Nuclear magnetic resonance investigation of H, H2 and dopants in hydrogenated amorphous silicon and related materials

Nuclear magnetic resonance investigation of H, H2 and dopants in hydrogenated amorphous silicon and related materials

Physica B 170 (1991) 3()5-319 North-Holland Nuclear magnetic resonance investigation of H, H 2 and dopants in hydrogenated amorphous silicon and rela...

1MB Sizes 0 Downloads 13 Views

Physica B 170 (1991) 3()5-319 North-Holland

Nuclear magnetic resonance investigation of H, H 2 and dopants in hydrogenated amorphous silicon and related materials J.B. Boyce and S.E. R e a d y Xerox Palo Alto Research Center, Palo Alto, CA 94304, USA

Nuclear magnetic resonance has been successfully applied to the study of the microstructure of hydrogenated amorphous silicon and related materials. It has been used to determine the local bonding and structural environment of the host atoms, the hydrogen, and the dopants. First, we review some of these NMR experimental results on the hydrogen microstructure in hydrogendated amorphous semiconductors and compare the results on plasma deposited hydrogenated amorphous silicon (a-Si:H), remote hydrogen plasma deposited a-Si:H, thermally annealed a-Si:H, doped a-Si:H, microcrystalline Si and amorphous (Si, Ge):H alloys. A common feature is that these materials exhibit a heterogeneous distribution of hydrogen bonded to the semiconductor lattice in dilute and clustered phases. In addition, the lattice contains voids of varying number and size that contain non-bonded molecular hydrogen whose quantity is altered by deposition conditions and thermal treatment. Second, we review some aspects of the local bonding structure of dopants in a-Si:H. A significant fraction of the dopants are found to be in dopant-hydrogen clusters similar to those proposed to explain hydrogen passivation in crystalline silicon. Implications of the determined local structure on the doping efficiency are discussed.

I. I n t r o d u c t i o n

Nuclear magnetic resonance (NMR) studies have provided nmch useful information on the local atomic bonding and microstructure in hydrogenated amorphous silicon and related materials [1-25]. Here we review some of the resuits of NMR measurements that have provided information on two different structural aspects of these materials: (1) the H distribution and bonding, and (2) the local bonding structure of dopants in a-Si:H. The role of hydrogen in amorphous semiconductors has been much discussed [26, 27]. This is due, in large part, to its important role in converting an electrically poor amorphous semiconductor with a large density of defect states in the gap to a low-defect material that can be doped and made into devices. Since NMR is sensitive to H, it has played a significant role in determining the hydrogen microstructure in these materials. It has been used to study plasma-deposited a-

Si:H, thermally annealed a-Si:H, remote hydrogen plasma deposited a-Si:H, doped a-Si:H, microcrystalline Si (ixc-Si:H) and amorphous (Si, Ge):H alloys. In plasma deposited a-Si:H, most of the hydrogen is bonded to the Si in a dilute phase, that gives rise to a narrow NMR line, and in a clustered phase, that gives rise to a broad NMR line [1, 2]. But, in addition, a small fraction ( - 1 % ) of the H is in the form of H~ molecules. These molecules reside in small internal voids and can be observed directly in the NMR spectrum at low temperatures [8] or indirectly in the temperature variation of the H nuclear spin-lattice relaxation time, T 1, which exhibits a minimum near 40 K [1, 5[. This TI minimum can be accounted for by the presence of a small amount of H~ in voids in the material [25]. The fraction of the H in the form of molecular H 2 is larger in the following materials over that in a-Si:H: annealed a-Si:H, a-Si prepared by remote hydrogen plasma deposition, Ixc-Si:H, and Ge-rich a-Si t ,Ge~:H alloys [3]. In

0921-4526/91/$03.50 © 1991- Elsevier Science Publishers B.V. (North-Holland)

.I.B. Bovc~', S. l:. Ready

NMR inre~li
all these materials, however, one observes a broad and narrow H N M R line corresponding to a clustered and a dilute phase. In addition, they all exhibit a T~ minimum. This implies that, despite substantial differences in preparation, composition and properties, they have similar H bonding and microstructure. Differences, however, do exist, and these can be accounted for by a differing microvoid structure and a corresponding difference in the amount of molecular H~ contained in the films [3]. The details of the local bonding structure of dopants in crystalline and a m o r p h o u s silicon play a key role in determining the electrical properties of these semiconductors [26], The specific parameters that are pertinent are the n u m b e r of neighbors (three-fold or four-fold coordinated). the type of neighbors (St, dopant, hydrogen), the local order (random or clustered), and the nearneighbor distances. The local environment ol dopants in crystalline silicon has been determined to be four-fold coordinated substitutional sites with a local distortion of the surrounding Si tetrahcdron that depends, in part, on thc differenec in the atomic radius of the Si host and impurity. When H is added to the crystalline St. passivation of the dopants occurs [28-321. Several theoretical d o p a n t - h y d r o g e n bonding structures have been proposed to account for the passiwttion [30, 33-35]. In contrast, the local environment of dopants in hydrogenated a m o f phous silicon (a-St:H) is not known [26]. The doping efficiency is quite low. leading to thc suggestion that the majority of the dopants are three-fold coordinated rather than four-foM. This is a likely structure, but the location of H is unspecified. N M R double resonance [18, 191 shows that H plays a significant role and may account, in part, for the low doping efficiency in a-Si:H. Conventional N M R studies also provide complimentary information on the local structure of dopants in a-Si:H [19, 21-24]. This paper is organized as follows. In section 2 some pertinent experimental details are presented. Section 3 describes the results on the hydrogen microstructure of a-Si:H and related materials. Section 4 gives a discussion of the local bonding structure of dopants in a-Si:H.

2. Experimental technique 2.1. Sample,s

For the most part, the N M R results described here were obtained on samples that have good electrical properties, the materials of most interest for electronic applications. These samples were prepared by RF glow discharge ( G D ) deposition, also known as plasma-enhanced chemical wipor deposition ( P E C V D ) . The source gases are appropriate mixtures of silane, germane, hydrogen, diboranc and phosphine. Low power density (0.02-0.3 W/cm e) and heated substrates (230 250°( ` ) arc used in order to produce high density, good quality tilms. One set of a-Si:H was prepared using a remote hydrogen plasma under conditions that also produce electronic grade material. For N M R samples, the substratc is usually aluminium foil which is etched away in dilute HCI. 2.2. N M R Sl)e(lro.s~'OlLV

The N M R signal provides a passive, local probe of the environment of the nuclear spins of interest. It is passive since the interactions of the nuclear moments with the lattice and other spins is weak. It is local since these interactions fall off rapidly with distance, typically as 1/r ~'. it is structural since these interactions depend on the atomic configuration and relative distances of the interacting species. The total spin Hamiltonian is

I>1 tt-

H z + tt<1 + H c + H~.) + H wi,, I~,~i<~.,

( 1)

where H z is the Z e e m a n interaction with the externally applied magnetic field, 1t z yhl" H. with "y the nuclear g}ronmgnetic ratio, h l'lanck's constant, / the nucler spin operator and H the applied magentic field. The other terms make much smaller contributions and so arc perturbations on H z. These terms tire, from left to right, the dipolar, the chemical shift, the quadrupolar, and ihc spin-lattice interactions. This latter term is included to represent all interactions with the lattice that are not contained in

