Silicon–germanium heterojunction bipolar transistors (HBTs) are of interest as low-noise microwave amplifiers due to their competitive noise performance and low cost relative to III–V devices. The fundamental noise performance limits of HBTs are thus of interest, and several studies report that quasiballistic electron transport across the base is a mechanism leading to cryogenic non-ideal IV characteristics that affect these limits. However, this conclusion has not been rigorously tested against theoretical predictions because prior studies modeled electron transport with empirical approaches or approximate solutions of the Boltzmann equation. Here, we study non-diffusive transport in narrow-base SiGe HBTs using an exact, semi-analytic solution of the Boltzmann equation based on an asymptotic expansion approach. We find that the computed transport characteristics are inconsistent with experiments, implying that quasiballistic electron transport is unlikely to be the origin of cryogenic non-ideal IV characteristics. Our work helps to identify the mechanisms governing the lower limits of the microwave noise figure of cryogenic HBT amplifiers.

Silicon–germanium heterojunction bipolar transistors (HBTs) are widely used in microwave applications such as radar and communication systems1,2 and show potential in space science3 and imaging applications4–6 owing to their competitive microwave performance combined with ease of integration with the CMOS process, high yield, and low cost relative to III–V technologies.7,8 SiGe HBTs are now approaching the THz domain due to fabrication process advancements that enable the continued scaling of key parameters such as base width and resistance.2,9 Moreover, at cryogenic temperatures, they have demonstrated DC gains exceeding 45 dB, noise temperatures 5 K at 3 GHz with bandwidths of 2–3 GHz,10,11 lower 1/f noise relative to III–V devices,12 and power consumption in the hundreds of microwatts.13 These competitive figures of merit make SiGe HBTs of interest for low-noise amplifiers and cryo-CMOS circuits14,15 in radio astronomy and quantum computing applications13,16 and in oscillators with low phase noise.12,17,18

Following the first reports of SiGe HBTs grown by molecular beam epitaxy,19,20 studies of the cryogenic performance of HBTs were performed in the early 1990s.19,21 Subsequent work focused on understanding and optimizing base doping and Ge grading profiles for cryogenic performance.22 While these works demonstrated the enhanced collector current and transconductance expected at cryogenic temperatures, later studies reported marked deviations from the DC current–voltage characteristics predicted from drift–diffusion theory.23,24 Specifically, it was observed that below 77 K, the collector (IC) and base current (IB) exceeded the predicted values at a given temperature and bias, and were independent of temperature, with the transconductance gm plateauing instead of increasing as T1. These non-ideal IV characteristics limit the gain, cutoff frequency, and ultimately the microwave noise figure.

The origin of this behavior has been attributed to several mechanisms, including trap-assisted tunneling at base–emitter voltages (VBE) well below the built-in potential;23 and non-equilibrium carrier transport, in which electrons quasiballistically traverse the base, at biases comparable to the built-in potential.24 The latter effect has been phenomenologically described using an elevated electron temperature that is taken as the effective temperature for thermionic emission at the base–emitter junction.10,11,24 As base widths were scaled further down to 20 nm in recent years, direct tunneling of electrons was reported to contribute to the collector current at biases approaching the built-in potential.25–27 

There are several discrepancies with these explanations, however. First, non-ideal collector currents have been reported in first-generation devices with base widths on the order of 100 nm,27 for which direct tunneling is unlikely. Furthermore, at base doping levels above 3×1018 cm3 common in modern devices,25,27 ionized impurity scattering is expected to dominate. Therefore, quasiballistic transport is not expected to be markedly more pronounced at cryogenic temperatures relative to room temperature as evidenced by the weak temperature dependence of the minority carrier mobility.11,28–30 Finally, while electrons may be heated as they traverse the base by the built-in field, the collector current and transconductance depend on the electron temperature at the base–emitter junction prior to transport across the base and thus cannot be affected by the built-in field.

A quantitative, non-empirical description of the quasiballistic transport process would allow a more thorough test of whether quasiballistic transport is a possible origin of the cryogenic DC non-idealities. Steady particle transport across a slab of thickness comparable to the MFPs of the particles is described by the Boltzmann equation.31,32 The slab problem has been extensively investigated for radiative,32–34 neutron,35 phonon,32,36 and electron transport37–40 as well as for rarefied gases.41–43 Although analytical solutions under the “one-speed” or “gray” approximations are possible,34,37 solutions considering energy-dependent relaxation times are less tractable analytically. Alternative numerical approaches are computationally expensive owing to the need to discretize both the spatial and reciprocal space coordinates.37,38,44 Several works have reported asymptotic series expansion methods, based on the original theory by Grad,45 for rarefied gas dynamics46 and phonon transport.47 These methods could enable the efficient solution of the present problem of electronic transport in a narrow base, but they have yet to be adapted for electronic transport.

Here, we report a study of quasiballistic electron transport across a narrow base using an exact, semi-analytic asymptotic expansion approach to solve the Boltzmann equation. We show that quasiballistic transport leads to decreased collector current for a given base–emitter voltage and does not alter the transconductance. Both of these findings contradict experimental observations, indicating that quasiballistic transport is unlikely to be the origin of cryogenic non-idealities in SiGe HBTs. Alternatively, we hypothesize that the origin of these non-idealities could arise from local inhomogeneities in built-in potential, which are readily observed in Schottky diodes48,49 and have been postulated to exist in SiGe HBTs.50 Our work advances the understanding of cryogenic transport processes in HBTs that limit gain, transconductance, and thus ultimately microwave noise performance.

