I-V and FPP measurements

Fig. 4.3 shows the I-V measurement results of samples fabricated in section 4.2.1. The forward current (IF) is defined at sample with 1V forward bias voltage, and the reverse (I_R) current is defined at sample with 2V reverse voltage. From the I_F/I_R ratio, it seems that the sample with 2^nd RTA 550℃ 60s has the best performance at this measurement. The better performance of 2^nd RTA 550℃ 60s than those samples with lower 2^nd RTA temperatures can be contributed to two reasons. First, the activation process will also re-crystallize the amorphous layer at silicide/silicon interface caused by silicide formation process. With higher thermal budget, the better

re-crystallization interface can be formed. Secondly, as shown in Fig. 4.1 and Fig. 4.2, the P⁺/N junction position of 2^nd RTA 550℃ 60s sample might be beyond the point 2 appeared in Fig. 4.1, where less doping defects (unactivated dopant and silicon interstitials) are present. However, the I-V behaviour above 550℃is different to those observed about N⁺/P junction in Fig 3.6, in N⁺/P case, above 550℃, I_F/I_R ratios decreased with increasing 2^nd RTA temperature.

To explain the I-V behaviour above 550℃, some other information about the junction formation should be measured as the reference. In Fig. 4.4, the relative resistance of these samples (after NiSi removed) measured with FPP test structure is displayed. The resistance reflects the combination interactions of the doping activation level and the junction depth. All resistances are normalized to the lowest measured value (2^nd RTA 650 ℃ 60s). It exhibits that the resistance decreases gradually but has an abrupt decrease at 650℃ for N30 samples (treated with 2^nd RTA 30s) and at 600℃ for N60 samples (treated with 2^nd RTA 60s). In addition, the lowest resistance values measured in N30 and N60 samples are almost the same at the original data. This implies that above 2^nd RTA 650℃ 30s or 2^nd RTA 600℃ 60s, the bulk activation of boron is higher than the substrate doping density, and the p-type junction extended far into the substrate. As a result, the resistances measured under this bulk activation [7] condition are very similar. Bulk activation behaviour dominates the P⁺/N junction formation at high thermal budget conditions. This explains the difference of the N⁺/P and P⁺/N junction’s I-V behaviours above 550℃.

In N⁺/P case, the performance of junction formed at high thermal budget condition is dominated by defect formation or dopant deactivation, but for P⁺/N case, bulk activation behaviour dominates.

38

The low IF/IR ratio of 600℃ 2^nd RTA sample (N600) in this study is due to the low forward current density measured at VA = 1V. Fig. 4.5 shows the J-V curve of some samples treated at 2^nd RTA 550℃ 60s (N550-1), 600℃ 60 s (N600-1), and 650

℃ 60 s (N650-1). The ideality factors of these three samples are all round 1.02. The low J_A of 1 is explained as follows: the law of junction is violated first of N600-1, which implies the voltage drop in the bulk regions occurring at lowest forward voltage in N600-1 sample. This means that the built-in voltage (Vbi) of the N600-1 sample is lowest among these samples. This observation confirmed the measured C-V data in Fig. 4.2, the N_eff is lowest at 2^nd RTA 600℃ 60s sample.

Besides, from Fig. 4.5, it is showed that J-V behaviour of N550-1 sample is different to the others. This doubly verified that higher than 2^nd RTA 550℃ 60s activation condition, the junction formation mechanism is much different. The N600 and N650 samples have lower leakage current at low reverse biased voltage due to lower defect densities presented at the junction depletion region. Bulk activation behavior extended the P/N junction far from the implantation caused defect region (i.e.

Peak 2 in Fig. 4.1). Although the N600 and N650 samples take the advantage of the good electrical properties, they are not met the shallow junction requirement.

For more comprehensive understanding the P⁺/N junction formation process, comparing with N⁺/P junction formation in chapter 3 and the silicide/silicon interface formation processes in chapter 2 is needed. First, the boron activation densities are about 10¹⁷ to 10¹⁸ cm^-3 in this study, however, for the N⁺/P case, phosphorous could achieve 10¹⁹ to 10²¹ cm^-3 doping activation level depending on different process conditions (Fig. 3.2). On the other side, boron activation densities near the silicide/silicon surface could also achieve 10¹⁹ to 10²¹ cm^-3 doping activation level as shown in Fig 2.2 [8]. These two observations suggested that the activation abilities for

boron and phosphorous are similar in metal assisted re-crystallize region (near the silicide/silicon interface). But a distance from the M/S interface, near the projection range located in silicon side in this study, where activation mainly correspond to the SPER [9] technique only, the dopant activation ability (10¹⁷ to 10¹⁸ cm^-3) is much lower than which with the metal assisted activation region (10¹⁹ to 10²¹ cm^-3). In short, metal assisted activation is more effective than SPER region at the temperatures below 600℃.

