Advances in Electronics Manufacturing Technology www.vertilog.com
V-EMT 1:27 (6 December 2004)
Enhanced copper bath monitoring using electrochemical impedance spectroscopy Abstract A. Zdunek Air Liquide Chicago Research Center D.N. Sanz Air Liquide Centre Recherche Claude Delorme C. Gabrielli, P. Mocoteguy, H. Perrot CNRS, Laboratoire Interfaces et Systemes Electrochimiques C. Mace, A. Quennoy, E. Chabal Altis Semiconductor
Methods are required to monitor the composition and ageing of the chemicals used in copper electroplating baths in order to better guarantee the quality of the deposited copper. We have used electrochemical impedance spectroscopy measurements on copper interconnect deposition baths in a wafer fab to characterize changes in the condition of those baths. The method is able to clearly distinguish between tools using different plating chemistries. Results from experiments to monitor multiple process tools indicate the shape of the impedance scan is sensitive to changes in the steady state of the bath chemistry and may be an indicator of overall bath quality.
Introduction The ability to perform superconformal deposition of copper in vias and trenches depends strongly on organic additives to the copper sulfate/sulfuric acidbased electroplating bath. Monitoring or controlling the concentration of these additives as well as their byproducts is therefore crucial to guaranteeing the properties of the copper deposit. Also, changes in the overall bath quality, which is dependent on manufacturing conditions, additive degradation, and byproduct accumulation, can profoundly affect the copper deposit, are not presently controlled. State-ofthe art copper interconnect plating tools use online monitoring to quantify bath component concentrations to achieve tighter bath control. At present, bath quality is maintained through bleeding a proportion of the solution each day (generally about 10%), and methods such as high performance liquid chromatography (HPLC) or mass spectroscopy have been shown to be useful for bath analysis of byproducts. A more quantitative determination of bath quality would enable out-of-spec bath conditions to be readily measured, identified and controlled, and more accurate preventative maintenance performed. Recent progress to quantify bath ageing and measure byproducts has been reported although implementation remains difficult, partly due to the proprietary nature of copper bath formulations [1-3]. Previously we have proposed electrochemical impedance spectroscopy (EIS) as a possible method For more papers by this company, or to contact the author, go to www.vertilog.com and retrieve paper V-EMT 1:27
for following bath quality during wafer processing [4]. Experiments in the laboratory during simulated bath ageing concluded that the impedance technique follows copper bath ageing during and after plating [4]. Furthermore, these results appeared to correlate with deposit morphology and thus may indicate overall quality of the bath during plating operation [5]. In this paper, impedance measurements were made on actual process bath solutions at a semiconductor-manufacturing site to validate the laboratory experiments, demonstrating that EIS could detect differences in bath solution types and also changes with time in a production environment. EIS is a steady-state technique capable of observing phenomena in electrochemical systems where relaxation times vary over many orders in magnitude [6]. A small-amplitude sinusoidal voltage, E(ω) is applied at a number of discrete frequencies, ω, ranging from 0.001 to 100,000 Hz, to a working electrode immersed in the fluid as shown in Figure 1. At each of these frequencies, a sinusoidal current results, I (ω), that is out-of-phase with the applied sinusoidal voltage signal. The electrochemical impedance, Z(ω), is the frequency-dependent proportionality factor between the voltage signal and current response. Z(ω) = E(ω)/ I (ω) Z(ω) is a complex-valued vector quantity with real and imaginary components, whose values are frequency-dependent. The real and imaginary impedance components can be plotted against each other to generate a Nyquist plot with characteristic © 2004 Vertilog Ltd. All rights reserved.
Air Liquide Chicago Research Center
Advances in Electronics Manufacturing Technology
semicircle shapes as shown in Figure 2. This plot is derived from a model of copper deposition that accounts for organic additive adsorption and described elsewhere [7]. Each point on the plot is the impedance at one frequency, and each semicircle is characteristic of a single time constant. The ohmic, or electrolyte solution resistance, is determined at the high frequency intercept on the real (horizontal) axis. The semicircle diameter gives the charge transfer resistance, which is related to the rate of the electrochemical reaction. As capacitors have only an imaginary impedance component, the semicircles above the horizontal axis are capacitive, while those below are inductive.
V-EMT 1:27
Figure 1. AC voltage signal imposed on an electrochemical system and the resulting AC current and impedance.
Methodology
Figure 2. A modeled EIS plot of copper deposition (extracted from Ref. [7]).