J.B. Boyce, S.E. Read)' / NMR investigation of hydrogenated amorphous silicon

the other terms, such as, paramagnetic impurities and phonons. There is also a term, similar to H z, that describes the interaction of the spins with the radio frequency field applied to o b s e r v e the N M R signal. We first consider the spectral information and then, briefly, s p i n lattice relaxation. For the important case of hydrogen, which has l = 1 2, Ho = 0 and H c small, the relevant structural information is contained in H a which provides a width to the N M R spectral line. The second m o m e n t of this line, due to the like-spin H - H dipolar interactions, is [36] M~ = ~9' 4~-,,2 It 2 1/r,j0 ,

307

(a)

I

I' I

-40

0

40

(V-Vo) (kHz)

(2)

where r~j is the distance between H nuclei i and j. In a-Si:H, the 24Si nuclear m o m e n t is smaller than that of H and its natural abundance is low (4.7%), so that the H dipolar width should be well represented by eq. (2). For a concentrated spin system, the line shape is Gaussian, for which M 2 is the Gaussian width squared. So the full width at half m a x i m u m ( F W H M ) is ~ = ( 8 ( l n 2 ) M , ) 1~2. For hydrogen, this gives cr (kHz) = 190(21/rij6) ~/2, with r in ,~. The structural information is contained in the sum over 1/r ~. For a dilute spin system, the line shape is more closely approximated by a Lorentzian line whose width will also vary as 1/r 3. Both line shapes are observed in a-Si:H, a Gaussian broad line and a Lorentzian narrow line, as seen in fig. l(a). In addition to the central lines broadened by H - H dipolar interactions, the N M R spectrum also contains a Pake doublet due to molecular H 2 trapped in microvoids in the a m o r p h o u s matrix. This is shown in fig. l(b) and discussed in the next section. For P and B dopants, the other terms in eq. (1), namely, H c for P(l = ½) and H o for B ( I = ) can contribute significantly to the spectral line shape and width and provide additional probes of the structural information. In addition to the N M R spectrum, measurements of the H nuclear spin-lattice relaxation time, T~, have provided additional insight into the H microstructure. It is now well established that the T j ( H ) in these materials is dominated,

(b)

J

I

't ,' ",,

-200

,j¢

"~

0

J' I

200

(V-Vo)(kHz) Fig. 1, (a) H NMR spectrum of electromc grade a-Si:H at 3(//) K, showing the broad and narrow lines. (b) A blow up of the NMR spectrum at 1.43 K, showing the Pake doublet due to H,. Note the different horizontal scales.

at least in the 4-200 K temperature range, by relaxation to molecular hydrogen trapped in microvoids in the a m o r p h o u s matrix [1]. In standard, electronic grade a-Si:H, about 1% of the H is in the form of H 2. For this case, the T~ of the bonded H, the majority of the H in the material, is related to the T~ of the nonbonded H 2 by T I ( H ) = A T , ( H : ) + Tt~SD~ ,

(3)

where 1/r,(H~)-Cm/(r~-m + o~), with F m the temperature-dependent molecular angular m o m e n t u m relaxation rate and w0 the L a r m o r frequency. A is proportional to the ratio of the H concentration to the H, concentration, A--[n(H)/n(H2) ]. TI(SD ) is the spin-diffusion

J.B, Boyce, S.E. Ready ' NMR invevti~ation o! hydrogenated amorptum,s wlic(m

308

3. Hydrogen microstructure of a-Si:H and related materials 10,000

3.1. a-Si: t l

I000

v

I00

I0

\

1

0.I I0

I00

I000

TIK) Fig. 2. 7'~ versus T f o r electronic gradc a-Si:H, showing the T, m i n i m u m near 40 K. The solid line is a fit to eq. (3) and ~iclds n ( H e) ~ 0 . 0 6 air; ".

bottleneck term. If it is small, then 1/ T~(H) ~ n ( H , ) , so that the amount of H e trapped in voids in the material can be estimated by measuring T I ( H ). Equation (3) predicts the prominent feature of the measured T~ shown in fig. 2, namely, a minimum for I~,, ~ ~o,,. It also describes the shape and magnitude well. It turns out, however, that not all the H, will provide efficient relaxation of the bonded H via the mechanism of eq. (3). Only H 2 in good contact with the a-Si host phonon spectrum contributes to eq. (3), i.e. those H e molecules adsorbed on the surfaces of voids [8, 37]. If the w)ids are large and contain substantial amounts of H e . then a substantial fraction of the H~ can be in the interior of the void. This H, has a long T~(H~) which is determined by the electric q u a d r u p o l e quadrupole ( E Q Q ) interactions of the H~ molecules and is nearly t e m p e r a t u r e indepenent. These H 2 do not contribute to the relaxation of the bonded H according to eq. (3) but are observable in the N M R spectrum, as discussed in the next section.

7.1, 1. Bonded hydrogen Thc N M R spectrum of H in a-Si:H has becn extensively studied [1,21 . It is lkmnd that dcspitc a wide range of deposition conditions and preparation techniques, a two c o m p o n e n t central line is always observed, a broad line and a narrow line as seen in lig. 1. The narrow line is Lorentzian in shape with F W H M ~ 3 k H z , while the broad line is Gaussian with F W H M ~ 2 5 k H z . Also the total hydrogcn content for the samples prepared on heated substrates is found to be ~-10atC~-, with about 4at(";- in the narrow line and approximately 6 a t G in the broad line. This is evident from inspection of the line shape information for several representative samples in table l. The narrow line corresponds to isolated S i - t t units randomly distributed throughout the amorphous Si network. Estimates [4] from the 3 kHz line width using eq. (2) place the mean spacing between the H atoms in this dilute phase at about 8 A. Also application [51 of a statistical theory to this narrow Lorentzian line arrives at a similar conclusion, namely, that it is due to a random or slightly clustered distribution of nlonohydride units over about 75C~ of the a-Si lattice. The origin of the broad line is less certain, but it does arise from regions with more clustered H. Various types of structures have been considered. H y d r o g e n a t e d monovacancies and divacancies are ruled out since their line width is too large ( ~ 1 0 0 k H z ) . SiH~, SiH~ and polysilanc chains give line widths that are too small but could yield a 25 kHz line if clustered. Hydrogenated internal surfaces are a preferred candidate since internal voids and platelets are known to exist. However, no detailed structure of the hydrogenated internal surfaces has been proposed. We now describe three recent results which add some additional information to the nature of the broad line.

J.B. Boyce, S.E. Read), / NMR investigation of hydrogenated amorphous silicon

3(19

Table 1 Room temperature line width parameters and T~ of a-Si:H samples prepared by plasma-enhanced CVD. The doping and substrate temperatures are listed. Note that all samples prepared on elevated-T substrates have nvot,~lOat%, ny~,,, ~ 4 a t % , Av N 3 k H z , and Avm~,~,~~~ 25 kHz. For samples prepared on RT substrates, nxo,, increases to =30 at% but the line widths and n N....... remain about the same. .......

