Contents

1 Deep levels and DLTS 2

1.1

Deep states 1.1.1

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The trap signature . . . . . . . . . . . . . . . . . . . . . . . . . . Electron trap and hole trap . . . . . . . . . . . . . . . . . . . . . Majority and minority carrier traps . . . . . . . . . . . . . . . .

2 3 4 5 5 6 8 8 9 10 11 12 13 13 14

Transient response 1.1.1.1 1.1.1.2 1.1.1.3

1.2

Basis of transient depletion experiment for a majority carrier trap . . . . . . . . . 1.2.1 Rectangular transient charge model . . . . . . . . . . . . . . . . . . . . . . 1.2.1.1 1.2.1.2 The trapped charge spatial distribution . . . . . . . . . . . . . . Charge transient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.3

Deep Level Transient Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.1 1.3.2 1.3.3 The two-point subtraction rate window principle DLTS & quantum wells 1.3.3.1 1.3.3.2 Diculties aecting DLTS experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A brief overview of quantum wells

DLTS probe of a Schottky barrier with a single quantum well . .

Chapter 1

Deep levels and DLTS

1.1 Deep states
An electron and a hole in a semiconductor may recombine, if an electron transition from the conduction band to the valence band occurs, or be generated if the electronic transition is in the opposite direction. We may classify the Generation-Recombination processes in various ways: we speak of direct or band-band GR if the carrier transition occurs in a single process, of indirect or trap assisted GR if the carrier transition is via energy levels within the bandgap. If the energy involved in the transition is given by/transfered to lattice phonons it is non-radiative or thermal, it is radiative or optical if the energy is balanced through a photon; nally let's hint to Auger mechanism, in which energy is losed or acquired by another carrier. Radiative and non-radiative recombination processes compete between each other, but clearly in light-emitting devices one tries to favour the former process. On the left we show the four possible nonradiative processes via a generic trap level

Et .

Since the transition probability is exponentially related to the energy jump, it is aected by the level position within the bandgap. Obviously if the level is next to the valence band the (1) and (3) processes are more probable than the (2) and (4) ones; if the level is next to the conduction band it happens vice versa; Figure 1.1:
Emission and capture processes at a deep level: (1) hole emission to the valence band, (2) electron capture from the conduction band, (3) hole capture from the valence band, (4) electron emission to the valence band.

only when the level is close the midgap all the processes have the same probability. Because of generation of an electron-hole couple implies the occuring of processes (1) and (4), while its recombination the processes (2) and (3), then we speak of GR centre when the level is nearby midgap situated. The most important indirect

GR mechanism is the non-radiative or Shockley-Read-Hall (SRH) one; there also exist indirect radiative mechanisms. We will not treat extensively the SRH theory; let's only say that, in a doped semiconductor, if the deep level is at or close to the midgap, in low level injection condition (: the excess carriers concentrations are the carrier capture section, lifetime is rate. If than the majority carrier concentration), set that

is

min0

< v >is the carrier average thermal velocity then the minority carrier 1 Nt <vmin >min , i.e. it depends on the deep level density Nt and minority
then the non-radiative lifetime is minimized if trap level is at or

carrier characteristic quantities; the recombination rate is limited by the minority carrier capture

n < vn >= p < vp >

close to the midgap energy: deep levels are eective recombination centers if they are near the middle of the gap.

CHAPTER 1.

DEEP LEVELS AND DLTS

Defects in the crystal structure are the most common cause for deep levels within the bandgap, and so for non-radiative recombination. These defects include unwanted foreign atoms, native defects, dislocations, and any complexes of defects, foreign atoms, or dislocations; in compound semiconductors, native defects include interstitials, vacancies, and antisite defects. All such defects form one or several energy levels within the forbidden gap of the semiconductor. Owing to the promotion of non-radiative processes, such deep levels or traps are called luminescence killers.

1.1.1 Transient response

Let's consider a deep state with energy concentration

Et

and

Nt

for a n-type semiconductor;

the four basic processes which determine its population dynamic are the capture and emission (from and to the relative band) for both carriers. Let a deep centre be exposed to a ux of Figure 1.2: if

Electron and hole capture and emission processes for an electron trap in a n-type semiconductor

n < vn >electrons per unit area per unit time: nt deep states out of Nt are occupied by t the number of electrons captured pt Nt nt unoccupied states is nt =

electrons at any instant, in a successive time interval by the

n < vn > n(Nt nt )t,

for

the electron capture cross section; so the electron capture

rate per unoccupied state is

cn

1 nt = n < vn > n Nt nt t cp .

