Controlling Germanium Oxidation With a Vacuum Substrate Carrier

Queue time is unavoidable for sub-65 nm technology development. It is caused by a lack of equipment, transfer between semiconductor tools or transfer to and from remote sites. Molecular contamination has the potential to cause damage equivalent to today’s particle contamination. Furthermore, outgassing from a wafer or substrate carrier box will be extremely critical for neighboring wafers.

The quality of wafers can be affected during queue time by three major contamination issues: oxidation, which can occur between wafer process steps; the accumulation of organic and airborne molecular contaminants (AMCs) on silicon wafers during storage and shipping; and particle-generated defects.^1-5 These issues, which become more critical as feature size shrinks, will degrade the expected yield regardless of other parameters. For instance, the presence of native oxide on a wafer can affect the quality of different processes, such as dielectric thin-film deposition (high-k), atomic layer epitaxy or metal gate deposition. Corrosion is one of the visible effects of this phenomenon. Reducing or eliminating these effects during queue time can significantly increase yield.

In current fabs, the solutions are either to add cleaning steps or apply very stringent procedures to reduce batch queue time. These solutions increase the cycle time and complexity of the process flow. Moreover, the added cleaning steps imply etching of underlying layers, which could delay the introduction of new materials for <65 nm technologies. One of the ultimate solutions is to transport contamination-sensitive substrates under vacuum without any contact with the atmospheric environment.

In this article, a new method of oxidation and AMC prevention is presented, using ultraclean vacuum wafer mini-environment technology developed by Alcatel Vacuum Technology France (AVTF) in collaboration with LTM-CNRS within the framework of MEDEA+ Hymne project. AVTF is developing a vacuum substrate carrier — or “mobile loadlock concept.” The interior of the vacuum carrier stays within sub-vacuum pressure range, maintaining an ultraclean, oxygen-free mini-environment.

Germanium (Ge) has the potential to replace silicon in certain applications. It has the advantage of higher carrier mobility than silicon for both negative and positive charges, and a better compatibility with high-k materials. Also, a lower dopant activation temperature eases the formation of shallow junctions. However, germanium oxidizes significantly faster than silicon,⁶ which could create manufacturing challenges. We tested germanium wafer oxidation to qualify our experimental setup. The results show a significant benefit using vacuum transfer and storage, as opposed to standard front opening unified pods (FOUPs).

Contamination control, cycle time and footprint can all have a large impact on manufacturing productivity. Today, 30% of cycle time is due to loadlock pump-down time and substrate transfer time to the next piece of equipment. These losses can be minimized by using the vacuum substrate carrier. It will also have a secondary impact on pump operation energy savings. Footprint will be reduced because the vacuum carrier provides the equivalent of three elements — loader/carrier, robot transfer and loadlock.

As defined by Alcatel, a vacuum substrate carrier is a transportable metallic pod that interfaces directly to a vacuum transfer chamber without breaking the vacuum environment. The system replaces the standard plastic pod, its loader and loadlock chamber. Furthermore, the vacuum substrate carrier has enabled new operations that were not possible with the traditional substrate carrier under atmospheric pressure. The processed substrates can be stored under vacuum, avoiding any contact with volatile organic compounds (VOCs), oxidation or undesired outgassed materials from carrier walls or the substrate itself. Cross-contamination is also suppressed.

X-ray photoelectron spectroscopy (XPS) measurements were performed on a customized Theta 300 spectrometer, which was directly interfaced to the vacuum carrier pod via a vacuum transfer chamber (10^-7 mbar range). All measurements were performed on 200 mm wafers.

Two kinds of wafers were used in this study: silicon (001) blanket wafers and silicon (001) wafers with 2.5 µm thick germanium layers epitaxially grown by chemical vapor deposition (CVD). Chemical treatments were performed in the SP203 200 mm spin-on single-wafer tool. In order to limit the reoxidation of the germanium surface during cleaning, the following HF-based sequence was used: After being deoxidized in a HF 2% solution for 60 sec, silicon or germanium surfaces were rinsed using a deionized water solution. The samples were then dried with nitrogen gas and put into the vacuum carrier pod immediately.

The aluminum Kα X-ray source (1486.6 eV photons) has been used for all analyses at a pass energy of 60 eV. The concentrations of silicon, germanium, oxygen and carbon atoms were extracted from the Si₂p, Ge₂p_3/2, O1s, and C1s core-level energy regions, respectively. For germanium quantification, the Ge₂p_3/2 level (1220 eV binding energy) was used because of the high surface sensitivity obtained at this low level of photoelectron kinetic energy. The angle-resolved (AR) capability of the Theta 300 was used for all analyses, with angles of between 23.5° and 72.5°. This gives access to depth-resolved information. Note that the angles values are referred to the normal value of the substrate. Using a numerical fitting procedure, spectral deconvolution was performed to extract the peak contributions in the acquired energy regions. Individual line shapes are simulated with a combination of Lorentzian and Gaussian functions. The background subtraction was performed using a Shirley function calculated from a numerical iterative method. After correction of the lens transmission factor, each element concentration is obtained by dividing calculated peak areas by the corresponding Scofield cross-section (Si₂p: 0.87; Ge₂p_3/2: 24.5; O1s: 2.93; C1s: 1.0). The sum of the concentration of different elements present on the analyzed surfaces is equal to 100%.

