[PPT] - Active Electronic Impurity Doping of Silicon Nanovolumes: Failure PowerPoint Presentation

SLIDE 1

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Active Electronic Impurity Doping

f Silicon Nanovolumes:

Failure and Alternatives

Dirk König Integrated Material Design Centre (IMDC) and School of PV and Renewable Energy Engineering (SPREE), University of New South Wales (UNSW), Sydney/Australia

1

http://www.imdc.unsw.edu.au/ http://www.engineering.unsw.edu.au/energy-engineering/

SLIDE 2

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

It is not about

co-doping to improve luminescence properties
plasmonics which requires (semi-)metallic properties
chemical activation of solid surfaces, e.g. for catalysis

which use doping densities in the ≥ 1 atom-% range

Definition of Scope

2

This presentation is about inducing electronic p- or n-type behaviour into ultrasmall Si nanovolumes such as nanocrystals (NCs), fins (for ULSI-FETs), nano-wires and -wells.

SLIDE 3

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Outline

1. Conventional Doping, Theory

1.1 Broader Perspective 1.2 Dopant Formation Energies 1.3 Dopant Ionization Energies

2. Conventional Doping, Experiment
3. Phosphorus (P) in SiO2/Si NC Systems: h-DFT, APT, XANES
4. Interface Impact of Dielectric – Alternative 1
5. Excess Si (Ge) as Donor in Adjacent Barriers – Alternative 2
6. Modulation-Doped SiO2; Preliminary Results – Alternative 3
7. Conclusions

3

SLIDE 4

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

macroscopic scale: Dopant, foreign atom segregation at high

temperatures, cm range → Si float zone refinement, 1450 °C

microscopic scale: Dopant segregation to grain boundaries,

µm range → Si solid phase crystallization, 600 to 800 °C What prevents doping of Si-NCs? # doping of Si-NCs requires energy (mechanical stress & surface tension, electrostatic interaction); NCs build up counter-stress → self-purification [1−6] # DBs saturating dopants (being inactivated) delivers energy → DB passivation at Si NC interface (||) by fully saturated dopants # ionization energy (Eion) of dopants >> kT for NCs showing quantum confinement (QC)

segregation anneal of Si nanocrystals (Si-NCs) in Si-rich SiO2 or

Si3N4 carried out at 1050 to 1200 °C …

1. Conventional Doping, Theory

1.1 Broader Perspective

Theory Experiment [1] PRB 75, 235304 (2007) [3] PRB 72, 113303 (2005) [5] PRL 100, 026803 (2008) [2] PRL 100, 179703 (2008) [4] Nano Lett. 8, 596 (2008) [6] PRB 80, 165326 (2009) 4

SLIDE 5

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015 [1] PRB 75, 235304 (2007) [2] PRL 100, 179703 (2008) [3] PRB 72, 113303 (2005)

1. Conventional Doping, Theory

1.2 Dopant Formation Energies

small NCs: much increased dopant formation energy Eform for many materials
atom size difference   stress   Eform , triggering self-purification
Eform(bulk Si) ≈ 0.1 eV[1] = kBT at T ≈ 900 °C,

Eform(Si NC) = 6 to 14 × Eform(bulk Si)

studies so far do not include anions of dielectric
r DBs which are likely to getter dopants

[2]

r

Si146BH100 dQD = 17.8 Å [3]

r

5

SLIDE 6

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

1. Conventional Doping, Theory

1.3 Dopant Ionization Energies

small Si NCs experience quantum confinement (QC) for dNC ≤ 2 rexc ≈ 9 nm
dopant is point defect (analogy to H atom) – QC only for d << 9 nm

 Eion(ND) with dNC for donor on lattice site  tiny ionization probability (Ρion) of dopants in Si NCs with notable QC, cf. Si bulk, 300 K: Ρion = exp(-Eion/kT) = 0.15 with Eion(P as ND) = 0.049 eV, Ρion increases further for ND ≥ 1017 cm-3

[1] D. König, Chapter 8, Nanotechnology for Photovoltaic Devices, J. Valenta & S. Mirabella (Eds.), Pan Stanford, 2015

[1]

6

SLIDE 7

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Outline

1. Conventional Doping, Theory
2. Conventional Doping, Experiment

2.1 Free-Standing NCs 2.2 ULSI MISFETs 2.3 Sample Preparation Issues 2.4 Characterisation Strategy for Active Dopants

3. Phosphorus (P) in SiO2/Si NC Systems: h-DFT, APT, XANES
4. Interface Impact of Dielectric – Alternative 1
5. Excess Si (Ge) as Donor in Adjacent Barriers – Alternative 2
6. Modulation-Doped SiO2; Preliminary Results – Alternative 3
7. Conclusions

