrics and copper, requires the use of several complementary metrology tool sets. For example, for initial screening, information on stress hysteresis, thermal stability, outgassing, film shrinkage, thermal expansion coefficient, adhesion, and other material and electrical property changes after heat cycling needs to be determined. Following curing, cleaning, etching, stripping and CMP steps, outgassing, adhesion and electrical properties need to be remeasured. For materials where the information base of such properties is sketchy or limited, conclusions may be ambiguous, making data correlation time-consuming. With new materials in particular, samples often are limited in quantity, and process conditions may not be repeatable. To overcome these sample-to-sample and tool-to-tool variations, an in situ metrology tool, the FSM900TC-VAC, has been developed that simultaneously measures several of these physical, optical, chemical and electrical properties changes during a heat cycle. Evaluation and screening of these new materials would improve dramatically with this tool, resulting in shorter process integration time and time to market for the products. Key to this system is in situ monitoring using thermal desorption spectroscopy (TDS) in a sealed environmental/vacuum chamber.

System features

Film shrinkage customarily is measured using a spectrometer or ellipsometer, measuring film thickness at room temperature, before and after a thermal process. While this approach is adequate for evaluating total change, it cannot determine the rate of shrinkage at the various temperatures, nor provide information about abrupt changes in thickness at certain temperatures. By incorporating a spectrometer that monitors film thickness changes in situ, for example, through an optical window on top of the chamber, useful information can be extracted at essentially the same rate as material loss. Then, correlating data obtained from stress hysteresis and TDS allows fast verification of distinct thermal load limitations on the materials under study.

The instrumentation consists of a sealed annealing chamber based on a rapid thermal processing (RTP) type chamber design. The system can operate either under high vacuum (up to 1x10^-6 Torr range) or in a controlled inert gas environment to closely simulate processes. By incorporating multiple metrology probes either inside the chamber or through optical windows, information like film stress hysteresis, thermal stability, TDS, film shrinkage, reflectivity, resistivity and CV changes may be monitored simultaneously in situ.

4. Film shrinkage monitoring of the low-k sample confirms that shrinkage occurred at 350°C.

Low-k and copper thin films pose a characterization problem for metrology tools, such as traditional high-temperature stress tools. This is because these oxidation-sensitive materials need to be processed or annealed in an atmosphere with very low partial pressure oxygen, in many cases <10 ppm. This low oxygen criterion is impossible to achieve with traditional high-temperature stress tools because of their relatively open (or unsealed) heat chuck design. In contrast, the FSM900 chamber is designed to operate in a high vacuum environment or in a controlled inert gas environment, where <10 ppm oxygen can be realized easily for simulating actual thermal processes.

5. In situ film reflectivity of the CVD low-k sample is shown at various wavelengths.

Most outgassing studies on the new low-k materials currently are done on small, broken pieces of wafers on a tool, like TDS. The results may raise correlation issues, as they may not reflect whole wafer desorption under actual thermal processes. The new integrated tool works with whole wafers. With a highly reflective, cold-walled chamber, radiant heat from the lamps is focused directly and predominantly on the sample. That means background desorption from the chamber wall is minimized during outgassing measurements. Full wafers, up to 300 mm, can be characterized.

Experimental Setup

6. The first thermal stress cycle for the plated copper film was taken over a temperature range exceeding 400°C, in a vacuum.

Two types of 200 mm samples were evaluated: a CVD low-k (carbon-doped oxide) wafer and four copper wafers. The CVD low-k sample already was cured, with nominal thickness of 7000 Å. Copper wafers include electroplated copper (samples C, W1 and W2, with nominal thickness of 9900 Å) and a PVD-deposited seed layer of copper on TaN (sample D). The experiment was performed using the FSM900, with chamber base vacuum ~1x10^-5 Torr. Metrology probes included in situ film stress, TDS, film shrinkage, reflectivity and resistivity. Temperature ramp rate was set at 10ºC/min. Film stress and thickness and reflectivity were observed optically, through an optical window on top of the chamber. The former used laser scanning to track wafer curvature changes, while film thickness and reflectivity changes were monitored with a spectrometer (400 to 850 nm). TDS was measured with a quadropole/mass spectrometer, and sheet resistivity was determined with an in-chamber, four-point probe.

