Optimization of Cupric Chloride Subtractive Etching for Cu High Density Interconnects

As microelectronics continue to shrink in size while still requiring more functionality in the same small spaces, industry has pushed for increasing innovation in miniaturization. This has led to the development of high density interconnect (HDI) printed circuit boards, which require very fine and very reliable controlled Cu feature etching [1]. The primary fabrication pathway for these features is subtractive copper etching, which can be accomplished with a variety of etchants, including ferric chloride² and ammonium hydroxide [3]. The most widely used method for etching, however, is using acidic cupric chloride [4–6].

Acidic CuCl₂ as a Cu etchant has several advantages. Primarily, it has the ability to be easily regenerated via the addition of an oxidant such as hydrogen peroxide or sodium chlorate [7], [8]. The oxidizer converts the Cu(I) complexes back to Cu(II) complexes thus regenerating the original etchant. The overall process of reclaiming the original etch bath is generally divided into two distinct steps, the first of which is the regeneration of the CuCl₂ etchant. The second step is the restoration of the original etch bath, via ion additions to make up for ion losses during etching and regeneration, so that the used quantity of etch bath can then be reused for etching [9]. This regeneration and restoration eliminates vast quantities of etching waste and improves overall efficiency, making CuCl₂ an appealing choice of Cu etchant for PCB industry firms [10].

Quality control in fabrication of HDI PCBs relies on the patterned Cu features having very precise control of etch depth and undercut for the most consistent etching profile [11], [12]. As would be expected, both the etch rate and etch profile are dependent on the chemical composition of the etch bath. This requires that any method of monitoring of the etch bath not to disturb the chemical equilibrium. The complicated, interlocking picture of combination of Cu(I) and Cu(II) complexes that exist in solution can be shifted by any minor perturbations [13]. The non-perturbative standard parameters that are currently used in industry to monitor the acidic cupric chloride etch baths are oxidation-reduction potential (ORP), conductivity, and specific gravity [4], [14], [15]. Each of the tools correlates to some optimal aspect of the operational characteristics of Cu etching bath and taken together, they provide the standard picture of the condition of the etch bath. However, each tool is not chemically specific, and they can fall short in predicting the etch bath chemical equilibrium, especially at the extremely fine level required for next-generation device design and manufacture. In previous work, we reported on the use of thin-film UV-Vis as a monitoring tool to supplement those currently in use [16]. UV-Vis also does not disturb the chemical equilibrium, unlike the standard parameters, it is very chemically specific and can produce a snapshot of the etch bath chemical equilibrium during etching. Thin-film UV-Vis has the capability to analyze the solutions of very high concentrations that are encountered in Cu etch baths [17], [18]. Here, we expand on the use of UV-Vis as applied to the further regeneration and restoration of Cu etch bath to its original operating condition. We find that UV-Vis has the capability to effectively monitor the various changes to the chemical equilibrium of the etch bath during etching, regeneration and recovery. UV-Vis also sheds light on the decoupled roles of H⁺ and Cl⁻ to the Cu etching mechanism. Finally, UV-Vis can help to monitor the condition and predict the behavior of the etch bath in cases where the standard parameters fall short.

The etch bath solution prepared here was 2.0 M CuCl₂ / 1.0 M HCl, prepared using as-received CuCl₂·2H₂O (Acros) and 36% HCl (Fisher) and diluted to volume with >18.2 MΩ Millipore water. For etching measurements, as described in previous work [16], 2×2 cm coupons were cut from 35 μm Cu panels (epoxy core with 35 μm thick Cu on each side). The coupons were weighed and suspended for 6 min at a consistent depth and position in 30 mL of the etching solution that is kept under constant stirring and maintained at a constant 40 °C during etching by a thermal jacket. The coupons were removed, then rinsed, dried and reweighed with an analytical balance (± 0.1 mg), from which etch rates were calculated from total exposed surface area and the weight change of the coupons. The post-etch solutions were regenerated by using a calibrated 25 μL micropipette to deliver 30% H₂O₂ until the etch bath color changed from dark brown-green to light green, indicating full regeneration of Cu(I) to Cu(II), which was further confirmed via ORP and UV-Vis measurements. Further post-regeneration ion additions were made with HCl, KCl (Alfa Aesar) and 70% HClO₄ (Fluka). The etch bath was measured at each stage via UV-Vis, ORP, Conductivity, specific gravity, and pH. UV-Vis measurements were taken at room temperature with an Agilent 8453 UV-Vis Spectrophotometer using quartz cuvettes of ca. 70μm path length that were fabricated in-house by a dry-etching method using SiF₄/O₂ etching gases in an AGS Plasma Etcher (RIE MPS-150). The peak apex of the UV-Vis Spectra at the NIR region was identified to study the wavelength shift of the peaks during etching. ORP measurements were conducted at room temperature using a CHI 440 potentiostat and measuring the stable open circuit potential using a standard 3-electrode setup with Ag/AgCl reference. Conductivity measurements were conducted with a ThermoFisher Orion Star Benchtop Meter and Two-Electrode Cell. Conductivity measurements were made at constant 40 °C because of the intense temperature-sensitivity of conductivity. The pH measurements were made using the same ThermoFisher Benchtop Meter equipped with an Orion 8157BNUMD ROSS Triode. Specific gravity measurements were recorded gravimetrically at room temperature with 25 mL aliquots pipetted into 25 mL volumetric flasks. Calibration with water and CuCl₂ solutions showed that specific gravity measuring sensitivity is better than 1 g/L.