7

SLIDE 8

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

2. Conventional Doping, Experiment

2.1 Free Standing NCs

doping fails; 1.6×1019 cm-3  one donor every 4 nm Pearson, Bardeen[2]: Semiconductor – metal transition at ND ≈ 1.5×1020 cm-3 (≈ 0.3 atom-%)

[1] PRL 100, 026803 (2008) [2] Phys. Rev. 75, 865 (1949) [3] PRB 80, 165326 (2009)

[1] [3]

[P]EPR = P concentration which shows EPR

signal of unpaired e– (built-in donor) as required condition for active doping; ≤ 10-4 [P]nom

[P]EPR

s.c. = P concentration at the NC

after processing, incl. inactive donors

conductivity, 270 K[1]: 1 / 3 / 500 for ND =

0 / 1.6×1019 / 1.5×1020

8

SLIDE 9

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

2. Conventional Doping, Experiment

2.2 ULSI MISFETs

[1] JAP 104, 093709 (2008) [2] Nanotechnology 24, 275705 (2013)

[2], plasma doping + spike RTA [2], plasma doping + spike RTA

ultra-high density doping in Drain/Source (D/S) areas: dopant activation problem Entropy of lattice modification limits dopant formation: Profiles diffuse out ULSI: How can we introduce n- and p-type behaviour into ultrasmall Si nanovolumes without dopant clustering and out-diffusion?

dopant out-diffusion of 5 nm (flash anneal [FLA]) to 15 nm (spike RTA)[1]

 MISFET channel length diminished, or channel even shorted (spike RTA)[2]

dopant clustering and inactivation (spike RTA)[2]

9

SLIDE 10

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

kMC = kinetic Monte-Carlo SIM

2. Conventional Doping, Experiment

2.3 Sample Preparation Issues: Excess Si

TEM images only show NCs with right (low index) orientation to image plane
Full account of Si content requires ∈-specific imaging methods:

energy-filtered (EF) TEM, electron energy loss spectroscopy (EELS) or APT

Too much excess Si, forming a-Si/SiNC networks behaving like a-Si
In such networks, conventional doping works to some extent but

# No control over Si NC properties (energy gap, charge storage, PL response) # low carrier mobility ( ULSI), high recombination rate ( Solar Cells)

10

SLIDE 11

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

[1] Phys. Stat. Sol. B 248, 472 (2011) [2] JAP 118, 154305 (2015) [3] JAP 113, 133703 (2013) [4] APL 95, 153506 (2009) [1]

Si NC/SiO2 SL, CV data presented as evidence for Si NC donor (P) doping[1,2]:

# requires continuous SCR  continuous Si NC / a-Si network # Oxide capacity COx  with frequency  due to low majority carrier mobility and defect density typical for a-Si; “higher Si% P doped sample” = ??? # separate Si NCs in SiO2 behave very different[3]

2. Conventional Doping, Experiment

2.3 Sample Preparation Issues: Excess Si

∆COx

25 Si QD SL periods

[1]

similar issues with

# SiO0.66 /SiO2 vertical superlattice (SL) and CV for NC doping with Boron [2] # SiO0.7 NC SL solar cell [4]: Egap = 1.8 eV; Egap(a-Si:H) = 1.6 … 1.7 eV

11

SLIDE 12

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Too much P, resulting in SiPxOy ternary oxide Pearson, Bardeen: Si semiconductor-metal transition at ≈ 0.3 at-% dopant concentration[1]

[1] Pearson, Bardeen,

Phys. Rev. 75, 865 (1949)

[2] APL 87, 211919 (2005) [3] APL 92, 123102 (2008) [4] APL 102, 013116 (2013)

P concentration in most publications range

from ≈ 1[2] via ≈ 9 [3] to ≈ 12 [4] atom-%;

data from [4] below, using SiO0.9 as pre-cursor

for Si NCs in SiO2; rectification ratio (current densities at +/- 10 V) ≈ 9 to 35

2. Conventional Doping, Experiment

2.3 Sample Preparation Issues: Excess Dopant Densities

What do Photolumines-

cence (PL) data descibe?

How to obtain clear evi-

dence for active dopants (yielding charge carriers)?

