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

560 Principles of Radiation Interaction in Matter and Detection

by energy deposition in a critical volume which depends on the sensitive volume of

the device. This sensitive volume, which is the charge collection region, is smaller

than the physical volume. The evolution of modern technology goes toward the

creation of faster and less power consuming devices of increasingly small sensitive

volume and higher density circuits, achieving now sub-micron technology with an

increase of the number of memory elements. High density circuits with their smaller

feature size have less capacitance, lower operating voltage and the information is

stored with less charge or current. The device dimensions are now comparable to

and even lower than the range of the striking particles which generate electron-hole

pairs within the depth of the device. The occurrence of SEE is very sensitive to the

incidence angle of the striking particle upon the circuit device as particles with low

incidence angle may deposit more charge in the sensitive volume. This continuous

reduction in size and multiplication of cells decreases Q

and the amount of charge

required to generate SEE and opens this possibility to particles of lower energy. It

is observed that the relative system soft error rate (SER) for α-particle and neutron

eﬀects increases by three orders of magnitude when decreasing the range of sub-

micron process technology from 0.25 µm to 0.05 µm [Semico Research Corporation

(2002)].

A SEE is inﬂicted by a single particle, ion (direct ionization) or nucleon (indi-

rect ionization) responsible for damages, which can be temporary or permanent in

contrast to the permanent damage in electronics inﬂicted by an integrated radiation

dose, which is usually delivered by a high particle ﬂux, particularly near particle

accelerators or in space.

7.2.1 Classiﬁcation of SEE

One classiﬁes SEEs into three categories: Single Event Upset (SEU), or simply

called upset, causing temporary damage, Single Event Latchup (SEL) resp onsible

for temporary or permanent damage and Single Event Burnout (SEB) leading to

permanent damage.

Eﬀects caused by SEEs to particular devices depend on the type and energy

of the incident particle and the characteristics of the devices (material, geometry,

thickness, etc.). Then, the eﬀect of SEE are device speciﬁc and dep ends on how the

resulting corrupted information aﬀects the system including the device.

SEU occurs in memory circuits and logic circuits (DRAM, SRAM, microproces-

sors, etc.). Semiconductor memory devices store data as the presence or absence of

charge carriers in storage wells deﬁne the logic state ”0” or ”1”, respectively. The

amount of stored charges typically varies from ∼ 0.3 × 10

up to ∼ 3 × 10

elec-

trons. The number of electrons which diﬀerentiates between ”0” or ”1” is the critical

charge Q

. In a SEU, the incoming particles release in the proximity of a memory

cell an amount of charges exceeding Q

. This produces a voltage spike, which in-

duces a change to the cell status by a ﬂip 1 → 0 or 0 → 1 generating an error in a

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 561

bit. It is called a soft error (SE) when the content of the memory device is changed

without damaging the device. This occurrence makes the stored data unreliable and

the device must be rewritten with correct data. SEU may also change randomly a

program of a computer or confuse a processor to the point that it may crash. In

general the damage caused by SEU are non-permanent since their eﬀect is canceled

by data rewriting or system rebooting.

For high density circuits, charge collection by diﬀusion or charge sharing from a

single ion may lead to multiple-bit upsets. Multiple bit upsets (MBU) is a SEU that

brings simultaneously corruption of several memory cells (change of logic state)

physically adjacent. MBU can be induced by diﬀusion of charge carriers in the

sensitive volume before their collection or by ions striking at low incidence an-

gle [Reed et al. (1997)]. For instance, this is the mechanism responsible for MBU

in DRAM [Zoutendyck, Edmonds and Smith (1989a,b); Zoutendyck, Schwartz and

Nevill (1989)]. This eﬀect can be removed by a combination of error-correcting

codes, that work on a word-by-word basis [Bossen and Hsia (1980)], and lay-out,

that prevent physically adjacent bits from belonging to the same word of mem-

ory [Crafts (1993)]. Occurrence of MBU is expected to become more serious with

the use of increasingly smaller geometries by advanced very large scale integration

(VLSI) processes [Zoutendyck, Schwartz and Nevill (1989)].