4.4 Conclusions

From the experiment results discussed in section 4.3.1 and 4.3.2, the activation of P⁺/N junction formed with IIS method could be divided into three different mechanisms. First, near the silicide/silicon interface, metal assisted activation is presented, and secondly, at the heavily doped region where SPER caused activation behavior is observed. Finally, with higher activation temperature, above 600℃ in this study, bulk activation dominates the junction’s position and its behavior. Boron could be activated at an effective doping density about 7x10¹⁷ cm^-3 extracted from C-V measurement under 2^nd RTA 450℃ 60s with the SPER caused activation mechanism.

Among all process conditions in this work, samples treated at 2^nd RTA 550℃ 60s have the best junction electrical characteristics before the bulk activation behavior dominates the device’s properties.

40

References

[1] C.S. Kang, H.J. Cho, R. Choi, Y.H. Kim, C.Y. Kang, S.J. Rhee, C. Choi, M.S. Akbar, and J.C. Lee:

Electron Devices, IEEE Transactions on, Vol. 51, issue 2, pp. 220, (2004).

[2] F.C. Shone, K.C. Saraswat, and J.D. Plummer, IEDM Tech. Dig., pp. 407, (1985).

[3] C.C Wang, Y.K Wu, W.H Wu and M.C Chen: Jpn. J. Appl. Phys. Vol.44, pp. 108, (2005).

[4] C.P Lin; Y.H Hsiao, and B.Y Tsui: Electron Devices, IEEE Transactions on, Vol. 53, issue 12, pp.

3086, (2006).

[5] K.M. Chang, J.H. Lin, and C.H. Yang: Applied Surface Science, Vol. 154, pp. 6155, (2008).

[6] C. Y. Ting and S.S Iyer: Proc. 5^th IEEE Int. VLSI Multilevel Interconnection Conf., pp.307, (1995).

[7] R. Lindsay, K. Henson, W. Vandervorst, and K. Maex, B. J. Pawlak, R. Duffy, R. Surdeanu, and P.

Stolk, J. A. Kittl, S. Giangrandi, X. Pages and K. van der Jeugd: J. Vac. Sci. Technol. B, Vol. 22, issue 1, pp. 306-311, (2004).

[8] K.M. Chang, J.H. Lin, and C.Y. Sun: Applied Surface Science, Vol. 154, pp. 6151, (2008).

[9] M.J.P. Hopstakena, Y. Tamminga, M.A. Verheijen, R. Duffy, V.C. Venezia, A. Heringa: Applied Surface Science, Vol. 231-232, pp. 688, (2004)

0 20 40 60 80 100 1E19

1E20 1E21 1E22

Boron Concentration (cm-3 )

Depth (nm) 1

2

With 2^nd RTA 650^oC 60s

Fig. 4.1. SIMS profile of boron concentration with 2^nd RTA 650℃ 60s

42

400 450 500 550 600 650 10¹⁶

10¹⁷ 10¹⁸

7x10¹⁴ 8x10¹⁴ 9x10¹⁴ 1x10¹⁵

Boron Concentration (cm-3 )

Temperature (^oC) Boron Concentration

Substrate Concentration (cm-3 ) Substrate Concentration

Fig. 4.2. Boron and substrate concentrations extracted with C-V measurement

400 450 500 550 600 650 10⁵

10⁶ 10⁷ 10⁸

I F/I R Ratio (A/A)

2^nd RTA Temperature (^oC) IF / I

R Ratio where I

F is measured at V

F=1V and I_R is measured at V_R= 2V

Fig. 4.3. IF/IR ratio with 2^nd RTA 60s at different temperatures

44

Fig. 4.4. Relative resistance ratio measured with FPP test structures at different 2^nd RTA conditions

400 450 500 550 600 650

1 10 100

R trend (ohm/ohm) (normalized with N60 650 C)

2^nd RTA Temperature (^oC) 2^nd RTA 30s (N30)

2^nd RTA 60s (N60)

-2 -1 0 1 2 10^-10

10^-9 10^-8 10^-7 10^-6 10^-5 10^-4 10^-3 10^-2 10^-1 10⁰ 10¹

N550-1 N600-1 N650-1

J A (A/cm2 )

VA (V)

Fig. 4.5. The J-V characteristics of N550-1, N600-1, and N650-1 samples

46

Chapter 5

E LECTRICAL P ROPERTIES OF S HALLOW J UNCTION

5.1 Brief Introduction and Design Idea

Schottky-liked PN junction is adapted to some of the next generation devices, such as DSSMOSFET reported by Toshiba [1], IBM [2], TSMC [3] and ASM [4].

