Application of titanium silicide as an interconnect in deep submicron integrated chip manufacturing 2

Spike anneals are studied in three areas, namely, different spike temperatures with a fixed soak time at a lower temperature, a fixed spike temperature with different soak times at a low

Chapter 5 Study of Titanium Silicide Formation using Spike

Anneals

5.1 Introduction

To extend the usefulness of TiSi2, integrated chip manufacturers employ either

a pre-amorphizing implant (PAI) or an implant through metal (ITM) [5-8] technique to enhance the formation of the desired C54-TiSi2 phase However, both introduce damage to the substrate, resulting in device degradation like junction leakage and device driving current loss [9-11] Recently, refractory metals like Mo have been implanted in minute quantities into the Ti film [50] These act as nucleation sites, which results in smaller C49-TiSi2 grains However, introducing a new metal species into the IC manufacturing line may not always be desirable The effect of possible contamination to the existing process from the new metal species is unknown

Although TiSi2 can be replaced by alternative metal silicides namely, Co or Ni

as both metal silicides are proven at deep sub-micrometer [51], the change in silicide technology can be costly

In this chapter, an alternative option designed to extend the usefulness of TiSi2

is presented Spike anneals are studied in three areas, namely, different spike temperatures with a fixed soak time at a lower temperature, a fixed spike temperature with different soak times at a lower temperature, and different spike temperatures without any soak time

5.2 Experimental Procedure

Experiments were performed on 8” (200mm) wafers from 0.35 µm CMOS logic technology A 450/150 Å Ti/TiN stack is first deposited using PVD on patterned poly Si lines with SiO2 spacers These undoped poly Si lines are on the field SiO2 The deposition was carried out within 4 hours after a brief dip in a dilute HF solution that removed any native SiO2 Ti silicide was then formed in two RTA steps

All RTAs were carried out using an AST-SHS 3000 system with symmetrical double-sided heating on the front and backsides of the wafer During heating, the wafers spun at approximately 85 rpm in the chamber for better uniformity For better peak temperature control and repeatability, the heating rate in all the spike anneals was controlled at 125 Ks-1 ramp up, this is well within the maximum possible ramp up rate

of 300 Ks-1 Three peak temperatures for the spike anneals were used, namely 800, 850 and 900 °C Table 5-1 gives the details of the RTA1 conditions used in this study Specimen A, the control specimen, was prepared using the standard RTA1 recipe, 720°C for 30s The ramp-up rate for specimen A was 35 Ks-1 Except for specimen A, the soak times during RTA1 were reduced to compensate for the extra thermal budget

as a result of an initial spike anneal It should be noted that the total time above 720°C during each spike was approximately 10 s Kelvin and Serpentine Comb structures were used to measure the sheet resistance and gate-to-source/drain leakage current for various line widths For sheet resistance measurements, the line widths were varied from 1.0 to 0.275 µm

Table 5-1 Details of RTA1

Transmission Electron Microscopy (TEM) characterizations were performed on specimens G and H to observe the integrity of the Ti silicide on the 0.25 µm serpentine structure

5.3 Results

Different spike temperatures with a fixed soak time at a lower temperature

Figures 5-1(a) and 5-1(b) respectively show the sheet resistance of specimens

shown in Figures 5-1(a) and 5-1(b) for comparison purposes From Figure 5-1(b), the specimens that experienced a spike during RTA1 registered a lower sheet resistance as compared to specimen A Specimens with a higher initial spike temperature had lower sheet resistances The difference in the sheet resistance after RTA2 increases with decreasing line width suggesting that the initial high temperature spike reduces the area-dependence of the C49-to-C54 titanium silicide phase transformation When in contact with Si surface, the deposited 45nm of Ti likely to be fully consumed during reaction for all specimens

The sheet resistances of the specimen D after etchback and RTA2 do not show much difference from Figures 5-1(a) and 5-1(b), indicating that the dominant titanium silicide phase of specimen D after RTA1 is C54 Likewise for specimen C, C54-TiSi2

is the dominant phase after RTA1 As for specimen B, high sheet resistance and high standard deviation shown in Figure 5-1(a) suggests a mixed C49/C54 TiSi2 phase

Figure 5-1(a) Comparing sheet resistance after etchback for specimens with different spike temperatures plus a fixed soak time at a lower temperature during RTA1

Figure 5-1(b) Comparing sheet resistance after RTA2 for specimens with different spike temperatures plus a fixed soak time at a lower temperature during RTA1

It is interesting to note that the sheet resistance of specimen D remains low even at small linewidths This indicates that the high temperature spike at 900°C did not result in agglomeration of the thin silicide film

A fixed spike temperature with different soak times at a lower temperature

Figures 5-2(a) and 5-2(b) show the sheet resistance after etchback and RTA2 respectively For RTA1, specimens C, E and G experienced a spike at 850°C followed

by a lower temperature anneal at 720°C for 20, 10 and 0s, respectively It is noted from Figure 5-2(a) that except for specimen A, the formation of the C54-TiSi2 phase in specimens C, E and G after RTA1 is evident Even when the sheet resistance is somewhat higher at the smaller line widths, it is accompanied by a high standard

in Figure 5-2(a), the sheet resistance remains around 10 ohm indicating a dominance of C49-TiSi2 phase

