Using re-association kinetics to identify impurities in p-type silicon

09/01/2004

A new approach to in-line monitoring of impurities that pair with boron in p-type silicon is based on recombination lifetime measurements performed by a commercially available surface-charge profiler tool. The method monitors the re-association kinetics of metastable complexes that are formed between boron and individual atoms of impurities — such as iron and hydrogen — in the device region, near the surface of silicon wafers. This article describes the concept and proposes in-process control applications based on minority carrier recombination-lifetime degradation characteristics of impurities during the re-association process.

Lifetime measurement technology is currently recognized for clean control of silicon wafers and device manufacturing processes [1–8]. Process defects, most commonly heavy metals and/or oxygen precipitates, can cause severe lifetime degradation in regions where they occur. A new technique for measuring recombination lifetime near the surface of silicon has been developed to identify contaminants in the critical device region of wafers. We used a standard surface charge profiler (SCP), which utilizes illumination-induced surface photovoltage (ac-SPV). In contrast to other SPV techniques that measure the lifetime in bulk silicon, the SCP tool is capable of measuring recombination lifetime within about 1µm of the surface.

A unique feature of the SCP tool, supplied by QC Solutions Inc., is the capability to follow transient surface effects to determine the state of equilibrium with an ambient. Another benefit of SCP is the routine application of a special rapid optical surface treatment (ROST), which uses a proprietary light source to produce a photo-thermal effect [9, P. Newcomb et al. in 8]. It is also believed that the computed outputs represent the properties of the surface that are not modified in the course of the measurement itself, due to application of the low-intensity light [E. Kamieniecki in 8].

The new method consists of monitoring the lifetime degradation kinetics — i.e., lifetime degradation in the course of a time delay after the heating or illumination of silicon. Contaminants are revealed as a characteristic property on a kinetics curve. Every contaminant "X" that pairs with boron and forms complexes should have its own characteristic time constant of dissociation/association with boron. The tracing of iron contaminants in p-type silicon illustrates the method. The monitoring of other contaminants, such as copper or hydrogen, is also possible with this method.

Dealing with iron

Iron, as a common contaminant, has been studied the most [1–15]. Iron in p-type silicon has a unique property of being in three forms:

• Fe-acceptor pair (most commonly FeB), which is stable at room temperature;
• interstitial Fe_i⁺, which forms upon the breaking of Fe-B pairs during annealing; and
• iron precipitates, which are dissolved at elevated temperatures; Fe_i⁺ can be converted into iron precipitates at temperatures somewhat above 150°C [12].

Upon cooling from high temperatures, iron tends to stay in solution, forming electrically active defects. This means that Fe_i⁺ acts as an effective recombination center and can be detected by lifetime degradation. Iron precipitates and Fe-B pairs have much less impact on lifetime.

The variation of carrier lifetime between the values t_dis (seconds after a thermal or optical excitation) and t_assoc (after Fe-B association completion) helps to determine the iron concentration by means of expressions 1 and 2:

NFe = 3.4×1013 (1/tdis - 1/tassoc)	(1)
NFe-B = 2.9×1013/tassoc	(2)

This coefficient includes the dependence of the lifetime on the impurity energy level and may vary depending on the wafer resistivity/doping.

While storing at room temperature, the concentration of Fe-B pairs increases and interstitial iron density decreases. The recombination lifetime must follow the same time dependence. The Fe-B association should be completed after storage of wafers for 24 hrs at room temperature or by annealing at 80°C for 20 min [13].

The pairing reaction is diffusion-limited [10, 15] and highly dependent on the boron concentration, since this determines the mean distance between iron and the substitutional boron atoms, which are immobile at room temperature. At elevated temperatures, the pairing reaction is accelerated but the thermal equilibrium between the pairs and the isolated ions is shifted vs. that for isolated ions. At about 100°C, for example, the paired fraction is only 60% of the total ion amount [15].

The measurement of Fe contamination is complicated by several factors that impede the assessment of the Fe contributor. Other elemental pairs, such as B-Cr, B-H, Cu_i-Cu_s, Fe-Al, and so on, can also dissociate/associate during Fe activation by heating/illumination. This may introduce ambiguity in lifetime interpretation, especially for low contamination levels. The measurement of iron concentration is also affected by the presence of oxygen precipitates [10]. The iron interstitial donor can be passivated in the presence of transition metal impurities Ag, Au, Pd, and Pt [10]. An even more complicated situation arises in the presence of silicon dioxide. The Fe ion, for example, may exchange with Si in the native oxide, resulting in a negative charge [4].

