Leroy C., Rancoita P.-G. Principles Of Radiation Interaction In Matter And Detection

Подождите немного. Документ загружается.

January 9, 2009 10:21 World Scientiﬁc Book - 9.75in x 6.5in ws-bo ok975x65˙n˙2nd˙Ed

550 Principles of Radiation Interaction in Matter and Detection

7.1.2 Surface and Total Dose Eﬀects on the Gain Degradation of

Bipolar Transistors

Equation (7.13) accounts for the gain degradation of bipolar transistors when the

atomic displacement is the dominant damage-mechanism. However, it was found

that the gain degradation may result from damages in the silicon lattice and modi-

ﬁcations in the surface properties (i.e. the so-called surface eﬀects) of the crystal. For

instance, changes in the recombination properties of surface layers like those close

to the emitter-base junction may aﬀect the transistor current-gain. Surface eﬀects

may occur even at small radiation doses which, usually, are insuﬃcient to introduce

an appreciable bulk damage. Thus, these eﬀects can be investigated with particles of

very large D

Ion

NIEL

ratio between the doses deposited by ionization and atomic

displacement (e.g., see Sect. 4.2.3.1), like γ-rays from a

Co source.

This type of damage was considered to be responsible for the observed non-

linear dependence of ∆(1/β) in n − p − n transistors as a function of ﬂuence up to

≈ 10

Compton-electron/cm

2∗∗

(e.g., see Figure 60 at page 165 of [Vavilov and

Ukhin (1977)]). These electrons have an average kinetic energy of ≈ 0.53 MeV

and are generated by Compton eﬀect in irradiations with γ-rays from a

source

††

. The surface component of the damage grows rapidly and approaches sa-

turation for ﬂuences & 3 × 10

Compton-electron/cm

2‡‡

. Above ≈ 10

Compton-

electron/cm

, the transistor degradation was found to be dominated by the di-

splacement damage as expected from Eq. (7.12). After irradiations with 2 MeV

electrons, a similar behavior is observed for the same type of bipolar transistor

(e.g., see Figure 5.3 at page 215 of [Holmes-Siedle and Adams (2002)]). In irradi-

ations of n − p − n transistors with gamma-rays from

Co source up to a dose

of ≈ 0.5 Mrad, the excess base-current was observed to increase non-linearly with

increasing dose

∗

[Nowlin et al. (1993)]. Furthermore, this excess base-current was

observed to saturate at about 1 Mrad, almost independently of the dose rate [Kosier

et al. (1994)]. The linear increase of the inverse of the current gain was observed at

large doses (above 5–10 Mrad) after irradiations with gamma-rays from

Co source

and at large ﬂuences (& 10

e/cm

) with 4.1 MeV electrons [Cheryl, Marshall,

Burke, Summers and Wolicki (1988)].

The so-called lateral and vertical transistors are realized with an architecture

which favors the current ﬂow, mainly, parallel to the surface and in the vertical

direction, respectively. In irradiations with gamma-rays from

Co source up to

≈ 0.5 Mrad, a systematic comparison of dose eﬀects on p − n − p transistors has

∗∗

At this Compton-electron ﬂuence, the concentration of initial Frenkel-pairs is ≈ 4.9×10

−3

for E

= (21–25) eV.

††

The ratio of the Compton-electron ﬂux to the ﬂux of γ-rays from the

Co source is ≈

0.0162 [Summers, Burke, Shapiro, Messenger and Walters (1993)].

‡‡

At this Compton-electron ﬂuence, the concentration of initial Frenkel-pairs is ≈ 1.5×10

−3

for E

= (21–25) eV.

∗

To deliver a dose of ≈ 100 Mrad, a ﬂuence of ≈ 1.9 × 10

γ/cm

from a

Co source is needed

(e.g., see Figure 5.3 at page 215 of [Holmes-Siedle and Adams (2002)]).

January 9, 2009 10:21 World Scientiﬁc Book - 9.75in x 6.5in ws-bo ok975x65˙n˙2nd˙Ed

Displacement Damage and Particle Interactions in Silicon Devices 551

shown that vertical transistors exhibit the least gain degradation [Schimidt et al.

