Physica status solidi: Volume 24, Number 2 December 1, 1967 [Reprint 2021 ed.]
 9783112500224, 9783112500217

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

plrysica status solidi

VOLUME 2 4 • NUMBER

2 • 1967

Classification Scheme 1. Structure of Solids 1.1 Alloys. Metallurgy 1.2 Solid-State Phase Transformations 1.3 Surfaces 1.4 Films 2. Non-Crystalline State 3. Crystallography 3.1 Crystal Growth 3.2 Interatomic Forces 4. Microstructure of Solids 5. Perfectly Periodic Structures 6. Lattice Mechanics. Phonons 6.1 Mossbauer Investigations 7. Acoustic Properties of Solids 8. Thermal Properties of Solids 9. Diffusion in Solids 10. Defect Properties of Solids (Irradiation Defects see 11) 10.1 Defect Properties of Metals 10.2 Photochemical Reactions. Colour Centres 11. Irradiation Effects in Solids 12. Mechanical Properties of Solids (Plastic Deformations see 10)see 10.1) 12.1 Mechanical Properties of Metals (Plastic Deformations 13. Electron States in Solids 13.1 Band Structure. Fermi Surfaces 13.2 Excitons 13.3 Surface States 13.4 Impurity and Defect States 14. Electrical Properties of Solids. Transport Phenomena 14.1 Metals. Conductors 14.2 Superconductivity. Superconducting Materials and Devices 14.3 Semiconductors ' 14.3.1 Semiconducting Films 14.3.2 Semiconducting Devices. Junctions (Contact Problems see 14.4.1) 14.4 Dielectrics 14.4.1 High Field Phenomena, Space Charge Effects, Inhomogeneities, Injected Carriers (Electroluminescence see 20.3; Junctions see 14.3.2) 14.4.2 Ferroelectric Materials and Phenomena 15. Thermoelectric and Thermomagnetio Properties of Solids 16. Photoconductivity. Photovoltaic Effects 17. Emission of Electrons and Ions from Solids 18. Magnetic Properties of Solids 18.1 Paramagnetic Properties 18.2 Ferromagnetic Properties 18.3 Ferrimagnetic Properties. Ferrites 18.4 Antiferromagnetic Properties (Continued on cover three)

Classification Scheme 1. Structure of Solids 1.1 Alloys. Metallurgy 1.2 Solid-State Phase Transformations 1.3 Surfaces 1.4 Films 2. Non-Crystalline State 3. Crystallography 3.1 Crystal Growth 3.2 Interatomic Forces 4. Microstructure of Solids 5. Perfectly Periodic Structures 6. Lattice Mechanics. Phonons 6.1 Mossbauer Investigations 7. Acoustic Properties of Solids 8. Thermal Properties of Solids 9. Diffusion in Solids 10. Defect Properties of Solids (Irradiation Defects see 11) 10.1 Defect Properties of Metals 10.2 Photochemical Reactions. Colour Centres 11. Irradiation Effects in Solids 12. Mechanical Properties of Solids (Plastic Deformations see 10)see 10.1) 12.1 Mechanical Properties of Metals (Plastic Deformations 13. Electron States in Solids 13.1 Band Structure. Fermi Surfaces 13.2 Excitons 13.3 Surface States 13.4 Impurity and Defect States 14. Electrical Properties of Solids. Transport Phenomena 14.1 Metals. Conductors 14.2 Superconductivity. Superconducting Materials and Devices 14.3 Semiconductors ' 14.3.1 Semiconducting Films 14.3.2 Semiconducting Devices. Junctions (Contact Problems see 14.4.1) 14.4 Dielectrics 14.4.1 High Field Phenomena, Space Charge Effects, Inhomogeneities, Injected Carriers (Electroluminescence see 20.3; Junctions see 14.3.2) 14.4.2 Ferroelectric Materials and Phenomena 15. Thermoelectric and Thermomagnetio Properties of Solids 16. Photoconductivity. Photovoltaic Effects 17. Emission of Electrons and Ions from Solids 18. Magnetic Properties of Solids 18.1 Paramagnetic Properties 18.2 Ferromagnetic Properties 18.3 Ferrimagnetic Properties. Ferrites 18.4 Antiferromagnetic Properties (Continued on cover three)

physica status solidi B o a r d of E d i t o r s P. A I G R A I N , Paris, S. A M E L I N C K X , Mol-Donk, V. L. B O N C H - B R U E V I C H , Moskva, W. D E K E Y S E R , Gent, W. F R A N Z , Münster, P. GÖRLICH, Jena, E. G R I L L O T , Paris, R. K A I S C H E W , Sofia, P.T. L A N D S B E R G , Cardiff, L. N É E L , Grenoble, A. P I E K A R A , Warszawa, A. S E E G E R , Stuttgart, F. SE IT Z, Urbana, O. S T A S I W , Berlin, M. S T E E N B E C K , Jena, F. STÖCKMANN, Karlsruhe, G. SZIGETI, Budapest, J . TAUC, Praha Editor-in-Chief P. GÖRLICH Advisory Board M. B A L K A N S K I , Paris, P. C. B A N B U R Y , Reading, M. B E R N A R D , Paris, W. B R A U E R , Berlin, W. COCHRAN, Edinburgh, R. COELHO, Fontenay-aux-Roses, H.-D. DIETZE, Saarbrücken, J.D. E S H E L B Y, Cambridge,P.P. F E O F I L O V, Leningrad, J. H O P F I E L D , Princeton, G. J A C O B S, Gent, J. J A U M A N N , Köln, E. K L I E R , Praha, E. K R O E N E R , Clausthal-Zellerfeld, R. KUBO, Tokyo, M. M A T Y Ä S , Praha, H. D. MEGAW, Cambridge, T. S. MOSS, Camberley, E. NAGY, Budapest, E. A. N I E K I S C H , Jülich, L. P A L , Budapest, M. RODOT, Bellevue/Seine, B. V. R O L L I N , Oxford, H . M . R O S E N B E R G , Oxford, R. V A U T I E R , Bellevue/Seine

Volume 24 • Number 2 • Pages 391 to 764, K103 to K178, and A35 to A70 December 1, 1967

AKADEMIE-VERLAG•BERLIN

Subscriptions and Orders for Single copies should be addressed to AKADEMIE-VERLAG GmbH, 108 Berlin, Leipziger Straße 3 - 4 or to Buchhandlung KUNST U N D WISSEN, Erich Bieber, 7 Stuttgart 1, Wilhelmstr. 4 - 6 or to Deutsche Buch-Export und -Import GmbH, 701. Leipzig, Postschließfach 160

Editorial Note: "physica status solidi" undertakes that an original paper accepted for publication before the B1'1 of any month will be published within 50 days of this date unless the author requests a postponement. In special cases there may be some delay between receipt and acceptance of a paper due to the review and, if necessary, revision of the paper.

Schriftleiter und verantwortlich f ü r den I n h a l t : Professor D r . D r . h. c. P. G ö r l i c h , 102 Berlin, Neue Schönhauser Str. 20 bzw. 69 J e n a , Humholdtstr. 26. Redaktionskollegium: Dr. S. O b e r l ä n d e r , Dr. E . G u t s c h e , Dr. W. B o r c h a r d t . Anschrift der Schriftleitung: 102 Berlin, Neue Schönhauser Str. 20. Fernruf: 426788. Verlag: Akademie-Verlag G m b H , 108 Berlin, Leipziger Str. 3—4, Fernruf: 220441, Telex-Nr. 112 020, Postscheckkonto: Berlin 35021. — Die Zeitschrift „physica status solidi" erscheint jeweils a m 1. des Monats. Bezugspreis eines Bandes MDN 72,— (Sonderpreis f ü r die D D R MDN 60,—). Bestellnummer dieses Bandes 1068/24. Gesamtherstellung: V E B Druckerei „Thomas Müntzer" Bad Langensalza. — Veröffentlicht unter der Lizenznummer 1310 des Presseamtes beim Vorsitzenden des Ministerrates der Deutschen Demokratischen Republik.

Contents

Page

Original Papers M. ROUZEYRE

Contribution à l'étude des phénomènes de transport dans les structures aluminium-alumine-metal

399

H . SPREEN

Untersuchungen der Bereichstruktur von Nickeleinkristallen ( I I I )

413

H . SPREEN

Untersuchungen der Bereichstruktur von Nickeleinkristallen (IV)

431

G . M . BARTENEV a n d E . G . KORYAK-DOKONENKO

The Theoretical Strength of NaCl Crystals under Tension along its Body and Face Diagonals 2+

J.

STANKOWSKI

E P R S t u d y of Cu

Ion in Triglyoine Fluoroberyllate

J.

GEIGER

Inelastic Electron Scattering in Thin Films a t Oblique Incidence

443 451 457

A . S. NOWICK a n d G . J . D I E N E S

On t h e Inherent Curvature of the Arrhenius Plot in Diffusion E x periments J . B. PAGE JR. a n d

D . STRAUCH

Shell Model T r e a t m e n t of t h e Vibrations of Crystals Containing Substitutional Defects B . PISTOULET,

461

469

J . L . ROBERT, a n d M . MARQUES

S t u d y of GaSb Valence B a n d by Magnetoresistance a n d Hall Effect Measurements a t Low Temperature

481

J . M . TROOSTER a n d A . D Y M A N U S

Mossbauer Effect in G a 2 _ I F e I 0 3 and Related Compounds B.

AUTIN

Transient Acousto-Electric Phenomena in Semiconducting Electric Crystals

487 Piezo501

T . H . CHEUNG a n d S. WANG

Theoretical Studies on Z r B a n d s in Alkali Halide Crystals R . DUPREE,

. . . .

509

C. T . FORWOOD, a n d M . J . A . S M I T H

Conduction Electron Spin Resonance in Small Particles of Gold . .

525

C. HAMMAR a n d P . WEISSGLAS

H o t Electron Polar Scattering in I n S b

531

F . FROHLICH a n d G. H E N S E L

Low Temperature Conductivity of Ca-Doped KC1 Crystals Influenced b y Segregation Processes M. K.

SHEINKMAN, V . A . TYAGAI,

535

0 . V . SNITKO, I . B . ERMOLOVICH, G . L . B E L E N K Y ,

a n d V . N . BONDARENKO

Studies on t h e N a t u r e of CdS Single Crystal Sensitization b y Etching M . S. KAGAN,

543

S . G . KALASHNIKOV, a n d N . G . ZHDANOVA

Stationary State of a Semiconductor with Negative Bulk Differential Conductivity

551

R . E . NETTLETON

B. 26»

PIETRASS

Statistical P e r t u r b a t i o n Theory of Order-Disorder Ferroelectrics: Zeroth Approximation

561

Application of the S.C.F. Method of Lattice Dynamics to t h e Coulomb Lattice

571

Contents

394

Page F . FRÖHLICH u n d O . ADAMBTZ

Umwandlung von OH-Zentren in U-Zentren bei KCl-Kristallen durch elektrolytische Verfärbung

583

M . H Ö H N E a n d M . STASIW

ESR Studies of Different S e | - Centres in Silver Halides

H . D . KOSWIG,

T. L.

TANSLEY

H . DIEPERS

591

U . RETTER u n d W . ULRICI

Jahn-Teller-Effekt an Übergangsmetallionen in Silberhalogeniden (I) Forward Current Injection Modulation of Photocurrent in p-n Heterojunctions Elektronenmikroskopische Untersuchung von Fehlstellenagglomeraten in ionenbestrahlten Kupferfolien (II)

605 615 623

A . MARAIS a n d T . MERCERON

Orientation Superstructures and their Effects in Spinel Ferrites . .

635

J . HÖLZL u n d B . HEIMANN

B. A.

Experimentelle Ergebnisse zur Temperaturabhängigkeit der Sekundärelektronen-Emission an polykristallinem Nickel im Intervall 100 ^ T contre-électrode Mg-Al 2 0 3 .

403

Contribution à l'étude des phénomènes de transport

/ A ai-ai2o3-ai

J/ Jfi

e-WÂ

t 1| 1 I^

2

1 -2

/

/ ?

i

!

y / / / jm'

^

/

S /

-1 1

S/'sS '

/

2 Potententiel(V) —

-1

/ / -z A

hv=372eV

!

~3

»hv*3J9eV +hi>=3M8eV

H

°hv=2MeV

*hv2,27eV -5 Fig. 3. Caractéristiques courant photoélectrique tension de polarisation (R = 0: émission d'électrons par la contre-électrode)

Fig. 4. Variation du seuil avec la tension positive électrode-contre-électrode. V21 > 0

Fig. 5. Variation du seuil avec les tensions négatives électrode—contre-électrode. V21 < 0

404

M.

ROUZEYRE

5. Origine des photoélectrons 5.1 Sens du photocourant

et nature

de la

contre-électrode

Dans l'hypothèse d'une désexcitation rapide des porteurs dans l'alumine le courant d'électrons doit être dirigé de l'électrode à travail de sortie le plus élevé vers l'électrode à travail de sortie le plus faible c'est à dire : de la contreélectrode vers l'électrode pour des structures AU-A1203-A1, A1-A1203-A1, Ag-Al 2 0 3 -Al, et en sens inverse pour la structure Al-Al 2 0 3 -Mg. Au contraire si les quantités importantes sont les nombres d'électrons photoexcités, on attend le résultat inverse dans les structures A1-A1203-AU et Al-Al 2 0 3 -Ag, puisque l'or et l'argent sont moins absorbants que l'aluminium dans le domaine d'énergies exploré. Dans ces expériences l'électrode d'aluminium a une épaisseur > 1 0 0 0 A. 5.1.1 Contre-électrode d'aluminium, Dans le domaine spectral exploré, et quelles que soient l'épaisseur de la contre-électrode, variable de 20 à 300 À et de la couche d'oxyde 20 à 200 À le courant net d'électrons, en l'absence de polarisation appliquée est toujours dirigé de la contre-électrode vers l'électrode. Ce résultat est en accord avec les observations d'autres auteurs. Il correspond à un déplacement de porteurs dans le sens de la force électrique mais nous verrons qu'il ne s'agit en fait que d'un cas fortuit. 5.1.2 Contre-électrodes d'argent et d'or Pour les six radiations utilisées et dans le domaine d'épaisseurs exploré, 20 à 200 A environ, le résultat est inverse du précédent, c'est à dire que le déplacement s'éffectue dans le sens opposé à celui de la force électrique. 5.1.3 Contre-électrode de magnésium Pour les échantillons étudiés, épaisseur d'oxyde 8 « 35 Â et couche de magnésium avec un facteur de transmission voisin de 10%, l'aluminium émet la majorité des électrons pour les radiations de plus faible énergie 2,14 et 2,27 eV. alors que le magnésium est émetteur pour les radiations de 2,84, 3,08, 3,39 et 3,72 eV malgré l'effet antagoniste du champ électrique (Fig. 6). +10

5.2 Influence d'une polarisation extérieure sur le sens du courant photoélectrique

I" H et

L

01

%-1Mi r~ \ 2

;

;

; 3 hv(eV)

Pour s'affranchir des incertitudes sur les valeurs des travaux de sortie et le sens du champ interne on peut étudier l'influence d'une tension appliqué extérieurement sur le sens du courant photoélectrique.