Other contaminant issues

The behavior and effects of Fe, Cu, and Cr in silicon differ greatly. Copper and copper-associated defects degrade the minority carrier lifetime more in n-type than in p-type silicon. However, a process similar to that of Fe-B in p-type silicon was proposed [16, 17]. Copper impurity significantly affects recombination lifetime in p-type silicon after activation/.dissociation of the Cu-Cu pairs.

The optical activation process dissociates Cu interstitial-Cu substitutional pairs, a weak recombination center in p-type silicon, and the copper forms extended substitutional defects, which have much greater recombination activity. Unlike Fe-B pairs, no recovery of lifetime was observed following such an activation procedure in copper-doped silicon, and the reduction in lifetime due to copper contamination in silicon is permanent [16, 17]. The difference in Cu and Fe recovery properties after activation can be used to differentiate iron impurities from copper.

Copper is a more complex issue than iron because of its high diffusivity, even at low temperature. Cu forms precipitate in the near-surface region and can be gettered easily on various bulk and back-surface gettering centers. The interstitial copper in p-silicon is much less dangerous as a lifetime killer than interstitial iron or Fe-B pairs until copper precipitates occur, which can drastically reduce the yield of devices even at low copper concentrations, which still significantly degrade the recombination lifetime in n-type silicon.

The Cr-B pairs are also separated by heat treatment. In contrast to the Fe-B pairs whose dissociation causes the lifetime to decrease, however, the Cr-B separation appears as an increase of lifetime value. This introduces a possibility for direct distinction of Cr from Fe contamination [T. Pavelka in 1].

Hydrogen is also a known "bulk-compensating" donor that is capable of deactivating the boron acceptor with the formation of a B^-H⁺ pair [18].

Results from tests

The SPV signal and, accordingly, the lifetime value, are associated with the total charges accumulated in the near-surface region. The negative space charge Q_sc is balanced by the positive surface charge Q_s, and is determined as follows:

Qsc = qNscWd

(3)

The minority carriers recombine not only with the surface charges but also with all recombination centers and traps present in the depletion and inversion layers. The lifetime measurement conforms to any redistribution of charges on the surface and in the subsurface.

After heating, however, a single lifetime value does not provide definitive information about contamination. It is impossible to distinguish whether carrier-lifetime degradation results from a single impurity or from different impurities, since the inverse total lifetime is the sum of the inverse partial lifetimes. The level of the total subsurface contamination can be evaluated by the value of lifetime measured immediately after heat excitation (ROST), whereas the type of impurities may be recognized only by examination of impurity behavior during the relaxation period (i.e., measuring the recombination lifetime as a function of time at room temperature).

Figure 1. The typical curves of lifetime degradation are shown as a function of time delay after a single heating by ROST treatment. Each curve represents a series of more than 10 similar tests.Click here to enlarge image

Figure 1 represents the recombination-lifetime degradation measured on the p-type wafers that have been implanted with iron of 1E11cm^-2 and 1E12cm^-2. The first data point is related to a single ROST measurement that is associated with heating the wafer to 210°C, maintaining temperature for 3 min, and then rapidly cooling to room temperature. All other data points were obtained at room temperature over time after heat treatment.

The initial steep degradation of lifetime can be explained by adsorption of negatively charged molecules at the surface, the formation of a double charge layer due to ambient humidity, and thereafter a deposition of heavier organic molecules. The reduction of positive charges on the surface leads to a slow-going decrease of W_d, Q_sc, and the lifetime values. Saturation may occur only after several days of storing wafers in the cleanroom ambient.

After the steep slope-down, the lifetime changes only a little for several hours. During this time, the interstitial Fe_i⁺ atoms gradually diffuse through the silicon, approaching nearer to B^- atoms and combining with them electrostatically. About 20 hours after initial excitation, there is an apparent lifetime increase due to the reduction of Fei+ recombination centers. The Fe-B association process is complete in 24–25 hours. Finally, a new steep lifetime decrease is associated with continuing surface degradation and may last several days until saturation.

It is worth noting that Q_sc and lifetime values are measured with an accuracy of <1% (repeatability is <10%) and even small variations in the curve cannot be explained by the measurement noise.