12

SLIDE 13

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Spectroscopies: Photoluminescence (PL), Electron Paramagnetic Resonance (EPR), X-ray Photoelectron (XPS), X-ray Absorption Near Edge Structure (XANES)

1. Detects element (∈)-specific signal of dopant (donor ND)
2. Detects oxidation state ND/ND

+ or unpaired electron (e–) at T = 300 K

3. Sensitivity to detect ND concentrations < 0.3 atom-% [1.5×1020 cm-3]
4. Non-destructive to leave atomic configuration of sample intact
2. Conventional Doping, Experiment

2.4 Characterisation Strategy for Active Dopants[1]

[1] D. König, Chapter 8, Nanotechnology and Photovoltaic Devices, Valenta & Mirabella (Eds.), Pan Stanford, 2015 [2] König, Gutsch, Gnaser, Wahl, Kopnarski, Göttlicher, Steininger, Zacharias, Hiller, Sci. Rep. (Nature) 5, 09702 (2015)

PL (presence of dopant causes signal – origin unclear); EPR; XPS; XANES PL; EPR (Can detect ND via unpaired e–, but not their ionization at T = 300 K; T ≤ 20 K for reasonable ND concentrations [thermal noise]), XPS; XANES PL, EPR, XPS (limit ca. 0.5 atom-%), XANES (X-ray fluorescence: X-ray MFP  by ca. 103 over e– MFP  bigger volume  ND sensitivity below alloy conc.) PL, EPR, XPS (very surface sensitive, needs sputtering into material – artefacts like ∈-specific sputter rates, atomic re-arrangement), XANES XANES: only technique fulfilling all criteria to detect ND & ND

+ at T = 300 K

[Embedded] NCs: high resolution spatial scans; Atom Probe Tomography (APT)

13

SLIDE 14

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Outline

1. Conventional Doping, Theory
2. Conventional Doping, Experiment
3. Phosphorus (P) in SiO2/Si NC Systems: h-DFT, APT, XANES
4. Interface Impact of Dielectric – Alternative 1
5. Excess Si (Ge) as Donor in Adjacent Barriers – Alternative 2
6. Modulation-Doped SiO2; Preliminary Results – Alternative 3
7. Conclusions

14

SLIDE 15

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

3. Conventional Doping,

P in SiO2/Si NC System: DFT*, APT, XANES [1]

SiO2 [Si29O40(OH)36],

SiO2:P [Si28PO40(OH)36]: deep defects, just outside 1.5 nm NC gap  massive SiO2 barrier lowering  increased conductivity of samples with P is no evidence for active doping

*computational details: supplements to PRB 78, 035339 (2008) and Adv. Mater. Interfaces 1, 201400359 (2014) Si29O40(OH)36 [SiO2] Si28PO40(OH)36 [SiO2:P] Si74O55(OH)35H29 [SiO0.9] Si73PO55(OH)35H29 [SiO0.9:P]

Si, P, 1-nn O of P, O, H

[1] Sci. Rep. (Nature), doi: 10.1038/ srep09702 (2015)

SiO0.9 [Si74O55(OH)35H29],

SiO0.9:P [Si28PO40(OH)36]: deep defect, just outside 1.5nm NC gap  see above defects deep inside NC gap  recombination during transport

15

SLIDE 16

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Si83P(OH)3(OH)62 Si84P(OH)4(OH)63 Si84OP(OH)4(OH)63

*computational details: supplements to PRB 78, 035339 (2008) and Adv. Mater. Interfaces 1, 201400359 (2014) [1] Sci. Rep (Nature), doi: 10.1038/srep0970 (2015)

P at fully OH-terminated

1.5 nm Si NC [Si84(OH)64] for corner Si [>Si(OH)2  >P(OH)3], directly at corner Si [−OH  −P(OH)4], at corner Si via O bridge [−OH  −O−P(OH)4] as most stable configurations (gettering dangling bonds)

gettered P at Si NC

does not introduce defects into NC gap

Si84(OH)64

slight shift of highest
ccupied molecular
rbital (HOMO), lowest

unoccupied MO (LUMO) to higher binding energies with more O atoms groups, not due to P presence per se

Si, P, O, H

65 64 67 68

3. Conventional Doping,

P in SiO2/Si NC System: DFT*, APT, XANES [1]