(1995)]. Furthermore, it has been observed that lateral p − n − p transistors with

lightly-doped emitters degrade more rapidly than devices with heavily-doped emit-

ters [Wu et al. (1997)].

Surface and low dose-rate eﬀects in irradiated transistors are discussed, for in-

stance, in Section 6 of Chapter I of Part II of [Vavilov and Ukhin (1977)], [Cheryl,

Marshall, Burke, Summers and Wolicki (1988)], Section 6.14 of [Messenger and Ash

(1992)], Chapter 5 of [Claeys and Simoen (2002)] and Section 5.4 of of [Holmes-

Siedle and Adams (2002)] (see also references therein).

7.1.3 Generalized Messenger–Spratt Equation for Gain Degrada-

tion of Bipolar Transistors

As discussed in Sect. 7.1.1, the Messenger–Spratt equation [Eqs. (7.13, 7.14)] relates

the particle ﬂuence to the variation of the reciprocal of the common-emitter current

gain for particles ﬂuences suﬃciently large to allow the displacement damage in

the bulk to be the dominant damage-factor. In Section 4.3.1, we have already dis-

cussed that the concentration of Frenkel-pairs (FP ) is expected to be proportional

to the concentration of deep defects acting as recombination-centers, when these

traps are resulting from primary defects mostly created by cascading-displacement

processes. As a consequence, the variation of the reciprocal of the minority-carrier

lifetime [∆(1/τ

)] is expected to depend linearly on FP almost independently of

the type of incoming particle, i.e., to exhibit an approximate NIEL scaling.

Under the assumptions regarding the validity of the Messenger–Spratt equa-

tion (discussed in Sect. 7.1.1) and these latter about the creation mechanism of

recombination centers, we can combine Eqs. (4.131, 7.14) to obtain

∆

≈

τ,i

= λ

, (7.20)

where (as discussed in Sect. 4.3.1) λ is almost independent of the type and energy

of the incoming particle, but depends on the i) the type of substrate, ii) the dopant

concentration (slightly) and iii) the level of compensation. Thus, λ may dep end on

the type (n −p −n or p −n −p) of transistor and VLSI technology (e.g., base width

and doping-proﬁle of the graded base). However, the eﬀect on the gain degrada-

tion [∆(1/β)] is expected to depend almost linearly on FP . Equation (7.20) is the

generalized expression for the common-emitter current gain degradation in bipolar

transistors and it is termed generalized Messenger–Spratt equation

†

[Consolandi,

†

The treatment described in Sect. 4.3.1 and the generalized Messenger–Spratt equation were ﬁrst

derived by Rancoita et al. in [Colder et al. (2001); Codegoni et al. (2004b); Consolandi, D’Angelo,

Fallica, Mangoni, Modica, Pensotti and Rancoita (2006)] (see also [Leroy and Rancoita (2007)]

and references therein).

January 9, 2009 10:21 World Scientiﬁc Book - 9.75in x 6.5in ws-bo ok975x65˙n˙2nd˙Ed

552 Principles of Radiation Interaction in Matter and Detection

D’Angelo, Fallica, Mangoni, Modica, Pensotti and Rancoita (2006)]. This equation

extends the Messenger–Spratt equation (7.13) originally derived for ﬂuences of fast

neutrons. It relates ∆(1/β) to the concentration of Frenkel-pairs created in the si-

licon by the displacement damage and accounted for the energy deposited in NIEL

processes (treated in Sect. 4.2.1). In addition, it predicts that, similarly to ∆(1/τ

)

[see discussion in Sect. 4.3.1], ∆(1/β) is expected to follow an approximate NIEL

scaling. Furthermore, Eq. (7.20) can be rewritten as

∆

≈ k(I

) FP (7.21)

with

k(I

) =

. (7.22)

The ﬁrst exp erimental evidence of the validity of Eq. (7.20) [Eq. (7.21)] was ob-

tained in irradiations with fast-neutrons with ﬂuences up to ≈ 1.2×10

n/cm

, with

C-ions

of 95.0 MeV/amu up to ≈ 10

C/cm

and with

C-ions of 11.1 MeV/amu

up to ≈ 10

C/cm

[Colder et al. (2001)] (see also [Colder et al. (2002)]). ∆(1/β)

was found to follow an approximate NIEL scaling over the full range of the gen-

erated concentration of Frenkel-pairs

‡

[(6.4 × 10

–1.1 × 10

) cm

−3

]. It has to be

noted that for same amount of FPs the D

Ion

(Carbon)/D

Ion

(neutron) dose ratio is

larger than 10

(see Table 4.5 and [Consolandi, D’Angelo, Fallica, Mangoni, Mod-

ica, Pensotti and Rancoita (2006)]). Experimental values of ∆(1/β) as a function of