Figure 2. The arrowed lines show the sequence of lifetime values that were measured over time as follows: A) period of 10 hrs starting from excitation, B) next period from 10??25 hrs, and C) subsequent 26??35 hrs delay.Click here to enlarge image

To confirm the dissociation/association mechanism of this method, the recombination lifetime is plotted against the surface charge (Fig. 2). The Q_sc = Q_s is calculated as a product of the dopant concentration in silicon and the depletion width (Eqn. 3). The SCP tool at the point of maximum depletion initiated by ROST treatment determines a "real" doping concentration. The W_d was measured as is at every particular time delay.

The series A and C in Fig. 2 should be associated with the lifetime decrease due to the continuing surface degradation, as was stated previously. The series B manifests the lifetime increase that is the result of the continuing process of Fe-B association. This pattern was built for N_sc = 1E15cm^-3 and can be quantitatively changed for other doping levels.

The line B in Fig. 2 can be used to empirically establish a coefficient in Eqn. 2, which differs in different references [3, 11, A. Kempf in 1]. This coefficient is needed for a quantitative assessment of iron concentration at the near-surface area. Therefore, in the case of iron, the typical lifetime peak in the kinetics curves always appears in 20–25 hrs after excitation (ROST), as shown in Fig. 1. The lifetime peak forms the signature of iron contamination in the wafer.

From a practical standpoint, it is not necessary to wait multiple hours to identify the existence of a particular contaminant. A low-temperature anneal brings about the same effect of association during a shorter period than occurs at room temperature over a longer period. Fig. 3 shows the evolution of the recombination lifetime with the annealing time at 80°C. The result that the FeB association takes just about 20 min at 80°C annealing is in agreement with other research [13].

Figure 3. At 80??C, the dashed lines emphasize the transition from the beginning to the completion of Fe-B association, which happens in about 23 min annealing after initial ROST heating.Click here to enlarge image

Thus, tracking lifetime recovery values after annealing, i.e., after the pairing dissociation process, serves as a "time-spectroscopic" method that allows contaminant identification (iron in this particular case). The characteristic attributes of the carrier-lifetime recovery curve that follows the annealing — slope, peak occurrence, level of saturation, and curve regeneration after repeatable heat treatment — help to distinguish between different contaminants.

In the case of iron impurity, a rough assessment of the contamination level is possible while measuring the single lifetime value immediately after a heat/illumination treatment. Comparing the first data points in Fig. 1, the higher the concentration of impurity, the lower the initial lifetime value. A single ROST measurement of lifetime provides a reliable assessment of the iron contamination to distinguish between levels of 1E11cm^-3 and 1E12cm^-3.

One lifetime measurement may not necessarily indicate the contamination occurrence and the type of impurity, however; for example, a single ROST measurement does not show the lifetime degradation for hydrogen- and copper-contaminated wafers. The follow-up "time delay" test is needed to clarify the contamination type.

A similar approach to study the hydrogen behavior in the near surface of p-type silicon shows a higher recombination rate and the characteristic peak that occurs about 7.5 hrs after heating/dissociation. Our analysis showed that hydrogen atoms approach boron in 6 hrs and complete B-H pairing in 8 hrs after heating.

One of the features of SCP is that it measures average net doping concentration throughout the entire depletion layer. The depletion layer charges per unit area Q_sc includes all electrically active centers, and Eqn. 3 may be expanded as follows:

Qsc = q(NB + NFe + NH + ...) Wd

(4)

where N is the concentration of boron dopant, and, respectively, of the interstitial iron, of the hydrogen, and so forth. With the presence of embedded contaminants, the concentration of boron acceptors N_B does not remain constant but is consumed by other terms in Eqn. 4. After heat excitation, the numbers of electrically active impurities, such as Fe_i⁺, H⁺, etc., deactivate the boron acceptor in silicon so that the dopant concentration may be substantially reduced. A contribution of other centers to Q_sc can reach several percent for the given N_sc and N_Fe as shown in Fig. 2. Plotting Q_sc vs. time delay generally repeats the lifetime dependence as in Fig. 1.