Low-k results

7. During the second thermal cycle, starting and ending stress values are very similar for a plated copper film.

The CVD low-k sample was heat-cycled in a vacuum to temperatures up to 500ºC, with ramp rate of 10ºC/min. The results taken simultaneously are shown in Figures 1 through 5. From the stress hysteresis curves (Fig. 1 and Fig. 2), film stress is increasingly compressive and elastic up to 390ºC. Around 390ºC, the film becomes less compressive, signifying a phase change or the onset of the evolution of volatile compounds. This observation is confirmed by the TDS and film shrinkage data. TDS data showed outgassing tends to increase with an increase in temperature, even at a relatively low temperature. However, the rate of desorption appears to increase, notably above 350ºC. Figure 3 indicates these gases include hydrogen- and carbon-based compounds like methane (AMU 16 and 17) and carbon dioxide (AMU 44). In situ film shrinkage monitoring also confirmed that shrinkage occurred at 350ºC, during the heat-up cycle (Fig. 4). The increase in thickness during the heat-up cycle up to 350ºC is consistent with normal thermal expansion of the film. From the experiment, the film shrunk by about 150 Å by the end of the first thermal cycle. Reflectivity measurements (Fig. 5) were monitored at 550 and 650 nm. The data show a slight shift in reflectivity during the heat-up/cool-down cycle. The reflectivity data collected at 650 nm had a slightly better sensitivity.

8. Distinct changes in reflectivity in the plated copper film correspond to stress hysteresis data.

Figure 2 showed the stress hysteresis data of both the first and second thermal cycle. While the stress hysteresis of the film during the second thermal cycle showed much greater thermal stability, to evaluate the suitabilty and stability of this low-k material it may be interesting to remeasure its relative k value to see if the dielectric constant shifts after multiple thermal cycles.

Copper samples

In a separate experiment, two electroplated copper samples — samples W1 and W2 from the same batch as sample C — were measured for resistivity changes as a function of temperature.

9. Resistivity changes in the copper film correspond to the plastic flow phase portion of the stress curve, 220-250ºC.

From the stress hysteresis experiments (Figs. 6 and 7), the effect of a heat treatment on annealing copper are apparent. After the first thermal cycle, the starting and ending stress values for the copper film are very similar. However, to better understand the properties and the changes within the copper film, especially in the heat anneal process (first thermal cycle), in situ metrology was used to correlate the stress hysteresis (wafer curvature change) with the reflectivity, resistivity and TDS data. It was noted that there is a distinct reflectivity (Fig. 8) and resistivity change (Fig. 9), corresponding to stress hysteresis data, notably in the plastic flow phase portion of the stress curve (around 220-250ºC). The reason for the increased resistivity and reflectivity that occurred around the 220-250ºC region following the first heat cycle is uncertain. Since this occurred around the plastic flow zone of the copper film, it may be related to grain growth changes or other causes. However, from the second thermal cycle profile, it is apparent that the resistivity as a function of temperature is relatively linear and that the resistivity of the copper film is lower than the values from the first thermal cycle. This is consistent with the grain growth theory of electroplated copper where heat-annealed, or time-dependent, room-temperature self-annealed copper has much larger grain sizes than freshly electroplated copper. This also is consistent with the observation that the resistivity of an annealed copper film has decreased and its reflectivity has increased.

10. The TDS data of the plated copper show increased outgassing from AMU 44 and AMU 45, supporting results in Fig. 7 .

The TDS data showed some increase in outgassing for AMU 44 and AMU 45 (Fig. 10). These may possibly be carbon dioxide and isotopes of carbon. These may be byproducts from additives from the plating solution. The peaks for these outgassing species (AMU 44 and 45) occur around 120ºC and 230ºC, which showed good correlation to the unusual stress peaks at similar temperatures where the stress is changing from compressive to tensile in that portion of the stress temperature profile.

A subsequent experiment with sample D, PVD copper seed layer, shed a bit more light on the role of outgassing species in copper stress hysteresis (Fig. 11).

11. Copper stress hysteresis for the PVD copper seed layer was taken continuously to 400ºC.

In this experiment, it was noted that the AMU 44 showed a gas peak around 150ºC and 300ºC (Fig. 12), which also corresponded to the stress hysteresis peaks at the same temperatures. It is likely that outgassing had an influence on the stress hysteresis curves. Whether this has any influence on the reflectivity and resistivity changes of the copper, and whether this outgassing contributes to the reported premature failures of barrier layers, need further investigation. Both electroplated copper and PVD copper showed carbon desorption, which could be an indication of selective CO₂ adsorption onto the surface of Cu films. However, additional study also is needed.

12. TDS data show peaks near 150ºC and 300ºC, corresponding to the stress hysteresis peaks in Fig. 11 .

The impact of in situ integrated metrology has been demonstrated in terms of speed, ease and efficiency in evaluating data. By using a single sample within the same chamber, simulating a process thermal cycle, the sample-to-sample variations and tool-to-tool variations commonly seen in new material characterization can be overcome. The potential for shortened process integration time and compressed time to market can be realized. •