~

Sample

nx~,,,,

[Gas, T,]

(at%)

n (at%)

Intrinsic

9.5

Broad line

Narrow line

T~

FWHM (kHz)

n (at%)

FWHM (kHz)

(s)

5.3

22

4.2

2.9

2.5

9.3

5.5

22

3.8

3.0

2.4

11.4

7.1

24

4.4

3.2

2.7

13.6

10.2

24

3.4

2.1

2.5

10.0

7.5

23

2.5

3.0

1.8

12.0

7.6

24

4.4

2.8

2.8

10.0

8.3

27

1.7

3,5

1.4

11.1

7.4

24

3.7

3.4

2.4

13.(1

9.0

24

4.0

3.0

2.4

29.0

25.8

23

3.2

2.5

-

27.1/

23.2

24

3.8

3.4

-

24.0

17.5

21

6.5

2.8

-

230°C

Intrinsic 230°C

Intrinsic 188°C 111 -" P 10 z B 230°C 10 -~P 230°C 10 4 p 230°C 10 2 B 230°C lI) 4 B 188°C 10 a B 188°C Intrinsic 25°C

Intrinsic 25°C

Intrinsic 25°C

First, Sill 2 molecules have been directly observed in the N M R spectrum [9]. The dipolar interaction between the two H atoms on the molecule gives rise to a Pake doublet which, for a H - H spacing in SiH~ of 2.4 .~, is vp ~ 14 kHz. This splitting is rather small so that, if other H atoms are near the SiH2, i.e. within about 2.4 A, then the doublet will be indistinguishable from a Gaussian line. Nonetheless, a broadened doublet is observed. The determined extra spectral broadening due to neighboring H atoms is smaller than the H - H interaction on the Sill 2 molecule by about 20%, enough to increase the FWHM to about 21 kHz, close to the usual value of 25 kHz, but not large enough to eliminate the difference in line shape between a broadened

doublet and a Gaussian. This doublet is only observed, however, in poor quality films prepared on room temperature substrates (last three samples in table 1) and not for electronic grade material prepared on heated substrates. These results indicate that SiH~ molecules contribute substantially to the broad line in low grade aSi:H but can be ruled out as a major contributor to the broad line in good material. This conclusion is in agreement with infrared and Raman results [38] where the contribution of Sill, to the spectra decreases substantially with increasing substrate temperature. Second, multiple-quantum N M R ( M Q - N M R ) experiments [2, 23, 24] have been performed on a-Si:H prepared at different substrate tempera-

311)

,I.B. Hovce. S.I:. Ready : NMR inve,stiealion O/hvdro;,~'mm'd amorphous ~ilic(m

tures, l'. This technique provides a measure of the magnetic quantum n u m b e r of the spins coupled together by the dipolar interaction, thereby providing a measure of the n u m b e r of coupled spins. For the experimentally accessible time scale, the dominant contribution to the MQN M R comes from the broad line. The results have been interpreted to yield clusters of five to seven H atoms as the predominant bonding configuration for the broad H line. Differences w,'erc observed between samples prepared at different substratc temperatures. For the electronic grade a-Si:H prepared at 270-324°C, only these small clusters were found to exist, while for samples prepared at lower temperatures larger clusters exist along with the 5 - 7 atom clusters. As the H content increases, the n u m b e r of these clusters increases, but their size does not. When the H content increases even further, as for samples prepared on low t e m p e r a t u r e substrates, the cluster size also increases. These measurements provide no information on the specific structural model for these 5 - 7 atom clusters. Third, recent theoretical calculations [39] on hydrogen clusters in crystalline silicon have been used to construct a specific model for the clustered H responsible for the broad N M R line in a-Si:H. It is based on the H~ complex shown in fig. 3(a). A Si-Si bond is broken and one H bonds at the bond-center site to a Si atom which relaxes away from the original bond by 0.24 A. The second H bonds in an antibonding site to the other Si atom which relaxes toward the antibonding site by 0.79 A, putting it in the plane of the neighboring three Si atoms. This complex is neutral, ESR-inactive, lower in energy by 1.2 eV than two H atoms inserted into a Si-Si bond, and only 0.4 eV higher in energy than molecular H~ in a T a site. Since crystalline and a m o r p h o u s Si are similar locally, such a complex may also be appropriate for a-Si:H. However, the H - H distance at 3.38 J~ is too long to account for the broad N M R line, giving rr =-5 kHz. If two or morc H~,'s complex as first neighbors to one another, as suggested by Jackson and Zhang [40] shown in fig. 3(b), the shortest H - H distance becomes 2.26 ,~. The line width for a pair of H~ near neighbors is ~r= 17.4kHz, still somewhat

(a)

Si

i

,i

!H2 i 0.24

°,,A: 2!_C)

(

Fig. 3. (a) Structure of lhc H:~ diatomic hydrogen complex. (b) A higher level complex consislmg of four tt~, ,, thal c~ln account for the broad N M R line.

small. The width for tk~ur neighboring H'# is ~r = 2 1 . 4 k H z , in good agreement with the line width of the broad line. Thus higher level complexes of H f can account for the broad N M R line and provide a specific model for the clustered phase of H in a-Si:H. A cluster of eight or more H atoms is required, more than the 5 - 7 atom complex of the interpretation of M Q - N M R experiments.

3.1.2. Molecular hydrogen When a-Si:H is cooled to low temperatures, the N M R spectrum evolves into a three-component line: the central line with two width components, as observed at room temperature, plus a Pake doublet [3, 8, 9] with a splitting of 175 +_ 10 kHz, shown in fig. I(b). The observed doublet

J.B. Boyce, S.E. Read,; / NMR investigation of hydrogenated amorphous silicon

is due to orientational ordering of molecular H 2 or a slowing down of the molecular tumbling rate to less than the 175 kHz splitting. (Molecular H~ can exist in two states: ortho, with orbital angular m o m e n t u m J = 1 and nuclear spin I = 1, and para, with J = 0 and I = 0. The room temperature occupation of the two states is 3 ortho and ~ para. The o r t h o - p a r a conversion is slow, ~ 2 % / h at 4 K; only the ortho-H~ contributes to the NMR.) As T increases, the Pake doublet collapses due to the averaging by the rapid molecular tumbling and the signal due to H : is transferred to the broad line. As T increases further, the solid H 2 melts and the signal becomes motionally narrowed and contributes to the narrow line. This has a measurable effect [8] on the signal fraction in the two central line components in samples containing a large amount of H 2, i.e. n(H2)/n(H)>~O.05, and a small effect for samples with small amounts of H2, i.e. n(H2)/ n ( H ) ~< 0.05.

3.1.3. Spin-lattice relaxation The N M R spin-lattice relaxation time T~ exhibits a characteristic minimum in the vicinity of 4 0 - 5 0 K (see fig. 2) due to the presence of a small number of H~ molecular relaxation centers on the surfaces of voids in the a-Si host. The H atoms in the molecule relax quickly due to interactions between the rotational angular momentum of the ortho-H, and the phonons of the host. The H~ molecules thereby serve as relaxation centers for the H atoms bonded to the Si, relaxing them by spin diffusion. This relaxation process is described by eq. (3) above and yields, for typical electronic grade a-Si:H, n(H2)/ n ( H ) ~ 0 . 0 1 or n ( H 2 ) ~ 0 . 1 at%.

