Characteristic energy losses of injected low energy electrons in ionic solids

Characteristic energy losses of injected low energy electrons in ionic solids

J Phw Chem Solidr. IY?J. Voi WI. pp. t6-fhh Pcrgamon Press. Pnnted m Great Britam CHARACTERISTIC ENERGY LOSSES OF INJECTED LOW ENERGY ELECTRO...

453KB Sizes 0 Downloads 42 Views

J Phw

Chem Solidr.

IY?J.

Voi

WI.

pp.

t6-fhh

Pcrgamon Press.

Pnnted m Great Britam

CHARACTERISTIC ENERGY LOSSES OF INJECTED LOW ENERGY ELECTRONS IN IONIC SOLIDS TIMOTHY HUANGand W. H.

Department of Chemistry

and the Radiation (Received

Laboratory,

University

19 J~tly 1974; in reuisedform

HAMILL

of Notre Dame, Notre Dame, IN46556.

U.S.A.

IO Ocrober 1974)

Abstract-Characteristic energy losses of O-18 V electrons have been measured for thin films of alkali and alkaline earth fluorides as well as NaBr, RbBr and RbCI. In addition lo losses in the optical region there are also several losses at lower energy for each system, some of which may be due lo triplet exciton states. Results for the fluorides, in particular, can be correlated with each other and with optical data.

1. IhTRODtiLTION

Characteristic

energy

losses

One procedure for further testing the capabilities of this procedure is to examine characteristic loss spectra at resonance in the energy range below core excitations of cations. Examination of a series of lattices with a common anion should reveal common structure. The spectra of the five alkali iodides were examined previously[ I]. They are very similar but uniformly poorly resolved except for halogen doublet structure. Since the spectrum of KF was best resolved of the halide ion lattices examined previously, two series of fluorides have since been examined, three alkalis and three alkaline earths. Three additional alkali halides (RbCI, RbBr and NaBr) have been added for completeness. (Only the hygroscopic f.c.c. alkali halides have not yet been examined.) The fluoride energy loss spectra contain more detail than their optical absorption spectra, besides adding new structure. This is most important for low-lying, optically inaccessible states.

at resonance

for electron impact on thin-film alkali halides in the range O-35eV have been observed for energies below the lowest optical absorption as well as for higher energy[ 11. For some of these systems the intrinsic luminescence has been similarly excited and excitation peaks correlate with loss peaks, both in the region of optical absorption and below it [2,3]. Corresponding experiments with several Tl’doped alkali halides provide evidence for host-sensitized activator luminescence[?d]. Again the excitation spectra correlate with characteristic loss spectra, some of the excitations occurring at energies below the first optical exciton. Emission spectra, band widths, lifetimes and activation energies of both undoped and doped systems agree with those for uv-excited single crystals[3]. There is no evidence that thin-film defects contribute measurably to energy loss spectra. The range of very low energy electrons in solids is not known, but it is certainly very much less than -10’ atomic diameters. A value of about three diameters or less is often assumed[5]. Consequently, defects should not contribute measurably to energy losses which are therefore characteristic of the lattice. In two of the systems examined previously vacancies were detected by a-center luminescence (although not by energy losses), but it was readily removed by annealing[3]. The measurement of characteristic energy losses in solids by low energy electron impact is somewhat limited by low resolution. There is no alternative procedure in the sub-optical range, however, because of spin and symmetry optical selection rules which also apply to fast electron impact. Another possible limitation can arise from multiple loss events, but they have not yet been detected for alkali halides by the methods used in the present work. Still another possible limitation of the method is charge trapping, but this is undetectably small in the undoped insulator thin films used in this work.