16

SLIDE 17

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015 *computational details: supplements to PRB 78, 035339 (2008) and Adv. Mater. Interfaces 1, 201400359 (2014) [1] König, et al., Sci. Rep (Nature), doi: 10.1038/srep0970 (2015) [2] PRB 72, 195323 (2005)

fully OH-terminated 1.5 nm Si NC with P on

substitutional (Si) lattice site [Si83P(OH)64]: # extremely rare; Eform ≈ 10 Eform(bulk Si) # HOMO is donor MO, but Eion = 0.51 eV … Ρion = 2.9×10-9 at T = 300 K = 1.9×10-8 Ρion(bulk Si) Eion  with dNC , but not enough for P to provide e– to NC showing QC as evident from XANES

interstitial P, valid for virtually 100 % of P,

experimental P coordinates used [2]: # HOMO of P is 2.18 eV below LUMO of Si84(OH)64 # LUMO of P is 0.46 eV below LUMO of Si84(OH)64 active PL transition, 720 nm

3. Conventional Doping,

P in SiO2/Si NC System: DFT*, APT, XANES [1]

17

SLIDE 18

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

APT scan of 3 nm Si NC [top] and complete SL [right]; iso-surfaces shown at 79 mol-% Si (SiO0.3). Si = red, P = green. Length scales in nm.

[1] JAP 115, 034304 (2014) [2] S. Gutsch, PhD thesis, University of Freiburg (2014) [3] Sci. Rep (Nature), doi: 10.1038/srep0970 (2015)

max. P concentration at interface  supports thermodynamics of gettering
55 % of P reside in SiO2 matrix (85 % of tot. volume)
30 % of P trapped at interface
only 15 % of P diffuse into Si NC

 self-purification

3. Conventional Doping,

P in SiO2/Si NC System: DFT, APT [1-3], XANES

18

Proxigram, radial atomic concentrations of Si NCs from distance at interface (SiO0.3).

SLIDE 19

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

XANES of P-doped Si wafer reference, 30 bi-layer Si NC SLs (SiO0.93 precursor, 5 nm SiO2 barriers) and bulk SiO0.93 (300 nm thick). SiOx samples have 30 nm buffer-SiO2 for separation from Si substrate. Grey symbols & lines show P oxidation numbers.

[1] JAP 115, 034304 (2014) [2] S. Gutsch, PhD thesis, Freiburg University (2014) [3] König, et al., Sci. Rep (Nature), doi: 10.1038/srep0970 (2015) [4] Pearson, Bardeen, PR 75, 865 (1949)

bulk Si, ND(P) = 0.2 to 1×1019 cm-3 = 17- to 80-fold below

alloy limit[4]: P0 = – 0.09, P+ = + 0.91 (Ρion = 0.15), avg. P = +0.06  oxidation number (ON) ≈ 0, E(P, K shell) = 2144.8 eV P ionization shifts E(P, K shell) scanned by XANES, DFT with all-e– MO-BS used to interpret XANES data[3]

bulk SiO2: P+1.29 (SiO2) … P+1.98

(P4O10 cage), both from DFT  ON = +4 ... +5; E(P, K shell) = 2148.6 (as in DFT) and 2152.4 eV P2O5 phase segregation in SiO2?

Si NC SLs (2-5 nm), bulk SiOx:

E(P, K shell) = 2143.7 eV = E(P, K shell, bulk Si) – 1.1 eV  ON = –1, charge of P much more negative than in bulk Si  e– attached to P; no doping

3. Conventional Doping,

P in SiO2/Si NC System: DFT, APT, XANES[1-3]

19

SLIDE 20

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Outline

1. Conventional Doping, Theory
2. Conventional Doping, Experiment
3. Phosphorus (P) in SiO2/Si NC Systems: h-DFT, APT, XANES
4. Interface Impact of Dielectric – Alternative 1

4.1 Energy Offset – Oxide vs. Nitride Embedding 4.2 Interface Impact of Dielectric, Nanoscopic Field Effect 4.3 A Glance at Device Concepts

5. Excess Si (Ge) as Donor in Adjacent Barriers – Alternative 2
6. Modulation-Doped SiO2; Preliminary Results – Alternative 3
7. Conclusions

20

SLIDE 21

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

4. Interface Impact of Dielectric

4.1 Energy Offset – Oxide vs. Nitride Embedding[1,2]

[1] PRB 78, 035339 (2008) [2] Adv. Mater. Interfaces 1, 201400359 (2014) [3] APL 85, 3408 (2004)

Egap[Six(NH2)y] ≈ Egap[Six(OH)y] + 0.6 eV - as for NCs in SiO2 vs. Si3N4 (expr.)[3]

DFT modelling, fully X-terminated SixXy, X = H, NH2, OH, F (CH3)[1] - Si10X16 (7 Å) to Si165X100 (18.5 Å), Si286X144 (H, F, OH; 22.2 Å)[2], Si680H196 (29.6 Å)