We begin by describing the semi-analytical asymptotic approach used to solve the one-dimensional Boltzmann equation describing electron transport across the base. Focusing on the DC characteristics, we assume steady transport across a base of width L with prescribed forward and reverse going electron distribution functions at the left and right boundaries, respectively. Then, the Boltzmann equation for the electronic distribution function fλ(x) is given by

fλx¯=LvxλλCλλfλ=λMλλfλ,
(1)

where Cλλ is the collision matrix, vxλ is the group velocity for electronic state λ, Mλλ is a dimensionless matrix defined as above, x is the spatial coordinate, x¯x/L, and fλ is the desired distribution function normalized so that V1λfλ(x)=n(x), where V is a normalizing volume. Macroscopic quantities such as electric current are computed using the appropriate Brillouin zone sum.32 

In this work, we will solve this equation using an exact, semi-analytic asymptotic expansion method originally reported for rarefied gases46 and recently adapted for phonons.47 The theory for phonons is given in detail in Ref. 47 for an isotropic crystal under the relaxation time approximation. Here, we generalize this theory to incorporate the full electron collision matrix and arbitrary crystals and extend its applicability beyond linear deviations from a reference distribution.

In the asymptotic expansion approach, fλ and associated observables are written as series expansions with the average Knudsen number ϵΛ/L as the expansion parameter, where the average mean free path Λλkvxλτλfλeq/neq. Here, fλeq and neq are the reference equilibrium distribution function and electron density at zero base–emitter voltage, and τλ is the wave vector dependent relaxation time. As an example, fλ is written as

fλ=lϵlfλ,l,
(2)

where fλ,l is the distribution function for order l. Substituting this expansion into Eq. (1), we get

lϵlfλ,lx¯=lϵlλMλλfλ,l.
(3)

Since Mλλ scales with ϵ1, the left and right side of Eq. (3) are offset by one order of ϵ. Equating terms with equal order of ɛ for l=0, we find that the zeroth-order solution satisfies

λMλλfλ,0(x)=0=g0(x)λMλλe0,λ,
(4)

where we have split the solution into a pure x-dependent term g0(x) and wave vector dependent term e0,λ; i.e., fλ,0(x)=e0,λg0(x). We note here that the null eigenvector e0,λ can be any Boltzmann distribution function, but for this work, we choose e0,λ=fλeq corresponding to the Boltzmann distribution for the case where no bias is applied.

We now consider the first-order terms. From Eq. (3), the solution can be written using the zeroth-order solution,

fλ,1(x)=λϵ1Mλλ1(fλ,0x¯)+g1(x)e0,λ.
(5)

This solution is also determined only up to an x-dependent multiple of the null eigenvector, denoted g1(x), which will be obtained after deriving the governing equation for x-dependence. The higher-order solutions proceed similarly.

The x-dependence of the zeroth-order solution can be derived by enforcing current continuity for the electric current density Jx in the x-direction for a 1D steady slab,

Jxx¯=0=V1x¯λqvxλlϵlfλ,l,
(6)

which must be satisfied at each order of the expansion. Substituting from Eq. (5) for the first-order term, we find that

λvxλϵx¯(ϵ1Mλλ1fλ,0x¯+g1(x)e0,λ)=0.
(7)

The second term in the parentheses vanishes since vxλ is odd in λ, whereas e0,λ is even. Therefore, we find

2fλ,0(x)x¯2=0.
(8)

Applying V1λfλ,0(x)=n0(x) to the above equation, we see that the zeroth-order electron density n0(x) satisfies the diffusion equation. Similarly, it can be shown that higher-order terms of fλ,l(x) also satisfy the diffusion equation (see Appendix A of Ref. 47).

To solve these equations, the boundary conditions at each order must be specified. The boundary conditions at zeroth order are simply the prescribed isotropic, hemispherical electron distribution functions at the edges of the slab specified by nL and nR. Note that in general, the actual carrier density at the edges of the slab will differ from these values owing to quasiballistic transport. Thus, the solution at zeroth order is just the diffusion equation solution. Note that the boundary conditions of the original problem are completely satisfied at zeroth order.

We now discuss the boundary conditions at higher orders. Because the zeroth-order solution completely satisfies the boundary conditions of the original problem, to enforce the boundary conditions at higher orders, we must introduce a boundary correction term Φλ(x) that satisfies the Boltzmann equation in the boundary region. This boundary correction must exactly cancel the contribution from fλ(x) at the boundaries and vanish in the bulk.

The Boltzmann equation for the boundary solution at first-order is

Φλ,1(η)η=λϵMλλΦλ,1(η),
(9)

where we introduce the stretched boundary-region coordinate ηx/ϵ. Focusing on the left boundary at x¯=0, the condition enforced on this first-order boundary term is that it must cancel the first-order bulk term: Φλ,1|0=fλ,1|0. Using Eq. (5), the condition for the left boundary is

Φλ,1|0=c1e0,λg0x¯|0λϵ1Mλλ1g0x¯|0e0,λ,
(10)

where, as in Ref. 47, we anticipate the boundary term to scale as the gradient of the previous order solution,

g1|x=0=c1g0x¯|x=0.
(11)

An analogous boundary condition applies to the right boundary. The unknown constant c1 captures the jump-type boundary condition for non-diffusive corrections at first order. Solving for c1 involves obtaining the eigenvalues and eigenvectors of M and is described in detail in Ref. 47. In brief, the general solution to Eq. (9) can be written as a linear combination of the eigenvectors corresponding to the negative eigenvalues of M,

Φλ,1=iAihλ,ieρiη,
(12)

where Ai are unknown coefficients and ρi and hλ,i are the negative eigenvalues and the corresponding eigenvectors, respectively. Only the negative eigenvalues are used so that the boundary solution tends to zero in the bulk. Then, Eq. (10) gives a linear system of equations for Ai and c1. We note that in the formulation in Ref. 47, c1 represents a deviation in a linearized quantity, such as temperature, relative to an equilibrium value. However, in the present formulation, c1 multiplies the absolute distribution function rather than a deviation and is not restricted to linear deviations from that distribution.