2 M émetteur ¿Î d'électrons l ' i g . 6. D i o d e A l - A l 2 0 3 - M g .

e = 35 Â, s a n s polarisation

Contribution à l'étude des phénomènes de transport 5.2.1 Contre-électrodes

405

d'argent et d'or

L'application d'une f o r t e tension positive F 2 1 sur l'électrode d'aluminium diminue la contribution de celle-ci au courant photoélectrique par augmentation du seuil correspondant: = q V21 et par augmentation de la force électrique antagoniste. On obtient bien l'annulation du courant photoélectrique mais les tensions nécessaires sont supérieures à 1 V avec les radiations de plus courte longueur d'onde. Comme il est impossible d'admettre, pour les t r a v a u x de sortie argent^alumine et or-alumine, une valeur inférieure de 1 e V à celle du travail de sortie A1-A1 2 0 3 on doit donc bien conclure à un déplacement des électrons en sens inverse de la force électrique. 5.2.2 Contre-électrode

d'aluminium

L a diminution du courant photoélectrique s'obtient cette fois en polarisant positivement la contre-électrode, qui est émettrice en l'absence de polarisation appliquée, mais la valeur de cette tension d'annulation n'est ni constante ni égale à la valeur — (PJq, qui annule le champ électrique interne, prévue dans l'hypothèse d'une désexcitation très rapide dans l'alumine. Dans le Tableau 1 sont rassemblées les valeurs obtenues avec une couche d ' o x y d e voisine de 70 Â et des contre-électrodes d'épaisseur variable. L a valeur de cette tension d'annulation dépend de l'épaisseur de la contre-électrode et est considérablement plus grande que celle que l'on peut raisonnablement attribuer à la différence des potentiels de sortie aux deux contacts aluminium-alumine. D'autre part une variation de cet ordre de grandeur de la différence 01 — í>2 avec l'épaisseur de la contre-électrode serait décelable dans un diagramme de F o w l e r et nous n'avons jamais pu la mettre en évidence (voir Section 5.1 et F i g . 2). 5.3 Variation de la tension d'annulation avec l'énergie excitatrice et l'épaisseur d'oxyde dans une structure Al-Al203-Al L e s résultats obtenus (Tableau 2) sur une série d'échantillons aluminiumalumine-aluminium montre que a) la valeur de la tension d'annulation varie considérablement tant avec l'épaisseur de la contre-électrode et de la couche d ' o x y d e qu'avec l'énergie des photons incidents ; b ) pour des épaisseurs d'électrode et de contre-électrode constantes, la valeur de cette tension d'annulation croît généralement avec l'épaisseur d ' o x y d e ; c) à quelques rares exceptions près, qui se produisent toujours pour la radiation X = 4360 A et restent souvent dans le domaine des erreurs de mesure, la valeur absolue de la tension d'annulation est une fonction croissante de l'énergie excitatrice h v. Cette dernière variation qui ne peut être imputée à des variations concomittantes des t r a v a u x de sortie permet de rejeter définitivement l'hypothèse d'une désexcitation très rapide des porteurs dans l'alumine. 5.4 Influence du sens de la lumiere incidente sur le sens du photocourant dans une structure Al-Al203-Al On peut encore démontrer expérimentalement l'importance du nombre de photoélectrons engendrés dans l'une et l'autre électrode en choisissant une structure à géométrie symétrique aluminium-alumine-aluminium dont les

406

M. ROTJZE VRE (M O O -H

A rJÍ

3,72

o I

lO^OOiOiOlOíOWi ©©cT©©©©© ^ N H CC IC O CÛ co CO ^ h M a © o © ©* ©*" o o o

co a oo ^ N HM« o" © ©" ©"

S

® 13 tí u o « 2 13 « o ÉL53 H

o -H o

o N Oí w H © Hw © ©

00

o

co s



c3

0,13

Hcá 1,9 eV

O +1

t> © C^ CO ©" ©

Électrode

>o

C8 H

o-«^ o o o o o o o o o o o o o o o o o o o o o o o o I—(I—Ir—(1—(I—l t—(i—1 rt

Alumine

«rj

o 41

-

02 .

La valeur de cette tension d'annulation est bien une fonction de l'énergie excitatrice h v de Q1¡Q2 donc de l'épaisseur des électrodes, et de P 1 2 /P 2 1 donc de l'épaisseur d'oxyde. Si l'épaisseur d'oxyde est assez faible pour que les probabilités de transfert varient peu avec l'énergie excitatrice la tension d'annulation est une fonction linéaire de h v dont l'extrapolation doit permettre une détermination séparée du seuil. Ce point parfaitement confirmé expérimentalement fera l'objet d'une publication séparée [16]. 6.3 Longueur d'atténuation

en énergie dans

l'alumine

Il est intéressant de remarquer, dans les expériences de renversement du sens du courant avec le sens de la lumière incidente, que le rapport des courants JJJC est minimum pour la radiation h v = 2,84 eV. 27!

410

M.

ROUZEYRE

Le fait que ce rapport augmente, d'une part quand on se rapproche du seuil et d'autre part pour les fortes énergies, suggère l'existence de deux méchanismes possibles d'interaction dans l'isolant. Le premier, quasi élastique, n'affecterait que les électrons excités à une énergie voisine du seuil et serait responsable de l'augmentation du rapport JJJe aux faibles énergies. Le deuxième très inélastique, processus d'ionisation par choc par exemple, se manifesterait d'autant plus que l'énergie initiale serait plus grande. On est donc tenté d'écrire pour les probabilités de transfert des électrons des expressions de la forme : P12 = T1 dans le sens de la force électrique, P 2 1 = T2 exp (— s/L) en sens contraire de la force électrique. L représente la longueur d'atténuation en énergie associée au mécanisme inélastique. T1 et T2 l'effet des réflexions quantiques et des réflexions par chocs élastiques. Pour des énergies très nettement supérieures au seuil, T2 doit être peu différent de T1 de telle sorte que, dans cette hypothèse, on peut remplacer le rapport P 21 /P ]2 par e ~ s ! L et calculer L. La détermination de e s l L est très imprécise car la valeur absolue du rapport JJJe n'est, dans le mailleur cas, connue qu'avec une précision de l'ordre de 15%. Par contre, toutes les mesures de 30 Oe auch als F u n k t i o n von L gemessen worden. Über die Abschlußstruktur I I . u n d höherer Ordnung auf der (OOl)-Fläche haben wir noch keine experimentellen Ergebnisse. I m R a h m e n modellmäßiger Vereinfachungen h a t der Kristall n u r eine beschränkte Anzahl von Möglichkeiten, Abschlußstrukturen zu bilden. Für Hi = 0 haben wir mehrere denkbare Modelle mit verschiedenen Abschlußstrukturen höherer Ordnung durchgerechn e t ; n u r das in Fig. 13 skizzierte Modell gibt die gemessene Temperaturabhängigkeit [5] von d richtig wieder; d n i m m t mit fallender Temperatur zu. Herrn Prof. Dr. Ch. Schwink danke ich f ü r die interessante Aufgabe. Besonderer Dank gilt ihm, aber auch den Mitarbeitern seiner ferromagnetischen Arbeitsgruppe f ü r die vielen anregenden Gespräche. H e r r n Prof. Dr. W . Rollwagen danke ich f ü r die Möglichkeit, diese Arbeit in seinem I n s t i t u t auszuführen, u n d f ü r die Mittel, die mir vom I n s t i t u t zur Verfügung gestellt wurden. Literatur [1] CH. SCHWINK und H. SPREEN, phys. stat. sol. 10, 57 (1965). [ 2 ] CH. SCHWINK u n d O . GRÜTER, p h y s . s t a t . sol. 1 9 , 2 1 7 ( 1 9 6 7 ) .

[3] J. KOCINSKI, Acta phys. Polon. 18, 169 (1959). [4] CH. SCHWINK und H. SPREEN, Z. angew. Phys. 15, 234 (1963). [ 5 ] H . F R E Y , O. GRÜTER, D . KRAUSE u n d CH. SCHWINK, Z. a n g e w . P h y s . 1 8 , 4 6 1 ( 1 9 6 5 ) .

[6] CH. SCHWINK und G. DEDIE, Z. angew. Phys. 21, 338 (1966). [ 7 ] CH. SCHWINK, H . SPREEN, E . BUNGENSTAB, G. D E D I E u n d O. GRÜTER, I E E E

Trans.

Magn. MAG-2, 503 (1966). [8] M. L. NEEL, J. Phys. Radium 5, 241 und 265 (1944). [9] U. M. MARTIUS und K. V. Gow, Canad. J. Phys. 33, 225 (1955). [ 1 0 ] E . TATSUMOTO u n d T . OKAMOTO, J . P h y s . S o c . J a p a n 1 4 , 1 5 8 8 ( 1 9 5 9 ) . [ 1 1 ] M . YAMAMOTO u n d T . IWATA, S e i . R e p . R e s . I n s t . T ö h o k u U n i v . A 5 , 4 3 3 ( 1 9 5 3 ) .

[12] [13] [14] [15] [16]

A. HUBERT, Z. angew. Phys. 19, 521 (1965). CH. SCHWINK, Z. Phys. 173, 542 (1963). O. SCHÄRPF, erscheint demnächst. E. LIFSHITS, J. Phys. (USSR) 8, 337 (1944). J. KACZER, Soviet Phys. — J. exper. theor. Phys. 19, 1204 (1964).

[ 1 7 ] B . A . LILLEY, P h i l . M a g . 4 1 , 7 9 2 ( 1 9 5 0 ) . [ 1 8 ] CH. KITTEL, R e v . m o d . P h y s . 2 1 , 5 4 1 ( 1 9 4 9 ) . (Received

September

12,

1967)

H . SPREEN: Untersuchungen der Bereichstruktur von Nickeleinkristallen (IV)

431

phys. stat. sol. 24, 431 (1967) Subject classification: 18.2; 4; 21.1 Sektion Physik der Universität

München1)

Untersuchungen der Bereichstruktur von Nickeleinkristallen IV. Die Grundstruktur in zylindrischen Kristallen mit als Achse 2 ) Von H.

SPKEEN3)

Wir untersuchen an zylindrischen Nickeleinkristallen mit vcvD.

(12b)

In these conditions J^t) =

VS

-*L-exp ^ o

I

A

C

VD

9

f l

A

J

V

c

*.

(13)

"D

Fig. 3. Variations of i j versus vc at given y

Transient Acousto-Eleotric Phenomena in Semiconducting Crystals

Then, we can evaluate the time

505

at which (14)

where J0

=

c

a E , c

=

— = 4 712

a

n

kT e v0

v

n

(15)

qv.

By substituting

(i2c)

where Vk is given by

= -4 ( W )



(12d)

Theoretical Studies on Zj-Bands in Alkali Halide Crystals

513

a>jc is the frequency of the longitudinal optical phonons of momentum k and = — — , es being the static dielectric constant for the crystal (see [12] for £00 £s details). Following the canonical transformation used for theexciton part, we can write £

Ht+

Vq^ = E A*kAkhcok h

(13)

k h™k

F e _ i = E (VtAkeikr + c.c.) + E nw ( e i f c r + e _ i f c ' r ) • (14) h h k The last term in (13) is a constant resulting from the canonical transformation and can be omitted in the present calculation. The last term in (14) can be reduced to [ — 1 " ( " - £)s / r \£oo Using the new operators for exoitons and phonons and taking |0) defined by B w \0) = 0 as the vacuum state for excitons and ||0> defined by A^l |0) = 0 as the vacuum state for phonons, we have H

for r < R, and with

=

J*_ M*_ 2 m- ~ d

2m

esr

r3

r

(15) y

'

r*

Hex = E 5 * Bw Ew + E (Vw Bw e * » " + c.c.) ,

(17)

ik r

Hvh = E A*hAk h ojk + E (VkAk e + c.c.) (18) k k for r > R. In the polaron model the effective Hamiltonian of the problem has the same form as that of (15) for r < R and has the following form: H a c =




+

(19)

for r > R, where F and 0 are, respectively, the wave functions of the excitons and the phonons. Using the Hartree-type self-consistent-field (SCF) method developed in the polaron model, we obtain < F \ n ^ F and

}

= - ( l ~ 71 \

y

) [/x £co/

2 /2]

=^;[yi-2yJ> where

,

k =

Mimas /1 = / Qldw , 0 Wmax C sin r 10 4 . J ^ 0

(20) (21) (22)

'

^'max

^ = f g l d k ,

(23)

(24)

0

¿mas

R. Here Hi has the same form as that appeared before. F q _ i , F p _ j , F d _ i , and F e _ j represent the interactions of the ionic polarization field of the crystal with the negative-ion vacancy, with the positive-ion vacancy, with the divalent positive ion, and with the trapped electron, respectively. F e _ j is omitted for r = 0, we obtain ht = K for r < R and h, = j ; At Akhmk k

+ £(Vk k

Akeik*

+ c.c.)

+

± E \r

l + e±) + K Tp /•(j /

e

(44b)

for r > R, where K is a constant resulting from the canonical transformation of hj and can be neglected in this calculation. Thus the Hamiltonian is of the form 2 me for r < R, and

d /v)2

H = ~

1

\

£00/ r

pi f>2 - — + -

r

1 £oo p

«00 ^

fi?, p2 -- + AC -r+Hex

'

r\ ^ + Hvll

K

'

(46)

Theoretical Studies on Z,-Bands in Alkali Halide Crystals

519

for r > R, where Hex and Hph are given by Hex = 2

B*w Bw Ew + £

HPh = £ At Ak h cok

2

k

Here we have replaced £

(VWBW (VkBk

k

V' (r — Rx0)

e,

+

e

e12 / 1/1 1\ £oo V ' l ( r 7 - r T ) + A C e2

25088

1

V v "

i

e

/

v

g

l

I (51)

A term like the last term in (51), which is associated with the alkali-earth impurity, is very small and thus we neglect it in calculating the energy levels for a given crystal, because the wavefunction is centered on the negative-ion vacancy and the term AC e2/r^ vanishes as r¿"2. Neglecting the last term of (51) and minimizing Eg with respect to the variational parameter A in (49), we find the ground-state energies of Z r centers de34 physica 24/2

520

T . H . CHEUNG a n d S .