In a SEL [Bruguier and Palau (1996); Dodd (1996); Pickel (1996)], the energy

released by a single particle is injected into parasitic p − n − p − n structures

inherent to CMOS technology (proximity of NMOS and PMOS transistors), possibly

triggering a short circuit and at the end resulting into the destruction of the device

if the feeding voltage is not switched-oﬀ immediately. A SEL may be cleared by a

power oﬀ-on reset.

The SEB is a condition that can cause device destruction due to high current

state in a power transistor. In particular, the passage of a particle through the

dielectric of a power MOSFET may create a plasma in between silicon and the grid

with subsequent diﬀusion towards the Si-SiO

interface and increase of the electric

ﬁeld in the gate oxyde leading to destructive burnout. This non-reversible process

is called a single-event gate rupture (SEGR). For GaAs circuits, SEL and SEB do

not occur but GaAs devices are more sensitive to SEU than Si devices [Karp and

Gilbert (1993)].

This review will focus on the study of SEU, and soft error (SE), in particular.

SEEs occur randomly since ions, protons or neutrons of various energies and

incidence angles hit the circuit randomly. Due to their random nature it is diﬃcult to

predict their occurrence simply based on theoretical analysis. However, it is possible

to calculate the rate of SEE events from the combination of measurements of their

cross sections (the meaning of this cross section will be deﬁned in Sect. 7.2.8) for a

speciﬁc device exposed to a given radiation environment with models describing the

physical mechanisms taking place inside the device. One needs to know the ﬂux or

ﬂuence of the irradiating particles. As the sensitivity of a device to SEE depends on

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

562 Principles of Radiation Interaction in Matter and Detection

its geometry, technology, size, thickness, circuit and memory densities (and speed),

rate estimates may have impact on device design.

SEE occur both in space, atmospheric and terrestrial environment.

7.2.2 SEE in Spatial Radiation Environment

Anomalies in satellites orbiting around the Earth or in spacecraft operating in

deeper space are the result of their exposure to the spatial radiative environment

made of several components (a detailed review of these components is presented

in Sects. 4.1.2, 4.1.2.4, 4.1.2.5). The primary cosmic rays includes galactic parti-

cles entering the solar system. The galactic ﬂux of primary cosmic rays is very low

∼ 5 particles cm

−2

−1

. These cosmic rays are made of very energetic charged par-

ticles, dominated by protons (≈ 87%) with an energy range from 100 MeV up to

1 GeV, α-particles (≈ 12%) and heavy-ions representing ≈ 1% of this radiative com-

ponent with energy ranging from 1 MeV up to 10

TeV. The abundance of heavy ions

rapidly falls with the atomic number. Iron, Z = 26, is the heaviest most signiﬁcant

component, abundances of species heavier than iron are 2–4 orders of magnitude

smaller than iron. The large electric charge of these heavy ions gives them an ioniza-

tion power much higher than that of protons and α -particles. Primary cosmic rays

of the highest energy may hit the Earth. The ﬂux of these galactic cosmic rays in

the solar system is modulated by the solar activity (Sect. 4.1.2.3). Large solar wind

shields the solar system from these particles. The solar wind is made of particles

(electrons, protons and α -particles) of energy (< 100 keV) much lower than that of

galactic particles. The Sun has an eleven-year cycle in average with four years of

low activity and seven years of high activity during which solar ﬂares of variable du-

ration (from a few hours up to several days) occur (Sect. 4.1.2.1). These solar ﬂares

produce a ﬂux of particles made of protons (10

p cm

−2

−1

) of energies from a few

MeV up to several hundreds of MeV and heavy ions (10

particles cm

−2

−1

) with

energies ranging from 10 MeV up to 100 GeV. The charged particles trapped in the

Van Allen belts are another component of the spatial radiation environment. The

belts exist in two regions: the inner zone which extends over a region close to Earth