The devices reported from reference 1 to reference 3 were fabricated with DSS method, and ASM adopted the IIS technology. Although each device processed with different conditions, all of these devices showed interesting electrical properties. Most devices had special junction behaviors characterized between traditional PN junction and MS junction. In this chapter, we want to devote our effort to study the shallow junction behavior.

Based on the understanding of the junction formation mechanisms described in previous chapters, we know that if we want to fabricate a shallow junction using IIS method, we must activate the junction in the metal assist activation region [5].

And if we want to optimize the junction’s electrical properties, to lower the defect density near the junction position is required. There are some defects might be generated in the junction formation process, like implantation defects and metal pollution. Some people wondered if there exist deep level defects by metal pollution originated from the implantation process: implanted dosage collides with the metal atom in the silicide and let the metal ion penetrate into the junction region. There are some reports showed that the second doubt listed above is not observed using the DLTS (Deep-Level Transient Spectroscopy) measurement [6,7] after two-step RTA.

The best way to reduce the implantation defect is to lower the implantation energy.

Also, same as in chapter 2 and chapter 3, in this chapter we choose the low implantation energy to let the projection range falls in the silicide to minimize the implantation defects. In addition, since the un-activated dopant and the silicon interstitial replaced by the activated dopant are all playing the roles as the impurities in the junction, we lower the implantation dosages in order to lower the potential defects in the junction to improve the junction’s performance.

5.2 Experiments

Two-port vertical diode structure was adopted in this study. Two different junction types were studied, M/N⁺/P, and M/P⁺/N. The starting material was (100) n-type and p-n-type silicon wafers, after thermal oxidation and active regions definition, backside of the n-type wafer was implanted with phosphorous 5x10¹⁵ cm^-2 and backside of the p-type wafer was implanted with BF₂ 5x10¹⁵ cm^-2. All samples were then be annealed in furnace at 1000℃ for 30 minutes. TaN 100 nm was sputtered on as the back contact. After dipped dilute HF to remove the native oxide at the active region, 20nm nickel was deposited by e-gun system. All samples were then treated with 1^st RTA at 400℃ for 30s. After un-reacted Ni removed, n-type substrate samples were ion implanted with BF₂ 5x10¹³ cm^-2, and p-type substrate samples were ion implanted with phosphorous 5x10¹³ cm^-2. Finally, samples were treated with 2^nd RTA 60s at different temperatures from 450℃ to 600℃, 50℃ per step.

The major measurement methods used in this study were C-V, I-V, and I-V-T (I-V measurement in different temperatures is used to discuss the activation energy [6,8]) measurements. C-V measurements were processed with HP4284 at 50 kHz, and I-V measurements were measured with HP4156C.

48

5.3 Experiment Results

5.3.1 C-V measurement for M/P⁺/N and M/N⁺/P junctions

At this study, the implanted doses were one hundredth of that in chapter 3 and chapter 4. According to the previous knowledge, there are two possible results that might turn up in this study instinctively. First, since the implanted dose was lowered, the measured N_eff from the C-V data should not be higher than that obtained in chapter 3 and 4. On the other hand, from chapter 2, it was shown that even at very low implanted dosage (1x10¹³ cm^-2), dopant activation at the silicide/silicon interface was high and compatible to the measured data with high implant dosage case (5x10¹⁵ cm^-2, Chapter 3 and 4). Since junction fabricated in this study is supposed to have very shallow junction, as the result, the measured N_eff at this study should not be lower than what measured in chapter 2. Interestingly, the measured results in this study, as illustrated in Table 5.1, did not meet both intuitions just stated above.