After calculating c1, the first-order solution is completely specified using the diffusion equation for g1(x) and the boundary condition from Eq. (11). The analysis progresses similarly for higher-order solutions. For the specific slab problem here, it can be shown that the jump coefficients associated with second-order derivatives vanish, as in Appendix C of Ref. 47. Therefore, c1 is the only required coefficient, allowing the asymptotic expansion to be summed over all higher-order terms. After summing fλ over the Brillouin zone to obtain the carrier density, we obtain

n(x)=n0+(nRnL)(12x¯)n=1ϵn(2)n1c1n=n0+ϵc11+2ϵc1(nRnL)(12x¯),
(13)

where n0(x) is the zeroth-order carrier density and the higher-order terms correct for non-diffusive transport. Because the macroscopic constitutive relation between the electric current density and the number density gradient applies in the bulk,47 the current density including the effect of quasiballistic transport is given by

Je,x=qDe,Sin(x)x.
(14)

Here, De,Si is the bulk diffusivity. These results have been extensively validated by comparison to Monte Carlo simulations in Ref. 47. We also observe from Eq. (13) that the dependence of carrier density and thus electric current on base–emitter voltage VBE is only through the left boundary condition nLeqVBE/kT, allowing the current–voltage characteristics for arbitrary VBE to be obtained once c1 is computed.

We briefly discuss the features of the present asymptotic approach compared to those reported previously for electronic transport. An early study of the slab electron transport problem employed a direct scattering matrix solution to the Boltzmann equation38 that required discretization in the wave vector and spatial coordinates and was thus numerically expensive. Approximate methods such as the McKelvey flux method39,51 and the Landauer approach44,52 were next applied for electron and phonon transport. These flux-based methods provide approximate solutions to the Boltzmann equation with minimal computational expense due to simplifications to the collision term. The advantage of the present approach is that it provides the exact solution to the Boltzmann equation with the full collision matrix while requiring only discretization in wave vector space rather than in both real space and wave vector space. The resulting calculation is thus orders of magnitude faster than the purely numerical solution of the Boltzmann equation. Furthermore, compared to Ref. 47, this work does not require linearization around an equilibrium distribution and can accommodate arbitrary crystals, rather than isotropic ones, and the full collision matrix, rather than the relaxation time approximation. Finally, with a standard numerical treatment, the current density would need to be recomputed when any variable is changed. With the present method, the expensive calculation for c1 only needs to be performed once at each temperature because the average Knudsen number and VBE are independent of c1 and included directly in Eq. (13).

We apply this approach to study 1D steady-state electron transport in slabs of non-degenerate silicon (NA=3×1018 cm3) and of width L40 nm–100μm. The hemispherical distribution function at the left and right boundaries is a Boltzmann distribution with carrier density nL=nL,eqexp(qVBE/kT) and nR=0, respectively, where nL,eq is the equilibrium minority carrier concentration. Note that at high biases, the law of the junction used for nL may not be strictly satisfied due to degenerate carrier statistics, but this approximation will not affect our conclusions as discussed later. The junction is assumed to be a step junction with a uniform Ge fraction of 0.2, with bandgap values for SiGe obtained from Ref. 53. Owing to lack of precise knowledge of the electronic structure and scattering mechanisms in the strained SiGe:C films used in devices, in this work, we assume parabolic bands and employ the relaxation time approximation, with energy-dependent impurity and acoustic phonon scattering rates based on the forms given in Chap. 3 of Ref. 31. Several works report that the temperature dependence of minority carrier mobility in p-doped SiGe does not obey the T1.5 scaling typical of impurity scattering at cryogenic temperatures.28–30 Therefore, we extract coefficients and exponents for relaxation time τ=μ0TpEs by fitting the computed mobility to minority carrier mobility data reported in Fig. 3 of Ref. 30 and combining the relaxation times for acoustic and impurity scattering using Matthiessen’s rule. The fitting coefficients and exponents for acoustic (impurity) scattering rates are μ0=350(700) cm2 V1 s1, p=1.5(0.25), and s=0.5(0.5), respectively. Our conclusions are robust to these assumptions. We solve for c1 at the temperatures corresponding to Ref. 25, assuming an isotropic crystal using a 2200×2200 grid in energy and an angular coordinate space. The collision matrix used in this work corresponds to Eq. (B.5) of Ref. 47. For example, at 300 K, we find c1=0.7341.