about 1.0 Re up to . 2.4 Re (Re = 6371 km is the radius of the Earth) and the

outer zone which extends over a region & 2.4 Re) up to 9.0 Re. The inner belt is

ﬁlled by electrons, protons and some heavy ions. Electrons and protons are pro-

duced by the β-decay (n → p e

−

¯ν

) of neutrons produced by the interaction of

cosmic rays with atmospheric particles [Singer (1958); Hess (1959)]. The electrons

and protons are trapped by the Earth’s magnetic ﬁeld. The neutrinos, being electri-

cally neutral and with a very small mass, evade the belt into the cosmos. The inner

belt reaches a minimum altitude of 250 km, locally, above the Atlantic Ocean oﬀ

the Brazilian coast, the South Atlantic Anomaly (SAA) as the result of the oﬀset

between the Earth’s geographical and magnetic axes. The inner zone is dominantly

populated with protons of energies in the (10–100) MeV range, their population is

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 563

aﬀected by the solar wind and ﬂares. The trapped protons ﬂux varies between 1

to 2×10

particles cm

−2

−1

with a maximum around 2.0 Re. In the inner zone, the

trapped electrons have energies . 5 MeV. The electron content of the inner zone

may vary considerably in time due to phenomena like solar wind or like nuclear ex-

plosions in upper atmosphere. A low ﬂux of trapped ions of rigidities corresponding

to a few MeV/nucleon and around (3–4) Re, mainly He and O with traces of C and

N, has been observed. These particles are believed to be extracted from the up-

per layers of the atmosphere during solar storms. SAMPEX data [Cummings et al.

(1993)] have also shown the existence of belts included in the inner zone, containing

nuclei such as N, O (of ≈ 10 MeV/nucleon) and Ne [of (10–100) MeV/nucleon]. The

ﬁlling mechanism of these SAMPEX belts is likely due to the interaction of the

anomalous cosmic rays with the Earth atmosphere (in anomalous cosmic rays there

are more α-particles than protons and much more oxygen than carbon while in the

galactic cosmic rays there are more protons than α-particles and equal amounts

of oxygen and carbon). Recently, belts below the inner belt zone, at ∼ 1.0 Re,

have been observed by the AMS experiment [AMS Collab. (2000a,b,d, 2002)] with

a relatively large content of electrons, positrons, deuterium and

He with rigidi-

ties of 10

–10

MeV/nucleon. The origin of the AMS belts is believed to result

from the interaction of primary cosmic rays with the atmosphere. The outer b elt

zone is mainly populated with electrons of energies ∼ 7 MeV (up to 10 MeV), with

a ﬂux of an order of magnitude higher than in the inner zone and peaking at

about 5 Re. The trapped electrons ﬂux in the outer zone varies between 10

and

2 × 10

particles cm

−2

−1

. This outer zone is ﬁlled by the solar wind.

7.2.3 SEE in Atmospheric Radiation Environment

The atmospheric environment originates from the cosmic ray cascades in the Earth’s

atmosphere and are a concern for avionics. This environment results from the inte-

raction of cosmic rays with atmospheric atoms through ionization or nuclear re-

actions. The isotopic content of the atmosphere is predominantly made of oxygen

(21.8%, Z = 8), nitrogen (76.9%, Z = 7) and argon (1.3%, Z = 18). These nu-

clear reactions are mostly induced by the primary protons which dominate (≈ 87%)

the incident cosmic ray ﬂux and generate secondary particles through elastic and

inelastic scatterings, mainly neutrons (90% and more), pions, kaons, muons (from

charged pion and kaon decays), photons (from neutral pion decays and photopro-

duction reactions), muons from pion decays and electrons from muon decays and

gamma conversion. The neutron component is of interest for avionics although these

neutrons can penetrate the whole atmospheric depth. However, the ﬂux of particles

