Query processing and data integration

In document Vanja Josifovski (Page 32-38)

The simulated RTH of the reference RF DUT RTH,ref with four emitters (Subsection 2.1.1) was found to be 436.6 K/W.

As previously shown with the single-emitter devices, a first study was aimed to estimate the influence on RTH of technology solutions based on alternative emitter materials. Table 4.4 shows that replacing the thermally-resistive ternary InGaAs emitter cap layers with GaAs (without altering the cap thickness) yields a -0.7% RTH reduction (but the emitter resistance may significantly grow up); similarly, adopting an Al0.25Ga0.75As layer for the deep emitter (as in ‘old’ HBT technology) in lieu of In0.49Ga0.51P provides a marginal cooling action (-0.5% RTH lowering). Simulations were also performed to ‘emulate’

the detrimental effect due to heavy process strain on the thermal conductivity of the InGaAs cap, which was set to the reference conductivity of In0.5Ga0.5As divided by factors 2.5 and 5;1 it was found that RTH increases by 0.7% and 1.2%, respectively. The same analyses were repeated by assigning a much lower thermal conductivity (0.015×10-4 W/µmK [Anh98]) to Si3N4, which surrounds the active device region (Fig. 2.3a). In this case, the RTH (474.3 K/W) grows by 8.6% compared to RTH,ref (436.6 K/W) due to the drastic shrinking of the path for the upward heat flow. A further study evidenced that with this poorly-conductive Si3N4 the layers of the emitter stack play a slightly more important role with respect to the reference case.

Nevertheless, the value 0.185×10-4 W/µmK is chosen as a reference since it is more commonly encountered in the literature.

1 Such a choice stems from the fact that the mobility is known to undergo a strain-induced degradation by similar factors.



reference DUT 436.6

GaAs (instead of InGaAs) cap 433.7 (-0.7%) Al0.25Ga0.75As instead of In0.49Ga0.51P 434.6 (-0.5%) k(cap)=0.0192×10-4 W/µmK due to strain 439.8 (+0.7%) k(cap)=0.0096×10-4 W/µmK due to strain 441.7 (+1.2%) thermal conductivity of Si3N4: 0.015×10-4 W/µmK


(reference: 0.185×10-4 W/µmK)

474.3 (+8.6%)

Another analysis was conceived to quantify the influence of the laminate design; the main results are reported in Table 4.5. It is shown that the laminate weakly contributes to the RTH of the whole structure thanks to the high thermal conductivity offered by the wide (the diameter being 125 µm) 3×3 Cu vias (i.e., it almost behaves as a thermal short-circuit): fully removing it and applying the thermal ground to the GaAs substrate bottom favors indeed only a -2.8% RTH reduction.

Conversely, shrinking the vias to a diameter of 60 µm perceptibly hinders the downward heat propagation, and RTH increases by 5.2%.

Further findings are: (i) little impact is induced by the thickness of the vias; (ii) RTH is almost insensitive to the thickness of the Cu plates due to a compensation between heat spreading and longer heat path to the sink; (iii) the role played by the (uncertain) thermal conductivity of the laminate dielectric can be safely disregarded since the heat flows mainly through the Cu vias. In a practical sense, this ‘lack of sensitivity’

is important for design support since these features often change from a laminate technology to another.

The individual and combined sensitivity to all parameters reported in Table 2.1 was then quantified from expansion (4.1), the coefficients of which, listed in Table 4.6, were determined from the few FEM RTHs required by DOE. Fig. 4.7 illustrates the percentage RTH

variation with respect to RTH,ref; also in this case, only a parameter was varied in the assigned range while keeping the others equal to the reference values, or equal to values that minimize (maximize) RTH, i.e., the minimum (maximum) thicknesses for the semiconductor layers and

maximum (minimum) for the metal ones. This allows gaining an in-depth insight into the influence of the parameters over the assigned ranges.



reference DUT 436.6

DUT without laminate 424.3 (-2.8%)

laminate vias thickness: 15 µm (reference: 25 µm) 434.7 (-0.4%) laminate vias thickness: 35 µm (reference: 25 µm) 438.5 (+0.4%) laminate vias diameter: 60 µm (reference: 125 µm) 459.2 (+5.2%)

laminate metal layers thickness: 9 µm

(reference: 12 µm) 436.7 (<0.1%) laminate metal layers thickness: 19 µm

(reference: 12 µm) 436.7 (<0.1%) thermal conductivity of the laminate dielectric:

0.1×10-4 W/µmK

(reference: 0.0065×10-4 W/µmK)

435.6 (-0.2%)