NH2 and OH termination emulate Si3N4 & SiO2 embedding

very well[2]

N, O, F

dominate electronic structure for dNC ≥ 4 nm[1], NC size has little influence on Egap

electronic structure is f(Χ,EN,Eion) of anion relative to Si
HOMO  1.5eV, LUMO  2.1eV for Six(NH2)y vs. Six(OH)y

21 Si165(NH2)100 Si165(OH)100

SLIDE 22

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

ERDA: ca. 1 atom-% N in Si NC SL stack
XPS: K shell signal of N(Si3N4) at 397.5 eV[2]

shifted by +0.55 eV for interface N, indicating O for N as 2-nn atoms:

Si scan, HAADF

STEM

[1] Adv. Mater. Interfaces 1, 201400359 (2014) [2] D. Briggs, M. P. Seah , Practical Surface Analysis (Vol. 1) , 2nd ed. , Wiley& Sons , Chichester 1993

Experiment confirms DFT results: SiOx/SiO2 layer stacks annealed in Ar vs. N2, with some N built into Si NC/SiO2 interface derived from

Si(NC), N, O, Si(SiO2)

PL: Egap ,

∆ Egap  with dNC 

Interface impact

22

4. Interface Impact of Dielectric

4.1 Energy Offset – Oxide vs. Nitride Embedding [1]

∆Egap = 2.8 % relative

SIMS (HAADF-STEM):

SiN signal in phase with Si3 signal

SLIDE 23

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Si10 NC in 3 ML pure SiO2 (a) vs. ¼ ML N at interface (b) reproduce

Egap; HOMO & LUMO shift towards Evac(c)  with Si165X100 (X = OH, NH2) data, interface coverage estimate is 0.1 to 0.2 ML N What does ca. 1 atom-% N mean in terms of Si NC/SiO2 interface coverage?

[1] Adv. Mater. Interfaces 1, 201400359 (2014)

Si(NC), Si(SiO2), N, O Si10, 3ML SiO2 Si10, 3ML SiO2, ¼ML N at interface

(b) (a) (c)

Si35(OH)x(NH2)y test (x+y = 36)

(d)

Egap change not linear with O/N ratio (d) 

interface coverage is 0.1 … 0.2 … 0.5 ML N

23

4. Interface Impact of Dielectric

4.1 Energy Offset – Oxide vs. Nitride Embedding [1]

SLIDE 24

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015 [1] Adv. Mater. Interf. 1, 201400359 (2014) [2] JAP 113, 033528 (2013)

ICT linked to sum of volume elements (dV) over constant volume V, dV are (a)
1 cubicle for no QC (bulk); dV = AG × h
2 cubicles for 1d QC (QW); dV = AG × h
2 wedges for 2d QC (Q-wire); dV = ½ AG × h
2 pyramids for 3d QC (QD); dV = 1/3 AG × h

 dICT = 1 : 2 : 4 : 6 for bulk : QW : Q-Wire : QD How can we describe & determine interface charge transfer (ICT) & its impact?

(a)

experiment[2]: smaller Egap for Si NCs of same

size in SiO2 vs. Si3N4 for dNC ≤ 7 nm,  dICT(QD) = 70 Å, confirmed by Si NC ionization [DFT, (b)] massively exceeding NSi/Nx for O, N, F; ICT ~ dNC

2 (interface

area), ICT is far from being saturated

ICT very strong, 81 e– (1 e– at Coulomb

blockade) [Si680F256, (b)]; ICT needs ≤ 3 MLs of dielectric[1]  u-thin barriers  Nanoscopic Field Effect; controls electronic structure of ≥ 70 Å Si QDs, ≥ 47 Å Si Q-wires, ≥ 23 Å Si QWs

24

4. Interface Impact of Dielectric

4.2 Interface Impact of Dielectric, Nanoscopic Field Effect[1]

(b)

SLIDE 25

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

4. Interface Impact of Dielectric

4.2 A Glance at Device Concepts[1]

[1] Adv. Mater Interfaces 1, 201400359 (2014)

u-thin barrier QW-SL as un-

doped photodetector on-chip; fast response, low heat loss Band structure shift: n/p-type Si, CMOS-compatible materials & technology

SiO2 / Si3N4 / SiO2  n / p (depleted) / n; for

p / n (depleted) / p, swap SiO2, Si3N4 locations

compatible with Fin-FET design

25

SLIDE 26

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Outline

1. Conventional Doping, Theory
2. Conventional Doping, Experiment
3. Phosphorus (P) in SiO2/Si NC Systems: h-DFT, APT, XANES
4. Interface Impact of Dielectric – Alternative 1
5. Excess Si (Ge) as Donor in Adjacent Barriers – Alternative 2