(neutrons, protons, pions and kaons) is decreasing with altitude and latitude. For

instance, the ﬂux of neutrons is attenuated by their collisions with atmospheric

atoms, thus decreasing the neutron ﬂux at lower altitudes. The Earth’s magnetic

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

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

564 Principles of Radiation Interaction in Matter and Detection

ﬁeld traps the cosmic particles which decreases the probability of their interactions

with the atmosphere atoms. As the ﬁeld lines are closer at the poles, trapping is more

eﬀective, and cosmic particles can approach closer to the Earth’s surface at polar

latitudes compared to equatorial latitudes. Thus, the ﬂux of cosmic particles, in par-

ticular neutrons, is larger at polar latitudes compared to equatorial latitudes. The

neutron peak ﬂux is about 4 neutron cm

−2

−1

at 20,000 m. The neutron ﬂux is re-

duced to about 1/3 and 1/400 of the peak ﬂux at 10,000 m and on the ground,

respectively [Normand and Baker (1993); Taber and Normand (1993); Sims et al.

(1994)]. The SE rate (SER) due to atmospheric neutrons (neutrons with energy

higher than 1 MeV) is estimated for a range of sub-micron CMOS SRAM circuits

following the formula [Shivakumar, Kistler, Keckler, Burger and Alvisi (2002)]:

SER(numb er) ∼ φ × A × e

−Q

, (7.27)

where φ is the neutron ﬂux with energy > 1 MeV, in particles cm

−2

−1

, A is the

area of the circuit sensitive to particles strikes, in cm

. The critical charge of the

SRAM cell, Q

(in fC), depends on the circuit characteristics, namely the supply

voltage and the eﬀective capacitance of the drain nodes. Q

(in fC) measures the

magnitude of the charge generated by a particle strike, it represents the charge

collection eﬃciency of the circuit. The SER depends on the ratio Q

and

on the area of the sensitive region of the device exposed to striking particles and

then decreases with the size (length × width) of the device. This model is used

in [Hazucha and Svensson (2000)] to evaluate the eﬀect of device scaling on the

SER of memory circuits. In [Hazucha and Svensson (2000)] the conclusion is reached

that SER/chip of SRAM circuits should increase at most linearly with decreasing

feature size.

7.2.4 SEE in Terrestrial Radiation Environment

The terrestrial environment is made of the particles that ﬁnally hit the Earth. Par-

ticles of very high energy (> 1 GeV) can induce cascades which can penetrate the

atmosphere down to the sea-level. These particles are usually galactic particles

but particles of the highest energy produced during active Sun periods may also

reach the Earth level and increase the intensity of cosmic rays at the Earth’s sur-

face. However, the additional magnetic ﬁeld created by the solar wind around the

Earth acts as a shield against cosmic rays and reduces their sea-level ﬂux. The

most abundant particles are muons. Their large number results from the decay

of pions [π

(π

−

) → µ

(µ

−

) ν

( ¯ν

)] generated in the cascading process initia-

ted by cosmic rays collisions with Earth-atmosphere nuclei (Sect. 9.11.2). The

muon has a lifetime of 2.2 µs and decays into electron and neutrino-antineutrino

[µ

−

(µ

) → e

−

) ¯ν

(ν

) ν

( ¯ν

)] creating a sea of electrons and positrons in the

atmosphere even close to sea-level. Neutrons are the next most abundant particles

of the terrestrial environment. Being electrically neutral, they do not lose energy to

the sea of electrons in the atmosphere and may reach the sea-level. Protons equally

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 565

present to neutrons during the ﬁrst steps of the cascading process, being electri-

cally charged constantly lose energy to the atmospheric electrons and their number

is depleted compared to neutrons at sea-level.

7.2.5 SEE produced by Radioactive Sources

Another radiation environment to be faced by integrated circuits (IC) is generated

by radioactive sources (α-particles emitters such as Uranium [

238

U] and Thorium

[

232

Th]) present in their package molding compounds. It has been shown that the

α-particles packaging materials can produce upsets [May and Woods (1979)].