Figure 1(a) shows the carrier density vs normalized distance at 300 K for L=40 nm and VBE=0.8 V calculated at different orders of the expansion. The zeroth-order solution is by definition the drift–diffusion solution and exhibits a linear profile with a carrier density at the left edge of the slab of nL=nL,eqexp(qVBE/kT)=0.87×1017 cm3. The higher-order non-diffusive corrections predict a reduced carrier concentration gradient across the base relative to the zeroth-order case. This feature occurs because at ϵ=0.38, transport is quasiballistic, with electron mean free paths being on the order of the base width. Under these conditions, the ballistic propagation of some electrons across the base lowers the carrier density at the left boundary and increases it at the right boundary, resulting in jump-type boundary conditions at each edge and a shallower carrier density across the base. The solution at the first-order over-corrects for non-diffusive effects, but the infinite series accounts for higher-order terms and lies in between the zeroth and first-order solutions. This non-monotonicity occurs because the closed form solution switches signs at each order in Eq. (13).

FIG. 1.

(a) Electron density n(x) vs normalized distance x¯ at 300 K and VBE=0.8 V at zeroth order (blue dashed–dotted line), first order (orange dashed line), and infinite order (green solid line). (b) Normalized electric current density vs Knudsen number ϵ at 300 K (green dashed–dotted line) and 36 K (blue solid line) derived from a carrier concentration gradient. The current density is normalized to the diffusion current density obtained at zeroth order (dashed line). Quasiballistic transport predicts a comparable decrease in current relative to drift–diffusion at both room and cryogenic temperatures as Knudsen number increases.

FIG. 1.

(a) Electron density n(x) vs normalized distance x¯ at 300 K and VBE=0.8 V at zeroth order (blue dashed–dotted line), first order (orange dashed line), and infinite order (green solid line). (b) Normalized electric current density vs Knudsen number ϵ at 300 K (green dashed–dotted line) and 36 K (blue solid line) derived from a carrier concentration gradient. The current density is normalized to the diffusion current density obtained at zeroth order (dashed line). Quasiballistic transport predicts a comparable decrease in current relative to drift–diffusion at both room and cryogenic temperatures as Knudsen number increases.

Close modal

To study how the electric current density is impacted by quasiballistic transport, we calculate the current density vs Knudsen number in Fig. 1(b) for two temperatures values reported in Ref. 25. The electric current density from Eq. (14) is proportional to the carrier concentration gradient in Fig. 1(a) for a given voltage, temperature, and Knudsen number. At 300 K, the current density is observed to decrease from the diffusive limit as ϵ1 (quasiballistic limit). As seen from Fig. 1(a), quasiballistic corrections lower the concentration gradient, and the contribution of the correction terms is multiplied by Knudsen number, resulting in greater deviation with increasing Knudsen number. Therefore, a shallower concentration gradient and consequently a decreased current density are expected with increasing Knudsen number. At 36 K corresponding to the temperature reported in Ref. 25, the calculation for c1 is repeated to give a value of 0.733. At this temperature, we see that the deviation from the diffusive limit is comparable to that at 300 K. The difference between the two curves is due to changes in bulk diffusivity at 36 K that are independent of c1 and predict a slightly smaller current density than at 300 K. This calculation shows that even at cryogenic temperatures, quasiballistic effects yield a lower collector current density relative to the drift–diffusion prediction, opposite to that observed experimentally.

We now compare the predictions of our calculation with experimentally reported DC IV characteristics. The data from Ref. 25 are selected due to the availability of Gummel curves for a state-of-the-art device at a reasonably dense concentration of temperatures. Figure 2(a) plots calculated and measured collector current density vs base–emitter voltage VBE. We observe that at all temperatures, quasiballistic transport predicts a weakly reduced current density compared to the drift–diffusion prediction, as expected from the discussion of Fig. 1(b). The two calculations nearly coincide in Fig. 2(a) because the deviation of the quasiballistic current density from the diffusive current density is small on the logarithmic scale. At 300 K, 200 K, and 77 K, both predictions agree well with measured data. However, at 36 K, quasiballistic and drift–diffusion current density predictions are similar in magnitude but are both orders of magnitude lower than the measured data for a given VBE. Furthermore, the predicted slope of the JV curve at 36 K is unchanged from the drift–diffusion prediction and does not saturate to a value similar to that at 77 K as observed in the measured data.

FIG. 2.

(a) Collector current density Jc vs base–emitter voltage VBE from drift–diffusion (dashed line) and quasiballistic calculations (solid line) compared to measured data (triangles, Ref. 25) at various temperatures. (b) Transconductance per unit area gm vs temperature for an ideal diode (dashed line), from calculation (circles), and experiment (triangles, Ref. 25) for a fixed JC derived from the JV characteristics in (a). The present calculations do not predict the increased current density or decreased transconductance observed in measurements at cryogenic temperatures.

FIG. 2.

(a) Collector current density Jc vs base–emitter voltage VBE from drift–diffusion (dashed line) and quasiballistic calculations (solid line) compared to measured data (triangles, Ref. 25) at various temperatures. (b) Transconductance per unit area gm vs temperature for an ideal diode (dashed line), from calculation (circles), and experiment (triangles, Ref. 25) for a fixed JC derived from the JV characteristics in (a). The present calculations do not predict the increased current density or decreased transconductance observed in measurements at cryogenic temperatures.

Close modal

To further examine this result, we present transconductance vs temperature at a fixed value of collector current density in Fig. 2(b). The transconductance per unit area gm is calculated as the slope of the current–voltage curve at a fixed bias, and for an ideal SiGe HBT, it is given by gm=qJC/kT. The measured transconductance exhibits a plateau to a lower value than the ideal value as the temperature decreases. However, the transconductance values calculated from the slope of the calculated results in Fig. 2(a) are identical to the ideal values. Therefore, quasiballistic transport does not change the temperature dependence of the transconductance. The reason can be seen in Eq. (13) in which VBE affects the electron density only through the left boundary condition nL; therefore, a change in voltage affects current density in a way that is independent of the mechanism of carrier transport and thus preserves the diffusive temperature dependence regardless of the Knudsen number.