3.2. Annealed a-Si:H Annealing studies have provided much information on the H content, bonding, microstructure and diffusion in a-Si:H. For example, evolution studies [27] show a low-T (~400°C) and high-T evolution peak (~600°C) for room temperature substrate films but only a high-T evolution peak for substrate temperatures above 200°C. From such measurements it has been

31l

concluded that low-T~ material has an open, void-rich structure that allows for rapid diffusion of molecular hydrogen. Annealing also increases the H~ content of the film. This H~ collects in voids before totally evolving from the film. For electronic grade a-Si:H, no substantial changes occur in the H NMR until the anneal temperature, T A , is above 450°C, as shown in fig. 4. This is to be contrasted with the changes observed in the hydrogen [3, 6] and deuterium [11] NMR spectra at about T A ~ 200°C for samples deposited on room temperature substrates, i.e. electronically poor material. The H , content increases with T A up to about 500-550°C as the total H content drops. Above this TA, the H , in the voids also begins to evolve. The RT line width of the narrow line narrows for T A ~ 5 0 0 ° C , evidence of the mobility of the H~ in the voids. This implies that the void size grows with T A to accommodate the larger amount of H~ and to provide the space for the H~ to move and motionally narrow the line width. The signal fraction and F W H M versus measurement temperature for a sample annealed

o n(H) Tot ~ n(H) Broad line _ V n(H) Narrow line n(H2)

............. ...........

8

,--

"""

-

7

0

6 ,,&

Q

o "0 >,

-r-

4

-4

- 4..

'~ .....

c3- - - "

3

,

,

o',

2 1

0 200

~''~ I 300

I .--~'" 400

I 500

×"-x 600

Tanneat (C)

Fig. 4. The hydrogen content in the w m o u s components of the NMR line for annealed a-Si:H versus T~.

312

,I. I~, lJovcc, ,S. 1;. Ready I

,'\'MR mvc~li
about 2()kHz, reduced from 25 ktlz duc to the loss of bonded H in lhc chistcred phase (fig. 4).

I

i

L

0.6~.<,

OZ

,i

.

.7.7. Remote lryUro
:.

~

0+I4

-o !

o

i

10



000~

f

Z

,!

C5 0.2~

/, i

+

<'p

0L

+°t 2O

N



llolili

•o

o•~llooilmOoo

le



l

10 ~ 5

2 c'~'z2%~

1

"

Z-,': Z~ t .

F.

[

I

I

!

I

2

5

10

20

50

14?. .i .,'.. 100 200 500

T(K)

Fig. 5. l'ilc signal lracli(m el the three COillponclllS, el the line and FWHM of the broad and nalrov¢ lines ~ursw, nlcasuronlcril

7 for

a-Si:lt

allllCait.,d

~il 7' x

5(111'('.

at 5(1()°C arc shown in fig. 5. The narrow line width is ~1 kHz at 30(l K but ~ 2 kHz al 2 K, As the san]pie is cooled, the narro\v linc. which consists of about half b o n d e d H alid half H , , broadens as the translational motion of the I ~ freezes ()tit. Cooling further hclow ahotlt l()K causes the molecular tumbling motion to freeze out and the Pake doublet l() a p p e a r in the spectrunl, pulling its signal inicnsit} fronl the narrow line. This transition is rcsponsihlc for the b u m p m the F W H M versus 7 for the narrow line. The narrow line width of 1.5 kHz fit low 7' corresponds to the remaining h o n d e d H in the dihitc phase. Its ~idth has decreased from 3 k H z in as-deposited a-Si:H to about 1 . 5 k H z in the annealed sample duc to the reduction in the a m o u n t of b o n d e d H in the dilute phase (lig. 4). The t~road line width stays constant with 7 at

The tt N M R spectra 12(I I of electro•lie grade a-Si:ti p r e p a r e d by I'CIII(){O h y d r o g e n ph_lSlllll ( R H P ) deposition lit 7" - 4()0°C is similar to thai for a-Si:H made hv the conventional glm~ discharge ( G I ) ) technique (see section 3.1 ). The RHP sample has a total It content of ahout 1<1ill,(<; , ~ith 4 ill(;} in the n~,lrl-ow line (FWHM-=3kltz) illld 6arC<; in the broad line ( F W t t M ~ 2 f ~ k H z ) . Also a 7 t mininlunl is ohserved at about -l()K. ' l h c s c paralllcters are alnlosl identical with those for good-qualit} (}D nlatcria[. A difference, however, is evident in lhe N M R specml when the R H P sample is cooled to helow 30 K where a Pake doublel duc to the non-hondcd H , is ohserved. This doublet conlpliScs l()rr: Of the tolal signal lit 4.2 K or ().5 at+:7 I 1,. This is a factor of ~5 larger than lhc altlOtlnt t)f H+ in GL) material. Such a large fraction of 1-I, in tile sample is usually an indication thai thc nlatel-ia[ is highly defective, such fls observed for annealed films or a-Go. However, the R f I P films have low ctefcct densii), leading Io the proposal that ttlc material is SilllUllancously deposited, annealed and hydrogenated during deposition. ( ' o m p e t i n g processes of It difl:usion and cxolution fit the high deposition t e m p e r a t u r e (40IF(' versus 230%' for G D ) , ahmg with continuous 1--t rcincorporation, cause the film to have the usual aillounl lind distribution of honded tt hut a lar,,er ilnlotln{ ()f non-bonded H,.

3.4. txc-5"i:lf In the growth e l silicon thin films h) i>hlsma enhanced chonlica] vapor deposition ( P L : ( ' V D ) , I*+-Si:H rather than a-Si'H results when the inaIcrial is prepared with tI_JSiH~ gas phase ratios greater ttmn about 2(). For the +c-Si:H phase, the H N M R spectra exhihit the ubiquitous broad and narrow lines and 7~ has a m i n i m u i i l near 40 K 1141 A major difference betwocrl the spectra for a-Si:H and +c-Si:H is in the nature of tile

J.B. Boyce, S.E. Ready 50

i



20



!





!

O ~ Q

]

0000

NMR investigation of hydrogenated amorphous silicon

]











\

10 '~'5 ...tJ<

-'t- 2 u.

1

98/1 H2/SiH 4

0.5

' \~,..~

(a) 0.2 t 0.1

A _ _ I

5

1

1.0

10

t

20 T (K)

,

50

r ~ - - I

;

100 200

I

500

I

98/1 H2/SiH 4 Q

eee,,~- - °~ B

0.8

B•

¢-





._o 0.6 LL C

0.4

.(X N 2.

O~p

0.2

(b)

-~

0.0

_~__ 2± 2

5

t 10

I ~_ 20 50 T(K)

I

J

100 200

J

500

Fig, 6. (a) Temperature dependence of the full width at half maximum (FWHM) of the broad line (O) and the narrow line (A) of the hydrogen NMR signal for a p,c-Si:H sample prepared with a H_,/SiH 4 gas phase ratio of 98:1. (b) Temperature variation of the signal fraction for the broad (B), narrow (N), and Pake doublet (P) (due to H2) line components in the 98:1 p,c-Si:H sample.

narrow line where two significant differences can be noted, as displayed in the data of fig. 6. First, for t~c-Si:H, the R T width is about 0 . 4 k H z versus about 3 kHz for a-Si:H and about 1.5 kHz for a-Si:H annealed to 500°C. This narrow line