5.1 Principle, Material Choice 5.2 h-DFT on Si-/Ge-doped Barrier Material 5.3 Technological Implementation

6. Modulation-Doped SiO2; Preliminary Results – Alternative 3
7. Conclusions

26

SLIDE 27

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

use excess Si or Ge as shallow donor in

adjacent barrier layer (matrix)[2]:

desired band offsets to Si or Ge QDs (wires,

wells): confinement and tunnelling [type I band offsets]

max. anionic Gr. V element → min. auto-compensation

when doping → Gr. III nitrides

no NC doping? – modulation doping (III-V)[1]

 no defects by dopant species at Si NCs  Doping by minute Si/Ge amount diffusing into barrier material → NC formation OK  doping matrix much easier: less restrictions

solid Evac - EC [eV] Evac - EV [eV] ∆EC(Si) [eV] ∆EV(Si) [eV] Egap [eV] remarks AlN [2-5] 2.1 8.2 – 8.4 2.0 3.0 – 3.2 6.1 – 6.3 Würtzite (Zinc blende) GaN[ 6,7] 3.3 6.7 0.75 1.55 3.4 – 3.5 Würtzite (Zinc blende)

[1] APL 33, 665 (1978) [2] AIP Advances 3, 012109 (2013) [4] JJAP 33, part 1, 2453 (1994) {open access} [5] Czech. J. Phys. B 30, 586 (1980) [2] JAP 50, 896 (1979) [6] MRS Symp. Proc., Pittsburgh/PA. 423, 69 (1996) [3] APL 68, 2541 (1996) [7] S. Adachi, Handbook on Physical Properties of Semi- conductors, Vol. 2, Kluwer Academic Publishers (2004)

5. Excess Si (Ge) Doing Adjacent Barriers

5.1 Principle, Material Choice[2]

27

SLIDE 28

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Ge? – use Ge3N4 or Si3N4 with excess Ge
Al107GeN108H126 (MOs of H eliminated): LUMO(AlN) – UMO(Ge)

= 0.16 eV and 0.36 eV [*] (expr. = ?) → donor Ge in AlN not as attractive as Si

Al107SiN108H126 (MOs of H eliminated): LUMO(AlN) – UMO(Si)

= 0.11 eV and 0.17 eV [*] (expr.: ≈ 0.1 ... 0.3 eV) ... works

HF/3-21G(d)// B3LYP/6-31G(d)

[*] AIP Advances 3, 012109 (2013) {open access}

HF/3-21G(d)// B3LYP/6-31G(d)

5. Excess Si (Ge) Doing Adjacent Barriers

5.2 h-DFT on Si-/Ge-doped Barrier Material[*]

28

SLIDE 29

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015 [a] Diamond & Related Mater. 17, 1273 (2008) [b] APL 81, 1255 (2002) [c] APL 98, 092104 (2011) [d] PRB 61, R16283 (2000) [e] Mat. Sci. Eng. B 50, 212 (1997) [f] Mat. Sci. Eng. B 91, 285 (2002) [g] Sol. Stat. Electron. 42, 627 (1998)

Band offsets of Si to AlxGa1-xN as function of x (left), with experimental Si donor energies from literature. HF-DF values for Si & Ge in AlN from our work shown in orange & mangenta. Band gap of AlxGa1-xN as function of x (right).

) 1 ( 3 . 1 ) 1 ( 42 . 3 13 . 6 ) ( Egap x x x x x − − − + =       − − × ∆ − ∆ + ∆ = ∆ ) GaN ( ) AlN ( ) GaN ( ) ( )] GaN ( ) AlN ( [ ) GaN ( ) (

, , , , , , V C gap V C gap V C V C V C V C

E E E x E E E E x E

Egap(AlN)

AlxGa1-xN band gap as f(x)[1]:

Egap(GaN)

[1] APL 81, 5192 (2002)

use Egap(x) for determining

band offsets ∆EC, ∆EV [2]:

Egap(AlxGa1-xN) = 3.4 to 6.1 eV[1]:

control band offsets ∆EC, ∆EV

Si shallow donor if x ≤ 0.76,

else auto-compensation for ND(AlN) > 1×1017 cm-3

Ge shallow donor if x ≤ 0.3[3]

[3] PRB 56, 9496 (1997)

5. Excess Si (Ge) Doing Adjacent Barriers

5.2 h-DFT on Si-/Ge-doped Barrier Material[2]