Our analysis has shown that the predicted influence of quasiballistic transport on the IV characteristics of HBTs is inconsistent with the experimental observations. We now address several mechanisms that we have not included in our analysis. First, we have not incorporated the built-in field in the base region due to Ge grading. However, as in Ref. 54, the only effect of a field term is to modify M and thus change the value of c1, and our conclusions are robust to such changes. Second, modern devices may have base doping levels exceeding 1019cm3 to minimize base resistance and prevent carrier freeze-out.25,27,55 This high level of doping leads to degenerate carrier statistics, while we have assumed non-degenerate statistics. However, non-idealities in the IV characteristics are observed at temperatures up to 80 K for which the electrons are non-degenerate, and the present analysis thus applies under relevant conditions. Furthermore, in degenerate conditions, the law of the junction overpredicts the minority carrier density,56 implying that accounting for degenerate statistics would predict a further decreased collector current that again contradicts experiment. Last, the asymptotic solution is restricted to values of the average Knudsen number such that 2ϵc1<1, which limits our analysis to a minimum base width of 40 nm for the chosen scattering rates. However, DC non-idealities were reported in devices with base widths on this order,25–27 and again, our analysis applies. Given these considerations, we conclude that quasiballistic transport is not responsible for non-ideal cryogenic DC characteristics of SiGe HBTs.

Finally, we discuss alternate explanations for the cryogenic non-idealities. Prior works have suggested that direct tunneling from the emitter to the collector is a possible origin of non-idealities in devices with narrow-base widths 10 nm.25,26 However, non-ideal cryogenic transconductance has been observed in first-generation SiGe HBTs with base widths on the order of 100 nm24,27 for which the direct tunneling current is expected to be negligible. Other reported evidence for a transition from trap-assisted transport to a tunneling or quasiballistic mechanism is the change in the slope of the collector I–V characteristics.26,27 However, similar trends are observed in the forward regime of Schottky diodes49,57 for which direct tunneling or quasiballistic transport is not relevant, and the possibility of a similar mechanism occurring in SiGe HBTs has not been excluded.

We instead offer a hypothesis for the origin of cryogenic non-idealities as originating from spatial inhomogeneities in the base–emitter potential barrier. The earliest works on heavily doped p–n junctions reported temperature-independent slopes of cryogenic forward-bias I–V characteristics for voltages where band-to-band tunneling is unlikely to occur.58–60 Extensive studies of Schottky diodes have reported a variety of cryogenic IV non-idealities including the so-called T0 anomaly61 and temperature-dependent ideality factors.62–64 These non-idealities have been explained by inhomogeneities in barrier height49 that exist even in high-quality epitaxial junctions such as NiSi2/Si junctions and have been linked to the atomic structure of the interface.48,65 In SiGe HBT junctions, non-ideal base currents and the resulting random telegraph noise have been attributed to voltage barrier height fluctuations arising from trap states in the base–emitter space charge region50 such as those of electrically active carbon defects.66 Future work will examine whether such barrier inhomogeneities are able to explain the collector current non-idealities of SiGe HBTs.

We have reported a study of quasiballistic transport in SiGe HBTs using an exact, semi-analytic solution to the Boltzmann equation based on an asymptotic expansion method. We find that the predicted IV characteristics including quasiballistic transport are inconsistent with experiment. Specifically, our calculations including quasiballistic transport predict collector currents that are orders of magnitude smaller than the measured currents for a given base–emitter voltage, and an unaltered temperature dependence of transconductance relative to the ideal value, both of which contradict experimental observations. We suggest that local fluctuations in the base–emitter barrier height could account for the non-ideal collector current as has been observed in Schottky diodes. Our work advances the understanding of electronic transport in cryogenic SiGe HBTs that is required to further improve their microwave noise performance.

The authors thank Mark Lundstrom, J. P. Peraud, and Nicolas Hadjiconstantinou for useful discussions. This work was supported by the National Science Foundation (NSF) (Award No. 1911926).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

The authors have no conflicts to disclose.