Figure 5-2(a) Comparing sheet resistance after etchback for specimens with a fixed spike temperature plus different soak times at a lower temperature during RTA1

Figure 5-2(b) Comparing sheet resistance after the RTA2 treatment for specimens with a fixed spike temperature plus different soak times at a lower temperature during RTA1

From Figure 5-2(b), the increase in the sheet resistance with decreasing line widths for specimen A can be observed, indicating the area-dependency of the C49-to-C54 TiSi2 phase transformation However, the specimens C, E and G show low sheet resistance down to a line width of 3.0 µm At a line width of 0.275 µm, the specimen C reveals a higher sheet resistance of 4.8 ohm whereas the resistance of both specimens

E and G remains low Moreover, as shown in Figure 5-2(b), the standard deviation for the specimen C at a line width of 0.275 µm is much larger, indicating an incomplete phase transformation

Different spike temperatures without any soak time

Figures 5-3(a) and 5-3(b) show the sheet resistances of the specimens A, F, G and H after etchback and RTA2, respectively Although the specimens F, G and H did not receive an isothermal soak, the duration at which the temperature of these specimens was above 720 °C was approximately 10 s

Figure 5-3(a) Comparing sheet resistance after etchback for specimens with different spike temperatures without any soak time during RTA1

Figure 5-3(b) Comparing sheet resistance after RTA2 treatment for specimens with different spike temperatures without any soak time during RTA1

Figures 5-3(a) and 5-3(b) show the sheet resistance of the specimen H to be 2 ohm down to 0.275 µm Hence, C54-TiSi2 is clearly the dominant phase for the specimen H after etchback and RTA2 For the specimens F and G after etchback, Figure 5-3(a) suggests the presence of a mixed C49/C54 TiSi2 structure This is evident both from the sheet resistance measurements and also from the high standard deviation

It is interesting to note that the specimen G shows a low sheet resistance for larger line width after etchback (Figure 5-3(a)), suggesting the dominance of C54-TiSi2 In Figure 5-3(b) all specimens, except the specimen H, show an area dependency of C49-to-C54 TiSi2 phase transformation However, the area-dependency effect is somewhat delayed for the specimens F and G compared to the specimen A At

a line width equal to 0.275 µm as shown in Figure 5-3(b), the specimen G has a sheet resistance of 2.37 ohm compared to 3.40 ohm for specimen F and G This suggests a higher spike temperature is more resistant to the area-dependency of the phase transformation

5.3.2 Gate-to-source/drain leakage current

Measurements were taken on serpentine comb structures with line widths 0.35, 0.33, 0.30, 0.28 and 0.25 µm with a total length of 8400 µm and a 0.7 µm pitch The applied voltage across gate-to-source/drain was swept from –5 to +5 V Should there

be excessive Si diffusion over the spacers leading to bridging, the profile of the measured current would indicate a resistive (ohmic) connection

Gate to source/drain leakage current for 0.25 µm poly serpertine comb

-1.00E-09 -8.00E-10 -6.00E-10 -4.00E-10 -2.00E-10 0.00E+00 2.00E-10 4.00E-10 6.00E-10 8.00E-10 1.00E-09 1.20E-09

Figure 5-4(b) Gate-to-source/drain leakage current versus applied voltage on a line width equal 0.25 µm taken from specimen D after the RTA2 treatment

Figures 5-4(a) and 5-4(b) show the leakage current versus applied voltage on line width equals 0.25 µm taken from specimens A and D respectively after etchback From the two Figures, it is evident that the leakage current of specimen D (Figure 5-4(b)) is many orders of magnitudes larger than that of the specimen A (Figure 5-4(a)) The linear dependence in Figure 5-4(b) also indicates a resistive contact across gate-to-source/drain, thus bridging occurred on specimen D, whereas for specimen A, bridging did not occur

Gate to source/drain leakage current for 0.25 µm poly

serpertine comb

1.00E-11 1.00E-10 1.00E-09 1.00E-08 1.00E-07 1.00E-06 1.00E-05 1.00E-04 1.00E-03 1.00E-02

G: 850C spike H: 900C spike

is clear that bridging occurs when the RTA1 spike temperature reaches 900°C It is also observed that the leakage currents for specimens C, E and G are higher at a low applied voltage (< 1.0 V) as compared to specimen A On the other hand, the leakage currents for specimens B and F are similar to that of specimen A These results

indicate that the leakage currents of specimens B and F, which had a spike temperature

of 800 °C during RTA1, are the same as that of specimen A For specimens C, E and

G which have a spike temperature of 850°C during RTA1, the leakage currents are slightly higher as compared to that of specimen A

5.3.3 Microscopic characteristics

Figures 5-6 and 5-7 show TEM images of specimens G and H respectively after RTA2 Both Figures show that the TiSi2 formed on the poly Si gate had a smooth interface No signs of agglomeration are observed in either Figures, especially for specimen H, which had received a spike temperature of 900 °C It is noted that both Figures show some formation of TiSi2 over the spacers near the top, with specimen H showing a larger amount of TiSi2 forming over its spacer region