29

EC – 0.11 eV EC – 0.16 eV [2] AIP Advances 3, 012109 (2013) {open access}

SLIDE 30

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

AlN can be deposited by sputtering,
incl. stoichiometry control[3]; co-sputtering

with GaN possible[4]

Tmelt ≥ 2750 ºC (AlN)[1], = 2518 ºC (GaN)[2] [1] Y. Goldberg in Properties of Advanced

Semiconductor Materials, Wiley & Sons, New York (2001), pp. 31-47; Mater. Res. Bul. 5, 783 (1970) [2] PRB 7, 1479 (1973) [3] Tamkang J. Sci. Eng. 7, 1 (2004) [4] J. Cryst. Grwth. 311, 459 (2009) [5] JAP 97, 093516 (2005) [6] J. Cryst. Grwth. 234, 637 (2002) [7] JAP 97, 083505 (2005);

J. Cryst. Grwth. 311, 2899 (2009)

[8] PRB 56, 9496 (1997) [9] Diamond & Related Mater. 17, 1273 (2008)

GaN deposited by sputtering[5]

and metal-organic (MO) CVD[6]

donor Eion of Si in GaN = 12 meV[7],

Ge is ”excellent donor”[8]

pre-anneal post-anneal

Process Flow[*]:
1. replace Si3N4 by

AlxGa1-xN for barrier layer deposition

2. in-situ activation

during segregation anneal

∃x. UV LEDs based on Si-doped AlN[9]

[*] AIP Advances 3, 012109 (2013) {open access}

5. Excess Si (Ge) Doing Adjacent Barriers

5.3 Technological Implementation[*]

30

SLIDE 31

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Outline

1. Conventional Doping, Theory
2. Conventional Doping, Experiment
3. Phosphorus (P) in SiO2/Si NC Systems: h-DFT, APT, XANES
4. Interface Impact of Dielectric – Alternative 1
5. Excess Si (Ge) as Donor in Adjacent Barriers – Alternative 2
6. Modulation-Doped SiO2; Preliminary Results – Alternative 3

6.1 Principle, hybrid Density Functional Theory (h-DFT) 6.2 High Frequency – Capacitance-Voltage (HF-CV) 6.3 Deep Level Transient Spectroscopy (DLTS) 6.4 ULSI Devices and HIT Solar Cell Concepts 6.5 Lifetime Measurement

7. Conclusions

31

SLIDE 32

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

6. Modulation-Doped SiO2; Preliminary Results

6.1 Principle, hybrid Density Functional Theory (h-DFT)

SiO2 is best material for Si regarding

# chemical surface passivation # phase separation of SiOx → SiO2 + Si NCs and NC embedding  Modulation doping

f Si by SiO2
SiO2:X modulation doping of Si10 NC (dQD =7 Å) is energy limit

– doping works for all bigger Si [& Ge] NCs (QD confinement )

high e– transfer Si10  X complex in

SiO2: β-LUMO, β-HOMO at X and Si10 NC, overlap through 9 Å perfect SiO2

hybrid-DFT simulation
f Si10 NC in 3 ML SiO2

(a) pure (b) with modulation acceptor complex (X) located 3ML from Si10 NC

32

(b) (a)

SLIDE 33

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

CV of X mono layers (MLs) embedded in SiO2: flat band voltage (Vfb) shift to

more positive values show ionized modulation acceptor complex X (X−)

6. Modulation-Doped SiO2; Preliminary Results

6.2 High Frequency – Capacitance-Voltage (HF-CV)

33

SLIDE 34

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

DLTS of 1 ML X in SiO2 and reference w/o X, showing e− escape from X into Si [scheme (b)]. Deconvoluted sub-peaks due to acceptor X moving ca. 1 ML within SiO2 during anneal. Measurement schemes used for DLTS

low-T DLTS [scheme (b)]: few X release e− into Si, sharp peak of X signal,

tunneling dominant, thermal scattering and hopping suppressed; E(X) ≈ EV(Si) − 0.6 eV, calculated from band bending at VR (Poisson solver)

X moves ≤ ± 2 Å in SiO2 by activation anneal;

ultrathin layers possible w/o out-diffusion of X

6. Modulation-Doped SiO2; Preliminary Results

6.3 Deep Level Transient Spectroscopy (DLTS)

34

SLIDE 35

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

DLTS of 1 and 2 ML X in SiO2 and reference w/o X, showing e− capture of X from Si [scheme (a)]. Different charge densities and kinetics of X MLs at different depths are clearly visible.