1.
H.
Rücker
,
B.
Heinemann
, and
A.
Fox
, “Half-terahertz SiGe BiCMOS technology,” in 2012 IEEE 12th Topical Meeting on Silicon Monolithic Integrated Circuits in RF Systems (IEEE, 2012), pp. 133–136.
2.
P.
Chevalier
,
M.
Schroter
,
C. R.
Bolognesi
,
V.
D’Alessandro
,
M.
Alexandrova
,
J.
Bock
,
R.
Flickiger
,
S.
Fregonese
,
B.
Heinemann
,
C.
Jungemann
,
R.
Lovblom
,
C.
Maneux
,
O.
Ostinelli
,
A.
Pawlak
,
N.
Rinaldi
,
H.
Rucker
,
G.
Wedel
, and
T.
Zimmer
, “
Si/SiGe:C and InP/GaAsSb heterojunction bipolar transistors for THz applications
,”
Proc. IEEE
105
,
1035
(
2017
).
3.
J. C.
Bardin
,
P.
Ravindran
,
S.-W.
Chang
,
R.
Kumar
,
J. A.
Stern
,
M. D.
Shaw
,
D. S.
Russell
, and
W. H.
Farr
, “A high-speed cryogenic SiGe channel combiner IC for large photon-starved SNSPD arrays,” in 2013 IEEE Bipolar/BiCMOS Circuits and Technology Meeting (BCTM) (IEEE, 2013), pp. 215–218.
4.
J.
Yuan
,
J. D.
Cressler
,
R.
Krithivasan
,
T.
Thrivikraman
,
M. H.
Khater
,
D. C.
Ahlgren
,
A. J.
Joseph
, and
J. S.
Rieh
, “
On the performance limits of cryogenically operated SiGe HBTs and its relation to scaling for terahertz speeds
,”
IEEE Trans. Electron Devices
56
,
1007
(
2009
).
5.
J. D.
Cressler
, “New developments in SiGe HBT reliability for RF through mmW circuits,” in 2021 IEEE International Reliability Physics Symposium (IRPS) (IEEE, 2021), pp. 1–6.
6.
M.
Frounchi
and
J. D.
Cressler
, “A SiGe millimeter-wave front-end for remote sensing and imaging,” in 2020 IEEE Radio Frequency Integrated Circuits Symposium (RFIC) (IEEE, 2020), pp. 227–230.
7.
J. D.
Cressler
, “
On the potential of SiGe HBTs for extreme environment electronics
,”
Proc. IEEE
93
,
1559
(
2005
).
8.
J. D.
Cressler
and
G.
Niu
,
Silicon-Germanium Heterojunction Bipolar Transistors
(
Artech House
,
2003
).
9.
B.
Heinemann
,
H.
Rucker
,
R.
Barth
,
F.
Barwolf
,
J.
Drews
,
G. G.
Fischer
,
A.
Fox
,
O.
Fursenko
,
T.
Grabolla
,
F.
Herzel
,
J.
Katzer
,
J.
Korn
,
A.
Kruger
,
P.
Kulse
,
T.
Lenke
,
M.
Lisker
,
S.
Marschmeyer
,
A.
Scheit
,
D.
Schmidt
,
J.
Schmidt
,
M. A.
Schubert
,
A.
Trusch
,
C.
Wipf
, and
D.
Wolansky
, “SiGe HBT with fx/fmax of 505 GHz/720 GHz,” in Technical Digest—International Electron Devices Meeting (IEDM) (IEEE, 2017), Vol. 2, p. 3.1.1.
10.
J. C.
Bardin
,
S.
Montazeri
, and
S. W.
Chang
, “
Silicon germanium cryogenic low noise amplifiers
,”
J. Phys.: Conf. Ser.
834
,
012007
(
2017
).
11.
J. C.
Bardin
, “Silicon-germanium heterojunction bipolar transistors for extremely low-noise applications,” Ph.D. thesis (California Institute of Technology, 2009).
12.
B.
Van Haaren
,
M.
Regis
,
O.
Llopis
,
L.
Escotte
,
A.
Gruhle
,
C.
Mahner
,
R.
Plana
, and
J.
Graffeuil
, “
Low-frequency noise properties of SiGe HBT’s and application to ultra-low phase-noise oscillators
,”
IEEE Trans. Microw. Theor. Tech.
46
,
647
(
1998
).
13.
S.
Montazeri
,
W. T.
Wong
,
A. H.
Coskun
, and
J. C.
Bardin
, “
Ultra-low-power cryogenic SiGe low-noise amplifiers: Theory and demonstration
,”
IEEE Trans. Microw. Theor. Tech.
64
,
178
(
2016
).
14.
W. T.
Wong
,
M.
Hosseini
,
H.
Rucker
, and
J. C.
Bardin
, “A 1 mW cryogenic LNA exploiting optimized SiGe HBTs to achieve an average noise temperature of 3.2 K from 4–8 GHz,” in IEEE MTT-S International Microwave Symposium Digest, 4–6 August 2020 (IEEE, 2020), p. 181.
15.
B.
Patra
,
R. M.
Incandela
,
J. P.
Van Dijk
,
H. A.
Homulle
,
L.
Song
,
M.
Shahmohammadi
,
R. B.
Staszewski
,
A.
Vladimirescu
,
M.
Babaie
,
F.
Sebastiano
, and
E.
Charbon
, “
Cryo-CMOS circuits and systems for quantum computing applications
,”
IEEE J. Solid-State Circuits
53
,
309
(
2018
).
16.
S.
Weinreb
,
J. C.
Bardin
, and
H.
Mani
, “
Design of cryogenic SiGe low-noise amplifiers
,”
IEEE Trans. Microw. Theor. Tech.
55
,
2306
(
2007
).
17.
R.
Plana
and
L.
Escotte
, “Noise properties of micro-wave heterojunction bipolar transistors,” in 1997 21st International Conference on Microelectronics. Proceedings (1997), Vol. 1, pp. 215–222; available at http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=91521052&fileName=GC91521052.pdf
18.
G.
Niu
, “
Noise in SiGe HBT RF technology: Physics, modeling, and circuit implications
,”
Proc. IEEE
93
,
1583
(