(1.1) Due to

and an analogue expression holds for the hole capture rate per occupied state the minority carrier lifetime: for a p-type semiconductor

recombination of electrons and holes at the deep centre, the carrier capture rate is related to

1 n

lifetime associated with electron capture at the deep state is

1 n nt , thus the non radiative t nnrad = n < vn > (Nt nt ).

The electron occupancy of the deep state is determined by the competing processes of emission and capture: electrons are emitted/holes are captured at the emitted/electrons are captured at the occupancy is

nt

occupied states, holes are

pt

unoccupied states, and the net change rate of electron

dnt = (cn + ep )(Nt nt ) (cp + en )nt dt

for

(1.2)

en ,ep

the electron, hole emission rates.

Let's determine solutions to the above, so called, rate equation when density

Nt

the net doping

Nd Na ,

and thus the free carriers concentrations

n, p

are not inuenced by the GR

processes and can be considered constant. If we set the boundary condition then its general solution is

nt (t = 0) nt (0),

nt (t) = nt () [nt () nt (0)] exp

set that the steady state occupancy is

nt () a = Nt a+b

(1.3)

1 a+b is the time constant, dened the parameters a cn + ep , b en + cp . The initial nt (0) nt () occupancy Nt is usually set experimentally: when it diers from the equilibrium value Nt ,

then the occupancy relaxes exponentially to this latter with rate constant are:

a + b; two special cases

CHAPTER 1.

DEEP LEVELS AND DLTS

for

nt (0) = Nt ,

then

nt (t)

decays

nt (t) =
for

t Nt a + b exp a+b

(1.4)

nt (0) = 0,

then

nt (t)

increases with time

nt (t) = nt () 1 exp

(1.5)

We must be careful and distinguish between the rate of change of rate of emission of electrons from the trap is emitted.

nt , and the emission and capture

time dependent, while

rates: for instance, when the only process is that of electron emission, then the instantaneous

dnt = en nt (t), dt

en

is the

emission rate per trapped electron, i.e. the probability per unit time that a single electron is

1.1.1.1

The trap signature

At thermal equilibrium, it holds: the steady state condition:

dnt dt

= 0,

the principle of detailed balance: the rates for a process and its inverse must be equal and balance in detail (so that it does not occur a net electron transfer from a band to another):

en nt = cn (Nt nt ) ep (Nt nt ) = cp nt
By combining the above requirements with the rate Equation (4.2), we nd that the thermal equilibrium trap occupancy is

ep cn nt = = cn + en Nt ep + cp
statistics, and so

(1.6)

on the other hand, at thermal equilibrium, the electron occupancy is ruled by Fermi-Dirac

nt Nt

1 1+g exp
Et EF kT

, where

g0 g1 is the degeneracy ratio,

g0

being the

degeneracy of the trap level

Et

when it is electronless and

g1

that when the trap level is occupied

by one electron. Thus we can relate the capture and emission rates for both carriers:

en = g exp cn
Since

Et EF kT

cp ep

(1.7)

g is of order of unity, then, roughly speaking, if EF > Et cn > en , ep > cp , while if EF < Et cn < en , ep < cp ; if EF = Et cn en , ep cp : strictly speaking en = cn when EF = Et kT ln g . Because of Equation (4.1), the capture rate cn is dependent on the doping density; the analogue is valid for cp too. Instead, the emission rates and the capture cross sections are
intrinsic characteristics of the trap. We remember that for a non degenerate semiconductor holds the Boltzmann approximation:

E n = NC exp ECkT F

E p = NV exp EFkT V

Unlike shallow hydrogenic levels, deep state centres may be occupied by more than one electron. Fynally, through equations (4.1) and (4.7), we derive

en = n < vn > g exp

EC Et kT

(1.8)

CHAPTER 1.