TiSi 2

specimen G

grains create more triple grain boundary junctions, which act as nucleation sites for the thin film C49-to-C54 TiSi2 phase transformation

Other than increasing the nucleation rate of the C49-TiSi2 phase, the results also suggests other factors arising from the spike anneal which contribute to enhance the C49-to-C54 TiSi2 phase transformation at smaller line widths

Table 5-2 Sheet Resistance after etchback for varying spike anneal temperatures

Specimens Spike temperature

[2], this result indicates the presence of a dominance of, if not total, C49-TiSi2 phase in specimen A after RTA1 This is not the case for specimens after a spike anneal For specimens with a spike anneal at temperatures of

800, 850 and 900 °C in RTA1, the sheet resistance measured after RTA1 ranges from 2.24-12.25, 1.83-10.69, 1.74-3.75 ohm respectively This can only be due to the presence of a mixture of the C49/C54-TiSi2 phases after RTA1 as the thickness of the silicide film is similar On larger line widths, for example 1 µm, a complete C49-to-C54 TiSi2 phase transformation occurs after the RTA2 treatment Furthermore, the sheet resistances of all specimens at linewidth equals 1 µm are comparable, indicating that the thickness of their silicide films is similar Hence it is believed that the lower sheet resistances of specimens other than specimen A after the RTA1 treatment are not due to thicker silicide formation but rather to the presence of the C54-TiSi2 phase; with higher spike anneal temperatures resulting in larger amounts of C54-TiSi2

After the formation of the C54-TiSi2 phase during the spike anneal, the growth

of C54-TiSi2 grains are thermodynamically more favorable over that of the C49-TiSi2

phase However, this is not to claim that C49-TiSi2 does not continue to form after the spike anneal Even if there was insufficient thermal budget after the spike anneal to convert all the C49-TiSi2, the presence of these C54-TiSi2 grains after the RTA1 treatment would then act as seeds for the C49-to-C54 TiSi2 phase transformation during the RTA2 treatment This is observed in Figures 5-2(a) and (b) The specimens with the same spike anneal temperature of 850 °C but varying soak times at 720 °C show varying sheet resistances after etchback However, after the RTA2 treatment these specimens show very similar results, with the exception of specimen C with a line width of 0.275 µm

It is noted from Figures 5-1 and 5-3, that a high spike anneal temperature of

900 °C did not result in agglomeration of the thin TiSi2 film, instead the sheet resistance measured was lower than that for specimens spike annealed at 800 and 850

°C The absence of agglomeration after a spike anneal at 900 °C is confirmed by the TEM picture in Figure 5-7 This observation contrasts with previous findings that claimed a high temperature treatment at 900 °C resulted in agglomeration of thin TiSi2

films [34] It is believed that two factors could be attributed to preventing agglomeration from occurring during the RTA1 treatment Firstly, in all the RTA1 treatments, there was no thermal annealing before the spike anneal This means that prior to the peak of the spike anneal, a minimal of TiSi2 was formed Since the rapid thermal processing chamber did not have any forced cooling other than the constant N2

flow, the ramp-up rate (of 125 Ks-1) is much higher than the ramp-down rate The thermal budget on the ramp-up would be lower than that of the ramp-down Hence, it

is reasonable to say that the formation of the TiSi2 film did not stop at the peak of the

spike anneal Even if the high spike anneal temperature had resulted in agglomeration, the continuing growth of the TiSi2 film during ramp-down would have filled up any voids formed due to agglomeration

Secondly, during the RTA1 treatment the TiN cap layer has not been etched away, and therefore would have acted as a compressive layer on any TiSi2 formed This is because both the C54-TiSi2 phase (14.3 x 10-6 K-1) and the C49-TiSi2 phase (12.0 x 10-6 K-1)[52] have higher thermal expansion coefficients than the TiN phase(9.35 x 10-6 K-1)[53] During heating, TiSi2 expands more than TiN, and therefore gets compressed by the TiN The presence of a compressive capping layer would then help prevent agglomerations from occurring in the TiSi2 regions

In a normal salicide process, the low RTA1 temperature (typically at 700~750

°C) treatment forms the C49-TiSi2 phase while a high RTA2 temperature (typically at

850 °C) treatment converts the C49-TiSi2 formed to the desired C54-TiSi2 phase The main reason for the 2 steps of the RTA process is to prevent gate-to-source/drain leakage due to excessive Si diffusion over the spacers during the formation of the TiSi2phase Si was previously identified to be the main diffusing species for TiSi2 formation [43] Similarly a high temperature segment of 800 °C or higher due to the spike anneals in the RTA1 treatment would also lead to a higher rate of Si diffusion Having

a high temperature segment at the beginning of the RTA1 treatment was chosen so that much of the Ti remained unreacted during the high rate of Si diffusion In the present study, specimens with spike anneal temperatures of 800 and 850 °C did not result in excessive leakage current although specimens with a spike anneal temperature of 900

°C during the RTA1 treatment clearly show bridging across gate-to-source/drain For a