313

broadens as the sample is cooled, reaching 5 kHz, comparable to the narrow line width of a-Si:H, below 20 K. The width of the broad line, on the other hand, does not change significantly with T and has the usual 2 5 k H z width. This narrow line has also been observed by others [12, 13] and has been attributed to mobile, bonded H,such as Sill 2 chains. As seen in fig. 6, it is due, in large part, to molecular H~. The motion of the H , in ~c-Si:H has a more substantial effect on the line width than does the motion in a-Si:H annealed to 500°C. A reason may be that the low-density, intergranular regions in p,c-Si provide more open, interconnected space for H , motion. The second difference is in the amount of H that makes up the immobile portion of the narrow line as T--~ 0. This is bonded hydrogen in the dilute phase and consists of only ~ 1 0 % of the H in i~c-Si:H versus ~ 4 0 % of the H in a-Si:H and -~30% for a-Si:H annealed to 500°C. In each case, however, the broad line contains about half of the total H content. A possible explanation for both of these observations is that the microcrystalline regions contain little or no H and the a m o r p h o u s matrix regions are highly defective, containing large interconnected voids in the intergrain regions. The hydrogen bonded to the Si on the surfaces of these voids accounts for the H in the broad line. These open intergranular regions contain the molecular H~, a small fraction of which is adsorbed on the surfaces of these regions and serves as relaxation centers for the remaining hydrogen (see section 2.2) and yields the T 1 minimum near 40 K, The H~ in the open regions is able to diffuse readily at RT, giving rise to the narrow line, but its translational motion stops as the H~ solidifies near 20 K. The small amount of dilute-phase bonded H can be contained at the surfaces of the crystalline regions or in regions of the amorphous matrix that have low H density. The fact that it has a narrow-line line width close to the standard value as T--~ 0, argues against it being distributed in the microcrystalline regions at, say, defect sites. Rather it appears that the microcrystalline regions are relatively free of H and most of the H, bonded and non-bonded, is contained in the a m o r p h o u s matrix regions [41].

314

.I.B. B e y ' c o , 5./;'.

I?ead~

.VMR I#lrd.wl
,;',5. a-(St, Ge):H allots F o r a-Si~ G e , : H alloys, the H N M R s p e c t r a exhibit the usual b r o a d a n d n a r r o w lines and T, m i n i m u m n c a r 4 0 K . T h e H c o n t c n t is f o u n d to be a b o u t thc S a l l l e ( ~ - 1 5 tiit¢7: ) for tile Si and G0 e n d s and to m c r e a s c to a b o u t 27 atr{ at 5t):5(i, as shown in fig. 7. T h e b o w e d b e h a v i o r differs from the m o n o t o n i c d c c r e a s o in t t ( H ) w,ith mcreasing ( i t c o n t e n t , x, o b s e r v e d tB others [15, 17]. This mcrease in H contcnt that we observe occurs entirely' in the chlstered phasc since the dilute phasc content remains constant at = S a t G until the Gc end, whore it d o u b l c s t o ~ 1 0 a t C f . In addition to doubling in contcnt at the Go-rich end of the diagram, the width of the n a r r o w line d c c r e a s e s from ~ 3 k H z for x < 0.6 to --~1 k H z for x > 0 . 9 . T h e b r o a d line, on the o t h e r hand. r e m a i n s e s s e n t i a l l y c o n s t a n t with x at the usual value of ~ 2 5 kHz. A significant fraction of the n a r r o w line in the G o - r i c h alloys is duo to l n o l c c u l a r H,. As tile a - G c : H is c o o l e d , the n a r r o w line b r o a d e n s quickly, and a Poke d o u b let a p p e a r s b e l o w 25 K and c o n t a i n s = 2 at
[3, 16l

30

I

I

a-Sil ,Oe×:H

I

I

n(H), /



25

-o

¥

20_

• I

\

- "

"" "

n ( H ) ...... o

15"--

(b

10

2 ,,~. Y•

2_

n(H),

/ ',

0 00

I 02

l 04

If-

1 06

I 08

10

x fn S o l i d

Fig. 7. Tt]c h y d r o g e n c{}ntcnt in t h e v a r i o u s c o m p o n e n t s l h c N M R line in a - ( S t , G c ) r H vcrsu<, ( i t c { m t c r l t ,

{}1

with the conclusions stated above for annealed cl-Si:tt and d r a w n fronl h y d r o g e n e v o l u t i o n studies [271, namely, that the a d d i t i o n of (}c c r e a t e s alh}vs possessing a void-rich slrtlctnrc which b e c o m e s m o r e pr{}nounccd with i n c r e a s i n g (}e c o n t e n t . T h e N M R shows that thcsc ~oids contain a signilicant an]ount {}1 H , and that the total PI~ c{}ntent is larger for a-Go than for a-St. ,7,0. ('omparLwm

ql dm Itydrogenaled s rwem.s

All of the materials systems discussed here arc similar in that the'~ all contain a hirge quantity of It (=l(}at~< :) that resides in throe distinct phases: a clustered phase exhibiting a broad N M R line ( = 2 5 k H z F W H M ) , a dilute phase exhibiting a nat-row N M R line ( = 3 k H z f : W H M ) , and n o n b o n d c d m o l e c u l a r h y d r o g e n that c o n t r i b u t e s to the narro\~ N M R line n e a r t~/I but is split off as a 175 k H z Pakc d o u b l e t at low t e m p e r a t u r e s ( , 1 0 K ) . In a d d i t i o n , all these materials exhibit a ininiinunl ill thc T, tersus T curvos at low tenlperatuies due to relaxation by tt~ m o l e c u l e s t i d s o r b c d on the surfaces of voids. l ) i f f e r c n c e s bctx~ccn G I ) a - S i : H and thcsc o t h e r m a t e r i a l s do occur. O n e i n l p o r t a n t diffcrcricc is that thc m o l c c u t a r H , c o n i c n l is 5 - 1 0 liines that of G I ) m a t e r i a l . The ( i t - r i c h alleys and the allncalod a-Si:ll ( T \ ~-=500°( '} have illorc {llld larger voids, c o n t a h l i n g larger q u a n t i t i e s of nonb o n d c d m o l e c u l a r H~. Simihlr s t a t e m e n t s appl> to bCC-Si:ll, but the o p e n regions are hug0r or m o r e i n t e r c o n n e c t e d , l e a d i n g to an even narr o w e r narrow line. Also the amount of tl bonded in thc dihitc p h a s e is less, c o n s i s t e n t with the suggestion that the crystalline regions c o n t a i n little or no hy'drogcn. T h e s c hlrgc q u a n t i t i e s of H , m a k e a s u b s t a n t i a l c o n t r i b u t i o n to the narrow N M R linc at r o o m t e m p c r a t n r c . In fact, the n a r r o w line consists actually of two n a r r o w lines. one c o r r e s p o n d i n g to the dilute p h a s e b o n d e d H ( F W H M ~ 3 kHz) and o n e c o r r e s p o n d i n g to f]~ ( F W H M < 3 kHz). F o r the a n n e a l e d a - S i : H w h e r e significant q u a n t i t i e s of H have e v o l v e d , the n a r r o w linc width for the b o n d e d H is red u c e d , as e x p e c t e d . In till these m a t e r i a l s systems, h o w e v e r , the b r o a d N M R lines tire similar t{} each o t h e r and to that of a - S i : H ( F W H M

J.B. Boyce, S.E. Ready / NMR investigation of hydrogenated amorphous silicon

25kHz). Again, for the annealed a-Si:H, this width is somewhat reduced.