WANG

Table 2 Theoretical values of the variational parameters and energies of the Z r c e n t e r and the experimental frequencies (expressed as transition energies) of the Zj-bands for the crystals of interest Model

Crystal

X (A- 1 )

(eV)

P (A" 1 )

(eV)

AE (eV)

a

KC1 NaCl

0.70 0.77

-1.78 -2.15

0.28 0.35

0.07 0.13

1.85 2.28

b

KC1 NaCl

0.69 0.76

-1.61 -1.96

0.23 0.29

0.11 0.20

1.72 2.16

Experimental A E (eV) of the Z r b a n d maximum Crystal

AE

1.9 to 2.1 KCl:Ca NaCl: Ca 2.4 to 2.5

scribed in Fig. 1 for both KC1 and NaCl. These values and the corresponding A are summarized in Table 2. The values used for m and constants in the calculation are the same as those of [12], We shall now consider the first excited state to which the transition can be made from the ground state of the Z r center. The Bushnell model implies that the Zj-transition is the perturbed F-transition. So we expect that a) the electronic orbit of the first excited state of the Z r center in the Bushnell model is not much different from that of the first p-like excited state of the F-center and b) the function of (50) is a good trial wavefunction for the present excited state. The results of [12] showed that an approach used in evaluating (F\Hex\Fy for deep states such as that used for the present ground state (i.e. a type of selfconsistent field (SCF) method) is invalid for large-orbit states such as the excited states of the F-center and the present system. Therefore we follow the perturbation method described in [12] to treat E'h. 2L

Note t h a t there is no contradiction to what has been said in Section 2 as there the assumption was made t h a t the space charge layer was located far away from the electrodes, this being not valid at the ends of the saturation region. Acknowledgements

We should like to thank V. L. Bonch-Bruevich, A. F. Volkov, and A. Ya. Shulman for many very useful discussions. We should like as well to thank V. L. Bonch-Bruevich for informing us of his new results on the static field distributions in short samples [16]. References [ 1 ] H . KIESS, p h y s . s t a t . s o l . 4 , 1 0 7 ( 1 9 6 4 ) . H . K I E S S a n d F . STOCKMANN, p h y s . s t a t . s o l . 4 , 1 1 7 ( 1 9 6 4 ) . [ 2 ] K . W . BOER a n d W . E . WILHELM, p h y s . s t a t . s o l . 4 , 2 3 7 ( 1 9 6 4 ) .

[3] A. YAMASHITA and R. NIL, Japan. J. appl. Phys. 5, 263 (1966). [ 4 ] M . S. KAGAN, S. G. KALASHNIKOV, a n d N . G. ZHDANOV A, p h y s . s t a t . s o l . 1 1 , 4 1 5 ( 1 9 6 5 ) . [ 5 ] N . G. ZHDANOVA, M . S. KAGAN, a n d S. G. KALASHNIKOV, F i z . t v e r d . T e l a 8 , 7 7 4 ( 1 9 6 6 ) . [ 6 ] A . A . ANDRONOV, E . A . LEONTOVICH, J . J . GORDON, a n d A . G . MAYER, Q u a l i t a t i v e

Theory of Dynamic Systems, Izd. Nauka, Moscow 1966. [ 7 ] K . W . BOER a n d W . E . WILHELM, p h y s . s t a t . s o l . 3 , 1 7 0 4 ( 1 9 6 3 ) . [ 8 ] K . W . BOER a n d P . L . QUINN, p h y s . s t a t . s o l . 1 7 , 3 0 7 ( 1 9 6 6 ) . [ 9 ] K . W . BOER, P h y s . R e v . 1 3 9 , A 1 9 4 9 ( 1 9 6 5 ) .

[10] V. L. BONCH-BRUEVICH, Fiz. tverd. Tela 8, 356 (1966). [11] V. L. BONCH-BRUEVICH and SH. M. KOGAN, Fiz. tverd. Tela 7, 23 (1965). [12] M. S. KAGAN and S. G. KALASHNIKOV, Proc. Intern. Conf. Phys. Semicond., Kyoto 1966 (p. 537). [ 1 3 ] S. G. KALASHNIKOV, M . S. KAGAN, a n d V . A . VDOVENKOV, F i z . T e c h n . P o l u p r o v . 1 , 1 1 6

(1967). [14] A. F. VOLKOV, Phys. Letters (Netherlands) 20, 598 (1966); Fiz. tverd. Tela 8, 3187 (1966). [ 1 5 ] V . L . BONCH-BRUEVICH, p h y s . s t a t . s o l . 2 2 , 2 6 7 ( 1 9 6 7 ) . [ 1 6 ] V . L . BONCH-BRUEVICH, p h y s . s t a t . s o l . , 2 3 , 7 6 1 ( 1 9 6 7 ) . (Received

October

5,

1967)

R. E.

NETTLETON

: Statistical P e r t u r b a t i o n Theory of Ferroeleotrics

561

phys. stat. sol. 24, 561 (1967) Subject classification: 13; 14.4.2 Sandia Laboratory, Albuquerque,

New

Mexico

Statistical Perturbation Theory of Order-Disorder Ferroelectrics: Zeroth Approximation1) By R. E.

NETTLETON

Approximate spin-wave energies are calculated b y constructing approximate eigenvalues of t h e commutator operator [H, ]. One linearizes commutators b y replacing t h e Z spin operator b y its expectation (Z) and then determines the latter self-consistently. Such a calculation is discussed for t h e Blinc formulation of the Slater-Takagi-Senko model of K H 2 P 0 4 which is treated by a coordinate rotation similar to t h a t used b y DeGennes. F o r sufficiently large tunneling of protons on hydrogen bonds, t h e rotation angle will approach jt/2 a t a temperature defining a second-order transition. If tunneling is weak, t h e transition is firstorder, unless there is no tunneling. I n the latter case, t h e model reduces to t h a t of t h e Weiss molecular field, and t h e transition is again of second order. Nous calculons des énergies approximatives des ondes de spin en construisant des valeurs caractéristiques approximatives de l'opérateur commutateur [H, ]. Nous linéarisons les commutateurs en remplaçant l'opérateur de spin Z p a r sa valeur moyenne (Z), après quoi nous déterminons la dernière d'une manière conséquente en elle-même. Ce calcul est appliqué à la formulation de Blinc du modèle Slater-Takagi-Senko de K H 2 P 0 4 , qui est traitée p a r une rotation de coordonnées, semblable à celle employée p a r DeGennes. Dans le cas d ' u n effet de tunnel suffisamment grand de protons sur les liens d'hydrogène, l'angle de rotation s'approchera de m/2 à une température définissant une transition du second ordre. Si l'effet de tunnel est faible, la transition est du premier ordre, à moins qu'il n ' y ait pas d'effet de tunnel. Au dernier cas, le modèle revient à celui du champ moléculaire de Weiss, et la transition devient encore une fois celle du second ordre.

1. Introduction The object of this paper is the derivation of a self-consistent equation determining the polarization as a function of temperature in the ordered phase of a ferroelectric of KH 2 P0 4 type. Such an equation yields the saturation polarization at absolute zero, and also permits a simple analytic discussion of the effect on the order of the transition produced by varying the magnitudes of the several terms in the Hamiltonian. The most important terms are exchangetype interactions between electric moments and also tunneling. Tunneling produces spin wave dispersion and can cause the transition to be of first or second order depending on its magnitude. To determine approximate excitation energies for the system of electric moments we use a simple method which has been shown to be equivalent to that of thermodynamic Green's functions for the Heisenberg ferromagnet [1], If H is the Hamiltonian for the moments in question, we construct non-zero operators el which satisfy [H, el] = ek6l + Rk, (1) 1

) This work was supported b y t h e U.S. Atomic Energy Commission.

562

R . E . NETTLETON

where R k is small and can be neglected in the lowest, or zeroth approximation. T o make Bk small, the excitation energies efc are taken to be functions of the temperature-dependent polarization. The latter in turn is calculated in terms of the ek by introducing spin operators Xlt Yj, Z} for the j'-th lattice site such that Z j = +1/2 represents the two possible directions of polarization on that site. Then (cf. [1], equation (3.11)) < Z , >

=

1

-

< S j

S j y

;

S t

=

X ,

+

i Y , ,

S y

=

X ,

-

i

7 ,

.

(2)

The ^-operators will be found to be linear sums of 0-operators, and the average Sty can be computed from (cf. [1], equations (2.9)) < 0 ^ Qy

=

0

k

( . W ,

eiiy

O ( B ) ;

+

®

=

k

[ e x p ifi

ek) -

l ] "

1

,

(3)

where Q is any operator which can be expanded in the complete set of d's. Since ek is a function of (Z"), equations (2) and (3) give the basic self-consistent equation. The Hamiltonian which we use in applying equation (1) to K H 2 P 0 4 is that of Blinc and Svetina [2], This is based on a model introduced by Slater [3] and Takagi [4] which assigns energies ec to the various configurations c of protons on the four hydrogen bonds which connect a given P 0 4 group to its neighbors. These configurations and energies ere diagramed in Fig. 1. The protons interact with the P 0 4 ions to polarize the latter. Thus, if a proton is in the potential well closer to or farther from the top of a P 0 4 group, we denote the associated polarization by j, or respectively, and use the opposite convention for protons at the bottom of the P 0 4 group. The contribution of Fig. l c to the four-body interactions Hs B in H is then N e2

2 J j = i

n

j u

21 n}

3

1

4

j,

and similarly for other proton arrangements, where the w's are 1 or 0 according as the indicated proton on the j-th P 0 4 group has or does not have the designated configuration. In addition to four-body interactions, there are two-body terms (Hj, R) represented by an exchange constant ./,• ¿< ¡y (index i denotes the P 0 4

7

*

I

3

n ' t

ti

M

Fig. 1. Proton configurations and energies ec on the four H-bonds (numbered 1 to 4 in a)) surrounding a P 0 4 group (denoted by a square). A dot signifies that the proton is in the potential well nearer to the P 0 4 group, while absence of a dot means the proton is nearer to an adjacent P 0 4 group. Arrows show the direction in which interaction with each of the four protons tends to polarize the group

Statistical Perturbation Theory of Order-Disorder Ferroelectrics

563

group and j one of the four protons). Finally, there is a tunneling term Ht with constant r , which shifts a proton from f to j and was introduced [2] to explain the isotope effect on Curie temperature Tc. The t o t a l H is t h u s H = H t + Hsr 1 Ht = — -7r 4 N

^SR = ^LU

=

+ N

HLR; 4

t £ £ (Pij t i=lj=l

t + bin bij t) ,

4

£klkik,k1 Wj 1 lc, niilCt niska W; 4 ] , 2 Ct i = 1 kukt = j tf 4 nijl) (ni'j't ni'j'l) — ~r S Jii'jj'(niH ** i, ?' = 1 j,f = 1

W

The operator b]^ is a Fermi operator which creates an f proton a t the j-th site on the ¿-th P 0 4 group, and niik = b\ik (k = f , with ni)it= 1. I n writing H entirely in terms of proton operators, we assume with Slater [3] and Blinc [2] t h a t the cell polarization is rigidly coupled to the proton configuration. To apply equations (1 to 3) to (4), we want to express H in terms of spin operators Xu Y,, Z, .which is accomplished [5] by defining (we use a single subscript henceforth for the index pair i, j in (4)) Xi

= Y (bh bH +

bi[

&if) ,

Yi

= Y (bh bH — bU 6

Zi

= J (bh bi\ — bU bil) •

I n terms of spin operators, equation (4) assumes the form Ht = - i - r z ^LR + Hsu ~ — IT £ 7i1 j

xt,

+ 5T E Vifkm Zt Zj Zk Zm . ^ i,j,k,m

(5)

We drop the three-spin terms on the ground t h a t contributions thereto t e n d to cancel and t h a t these terms complicate the analysis while introducing little qualitative effect. The / ' s are sums of J ' s and e's, while yw =

£o

- 4 £ l + 2 f 2 + 4 ^ - 4 £3 + e 4

whenever the four indices denote distinct protons on the same P 0 4 group, and VW = 0 otherwise. The Xt in the tunneling term IIt are spin-flip operators which create higherenergy states from the totally aligned configuration, and so we expect the excitations 6 k to be linear sums of these. I n the next section, these sums will be constructed so t h a t equation (1) is satisfied when we use the H given in (5). The corresponding self-consistent equation for (Z~y will be derived in Section 3, where we also solve the equation at absolute zero. By differentiating the selfconsistent equation with respect to temperature in Section 4, we shall locate the point where the slope of polarization becomes negatively infinite, indicating a first-order transition for certain values of F and These properties will be

564

R . E . NETTLETON

proved more explicitly for a simple model having one spin per unit cell ( K H 2 P 0 4 has four) in Section 5, and a summary of qualitative conclusions will be given in Section 6. 2. Excitation Energies and Operators We proceed to form the commutator of the Hamiltonian (5) with Sf and Sj~ with a view to identifying the 6k, sk in equation (1). As a preliminary, we introduce two more approximations. Firstly, we restrict the exchange type interactions to nearest and next-nearest neighbors denoted by E and represent nnn

the effect of more distant neighbors by a molecular field with interaction cons t a n t y. Thus •

7i i Zt

Zj

2

7uZiZi

+

2 y < Z > Z

(

(6)

.

Secondly, following DeGennes [6], we rotate the coordinates through an angle — cos - 1 & about Y, and determine •& so t h a t the terms linear in X( drop out in the rotated frame, provided we replace Zt by C (1) = (Z") and Zt Zt Zk b y (Zi Zj Zky = 0 ( 3 ) everywhere in these terms. W i t h all operators referred to the rotated frame, H is now H » -

1

L 7n nnn

+

S ijkm

^

\T (1 - tf2)1'2 + y2

= [&*ztz,

+

(i

-

& * ) X

Dijkv

Xt

X]

C (1) ] E Zt

x< x t ]

t

X

]

Z

k

Z

+

m

- - & ( 1

Zj, Z j z

xk

-

k

#2)3/2

x

t

x

1

x

k

z

n

z„

(7)

Terms linear in X have been dropped from H0 by requiring t h a t § satisfy - r & + Z c —oo while and thus P , are still > 0. To see this, differentiate (15) with respect to temperature and approximate: 8 fìke 8T

so that from (15) 16 J 0

where

cm

Qks 8 J 1 ë'r

0C oo. Accordingly, unless the parameters are so chosen that where > 0 which would be unphysical, there must be a point