The fundamental Cu etching reaction with acidic CuCl₂ etch bath is:

As Cu etching progresses, the accumulated spent CuCl needs to be regenerated back to CuCl₂, by adding oxidizing agent like H₂O₂, to recover its Cu etching potency. During the H₂O₂ regeneration process, some of HCl is “consumed” via the following equation:

In order to maintain the targeted Cu etch rate, controlled amounts of HCl is added periodically to the Cu etch bath in order to “restore” back to the original etching potency. Currently, the regeneration and restoration steps of the acidic cupric chloride etch bath are controlled in real time using three standard monitoring tools including oxidation-reduction potential (ORP), conductivity, and specific gravity. However, these monitoring tools are not chemically specific enough, often leading to poor etching control and resulting in wasteful premature dumping of Cu etch bath. In the following section, we explored UV-vis spectroscopy to characterize Cu etching baths during etching, regeneration and restoration with an aim of achieving better process control of Cu etching process.

Fig. 1 displays the etch rate and visible color changes in the etch baths of original pre-etch solution, post-etch and subsequent regeneration processes by adding H₂O₂. There is a constant increase in Cu etch rate as the solution is regenerated, but it reaches a limit at full regeneration (4.8 μm/min) that is significantly lower than the original etch bath (5.5 μm/min). Additionally, the final regenerated post-etch bath is visually more blue in color than the original, indicating a different chemical equilibrium state. This simple visual inspection points to the necessity of additional steps like addition of HCl in order to restore etch bath to its original chemical equilibrium for the targeted Cu etch rate.

The UV-Vis spectra of pre-etch, post-etch, and regenerated solution are displayed in Fig. 2. The most striking feature is the elimination of the visible shoulder feature in the 450–600 nm range of the spectra after adding H₂O₂ for regeneration. During etching, this 450–600 nm shoulder increases significantly, and it is likely directly related to the Cu(I) in the etch bath, as discussed in previous work [16]. The large absorption increase across the visible range leads to a visual color change from green to brown (see Fig. 1). As H₂O₂ is added, the 450–600 nm feature shrinks until full regeneration, when this peak returns to nearly the original shape of the pre-etch bath (dashed line in Fig. 2). Visually, the etch bath returns to nearly its original color, except now with a distinctly blueish tint. This color difference is reflected in the UV-Vis, as the visible range shoulder (350–450 nm) in the regenerated post-etch bath is mildly blue-shifted compared to the pre-etch bath, indicating a deficiency in HCl, which will be further discussed later. The 400–600 nm shoulder feature has utility as a monitoring component, as it monitors the Cu(I) concentration in solution, and can be used as a guide for achieving precise regeneration control.

The other major feature to note in the spectra is the broad NIR peak, from 750–1000nm. This peak arises from Cu(II) chloride complexes, and is typically ascribed to an optical d-d transition [19], [20]. During etching, the NIR absorbance peak intensity decreases linearly with etch rate¹⁶ and blue-shifts the peak by ~10 nm (see inset of Fig. 2). During regeneration, nearly the inverse effect is observed. The intensity increases all the way to full regeneration, ending more intense than the original pre-etch solution. This result is explained by accounting for the copper ions added to the solution via Cu etching that, after post-regeneration, is now Cu(II), making the fully regenerated solution having a higher Cu²⁺ concentration. Both Cu(I) and Cu(II) exist in multiple equilibrium of chlorinated and hydrated complexes in Cu etch bath solution, the equilibrium of which is dominated by chloride concentration due to its much higher complexing tendency [21]–[23].