FP are shown in Fig. 7.3 for collector currents of 50 µA. In the ﬁgure, the data are

for n − p − n transistors with (a) small and (b) large emitter area, and (c) vertical

and (d) lateral p −n −p transistors

‡‡

manufactured according to the HF2BiCMOS

technology [Gola, Pessina and Rancoita (1990)].

As can be seen in Fig. 7.3, these results have been conﬁrmed in irradiations with

Ar- and Kr-ions [Codegoni et al. (2004b); Consolandi, D’Angelo, Fallica, Mangoni,

Modica, Pensotti and Rancoita (2006)]. Irradiations with 9.1 MeV electrons for ﬂu-

ences up to 2 × 10

e/cm

have provided a further conﬁrmation of Eq. (7.20) (see

Fig. 7.3) [Codegoni et al. (2006); D’Angelo, Fallica, Galbiati, Mangoni, Modica,

Pensotti and Rancoita (2006)]. In fact, the ratio D

Ion

(electron)/D

Ion

(neutron) of

the ionization dose of electrons with respect to that of fast-neutrons is larger than

four orders of magnitude for the same amount of FP created by displacement da-

mage (see Table 4.5 and [Consolandi, D’Angelo, Fallica, Mangoni, Modica, Pensotti

and Rancoita (2006)]). It has to be noted that, as discussed in Sect. 4.2.1.4, the recoil

silicon is able to produce a small cascade development for incoming electrons with

The reader can ﬁnd the deﬁnition of kinetic energies per amu in Sect. 1.4.1.

‡

The concentrations of Frenkel-pairs for C-ions were computed by means of the TRIM simulation

code [Ziegler, Biersack and Littmark (1985a); Ziegler (2001)] and for fast-neutrons by the data

available in literature for the damage function (e.g., see Sect. 4.2.1.5).

‡‡

The devices were characterized using an HP4142B modular DC source-monitor and the IC-CAP

code (version 2004 [IC-CAP (2004)] or previous versions) controlled by a workstation.

January 9, 2009 10:21 World Scientiﬁc Book - 9.75in x 6.5in ws-bo ok975x65˙n˙2nd˙Ed

Displacement Damage and Particle Interactions in Silicon Devices 553

Fig. 7.3 ∆(1/β) as a function of the concentration of Frenkel-pairs (FP ) (from [Leroy and Ran-

coita (2007)]; see also [Consolandi, D’Angelo, Fallica, Mangoni, Modica, Pensotti and Rancoita

(2006)]): for n − p − n transistors with (a) small [5 µm ×5 µm] and (b) large [50 µm ×50 µm]

emitter area, and for (c) vertical and (d) lateral p − n − p transistors manufactured according

to the HF2BiCMOS technology [Gola, Pessina and Rancoita (1990)]. The collector current is

50 µA. The incoming particles are fast-neutrons,

C-ions of 95.0 MeV/amu (high-energy),

C-

ions of 11.1 MeV/amu (medium-energy),

Ar-ions of 13.6 MeV/amu,

Kr-ions of 60.0 MeV/amu

and electrons of 9.1 MeV [Colder et al. (2001, 2002); Codegoni et al. (2004b, 2006); Consolandi,

D’Angelo, Fallica, Mangoni, Modica, Pensotti and Rancoita (2006); D’Angelo, Fallica, Galbiati,

Mangoni, Modica, Pensotti and Rancoita (2006)]. The line represents the linear dependence ex-

pected from Eq. (7.20).

this kinetic energy. Furthermore, for any value of the collector current, the slope

of the ﬁtted straight-line to ∆(1/β) as a function of FP [e.g., see Eqs. (7.21, 7.22)]

determines the quantity k(I

) and, thus, the value of

)

2 π k(I

)