Fig. 3. Polarization vs. temperature for first-order transition. 8P/82 7 - oo at Tc which is below T„, the temperature at which should vanish according to equation (15). F o r Tc < T < T„, S a n d P are imaginary. Raising r increases Tc until it coincides with T, 37

physica 24/2

(19)

as •& -> 0, e01 -> 0 the curly bracket 06 ,(I) /8T is everyT c where the left-

R. E. Nettliiton

568

hand side changes sign and where the curly bracket vanishes while is still > 0. Since the right-hand side of (19) is < 0, dC^/dT -> - o o at Tc. This behavior is diagramed in Fig. 3, where the finite jump in polarization determined by (13) and (14) at Tc characterizes a first-order transition. From equation (12), we see that increasing y will gradually raise e01 and the entire lowest spin-wave branch corresponding to a given until the divergence of 6C,(1>/6T coincides with the temperature where •&• = 0. The transition is then second-order. If _T is now reduced, e.g., by deuteration, the negative C^/1-1) terms in the curly bracket of (19) are decreased, and this bracket will now vanish for higher values of s/T ,i.e., while is still > 0. This can be seen more easily in the simplified model of the next section, where we can prove that reducing r must convert the transition from second to first order. A t a second-order transition where •& vanishes, the four-body forces, multiplied by do not contribute to determining T c . Near a first-order transition these forces contribute to the curly bracket in (19) the term

which will be > 0 if C ^ is small, thus tending to raise the temperature at which the bracket vanishes. I t is thus possible for H S R to narrow the gap between T c and the point where § tends toward zero, and thus to increase somewhat the apparent Te in K D 2 P 0 4 . Accordingly four-body forces appear to increase the isotope effect at the transition and to decrease the effect predicted for P 0 . The divergence of QC^/dT given by (19) is not the logarithmic one we should expect from measurements of Cv [9]. This can be seen by differentiating again and showing that (19) implies Q^CW/dT2 = 0{[dC^/dT]3) suggesting C' ~ l 2 *—• (Tc — T) l as T TQ. This is not surprising, since the zeroth approximation may well break down near a second-order transition. Also, the volume dependence of the Jt and y arising from spin-lattice interaction, which has been neglected, may have an appreciable effect on the behavior of P near Tc [9]. 5. Simplified Model Calculations The properties shown to be implicit in equation (19) can be demonstrated explicitly for a model with one spin per unit cell. This is not a good representation of K H 2 P 0 4 , but it should be relevant to NaN0 2 . Since the transition in this substance is believed first-order, we retain r but drop ?/;(4) which has no meaning in this model. The basic equation (15) is changed only by the fact that we no longer sum on s and that the calculated dispersion relations for NaN0 2 -type lattice suggest we replace 16 J - » 10 J. Let us assume initially that r > 0 is so determined that the model gives a second-order Tc, at which OC^/Si7 -> —oo, coincident with T0, defined to be the temperature at which # = 0. Since & is small near this point, the Independence of C(r> can be determined by expanding the basic equation in powers of Thus, evaluating all coefficients at $ = 0, we have n + 2 0k] = F = A1 ft2 + A2

+ 0(&6) ; (20)

l —

2

efc

2

-y

[l+2 0 + as T T~, since otherwise d&^dT, and thus dC^joT [cf. equation (13)], is finite at Tc. By its structure, A1 = 0 (Tc — T), implying dA proportional to Tc — T near the transition provided ^42(TC) > 0. Subject to A2 > 0, then, we shall have ~ (Tc - Tf>2 and dC^/QT {Tc~ T)-1'2 as predicted in the previous section. Explicitly, 9 r2 3 r* A2 = N-1 Z — T)'1 &k (0k + l)

j(x

k Bk [

l & + J 2 _£!_ (i + 2 0 k ) + ~ [1 + 2 0 k ] 4 £fc X T Ek

9 r2

e To show A2(TC) > 0, we can use A^T^ = 0 and apply the mean value theorem to both u4's. Equation (20) determines at any T near the point T0 where = 0 and in particular its value (?2(_T) at T 0. This result can explain why deuteration changes the character of the Curie point in K H 2 P 0 4 and why we should also expect a first-order transition in NaN0 2 , where J1 should be small because of the size of the rotating N 0 2 group. Since A1 vanishes when Tcoincides with # = 0, equation (22) implies that A^ 0 if r is reduced slightly to make Tc < T0. According to equation (20), this means that # 2 < 0 in the region I\. < T < T0 so that 0 is imaginary immediately above Tc. This supports the conclusion that Tc must be a first-order transition point whenever r > 0 and Tc 0 so that for KH 2 P0 4 , where tunneling is large, we cannot rely on experience with analogous magnetic problems to assess the value of the approximation. An investigation of the effect of the 0(B) correction to these calculations, i.e., of the first order of statistical perturbation theory, will be undertaken in a later paper. Acknowledgements

The author is indebted to D. C. Wallace of Sandia Laboratory for numerous consultations during the course of the work and to J. L. Bjorkstam of the University of Washington for communication of his measurements on the isotope effect in K H 2 P 0 4 prior to publication. References [ 1 ] D . C. WALLACE, P h y s . R e v . 1 5 2 , 2 6 1 ( 1 9 6 6 ) . [ 2 ] R . BLINC a n d S. SVENTINA, P h y s . R e v . 1 4 7 , 4 3 0 ( 1 9 6 6 ) .

[3] J. C. SLATER, J. chem. Phys. 9, 16 (1941). [ 4 ] Y . TAKAGI, J . P h y s . S o c . J a p a n 3 , 2 7 1 ( 1 9 4 8 ) . [ 5 ] M. TUKONAGA a n d T . MATSTJBARA, P r o g r . t h e o r . P h y s . 3 5 , 5 8 1

C1966).

[6] P. G. DEGENNES, Solid State Commun. 1, 132 (1963). [7] F. JONA and G. SHIRANE, Ferroelectric Crystals, The MacMillan Comp., New York 1962 (p. 8 9 ) . [ 8 ] J . L . BJORKSTAM a n d R . E . OETTEL, p r i v a t e c o m m u n i c a t i o n .

[9] R. H. DONALDSON, Phys. Rev. 157, 366 (1967). [10] M. E. SENKO, Phys. Rev. 121, 1599 (1961). [ 1 1 ] H . B . SILSBEE, E . A . UEHLING, a n d V . H . SCHMIDT, P h y s . R e v . 1 3 3 , A 1 6 5 ( 1 9 6 4 ) .

(Received August 30, 1967)

B. PIETRASS : Application of t h e S.C.F. Method t o t h e Coulomb Lattice

571

phys. stat. sol. 24, 571 (1967) Subject classification: 6 Institut für Theoretische Physik der Technischen

Universität

Dresden

Application of the S.C.F. Method of Lattice Dynamics to the Coulomb Lattice By B.

PIETRASS

Anharmonic effects of the Coulomb-lattice vibrations are investigated b y the time-dependent s.c.f.-method. I n contrast to the traditional t r e a t m e n t using the lattice theory of Born there are higher branches in t h e dispersion relation of the lattice vibrations which represent anharmonic elementary excitations. I n t h e harmonic oscillator approximation for the lows, c.f.-states, whichisjustifiedforsufficientlylowdensity,Kohn's s u m r u l e is generalized to include t h e anharmonic branches. Taking into account anharmonic branches, t h e instability of the lattice state above a certain density is indicated b y imaginary eigenfrequencies in the transverse vibration mode. The contribution of the anharmonic branches to the zero-point energy is evaluated. Anharmonische Effekte der Coulombgitter-Schwingungen werden auf der Grundlage der zeitabhängigen s.c.f.-Methode untersucht. I m Unterschied zur Behandlung mit der traditionellen Bornschen Gittertheorie treten in der Dispersionsrelation der Gitterschwingungen höhere Zweige auf, die anharmonische Elementaranregungen darstellen. F ü r genügend kleine Dichten können die s.c.f.-Zustände durch Zustände des harmonischen Oszillators angenähert werden. I n dieser Näherung gilt eine verallgemeinerte Kohnsche Summenregel, die die anharmonischen Zweige einschließt. Bei Berücksichtigung der anharmonischen Zweige wird die Instabilität des Gitterzustandes oberhalb einer bestimmten Dichte durch imaginäre Eigenfrequenzen im transversalen Zweig der Gitterschwingungen angezeigt. Der Beitrag der anharmonischen Zweige zur Nullpunktsenergie wird untersucht.

1. Introduction The s.c.f. method of lattice dynamics developed by Brenig [1] and Fredkin and Werthamer [2] shows some features which are different compared with the standard method of lattice dynamics known as Bora's lattice theory [3]. The purpose of this paper is to exhibit these new features for a special model system with well defined interactions which allow quantitative calculations to be carried out. The Coulomb lattice with its pure Coulomb interactions between the lattice particles is not a very realistic model for solids. However, it has been investigated by several authors [4] in relation with lattice dynamics of metals [5,6] and also as the low-density limit of the electron gas [7 to 11], Up to now the dynamics of the Coulomb lattice has been studied only on the basis of Born's lattice theory [6, 9, 10]. The characteristic features of Born's theory are the expansion of the potential about the equilibrium positions of the lattice particles, the introduction of normal coordinates in the harmonic approximation, the quantization of the normal mode oscillations and the t r e a t m e n t of the anharmonic terms of the expansion as phonon-phonon interaction.

572

B . PlETRASS

The s.c.f. method does not start with an expansion of the potential energy. I t uses the concept of the average field governing the motion of the particles. The average potential near the lattice sites and the low-lying s.c.f. states are harmonic-oscillator-like. Brenig observed that particle-hole excitations propagating through the lattice turn out to be phonons. Taking into account excitations to the lowest excited state which is threefold degenerated he has shown that one obtains the harmonic approximation of Born's lattice theory. The natural extension of Brenig's approach is to take into account also excitations to higher states. Consequently, in the dispersion relation of the elementary excitations there appear higher branches which are not present in Born's theory. In this paper the higher branches of the Coulomb lattice are investigated. We limit ourselves to the case of zero temperature which leads to some simplifications in comparison to the finite temperature case considered by Fredkin and Werthamer. Assuming the particles to be in harmonic oscillator ground-states, the average potential is calculated using Ewald's technique in Section 2. From the density dependence of the average potential it is concluded that for sufficiently low densities also higher s.c.f. states can be approximated well by harmonic oscillator states. In the oscillator approximation for the s.c.f. states, in Section 3 Kohn's sum rule is generalized to include the higher branches. This has recently been reported in a short note [12]. Brenig has pointed out that exchange effects can be treated in principle by the s.c.f. method. He has proposed to use Wannier functions which are orthogonal with respect to different lattice sites for the formulation of the s.c.f. method including exchange. However, no suitable Wannier functions are known for practical calculations. In Section 3 we outline an alternative possibility to treat exchange using orbitals which are not orthogonal with respect to different lattice sites. This approach is suggested by the use of non-orthogonal orbitals in magnetic problems [13, 14]. In Section 4 it is shown that above a certain density there occur imaginary frequencies in the transverse vibration mode, indicating lattice instability, if one goes beyond the harmonic approximation in the s.c.f. method. Taking into account the approximations in the present calculation, the electron lattice is estimated to be stable for r s > 5. The rB-value of the melting density found here is lower than those given earlier by other authors [8, 9, 11], but is in accordance with Carr [10], who came to the same conclusion on a different way. The derivation of the instability from lattice dynamics seems to be more satisfactory than the methods used by the authors mentioned in the problem of electron lattice instability. This problem was raised by Wigner's fundamental work on the electron gas [7]. Nozieres and Pines [8] used Lindemann's empirical criterion, which describes the melting of solids. In the Coulomb lattice, however, the interaction between the lattice particles is different from that in a solid, so that Lindemann's criterion leads to incertain results. This has already been pointed out by Coldwell-Horsfall and Maradudin [9] and by Carr [10], De Wette [11] has estimated the melting density from the investigation of the bound states in the average potential of the lattice particles. But his assumptions on the average potential are not convincing. In Section 5 the zero-point energy of the Coulomb lattice vibrations is evaluated, proceeding like Fredkin and Werthamer. In the case of zero temperature and in the harmonic oscillator approximation, however, the calculations can be

Application of the S.C.F. Method to the Coulomb Lattice

573

put forward further. I n the harmonic approximation, the usual expression for the zeropoint energy is rederived. Taking into account higher branches, there arise correction terms similar to the anharmonic corrections in Born's theory which have been considered by Carr [10]. As far as numerical values are given they are related to lattice particles which have electron mass. I n this case we adopt the notation used for the electron gas. The particle density JV is characterized by the dimensionless parameter rs which is the radius of the unit sphere (Volume JV' 1 ) measured in Bohr radii a0. The energy is given in Rydberg units (1 ryd = e2/2 a0). I n the present state of investigation no final answer can be given to the question whether elementary excitations belonging to the higher branches of the dispersion relation exist in real solids, or whether they are spurious, i.e. due t o the approximations of the s.c.f. method. The anharmonic quasi-particles follow directly from the concept of the time-dependent average field. Hence one has to investigate to what extent this concept corresponds to reality. I n real solids the higher branches would be of a different form because of the different interaction of the lattice particles. I t may be that in solids the higher s.c.f. states are not harmonic-oscillator-like. In addition, the lifetime of the anharmonic quasi-particles and their relation to multi-phonon excitations has to be investigated. The s.c.f. method suggests that the anharmonic quasi-particles could be excited by quadrupole fields of scattering particles, e.g. thermal neutrons. I t is hoped that a more detailed discussion of this point can be given in the future.

2. Effective Potential and S.C.F. States From investigations of the electron gas [7 to 10] it is known that in the lowdensity limit the particles tend to forma lattice. A t sufficiently low density the lattice state has lower energy than states with uniform particle density. The lattice state has not the full translational symmetry of the Hamiltonian which is given by

H = E h + E vtf + E i

i/4

(22a)

give already a qualitative picture of the change of the dispersion relation with density. The dimensionless quantity Mnn-j(h eo0)2 determines the magnitude of the splitting of the energy levels n and n' due to their coupling. From (22a) it follows that in the harmonic approximation (n = n' = 1) the picture of the dispersion relation does not change with density (a>0is held fixed). Going beyond the harmonic approximation, the form of the dispersion curves depends on density. The splitting of the higher levels goes to zero as rs oo. Diminishing r s the splitting of the higher levels is increasing, and the harmonic branches are lowering, so that at a certain critical value of rs negative o'kx will appear, indicating lattice instability. The dispersion curves, which have been calculated for (0,0, k), n = 1 , 2 and several r s , show that negative a»^ appear first in the transverse modes at km n\2 a and for a rs-value between 1 and 2. To estimate ? s the following remarks are necessary: 1. The dispersion curves were calculated only for wave vectors (0, 0, k) with moderate accuracy. I t cannot be excluded that at other fc-direction imaginary frequencies may occur already at a somewhat greater r s -values. 2. The harmonic oscillator approximation for the s.c.f. states is justified only for rs 5. 3. The coupling to levels higher than n = 2 has been neglected. 4. In the evaluation of the matrix elements, the approximation 0 & 1 has been made (cf. Appendix). The largest terms neglected thereby contain a factor e x p ( — 1.55 r\>2).

5. Exchange effects were neglected. Their order of magnitude has already been estimated in Section 3.

579

Application of the S.C.F. Method to the Coulomb Lattice

Therefore, from the present calculation it can only be concluded t h a t t h e electron lattice is stable for r s S; 5. 5. Zero-point Energy of the Lattice Vibrations To evaluate the zero-point energy we follow Fredkin and Werthamer [2], b u t we limit ourselves to the case of zero temperature which leads t o some simplification. The zero-point energy is expressed by the linear response function

ZrjSF:

oo E = E

c l

- h

Z

fdm'

k J m

r ^ ^ I m ( g )

J ¿ n

\°F/k,m+iE

.