Applications and conclusion

The re-association kinetics approach has demonstrated a capability that is well suited for monitoring process-induced contamination. Potential applications include tracking the quality of incoming wafers; evaluating the surface stability of bare silicon wafers after exposure to the cleanroom environment, while stored in boxes, etc.; in-line monitoring of contaminants in oxidation/diffusion furnaces; qualifying wet clean processes; evaluating surface condition after etch and pre-oxidation processes; and controlling epi layer characteristics due to parasitic contribution from the substrate, epitaxial layer defects, and variations in doping concentration in the near-surface region.

Process-induced contaminants that are unstable in p-type silicon at elevated temperatures and tend to form complexes that affect their effectiveness as recombination centers are the best candidates to be readily recognized with the approach of association-related kinetics. Most metals and hydrogen have such properties.

Acknowledgments

The author is particularly indebted to Herb Robertson for the problem statement, helpful discussions, and continuous support, and wishes to thank Kathy Jenkins and Amber Ngo for their active contribution to experiments. In addition, the author would like to thank Dr. Dieter Schroder from Arizona State U. and Dr. Edward Tsidilkovski from QC Solutions Inc. for many helpful discussions.

References

1. Recombination Lifetime Measurements in Silicon, D.C. Gupta, et al., ed., STP:1340, pp. 1–390, 1998.
2. Lifetime Factors in Silicon, ASTM Spec. Techn. Publ., STP:712, pp. 1–250, 1980.
3. J. Lagowski, et al., "Non-Contact Mapping of Heavy Metal Contamination for Silicon IC Fabrication," Semicond. Sci. Technol., Vol. 7, pp. A185–A192, 1992.
4. P. Roman, et al., "Monitoring of Fe Contamination on Si Surfaces Using Non-Contact Surface Charge Profiler," Proc. Electrochem. Soc., Vol. 95-20, pp. 344–349.
5. F. Tardif, et al., "Monitoring of Metallic Contamination by Direct and Indirect Analytical Methods," Proc. ALTECH 95 Conf.
6. A. Danel, et al., "Sensitivity of the SCP Method For Some Monitoring Applications," Proc. Electrochem. Soc., Vol. 97-22, pp. 394–403.
7. A. Danel, "Monitoring of Noble Metals in HF Based Chemistries by µ-PCD, SPV, SCI and SCP," Proc. Fifth Intern. Symp. on Cleaning Technology in Semiconductor Device Manufacturing, 1997.
8. Proc. Electrochem. Soc., Vol. 99-16.
9. F. Tardif, et al., Solid State Phenom., Vol. 65-66, 1999.
10. A.A. Istratov, et al., "Iron and its Complexes in Silicon," Appl. Phys. A 69, pp. 13–44, 1999.
11. G. Zoth, et al., "A Fast, Preparation Free Method to Detect Iron in Silicon," J. Appl. Phys., Vol. 67(11), pp. 6764–6771, June 1990.
12. X. Gao, et al., "Annealing and Profile of Interstitial Iron in Boron-Doped Silicon," Appl. Phys. Lett., Vol. 59, pp. 2133–2135, Oct. 1991.
13. J. Lagowski, et al., "Iron Detection in the Part Per Quadrillion Range in Silicon Using SPV and Photodissociation of Iron-Boron Pairs," Appl. Phys. Lett., Vol. 63(22), pp. 3043–3045, Nov. 1993.
14. K. Graff, et al., "The Properties of Iron in Silicon," J. Electrochem. Soc.: Solid-State Science and Technol., Vol. 128, No. 3, pp. 669–674, 1981.
15. K. Graff, Metal Impurities in Silicon-Device Fabrication, 2nd revised edition, Springer-Verlag, 2000.
16. D.A. Ramappa, et al., "Quantitative Analysis of Copper Contamination in Silicon by SPV Minority Carrier Lifetime Analysis," Characterization and Metrology for ULSI Technology Intern. Conf., pp. 215–219, 2000.
17. W.B. Henley, et al., "Detection of Copper Contamination in Silicon by Surface Photovoltage Length Measurement," Appl. Phys. Lett., Vol. 74(2), pp. 278–280, Jan. 1999.
18. C.-T. Sah, et al., "Deactivation of the Boron Acceptor in Silicon by Hydrogen," Appl. Phys. Lett., Vol. 43(2), pp. 204–206, July 1983.

Yuri Sokolov is a staff engineer at Fairchild Semiconductor Corp., 3333 West 9000 South, West Jordan, Utah 84088; ph 801/562-7158, e-mail [email protected].