Table 5.1 illustrates the N_eff extracted from C-V measurement results without considering bandgap narrowing effect. It seems that the Neff obtained by C-V method with different implanted dosages activated at the same thermal budget does not relate to their implanted dosages. This phenomenon could be explained by the junction activation behavior discussed in previous chapters. The Neff is not only represented the dosages but also the information of the junction position. The N_eff extracted from C-V measurement results adjusted with bandgap narrowing effect [9] is present in Table 5.2. It is interesting that for heavily boron doped samples, the higher the implanted dosages, the lower the Neff concentrations near the P⁺/N interface that we obtained. Since Boron has relative higher diffusion coefficient than phosphorous does [10,11] at the temperature lower than 600℃. The schematic graphs of boron and phosphorous with different implanted dosages diffused from silicide interface at same

thermal budget are shown in Fig. 5.1. At the same thermal treatment, heavily doped sample diffused deeper than lightly doped sample, and at the same thermal treatment and same doping density, boron doped sample has almost double diffused region than phosphorous doped sample. For samples implanted with phosphorous 5x10¹³ (5E13) and 5x10¹⁵ (5E15) cm^-2, it could be found, in table 5.1, the highest activation level of these two different implanted samples was both occurred when treated at 500℃. It means that under this thermal treatment, devices are at their own best doping activation ability. Since these two samples were treated at the same temperature and had the same thermal budget, if we assume that the activated regions were the same and the diffusion coefficient was not different too much [12], 5E15 implanted samples should have higher Neff than 5E13 implanted samples. But what is interesting is that the Neff measured for 5E13 implanted samples are higher than 5E15 implanted samples at this temperature. Does this imply that that the 5E13 implanted samples have higher activation ability? Thus, this conclusion is not correct because with higher implanted dosages the samples become more amorphous and should have better activation ability. So the most rational explanation is that: Comparing with lightly implanted samples treated with the same thermal budget, the heavily doped samples have a more wide activated region, a deeper junction. This statement is true limited for 500℃ annealed sample, since below this temperature, samples are not really at there most stable condition (i.e. activation is still not complete). And samples with higher temperature treatment, some of the deactivation behavior might occur. On the other hand, the behaviors of samples doped with three different BF₂ dosages are easy to be understood. Since boron is fast diffuser in silicon, the Neff measured with MS diode (1x10¹³ cm^-3, N_eff near the silicide surface) are higher than the PN diode device. (5x10¹³ cm^-3, Neff extracted in the silicon). And the activation mechanism of

50

the heavily doped one (5x10¹⁵ cm^-3, with the projection range inside the silicon) is most due to the SPER and bulk activation [5], hence, it has the lowest effective doping densities.

5.3.2 I-V measurement for M/P⁺/N and M/N⁺/P junctions

By the analysis stated in the 5.1, we know that lowering the implanted dosage should result in the lower defect densities in the junction (both MS and PN junctions).

For a general junction, the saturated leakage current I0 is given by [13]

)

The dominant term is the minority carrier densities in the lightly doped region, the substrate. Since the doping of the substrate is the same in our study, the devices with the lower defect density should have better performance in the leakage current.

The off state current densities (measured at reverse bias V_R = 2 volt) of 5x10¹³ cm^-2 implanted and 5x10¹⁵ cm^-2 implanted samples treated with different 2^nd RTA temperatures are demonstrated in Fig. 5.2. It shows that the 5x10¹³ cm^-2 implanted samples have higher reverse current densities than 5x10¹⁵ cm^-2 implanted samples for both phosphorous and boron doped samples. Comparing the on/off ratios to the 5x10¹⁵ cm^-2 implanted samples showed in chapter 3 and 4, the 5x10¹³ cm^-2 implanted samples also have worse performance. The Jon, Joff and on/of ratio of 5x10¹³ cm^-2 implanted samples are showed in Fig. 5.3. This negative behavior might due to the shallow heavily doped region (i.e. the P⁺ layer of the M/P⁺/N structure), not only play the role as the PN junction but also like the schottky barrier modulation layer of the silicide/silicon interface [1,14]. The schottky-like behavior could be observed from the I-V-T measurement at the junction with the reverse bias condition [8,15]. Fig. 5.4

shows the I-V-T measurement result of the 5x10¹³ cm^-2 implanted M/N⁺/P sample treated with 2^nd RTA at 450℃ for 60s. The slope of the line is about –6.63, which corresponds to the activation energy at 0.58 eV. The activation energies (it is used as schottky barrier height, SBH, in the following paragraphs) of the 5x10¹³ cm^-2 implanted samples extracted from I-V-T measurements are summarized in Fig. 5.5 (for M/P⁺/N samples) and Fig. 5.6 (for M/N⁺/P samples). Interestingly, from the I-V-T measurement, M/P⁺/N samples showed the similar SBH versus Bias dependence through different 2^nd RTA conditions, in contrast, M/N⁺/P did not. This might be due to the difference in diffusion ability of phosphorous and boron. As mentioned in 5.3.1, boron is faster diffuser than phosphorous and would make a relative deeper junction than phosphorous did. The deeper junction might be more insensitive to the variation of the junction depth. As a result, M/P⁺/N samples behave in a more similar way.