0.9
J É Ñ H FB
0.5 0.3
H FBJ É Ñ
É Ñ J B H F
0.7
-Im(∆E∆I) (Ω.cm2)
EIS measurements were performed on solutions extracted from copper plating tools in a wafer fab. Two different bath chemistries were studied; solution 1 contained a two additive system with an inorganic part consisting in copper sulfate, sulfuric acid and Cl-, and solution 2 was a three additive system with an inorganic part consisting in the same constituents as solution 1 but with different concentrations. Tools A1 and A2 were operated using bath solution 1, and tools B1, B2, B3 and B4 were operated using bath solution 2. Using a sampling port in each of the tools, copper bath solutions were periodically obtained from the process tools during a 10-day campaign and subjected to EIS measurement to determine the steady state stability of the bath solution as a function of processing time. The EIS measurement time depends on the frequency range being used. In this work frequencies between 0.01 Hz and 62.5 kHz were used, giving a measurement time of approximately 20 mins. The EIS equipment and setup consisted of a Sotelem-Vinci PGstat-Z potentiostat and Solartron 1250 frequency generator/analyzer for imposing and acquiring the electric signal, as well as an electrochemical cell and ancillary components (rotating disc electrode assembly, SMSE reference electrode, counter electrode). The disc electrode consisted of a 5 mm in diameter Goodfellow 99.99%+ copper rod embedded in an inert and insulating Presi Allylic Glass Fiber resin. It was polished with a 1200 grade SiC paper and rinsed with deionised water to clean the surface before measurement. The counterelectrode was a Goodfellow 99.99%+ copper plate with about 30 cm2 surface area. A small DC current was imposed on the system in addition to the AC current signal. In this way the conditions during deposition were closely followed during measurement, with the small DC current calculated to provide about the same current density (mA/cm2) that would be experienced by a wafer during plating.
B H FJ ÑÉ
FÑB H ÉJ
-0.3
1.0
1.5
2.0
H
B
F
2.5
3.0
J É Ñ
Tool A1
J
Tool A2
H
Fresh Solution 1
F
Fresh Solution 2
Ñ
Lab Aged (Vieil63)
É
Lab Aged (Vieil 53)
J
J
H BJ F É Ñ ÉÑ B J F B H J Ñ F H É Ñ É Ñ Ñ ÉBÉJ Ñ J É FH BÑ HBÉ HB BJ Ñ J F F H J J Ñ B F Ñ J FH J FH ÉBÑ FHÉÑ ÉJ Ñ HÉB FHÉB HB Ñ JÉ J F ÑB F J Ñ ÉJB B Ñ H Ñ Ñ B É H Ñ F FHÉH J B J B B F H H B J F Ñ J J F É J B É ÉÉJ Ñ É É HFH Ñ F Ñ ÉB É F HH F F
Ñ ÉJ F H B
0.0 0.5
B
B H J F ÑÉ
0.1 -0.1
B
H B HJ F ÑÉ
Ñ É J B F H
B
F
3.5
4.0
F
4.5
H
Re(∆E/∆I) (Ωcm2)
Figure 3. Impedance spectra comparison between process tool, fresh solution and lab-aged solution: Bath solution 1.
www.vertilog.com page 2
© 2004 Vertilog Ltd. All rights reserved.
Air Liquide Chicago Research Center
Advances in Electronics Manufacturing Technology
Results and Discussion
1.0 0.8
H Ñ JÇ É F B
-Im(∆E∆I) (Ω.cm2)
0.4 0.2
JÇ H Ñ É F B
Ñ É H B JF Ç
0.6
ÇH J Ñ É F B Ç J H F B É Ñ
Ñ É B H Ç JF
Ç H J B F Ñ É B B Ç F BH F É Ñ ÇHF J B É É F Ñ ÑHÇF É B ÑH BH J ÇH ÇJ JÇ É Ñ É Ñ BH J ÇB ÇÇ Ç F Ñ BJ JF H F É Ñ É É Ç J HÇÇ ÇÇ Ñ H B J H B F B F Ñ Ñ Ç JF Ç É Ç É J H Ç Ç H B É B J F F HÇ Ñ Ñ É J É Ñ JF B É H F Ñ JF HÇ H B É H É J BH Ñ F JÑ FÉB JÑ ÉB ÑF 1.5 2.0 2.5 H J B Ñ FÉ J B H Ñ
Ñ B F É H Ç J Ñ B F H É Ç J
0.0 -0.2
0.0
0.5
1.0
-0.4 -0.6
J Ñ
Ñ É J F H Ñ
-0.8
B
-1.0
J FÉB Ñ F É B H
Ç Ç 3.0
É FB
3.5
B
Fresh Solution
J
Tool B1
H
Tool B2
F
Tool B3
Ñ
Tool B4
É
Tool B4
Ç
Lab Aged Bath
Ç
H
Re(∆E/∆I) (Ωcm2)
Figure 4. Impedance spectra comparison between process tools using solution 2 and a fresh batch of the solution.