RTHa 3.08 RTHbe 0.07 RTHbce 0.02 RTHb 1.49 RTHcd 0.31 RTHbde -0.02 RTHc 4.28 RTHce 0.14 RTHcde 0.1 RTHd -20.76 RTHde -0.34 RTHabcd -0.08 RTHe -18.9 RTHabc 0.02 RTHabce 0.02 RTHab -0.07 RTHabd -0.02 RTHabde 0.01 RTHac -0.3 RTHabe -0.02 RTHacde -0.04 RTHad 0.16 RTHacd -0.01 RTHbcde 0.01 RTHae 0.15 RTHace -0.02 RTHabcde 0.05 RTHbc -0.13 RTHade -0.03

Simulations were also carried out to examine the behavior of the four-emitters RF DUT under a large variety of operating conditions, which is particularly important for reliability studies (e.g., [Zam13a]).

Nonlinear thermal effects were activated by enabling (3.3) and (3.4) in COMSOL with the coefficients reported in Table 3.1.

20 30 40 50 60 70 80

-4 -3 -2 -1 0

1 (a)

Percentage RTH variation [%]

Thickness of semiconductor layers [nm]

In0.5Ga0.5As layer M1 M2 InxGa1-xAs layer

In0.49Ga0.51As layer

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Thickness of metal layers [µm]

20 30 40 50 60 70 80

-6 -5 -4 -3 -2


Percentage RTH variation [%]

Thickness of semiconductor layers [nm]

0.5 1.0 Thickness of metal layers [µm]1.5 2.0 2.5 3.0 3.5 4.0

20 30 40 50 60 70 80 -3

-2 -1 0 1


Percentage RTH variation [%]

Thickness of semiconductor layers [nm]

0.5 1.0 Thickness of metal layers [µm]1.5 2.0 2.5 3.0 3.5 4.0

Fig. 4.7 WB technology: percentage RTH variation with respect to RTH,ref, as obtained by modifying only one parameter, and keeping the others equal to (a) the reference values, or equal to the values ensuring the (b) lowest and (c) highest RTH. Expansion (4.1) (solid lines) with coefficients reported in Table 4.4 is compared to

validation data (symbols).

Wide ranges were explored for the temperature of the laminate backside (which in practical cases might not be a good thermal ground) and PD (corresponding to VC=5 V and JC=1, 10, 20, 30, 40 kA/cm2), both enhancing nonlinear thermal effects [Wal01]. Fig. 4.8 shows the average temperatures over the emitter M2 top and the base-emitter junction under the previously described conditions, while Fig. 4.9 illustrates the temperature distribution over a cross section taken along the emitter center for a backside temperature equal to T0=300 K and power amounting to 164 mW.2 This analysis is intended to provide clear guidelines concerning the limits (in terms of backside temperature and power) beyond which the device may experience thermally-induced reliability issues. As an example, the PD for T0

2 All simulations shown in this Chapter were performed by keeping unchanged the heat source geometry, which vertically coincides with the lightly-doped collector assumed to be fully depleted. However, it must be remarked that at high collector current density JC (high VBE) the Kirk (base push-out) effect takes place, which ‘pushes’ the electric field (and thus the dissipation region) toward the interface with the N+ subcollector. Additional simulations allowed determining that the thermal behavior is marginally mitigated by the Kirk effect in both packaging solutions since the thinner heat source is farther away from the base-emitter junction, whereas a slightly higher temperature is reached within the dissipation region due to the increased power density.

applied to the backside should not exceed 250 mW since the maximum operating temperature is usually assigned to 420 K.

300 350 400 450 500 550 600

300 400 500 600 700 800 900 1000

JC= 1 kA/cm2, PD= 8.2 mW JC= 10 kA/cm2, PD= 82 mW JC= 20 kA/cm2, PD= 164 mW JC= 30 kA/cm2, PD= 246 mW JC= 40 kA/cm2, PD= 328 mW

Temperature [K]

Temperature of laminate backside [K]

Fig. 4.8 WB technology: average temperatures over the base-emitter junction (solid lines) and the top surface of emitter M2 ‘seen’ by IR (dotted) as a function of

temperature of laminate backside for various PDs.

Fig. 4.9 WB technology: temperature distribution over a cross section through the emitter center for a backside temperature equal to 300 K and power of 164 mW.

Additionally, it is found that the IR-based imaging, often exploited to get the temperature map over the M2 top, negligibly

underestimates the base-emitter junction temperature (which directly impacts the collector current), unless very high backside temperatures and/or powers are applied – which can be encountered during reliability testing.

In document Vanja Josifovski (Page 32-38)