4. Local bonding structure of dopants in a-Si:H The local bonding structure of dopants in amorphous silicon is important since it determines the doping mechanism and efficiency in this semiconductor. In one model [42] for doping of a-Si:H, the P, for example, can bond to the Si host in two ways: (1) as a neutral three-fold coordiated P, designated P~,~, giving an inactive site; or (2) as a four-fold coordindated, ionized donor, P4, plus an equal number of charged dangling bonds, Si3. This model predicts that the number of donors, N D, varies as the square root of the total number of dopant atoms in the material, N o. From a number of optical, ESR, and transport measurements, this model is found to hold, and, for phosphorous, N D -~ 10 ~N0~'2. The doping efficiency is seen to be low and given by r I = ND/N~= 10 3 Not,2. So for high impurity concentrations appropriate for structural studies, say 1 at%, the doping efficiency is -q ~ 10 -2. Only 1% of the impurity atoms introduced are then expected to be in four-fold sites, while the majority are in three-fold sites. Obtaining information on the local bonding structure of dopants requires a local structural probe and two have been used. These are XAFS (X-ray absorption fine structure spectroscopy) and NMR. XAFS has been applied to As and P-doped a-Si:H. The first study [43] supports the idea that the inactive dopants are three-fold coordinated, whereas the active dopants are four-fold. However, a very high fraction of fourfold As atoms were found, i.e. 20% for a 1 at% As sample. The second study [44] concludes that the P atoms are 3- and 5-fold coordinated. XAFS, however, is not sensitive to the light hydrogen atoms, which are an essential component in good electronic quality a-Si. Hydrogen most likely plays a significant role in determining the doping mechanism and efficiency, in addition to the Si coordination and possible clustering. In crystalline Si, it is known that hydrogen passivates acceptor J28, 29, 31-34] and donor states +

315

[30, 34, 35]. This is likely to be the case in a-Si:H also because of its high H content ( ~ 10 at% H) and since the local structure of the amorphous and crystalline phases are similar. Both standard NMR and a double NMR techniques have been used to probe the local bonding environment of boron and phosphor impurities in singly-doped and compensated a-Si:H and to comment on the issues of doping efficiency. 4.1. Standard N M R results

Single NMR experiments have been used to extract structural information on dopants. In the analysis of P NMR in n-type a-Si, the majority of P was found to reside in three-fold coordinated structures [23, 24]. The spin-spin decay time, T e, of the spin echo is due to P - P dipolar interactions for a random distribution of P. No substantial clustering occurs and the bulk of the P is in the dilute-H phase, not the clustered-H phase [19]. From chemical shift and T: decay results, 20% of the P in material with ~1 at% P were determined to be four-fold coordinated [23]. No evidence of such a signal is observed in another study [19]. Also, as discussed above, the doping efficiency for a 1 at% P-doped sample is about 0.01, i.e. only 1 in 100 (not 20 in 100) of the P should be four-fold coordinated. It appears, however, that this four-fold coordinated P is metastable [24]. It requires a dark anneal to be present and can be eliminated by light soaking. In p-type material, two quadrupole-broadened (from H o in eq. (1)) boron lines are observed. From this it can be concluded [21, 22] that the B atoms reside in two distinct three-fold sites: one consisting largely of axially symmetric BSi, structures and another consisting of Si~BH sites with the B - H distance somewhat relaxed. Another study [19] supports this conclusion of two threefold sites and, in addition, finds a much narrower line in compensated material. The width of the narrow boron line is quite comparable to the line width of tetrahedrally coordinated boron ( < 2 k H z ) measured in crystalline silicon J45]. This signal, which corresponds to about ~ of the B in a 2 a t % sample, is thereby attributed to four-fold sites. This is a very large fraction con-

316

Y.B. Bovce. S. t:. Reudv : NMR im'c~tigation ql hydrogenated amou;houv sil/con

s i d e r i n g the fact that the d o p i n g efficiency is quite low. H o w e v e r , the d o p i n g efficiency in c o m p e n s a t e d s a m p l e s can be two o r d e r s of magn i t u d e l a r g e r than in s i n g l y - d o p e d m a t e r i a l [46]. But, as discussed b e l o w , a b o u t half of the B have a P n e a r n e i g h b o r . So these f o u r - f o l d c o o r d i n a t e d B a t o m s , only o b s e r v e d in the c o m p e n s a t e d m a t e r i a l , most likely consist of B with t h r e e Si and o n e P, i.e. BPSi~ units,

c e i v e r coil t u n e d to the / - s p i n r e s o n a n c e while also a p p l y i n g a third pulse, P~, at a n o t h e r r a d i o f r e q u e n c y to a s e c o n d N M R coil t u n e d to the ./-spin r e s o n a n c e . T h e signal is m o n i t o r e d both w i t h o u t P,, giving S . , a n d with P~, giving S. S is r e d u c e d from S. by the flipping of the Y-spin, and the size of the r e d u c t i o n d e p e n d s on r and the I - , I d i p o l a r c o u p l i n g a c c o r d i n g to

5 S ( 2 r ) = (S,, - S)/,5,~ 4.2.

Double

NMR

results

,,ll - exp(-2

Spin e c h o d o u b l e r e s o n a n c e ( S E D O R ) has b e e n used to p r o b e the local e n v i r o n m e n t a r o u n d the d o p a n t s in s i n g l y - d o p e d and c o m p e n s a t e d a-Si [18, 19]. T h e s e d o u b l e r e s o n a n c e studies differ from single r e s o n a n c e studies of the d o p a n t a t o m s t h e m s e l v e s in that the d o u b l e r e s o n a n c e yields direct i n f o r m a t i o n on the relative l o c a t i o n o f a p a r t i c u l a r p a i r of a t o m s selecte d s p e c t r o s c o p i c a l l y by the two N M R f r e q u e n cies used. T h e single r e s o n a n c e e x p e r i m e n t s must c o n t e n d with the c o n t r i b u t i o n s of all the a t o m s p r e s e n t in the m a t e r i a l . T h e S E D O R m e a s u r e m e n t s are p e r f o r m e d using a s t a n d a r d spine c h o pulse s e q u e n c e ( P ~ - r - P e ) a p p l i e d to a re-

2

2

,Mo~,,r-~)],

2

I,

with ~ w ~ = [ y / y j / ~ / S ] ( N ~ / r ). c~ is a p a r a m e t e r that is ~<1 and d e p e n d s on the c h a r a c t e r i s t i c s of the pulse P~. This t e c h n i q u c yields a signal that d e p e n d s very s t r o n g l y on the d i s t a n c e b e t w e e n the spins I and ,I. F o r the d a t a on thc d o p e d a - S i : H , it is f o u n d that m o r e than o n e i n e q u i v a l ent site with d i f f e r e n t e n v i r o n m e n t s , such as 1 spins c o u p l e d to .1-spin first n e i g h b o r s a n d I spins with no ,/-spin first n e i g h b o r s , occur. A twoe n v i r o n m e n t a p p r o x i m a t i o n was used, i.e. two e x p r e s s i o n s o f the t y p e o f eq. (4). T h e results of such fits to the S E D O R d a t a on c o m p e n s a t e d , p - t y p e , and n - t y p e a - S i : H are given in table 2.