DEEP LEVELS AND DLTS

and by similar proceding

ep = p < vp >

1 Et EV exp g kT

(1.9)

We may determine the temperature dependence of the emission rates; in the case of electron emission:

< vn >= NC = 2MC

3kT m , for

the conduction band eective mass,

2m kT 3/2 , for h2

MC

the number of conduction band minima, ,

if we assume then

(T ) = exp E kT

en (T ) T 2 na exp
A plot of

Ena kT

(1.10)

1 T , called trap signature, is a straight line; experimental data for most traps t the above curve over many orders of magnitude of en . Let's observe that the apparant
versus

e ln Tn 2

Ena EC Et + E is not the trap level, and that the apparent capture cross section na is to the extrapolated (T = ) value of the capture cross section through the degeneracy factor; both the identications are valid only if EC Et is temperature independent. We may catalogue the traps according the Ena and na values.
activation energy 1.1.1.2 Electron trap and hole trap

Because of Equations (4.8)&(4.9), we nd that

en = ep

if

Et = E1 EF i +

kT ln 2

p < vp > n < vn >

EC +EV 2

(1.11)

We remember that the intrinsic Fermi level is

EF i

kT 2

ln

NV NC

, so the energy

E1
i.e. i.e.

is located near the midgap energy; we call the deep level an electron trap if it is it is located in the upper part of the gap (Et it is located in the lower part of the gap.

> E1 ),

while it is an hole trap if it

en > ep , is en < ep ,

1.1.1.3

Ma jority and minority carrier traps

Let's make a short resume of what just seen: the relation

en

by the position of the trap level the ratio that

ep , which establish if the trap is electron-kind or hole-kind, is determined Et respect to the energy E1 (Equation (4.11)),

Nt

e c depends on the relative positions of Et , EF (Equation (4.7)). But, if we assume the net doping density then EF is controlled by the doping, i.e. it is close to:

 

the conduction band edge for a n-type semiconductor the valence band edge for a p-type semiconductor

EF > Et ,

EF < Et .

By combining the two points above, we nd that: an electron trap (en

cn > en

ep ) in a n-type semiconductor (cn > en , ep > cp ) is such that 1 ep > cp nt (t) relaxes with time constant = cn +en , and the equilibrium
nt () Nt

occupancy is

cn cn +en ; if

cn

en ,

then at the equilibrium

nt = Nt .

CHAPTER 1.

DEEP LEVELS AND DLTS

a hole trap (en

ep ) cp

in a p-type semiconductor (cn

ep < cp nt (t)
nt () Nt

relaxes with time constant

< en , ep < cp ) is such that cn < en 1 cp +ep , and the equilibrium occupancy is nt = 0 pt = Nt .

ep ep +cp ; if

ep ,

then at the equilibrium

Figure 1.3:

Majority carrier traps

Let's generalize: we speak of majority carrier (electron for n-type material, hole for p-type one) trap when

emaj

emin

and

Et

is close to the majority carrier band edge; the majority

carrier equilibrium occupancy is equal to

carrier trap for:

Nt

if

cmaj

emaj .

By analogy, we speak of minority

an electron trap in a p-type semiconductor; now only

nt (t) relaxes with time constant = ep ep +en +cp the relaxation process involves the exchange of carriers with both bands, and the equilibrium value pt = Nt is reached when both hole capture and electron emission
dominate over hole emission. an hole trap in a n-type semiconductor; and the equilibrium occupancy is

cn is negligible (because cn n) nt () 1 ep +en +cp , and the equilibrium occupancy is Nt =

nt (t) =

relaxes with time constant

1 cn +ep +en ,
is

nt () Nt

cn +ep cn +ep +en

the equilibrium value

nt = Nt

reached when electron capture and hole emission dominate over electron capture.

Figure 1.4:

Minority carrier traps

Again

Et

is close to the appropriate (: minority carrier) band edge. The distinctive feature of

minority carrier traps is that the equilibrium occupancy is established through interaction with both bands: it follows that the trap occupancy cannot be perturbed by its equilibrium value by simply acting on the majority carrier population, as in the case of majority carrier traps.

1.2

Basis of transient depletion experiment for a majority carrier trap

Let's examine the sequence in a transient depletion experiment for a Schottky barrier on a n-type semiconductor, the latter containing a donor-like majority carrier trap with energy level With no bias applied, deep states in neutral material (: depth at zero bias applied) are occupied because

Et .

x > x0 ,

set that

x0

is the depletion

cn > en .

CHAPTER 1.