1.0 0.8 BÇ H J ÉF Ñ
-Im(∆E∆I) (Ω.cm2)
0.6
Ñ É H B FJ Ç Ñ ÉJ F H Ç B Ñ ÉJ F H Ç B
B HÇ J F É Ñ
0.4 0.2
BÇ ÉFJH Ñ
Ñ É FB J H Ç Ç F Ñ J É F F Ñ É Ç ÇB J HH J B H F J Ç H Ñ É É Ñ B B B Ç H Ç F J F Ñ É J Ñ É Ñ H F É B J Ç H B B F H Ç Ç J É B H J F Ñ É F Ñ É J F J B H Ç ÉÑ Ç H B Ñ F Ñ É F B H JÇ É J F F Ñ Ñ H H É Ç É Ç B J B Ç B JÇ H É Ñ F B J H É Ñ F B J É Ç É H Ñ F J B Ç Ñ H F É J Ñ B H Ç F ÉJ 1.5 2.0 Ñ HÇ FB É ÑJ B H FÇ
B HÇ J F É Ñ B J Ç H F É Ñ
0.0 -0.2
0.0
0.5
1.0
-0.4 FÇ É H BÑJ
-0.6 -0.8 -1.0
2.5
B
Day 1 14h17
J
Day 2 11h09
H
Day 3 13h54
F
Day 7 10h57
Ñ
Day 8 11h19
É
Day 9 10h40
Ç
Day 14 10h53
É ÑJ BÇ FH FÇ HB ÑJÉ
É FHBÇÑ J
Re(∆E/∆I) (Ωcm2)
Figure 5. Evolution with time of the impedance spectra recorded on tool B1 (solution 2) over a period of 14 days.
1.0
B
Day 1 09h25
0.8
J
Day 2 14h45
H
Day 5 15h00
F
Day 6 11h00
Ñ
Day 7 14h55
É
Day 8 15h23
Ç
Day 9 08h30
H Ç É BÑ F J H Ç Ñ F J É B
0.6
-Im(∆E∆I) (Ω.cm2)
Figure 3 compares the impedance spectra of solution 1 obtained from tools A1 and A2 and the spectra from fresh and aged samples of the same solution. The impedance spectra from the process tools are very similar and the baths appear to be at the same steady state. The tool bath impedance spectra were compared with those obtained in the laboratory for a fresh sample of solution 1 (no ageing or processing). It can be seen that the spectra obtained from the tool baths are close to the fresh solution but their third loops are slighter larger in diameter than the fresh sample. In addition, two spectra previously obtained in the laboratory on partially aged solution 1 were also added to Figure 3. The shape of the spectra is similar to those obtained from the process tool samples, however the third capacitive loop diameter is half the one obtained for the process tool baths. The bath solutions appear therefore not to be aging and to be staying close to the fresh solution. A possible explanation is that the process tool bleed and feed operation is compensating the bath so that the “aged” appears nearly like a fresh bath. This suggests that bath compensation could be optimized to reduce bath bleeding without affecting deposit quality. Figure 4 compares the impedance spectra obtained on the process tools using solution 2. Differences between the second (lower frequency) capacitive loop and the inductive loop diameters were observed between the spectra obtained from the different tools. The process tools appear to be at slightly different steady states from an impedance spectra standpoint. These differences are due to different wafer production levels or different concentration control. Indeed, as is shown in Figure 5, the impedance spectra can be very reproducible when the steady state of the bath is stable. However, a fresh sample of solution 2 was measured by EIS for comparison. The impedance spectra obtained from this fresh solution was close to those obtained from the process baths. For comparison, an EIS data obtained from a lab-aged sample of solution 2 is included in Figure 4. This is very different, showing a clear third capacitive loop. Again, it can be tentatively concluded that the steady state of the new chemistry bath in the process tool is close to the fresh solution. Comparison between Figures 3 and 4 shows that the impedance method is able to clearly distinguish between the two types of solution. Interestingly, solution 2, which is designed for better superfilling of narrow vias, does not exhibit a third capacitive loop, but rather a large inductive loop. As reported elsewhere, these spectral characteristics have been
V-EMT 1:27
0.4 0.2
Ç H F Ñ J ÉB
0.0 -0.2 -0.4 -0.6 -0.8 -1.0
ÉB Ñ F H J Ç B F É Ñ J H Ç
Ç F H J Ñ É B F Ç J Ñ H É B
0.0
Ñ BJ É H Ç F
0.5
F É Ñ JB H Ç H F Ç B É Ñ É J J BFÉ H HÑ Ñ Ç Ç BFJ J É H Ñ BF Ç HÑF J H É Ç FJ Ç B Ñ BF É É Ñ J FB H Ñ J B H Ç É Ç H Ñ F F Ç B ÉJ É Ñ B F Ñ J B É J H H Ç Ç F B É Ñ F Ñ H J Ñ É B F J J B Ñ H É Ç H Ç ÉÇ B ÑÑ Ñ H F Ñ J ÉÑ Ç B HF FF É J Ç É B JHH ÉÉ B Ç B Ç J Ñ B Ç HF 2.5 J F 1.0 1.5 2.0 É B F Ç J HF É Ñ F B J Ç F É HF B Ç J JÇ B É Ñ H FÑ Ç Ç BÉ HF J B J É HF Ñ
3.0
Re(∆E/∆I) (Ωcm2)
Figure 6. Evolution with time of the impedance spectra for tool B4 (solution 2) over a period of 9 days.