(23)

r is the displacement, F is the external field, = r F . E c l is the classical lattice energy and the second term on the right-hand side of (23) represents the zeropoint energy of the lattice vibrations. The integrand in (23) is t o be taken a t variable mass m'. The linear response function is calculated from (18), )

\ o f /km

= 2 h ~ * Z (ft)2

X

XX'

f i* r« o t v o / t . o « . ' o •

(24)

I n the oscillator approximation for the s.c.f. states we have (0|x|n T i ) = m oJq d n i dn and the corresponding expressions for the m a t r i x elements of y and z. Thus,

Z —- m, ryd •—' m, we get for the energy per particle the expansion = £h + ( Cl r~* + c2 a-^ + c 3 r-3 + • • •) ryd . (30) The correction terms of the harmonic approximation depend on r s in the same way as the anharmonic contributions do in Born's theory. The latter have been considered by Carr [10]. This indicates that there is a correspondence between the anharmonic effects in Born's theory and the higher branches in the s.c.f. method. e

Acknowledgements

The author wishes to thank Prof. G. Heber and Prof. W. Macke for their help and encouragement. Discussions with Dr. G. Lehmann, W. John, and H. Wonn are gratefully acknowledged. Appendix Matrix

elements

In this appendix the matrix elements appearing in (17) are evaluated in the harmonic oscillator approximation for the s.c.f. states: /0

n

( vk \o

T

i\

n' r

i'j

^

{R

0

r'

\R'

0

E^e^«"»»

«

\R n T

i\

I R'

i'f

n' P

.

(Al)

It is convenient to express the matrix element between the states |n jV i) as linear combinations of the matrix elements between the states n2 n3) and to evaluate /0

% n2 n3\

\0

r

=

drd r'e°

J

«

|r-r'-H|

(0|f) (i"|Wi n2

n3)

(0|r')

{r'\n[ né n's) =

e2 &»+»'( — l) n (2»+»' nx! n2! n3! n{! n2! n'z J) 1/ 2 ' where n = % +

-f n3, R = (X, Y, Z), and

J(R) = J

r

drr ^

R|

(0|r)> (0|f)» =

R~l.

(A6)

All matrix elements may be expressed as linear combinations of such lattice sums and thus the calculation of the matrix elements (Al) is reduced to the calculation of certain lattice sums. The lattice sums have obvious symmetry properties. As far as numerical values for the matrix elements were needed in the present calculations, the corresponding lattice sums were calculated for the b.c.c. lattice by direct summation over 64 points which are nearest to the origin. The values obtained for the lattice sums are accurate to about 1 0 % . More accurate values can be obtained by Ewald's method [18, 19]. References [1] W . BRENIG, Z. P h y s . 171, 6 0 (1963). [2] D. R . FREDKIN and N. R . WERTHAMER, Phys. R e v . 1 3 8 , A 1 5 2 7 (1965).

[3] M. BORN and K. HUANG, Dynamical Theory of Crystal Lattices, Oxford 1954. [4] D. PINES, Elementary Excitations in Solids, W. A. Benjamin, New York 1963 (p. 28, 91, and 231). [5] J . BARDEEN and D. PINES, P h y s . R e v . 9 9 , 1 1 4 0 (1955).

[6] C. B. CLARK, Phys. Rev. 109, 1133 (1958). [7] E. P. WIGNER, Trans. Faraday Soc. 34, 678 (1938).

[8] P . NOZIERES and D. PINES, P h y s . R e v . Ill, 4 5 5 (1958). [9] R . COLD WELL-HORSF ALL and A. A. MARADUDIN, J . m a t h . P h y s . 1, 3 9 5 (1960).

[10] W. J. CAR® JR., Phys. Rev. 122, 1437 (1961). [11] F . DE WETTE, P h y s . R e v . 1 3 5 , A 2 8 7 (1964).

[12] B. PIETRASS, Phys. Letters (Netherlands) 24A, 555 (1967). [13] W . J . CARR JR., P h y s . R e v . 9 2 , 2 8 (1953). [14] W . J . MULLIN, P h y s . R e v . 136, A 1 1 2 6 (1964). [15] K . FUCHS, P r o c . R o y . Soc. (London) A 1 5 1 , 5 8 5 (1935).

[16] JAHNKE-EMDE, Tables of Higher Functions, Teubner-Verlag, Leipzig 1960. [17] H. JONES, The Theory of Brillouin Zones and Electronic States in Crystals, NorthHolland, Amsterdam 1960. [18] P . P . EWALD, Ann. P h y s . (Germany) 6 4 , 2 5 3 (1921). [19] M. H . COHEN and F . KEFFER, P h y s . R e v . 99, 1128 (1955). (Received

July 14,

1967)

P . FRÖHLICH und 0 . ADAMETZ: Umwandlung von OH-Zentren in U-Zentren

583

phys. 8tat. sol. 24, 583 (1967) Subject classification: 10.2; 22.5.2 Institut für experimentelle

Physik der Martin-Luther-Universität

Halle

Umwandlung von OH-Zentren in U-Zentren bei KCl-Kristallen durch elektrolytische Verfärbung Von F . FRÖHLICH u n d 0 .

ADAMETZ1)

Mit Hilfe von Absorptionsmessungen im U V und I R wird gezeigt, daß sich durch elektrolytische Verfärbung von KCl-Kristallen OH-Zentren in U-Zentren umwandeln lassen, falls die OH-Ionen an Ca-Ionen gebunden im Kristall vorliegen. Die Zahl der gebildeten U-Zentren ist der Ausgangskonzentration der OH-Ionen im Kristall proportional und n i m m t mit der durch die Farbzentren übertragenen Ladungsmenge und mit steigender Temperatur zu. Gleichzeitig mit den U-Zentren entstehen Absorptionsbanden im UV, die dem Sauerstoff zugeordnet werden können. B y absorption measurements in the ultraviolet and infrared region it is shown that electrolytic coloration of KCl crystals transforms OH centres into U centres if the OH ions in the crystal are present in the form of a complex with Ca ions. The number of U centres formed is proportional to the initial concentration of the OH ions in the crystal and increases with the electrical charge transported by the colour centres and with increasing temperature. Simultaneously with the U centres, absorption bands are formed in the ultraviolet which can be related to oxygen.

1. Einleitung Die elektrolytische Verfärbung von Alkalihalogenidkristallen wurde bereits von Pohl und Mitarbeitern [1] zur Aufklärung der Eigenschaften der F-Zentren verwendet. Ebenso h a t von Lüpke [2] Untersuchungen bekanntgemacht über die N a t u r der U-Zentren in fremdionendotierten KCl-Kristallen, die während der Züchtung der Kristalle an L u f t gleichzeitig elektrolytisch verfärbt wurden. I n neuerer Zeit wurden von Heiland [3] ausführlich die Bedingungen, u n t e r denen überhaupt eine elektrolytische Verfärbung möglich ist, und insbesondere der zeitliche Verlauf des Stromes beim Einwandern der Farbzentren u n d deren Konzentrationsverteilung in KCl-Kristallen untersucht. Eine Reduktionswirkung der durch den Kristall bewegten Elektronen auf Kationenverunreinigungen im Kristall haben vor einiger Zeit Hersh und Bronstein [4] an Alkalihalogenidkristallen und kürzlich Fong [5] an fremdionendotierten Erdalkalihalogenidkristallen beobachtet. Weiterhin gelang vor kurzem Fischer, Gründig und Hilsch [6] die Herstellung von definiert K0 2 -dotierten KCl-Kristallen und durch nachfolgende elektrolytische Verfärbung („F-Zentren-Spülung") die Reduktion von K 0 2 zu K 2 0 im Kristall, wodurch eine eindeutige Zuordnung der durch O2und 0 -Zentren hervorgerufenen Absorptionsbanden möglich wurde. I n der vorliegenden Arbeit sollte an KCl-Einkristallen der Einfluß einer elektrolytischen Behandlung auf das Absorptionsverhalten der Kristalle im ultravioletten und infraroten Spektralgebiet untersucht werden. Dabei wurde gefunWissenschaftlicher Mitarbeiter am Institut für Einkristalle, Turnov, CSSR. 38

p h y s i c a 24/2

584

F . FRÖHLICH u n d O . ADAMETZ

den, daß in OH-haltigen Kristallen während der Elektrolyse OH-Zentren in U-Zentren umgewandelt werden können. Nachfolgend werden deshalb insbesondere an geeignet Ca- und OH-dotierten KCl-Kristallen die Bedingungen untersucht, unter denen eine solche OH-U-Umwandlung möglich ist. 2. Experimentelles Die benutzten KCl-Kristalle wurden nach dem Kyropoulos-Verfahren an Luft oder bei 30 Torr Reinst-Argon in Platintiegeln aus analysenreinem Material gezogen. Aus größeren Kristallstücken wurden Proben von 10 X 10 X 20 mm 3 herausgespalten. Diese Proben wurden der Elektrolyse zwischen einer spitzen Molybdänkathode und einer ebenen Graphitanode bei Temperaturen zwischen 530 und 700 °C in Argonatmosphäre ausgesetzt. Der zur Verfärbung notwendige hermetisch abgeschlossene Kontakt zwischen Kathode und Kristall wurde durch eine Feder gewährleistet, die außerhalb des Ofens angebracht die Kathodenspitze während der Elektrolyse mit regulierbarer K r a f t in denKristall drückt. Dadurch werden unerwünschte Sekundärreaktionen des an der Kathode gebildeten Kaliums verhindert. Die von der Kathode ausgehende Farbzentrenwolke bewegt sich im elektrischen Feld, bis das ganze Kristallvolumen verfärbt ist. Zur ursprünglichen Ionenleitung kommt dabei ein durch die F-Zentren bedingter Elektronenstrom hinzu, der bei den zu beschreibenden Experimenten stets wesentlich größer war als der Ionenstrom. Durch Umpolen des elektrischen Feldes lassen sich die F-Zentren wieder aus dem Kristall entfernen. Eine solche Entfärbung wurde bei den nachfolgenden Messungen — wenn nicht anders vermerkt — stets am Ende der Verfärbung durchgeführt. Die Absorptionsmessungen wurden bei Zimmertemperatur durchgeführt, im Bereich zwischen 190 und 1100 nm mit Spiegelmonochromatoren und geeigneten Multipliern, in der Umgebung von 2,8 ¡i.m mit dem Unicam Spektralphotometer S.P. 700. Die Zentrenkonzentrationen lassen sich nach den bekannten Formeln aus der Größe des Absorptionskoeffizienten im Bandenmaximum bestimmen. Wegen der Unsicherheit der jeweiligen Oszillatorenstärken wird nachfolgend jedoch als relatives Maß für die Konzentration immer nur der Absorptionskoeffizient angegeben. 3. Ergebnisse und Diskussion 3.1 Die Umwandlung

von OH-Zentren

in

U-Zentren

Zunächst wurden nicht zusätzlich vorbehandelte (,,as cleaved") Proben eines an Luft hergestellten („air-grown") Kristalles untersucht. Diese zeigten im Ausgangszustand im UV eine schwache OH-Bande bei 204 nm. Nach elektrolytischer Verfärbung ergab sich eine kräftige U-Bande (vgl. Fig. 1). Da der Defekt, der für die U-Bande verantwortlich ist, als ein Hydridion auf einem Anionenplatz angenommen wird, kann geschlossen werden, daß das U-Zentrum während der Elektrolyse aus dem OH-Zentrum gebildet wurde, was durch die gleichzeitige Andeutung einer Sauerstoffionenbande unterstützt wird. Zum empfindlichen Nachweis dieser neben der U-Bande auftretenden Absorption wurde an einem unbehandelten Kristall von 20 mm Dicke das Absorptionsspektrum gemessen und danach der Kristall 8 min bei 650 °C elektrolytisch verfärbt. Nach Abspaltung der durch die Elektroden mechanisch beschädigten Oberflächen wurde erneut die Absorption gemessen. Die Differenz der Absorptionskoeffizienten aus den beiden Messungen ist in Fig. 2 dargestellt. Sie zeigt direkt die durch die elektrolytische Verfärbung bewirkte Absorptionszunahme

585

Umwandlung von OH-Zentren in U-Zentren bei KCl-Kristallen '-EM 60 5,5

—E(eV) 5,0 4,5

4,0

2 °240 260

1 300 320 (nm)

Wellenlänge (nm) l'ig. 1. Extinktion eines KCl-Kristalles als Funktion der Wellenlänge

Fig. 2. Zunahme Ak des Absorptionskoeffizienten nach Elektrolyse eines KCl-Kristalles als Funktion der Wellenlänge im UV

an. Bei 240 n m ist der langwellige Ausläufer der U-Bande zu erkennen. Die sich anschließende breite Absorptionsbande k a n n man nach der Lage des Maximums bei 295 n m nach Rolfe, Lipsett und King [7] u n d Fischer, Gründig und Hilsch [6] O—-Ionen zuordnen. Auch die Breite der Bande ist mit der aus [6] f ü r Zimmertemperatur abzuschätzenden Halbwertsbreite verträglich. Der langwellige Ausläufer der bei 215 n m zu erwartenden 0 — - B a n d e geht in der U-Bandenabsorption unter. 2 ) U m die zu vermutende OH-U-UmWandlung genauer verfolgen zu können, wurden die Experimente mit definiert OH-dotierten Kristallen wiederholt. Durch die Zugabe von 4 x 10~3 Molbruchteilen OH/KC1 zur Schmelze ergab sich im Kristall eine Absorptionskonstante im Maximum der UV-Absorptionsbande von 2,2 m m - 1 , was einer Konzentration von etwa 4 x 10~4 OH/KC1 entspricht. Die nach der gleichen elektrolytischen Behandlung durchgeführte Absorptionsmessung zeigte jedoch keine OH-Umwandlung, sondern die OH-Bande war sogar u m etwa 10% vergrößert. Ein Kontrollversuch mit einer Probe aus demselben Kristall, der nicht elektrolysiert, aber der gleichen Temperbehandlung ausgesetzt wurde, zeigte ebenfalls eine um den gleichen Betrag vergrößerte OH-Bande. Diese Erhöhung der OH-Bande wird offenbar also nur durch die Temperaturbehandlung und nicht durch die Elektrolyse bewirkt. Der Grund hierfür sind möglicherweise die auf den Probenoberflächen immer vorhandenen adsorbierten Wassermoleküle (die sich nur durch Tempern im Hochvakuum entfernen lassen). Während der Temperung können daraus durch thermische Dissoziation gebildete OH-Ionen in die Oberflächenbezirke des Kristalles eindiffundieren.

Es erhebt sich hiernach die Frage, wodurch das unterschiedliche Verhalten des air-grown Kristalles und des bewußt mit OH dotierten Kristalles nach der gleichen Behandlung bedingt ist. Bereits bei Untersuchungen über die Tieftemperaturleitfähigkeit und Röntgenverfärbbarkeit von Ca- u n d OH-dotierten KClKristallen [8] wurde gefunden, daß die Experimente mit OH-dotiertem Kristallmaterial im starken Maße durch andere Fremdionen, insbesondere zweiwertige 2 ) Kürzlich konnte gezeigt werden, daß auch die umgekehrte Reaktion im Kristall möglich ist: durch eine Temperung in Sauerstoff werden U-Zentren in OH-Zentren umgewandelt [10].

