Davide Codegoni

Molybdenum Contamination in Silicon: Detection and Impact on Device Performances

In this paper, a case of molybdenum contamination from wet cleaning is discussed, and various techniques are compared for their ability to detect molybdenum. In addition, the impact of this sort of contamination on the electrical results of a bipolar device is studied.

Niobium Contamination in Silicon

In this paper niobium is characterized as a silicon contaminant. It is shown that niobium is a relatively slow diffuser, with an intermediate diffusivity between very slow diffusers such as molybdenum and fast diffusers such as iron. Niobium is found to be very effective as a recombination center, and in addition prone to surface segregation. In addition, niobium shows optical activation, but no thermal activation. Three deep levels are revealed in niobium contaminated silicon, plus an additional level observed in high contamination dose samples only. One of these levels is located very close to midgap, and consistently niobium was also found very effective in degrading the generation lifetime.

Highly Sensitive Detection of Inorganic Contamination

As the detection of inorganic contaminants is of steadily increasing importance for the improvement of yields in microelectronic applications, the aim of one of the joint research activity within the European Integrated Activity of Excellence and Networking for Nano- and Micro-Electronics Analysis (ANNA, site: www.ANNA-i3.org) is the development and assessment of new methodolo¬gies and metrologies for the detection of low concentration inorganic contaminants in silicon and in novel materials. A main objective consist in the benchmarking of various analytical techniques avail¬able in the laboratories of the participating ANNA partners, including the improvement of the res¬pective detection limits as well as the quantitation reliablity of selected analytical techniques such as total-reflection x-ray fluorescence (TXRF) analysis.

Nickel Contamination in Silicon: Electrical Activity Study and Microscopy Analysis

It is well known that nickel is a very fast diffuser, with a significant diffusivity even at low temperatures. For this reason, nickel precipitates at the silicon surface during wafer cool down, and it is therefore extremely harmful for devices. For the same reason, nickel is often missed by recombination lifetime techniques and by DLTS. However, in this paper we show that nickel remains in the solid solution after RTP, so it can be detected by recombination lifetime techniques such as SPV and Elymat (based on photocurrent measurements). A quantification of nickel contamination based on lifetime measurements is provided. In addition, the results about the electrical activity of nickel in carrier recombination and generation are compared to TEM analyses of nickel precipitates at wafer surface. MCZ wafers were oxidized to grow 1000 A oxide and implanted with nickel at wafer backside, in order to have exact knowledge of the nickel dose in silicon, and after the implantation the samples were annealed by RTP at 1100 °C for 180 s. For a comparison, furnace annealed samples (800°C, 30 min, or 780°C, 1h 50 min + 800°C, 1h 40 min) were also studied. SPV, Elymat, DLTS, capacitance vs. voltage (C-V) and generation lifetime measurements were obtained. The wafers for C-V and generation lifetime measurements received an additional annealing at 430oC in N2/H2 to reduce interface states. After RTP, the nickel implanted region is clearly identified by SPV. The measured carrier diffusion length decreases as 1/QNi (see fig.1), where QNi is the implanted nickel dose, showing that a constant fraction of the implanted dose is active for recombination in the silicon volume after RTP. No modifications of the SPV data were observed upon illumination in the nickelimplanted region, opposite to what happens for iron and for cobalt. Similar results are also obtained by Elymat measurements. The injection level spectroscopy features of the Elymat technique were used to obtain a fingerprint of the nickel activity as a recombination center (fig.2). No relevant change was detected upon baking at 250 °C for 90 min, indicating that nickel in the bulk is in some stable state. Opposite to what observed after RTP treatments, no effect of nickel on bulk recombination lifetime was detected after a furnace treatment, whether by SPV or by Elymat. The Elymat technique offers the possibility to estimate the contaminant segregated at the wafer surface through measurements of surface recombination velocity. Nickel segregation at wafer surface was actually detected by this method (fig.3). It is interesting to remark that the dependence of the surface recombination velocity on the nickel dose is essentially independent of the thermal treatment at high nickel dose. On the contrary the results are strongly affected by the thermal treatment at low dose. Though nickel is partially in the silicon volume after RTP treatments, no DLTS signal could be revealed. This fact is probably due to a depletion of the near-surface region because of surface segregation even during the very fast RTP ramp. Generation lifetime measurements are also sensitive to nickel contamination, but the dependence of generation lifetime on dose is affected by the thermal treatments (whether furnace annealing or RTP) used for metal drivein. In the case of a furnace annealing generation lifetime is the most sensitive quantity to detect surface-segregated metals. TEM analyses were carried out to investigate the morphology of nickel precipitates at the silicon surface and to relate this morphology to the observed electrical properties of the precipitates. As an example, fig. 5 shows plan TEM images of nickel precipitates in furnace annealed wafers ((a) 800°C, 30 min, (b), 780°C, 1h 50 min + 800°C, 1h 40 min).