DEEP LEVELS AND DLTS

Figure 1.5:

Conduction band diagram at zero bias applied

By applying suddenly a reverse bias

VR ,

within the enlarged depletion region the capture

process of electrons from the conduction band to the trap level is turned o ; the traps empty, and the emitted electrons are rapidly swept out of the depletion region by the associated eld.

Figure 1.6:

Conduction band diagram at reverse bias applied

(Equation 3.21) as the dierence xn x1 , x1 the point in which the trap level crosses the Fermi level; if the bias applied is so large that xn , the net space charge density during the emission process is (x) = e [Nd + Nt nt (t)]. We also remember that we can express the depletion depth xn in function of the total potential across the junction (Equation (2.6) for Nd Nd + Nt nt (t)), then we substitute in the parallel
We remeber we dened the transition distance being plate expression (Equation (3.6)), nally we obtain the time dependent capacitance

C (t) = A
set that

s e 1/2

[Nd + Nt nt (t)]1/2 V 1/2 C () 1

nt (t) Nd + Nt

1/2

C () is the steady

state (since

In the dilute trap concentrantion limit

Nt

nt () = 0 within the depletion region) capacitance. Nd ( nt Nd ) the square root term in the above nt (t) C () 2Nd

equation can be expanded, so that the change in capacitance is

C (t) C (t) C ()
Because the initial occupancy is

nt (0) = Nt , then nt (t) follows a decaying law expressed by Equation (4.4). For a majority carrier trap cmaj > emaj are the only not-negligible capture and emission rates, so a + b = cmaj + emaj ; in our case (: electron trap) the only possible process being that of electron emission a + b = b = en . Finally nt (t) = Nt exp (en t) and thus: C (t) Nt = exp (en t) C () 2Nd
(1.12)

If the reverse bias is removed then the empty traps are relled by electron capture; as already underlined, we use a hybrid depletion approximation, since we neglect the free carriers within the depletion region ar regards the space charge density, but we consider them as a trap relling mean.

CHAPTER 1.

DEEP LEVELS AND DLTS

The above equation, which states that the capacitance increases exponentially with time as carriers are emitted from the trap, is the basis of the transient capacitance techniques and the DLTS ones: according to it, the time constant of the capacitance transient gives the thermal emission rate, and

C0 C (0) =

Nt C () 2Nd

(1.13)

1.2.1 Rectangular transient charge model

Let's briey resume the assumptions made for determining Equation (4.13): all the traps lled prior to the starting of the emission process, a circumstance which cannot apply if the capture cross section is very small, the emission process occurs all over the depletion region, while the traps within belong to the transition region below the Fermi level), we adopted the depletion approximation, instead of considering the distance over which the trap occupancy changes near

x2 x0

from the surface are never lled (they are always above the Fermi level), and those which

xn x1 < x < x n

are never empty (they are always

xn .

In particular, the assumption that the emission occurs all over the depletion region does not aect seriously the relationship between the transient time constant and the emission rate, instead it does inuence that between the transient amplitude and the deep states concentration: for trap proling experiments it is thus important to understanding the factors controlling the extent of the region over which the traps emit.

1.2.1.1

The trapped charge spatial distribution

This model, of the spatial distribution of trapped charge producing the transient response in a majority carrier trap experiment, is in many respects an extension to the depletion approximation; from it we will determine a bettered version for the capacitance transient. Let's consider a Schottky barrier on n-type semiconductor, the latter containing an electron trap level

n Et

and a hole one

p Et ,

at zero

applied bias; let the depletion layer edge be

x0 . cn

Figure 1.7:

Energy band diagram for zero applied bias

x > x0 it is en and so the traps are fully occupied. For x < x0 , the the free carrier concentration decreases with distance from x0 (since Equation (3.7) for xn x0 ), and so the capture rate (Equation (4.1) cn (x) n(x)); instead
Since Equation (4.7), for the emission rate does not depend on distance,

nt (x, t = 0)) also decreases with distance from x0 in the same manner as n(x). Provided that LD x0 , the transition in trap occupancy may be considered to occur abrutply at depth x2 x0 , i.e. the
thus nally from the equilibrium trap occupancy Equation (4.6) we desume that location where the trap level crosses the Fermi level at zero bias applied. From Equation (4.7) we determine that

cn (x2 ) en ,

and so again for Equation (4.6) we get

cn (x), nt (x) vary fastly with distance, being the degeneracy factor (x 1 cn (x2 ) = en , ntNt2 ) = 2 , and for uniformly doped material the two relationships
by the bias applied.

nt (x2 ) Nt ; since both 2 g 1, it is usual to consider

are independent

CHAPTER 1.