38'

586

F . FRÖHLICH u n d O . ADAMETZ

Ol ohne OH

b) OH-dotiert

1

c> A / !.\0Hmaskiert Ia 1

1 1

l" \durchCa"_

/ \

-

\

2

' V

OH

200

Fig. 3. Absorptionskoeffizient von K C l - K r i s t a l l e n unterschiedlicher Dotierung als F u n k t i o n der Wellenlänge im U V . K u r v e 1 ( O ) vor der Elektrolyse, K u r v e 2 (A) nach der Elektrolyse gemessen

220

1

240 200

l

l

220

l

l

1

240 200 220

Metallionen, beeinflußt werden. Tatsächlich ergab bei den hier beschriebenen Experimenten eine Bestimmung der Ca-Verunreinigung, daß der Ca-Gehalt wesentlich größer war in dem Kristall, in dem die OH-U-Umwandlung stattgefunden hatte. Nach den Untersuchungen von Fritz, Lüty und Anger [9] können Ca-Ionen im Kristall OH-Ionen im Verhältnis 1 : 2 an sich binden, wahrscheinlich unter Bildung von Ca(OH) 2 . Das in diesem Komplex gebundene OH gibt keinen Beitrag zur ultravioletten OH-Bande, läßt sich aber durch die ultrarote OH-Ionen-Schwingungsbande bei 2,8 [xm nachweisen. Die Ergebnisse der elektrolytischen Behandlung in Kristallen mit unterschiedlicher Ca- und OH-Ionen-Konzentration sind in Fig. 3 für den ultravioletten und in Fig. 4 für den infraroten Spektralbereich in der Umgebung von 2,8 [im zusammengefaßt. Im ersten Kristall war durch spezielle Vorbehandlung des Ausgangsmaterials und durch Züchtung unter 30 Torr Reinst-Argon der Gehalt an OH-Ionen weitgehend beseitigt. Weder vor noch nach der Elektrolyse ist an diesem Material irgendeine Bande im UV (vgl. Fig. 3a) oder im I R (vgl. Fig. 4a) zu beobachten. In Fig. 3 b und Fig. 4 b sind die Ergebnisse für bewußt mit OH-Ionen dotierte Kristalle dargestellt. Man erkennt die charakteristischen OH-Absorptionsbanden im UV und I R . 3 ) Durch die Elektrolyse wird keine OH-U-Umwandlung bswirkt. Die Zunahme der OH-Banden wird — wie bereits erwähnt — durch die Temperbehandlung verursacht. In Fig. 3c und 4c sind

Fig. 4. Dur chlässigkeit von KCl-Kristallen unterschiedlicher Dotierung als Funktion der Wellenlänge im I I I . K u r v e 1 v o r der Elektrolyse, K u r v e 2 nach der Elektrolyse gemessen. I n Teilbild a sind die K u r v e n 1 und 2 im Kähmen der Meßgenauigkeit identisch 3 ) In gesonderten Experimenten wurde die genaue Lage des Maximums der infraroten Absorptionsbande in OH-dotierten Kristallen bei 2,76 {im gefunden mit einem auf der langwelligen Schulter schwach angedeuteten Nebenmaximum bei «¿2,78 [im.

Umwandlung von OH-Zentren in U-Zentren bei KCl-Kristallen

587

die Ergebnisse für „air-grown" Kristalle dargestellt, in denen die OH-Bande im UV durch Ca „maskiert" ist. Durch die Ca-Zugabe ist hier die Form der infraroten OH-Absorption in charakteristicher Weise verändert. 4 ) Nach der Elektrolyse zeigte sich eine starke U-Bande im UV. Im I R ist die OH-Absorption völlig verschwunden. Um zu entscheiden, ob der Ionenstrom an der Umwandlung beteiligt ist oder nicht, wurde eine Probe zwischen zwei Graphitelektroden elektrolysiert, was die Bildung von F-Zentren verhindert und nur einen konstanten Ionenstrom zuläßt. Eine anschließende Absorptionsmessung zeigte, daß keine OH-U-Umwandlung stattgefunden hatte. Der bestimmende Faktor der Umwandlung sind also die bewegten Elektronen.

Aus den Ergebnissen kann gefolgert werden, daß eine Umwandlung von OHZentren in U-Zentren offenbar nur stattfindet, wenn die OH-Ionen mit CaIonen im Kristall gebunden vorliegen. Deshalb findet auch bei dem in Fig. 1 dargestellten OH-U-Umwandlungsprozeß nicht — wie ursprünglich vermutet — eine Umwandlung der in Kurve 1 angezeigten OH-Bande statt, sondern die U-Bande wird auch hier durch — in Kurve 1 naturgemäß nicht erkennbare — Ca-maskierte OH-Ionen bewirkt. Die OH-Bande bleibt erhalten. Sie bewirkt eine Anhebung des kurzwelligen Ausläufers der U-Bande in Kurve 2. Die Zersetzung des OH-Zentrums kann schematisch also nicht durch die einfache Gleichung OH" + 2 e- -> H " + 0 — (1) beschrieben werden, weil bei freien OH-Ionen im Kristall keine U-Zentren (H _ -Ionen) entstehen. Unter Berücksichtigung des für die Umwandlung notwendigen Ca-OH-Komplexes läßt sich die Zersetzung des OH-Zentrums formal beschreiben durch Ca(OH) 2 + 4 er -> CaO + 2 H " - f 0 ~ . (2) Die Reaktionsprodukte H~ und 0 — werden durch die zugehörigen Absorptionsbanden tatsächlich nachgewiesen. Über die spektral-photometrischen Nachweismöglichkeiten von CaO in KCl-Kristallen liegen bisher keine Angaben vor. 3.2 Einfluß

der OH- und

Ca-Konzentration

Zur Untersuchung der OH-U-Umwandlung als Funktion der OH- und CaKonzentration wurden drei weitere Kristalle an Luft hergestellt, mit den in Tabelle 1 angegebenen Dotierungen zur Schmelze. Tabelle 1 Kristall

OH-Dotierung

Ca-Dotierung

Probe

A

keine zusätzliche Dotierung

2 x IO"3 Ca/KCl

1 2

oben unten

B

2 x 10" 1 0H/KC1

1 X IO"4 Ca/KCl

3 4

oben unten

C

4 x 10" 3 0H/KC1

2 x l 0 " 3 Ca/KCl

5 6

oben unten

Lage im Kristall

4 ) Die Absorption besteht aus einem steilen Hauptmaximum bei 2,75 ¡¿m und einem breiteren Nebenmaximum bei 2,81 (xm (vgl. auch [9]).

588

F . FRÖHLICH u n d O . ADAMETZ E(eV)

Fig. 5. XJ-Bandenabsorption von KCl-Kristallen mit verschiedenen 011- u n d Ca-Dotierungen nach elektrolytischer V e r f ä r b u n g ( f ü r die P r o b e n 1 bis 6 in Tabelle 1). K u r v e I zeigt die Absorption der P r o b e n im Ausgangszustand vor der Elektrolyse

Fig. 6. Absorptionskoeffizient der d u r c h die Elektrolyse b e w i r k t e n U-Bande als F u n k t i o n des ursprünglich im Kristall vorliegenden OH-Absorptionskoeffizienten im III bei 2,75 ¡xm

Die Zahl der f ü r die Messungen zur Verfügung stehenden Kristallbereiche mit unterschiedlichen Dotierungskonzentrationen konnte wegen der Segregation beim Kristallwachstum verdoppelt werden, indem jeweils Proben vom oberen u n d unteren Ende der etwa 10 cm langen Kristalle verwendet wurden.

Vor der Elektrolyse zeigten alle Proben keinerlei Absorptionsbanden in dem sich an die Eigenabsorption anschließenden UV-Bereich. Kurve I in Fig. 5 ist repräsentativ für die Absorption aller hierbei verwendeten Kristallproben. Die elektrolytische Behandlung erfolgte bei 650 °C. Die gesamte durch Elektronen übertragene Ladungsmenge betrug bei allen Kristallen etwa 3,5 mÄ min. Nach der Elektrolyse wurden in den Proben 1 bis 6 die in Fig. 5 dargestellten U-Banden gefunden. Deren Absorptionskoeffizient hängt von der Ausgangskonzentration der OH-Ionen im Kristall ab. I n Fig. 6 erkennt man einen linearen Zusammenhang zwischen dem Absorptionskoeffizienten im Maximum der durch die Elektrolyse gebildeten U-Bande bei 214 nm und dem für die OH-Konzentration im Kristall repräsentativen Absorptionskoeffizienten der infraroten Absorptionsbande bei 2,75 [im, gemessen vor der Elektrolyse. Der Konzentrationsgradient der OH-Ionen ist in den ,,air-grown" Kristallen sehr klein. Deshalb unterscheiden sich die resultierenden U-Zentren-Konzentrationen nur sehr wenig. Aus Gründen der Übersichtlichkeit ist daher in Fig. 5 nur eine Meßkurve für die beiden Meßproben 1 und 2 eingetragen. Bei allen Meßproben konnte nach der Elektrolyse in Analogie zu den Ergebnissen in Fig. 4c keine OH-Absorptionsbande im infraroten Bereich nachgewiesen werden. Es sei darauf hingewiesen, daß der Kristall C die gleiche OH-Dotierung zur Schmelze besitzt wie der allein mit OH dotierte Kristall aus Fig. 3 b und 4 b, in dem keinerlei OH —IJUmwandlung nachgewiesen werden konnte. 3 . 3 Einfluß

der Ladungsmenge

und der

Temperatur

Bei der Untersuchung der Abhängigkeit des Umwandlungsgrades von der Ladungsmenge und der Temperatur wurde dasselbe Kristallmaterial wie in

589

Umwandlung von OH-Zentren in U-Zentren bei KCl-Kristallen F i g . 7. Absorptionsltoeffizient im Maximum der IT-Bande als Funktion der durch Elektronen übertragenen Ladungsmenge für verschiedene Temperaturen



0

verfärbt verfärbt und entfärbt

50

100 150 Qe - (mA minj

ZOO

Fig. 3c (OH durch Ca maskiert) benutzt. Die Proben wurden bei verschiedenen Temperaturen und Elektrolysedauern behandelt. In jeder dieser Proben wurde der Absorptionskoeffizient im Maximum der entstandenen U-Bande ermittelt. In Fig. 7 sind diese Absorptionskoeffizienten über der durch die Elektronen übertragenen Ladungsmenge aufgetragen. Bei konstanter Temperatur nimmt die U-Zentren-Konzentration mit steigender Ladungsmenge zu. Bei den relativ niedrigen Temperaturen zwischen 500 und 600 °C wird dabei ein Sättigungswert erreicht. Die gestrichelte Kurve in Fig. 7 zeigt den Absorptionskoeffizienten der U-Bande in Abhängigkeit von der Ladungsmenge für den Fall, daß unmittelbar auf die Verfärbung eine elektrolytische Entfärbung folgte. 4. Zusammenfassung der Ergebnisse 1. Durch elektrolytische Verfärbung von KCl-Kristallen, in denen OH-Ionen mit Ca-Ionen zu einem Komplex gebunden vorliegen, entsteht eine U-Bande. Gleichzeitig verschwindet die infrarote OH-Ionen-Schwingungsbande: die OHBindung wird zerstört. 2. Zwischen der Ausgangskonzentration der an Ca-Ionen gebundenen OHIonen im Kristall und der resultierenden U-Zentren-Konzentration besteht ein linearer Zusammenhang. 3. Außer der U-Bande kann nach der Elektrolyse auch eine Bande nachgewiesen werden, die dem Sauerstoff zuzuordnen ist. 4. Mit steigender Temperatur und steigender Ladungsmenge nimmt die Zahl der gebildeten U-Zentren zu. Dagegen nimmt bei größeren Ladungsmengen nach elektrolytischer Verfärbung und anschließender Entfärbung die Zahl der gebildeten U-Zentren mit zunehmender Ladungsmenge ab. 5. In Kristallen, die keine OH-Ionen oder nur freie OH-Ionen enthalten, entstehen bei der Elektrolyse keine U-Zentren. Herrn Prof. Dr. W. Messerschmidt danken wir für sein förderndes Interesse, Herrn Prof. Dr. H. Bethge und Herrn Prof. Dr. A. Bohun (Institut für Festkörperphysik der Tschechoslowakischen Akademie der Wissensehaften in Prag) für viele wertvolle Hinweise.

590

F. FRÖHLICH und 0. ADAMETZ: Umwandlung von OH-Zentren in U-Zentren Literatur

[1] R. W. POHL, Phys. Z. 39, 36 (1938).

[2] A. D. VON LÜPKE, Ann. Phys. (Germany) 21, 1 (1934). [3] G. HEILAND, Z. P h y s . 1 2 8 , 144 (1950).

[4] H. N. HERSH und L. BRONSTEIN, Amer. J . Phys. 25, 306 (1957). [5] F. K. FONG, J . ehem. Phys. 41, 229 (1964); 41, 245 (1964); 41, 1383 (1964); RCA Rev. 25, 303 (1964). [6] F . FISCHER, H . GRÜNDIG u n d R . HILSCH, Z. P h y s . 1 8 9 , 7 9 ( 1 9 6 6 ) . [7] J . ROLFE, F . R . LIPSETT u n d W . J . KING, P h y s . R e v . 1 2 3 , 4 4 7 (1961).

[8] K . J . BERG, F . FRÖHLICH und G. HENSEL, Phys. Verh. 18, 95 (1967). [9] B . FRITZ, F . LÜTY und J . ANGER, Z. P h y s . 1 7 4 , 2 4 0 (1963).

[10] O. ADAMETZ, phys. stat. sol. 16, K161 (1966). (Received

September

27,

1967)

M . H Ö H N E a n d M . STASIW : E S R S t u d i e s o f D i f f e r e n t SE®

591

Centres

phys. stat. sol. 24, 591 (1967) Subject classification: 19; 10; 22.5.1 Institut für Kristallphysik

der Deutschen Akademie

der Wissenschaften

zu

Berlin

ESR Studies of Different Se|" Centres in Silver Halides By M. HÖHNE and M.

STASIW

The E S R spectrum of AgBr:Se crystals which have been irradiated a t 120 °K with green light is changed by heating the crystals to 170 °K. The spectrum obtained after heating was recently investigated and a t t r i b u t e d to a S e | ~ centre. I n t h e present paper t h e spectrum observed before heating is analyzed with regard to ¡^-values, and superhyperfine and hyperfine structure. This spectrum is also found to be due to a Se:] ~ centre of orthorhombic symmetry. The two centres differ in their surroundings a n d a model of this situation is developed. The investigations are extended to AgChSe. I n AgBr: Se-Kristallen, die bei 120 °K mit grünem Licht bestrahlt wurden, beobachtet man ein ESR-Spektrum, das sich durch Erwärmen der Kristalle auf 170 °K in ein anderes umwandelt. Das nach E r w ä r m e n erhaltene Spektrum wurde kürzlich untersucht und einem Se|"-Zentrum zugeordnet. I n der vorliegenden Arbeit wird das vor der E r w ä r m u n g beobachtete Spektrum bezüglich der g-Werte, der Superhyperfeinstruktur u n d der Hyperfeins t r u k t u r analysiert. E s erweist sich, daß auch dieses Spektrum auf ein S e ^ - Z e n t r u m von orthorhombischer Symmetrie zurückzuführen ist. Die zwei Zentren unterscheiden sich in ihrer Umgebung, f ü r die ein Modell entwickelt wird. Die Untersuchungen werden auf AgCl: Se erweitert.