The NIR absorption peak is best understood as an indicator of the concentration and distribution of various cupric chloride complexes in Cu etch bath solution. Although there is higher Cu²⁺ content in solution after fully regeneration, the availability of Cl⁻ to them is lower [13]. This leads to cupric chloride complexes that are less efficient at Cu etching in the fully regenerated post-etch bath than the original pre-etch bath, which is supported by ~15% lower etch rate data observed in Fig. 1. As illustrated in (2), the H₂O₂ regeneration process consumes some of the HCl in solution. The lower etch rate in the regenerated solution indicates the importance of adjusting HCl concentration (i.e. restoration step) to maintain a constant Cu etch rate. The shape of the NIR absorption peak is also helpful for understanding and indicates the effect of Cl⁻ concentration. Full regeneration does not red-shift the peak back to its original position (870 nm), which is expected since the Cu²⁺ increases without corresponding a Cl⁻ increase, and is consistent with the NIR peak shape indicating the Cl⁻ deficiency of cupric chloride complexes. Overall, the UV-Vis effectively captures the complex changes in the chemical equilibrium through etching and predicts the etching and regeneration behaviors. It can be analyzed in a two-dimensional manner, both in height and shape of peak absorbance, for even better chemical equilibrium monitoring and control.

For comparison, it is important to examine how the standard monitoring parameters behave during regeneration, and the data is summarized in Fig. 3. ORP, as previously examined, ¹⁶ behaves similarly to the visible range feature in the UV-Vis spectra, in that it is sensitive to the presence of Cu(I) in the etch bath. As expected from the Nernst equation, the ORP drops precipitously upon the addition of a small amount of Cu(I). This gives ORP excellent sensitivity to Cu(I), but the application to the etch bath is limited beyond that, since it is only very slightly sensitive to other species that are involved in the etching chemical equilibria, such as Cu(II), Cl⁻ and H⁺. Over-regenerating the solution by adding excess of H₂O₂ produces a higher ORP than the original etch bath, as expected, but this is not reflected by a significant corresponding change in the etch rate of the etch bath, as seen in Fig. 1.

Conductivity, used as a measure of HCl [24], predictably decreases as the solution regenerates, as the availability of free HCl in the solution is consumed by the regeneration mechanism. The most interesting result is specific gravity, which is commonly used by industry as a measurement of the Cu concentration in the etch bath [4]. As our data indicated in Fig. 3, addition of less dense 30% H₂O₂ (d =1.11) to post-etch solution (d=1.221), actually increased the overall specific gravity of fully regenerated post-etch solution (d=1.230). This is counterintuitive, as the peroxide molecules essentially convert to water by the regeneration reaction. This may be explained by considering that Cu etching, while generating Cu(I) ions, is likely also generating multi-valent, Cu(I)/Cu(II) complex [25]–[27], that matches the observed dark brown complexes. It is reasonable to assume that this complex could act to lower the specific gravity of the post-etch bath. Subsequently, adding H₂O₂ to oxidatively remove this Cu(I)/Cu(II) complex helps to restore the expected increase of specific gravity. These observations serve as an indicator of the complexity of the changes in the etch bath chemical equilibria that the UV-Vis is able to capture, but the standard parameters do not.

The restoration of the regenerated post-etch bath to its original Cu etching potency of pre-etch condition requires the addition of further reagents beyond the oxidizing regenerating agent. Since HCl is consumed during regeneration, a logical choice to recover the original etch bath parameters is the addition of concentrated HCl. The changes in the UV-Vis spectrum of adding HCl is displayed in Fig. 4 and the corresponding changes in the monitoring parameters and etch rate are in Fig. 5. As expected, the NIR feature increases in absorbance intensity and slightly red-shifts and the visible region (350–450 nm) shoulder red-shifts, as seen previously with increasing HCl content of the etch bath [13], [17]. At 0.5 M HCl addition, the peak of the NIR feature has returned to 870 nm, the same location as the original etch bath, and the corresponding etch rate has now returned to approximately the original etch rate (Fig. 5). In this simple peak shape analysis, the NIR absorbance peak profile does correspond well to the relative Cu etch rate of the etch bath. This confirms the central role of Cl⁻ and the [Cu(II)Cl_x]^2−x complexes to the fundamental Cu etch mechanism and demonstrates the ability of UV-Vis to effectively monitor the complex changes in the etch bath chemical equilibrium.

Since the NIR peak arises mainly from [Cu(II)Cl_x]^2−x complexes in the etch bath and is closely correlated to the etch rate, it might be reasonable to hypothesize that the etch rate is more sensitive to Cl⁻ concentration than H⁺. It is therefore worthwhile to explore the separate effects of H⁺ and Cl⁻ on the etch rate and various monitoring parameters, rather than on the addition of the whole of HCl. To accomplish this, the H⁺ and Cl⁻ were split into additions of KCl and HClO₄. K⁺ is largely a spectator ion, and perchlorate ion is non-complexing to Cu ions in solution [28], so this gives an approximation of the additions of H⁺ and Cl⁻ separately into the regenerated post-etch bath.