238

has a radioactivity of 6.85 × 10

−7

Ci g

−1

238

U decays to stable

206

Pb with the

emission of 8 α-particles with energies ranging from 4.15 to 7.69 MeV. The α-

particles having this range of energy can travel from 9.9 to 25.0 µm (average range

of 14.3 µm) [Ziegler, Biersack and Littmark (1985b)] in an alumina ceramic with

a density of 3.85 g cm

−3

. Only 25% of the generated α-particles escape from the

ceramic. Taking into account this escape factor, assuming that i) at equilibrium all

the decay isotopes of

238

U will have the same radioactivity as the parent

238

U, ii)

an average range 14.3 µm of α-particles in ceramic, one can relate the

238

U concen-

tration (in ppm) and the α-particle ﬂux in ceramic [Woolley, Lamar, Stradley and

Harshbarger (1979)]:

1 ppm

238

U/g of ceramic = 0.996 α cm

−2

hour

−1

. (7.28)

232

Th has a radioactivity of 1.11 × 10

−7

Ci g

−1

232

Th decays to stable

208

Pb with

the emission of 6 α-particles with energies ranging from 3.95 to 8.8 MeV. The α-

particles having this range of energy can travel from 9.2 to 31.2 µm (average range

of 19.1 µm) in an alumina ceramic [Ziegler, Biersack and Littmark (1985b)]. Per-

forming a calculation similar to that done for

238

U, one obtains the relation between

the

232

Th concentration (in ppm) and the α-particle ﬂux in ceramic:

1 ppm

232

Th/g of ceramic = 0.162 α cm

−2

hour

−1

. (7.29)

One can observe that the

238

232

Th α-particle ﬂux ratio is ∼ 6.0. This shows that

the α-particles produced by

238

U are the dominant cause of soft errors from IC

packaging compounds.

Considering the combined presence of

238

U and

232

Th in the packaging

compound, α -particles are emitted at energies ranging from 3.95 up to ∼

8.8 MeV. Alpha-particles with these energies can travel up to 57 µm in silicon sub-

strates (density of 2.33 g cm

−3

) [Ziegler, Biersack and Littmark (1985b)] and pro-

duce up to 2.4 × 10

electron-hole pairs in the silicon component (with a depth

≥ 57 µm) corresponding to a deposited charge up to 0.38 pC which may be possibly

higher than Q

for that device. Then, radioactive contamination of the circuit may

aﬀect its operation by causing soft errors. However, the number of the emitted α-

particles falls with the increase of their energy, decreasing the number of α-particles

capable to deposit a charge ≥ Q

. In addition α-particles produced near the surface

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

566 Principles of Radiation Interaction in Matter and Detection

Table 7.1 Stopping power and projected range in several material used in cir-

cuits for incident protons [dE/dx(p) and R(p)] and α-particles [dE/dx(α) and

R(α)] (from [Leroy and Rancoita (2007)]). The calculations have been made using

data from [Ziegler, Biersack and Littmark (1985b)]. The density of Si, SiO

and

GaAs are 2.33, 2.27 and 5.33 g cm

−3

, respectively. E

ion

is the electron-hole pair

generation energy.

Circuit Material Energy dE/dx(p) dE/dx(α) R(p) R(α) E

ion

(MeV) (keV/µm) (keV/µm) (µm) (µm) (eV)

Si 4 16 162 148 18 3.62

SiO2 4 17 178 138 16 17.0

GaAs 4 26 236 95 13 4.8

of the device will escape, contributing only a charge deposition very small compared

to Q

. Then, the soft error rate is decreasing with increasing Q

. If Q

is very large,

it may be that none of the emitted α-particles will have an energy large enough to

create a suﬃcient amount of electron-hole pairs to achieve a charge ≥ Q

and no soft

errors will be generated. If Q

is very small (let us say smaller than 0.05 × 10

) all

emitted α-particles will generate soft errors. The number of soft errors per α-particle

can be estimated [May and Woods (1979)] from knowing the device cross section (to

be discussed later), the depletion region depth, the charge collection areas used to

store charge, and the energy spectrum of incident α-particles. The sensitivity factor