2. EXPERlMJ3TAI. apparatus and experimental procedure have been described[l,4]. Comparison of the electron current I, transmitted by the sample at incident energy eV, with the backscattered current Ib at a retarding potential V, has shown that the spectra dL/dV, vs V, and dG/dV, vs V, convey the same information[l]. Electrons which undergo resonant losses are trapped by the potential well of the solid due to the positive bulk electron affinity and cannot backscatter to vacuum. They migrate to the anode and so contribute to I,. The cathode potential was modulated by 0.3 V at I kHz from the reference channel of a PAR Model l24A lock-in amplifier. The incident electron current was 10m9A and the transmitted component was amplified and X-Y plotted as dL/dV,. The zero of electron energy was established for each sample from the first optical exciton energy. Samples were evaporated in situ from an outgassed tungsten-wire evaporator, mounted on a rotary feedthrough, onto a stainless steel substrate at -300°K in a vacuum system maintained at 10-PTo~. Films are estimated to be -200 A thick. The spectrum was scanned, The

*The Radiation Laboratory of the University of Notre Dame is operated under contract with the U.S. Atomic Energy Commission. This is AEC Document No. COO-38-962.

MI

T. HUAKG and W. H. HAMILL

662

the evaporation was repeated, and the spectrum was scanned again to ensure film continuity. The spectrum was unchanged as to peak position by repeated evaporations. Peak amplitudes changed slightly because the transmitted current decreases weakly with increasing film thickness. All samples of ultrapure quality were obtained from Alpha Products, Ventron Corp. 3. RESULTS

The quality of electron characteristic energy loss spectra depends considerably upon the emission I - V characteristic of the electron gun. This can be demonstrated most easily by measuring I,, I*, and (I, + I,,) vs V, on an appropriate target. Such results appear in Fig. l(a) for evaporated gold on stainless steel. The gradual decrease in (I, + I,,) arises from large-angle back scattering out of the collecting system as I,, increases. The spectrum df, /d V, vs V, recovers loss peaks due to surface plasmon resonances at multiples of 6.1 eV, which are

6-

much too weak to detect in Fig. I(a), but no other structure. Consequently, the gun introduces no artifacts. The effect of a thin film of an ionic solid on a gold substrate on the I - V characteristics is illustrated by the results for BaF2 in Fig. I(b) on the same voltmeter energy scale. The point of inflection for I, vs V, is at the same incident electron energy as for gold, i.e.. there is no potential barrier. Backscattered electrons analyzed by d&,/d!/, vs V, show loss peaks at -0.6, 4.3, 6.5, ll.OeV and higher, but not well resolved. The inconstancy in (G + I,,) is due to loss of an approximately constant fraction (-20%) of I,, scattered out of the collecting system. The characteristic energy loss spectra of LiF, NaF, CaF*, SrK, BaF*, RbCI, NaBr and RbBr appear in Fig. 2(a)-(j) where 1, is the electron current transmitted by the film at a corrected incident energy eV,. The energy scale was calibrated from the energy of the first exciton band using the optical absorption spectra of Milgram and Givens for LiF[6], those of Eby et al. for the other alkali halides[7], and those of Tomiki and Miyata[S] for the alkaline earth fluorides, all at ambient temperature. Loss spectra measured at -70°K were slightly better resolved, but otherwise the same as those measured at

5-

a P 2

x

(a)

LiF

4-

3-

J

2

4

6

6

Electron

10

12

Energy

(eVJ

14

16

Fig. 2a.

Fig. I. (a) The transmitted and backscattered electron currents I, and I, for a gold target. (II) Same as the preceding with a film of BaFZ. Both energy scales are uncorrected for incident electron energy, eV,.

0

2

4

6 Electron

8 IO 12 Energy (evil Fig. 2b.

14

16

Energy losses in ionic solids

663

IfI

(cl CaF2

I

)

2

7

I 4

I 8

6

Electron

I

,

IO

12

Energy

(eV,

\,

I

14

BaFz

1

0”

16

1

Lll I

2

4

6

8

Electron

IO

Energy

12

(eVi

I4

16

1

,:,,,, Fig. 2f.

Fig. 2c.

(g) BaF,

!

2

4

6

6

Electron

IO

Energy

12

(eVi

I4

I6

0

2

4

6

8

Electron

)

IO

Energy

I2

I4

I6

(eVi)

Fig. 2g.

Fig. 2d.

(e) SrF2

(h) RbCl

:b,‘\:’ x4

01

I

I

,

2

4

6

l

8

Electron

IO

Energy

Fig. 2e.

I2

( eVi)

I4

I6

)

I

I

I

I

2

4

6

8

Electron

I IO

Energy

Fig. 2h.