Reference Samples for Ultra Trace Analysis of Organic Compounds on Substrate Surfaces

Reference samples were produced for development, benchmarking and comparison of analytical techniques based on mass spectroscopy as TD-GCMS and TOF-SIMS and x-ray analysis as TXRF-NEXAFS. Organic contaminants representing plasticizers, disinfectants and flame retardants were chosen. The contaminants were selected with respect to reliable detection using the above analytical techniques. The stability of the reference samples produced with dethylphtalate, triclosane, and tetrabrombisphenol A on silicon stripes or wafers with a diameter of 200 mm was found to be approx. 10 days. The comparison of the techniques showed that the mass spectroscopy methods allowed reliable qualification of organic surface contamination. TD-GCMS quantifies and identifies the volatile organic compounds whereas TXRF quantifies the carbon contamination, especially the non-volatile, on sample surfaces.

Proximity Gettering of Slow Diffuser Contaminants

In this paper, we test proximity gettering layers obtained by carbon or silicon implantation for their efficiency in molybdenum and tungsten gettering. DLTS was used to measure the impurity concentration in the solid solution and so to evaluate gettering efficiency. It was found that carbon implantation is effective in capturing these impurities, whereas silicon implantation is not. Extended defects seem not to play an important role in gettering these impurities. In addition, gettering was found to be most effective at high impurity concentration.

Packaging Trends and Technology in Wireless and SSD Applications

Hand-handled, wireless world represents the most challenging environment for package technology where all the system performances must be densely stacked. Chip scale package (CSP) is the generic term for packages approaching chip size, and the fine-pitch versions of BGA have become the most widely used kind of CSP for system miniaturization. With the introduction of the new feature, where heterogeneous semiconductor devices are stacked together within a single package working at extremely high frequencies, the package design in terms of system simulation – mechanical, electrical, and thermal – is becoming one of most important development activities for delivering robust system solutions. In this chapter, the package historical trends against the system evolution will be discussed, analyzing the principal integration challenging. Among them, this chapter will focus on thin die thickness trend, taking into account the new process technology and the related impact on the device characteristics. This is considered as one of the most important back-end technologies, enabling the new era of the package integration and miniaturization. New interconnection technologies among dies will also be reviewed and discussed by deeply analyzing the features of through silicon via process. This process will allow the new interconnection scheme for microelectronics for the next decade.

Cobalt Contamination in Silicon

The properties of cobalt as a contaminant in p-type silicon are studied by using cobaltimplanted wafers annealed by RTP or by RTP plus a low temperature furnace annealing. It is shown that after RTP most cobalt is under the form of CoB pairs. A quantification of cobalt contamination is provided based upon SPV measurements and optical pair dissociation. However, this quantification fails in furnace-annealed wafers because of the formation of a different level. It is shown that the CoB level is located near the band edges, whereas the level formed upon a low temperature furnace annealing is located near midgap. Besides, when the cobalt concentration is high enough a small fraction of cobalt is in a level different from the CoB pair even in RTP samples. This level can probably be identified with a previously observed midgap level. It is suggested that the same level is formed in RTP plus low temperature furnace annealed samples and in high concentration RTP annealed samples, and that this level may consist in some cobalt agglomerate.

Analysis of Contaminated Oxide-Silicon Interfaces

Methods for the analysis of the oxide-silicon interface were compared for their ability to reveal metal segregation at the interface and organic contamination. The impact of these contaminations on surface recombination velocity measurements, on capacitance vs. voltage, conductance vs. voltage and capacitance vs. time measurements and on MOS-DLTS spectra was studied. Niobium-contaminated wafers were used as an example of metal surface segregation, because it was previously shown that niobium is prone to surface segregation. Interface state density measurements obtained by the conductance method showed a limited impact of niobium implantation. Vice versa significant effects were found in MOS-DLTS spectra. For what concerns organic contamination, MOS-DLTS showed the most significant effects from the point-of-view of the intrinsic properties of the silicon oxide - silicon interface, and GOI tests demonstrate a clear impact of the organic contamination on MOS capacitors oxide breakdown events.