When reverse voltage is higher than 0.08V, the SBH seems to be decreased with the increasing reverse voltage bias (see Fig. 5.5). There are two common accepted theories used to explain the negative relationship between the SBH and the bias voltage. One is the image-charge-induced schottky barrier lowering [16], and the other is the inhomogeneous schottky barrier height induced electron dipole effect [8].

According to image-charge-induced lowering effect,

2

where ∆ is SBH lowering, and E is applied electric field. Fig. 5.7(a) is plotted as φ delta SBH versus V^1/2, where delta SBH, the value directly obtained from the I-V-T data subtracted by the highest measured value at each condition, is used to replace

φ

∆ , since ∆ is unknown. The result of the line fitting is not good, which implies φ

52

using image-charge-induced barrier lowering is not suitable to explain the data measured with VR higher than 0.08V alone. On the other hand, one expression to describe inhomogeneous schottky barrier height induced electron dipole effect is given by function 6 in reference 7.

) ] depletion width, and z is the depth from the MS interface. Ro is the radius of the circular patch that has a lower SBH. The effective barrier height (φ_eff) related to applied voltage is simplified as [8]:

3 shown in Fig. 5.7(b). Since the curvatures of the curves in Fig. 5.7(a) are positive and those in Fig. 5.7(b) are a little negative, it could be concluded that both schottky barrier lowering and inhomogeneous schottky barrier height induced electron dipole effects influence the behavior of the M/P⁺/N diodes described above.

Fig. 5.9 (M/N⁺/P) and Fig. 5.8 (M/P⁺/N) compare the samples with the V_bi extracted from C-V measurement, the SBH extracted from IVT measurement at 1.2V reverse bias, and the leakage current at 2V reverse bias at 25℃ under different 2^nd RTA treatments. There are three major factors could be found in these two figures.

First, in both Fig. 5.8 and Fig. 5.9, leakage currents are all negatively related to the SBH measured from IVT method (i.e. the higher the barrier height, the lower the leakage current). Secondly, in Fig. 5.8, it could be found that Vbi is positive related to SBH for the M/N⁺/P case. However, this phenomenon does not hold in the M/P⁺/N case shown in Fig. 5.9. It might be due to that phosphorous is less diffusive than

boron, and hence M/N⁺/P samples have shallower heavily doped region than M/N⁺/P samples do. This comes to the same observation in Fig. 5.5 and Fig. 5.6 just mentioned above. In the case of M/N⁺/P samples, the doping density in the PN junction interface, derived from V_bi, is much close to the real doping activation situation near the MS interface. Thirdly, referring to Fig. 5.8, it shows that the schottky barrier height will increase with the increasing dopant activation density of the N⁺ region in the M/N⁺/P samples. The N⁺ region plays a role like an interfacial layer that can modify the measured schottky barrier height.

The I-V curves at the forward biased region for the 5*10¹³ and 5*10¹⁵ cm^-3 implanted samples activated at different 2^nd RTA temperatures are summarized at figure 5.10. It could be seem that samples with low implant dosages have lower “on voltages”. Where samples with 5*10¹⁵ cm^-3 implant dosages are the samples fabricated in Chapter 3 and Chapter 4. Those “on voltages” lowered more significantly for M/P⁺/N samples than those of M/N⁺/P samples since M/P⁺/N samples with 5*10¹⁵ cm^-3 implant dosages have the deepest junction width.

5.4 Discussions

There are some interesting observations that should be given more attention.

5.4.1 Meaning of C-V measurement results

First of all, what needed to be made clear is the experiment results obtained by C-V measurement. Some research papers use the 1/C² –VR curve as a mean to obtain schottky barrier height of the MS interface and others treat it as a way to determine the built-in potential of the PN junction for the special device type that mentioned in section 5.2. The experiment results in this study showed that when the heavily doped

54

layer is very shallow, like the M/N⁺/P case, C-V measurement and I-V-T

layer is very shallow, like the M/N⁺/P case, C-V measurement and I-V-T