I

I

I

I2

14

16

(eVi)

T. HUANG and W. H. HAMILI

664

0.1 eV on the average.

Additional

electron

injection

of

2 x IO-’ C at 10 V caused peak shifts averaging +0.4 V. (A single scan of the spectrum

-2x

1O‘“C and no

measurable

damage is lo be expected.)

injects

There was no

appreciable

or systematic change in peak heights, nor did

new peaks appear. Even

though

substrate

are

possibility

of energy-dependent

tributing

evaporated also

films

on a polycrystalline

polycrystalline,

there

electron

to the measured structure,

derivative

spectrum. This possibility

is still

the

diffraction

particularly

con-

for the

was tested systemat-

ically by measuring the spectra for angles of incidence of the electron beam, referred 10 the normal. at 0.30 and 45”. The

relative

peak

intensities

changed

somewhat,

as

appears in Figs. 2(c, d) as an example, but there was no change in the peak position Electron

Energy

(evil

(i.e. energy).

instances a second-derivative

In a very few

spectrum improves

resolu-

tion. An example appears in Figs. 2(i. j) for BaF?.

Fig. 2i.

The results in Figs. 2(a)-(j) are summarized The arrangement (jl

exciton

RbBr

is arbitrary.

in Table I.

except for the first optical

energies which appear on one row, beginning

at

12.9 eV for LiF. Other peak energies have been organized by estimate with the assumed correlated excitations

in the

same row. The arrangement depends both on energy and, rather

roughly.

arrangement

on intensity.

The

plausibility

of this

rests largely on the more numerous

for the fluorides.

Data for KF, published

results

previously[l],

have been included.

\

An attempt results

with

possible,

has been made to correlate other

optical

data for

the present

single crystals.

measurements

Whenever

have been used. These

data appear in parentheses in Table I. Entire series of data have been used without

omissions, as far as possible. in

the energy range of interest. Single events have occasionI

I

I

I

I

I

I

I

2

4

6

8

IO

12

14

16

Electron

Energy

ally been included

(evil

detects

Fig. 2. Electron

ion

lattices at 300°K

characteristic energy loss spectra for halide from the transmitted electron current I, electron energy eV,. The angle of incidence of beam is 0” unless shown otherwise. Peak positions summarized in Table I.

at the are

It should be noted that the exciton present optical

work

thus

eliminating

transitions

absorption

provide

measurements.

no new information.

energies,

in

the

possibility

of

shallow

energies for the

have been zero-shifted

to coincide

with

By themselves,

they

Possible alternate

parentheses,

indicate

comparing data for a well-defined -3OO“K,

all possible

with adequate resolution.

Fig. 2j.

incident electron

to fill an apparent gap in the record,

since no one technique

the

exciton

limitations

transition

of

from different

types of measurement.

electron trapping. If deeply trapped electrons contributed to space charge, the film potential

would

slowly

This would

and somewhat

irreversibly.

J. DISCL’SSION

have changed shift the

The

observation

zero of energy from scan to scan, but no such effect was

by low-energy

observed. Moreover,

questions

electron trapping cannot account for

of previously

electron

unreported energy losses

impact on thin-film

about contributions

from defects,

at energies below

and did not change position

peaks are present on the first scan of an electron impact

after repeated

spectrum.

scans of a given film. The possibility impact

induced

of progressive defects

damage due to electron-

was examined

throughout

this

work. Since damage could not be detected under normal experimental presented.

conditions, After

only

one

the measurements

BaF: the spectra db/dV,

example

will

be

of Figs. 2(f,g)

on

vs V, were measured at many

values of V, from 3 lo 18 eV. The peaks at 0.4, 1.4,3.6,6.8

defects

According

to

in ionic crystals

self-trapped

Smoluchowski

lo

electron

impact,

of a they

thin films of ionic repeated

tests

disclosed no electron trapping in these experiments.

Also,

but

abrupt changes in I, due to charge trapping at - IO ’ A and

vs V, on the 17th scan of the

scan rates of I V/set

spectrum

those in Fig. 2(f) by less than

been observed with appropriately

from

point

by resonances below the band gap.