DEEP LEVELS AND DLTS

Now let's apply a reverse bias

VR

to the Schottky barrier, the de-

pletion region is enlarged up to again for from

xn ;

x < xn

the free carrier con-

centration decreases with distance

xn

(Equation (3.7)), and also Mutatis mutan-

the capture rate.

dis, we can repeat all what said be-

Figure 1.8:

x2 , now for x1 xn . In particular, in the region x < x1 , where the trap level
fore (zero bias case) for is above the Fermi level, Equation (4.7) implies that

Energy band diagram for reverse applied bias

the traps within relax by electron emission alone, while in the region

cn < en , and x > x1 cn > en , and

so so

the traps remain lled; we regard the transition between the two regions occuring abruptly at

x1 .

We remember that at zero bias applied the traps were lled for

x > x2 (see

Figure (4.7)). In

conclusion, not in the full width of the depletion region, but in the overlapping of the two above regions,

x2 < x < x 1 ,

it holds that

nt (0) = Nt , a = 0, b = en nt (t) = Nt exp (en t).

Figure 1.9:

Trapped electron densities at zero and reverse bias applied; the dotted line shows the rectangular approximation to nt (x, t)
1.2.1.2 Charge transient

Drawing uniform densities, space

on

the

trapped and deep get

charge for net state

spatial density (Figure (4.9)), donor we can the

charge density (x, t) = e {N + (x) + [Nt (x) nt (x, t)]} in the steady state under reverse bias (Figure (4.10a)), in the steady state (i.e. at the end of the lling process) with lower bias (Figure (4.10b)), during the emission process (Figure (4.10c)); we remember that at

N + (xn )

is the uncompensated donor density

xn ,

dened by Equation (3.22). steps have already

The

following

been seen in past occasions; by suitably apply Equation (3.8), we derive a relationship between Figure 1.10: the resulting changes in provided that

(x, t) and xn (t), V (t) ; dnt

into

Nd , Nt

are constants,

we can determine the general equation relating changes per unit time in trap occupancy

CHAPTER 1.

DEEP LEVELS AND DLTS

10

the

x2 < x < x 1

region, to changes per unit time in depletion depth and bias:

x2 (t) x2 dnt dxn e dx1 dVR 2 N + xn = + (Nt nt (t)) x1 (t) 1 dt dt dt 2 dt

If the reverse bias is constant so it is the transition distance equation in density

(1.14)

dVR = 0; Nt

we take

dx1 = dxn

(because of

Nd

is constant, and

see equation (3.21)) in equation (4.14), and we get a non-linear is small compared with the net background doping

xn (t).
then

When the trap density

N +,

N + Nd

(under the hypothesis the sample is uncompensated),

x1

becomes

eectively constant, so we nally obtain the linear equation:

Nd xn

dxn x2 x2 dnt 1 dC 1 x2 x2 dnt 2 = 1 = 1 2 2 dt 2 dt C dt 2Nd xn dt

dxn xn

(1.15)

Now we specify that we determine

xn = C

A C()

dxn = CA dC 2

xn by measurement of the high frequency capacitance, i.e. = dC . Finally, by integrating Equation (4.15), by taking C Nd ),
we come to the following form of the capacitance

to be eectively constant (for

Nt

transient:

C (t) 1 x2 x2 nt (t) nt () C (0) t = 1 2 2 exp C () 2 xn Nd C () nt (t) = Nt exp (en t) steady state one nt () = 0, so

(1.16)

We know that, for majority carrier-electron trap, the occupancy transient during the emission process is governed by

nt (0) = Nt

1 en , being the initial occupancy such that

the

C (0) 1 x2 x2 Nt 1 2 2 C () 2 xn Nd
In conclusion, what we have just showed is that in the dilute limit (Nt transient is exponential, with time constant When the bias applied is so large that

(1.17)

Nd ) the capacitance 1 , and aplitude to Nt . Equations (4.16) and en (4.17) are exact within the depletion approximation, with Nt , Nd uniform, set that Vbi is constant. x1 xn xn ,
and set that

x2

xn ,

then equation

(4.17) reduces to the simpler form (4.13).