2005
).
19.
G. L.
Patton
,
S. S.
Iyer
,
S. L.
Delage
,
S.
Tiwari
, and
J. M.
Stork
, “
Silicon-germanium base heterojunction bipolar transistors by molecular beam epitaxy
,”
IEEE Electron Device Lett.
9
,
165
(
1988
).
20.
A.
Gruhle
,
H.
Kibbel
,
U.
Konig
,
U.
Erben
, and
E.
Kasper
, “
MBE-grown Si/SiGe HBTs with high beta, fT, and fmax
,”
IEEE Electron Device Lett.
13
,
206
(
1992
).
21.
G. L.
Patton
,
J. H.
Comfort
,
B. S.
Meyerson
, and
F.
Emmanuel
, “SiGe-base heterojunction bipolar transistors,” in International Technical Digest on Electron Devices (IEEE, 1990), Vol. 1, p. 171.
22.
J. D.
Cressler
,
J. H.
Comfort
,
E. F.
Crabbe
,
G. L.
Patton
,
J. M.
Stork
,
J. Y.
Sun
, and
B. S.
Meyerson
, “
On the profile design and optimization of epitaxial Si and SiGe-base bipolar technology for 77 K applications—Part I: Transistor DC design considerations
,”
IEEE Trans. Electron Devices
40
,
525
(
1993
).
23.
A. J.
Joseph
,
J. D.
Cressler
, and
D. M.
Richey
, “
Operation of SiGe heterojunction bipolar transistors in the liquid-helium temperature regime
,”
IEEE Electron Device Lett.
16
,
268
(
1995
).
24.
D. M.
Richey
,
A. J.
Joseph
,
J. D.
Cressler
, and
R. C.
Jaeger
, “
Evidence for non-equilibrium base transport in Si and SiGe bipolar transistors at cryogenic temperatures
,”
Solid-State Electron.
39
,
785
(
1996
).
25.
H.
Rucker
,
J.
Korn
, and
J.
Schmidt
, “Operation of SiGe HBTs at cryogenic temperatures,” in Proceedings of the IEEE Bipolar/BiCMOS Circuits and Technology Meeting (IEEE, 2017), p. 17.
26.
D.
Davidović
,
H.
Ying
,
J.
Dark
,
B. R.
Wier
,
L.
Ge
,
N. E.
Lourenco
,
A. P.
Omprakash
,
M.
Mourigal
, and
J. D.
Cressler
, “
Tunneling, current gain, and transconductance in silicon-germanium heterojunction bipolar transistors operating at millikelvin temperatures
,”
Phys. Rev. Appl.
8
,
1
(
2017
).
27.
H.
Ying
,
J.
Dark
,
A. P.
Omprakash
,
B. R.
Wier
,
L.
Ge
,
U.
Raghunathan
,
N. E.
Lourenco
,
Z. E.
Fleetwood
,
M.
Mourigal
,
D.
Davidovic
, and
J. D.
Cressler
, “
Collector transport in SiGe HBTs operating at cryogenic temperatures
,”
IEEE Trans. Electron Devices
65
,
3697
(
2018
).
28.
S. E.
Swirhun
,
D. E.
Kane
, and
R. M.
Swanson
, “Temperature dependence of minority electron mobility and bandgap narrowing in p/sup +/Si,” in Technical Digest—International Electron Devices Meeting (IEEE, 1988), p. 298.
29.
D. B. M.
Klaassen
, “
A unified mobility model for device simulation—II. Temperature dependence of carrier mobility and lifetime
,”
Solid State Electron.
35
,
961
(
1992
).
30.
J. S.
Rieh
, “
Temperature dependent minority electron mobilities in strained Si/sub 1-x/Ge/sub x/ (0. 2 < x > 0. 4) layers
,”
IEEE Trans. Electron Devices
47
,
883
(
2000
).
31.
M.
Lundstrom
,
Fundamentals of Carrier Transport
, 2nd ed.
(Cambridge University Press,
2000
).
32.
G.
Chen
,
Nanoscale Energy Transport and Conversion: A Parallel Treatment of Electrons, Molecules, Phonons, and Photons
(
Oxford University Press
,
2005
).
33.
S.
Chandrasekhar
,
Radiative Transfer
(
Dover Publications
,
New York
,
1960
).
34.
M. F.
Modest
,
Radiative Heat Transfer
(
Academic Press
,
2013
).
35.
J.
Spanier
and
E.
Gelbard
, Monte Carlo Principles and Neutron Transport Problems, Addison-Wesley Series in Computer Science and Information Processing (Addison-Wesley Publishing Company, 1969).
36.
C.
Hua
and
A. J.
Minnich
, “
Semi-analytical solution to the frequency-dependent Boltzmann transport equation for cross-plane heat conduction in thin films
,”
J. Appl. Phys.
117
,
175306
(
2015
).
37.
A. A.
Grinberg
and
S.
Luryi
, “
Diffusion in a short base
,”
Solid-State Electron.
35
,
1299
(
1992
).
38.
M. A.
Stettler
and
M. S.
Lundstrom
, “
A microscopic study of transport in thin base silicon bipolar transistors
,”
IEEE Trans. Electron Devices
41
,
1027
(
1994
).
39.
S. I.
Tanaka
and
M.
Lundstrom
, “
A compact HBT device model based on a one-flux treatment of carrier transport
,”
Solid-State Electron.
37
,
401
(
1994
).
40.
M.
Özaydin
and
L. F.
Eastman
, “
Non-equilibrium carrier transport in the base of heterojunction bipolar transistors
,”
Solid-State Electron.
39
,
731
(
1996
).
41.
C.
Cercignani
and
A.
Daneri
, “
