3.3.1 P
+N Ge Junction Characteristics
Figures 3.2(a) (b) display the effects of 300°C FGA on the junction characteristics of Ge
p-MOSFETs for the two samples. The reverse bias leakage current density (JR) at -2V for no passivation sample are 1.4×10-2 A/cm2 and 6.9×10-3 A/cm2, while the forward bias current at 2 V are 9.2×102 A/cm2 and 4.7×102 A/cm2, without and with FGA respectively. On/off ratio44
of 4.8 orders is achieved, but junction leakage seems unacceptably high compared with the Si device. We suppose that large value of JR in this case originated mainly from the contribution of the generation current, which is dependent on the level of substrate doping (1.5E14 cm-3), residual metal contamination after surface cleaning, bulk defects in the Ge substrate and the Dit at the SiO2/Ge interface. Therefore, the lower JR after FGA is attributed to the effective reduction of the defects. On the other hand, an additional 300°C 90 minutes thermal annealing is definitely to cause dopant diffusion, and lower doping concentration verified by series resistance of the junctions (Figure 3.3) may account for the slight decrease of the forward current (JF).
To compare junction characteristics between 500°C 10s GeO2 passivation and no passivation sample, junctions with GeO2 passivation showing lower JF and larger JR is inferred to correlate with lower concentration of the S/D region. Lower concentration causing larger series resistance degrades the forward current, while it widens the depletion width to induce more generation current simultaneously. However, we are still not clear about the extent of oxidation enhanced diffusion (OED) of B in Ge, and it requires being under further investigation. From the result of larger JR for GeO2 passivation sample, it can be mentioned in advance that off current of ID-VG and the gated-diode current biased in accumulation are both larger for this sample.
3.3.2 Basic Device Characteristics
Several studies have demonstrated that FGA or pure H2 annealing can result in improved high-k/Ge interfaces, and so did we obtain the same results in the Chapter 2. Utilizing the conductance method, we did obtain a lower Dit value of about 5×1011 cm-2eV-1 and a better FLE of at least 15% enhancement for the MOS capacitors after performing FGA at 300°C. Hence, we anticipate that the fabricated p-MOSFETs would exhibit enhanced performance as a result of improvements on not only the p+/n junction but also the dielectric interface.
45
Fig. 3.4 displays the I
D-VD characteristics of the p-MOSFETs with and without FGA. As expected, the devices undergone FGA at 300°C exhibit superior performance with larger drive current. From linear region of ID-VD curves for different gate overdrive and various gate lengths, we are able to extract the S/D series resistance by the Terada and Muta method (see Fig. 3.5).For the two samples, source/drain series resistance (RSD) and G/S,D overlap (ΔL) are both increased due to more diffusion of dopants and lower S/D doping, caused by addition 300°C 90 minutes thermal annealing.
ID-VG characteristics of the p-MOSFETs are shown in Fig. 3.6, where higher on current (less coulomb scattering), lower off current (less defects), and better subthreshold swing (less Dit) are observed for both samples. However, GIDL becomes more severe after FGA for p-MOSFETs, with larger gate/drain overlapped region and lower doping concentration taken into consideration. The GIDL increases a half order as VD changes from -0.1V to -2.1V before FGA; in contrast, it increases two orders after FGA. Consequently, we believe the latter mechanism to be dominant, since only slight increase of ΔL should not be responsible for the huge increase of GIDL and moderate doping density rather than too high doping does GIDL occur (band bending much less than Eg for high doping).
In Fig. 3.7, we compare ID-VG characteristics of the two samples. No passivation sample shows higher on current (lower RSD), lower off current (narrower depletion width), and better subthreshold swing than the GeO2 passivation sample. It is not fair to jump to conclusions that GeO2 passivation sample has inferior performance; instead, thermal budget of S/D dopants should be kept the same at a more equal footing.
Fig. 3.8 demonstrates the gated-diode measurement to detect the interface state density
roughly. Igen,s are 6.6×10-8A and 4.0×10-8A for the no passivation and GeO2 passivation sample after FGA, respectively, indicating Dit of no passivation sample is larger than that of GeO2 passivation sample.Split–CV before FGA is measured to extract the effective mobility. Fig. 3.9 shows the
46
split-CV of the two samples. The effective inversion and depletion charge are the integration of Cgc and Cgb over VG respectively, eq. (3.1).
(3.1) The effective mobility and effective electric field can be derived from Qinv and Qb, eq(3.2).
(3.2) The μeff - Eeff plot is depicted in Fig. 3.10 [2] and no passivation as well as GeO2 passivation sample have their high field mobility 1.3X and 1.7X higher than the Si universal curve respectively, where higher mobility of GeO2 passivation sample is attributed to the better interface quality and less coulomb scattering.
An unusual phenomenon is observed where mobility seems to degrade after FGA, as demonstrated in Fig. 3.11. From our data, the slope of Cgc is much steeper after FGA so that Qinv becomes much larger at the same gate bias, while drain conductance ( ) only increases slightly. The much larger Qinv and little increase of lead to lower mobility after FGA, which is not consistent with the field effect mobility ( increases after FGA). We still need more time to figure out the mechanisms behind it.
3.3.3 Charge Trapping Behavior of Ge p-MOSFETs
For Ge PMOSFET, the GeO2 passivation sample shows more hole trapping than no passivation sample under static stress VG-Vth = -4V, caused by bulk traps in GeO2 and border traps at the GeO2/Al2O3 interface (see Fig. 3.12a). Besides, the γ value of zafar model [3] is extracted to be 0.17 and 0.22 for GeO2 passivation and no passivation respectively. Smaller γ represents wider distribution of the capture time and more short time constant slow traps existed, indicating ratio of early traps is larger for GeO2 passivation sample, as depicted in Fig.
3.12b. Also, GeO
2 passivation sample shows a little more severe SS degradation under static stress (Fig. 3.12c).47