There must be defects in evaporated prior

et al.,

hole with an electron(91. Consequently,

cannot be produced solids

These loss

arise from recombination

and lO.OeV for dl,/dV, differed

exciton.

particularly

resonances. These resonances appeared on the first scan or intensity

the first optical

solids raises

are quite impossible.

(Effects

have

doped insulators which

Energy losses in ionic solids Table

I. Electron

characteristic

LiF

NaF

2.8 4.4 5.1

3.6W

I.2 2.3~ 4.5 s.3w 6.5 8.5 9.8

s.3 7.2 9.1 10.4 (10.9)”

Il.3 12.9 (12.9) 14.1 (14.7)

(10.OJh I I .4w

(11.9) 13.2 IS.8 (lS.O)d

-ISw (IS,?)

losses and possibly correlated transitions indicated by w

KF

I.2

665

CaF,

SrF2

BaF,

I.1 -2.8 4.8 -5.6~ 7.4 9.2 I I.2 (11.2)’

I.0 -2.ow 4.3 5.2w 7.7 8.8 10,s (10.5) I I .sw (I 1.3)’

0.4 I .4 3.6 4.sw 6.8 9.3 IO.1 (10.1)’ 10.6~ ( 10.6)e I I.9 ( I ?.O)h 12.9w (12.7)” (13.7)h 14.8 (14.2)’

(11.0)’ I?.? (12.4)d 13.7w (13.6)”

(12.0)”

(14.5)

(I?,!)* 13.9w (13.7)”

16.0

16.1 (15.5)

IS.6 (16.0)’

(13.2)”

(16.5)’

( ). Weak

from other sources

RbCl’

NaBr”

RbBrh

0.4 2.lw 3.1 4.3w

0.3 I .4w 3.3 -4.4w

0.3 2.ow 3.3 -4.ow

6.3 7.3 (7.3)”

5.5 6.6 (6.6)” -7.2w

(8.3)’ 9.4 (9.5)’ I I.8 (12.0)’ (13.7)’ IS.5 ( 15.3)”

(7.2) 9.0 (8.7) I I.2 (11.1) (13.4) 14.8 (14.3)

5.S 6.5 (6.6)h -7.4w (7.0) 8.9 (8.4)’ II.4

bands are

(12.3) 14.4

‘Data from Ref. [6]. “Data from Ref. [7]. ‘Data from M. Creuzberg, Z. Phys. 196, 466 (1%6). ‘Data from D. M. Roessler, cited by M. Creuzberg, /or. cif. ‘Data from Ref.[ I], except for results in parentheses. ‘Data from D. Blechschmidt, R. Haensel, E. E. Koch, V. Nielsen and M. Skibowski. Phys. St&us Solidi 44, 787 (1971). #Data from Ref.[8]. “Data from G. Stephan. J. C. Lemonnier and S. Robin, 1. Opt. Sot. Amer. 57,486 (1%7); S. Robin-Kandare and J. Robin, Compr. Rend. ‘G. W. Rubloff, Phvs. Rev. 262, 1020 (1966); 85135, 662 (1972). ‘Y. Iguchi, T. Sasaki, H. Sugarawa, S. Sate, T. Nasu, A. Igiri. S. Onari and K. Kojima, Ph.w. Rec. Left. 26.82 (1971)

cause deep electron

trapping.

ble and predictions

were

The possibility

of observing

defect

site has been tested.

I-

KCI:

in

I-

characteristic

are predicta-

When

an electron

of

that

comparisons

possible

to occur

loss spectrum from

effects

A localized

is known

indistinguishable

These

confirmed.) energy

loss at a

unrelaxed

state of

being 0.2 eV. There

but

fail to detect reported

at 6.4 eV,

the

I per cent KI in KCI was

of undoped

KCI

in recent

24 instances

optical

were

than

structure

excitation

at

5.4 eV,

to the localized

In other

recent contain

undoped

alkali

the first

optical

sensitive

test is supplied

spectra

of

absorption

spectra,

tions of thallium those fore

conditions host

Rbi:TI’.