Tablc 2 Phosphor and boron NMR SEDOR data lot compensated a-Si:(H, f3, P) and samples singly-doped with P and B. The two-peaked experimental results are shown, with Frac being the fraction of the atom pairs with the given distance. The uncertainties in thc experimental results are about +0.3 A in distance and ÷ 15% in the fraction of pairs at a given distance. For comparison, the expected near-neighbor arrangement for a random distribution and a clustered local environment are also given. Both of these cases have the same distances. Sample

SEDOR Pair

Experiment Peak I

Expcriment Peak 2

Randon3 and ('lustered Cases

r~ (A)

Frac. ('4)

r. (A)

Frac. ((,:)

r (A)

Rand. Frac.

Clust. Frac.

a-Si:(H, B, P) 2 at% P 2 atSl B 13 at% tt a-Si:(H, B)

B P B-H P-I!

2.1 2.5 2.(~

40 4(I 5O

3,8 3,7 4.3

,'~(I 6{I 5O

1.9~ ~1.4 1.42

~ 43 43

100 I00 100

(1.5 at% B

B-|!

1.6

411

3. [

611

~] .4

35

100

P-H

2.6

53

4. I

47

35

10(I

I0 at% H a-Si:(H, P) (I.5 aff~ P I0 at
(4)

1.42

J.B. Boyce, S.E. Ready / NMR investigation of hydrogenated amorphous silicon

First, consider the B - P results for the compensated sample. The two inequivalent B sites that fit the data are 4 0 + 15% of the B with a B - P distance of 2.1 + 0.3 A and 60 +- 15% of the B with a P at a mean distance of about 3.8 + 0.3 A,. The first site has its B - P distance close to that in BP (1.96 A), showing that clustering occurs for 40 + - 15% of the pairs. This is well beyond the random value of 8%, indicating a chemical affinity of the B and P in a-Si. The B - H S E D O R decay and fit for aSi:(H, B) are shown in fig. 8(a). The results for the p-type material indicates that 40 +_ 15% of the B atoms have a hydrogen atom near neighbor located at a distance of 1.6+- 0 . 3 A . This agrees with the B - H bonding distance found in B_~H6 ( 1 . 4 A ) and with a theoretically derived B - H distance (1.63,~,) in crystalline Si [33]. Thus the data would indicate that about half of the B atoms have a hydrogen atom as a near neighbor. In contrast, the data for the compensated material indicate that there is no short B - H distance and neither the random nor clustered case apply. Possibly the clustering of the B

1.0

I B-H SEDOR a-Si:(H,B)

l

(a) 0

250

500

750

r ~sec) 1.0

06

_

250

50-0

-

-

750

7 (/~sec) Fig. 8. The (a) B - H S E D O R decay for p-type a-Si:H and (b) P - H S E D O R decay for n-type a-Si:H. The curve through the data points is a two Gaussian fit as described in the text, with the results listed in table 2.

317

and P in the compensated material prevents significant B - H bonding. The P - H results for both the compensated and singly-doped n-type samples yield a closest P - H distance of 2.6 +_ 0.3 A. This value is too large to be attributed to direct P - H bonding which is typically 1.4 A, similar to that of Si-H. Neither the random nor the clustered case applies to P-H.

4.3. Structural models and implications The B and P are found to cluster in compensated a-Si:H. Such a result may hint at an explanation for why the B and P dopant concentrations in the films tend to equalize during growth even if the dopant gas concentrations in the plasma are quite different [47]. Also the fact that the B and P tend to cluster has implications for the band structure and may give rise to neutral donor states above the valence band that are observed in compensated material [46, 47]. For the p-type material, the H is found to be a first neighbor to about half of the B dopants. This structure is very close to proposals for the B - H structure in passivated p-type crystalline silicon [33, 34], shown in fig. 9(a). For this crystalline case, it is found that the total energy minimum configuration corresponds to the hydrogen residing in a bond-center site between the boron and a Si neighbor. The antibonding sites of B and of Si were found to have a higher total energy. The B relaxes by about 0.5 A from the ideal four-fold site toward the three-fold site. This plus some Si relaxation provides a B - H distance of about 1.6 A, as does a cluster calculation (1.63 A) [33]. Both are in reasonable agreement with the N M R experimental value of 1.6 +0.3,&. It should be noted that ion channeling results arrive at this same atomic configuration but with a smaller B relaxation of 0.28A [31]. For the n-type material, the determined P - H distance of 2.6 + 0.3 A is too large to be attributed to direct P - H bonding which is typically 1.4A. This result is in qualitative agreement with the structure proposal for passivated n-tye crystalline silicon [30, 34, 35]. For this crystalline case, it is found that the total energy minimum

31~

.I.B. Boyce, S.E. Ready ' N M R invustication (~! hvdro.genated amotTgtou.s silicon

Passivated B-doped c-St

Si

O.

--

I'

:...

B-H~I.6,,g,

H

~

. . . . . . .

=~

-. . . . . . . . . . . . . . . . . . . . . .

Passivated P-doped oSi

St, _

,, •

"

. "

'.

.' •

" '

= 2,83,3, •

j

P-H =4.32A !

LL . . . . . . . . . . . . . . . . . . . . . . . . . . . .

j

Fig. 9. The theoretically determined H-passivation structures t\~r B-doped and P-doped crystalline Si. position for H is at the antibonding site of a Si nearest neighbor of a substitutional P atom, as shown in fig. 9(b). The direct bonding of H to a P has a significantly higher total energy than the bonding to a neighboring St. The detailed structure of this model, however, places the H 4.3,4 from the P, significantly further than measured. This distance is closer to the second P - H distance determined experimentally, namely, 4.1 ± 0 . 3 A . The calculations found that the other three antibonding sites that are closest to the P are less favorable than this further site. The amorphous case, however, has significantly more H, and the bulk of this H provides the fourth bond to a Si that is three-fold coordinated to other St. It ties up the Si dangling bond and is not thought to provide a fifth bond to the Si. A specific model, suggested by Jackson and Zhang [40], that does yield the experimentally determined distance is the structure of fig. 9(b) plus a near-neighbor H~, fig. 3. The H in the antibond-

ing site of the 1--I~: is 2.45 A from thc P. So the S E D O R data are consistent with the picture in which about half of the P has a neighboring H:~ (P-H-2.45A) and the other half are in the theoretically-derived passiwttcd structure for crystallinc Si ( P - H -- 4.3 A, rig. 9(b). Whether the half of the P that have a neighboring 1--I:~:also are in the passivated structure of fig. 9(b) cannot bc determined from the S K D O R data since the 2.45/X distance dominates the signal decay so that a second 4.3 A distance makes a negligible contribution. These results indicate that H plays a significant role in determining the doping efficiency of aSi:H. Approximately half of the B in p-type and half of the P in n-type a-Si:H are found to bc in H-dopant complcxes determined to passivatc dopants in crystalline St. The substantial relaxations that do occur in these H-dopant complexes place the dopants in a highly distorted environment, essentially three-fold coordinated. This then is consistent with the single N M R results that most of the dopants are three-fold coordinated and with models of doping that rely on three-fold coordination 142]. Nonethelcss, detailed models will have to include the signiticant role of H in doping efficiency in a-Si:H and consider the effects of different ~'three-fold'" configurations such as BSi~ and BSi~H.

Acknowledgements

The authors arc pleased to acknowledge the significant input of N.M. Johnson, R.E. Norberg, M. Stutzmann. and C.C. Tsai and informative discussions of structural models with W.B. Jackson and S.B. Zhang. This work was supported in part by SERI.