Flow of a rarefied gas between two parallel plates
,”
J. Appl. Phys.
34
,
3509
(
1963
).
42.
C.
Cercignani
,
Rarefied Gas Dynamics: From Basic Concepts to Actual Calculations
(
Cambridge University Press
,
2000
), Vol. 21.
43.
A.
Manela
and
N. G.
Hadjiconstantinou
, “
Gas motion induced by unsteady boundary heating in a small-scale slab
,”
Phys. Fluids
20
,
117104
(
2008
).
44.
M.
Lundstrom
, “Carrier transport in BJTs: From ballistic to diffusive and off-equilibrium,” in 2018 IEEE BiCMOS and Compound Semiconductor Integrated Circuits and Technology Symposium (BCICTS 2018) (IEEE, 2018), p. 174.
45.
H.
Grad
, “
Asymptotic theory of the Boltzmann equation
,”
Phys. Fluids
6
,
147
(
1963
).
46.
Y.
Sone
, “Asymptotic theory of flow of rarefied gas over a smooth boundary. I,” in Rarefied Gas Dynamics (Academic Press, 1969), p. 243.
47.
J. P. M.
Péraud
and
N. G.
Hadjiconstantinou
, “
Extending the range of validity of Fourier’s law into the kinetic transport regime via asymptotic solution of the phonon Boltzmann transport equation
,”
Phys. Rev. B
93
,
045424
(
2016
).
48.
R. T.
Tung
,
A. F. J.
Levi
,
J. P.
Sullivan
, and
F.
Schrey
, “
Schottky-barrier inhomogeneity at epitaxial NiSi2 interfaces on Si(100)
,”
Phys. Rev. Lett.
66
,
72
(
1991
).
49.
R. T.
Tung
, “
Electron transport at metal-semiconductor interfaces: General theory
,”
Phys. Rev. B
45
,
13509
(
1992
).
50.
M.
von Haartman
,
M.
Sandén
,
M.
Östling
, and
G.
Bosman
, “
Random telegraph signal noise in SiGe heterojunction bipolar transistors
,”
J. Appl. Phys.
92
,
4414
(
2002
).
51.
J.
Maassen
and
M.
Lundstrom
, “
Steady-state heat transport: Ballistic-to-diffusive with Fourier’s law
,”
J. Appl. Phys.
117
,
035104
(
2015
).
52.
J.
Maassen
and
M.
Lundstrom
, “
(Invited) The Landauer approach to electron and phonon transport
,”
ECS Trans.
69
,
23
(
2015
).
53.
See http://www.ioffe.ru/SVA/NSM/Semicond/Si/ for NSM Archives, Si—Silicon; accessed 29 April 2021.
54.
A. Y.
Choi
,
P. S.
Cheng
,
B.
Hatanpää
, and
A. J.
Minnich
, “
Electronic noise of warm electrons in semiconductors from first principles
,”
Phys. Rev. Mater.
5
,
044603
(
2021
).
55.
M.
Schröter
,
G.
Wedel
,
B.
Heinemann
,
C.
Jungemann
,
J.
Krause
,
P.
Chevalier
, and
A.
Chantre
, “
Physical and electrical performance limits of high-speed SiGeC HBTs—Part I: Vertical scaling
,”
IEEE Trans. Electron Devices
58
,
3687
(
2011
).
56.
V.
Jain
and
M. J. W.
Rodwell
, “
Transconductance degradation in near-THz InP double-heterojunction bipolar transistors
,”
IEEE Electron Device Lett.
32
,
1068
(
2011
).
57.
D. J.
Ewing
,
Q.
Wahab
,
R. R.
Ciechonski
,
M.
Syväjärvi
,
R.
Yakimova
, and
L. M.
Porter
, “
Inhomogeneous electrical characteristics in 4H-SiC Schottky diodes
,”
Semicond. Sci. Technol.
22
,
1287
(
2007
).
58.
L.
Esaki
,
Y.
Miyahara
, and
S.
Corporation
, “
A new device using the tunneling process in narrow p-n junctions
,”
Solid State Electron.
1
,
13
(
1960
).
59.
D. J.
Dumin
and
G. L.
Pearson
, “
Properties of gallium arsenide diodes between 4.2 and 300 K
,”
J. Appl. Phys.
36
,
3418
(
1965
).
60.
J. A.
Del Alamo
and
R. M.
Swanson
, “
Forward-bias tunneling: A limitation to bipolar device scaling
,”
Electron Device Lett.
EDL-7
,
629
(
1986
).
61.
F.
Padovani
and
G.
Sumner
, “
Experimental study of gold-gallium arsenide Schottky barriers
,”
J. Appl. Phys.
36
,
3744
(
1965
).
62.
F.
Padovani
and
R.
Stratton
, “
Field and thermionic-field emission in Schottky barriers
,”
Solid-State Electron.
9
,
695
(
1966
).
63.
R.
Hackam
and
P.
Harrop
, “
Electrical properties of nickel-low-doped n-type gallium arsenide Schottky-barrier diodes
,”
IEEE Trans. Electron Devices
19
,
1231
(
1972
).
64.
A. S.
Bhuiyan
,
A.
Martinez
, and
D.
Esteve
, “
A new Richardson plot for non-ideal Schottky diodes
,”
Electronics
161
,
93
(
1988
).
65.
R. T.
Tung
, “
Schottky-barrier formation at single-crystal metal-semiconductor interfaces
,”
Phys. Rev. Lett.
52
,
461
(
1984
).
66.
J.
Raoult
,
F.
Pascal
,
C.
Delseny
,
M.
Marin
, and
M. J.
Deen
, “
Impact of carbon concentration on 1/f noise and random telegraph signal noise in SiGe:C heterojunction bipolar transistors
,”
J. Appl. Phys.
103
,
114508
(
2008
).