(except

The

demonstrated

by the same

doped

fluorides

loo

weakly

The energy already

not

excited optical

for

in Table

are not unlike

of

lattices

fluoride

A,

ion lattices

measurement

with

for the same

experimental

Comparable resolved both

structure

range strong

there and

in

above

the

optical

are numerous

of doped In

may

rough

constancy

KF,

the

among

the fluorides. at lower may

energy indicate

for

cannot

ion lattices

uncomplicated

by halogen

doubling

conditions.

lowest

conduction

band,

KI and

example, singlets

for

plausibly band

There such

may is a pairs

be due to structure,

of hole states. in the

the

5.3, 6.5 and

transitions.

conduction

I

with

states below

at 4.5,

differences

a

in Table

By analogy

states, the related

of energy

electron

and undoped

be those

is

is clearly

the arrangement

at the first four optical

empty

There

in some details.

exciton.

states

and RbI

excitation[3].

that there are four excitonic

corresponding

and

for KI

weak.

although

for luminescence

RbI it is likely

excitons

above

advantage

immediately

by luminescence

may well be incorrect

Structure halide

KF,

reported are

correlates

are

exciton

of these

by Tomiki

in the results,

be excited

losses in the with

SrF? and BaR

better

first

most

structure for

energy

offer no evident

8.5 eV. If these are triplet C

I]. for other

exciton

to and

with values of E, at 300°K reported

resonances,

the

be

and

resonances

and Miyata(8J.

results

there-

thalliumB

I below the first optical those

experielectron

because

optical

pattern

with

cannot

first

adequately

impact

(6 instances)

be correlated

and

weak

the average

with losses which were weak.

of electron

0.4eV,

The

0.4 eV,

are

I, in 17 of

for electron

excitations

cannot

In the sub-optical

[3]. Excita-

and were

exhibit

undoped

luminescent

reported

were

the same

procedure

used in this work[l

events

ion

within

this region.

much

to the

confirmation

do

[ IO] and because

techniques

under

in fluoride

more

excitation

completely

temperature)

losses

bands

TI-

correlated

luminescence

characteristic alkali

If

the

lie below

An even

but this was not observed

sensitized.

for

of TI’

correspond

luminescence

of intrinsic

mental

observed

states

and

optical

which

in Table

is within

correlates

14 eV. Electron

by the luminescence

would

direct

of KI:Tl-

of the host[3].

and

spectra

not

although

exciton

KI:TI’

the

resonance

iodides,

by

at 6.4 eV.

the loss spectra

RbI:Tl’

directly,

exciton

work no

rather

region

measurements

exciton

is a tendency

are few instances

from this center

excited

the agreement

all but one failure There

work [2]. The 2.6 eV and 3.4 eV emissions first

with conventional

above the first optical

The

oversimplified

666

T. HUANG and W. H. HAMILL

tight-binding approximation, is constructed from s-like unfilled orbitals of the cation, followed by d-like bands[lZ]. REFERENCFS I. 2. 3. 4. 5.

Hiraoka K. and Hamill W. H., 1. Chem. Phys. 57,3881(1972). Hiraoka K. and Hamill W. H., J. C/tern. Phys. 57.4058 (1972). Huang T. and Hamill W. H., 1. C/tern. Phys. 62, (in press). Hiraoka K. and Hamill W. H.. 1. Gem. Phys. 58,3686 (1973). Somorjai G. A., Principles of Surface Chemisfry, p. 158. Prentice Hall, Englewood Cliffs, New Jersey (1972).

6. Milgram A. and C&ens M. P.. Phys. Rev. 125, 1506 (1%2). 7. Eby J. E., Teegarden K. J. and Dutton D. 9.. Phps. Rer. 116. 1099 (1959). 8. Tomiki T. and Miyata T.. 1. Phys. Sot. Japan, 27,658 (1%9). 9. R. Smoluchowski, 0. W. Lazareth, R. D., Hatcher and G. J. Dienes, Phys. Rev. Lett. 27, l288( 1971). IO. Forr6 M., Z. Phys. 58, 613 (1930). I I. Pooley D. and Runciman W. A., J. Phys. C: Solid 9. Phys. 3, 1815 (1970). 12. For discussion and references see Chap. I by Knox A. S. and Teegarden K. J. In Physics Color Centers (Edited by Fowler W. 9.). p. 25. Academic Press. New York (1968).

of