1.3

Deep Level Transient Spectroscopy

This technique is used to observe thermal emission from majority carrier traps by means of capacitance transients. The negative biased test diode is pulsed (for a short time interval

tf ill ) with positive tension Vf ill , consequently the depletion region is reduced from width the traps within

W to that W W W W

and are

lled with majority carriers. The parameter tf ill aects the lling level of the traps, while the relative magnitude of Figure 1.11:

VR + Vbi , Vf ill

the probed re-

gion, because the whole (VR

Repetetive lling and reverse bias pulse sequence, to which corresponds the diode capacitance transient

Vf ill ) Vf ill )

or only a section (VR of

+ Vbi + Vbi >

is thus repleted; in the

former case the instrument is probably overloaded with the high zero

CHAPTER 1.

DEEP LEVELS AND DLTS

11

bias capacitance. When the bias applied is restored to the negative value val

VR

(for a time inter-

tR ),

the trapped carriers are emitted with rate

en ,

and we consequently observe a transient

capacitance:

C (t) = C () + C0 exp
This sequence is repeated periodically, with time period measurement, the temperature is slowly increased.

(1.18) during the DLTS

tP tf ill + tR ;

1.3.1 The two-point subtraction rate window principle

At the core of the DLTS technique is the principle of the two-point subtraction rate window, whose im-

plementation we illustrate through the use of the double box-car or standard dual-gate signal averager system. Every capacitance signal, following the periodical bias pulse, is sampled by two gates set at times and

t1

t2

from the onset of the tran-

sient (t Figure 1.12:

= 0),

which is equivalent to

integrate the product of the signal and the weight function

Exemplication of the principle of the two-point subtraction rate window: above the capacitance transient sampled at times t1 , t2 , below the weighting function w(t)

w(t).

The

rst gate t1 should be set so to avoid the overload recovery part of the capacitance signal. With a small gate width

, t2

under the hypothesis

that transients are exponential, the box-car steady signal output is

S( ) C (t1 ) C (t2 ) = C0 exp

t1

exp

(1.19)

Figure 1.13:

Double box-car signal output

As we can desume from the Equation (4.19), and observe from Figure (4.13), the signal

CHAPTER 1.

DEEP LEVELS AND DLTS

12

output is zero for Set that

majority carrier traps,

t1 or t2 , not negligible only if t1 , t2 . S( ) > 0 for minority carrier ones.

For

t1 < t2 , S( ) < 0

for

1 en , the peak output is deterdS(T ) mined by the condition 0 = = dS(en ) dT den dS(en ) den = 0 since it dT , which is reduced to den den holds dT = 0 T (because of Equation (4.10)),

and thus it occurs when

ref
for

t2 t1 ln t2 t1

(1.20)

ref

the reference time constant of the rate

window preset

t2 t1 , which depends only by t2 , t1 ; ref at the beginning of the DLST measurement, the peak temperature Tpeak is characteristic of the trap, and the emission rate

Figure 1.14:

en (Tpeak )

is equal to

1 ref .

By repeating the

temperature scan with dierent values of sets of values

ref ,

Above deep level spectrum produced by a rate window with reference time constant ref , below the correspondent point on the trap Arrhenius plot

(Tpeak , en (Tpeak ))

are measured,

from which we can generate an Arrhenius plot of

ln

T2 en

versus

1 T , and so determine the trap parameters

Ena , na .
t2 t1 , in fact

Let's observe that

Speak S(ref )

is function only of the ratio

Speak C0 exp
thus changing of

ln 1

exp

ln 1

ref such that is constant then the peak height is independent from ref ; because Speak C0 , through Equation (4.13) we can estimate the trap concentration; a more correct evalutation takes into account the transition distance (Equation (4.17)
Since the transient is not recorded directly, it is possible to work with time constants as short as a few ms; the repetition time can also be short ( that

10ms), so that en

is eectively constant over

several cycles and thus the output signal is averaged without prolonging the scan time provided

ref

values are not too long.

Alternative rate window methods are those employing the lock-in amplier or an exponential correlator system.