(7.23)

as a function of the collector current [e.g., see Eq. (7.19)]. f

)/λ as a function of

is shown in Fig. 7.4 for the irradiated n −p −n and p −n −p transistors. These

curves indicate that i) the cutoﬀ frequencies reach their maxima above 100 µA as

January 9, 2009 10:21 World Scientiﬁc Book - 9.75in x 6.5in ws-bo ok975x65˙n˙2nd˙Ed

554 Principles of Radiation Interaction in Matter and Detection

Fig. 7.4 f

)/λ as a function of the collector current (in A) [see Eq. (7.23)] (adapted from

[Leroy and Rancoita (2007)]; see also [Consolandi, D’Angelo, Fallica, Mangoni, Modica, Pen-

sotti and Rancoita (2006)]): for n − p − n transistors with (a) small [5 µm ×5 µm] and (b) large

[50 µm ×50 µm] emitter area, and for (c) vertical and (d) lateral p −n −p transistors manufactured

according to the HF2BiCMOS technology [Gola, Pessina and Rancoita (1990)].

expected in this technology [Gola, Pessina and Rancoita (1990)] and ii) their ratios

are in agreement with those expected for the diﬀerent types of transistors in this

technology [e.g., ≈ 20 MHz for lateral p − n − p transistors versus (2–6) GHz for

n − p − n transistors]. It has to be noted that below ≈ 100 µA (e.g., see [Colder et

al. (2001); Consolandi, D’Angelo, Fallica, Mangoni, Modica, Pensotti and Rancoita

(2006)]), k(I

) can be approximated by

k(I

) = A

, (7.24)

where A

and n depend on the type of transistor. The values found for the coeﬃcient

n are ≈ −(0.2–0.4) [Leroy and Rancoita (2007)] (see also [Consolandi, D’Angelo,

Fallica, Mangoni, Modica, Pensotti and Rancoita (2006)]) and are in agreement with

those available in literature (e.g., see Section 7 of Chapter 1 of Part II in [Vavilov

and Ukhin (1977)]).

In summary, these experimental data indicate that, at large values of FP , the

dominant processes resulting in gain degradation are well accounted by the processes

of non-ionization energy-loss.

7.1.4 Transistor Gain and Self-Annealing

The defects induced by radiation, as discussed in Sect. 4.2.2, may be unstable and,

thus, their concentration may decrease with time. For instance (e.g., see Sect. 4.3.1),

an annealing eﬀect was observed for the lifetime of minority-carriers, thus decreasing

January 9, 2009 10:21 World Scientiﬁc Book - 9.75in x 6.5in ws-bo ok975x65˙n˙2nd˙Ed

Displacement Damage and Particle Interactions in Silicon Devices 555

the value of K

τ,n

(see pages 32–45 in [Srour, Long, Millward, Fitzwilson and Chad-

sey (1984)], also [Srour (1973)] and references therein). Therefore, the annealing is

expected to cause the (slight) increase of the current gain of the bipolar transistors.

To a ﬁrst approximation, this gain increase exponentially approaches a constant

value with time.

The annealing eﬀect on ∆(1/β) for large emitter area n−p−n transistors [Gola,

Pessina and Rancoita (1990)] has been investigated, at 25

◦

C, from immedia-

tely after the (electron) irradiation up to about 70 months after Argon irradi-

ation [Codegoni et al. (2004b); Rancoita (2005); Consolandi, D’Angelo, Fallica,

Mangoni, Modica, Pensotti and Rancoita (2006); Leroy and Rancoita (2007)] and

is shown in Fig. 7.5 for three values of the collector currents: 1 µA, 50 µA and

1 mA. The irradiations with electrons of 9.1 MeV (2 ×10

e/cm

) and

Ar-ions of

13.6 MeV/amu (10

Ar/cm

) created initially a concentration ≈ 7.2 × 10

−3

of Frenkel-pairs. The data show that i) there is evidence for a long-term annealing

with a decay time of (10.0 ± 2.98) months and ii) the eﬀect of annealing is similar

for devices irradiated with electrons and Ar-ions [Consolandi, D’Angelo, Fallica,

Mangoni, Modica, Pensotti and Rancoita (2006)].