Overall, as seen in Fig. 5, the etch rate increases significantly with the addition of KCl, but slightly decreases with the addition of HClO₄. This correlates very well with the UV-Vis, as seen in Fig. 6(b). The NIR (~700–1000 nm) peak increases significantly with KCl addition, but sees a very slight decrease upon addition of HClO₄. This behavior reinforces the hypothesis that the etch rate is more closely tied to the [Cl⁻] than the [H⁺] in the etch bath, and that the NIR peak is largely dominated by the [Cu(II)Cl_x]^2−x complexes in the etch bath. It is therefore vital that the monitoring of the etch bath includes sensitive measurement of the Cl⁻ concentration. UV-Vis provides some of that capability, and can clearly predict and explain the change in etch rate upon various ion additions. The pH of the solution certainly plays an important role in etching process, as low pH helps to break up the initial outer copper oxide layer and prevents copper hydroxides from precipitating out of the etch bath [29], [30]. However, the change in measured pH, even after the H₂O₂ regeneration consumes some H⁺ (see (2)), is only a difference of ~0.2 pH. Therefore, constant monitoring of [H⁺] is less vital as the typical Cu etching operation does not result in large change in measured pH.

The UV-Vis prediction of the etching behavior differences between KCl and HClO₄ is not similarly reflected in the standard monitoring tools. Conductivity is typically used as a measure of the HCl concentration in the Cu etch bath. However, the molar conductivity for H⁺ [349.6 S cm² mol⁻¹] is much higher than for Cl⁻ [76.35 cm² mol⁻¹] [31], which suggests that H⁺ has a larger effect on conductivity than Cl⁻. This is reflected in the conductivity measurements, seen in Fig. 7, as the increases in conductivity for HClO₄ addition are much higher than for KCl addition. In fact, the conductivity changes for HClO₄ addition are nearly identical to those of HCl addition. This indicates that the conductivity is mostly capturing the change in [H⁺] in the Cu etch bath. If conductivity monitoring data were capturing the HCl only, then increasing the conductivity would predict a higher etch rate. A larger increase in etch rate would then be expected for the HClO₄ than the KCl addition, which is the opposite of what occurs. As such, standard conductivity monitoring parameter picture does not predict the changes in Cu etch rate from KCl and HClO₄ addition to the etch bath, while UV-Vis does predict the etch rates very effectively. This is directly tied to the lack of chemical specificity in the current monitoring parameters, which UV-Vis can make up for as a complimentary tool to check the overall condition of the chemical equilibrium of the etch bath.

Over the complicated sequence of changes to the etch bath from etching, regeneration and restoration, the chemical equilibrium related to etching can be modified in many different ways. UV-Vis can serve as a chemically-sensitive way to check that the overall condition of the equilibrium of the etch bath is where it is supposed to be. This is particularly true of the NIR 860 nm feature, since it arises from the d-d transitions of the [Cu(II)Cl_x]^2−x complexes that drive the fundamental Cu etching mechanism. It is very likely that improper combination of regeneration and restoration steps can still produce standard monitoring parameter values, using currently used non-chemical specific probes, that are within targeted etching bath control tolerances, but have entirely different etch Cu rates from original etch solution. For example, when the final HClO₄ addition level is diluted down so that the specific gravity of the final bath matches that of the initial bath, all three monitoring parameters come to within standard process tolerances, as seen in Table 1. This would be expected to produce similar etch rates for the two etch baths, but they are, in fact, very different. UV-Vis, in this case, does predict the difference in Cu etch rate, as the absorbance of the NIR feature of the improperly regenerated/restored etch bath is much lower than that of the original etch bath (Fig. 8). UV-Vis is able to detect the difference in etch chemistries and thus aid in the prediction of the behavior of the etch bath beyond what the other parameters can accomplish together.

Thin-film UV-Vis has the capability to serve as both a monitoring tool for acidic Cu etch baths, and as pathway to reveal new insights about their behavior and mechanistic features. Through its use, the Cl⁻ concentration can be monitored in the etch bath apart from the H⁺ in solution. UV-Vis also helps to create a more comprehensive understanding of roles of H⁺ and Cl⁻ in the etching mechanism. Furthermore, it can act as a support for the current etch bath monitoring tools, helping to fill in where they fall short over the course of the complex system of changes to the etch bath chemical equilibrium, and help to achieve the high levels of etching control required for the ever-increasing demands of HDI PCBs development.

Authors (OC) acknowledge UNT and Intel MSR funding (contract # 2018-IN-2820) for financial support of this work and the Nanofabrication Cleanroom Facility in UNT Center for Advanced Research and Technology for their help in the fabrication of the quartz cell.