1.3.2 Diculties aecting DLTS experiments

The DLTS rate window technique draws on the hypothesis of exponential transient decaying, but this does not occur when:

Nt

is not so small compared with the net doping density

N +, en
becomes dependent

the thermal emission rate is dependent upon electric eld, and then

upon position within the depletion region due to the spatial variation of the electric eld, in a semiconductor alloy, the thermal emission rate at a particular trap site is dependent by the local crystal composition, and so its value is not uniquely determined through the sample. All above mechanisms lead to broadening of DLTS spectrum; by recording spectra for dierent pairs of

t1 , t2

values such that

ref

is constant, the peak should occur at the same temperature

in case of single exponential decay, while this does not happen with non-exponential one.

CHAPTER 1.

DEEP LEVELS AND DLTS

13

Also an interfacial layer can aect a DLTS measurement; due to a thin oxide which remains when the barrier metal is evaporated, or to other surface treatments during the diode fabrication, it can introduce peaks into the spectrum associated with interfacial states or modifying the capacitance transient originated by traps in the semiconductor bulk.

1.3.3 DLTS & quantum wells

1.3.3.1 A brief overview of quantum wells

Single or double heterostructures can create potential wells in the conduction and/or valence bands, which can conne carriers in the direction orthogonal to the well, leaving them free to move in the two other directions (i.e. lel to the heterojunction); the plane paralIf the potential well

is very narrow the allowed energy levels of the conned electrons and holes will be quantized: the resulting structure is called a quantum well Figure 1.15: (QW). Within the QW, assuming the eective mass approximation, the carrier wavefunction is solution of a 3D time-independent Schrodinger equation

Example of N-i-P double heterostructure in direct bias, which operates like the active region of a light-emitting diode or a semiconductor laser

+ V ( R ) ( R ) = E( R ); in this special case V ( R ) = V (z), which suggests that ( R ) exp (ik r) u(z), set that k (kx , ky ), 2m
2

2 r (x, y), u(z) is the solution of the 1D equation 2m z V (z) u(z) = u(z), E 2m k2 . Let it be un (z) the eigenfunction corresponding to the energy eigenvalue n for the given V (z), then the associated solution of the 3D original problem is n,k ( R ) exp (ik r) un (z); in 2 En (k) = n 2m k2 we recognize, for any set n value, the dispersion relation for a free 2D
electron gas with the energy ground state shifted to
band. Let's note that for

n : we call every parabola electric sub-

for

< E <

1 < E < 2 the allowed states are only in the lowest subband, while 2 there are allowed states in the subbands labelled by n = 1, n = 2; we also

note that for the same

value, the energy partitioning is dierent according to subbands: in the for the motion along the z axis is higher, so it is lower the transverse

higher subband the energy

kinetic energy. As energies increase there are more and more subbands from which to choose, and electrons with the same total energy have dierent transverse wave vector

k.

Finally, we remeber that the density of states of a 2D free electron gas is the step function

m (E), so each subbands contributes a step of equal height, starting at 2

of states is a staircase-like function, with jumps at the energies

n ; the total density m (E n ). n : g2D (E) = 2 n

The density per unit area of electrons trapped in a conduction band QW is found as usual by integrating, over all energies, the product of the density of states occupation function

g2D (E)

and the Fermi-Dirac

n2D

fF D (E): m m = fF D (E)dE = kT 2 n 2 n

ln 1 + exp
n

EF kT

(1.21)

An analogue reasoning is applicable to determine the density per unit area of holes trapped in a valence band QW.

CHAPTER 1.

DEEP LEVELS AND DLTS

14

Figure 1.16:

The electron wave functions (a), the dispersion relation (b) and the density of states (c) in the case of innitely deep square well of width 10nm
1.3.3.2 DLTS probe of a Schottky barrier with a single quantum well

DLTS is also applicable in studying the electron emission from a quantum well. The localized conned state in the quantum well, corresponding to the energy level separation of width

1 ,

may be regarded as

a giant trap, the thermal activation energy of the emission process interpreted as the energy

Eb

1 between the conned state and the top of the barrier.

the surface such that it lies within the reverse bias depletion zero bias one (xo

Let's suppose to use the DLTS technique upon a Schottky barrier with a single quantum well region (xw well.

Lz , located at xw below < xn ), but outside the

< xw ).