The efficiency of the silicon passivation was investigated using AR XPS measurements. Figure 1 shows the results obtained for a HF-passivated silicon wafer after 45 min. exposure in air. The top curve corresponds to a 76.5° take-off angle (grazing angle), whereas the bottom curve corresponds to a 23.5° take-off angle. The Si₂p energy level clearly shows the wafer’s monocrystalline Si-Si component with its typical spin-orbit coupling and the Si₂p_3/2 peak centered on 99.3 eV.⁷ If silicon surface SiO_x oxidation exists, it appears as a shoulder peak to the Si-Si component, yet it is not observed here. Thus, the HF passivation is very efficient here for silicon oxidation prevention, in accordance with previous studies where oxidation kinetics, in terms of days, have been reported.⁸ This result is enforced because of the AR mode. The top curve at 76.5° corresponds to a probed silicon depth of 1.7 nm, and the spectrum clearly shows that no oxidation took place, even at a grazing angle displaying great surface sensibility.

1. Si2p region of a HF-passivated 200 mm wafer after 45 min in air.

XPS analysis with germanium

Germanium, germanium on insulator (GOI) or SiGe substrates could be used with high-k gate stacks in order to enhance carrier mobility in the gate channel.⁹ However, germanium is known to be very reactive to air, giving undesired and non-reproducible oxidation.¹⁰

In order to quantify this surface reactivity and have reference data, an experiment similar to that for the silicon substrates was performed on passivated germanium wafers. Germanium wafers were loaded directly onto the vacuum carrier after various exposures to air for periods of time from 4 min to 2 months.

Figure 2 shows the evolution of the Ge₂p region of a germanium wafer with reference to air exposure time. The upper curves correspond to grazing angles with high surface sensitivity, and the lower curves correspond to bulk angles. The escape length λ (inelastic mean free path) was estimated using numerical computation based on the TPP2M formula.¹¹ With an escape length λ of 0.8 nm for the Ge₂p 266 eV photoelectrons, one can estimate the probed depth for the grazing and bulk angles. This results in a probed depth of ~0.6 nm for a grazing angle of 72.5° and 2.2 nm for a bulk angle of 23.5°.

2. Ge2p region of a HF-passivated substrate after 4 min (a), 15 min (b), 48 hr (c) and 2 months (d).

Figures 2a and 3 show a clear difference compared with the silicon surface. Only 4 min after cleaning, despite the HF passivation treatment, an oxidation evidence shoulder appeared around 1220 eV in the Ge₂p region. It showed an incomplete oxidation level of germanium. This GeO_x contribution comes in addition to the bulk Ge-Ge component at 1218 eV. Referring to the Si-H case,^12,13 and considering similarities between silicon and germanium in terms of electronegativity and position in the periodic table,¹⁴ a Ge-H contribution was fitted with a 0.3 eV shift from the Ge-Ge peak. Note that this Ge-H contribution disappears very quickly after a few hours in favor of the GeO_x component (for clarity, this has not been plotted in Fig. 3). Four minutes after passivation (Fig. 3a ), and despite the presence of this Ge-H peak at 1218, 3 eV, the spectral deconvolution led to a small 7% GeO_x contribution centered on 1220 eV.

3. Quantification results of a germanium HF-passivated substrate after several air exposure durations. Quantification has been done for the bulk angle (27.5° from the substrate’s normal).

It is important to note that this weak GeO_x component was enhanced at grazing angles, which indicated that the oxidation appears at the top of the substrate. Because of the high surface sensitivity obtained with AR measurements, we concludes that this GeO_x upper layer was very thin, with an approximate magnitude of 3-4 Å. This corresponds to only one monolayer of oxidized germanium.

Figures 2b and 2d and quantification results from Figure 3 showed that this oxidation continues slowly during the days and weeks following the HF treatment. AR spectra give depth-resolved information and information on the oxidation kinetics of germanium. Furthermore, with reference to the last measurement found on Figure 2d , where the oxide component is stable against the angle, one can observe that the oxide peak is always enhanced at grazing angles. This means that the oxidation progresses from the top of the germanium substrate. Slow oxygen incorporation was introduced with a more complete germanium oxidation state. It is also proved by a GeO_x component shift toward higher energies. Finally, after two months of exposure in air, this oxidized peak is centered around 1221 eV, which corresponds to a complete oxidized germanium bounding environment (GeO₂).⁷ Using previous considerations for λ and noting that the germanium substrate peak almost disappeared at a grazing angle of 72.5°, the GeO₂ layer thickness can be estimated to be more than 4 nm.