found to be the same as those from solutions that gave the smoothest deposits in the ageing studies performed with solution 1 [5]. In that work, the third capacitive loop had to be minimized to obtain the smoothest deposits. Figures 5 and Figure 6 show the evolution with time of the solution 2 impedance spectra for tools B1 and B4, respectively. The impedance spectra do not exhibit a third capacitive loop, but rather a large inductive loop. Figure 5 shows that the impedance spectra remain very similar over 14 days of measurements for tool B1, indicating good stability of the plating bath. A
www.vertilog.com page 3
© 2004 Vertilog Ltd. All rights reserved.
Air Liquide Chicago Research Center
Advances in Electronics Manufacturing Technology
change in the diameter of the inductive loop was seen in Figure 6 for three spectra (Days 5-7). This may indicate a change in the steady state of the bath on these days, perhaps due to some aging process. Also, in Figure 6, a change in the impedance spectra associated with the formation of a small third capacitive loop was observed (curve corresponding to the bath sampled on Day 7 at 14h55). This change could be correlated with a wafer breakage in the tool during manufacturing. After removal of the wafer, the EIS spectra returned to its previous steady state. These results could indicate that the impedance measurements made over time are able to detect changes in the steady state of the bath. However, these changes may or may not be significant to the overall quality of the bath but may represent the normal fluctuations associated with the dynamic plating process. Work is now in progress to correlate the EIS spectra with actual chemical changes. It has been reported that the 1st capacitive loop is due to the copper deposition reaction and that the second capacitive loop and inductive loop are related to the additives [7]. It is possible that the change in the second capacitive loop and the inductive loop are related to the additive breakdown mechanisms and possibly the complexation of the accelerator/suppressor with the copper ion that is occurring in the bath and at the electrode surface. Further measurements will be made to confirm this behavior and correlate with bath composition and deposit quality.
V-EMT 1:27
References [1] M. Pavlov, E. Shalyt, P. Bratin, D.M. Tench, Proc. Electrochem. Soc., PV 2003-10, pp53-62, (2003). [2] A. Jaworski, K. Wikiel, 203rd Electrochemical Society Meeting, Paper 678, (2003). [3] P. Bratin, G. Chalyt, A. Kogan, M. Pavlov, M.J. Perpich, D.M. Tench, 203rd Electrochemical Society Meeting, Paper 676, (2003). [4] C. Gabrielli, P. Mocoteguy, H. Perrot, A. Zdunek, D. Nieto Sanz, M.Clech, Proc. Electrochem. Soc., PV 2003-10, pp129-36, (2003). [5] C. Gabrielli, P. Moçotéguy, H. Perrot, A. Zdunek, P. Bouard, M. Haddix, Electrochemical and Solid State Letters, 7, ppC31-C34, (2004). [6] R.Varma and J.R. Selman, in ‘Techniques for Characterization of Electrodes and Electrochemical Processes’, p515, John Wiley & Sons, (1991). [7] C. Gabrielli, J.Kittel, P. Mocoteguy, H. Perrot, A. Zdunek, M. Haddix, P. Bouard, L. Doyen, 203rd Electrochemical Society Meeting, Paper 404, (2003).
Conclusions We have performed EIS measurements on copper interconnect deposition baths during copper metallization to characterize changes in the plating bath condition. A comparison between two bath solutions indicates that the impedance method is able to clearly distinguish between different bath compositions. In addition, the results show that for a given starting chemistry the impedance spectra can detect changes in the bath over time and can also detect process excursions. These results demonstrate that EIS is a possible candidate for an online monitor of bath quality. Work is now in progress to better correlate differences in EIS spectra with actual changes to the bath chemistry.
www.vertilog.com page 4
© 2004 Vertilog Ltd. All rights reserved.