When the bias is reduced to zero the carriers into the vicinity are entrapped by the quantum The electron capturing bends the nearby energy bands because of the localized charge, The carriers density near untill the lling process is interrupted since electrons are repelled.

the well is so reduced, consequently the capture rate starts falling, reaching a steady state value because of the compensating thermal emission process of electrons out of the well.

Figure 1.17:

Conduction band diagram immediately after reducing to zero the bias applied, and when the equilibrium is reached; d is the width of the depletion regions around the QW

CHAPTER 1.

DEEP LEVELS AND DLTS

15

By applying a reverse bias, the depletion region is emptied by the free carriers; the electrons early entrapped into the well are emitted by thermal process and then swept out, producing a capacitance transient.

Figure 1.18:

Charges entrapped by the well, thermally emitted at reverse bias applied

We note that the electrons rst emitted when the reverse bias is just applied are those whose quantized energies are greater than

1 , which lay under the Fermi energy level at zero bias;

because of the energy levels are dierent, the emission rates are also dierent and the initial decay is not exponential, but it is very fast and probably not detectable experimentally. What we observe is the emission of those electrons whose energy is the well ground state have similar emission rates, and so the slower decay is now exponential. By application of the principle of detailed balance, the electron capture process, from the barrier states (corresponding to energies above the band edge energies stay below

1 , which all

Eb )

to the well states (whose

Eb ),

must be compensated by the electron emission process, from the well The fraction of occupied and unoccupied states in the well and

states to the barrier states. per unit volume is

barrier is determined by the Fermi energy level. The total capture rate of electrons into the well

Rcapture cw pw = w < vn > nb pw

set that

nb

is the electron density in the barriers ,

assume that only the lower subband is occupied, so

pw the well unoccupied states density. We E pw mw z 1 b (1 fF D (E)) dE 2L

EEF kT

mw 2 Lz
that

1 (Eb E1 ), the factor Lz mw is the carrier eective

arising from the fact that the density is per unit volume, set mass; since

Eb EF
3/2

kT ,

to the electron density in the

barriers does not apply any more the Fermi-Dirac statistics, but the Boltzmann one, so

nb

E Nb exp EbkT F , band edge Eb . The

set that

Nb = 2

2mb kT h2

is the states density per unit volume at the

total emission rate of electrons from the well per unit volume is

Remission ew nw
set that

nw

is the well occupied states density; by remembering equation (4.21), and holding the

approximation From

ln (1 + x) x

Rcapture =

x 1 Remission , we determine

because of

EF

kT ,

then

nw

mw 2L z

kT exp

EF kT

ew = w < vn > 2
and assuming

2mb kT h2

3/2

Eb kT

exp

Eb kT

(1.22)

constant (remembering that

1

< vn > T 1/2 )

we derive that a plot of

Eb 1 versus T should have slope k

e ln Tw 2

CHAPTER 1.

DEEP LEVELS AND DLTS

16

Finally, by assuming that such that

x1 x2 = Lz

and

Lz xw , so that the volume in which the traps are observable is x1 + x2 2xw , then we derive the capacitance transient equation C (t) xw Lz nw0 = exp (ew t) C () x2 Nd n
(1.23)

set that than

nw0
if

is the initial carriers density into the well. The initial steady state carrier density for distances greater

is dened by the charge neutrality at large distance requirement, i.e.

wqw

wqw

is the width of the depletion region around the QW. Let it be

Nd

the donor

density in the barriers, then, by application of the depletion approximation, the charge per unit area inside the QW is equal to the charge per unit area in the depleted regions around the QW

nw0 Lz = 2wqw Nd , set that Eb EC = [EC ( 1 w )] (EF

we get

2 s wqw = e2Nd (Eb EC ) 2 1 ) (EC EF ) ( w being the

from Equation (3.5), and QW energy ground state)

follows by observing the steady state part of Figure (4.17). There are important features which make dierence between a QW and an usual deep state: the initial density of trapped electrons is ruled by the doping density of the material around the QW , and not by the trap density; as the bands bend with time, the density of free carriers near the QW reduces; the carriers are not emitted by each trap from the trap level of states; a constant emission rate

Et ,

instead from a continuum

ew is well dened only after as many electrons are emitted

1 . It is not clear

that all the remaining trapped electrons can be considered having energy capture cross section

if variation in the band bending around the QW during emission changes the eective

w ,

making

ew

time dependent.