Carbon contamination rises slowly during air exposure as well. The carbon fraction of Figure 3 includes C-C and C-O components, which explain the drop of carbon percentage after two months exposure in air caused by Ge/C oxidation competition. It favored germanium toward the end of the exposure period.

The conclusion for germanium oxidation analysis is that germanium remains highly sensitive to air, even after a few minutes, despite HF passivation treatment. This is not the case for silicon, where the passivations are very efficient. However, this high surface reactivity to oxygen leads to incomplete germanium oxidation that leads to unstable and poor surface-state quality. The oxidation reactions continue as time passes, and a complete oxidation state of the substrate surface was clear only after two months of air exposure.

Because of the unavoidable queue times in the manufacturing process, the high reactivity to air exposure is a problem for the quality and reproducibility of the following process steps, such as high-k deposition. It is further complicated by the oxidation reaction, as well as the surface chemical state that continues to evolve for weeks.

To qualify vacuum storage in the carrier, a HF-passivated germanium substrate was loaded, just after HF treatment, into the carrier. With the Theta 300 full-wafer XPS analysis capability, the elapsed time between the HF treatment and vacuum storage was <4 min.

The vacuum carrier was then separated from the interface and left in the cleanroom without any additional pumping for 24 hr. An XPS analysis was performed (an ultrahigh vacuum [UHV] step), and the carrier was left an additional 24 hr without further pumping. After another XPS measurement, the vacuum carrier was then left for four days without any pumping, then the last XPS measurement was taken (six days after the HF treatment).

Figure 4 shows an excellent result, compared with Figure 2 (germanium in the air). As for Figure 2, the germanium bulk contribution clearly appears at 1218 eV, associated with a weak shoulder centered around 1220 eV, attributed to a Ge-H and Ge-O bonding environment. Figures 2a and 4a are very similar because the HF passivation treatment was performed on an industrial cleaning tool with high reproducibility. Nevertheless, Figures 4b to 4d clearly show that the Ge-O/Ge-H component was stable during the storage time in the vacuum carrier. Even after six days of storage, there was no visible change in the Ge₂p spectrum shape. This result is reinforced by the fact that it is valid, even for a very sensitive surface-grazing angle of 72.5°, which only probes to a depth of 0.6 nm. It is a full demonstration of germanium oxidation in a vacuum carrier, showing bulk germanium substrate oxidation, which is considered to be very sensitive to air exposure.

4. Ge2p region of a HF-passivated substrate vs. four min in air (a), one day in carrier (b), two days in carrier (c), and six days in carrier (d). No Ge-Ox observed after six days.

The quantification results shown in Figure 5 confirm this absence of germanium oxidation, since the GeO component showed no variation at all during the storage period. Moreover, there is a significant increase in the surface oxygen content. This oxygen contribution is attributed to the extreme surface carbon contamination, known to be unavoidable, even in UHV chambers.¹⁵ The surface contamination is mainly composed of C-C bonds, associated with a few C-O bonds. That is why the oxygen contribution correlates with the global carbon amount on the surface (Fig. 5 ). However, this rising oxygen level remains much lower than in air: 13.5% of oxygen linked to carbon in the case of six days of vacuum storage vs. 24% of oxygen linked to carbon and germanium in the case of only 48 hr of air storage.

5. Quantification results of a HF-passivated substrate vs. storage time in a vacuum carrier. Quantification has been done for the bulk angle (27.5° from the substrate's normal).

Another good indicator of the overall surface oxidation and contamination is the substrate’s signal screening. The substrate’s photoelectrons (Ge-Ge peak) are screened by the upper-oxidized or carbon-contaminated layer. Comparing the curves of Figures 3 and 5 , Ge-Ge atomic percentage is <26% after two days of storage in air, while it is already >30% after only two days vacuum storage. This difference is attributed to the absence of germanium oxidation under vacuum.

XPS analysis was used to evaluate the oxidation of germanium films when the substrate was under vacuum storage. After six days, no oxidation was observed on germanium wafers stored in the vacuum carrier without additional pumping. The results were compared with traditional air exposure storage, and showed a clear difference. Applying these results to the current defectivity in semiconductor manufacturing caused by oxidation and AMC contamination, such a carrier has the potential to improve fab cycle time by drastically reducing pump-down times. It also provides the ultimate solution for an urgent need in the near future of the International Technology Roadmap for Semiconductors (ITRS) for high-quality, oxidation-sensitive, germanium-based substrates.

This work was performed in the frame of the MEDEA+ Hymne project with the support of the French Ministry of Finance and Industry (MINEFI).