References

[1] P.C. Taylor, in: Semiconductors and Scmimctals, Vol. 21, Part C (Academic Press. New York, 1084) p. qq. [2[ J.A. Rcimer and M.A. Pctrich, in: Advances m Disordered Semiconductors-Vol. 1: Amorphous Silicon and Related Matcrials, ed. tt. Fritzschc (World Scientilic, Singapore, 1989) p. 3.

J.B. Boyce, S.E. Ready / NMR investigation of hydrogenated amorphous silicon [31 J.B. Boycc. S.E. Ready, M. Stutzmann and R.E. Norberg, J. Non-Cryst. Solids 114 (1989) 2ll. [4J J.A. Reimer, R.W. Vaughn and J.C. Knights, Phys. Rev. B24 11981) 33611. [51 W.E. Carlos and P.C. Taylor, Phys. Rev. B26 (1982) 36115. [6] J.A. Reimer. R.W. Vaughan and J.C. Knights, Solid State Commun. 37 11981) 161. [7] D.J. Leopold, J.B. Boyce, P.A. Fedders and R.E. Norberg, Phys. Rev. B 26 (1982) 6053. [8] J.B. Boyce and M. Stutzmann, Phys. Rev. Lett. 54 (1985) 562. [91 J.B. Boyce, in: Hydrogen in Disordered and Amorphous Silicon, eds. G. Bambakidis and R.C. Bowman (NATO Advanced Study Inst. Proc. Rhodes, Greece, 1985) p. llJl. [10] V.P. Bork, P.A. Fedders, R.E. Norberg, J.B. Boyce and M. Stutzmann, J. Non-Cryst. Solids 77 & 78 (1985) 711. [11] V.P. Bork, P.A. Fedders, D.J. Leopold, R.E. Norberg, J.B. Boyce and J.C. Knights, Phys. Rev. B36 (1987) 9351. [12] S. Hayashi, K. Hayamizu, S. Yamasaki, A. Matsuda and K. Tanaka, J. Appl. Phys. 56 11984) 2658. [13] M. Kumeda, Y. Yonezawa, A. Morimoto, S. Ueda and T. Shimizu, J. Non-Cryst. Solids 59 11983) 775. [14] S.E. Ready, J.B. Boyce and C.C. Tsai, Mat. Res. Soc. Syrup. Proc. Vol. 118 (Materials Research Society, Pittsburgh, 19881 p. 103. [15] T. Shimizu, M. Kumeda, A. Morimoto, Y. Tsujimura and I. Kobayashi, Mat. Res. Soc. Symp. Proc. Vol. 70 (Materials Research Society, Pittsburgh, 1986) p. 313. [16] M. Stutzmann, R.A. Street, C.C. Tsai, J.B. Boyce and S.E. Ready, J. Appl. Phys. 66 (1989) 569. [17] E.J. Vanderheiden, G.A. Williams, P,C. Taylor, F. Finger and W. Fuhs, Mat. Res. Soc. Syrup. Proc. Vol. 149 (Materials Research Society, Pittsburgh, 1989), p. 5/13. 118] J.B. Boyce and S.E. Ready, Phys. Rev. B38 (1988) 11008. [19] J.B. Boyce and S.E. Ready, in: Advances in Disordered Semiconductors-Vol. 1: Amorphous Silicon and Related Materials, ed. H. Fritzsche (World Scientific, Singapore, 1989) p. 29. [2//] S.E. Ready, J.B. Boyce, N.M. Johnson, J. Walker and K.S. Stevens, Mat. Res. Soc. Syrup. Proc. Vol. 192 (Materials Research Society, Pittsburgh, 1990) p. 127. [21] S.G. Greenbaum, W.E. Carlos and P.C. Taylor, Solid State Commun. 43 (1982) 663. [22] S.G. Greenbaum, W.E. Carlos and P.C. Taylor, J. App. Phys. 56 (1984) 1874. [23] J.A. Reimer and T.M. Duncan, Phys. Rev. B 27 (1983) 4895. [24] M.J. McCarthy and J.A. Reimer, Phys. Rev. B36 (1987) 4525. [251 M.S. Conradi and R.E. Norberg, Phys. Rev. B24 ( 1981 ) 2285.

319

[26] See, for example, the reviews in: Advances in Disordered Semiconductors- Vol. 1: Amorphous Silicon and Related Materials, ed. H. Fritzsche (World Scientific, Singapore, 1989). [27] W. Beyer, in: Tetrahedrally-Bondcd Amorphous Semiconductors, eds. D. Adler and H. Fritzsche (Plenum, New York, 1985) p. 129. W. Beyer, H. Wagner and F. Finger, J. Non-Cryst. Solids 77&78 (1985) 857. [28] J.l. Pankove, D.E. Carlson, J.E. Berkreyheiser and R.O. Wance, Phys. Rev. Lett. 51 (1983) 2224. [29] N.M. Johnson, Phys. Rev. B31 (1985) 5525. [30] N.M. Johnson, C. Herring and D.J. Chadi, Phys, Rev. Lett. 56 (1986) 769. [31] A.D. Marwick, G.S. Oehrlein and N.M. Johnson, Phys. Rev. B 36 11987) 4539. A.D. Marwick, G.S. Oehrlein, J.H. Barrett and N.M. Johnson, Mat. Res. Soc. Sym. Proc. Vol. 104 (Materials Research Society, Pittsburgh, 1988) p. 259. [32] M. Stutzmann, Phys. Rev. B 35 (1987) 5921. [33] G.G. DeLeo and W.B. Fowler, Phys. Rev. B31 (1985) 6861. [34[ K.J. Chang and D.J. Chadi, Phys. Rev. Lett. 611 (19881 1422. [35] S.B. Zhang and D.J. Chadi, Phys. Rev. B41 (1990) 3882. [36] A. Abragam, Principles of Nuclear Magnetism (Oxford University Press, London, 1961). [37] P.A. Fedders, R. Fisch and R.E. Norberg, Phys. Rev. B31 (1985) 6887. [38] G. Lucovsky, R.J. Nemanich and J.C. Knights, Phys. Rev. B 19 (1979) 21164. [39] K.J. Chang and D.J. Chadi, Phys. Rev. Lett. 62 (19881 937, [40] W.B. Jackson and S.B. Zhang, in: Advances in Disordered Semiconductors-Vol. 1: Amorphous Silicon and Related Materials, ed. H. Fritzsche (World Scientific, Singapore, 1989) to be published, W.B. Jackson and S.B, Zhang, Physica B 170 (1991) 197 (these Proceedings). [41] N.M. Johnson, S.E. Ready, J.B. Boyce, C.D. Doland, S.H. Wolff and J. Walker, Appl. Phys. Lett. 53 11988) 1626. [42] R.A. Street, J. Non-Cryst. Solids 77 & 78 (1985) 1. [43] J.C. Knights, T.M. Hayes and J.C. Mikkelsen, Jr., Phys. Rev. Lett. 39 (1977) 712. [44] F. Boschcrini, S. Mobilio, F. Evangelisti and A.M. Flank, J. Non-Cryst. Solids 114 (1989) 223. [45] C.G. Brown and D.F. Holcomb, Phys. Rev. B 10 (1974) 3394. [46] M. Stutzmann, D.K. Biegelsen and R.A. Street, Phys. Rev. B 35 (1987) 5666. [47] R.A. Street, D.K. Biegelsen and J.C. Knights, Phys. Rev. B24 (1981) 969.