7.1.5 Radiation Eﬀects on Low-Resistivity Base Spreading-

Resistance

The bipolar transistors have a wide range of circuit applications, for example as

input device in fast front-end electronics

††

. However, the degradation of transi-

stor properties resulting from radiation damage may impair the expected opera-

tion of the whole electronic circuit. In Sections 7.1.1, 7.1.3, we have discussed

the decrease of the gain and the increase of the base current of bipolar transi-

stors with increasing ﬂuence. Furthermore from the base lead, a current must ﬂow

through the graded-base region up to the recombination centers, the emitter-base

junction and, also, the emitter region: the overall resistance, involving diﬀerent

parts of the base region, is referred to as base spreading-resistance {e.g., see Sec-

tion 16.4(d) of [van der Ziel (1976)], Section 7.2 of [M¨uller and Kamins (1977)] and

Section 14.4(c) of [Bar-Lev (1993)]}: the displacement damage may also cause a

variation of the eﬀective-doping concentration in the base. In high-resistivity de-

vices (e.g., silicon radiation detectors), the resistivity is largely modiﬁed by irradi-

ation (see Sects. 4.3.4, 4.3.5, 6.8.3). As the doping concentrations used in bipolar

transistors are several orders of magnitude larger, the radiation-induced change in

doping concentration is usually less important, but may not be negligible at large

ﬂuences. For instance when these transistors (see Section 5.1.2 of [Gatti and Man-

fredi (1986)]) are employed as input stage of charge-sensitive-preampliﬁer (CSP),

the increase of transistor base-current and slight variation of the eﬀective doping-

††

The reader may see, e.g., [Gatti and Manfredi (1986); Gola, Pessina and Rancoita (1990); Gola,

Pessina, Rancoita, Seidman and Terzi (1992); Baschirotto et al. (1997)] and references therein.

January 9, 2009 10:21 World Scientiﬁc Book - 9.75in x 6.5in ws-bo ok975x65˙n˙2nd˙Ed

556 Principles of Radiation Interaction in Matter and Detection

Fig. 7.5 Annealing eﬀect at 25

◦

C: dependence of ∆(1/β) for large emitter-area transistors [Gola, Pessina and Rancoita (1990)] on the time (in

months) after irradiation with electrons (2 × 10

e/cm

) and Ar-ions (10

Ar/cm

) (adapted from [Leroy and Rancoita (2007)]; see also [Rancoita

(2005); Consolandi, D’Angelo, Fallica, Mangoni, Modica, Pensotti and Rancoita (2006)]). The initial concentration of Frenkel-pairs is ≈ 7.2 ×

−3

. The lines show the ﬁtted curves for collectors currents exponentially decreasing with time. The currents are 1 µA (dashed line), 50 µA

(dotted line) and 1 mA (dashed and dotted line), respectively.

January 9, 2009 10:21 World Scientiﬁc Book - 9.75in x 6.5in ws-bo ok975x65˙n˙2nd˙Ed

Displacement Damage and Particle Interactions in Silicon Devices 557

Fig. 7.6 R

(in Ω) as a function of the concentration of Frenkel-pairs (in cm

−3

) (adapted from

[Leroy and Rancoita (2007)]; see also [Baschirotto et al. (1997)]). The line is to guide the eye.

concentration of the base spreading-resistance (R

) are among the relevant causes

of degradation of parallel and series noise of the CSP. This occurs even with very

short (≈ 20 ns) shaping-times

∗∗

The values of R

for n −p −n bipolar transistors, manufactured according to

VLSI technology in HF2BiCMOS process [Gola, Pessina and Rancoita (1990)], were

determined after irradiation with fast-neutrons up to a ﬂuence of ≈ 9××10

n/cm

by measuring the noise performance of CSPs with these devices employed as in-

put stage [Baschirotto et al. (1997)]: they are shown in Fig. 7.6 as a function

of the concentration of the Frenkel-pairs (FP ) created during irradiation [Leroy

and Rancoita (2007)]. R

is compatible with the value before irradiation up to

FP ≈ 6 × 10

−3

and is ≈ (70–75)% larger at the largest ﬂuence, i.e., for

FP ≈ 4.3 × 10

−3

, in agreement (see discussion in [Baschirotto et al. (1997)])

with data available in literature (e.g., see Section 5.13 of [Messenger and Ash

(1992)]).