S is deﬁned to be the fraction of all α-particles hitting the active cell at a speciﬁc

incidence angle (which depends on the geometry and α-particle ﬂuxes of the various

parts of ceramic package, possibly including lid and seal ring). An example of error

rate calculation in DRAM’s and CCD is given [May and Woods (1979)] (example

for 4K test structure):

SER = soft error rate = A × Φ

× S, (7.30)

where A is device active area, Φ

the α-particles ﬂux and S the sensitivity fac-

tor. For the speciﬁc example A = 0.027 cm

, Φ

= 3.8 α/(cm

h) and S = 0.015

errors/α giving an error rate of 1.5 ×10

−3

errors/h = 150 percent/(1000 h) in good

agreement with the observed soft error rate of 200 percent/(1000 h) obtained over

200,000 device hours of testing [May and Woods (1979)]. Puriﬁcation techniques are

needed to minimize the amount of α-particles source in materials. Nowadays, low-α

Mold Compounds are used [0.001 α/(cm

hour)] and produce FIT rates at least an

order of magnitude lower than FIT (Failures-In-Time, see Sect. 7.2.8) rates produced

by standard Mold Compound [0.04 α/(cm

hour)] [Actel (2006)]. In principle, the

SER component caused by cosmic rays can be separated from the component due

to α-particles emitted by the radioactive sources present in the IC packaging mate-

rials and the electronic noise in the measuring device by the measurement of SER

at several altitudes above sea level and repeating the measurement underground

which should largely eliminate the cosmic rays neutrons and protons ﬂuxes. The

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 567

−

diffusion

funneling

recombination

charges drift

under

electric field

(depletion region

and its extension)

Fig. 7.8 Direct ionization (from [Leroy and Rancoita (2007)]). The striking particle, incident at

angle θ, is an heavy ion (Z ≥ 2).

measurements at several altitudes above sea level allow the investigation of the al-

titude dependence of the cosmic rays eﬀect on IC. An example is provided by the

“Nitetrain” experiment where the presence of rather short-lived

210

Po, daughters

of radon gas, and long-lived daughters of

238

U and

232

Th on the chip (DRAM) is

taken into account together with cosmic rays striking the chip. The SER is expressed

as [O’Gorman (1996); O’Gorman et al. (1996)]

SER = A

exp

−t/t

1/2

+ B + C, (7.31)

where A

is the initial SER due to α-particles emitted by

210

Po (rather short-lived,

1/2

= 138 days), B is the SER contribution of the α-particles emitted by

238

U and

232

Th (long half-lives) and C is the SER contribution from cosmic rays. The SER

has been measured at four locations: altitudes of 0.1 km (2 measurements), 1.6 km,

3.1 km and 0.2 km underground [O’Gorman (1996); O’Gorman et al. (1996)]. For

the underground measurement, C can be set to zero in good approximation, since

cosmic neutrons and protons are stopped by the rock shielding of the underground

facility. The insertion of the SER data from the two sea level measurements and

the underground measurement into Eq. (7.31) along with the time of each of the

measurements provides a system of three equations which can be solved to determine

the three unknowns A

, B and C. That study has shown that the eﬀect of cosmic

rays is a strong function of altitude and that cosmic rays are a dominant cause, over

α-particle events, of soft errors observed in devices based on NMOS, CMOS and

bipolar technologies [O’Gorman (1996); O’Gorman et al. (1996)]. This observation

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

568 Principles of Radiation Interaction in Matter and Detection

Fig. 7.9 Current pulse shape I(µA) as a function of the collection time, with time constant T =

0.1, 0.3 and 0.5 ns (from [Leroy and Rancoita (2007)]; for the computed curves see also [Srinivasan,

Tang and Murley (1994)]). The pulses are normalized to Q = 100 fC total charge.

is important since ”historically” the α-emitters contamination of package molding

compounds was thought to be the main cause of SEU in IC. These results show

that memory circuits should be designed with the goal of immunity from upsets by

alpha-particles and cosmic rays.