1. Introduction In a recent paper [1] ESR studies of AgBr single crystals with impurities of Ag2Te and Ag2Se were reported. The non-irradiated crystals did not show resonance lines, apart from a weak and broad line. Irradiation with green light at 170 °K caused a spectrum, which was stable in a large temperature range. That spectrum will now be called M-spectrum.1) In the case of AgBr :Se other spectra of lower intensities were superimposed on the M-spectrum. These could be avoided in the following way: Irradiation of the AgBr:Se crystal at 120 °K yielded a different spectrum of comparable intensity, which will be called Tspectrum. After heating the crystal to 170 °K the T-spectrum had changed into the M-spectrum without underlying resonance lines of by-products. Irradiation of AgBr:Te crystals below 170 °K did not give any measurable resonance absorption. The M-spectrum was found to be due to a centre of orthorhombic symmetry with its principal axes parallel to [110], [110], and [001] directions. The centre turned out to consist of two equivalent impurity ions on lattice sites with one r

) This name for the spectrum and t h e corresponding centre is chosen in order to remind of t h e intermediate temperature range („mittlere Temperatur") of t h e formation. T h u s it can be distinguished from another Sejj" centre named centre of kind T („tiefere Temperatur").

592

M . H Ô H N E a n d M . STASIW

common hole, i.e. it is a Teij~ or Se^" pseudo-molecule. Thus the measured g-values, the hyperfine structure (hfs), and the superhyperfine structure (shfs) of the M-spectrum could be explained. Detailed knowledge of the surroundings of the centre and of the charge compensation was still missing. In the present paper the T-spectrum produced by irradiation of A g B r : S e at 120 ° K will be discussed. I t will be tried to get some further informations concerning the Seij" centre of kind M from the connections between the T- and M-spectra. Furthermore the investigations were extended to the system AgCl:Se. 2. Experimental The experimental equipment and the preparation of the crystals have already been described in [1], AgCl was prepared and cleaned in the same manner as AgBr. The distribution coefficient of selenium in AgCl is strongly different from 1. Nevertheless it was possible to prepare the samples from the monocrystals in such a way that the impurity concentration did not differ much from 100 ppm. 3. Results 3.1 The T-spectrum

in

AgBr:Se

As already described in [1] the T-spectrum is produced in the following way: The A g B r : S e crystals are irradiated at 120 ° K with green light of about 525 nm (combination of the Schott filters VG 8 and BG 19) and of the intensity 5 x l 0 1 6 quanta/cm 2 s for 1 hour. Then the production of centres is saturated. The concentration of the paramagnetic centres amounts to about 1/10 of the impurity. 3.1.1

g-values

The Fig. 1 to 3 show the first derivative of the absorption spectrum in the case that the external magnetic field H is parallel to a [110], f i l l ] , or [100] direction of the crystal respectively. J u s t like the M-spectrum the T-spectrum also consists of three lines for H |[ [110] and of two lines of different width for H || [111] and H || [100], From the angular dependence of the ¡/-values it may be concluded that the corresponding centre has orthorhombic symmetry : x and z are parallel to [110] and [1Ï0] directions, and one axis of symmetry, y, is parallel to a [001] direction. The principal values of the ¡/-tensor are given in Table 1. Table 1

!

9x

9v

9z

T-spectrum

AgBr: Se AgCl: Se

2.143 ± 0.001 2.101 ± 0.001

2.057 ± 0.002 2.078 ± 0.001

2.017 ± 0.001 2.023 ± 0.001

M-spectrum

AgBr: Se AgCl: Se

2.116 + 0.002 2.100 ± 0.001

2.086 ± 0.003 2.086 ± 0.001

2.005 ± 0.002 2.013 ± 0.001

In Fig. 4 the measured g-values are represented by circles. The full lines give the angular dependence of the ¡/-values calculated from the principal values given in Table 1. The deviation of about 0.003 observed during the approach

ESR Studies of Different S e |

-im

3.3

593

Centres in Silver Halides

3.0

36 H(k6) —

3.1

Fig. l

\\,



3.2

1

33

^

31 H(kS) — -

Fig. 2 Fig. 1 to 3. T-spectrum of an AgBr:Se crystal irradiated at 120 °K for different orientations of the magnetic field measured at 20 °K

¿75< 3.0

3.7

3.2

3.3

3.i

H(kû)—•

Fig. 3

Fig. 4. Angular dependence of the T-spectrum of AgBr:Se during rotation of the crystal around a [110] axis

0° 10° 20° 30° W [no] [m]

594

M . H O H N B a n d M . STASIW

of H to the [100] direction might be due to the influence of second-order effects causing a small lack of symmetry of the single lines, whereas the g-values were determined from decompositions into symmetrical lines. 3.1.2

Superhyperfine

structure

The line widths and line shapes at 4 and 20 ° K are equal. This independence of temperature suggests that the line width is caused by interaction with nuclei. As natural selenium contains only 8 . 3 % of an isotope with non-zero nuclear spin the nuclei taking part in the interaction must belong to the neighbouring silver or bromine ions. The Fig. 1 to 3 show that the shfs is not resolved for any direction of H. More exact statements concerning the interaction are therefore impossible, deviating from the case of the M-spectrum. Nevertheless the large nuclear moment of bromine compared with that of silver (their ratio ranges from 16 to 20 for the different isotopes) suggests the assumption that also in the case of the T-spectrum the line width is determined by interaction with bromine nuclei. This assumption is supported by the comparison with the corresponding measurements in AgCl:Se. There the shf interaction must be smaller, because the nuclear moment of chlorine is about three times smaller. Indeed narrower lines are observed in the T-spectrum of AgCl:Se represented in Section 3.2. The ratio of the halfwidths of corresponding lines range from 1.5 to 4. The angular dependence of the M-spectrum could be described to a far extent by assuming a shf interaction with two equivalent bromine nuclei. Only for some directions (cf. Fig. 11 in [1]) an additional line broadening was observed, which might have been caused by a small deviation from orthorhombic symmetry. In A g B r : T e the shfs of the line corresponding to H || z (g = gz) (cf. Fig. 1 in [1]) was resolved. Unfortunately there was no resolution in the case of the M-spectrum in AgBr: Se; therefore the shf constants could not be determined numerically. B u t as the line with g = gz was the most broadened one in this case, too (cf. Fig. 5 to 7 in [1]), an analogous interaction with two bromine nuclei could be assumed. Considering the line widths of the T-spectrum in the Fig. 1 to 3, one finds two remarkable facts. Firstly the line with g = gz is only a little broader than the line with g = gx. Therefore one has no reasons for a conclusion that an analogous shf interaction with two bromine nuclei exists. Secondly those lines show an especially strong broadening which correspond to the above mentioned additionally broadened lines in A g B r : T e . The broadest line of the T-spectrum in AgBr: Se is that for which the angles between H and the «-axis and the z-axis respectively amount to 45° (left line in Fig. 3). The observed angular dependence cannot be explained by mere shf interaction, even not by taking into account terms of second order or by assuming other directions of the tensor axes or interaction with other nuclei. But the angular dependence can be understood if one takes into consideration nuclear quadrupole interactions with bromine nuclei and if one assumes that the principal axes of the quadrupole interaction tensors coincide with those of the ^-tensor according to the additional term of the spin Hamiltonian (cf. Low [2]): XQ=Z

{Q'«> [ V

- 1

/ (/ +

l ) j + Q"(0 (/CO* _

JW)J .

(1)

The superscript i refers to different bromine nuclei; I x , I v , I z are the operators of the nuclear spin components, Q' and Q" the constants of the nuclear quadru-

ESR Studies of Different Se^

Centres in Silver Halides

595

pole interaction. Let the eigenvalue of the z-component Sz of the electron spin be denoted by M and that of / 2 by TO. Then the energy contribution AE caused by quadrupole interaction with one nucleus is given to the second order by A

TO«

4 / ( / + l) — 1 — 8 mW! \ \(Q' k%x gl sin sin2 2Ôà -- k% k% g% g% cos cos22 , \n'i + -J-). A possible contribution of «-functions is taken into account by adding |(7„ + = I-1- (Sj — s2) + -f) J/2 u As the eigenfunctions must again form a basis of E' , they turn out to be |±>

=

a ,

< + Y/ +

a2

k ± - J / ± «3

1

These eigenfunctions were used to determine the g-values from the Zeeman part of the Hamiltonian 36z = (i (L + 2 S, H) neglecting overlap. Bohr's magneton is denoted by ¡3. The result was = —4 + 4 i a2 a4 , Agy = 4 a\ — 4 a2 o3 , A.gz = — 4 «3 + 4 a\ — 4 i a3 a 4 ,

(6)

4 where 2J |aj|2 = 1. ai was chosen to be real. ¿=1 In the special case of a spin-orbit interaction small compared with the influence of the orthorhombic field the following relations will hold for a negative spin-orbit coupling constant X: Agx, A 0 , The condition

> |Ag,\ . bg* < 0

(7) (8) (9)

is not restricted to this special case. The conditions (7) to (9) depend on nothing but the symmetry of the molecule in the crystal. The assumption that the Mspectrum in AgCl is due to a Sejj - centre as in the case of AgBr explains the signs of Agx and Agv as well as their magnitudes compared with Agz in the same way. The positive value of Agz, which is observed in contradiction to the condition (9), is not unterstood. This difficulty had already been relevant to the centre of kind M in AgBr. Because of the smallness Agz = 0.003 and the accuracy of 0.002 we did not think this discrepancy to be very important. But the value of Agfz = 0.013 in AgCl forces us to look for the reason of the discrepancy. During the discussion of the g-values of the M-spectrum in AgBr it was supposed that the positive Agz could be caused by the overlap of the eigenfunctions. In order to investigate this influence the overlap of selenium functions and the admixture of eigenfunctions of the neighbouring bromine and silver ions were taken into consideration. I t turned out that no change in the sign is to be expected, if the hole mostly belongs to the selenium states, as it is really true. A small lack in the equivalence of the ions is a further possible reason for the positive sign of Agz. This possibility was supported by the observation of such a lack in equivalence in the hfs of the T-spectrum in connection with A^-values which were still stronger positive. If the impurity ions are not completely equivalent the symmetry group of the centre is no longer D2h, but it consists merely 3»'

M. Hohne and M. Stasiw

602

of the operations E, C2(z), I C\(x), I C2{y). According to this fact there are always two eigenfunctions belonging to the same representation, viz. Zp the photoelectron can drift into the wide gap material, an external photocurrent being manifest. If either of the conditions lp > lE or Ee less than the spike height is fulfilled the electron is expected to relax into the conduction band notch. The result may be recombination in preference to a tunnelling type emission, in which case no photocurrent arises. Presupposing favourable geometry in the band profile the mechanisms by which the excess energy of the non-equilibrium electron relaxes are important. The processes involved are electron-phonon interactions in the form of lattice and impurity scattering and electron-electron interactions, the so-called 'Maxwellization' of the electron gas. While the former is significant in momentum scattering the amount of energy exchanged per interaction is small when the rules for momentum conservation are taken into account [5]. Thus the electron may travel a large number of lattice spacings before losing a significant amount of its excess energy, provided no electron-electron collisions occur. However, since the momentum exchanges involved in phonon collisions are large the electron follows a random path if no field is present. With an impressed field, as in the transition region of a p-n junction a large drift is superimposed on the random motion and the electron may have a sufficiently large energy-relaxation length to cross the barrier. On the other hand a typical collision between a thermal electron and an excited photoelectron may result in the latter losing a significantly large fraction of its excess energy. Frohlich and Paranjape [5] have shown that electron concentrations of less than 1014 cm - 3 can result in non-equilibrium electrons relaxing in energy by electron-electron rather than electron-phonon exchanges. This is the model assumed by Gribnikov and Melnikov [6] in the analysis of 'hot' electron injection effects in isotope heterojunctions. Their assumption is that electron-electron interactions predominate in the energy relaxation process of the hot electrons while making a negligible contribution to momentum scattering although the overall excess energy of the electrons is finally liberated to the lattice by electron-phonon collisions. In the present work we shall adopt the same premise and demonstrate how the increasing concentration of injected (tunnelling) dark-current electrons into the conduction band notch will reach a point

Forward Current Injection Modulation of Photocurrent in Heterojunctions Notch Narrow

n

High energy photoeiectron

'

-©I

617

width gap

materiat

^s^ide

-Q-.J i •e—

gap

material

Effective scattering cross section, a

Tunnelling -injected population

Fig. 2. Maxwellization of hot photoeiectron

where high-energy photoelectrons thermalize rapidly and are trapped in the conduction band notch with a consequent loss of photocurrent. We consider, then, a free electron gas model in which an evenly distributed layer of injected electrons lies in the conduction band notch prior to their loss by recombination or migration (we shall employ a comprehensive lifetime r to describe this loss without being more specific). Defining an effective interaction cross-section, a, for the scattering process we say, in approximation, t h a t the photoeiectron is sufficiently hot for the injected population to be considered stationary during the time of transit. The problem in its simplest form is then as shown in Fig. 2 where the probability, P, t h a t an electron crosses the notch region without collision is just the probability t h a t no member of the injected population lies within a. If the total number of injected electrons within a junction of area A is n, we then have

Here we have taken one collision to be sufficient to cause the photoeiectron to become trapped. The situation for the case where several collisions are required is discussed in the Appendix. If we now regard the process on a macroscopic scale we can, in equation (1), let A, and hence n, become large (A a) and, maintaining the quotient njA at the level determined by the injection rate, rewrite (1) as

""(l-j)'-«p(-if).

0)

where the limit is taken as a/A 0, n -»• 00, such t h a t n ajA is invariant. Now n = N r where N is the number of injected electrons per second and r (s) is the carrier lifetime. The probability that an electron escapes thermalization by the injected population then determines the efficiency of the photon-photocurrent conversion and equation (2) becomes

where I , J p o are the photocurrents with and without forward dark current / D ; e is the electronic charge. 40*

618

T. L. TANSLEY

3. Results To test equation (3) against experimental results it was necessary to measure the variation of short circuit photocurrent with dark current level, temperature being taken as a further parameter. It was found more convenient in practice, to measure these two variables separately in terms of the independent variable F a , the applied forward bias. A large number of p-n heterojunctions (narrow gap p-type material) in the two systems Ga^In^^As-GaAs and GaAs-GaAs^Pj were prepared by vapour phase epitaxy techniques described elsewhere [3, 4, 7]. Every working device in the series exhibited a sharp cut-off of photocurrent at a well defined value of forward bias, and this cut-off value proved extremely sensitive to ambient temperature.1) Of the diodes for which detailed measurements were collected, the two presented here are for a typical GaAs-Ga(AsP) unit (forward characteristics, Fig. 3) and a (Galn)As-GaAs heterodiode with a double tunnelling forward characteristic (Fig. 4). Fig. 5 and 6 show the experimental values of relative photocurrent versus forward bias for the same two specimens at a number of different temperatures. The unbroken lines are the calculated curves produced by substituting values of / D , taken from Fig. 3 and 4, into equation (3). Fig. 5 demonstrates the applicability of the theory in the simple case while Fig. 6 shows clearly the dependence of loss of photocurrent on the injected dark current where this

Fig. 3. Forward characteristics of a GaAs-Ga(AsP) p - n heterojunction (GO-222) at eleven temperatures 1 ) The cut-off was found to be insensitive to wavelength of illumination (3), within the band of detection, a low-resolution monochromatic beam in the peak response region being used in the actual experiments.