Furthermore, R

is proportional to the base resistivity, thus to the reciprocal

of the carrier mobility [Eq. (6.19)]. For the p- (n-) type base, the mobility µ

(T )

∗∗

The reader can refer, for instance, to [Baschirotto et al. (1993, 1995a,b, 1996, 1997, 1999)] and

references therein.

January 9, 2009 10:21 World Scientiﬁc Book - 9.75in x 6.5in ws-bo ok975x65˙n˙2nd˙Ed

558 Principles of Radiation Interaction in Matter and Detection

[µ

(T )] is given

†

by [Arora, Hauser and Roulston (1982)]:

(T ) = 54.3 × T

−0.57

1.36 × 10

× T

−2.23

1 + 0.88 × T

−0.146

/ [2.35 × 10

× T

2.4

]}

, (7.25)

(T ) = 88.0 × T

−0.57

7.4 × 10

× T

−2.33

1 + 0.88 × T

−0.146

/ [1.26 × 10

× T

2.4

]}

, (7.26)

where T is the temperature (in kelvin) and T

= T/300; N

) is eﬀective

dopant concentration in the p- (n-) type base. Before and after irradiation with

the largest ﬂuence (see [Baschirotto et al. (1997)]), the (p-type) base spreading-

resistance was also determined as a function of the temperature T in the range

283 < T < 343 K. The temperature dependence of R

was observed to be de-

termined by values of µ

(T ) compatible with those calculated using Eq. (7.25) for

dopant concentration of ≈ 10

−3

before irradiation, i.e., with that employed for

the base in this technology, and ≈ 5×10

−3

after the largest irradiation [Baschi-

rotto et al. (1997)].

In Fig. 7.7, a radiation hard fast bipolar monolithic charge sensitive preampliﬁer

(CSP) and its associated monolithic shaper (with 20 ns shaping time) are shown:

both devices are implemented in HF2BiCMOS technology.

7.2 Single Event Eﬀects

As already mentioned, a Single Event Eﬀect (SEE) is due to an individual event

caused by a heavy-ion (i.e., a single charged particle with Z ≥ 2) traversing a semi-

conductor or semiconductor-based device (integrated circuit, power supply, etc.). A

SEE results in a failure of the device. This failure is the consequence of the charge

deposition resulting from the electron-hole pairs generated along the path of the

incoming particle within the sensitive volume of the device (direct ionization). SEE

can also be caused by the interactions (elastic, inelastic scattering or spallation)

of the incoming particle, neutron or proton, with atoms in the device producing

light particles such as protons, deuterons, α-particles, light nuclei such as lithium,

beryllium and heavy recoil nuclei such as magnesium. Most of the recoil species are

heavier than the original particle and have large ionization capabilities. Then, the

recoil particles can also generate tracks of electron-hole pairs along their path (in-

direct ionization). The charge deposition mechanisms are the same as in the case of

directly ionizing ions. In both situation, the energy deposition via direct or indirect

ionization creates electric charges which, if enough electric charge is created, may

modify the functional state of the device, leading to errors. The minimum electric

charge that causes the change of state in the device is called the critical charge (Q

)

and a SEE occurs if the collected charge exceeds Q

. The critical charge is formed

†

These expressions agree (within ±13%) with experimental data in the temperature range (250–

500) K and up to a dopant concentration of ≈ 10

−3

(for additional data see [Arora, Hauser

and Roulston (1982)] and references therein).

January 9, 2009 10:21 World Scientiﬁc Book - 9.75in x 6.5in ws-bo ok975x65˙n˙2nd˙Ed

Displacement Damage and Particle Interactions in Silicon Devices 559

Fig. 7.7 (left) Photograph of a monolithic CSP: four preampliﬁers are located on a single die, sharing the biasing network; the chip dimension

is 1.62 × 1.17mm

. (right) Photograph of the monolithic shaper associated to the CSP: the circuit area is 1.08 × 0.93 mm

. The photographs are

reprinted from Nucl. Instr. and Meth. in Phys. Res. B 155, Baschirotto A. et al., A radiation hard bipolar monolithic front-end readout, 120–131,