7.2.6 SEE in Accelerator Radiation Environment

Another source of SEE in the terrestrial radiation environment comes from the ex-

posure of the experiments electronic devices to the radiation environment created

by the operation of particle accelerators. Electronic devices located close to a beam

or to an interaction point have shown SEE vulnerability. Observation of SEE oc-

currences have been reported by the CDF experiment [Tesarek et al. (2005)] at

the Tevatron Collider. Failures of power supplies feeding readout electronics and

crates, loss of control function, data corruption etc. occurred in systems located

on the detector or close to it when beams were in operation. Similar failures are

anticipated at the LHC at even higher scale, since the high rate head-on collisions

of 7 TeV protons at high luminosity (luminosity peak of 10

−1

−2

) will create

an unprecedented high radiation ﬁeld (see Sect. 4.1.1). For instance, leakage out

of calorimeters of cascades generated by protons and pions produced by proton-

proton interactions will copiously produce neutrons and recoil nuclei through their

subsequent hadronic interactions with surrounding materials and equipment. Neu-

trons and heavy nuclear recoils will interact with silicon components in detector

and readout electronics, possibly causing SEE. In particular, heavy ions of rather

low energy (less than 10 MeV) and therefore of limited range produced locally in

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 569

an electronic chip or in an IC may induce a SEE (see Sect. 7.2.7.2). SEE will aﬀect

all kinds of semiconductor-based electronics components and devices used by the

experiments at LHC such as SRAM, DRAM, FLASH memories, microprocessors,

DSP, FPGA and logic programmable state machine and devices. Microelectronics

featuring high density of devices and sub-micron technologies increasingly used in

these experiments are especially SEE-sensitive.

7.2.7 SEE Generation Mechanisms

SEEs are caused by the release of charge by ionizing radiations in a semiconductor

device via two mechanisms depending on the striking particle species: direct ioniza-

tion if the striking particle is an heavy ion (Z ≥ 2, such as an α-particle or an iron

ion, for instance) and indirect ionization, i.e., ionization by heavy-ions secondaries

produced by the interactions of the striking particle, proton, neutron or pion, with

atoms of the device.

7.2.7.1 Direct Ionization

A charged particle (heavy ion, Z ≥ 2) traversing a semiconductor (e.g., a reverse-

biased p − n device in a memory cell or ﬂip-ﬂop) will loose its energy through

interactions with the material. The energy lost is primarily due to the interactions

of the ion with bound electrons in the material, causing an ionization of the material

and a dense track of electron-hole pairs. It is known that electron-hole pairs gener-

ation is an early process occurring within (2–3) µm of an α-particle track (e.g., see

Chapter 14 of [Leighton (1959)]) that will have a length of 18 µm in silicon for a

4 MeV α-particle. This α-particle will lose energy at a rate of dE/dx ≈ 162 keV/µm

in silicon. The early process of electron-hole generation and the high rate of energy

loss (see Table 7.1) of the incoming ion in the circuit materials (large stopping

power) are resp onsible for direct ionization. The total ionization produced in terms

of electric charges depends on the initial particle energy and on the average energy

needed to generate electron-hole pairs (3.62 eV for silicon, see Table 7.1). For a

4 MeV α-particle, 1.1 × 10

electron-hole pairs are created in a silicon depth of

18 µm. There are three contributions to the charge collected at the circuit nodes:

(1) when the particle path crosses the depleted region of the junction (Fig. 7.8), it

creates electron-hole pairs which will be separated by the junction electric ﬁeld

and the charge carriers (electrons and holes) will be collected at the correspon-

ding nodes generating a current pulse of short duration (Fig. 7.9) with a rapid

rise time, the prompt component of the current pulse.

(2) The excess charge created along the path causes a perturbation of the elec-

tric ﬁeld along the ionizing path (funneling) and the electric ﬁeld conﬁned to

the depletion region extends into silicon beyond the depletion region. Then,

the electric ﬁeld acts on a larger volume and permits the rapid collection of