Forward Current Injection Modulation of Photocurrent in Heterojunctions

619

Fig. 4. Forward characteristics of a (Galn)As-GaAs p-11 heterojunction (IQ-029C) at seven temperatures. The broken lines A, A', B , and B ' are extrapolations of the high and low current linear regions in the cases where a bend appears in the characteristic (see text)

0.5 0.6 0.7 Applied forward bias IV)-

The broken lines A, A', B , and B ' in Fig. 4 are extrapolations of the individual straight line components constituting the forward characteristics of IG-029C at 195 and 165 °K. These lines have been used in equation (3), giving the values of photocurrent shown by the broken lines superimposed on the actual cut-off characteristics in Fig. 6. These lines serve to highlight the way in which the cut-off depends fairly critically on the exact form of the dark current injection characteristics. The numerical values of the product a x which were used to obtain the best fit are plotted against temperature in Fig. 7. For the case of the GaAs-Ga(AsP) heterojunction, forward biasses near the flat-band value were required to obtain sufficient injection for cut-off at low temperatures. Under these conditions the value of / p 0 , the photocurrent un-

Fig. 5. Experimental points and theoretical curves (continuous lines) for GO-222. The broken line is the envelope of the results in the regions where injection induced cut-off is not important and is taken as the effect of band flattening on the value of / p 0 (see text)

Fig. 6. Experimental points and theoretical curves (continuous lines) for IG-029C. The broken lines A, A', B , and B ' describe the cut-off that would occur if the extrapolations of Fig. 4 represented the actual injection current (see text)

. 100 90 iso s 70 -g 60 5 50 | W I 10 f 10 % 10 0

f\

\

*m'K • 292'K » o 237°K * 195°K 162°K a W°K . 32 r ~~ûF

0.6 0.7 OS 0.9 1.0 forward bias tV! ——

620

T. L. Tastsley

y" 10

Fig. 7. Variation of parameter a T with temperature

x 00 222A, pûaAs-n 6a!AsPÌ o IO 029C, plOalnlAs-nfaAs

Tu-

rn

150 200 250 300 T!°K)-

modified by injection effects, is reduced by the large reduction in transition field. The broken line in Fig. 5 represents the variation in J p 0 deduced from the envelope of the curves in the regions where injection induced cut-off is insignificant. The model has a further success in accounting for the increase in the length of the photocurrent tail, or lack of definition of cut-off voltage, at low temperatures in Fig. 5. This is because the injection rate has become limited by the series resistance of the bulk materials (Fig. 3) and the externally applied bias is larger, at any given injection rate, by the ohmic voltage drop in these regions. 4. Conclusions All the heterojunctions in a series of p - n units grown by the deposition of either of two ternary derivatives of GaAs on GaAs substrates have been observed to show a loss of photocurrent when forward bias was applied. The voltage at which the photocurrent disappeared was further observed to be very sensitive to ambient temperature. Comparison of the photocurrent, 7 p , with the magnitude of the dark current, / D , showed the former to be related to the latter by the expression /p oo exp ( - ß ID) . Consideration of the energy relaxation of a hot photoelectron excited in the narrow gap material indicates that injected (dark current) electrons residing in the conduction band notch prior to recombination (or out-diffusion) absorb sufficient energy from the photoelectron to cause it to thermalize into the notch with a consequent loss of photocurrent. Quantitatively the photocurrent is significantly affected when the dark current density reaches about 1 0 - 3 A c m ~ 2 , corresponding to an interaction cross-section of about 10~ 12 m 2 assuming a lifetime of about 10~8 s, for the members of the injected population. This figure, corresponding to an effective scattering radius of about 1000 A seems excessively large for the scattering of a hot (photo-) electron by members of the population in thermal equilibrium. However, an assumed notch width of a few hundred Angstrom units, coupled with the above values of injection rate and lifetime, gives a quasi-equilibrium (notch-trapped) density of about 10 14 c m - 3 , which is about the level at which electron-electron energy relaxation appears, as discussed earlier [5].

Forward Current Injection Modulation of Photocurrent in Heterojunctions

621

Appendix The calculation outlined in Section 2 refers to a situation where one 'collision' represents the dividing point between electrons which cross the barrier and those which are trapped. We now make the argument one degree more general by saying that, during its traversal of the notch region, the photoelectron makes a number of collisions with tunnelling-injected electrons and that TO such collisions are necessary to lower it into the notch trap. Using the earlier notation, the probability that exactly m electrons occupy spatial positions within the region a is l

a

\

m

l - >

a \ « - ™

nl

. . . .

and the probability that less than m + 1 electrons occupy this region is thus D * 0, n oo) has been used to render equation (A3) in the form P

=

e x

V

[ - p n ] Z k =

o

(A4)

Ki

As one would intuitively forecast the effects of permitting the hot photoelectron a number of collisions with tunnelling-injected carriers are to shift the cut-off np-051 23156

Fig. 8. Theoretical cut-off curves for GO-222 at 325 °K

Applied forward bias

IV)—-

622

T . L . TANSLEY: Forward Current Injection Modulation of Photocurrent

to higher injection levels and to sharpen the cut-off. Comparison of Fig. 5 and 8 shows that the assumption that m = 0 gives the best fit to experimental results (this is also true when a r is optimized for each value of m tried). The implication seems to be that one electron-electron interaction between the hot photoelectron and the injected dark current is sufficient to thermalize the former into the trapping region of the conduction band notch.

References [1] [2] [3] [4] [5] [6] [7]

B. AGUSTA and R. L . ANDERSON, J. appl. Phys. 36, 206 (1965). T . B. RAMACHANDRAN and W . J. MORONEY, J. appl. Phys. 36, 2594 (1965). T. L. TANSLEY, phys. stat. sol. 23, 241 (1967). T. L. TANSLEY, phys. stat. sol. 18, 105 (1966). H . FROHLICH and B. V. PARANJAPE, Proc. Phys. Soc. 69, 21 (1956). Z. S. GRIBNIKOV and V . I . MELNIKOV, Fiz. tverd. Tela 7, 2912 (1965). C. A . FISHER, to be published. (Received July 31, 1967)

H. DIEPERS : Untersuchung von Fehlstellenagglomeraten (II)

623

phys. stat. sol. 24, 623 (1967) Subject classification: 11; 10.1; 21.1 Max-Planck-Institut für Metallforschung, Institut für Sondermetalle, Stuttgart

Elektronenmikroskopische Untersuchung von Fehlstellenagglomeraten in ionenbestrahlten Kupferfolien I I . E n t s t e h u n g der A g g l o m e r a t e 1 ) Von H.

DIEPEKS

Der Entstehungsmechanismus der in Cu-Polien durch Beschuß mit Ar-Ionen erzeugten Zwischengitteratomagglomerate wird untersucht [1]. Hierzu wird einerseits die Abhängigkeit der Tiefenverteilung der Agglomerate von der Orientierung der Folie und der Energie der Ionen, andererseits die Abhängigkeit der Agglomeratdichte von der Ionendosis (Dosiskurven) bei verschiedenen Folientemperaturen gemessen. Es ergab sich, daß die Agglomerate aus Eigenzwischengitteratomen des Cu aufgebaut sind, die durch gerichtete Stoßprozesse in Form dynamischer Crowdionen ins Gitter eindringen. Der Mechanismus der Agglomeration wird an Hand der Dosiskurven, der Tiefenverteilungskurven und weiterer Experimente zur Bestimmung der Agglomeratgröße in Abhängigkeit von der Dosis, der Temperatur und der Tiefenlage diskutiert. A study is made of the formation mechanism for the agglomerates generated in copper foils by ion bombardment [1]. The dependence of the depth distribution of the agglomerates on foil orientation and ion energy is measured. The dependence of the density of agglomerates on the ion dose (dose curves) is determined for different foil temperatures. The results show that the agglomerates consist of Cu interstitial atoms which penetrate the lattice by directed collision sequences as dynamic crowdions. The mechanism of agglomeration is discussed on the basis of the dose curves, depth distribution curves, and some further experiments in which the dependence of the size of the agglomerates on dose, temperature and depth position is measured. I.

Einleitung

I n Teil I [ 1 ] wurde nachgewiesen, d a ß die in Cu d u r c h B e s t r a h l u n g m i t 5 - k e V Ar-Ionen erzeugten und i m Elektronenmikroskop ( E M ) sichtbaren Defektagglom e r a t e v o m Z w i s c h e n g i t t e r a t o m t y p , d. h. v o n e i n e m K o m p r e s s i o n s f e l d , u m geben sind. N a c h R a u m t e m p e r a t u r b e s t r a h l u n g liegen sie in einer m i t t l e r e n Tiefe v o n c a . 1 4 0 A u n t e r h a l b der b e s t r a h l t e n Oberfläche. B e i m A b b r e m s e n d u r c h elastische S t ö ß e m i t den G i t t e r a t o m e n des Cu dringen 5 - k e V - A r - I o n e n in s t a t i s t i s c h e n S t o ß p r o z e s s e n 2 ) n u r bis m a x i m a l 4 0 A tief ins G i t t e r ein ( G e s a m t b r e m s w e g der A r - I o n e n n a c h Nielsen [2]). Die v o n einem 5 - k e V - A r - I o n auf ein C u - A t o m ü b e r t r a g e n e E n e r g i e b e t r ä g t m a x i m a l 4 , 7 5 k e V , so d a ß V e r l a g e r u n g s k a s k a d e n e r z e u g t werden, deren A u s dehnung i m Mittel, w e n n w i e d e r u m n u r s t a t i s t i s c h e Stoßprozesse in B e t r a c h t gezogen werden, ebenfalls wesentlich kleiner als die b e o b a c h t e t e Tiefenlage d e r !) Dissertation Universität Stuttgart (TH) 1967 (Teil I I ) . 2 ) Unter statistischen Stößen werden Stoßfolgen verstanden, bei denen die Stoßrichtungen in keiner festen Korrelation zu den Gitterrichtungen stehen.

624

H.

DIEPERS

Defektagglomerate ist. Es ist dehalb wenig wahrscheinlich, daß die Bildung der Agglomerate eine unmittelbare Folge statistischer Stoßprozesse ist. Vielmehr ist zu vermuten, daß die Punktdefekte über größere Reichweiten in einem Prozeß mit geringem Energieverlust transportiert werden. Als solcher kommt sowohl die Masse transportierende fokussierte Stoßfolge (dynamisches Crowdion) als auch die Kanalleitung der Ar-Ionen (Kanaleffekt) in Frage. In beiden Fällen erfolgt eine bevorzugte Ausbreitung in < 110)-Richtungen. Zur Aufklärung des Mechanismus der Agglomeratbildung sind daher in erster Linie die Fragen zu klären, ob die Agglomerate aus Eigenzwischengitteratomen des Cu oder aus Ar-Atomen bestehen 3 ), auf welche Weise die Defekte ins Gitter eindringen und schließlich, wie die Agglomeration der Defekte erfolgt. Über die ersten beiden Fragen soll eine Untersuchung der Orientierungs- und Energieabhängigkeit der Defektbildung Aufschluß geben. Zur Klärung der zuletzt genannten Frage wurden Untersuchungen über die Abhängigkeit der Agglomeratdichte und -Verteilung von der Dosis und der Bestrahlungstemperatur vorgenommen. Zur Durchführung dieser Experimente wurden folgende zusätzliche Experimentierverfahren und -einrichtungen benötigt. 2. Zusätzliche Experimentierverfahren und -einrichtungen Probe und Auffängerblenden (Fig. 1, Teil I) wurden auf T < 130 °K gekühlt. Die kalten Blenden verhinderten die Kontamination der Probe. Beim Transport und Einsetzen der bestrahlten, kalten Probe in die vorgekühlte Objektpatrone des Siemens Elmiskop I konnte durch eine Schutzschicht aus C0 2 eine Kondensation der Luftfeuchtigkeit auf der Probe und ein Temperaturanstieg über 194,7 °K, die Sublimationstemperatur des C0 2 , vermieden werden. Die C0 2 Schicht wurde im EM wieder abgedampft. Zur Messung der Probentemperatur im EM wurde zusammen mit B . Hertel [3] ein Verfahren erprobt, das auf Untersuchungen von Honjo et al. [4] über die kubische Modifikation des Eises aufbaut. Bei festgehaltener Temperatur der Objektpatrone wurde solange Wasserdampf in die Mikroskopröhre eingelassen, bis sich innerhalb des bestrahlten Bereiches eine Eisschicht auf dem Präparat bildete. Das Einsetzen der Eisbildung war an plötzlich auftretenden Doppelbeugungsreflexen zu erkennen. Gleichzeitig wurde der Dampfdruck in Probennähe gemessen und aus Dampfdrucktabellen die dem Dampfdruck beim Einsetzen der Eisbildung entsprechende Probentemperatur ermittelt. Sie lag bei allen Experimenten mit Probenkühlung im EM zwischen 206 und 209 °K. • 3. Das Eindringen der atomaren Einzeldefekte ins Gitter Sofern der Abstand der Agglomerate von der Folienoberfläche (Fig. 5, Teil I) unmittelbar korreliert ist mit der Reichweite der gerichteten Transportprozesse, sollte man eine Abhängigkeit dieses Abstandes von der Orientierung des Gitters zur Einstrahlrichtung der Ionen erwarten. Bei statistischen Stoßprozessen dagegen ist eine solche nicht zu erwarten. 3.1

Orientierungsabhängigkeit

(001)- und (HO)-Folien wurden bei einer Temperatur T < 130 °K bestrahlt und noch in der Bestrahlungsapparatur auf Raumtemperatur angelassen. So3 ) Letzteres muß berücksichtigt werden, da die Ar-Atome sich möglicherweise zu Agglomeraten zusammenlagern, die ebenfalls von einem Kompressionsfeld umgeben sind.

Elektronenmikroskopische Untersuchung von Fehlstellenagglomeraten (II)

625

Fig. 1. Tiefenverteilung der Defektagglomerate in einer (001)- und (llO)-Folie nach Bestrahlung mit 5 keV bei T < 130 ° K und Anlassen auf Raumtemperatur. Dosis: 1,6 x 10 18 Ar-Ionen/cm 2 . Die Kurven ergeben sich aus den Meßwerten von je 150 Agglomeraten

dann wurden in der in Teil I beschriebenen Weise Stereobilder aufgenommen und stereoskopisch ausgewertet. (OOl)-Folien enthalten vier