high-T DLTS [scheme (a)]: most X release e− in reasonable pulse time tp
probed transient is e− recapture into X from Si after their release at Vp,

correlates with X MLs as function of tp(d[SiO2 barrier]) and total charge density

6. Modulation-Doped SiO2; Preliminary Results

6.3 Deep Level Transient Spectroscopy (DLTS)

35

SLIDE 36

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

Structure (top) and band diagram (bottom) of standard HIT solar cell Structure (top) and band diagram (bottom) of HIT solar cell with ultrathin doped SiO2

X moves ≤ ± 2 Å in SiO2 during activation anneal; layers as thin as 10 Å

possible w/o out-diffusion of X

SiO2:X as passivation layer with majority (hole) conductivity
6. Modulation-Doped SiO2; Preliminary Results

6.4 ULSI Devices and HIT Solar Cell Concepts

CMOS compatible, no size limit, doping of oxide trench for MISFETs, SSDs, …

36

SLIDE 37

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

6. Modulation-Doped SiO2; Preliminary Results

6.5 Lifetime Measurement (unplanned test)

37

ordinary thick (µ-mechanics) DSP Cz-Si wafer, no H passivation, ca. 1.5 Ωcm,

… just a carrier substrate

SLIDE 38

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

7. Conclusions

[1] Sci. Rep (Nature) 5, 09702 (2015) [2] Adv. Mater. Interfaces 1, 201400359 (2014) [3] AIP Advances 3 , 012109 (2013)

size limit of conventional Si-nanovolume (NV) doping (NCs, ULSI):

# self-purification, failed formation and/or ionization, out-diffusion, clustering # dopant detection strategy: PL, EPR, XPS, XANES + APT (HAADF-STEM[1], EELS) # embedded NC doping with P fails (DFT, APT, XANES)[1]

Alternative 1: Interface charge transfer (ICT) to dielectric (SiO2, Si3N4)[2]

# very robust nanoscopic field effect, saturates with 3 ML dielectric, shifts Si nano-volume to n-type (SiO2 -coated) or p-type (Si3N4 -coated) # Undoped CMOS-compatible Si-MISFETs, on-chip photodetectors, etc.

Alternative 3: SiO2 (& Si3N4) modulation doping

# CMOS compatible without size limit, doping of oxide trench next to Si-NV # Highly carrier selective and passivated contacts for HIT solar cells

Alternative 2: Donor modulation doping of AlxGa1-xN as adjacent barrier

by excess Si from Si-rich Si3N4 (or Ge from Ge-rich Ge3N4 layer)[3]

Pearson, Bardeen, Phys. Rev. 75, 865 (1949): Semiconductor-metal

transition in Si around dopant densities of 0.3 at-% [1.5 × 1020 cm-3]

APL 103, 133106 (2013): SiOx, x < 1  interconnected a-Si/Si NC network

38

SLIDE 39

Dirk König slide / 39 IMDC and SPREE, Eng. Faculty, UNSW 19 Nov 2015

TEAM

Sebastian Gutsch Daniel Hiller IMTEK, Albert-Ludwigs-University Freiburg, Germany Sample preparation, PL, statistics of APT Michael Wahl Hubert Gnaser Jörg Göttlicher Ralph Steininger Institute of Physics and OPTIMAS, University

f Kaiserslautern,

Germany ANKA synchrotron, Karlsruhe Institute of Technology (KIT), Germany XPS, APT, SIMS XANES spectroscopy

[1] Phys. Rev. 75, 865 (1949) [2] PRB 72, 113303 (2005) [8] JAP 104, 093709 (2008) [13] JAP 115, 034304 (2014) [3] PRB 75, 235304 (2007) [9] PRB 80, 165326 (2009) [14] Chapter 8, Nanotechnology for [4] Nano Lett. 8, 596 (2008) [10] AIP Adv. 3, 012109 (2013) * Photovoltaic Devices, Valenta & [5] PRL 100, 026803 (2008) [11] Nanotech. 24, 275705 (2013) Mirabella (Eds.), Pan Stanford, 2015 [6] PRL 100, 179703 (2008) [12] Adv. Mater. Interfaces 1, [15] Sci. Rep. (Nature) 5, 09702 (2015); [7] PRB 78, 035339 (2008) 201400359 (2014) doi: 10.1038/srep09702 * 39

Kaining Ding IEK-5, Research Center Jülich, Germany Lifetime measurements

Literature: * = open access Thanks for HPC power – Leonardi (EE faculty) and Abacus (IMDC) clusters, both UNSW