Net Electron Number as an effective Electronic Descriptor

for Oxygen Carrier Performances of SrFeO3-based Oxygen

Carriers: A DFT Study

Xijun Wang, Emily Krzystowczyk, Jan Dou, Fanxing Li*

North Carolina State University

ACS Spring 2020 National Meeting

1
Outline

▪ Background introduction

▪ Vacancy formation energy calculations and experimental verification

▪ Mechanism study from the viewpoint of electronic structures

▪ Extend the application of net electron number (∆e) to more doped systems

▪ Conclusions and future work

2
 Background
Chemical Looping N2

Biomass & Fossil Hydrocarbons Clean Energy Carriers & Chemicals

1. Oxygen carrier performance is one of the most critical factors affecting commercial viability.
2. Introducing dopants is a common way to significantly improve the performance of perovskites.
3. However, the inner mechanism remains unclear.
4. By using DFT simulations, here we aim at revealing the inner mechanism of doping effect.

3
 Hohenberg-Kohn Theorems
𝜕
෡ 𝑟, 𝑡
𝑖ℏ Ψ 𝑟, 𝑡 = 𝐻Ψ
𝜕𝑡
Kohn-Sham Equation
𝐷𝐹𝑇

Key settings:
1. Functional/Pseudo-potential: PBE-PAW
2. Ferromagnetic state are applied for all the species.
3. GGA+U approach is used to correct the strong on-site coulomb interactions.

DFT provides a new way toward rational and efficient design and
optimization of high-performance sorbents. 4
Outline

▪ Background introduction

▪ Vacancy formation energy calculations and experimental verifications

▪ Mechanism study from the viewpoint of electronic structures

▪ Extend the application of net electron number (∆e) to more doped systems

▪ Conclusions and future work

5
 Method Verification
DFT 1st Vacancy formation energy vs. Experimental initial decomposition T
1
∆𝐸𝑉 = 𝐸𝑉 + 𝐸𝑂2 − 𝐸𝑝𝑒𝑟𝑓𝑒𝑐𝑡
2

DFT ΔEV: Cu < Co < K < Ni < Ba < Mn < undoped < La
Exp initial decomposition T: Cu < Co < K < Ni < Ba < Mn < La
A smaller ΔEV suggests easier oxygen emission, corresponding to lower initial decomposition temperature.
The 1st vacancy formation energy is a good indicator for the initial decomposition temperature.
Our DFT simulation methods and settings can be verified by the experiments.
6
Computed ΔEV VS. Experimental oxygen capacity

Due to the relatively high and low oxygen capacity as suggested by the
experiments, here we chose Cu and Y to represent the good and bad
samples for more detailed electronic structural calculations.
All oxygen vacancy formation energies increase with the increase in δ. The
ΔEV values of Cu doped samples are always smaller than the undoped ones,
while the ΔEV of Y doped samples are comparable with the undoped ones at
first and then go beyond the undoped ones as δ increases. These relative
trends are in consist with the experimental oxygen capacities.

7
Outline

▪ Background introduction

▪ Vacancy formation energy calculations and experimental verifications

▪ Mechanism study from the viewpoint of electronic structures

▪ Extend the application of net electron number (∆e) to more doped systems

▪ Conclusions and future work

8
 Charge distribution
(a) (b)
Firstly, we compared the charge donation by Cu
and Y dopant atoms (Figure a). Results show
that the doped Cu atom donates less electrons
than Fe in pristine SrFeO3, indicating a p-type
doping to the system. Therefore the number of
accepted electrons per O atom will decrease
(Figure b).

In contrast, Y donates more electrons to oxygen

and make the oxygen atoms more negatively
(c) (d) charged, providing a n-type doping. As a result,
for a given vacancy concentration (δ), the
number of electrons on O atoms follows a trend
of Y doped > Undoped > Cu doped.

The charge donation can be visually shown in

Figure c and d, where cyan and yellow bubbles
represent holes and electrons respectively.

Cu doped (p-type doping) Y doped (n-type doping) 9

Relationship between accepted electron numbers vs. vacancy formation energy

Further, we found that the number of accepted electrons

per O atom has a linear relationship with the vacancy
formation energy, suggesting that the less negatively
charged of O atoms, the lower the vacancy formation
energy.
To explain this finding, we move on to study the electronic
structures of the doped samples (see next slide).

10
 Why p- or n-type doping can influence vacancy formation energy?
Cu doped Undoped Y doped The p-band center (εp) can be defined as:
εp = -2.38 eV εp = -2.48 eV εp = -2.64 eV ‫ 𝐸 𝐷 ∙ 𝐸 ׬‬− 𝐸𝐹 𝑑𝐸
ε𝑝 =
Projected density of state (PDOS)

‫ 𝐸 𝐷 ׬‬− 𝐸𝐹 𝑑𝐸

εp = -2.47 eV εp = -2.57 eV εp = -2.71 eV

Total energy of Number of O

εp = -2.53 eV εp = -2.55 eV εp = -2.73 eV all O p-orbital p-orbital
electrons electrons

εp = -2.73 eV εp = -2.71 eV εp = -2.92 eV

εp = -2.76 eV εp = -2.91 eV εp = -3.27 eV

Different charge injection (p- or n-type doping) could have impact on the electronic structure
of the systems. For a given δ, Cu doping leads to less electron donation, making the Fermi
level down-shift. This make the p-band closer to the Fermi level and therefore increase the p-
band center. On the contrary, Y doping provide more electrons and thus up-shift the Fermi
level, lowing down the p-band center.
The relationship between Δe and εp should follow an inverse proportional trend. Therefore, p-
band center can be tuned by controlling n- or p- type doping.
11
 Why p- or n-type doping can influence vacancy formation energy?
n-type doping Discussion on the p-band model:

With p-type doping, the system will lose electrons and thereby
down-shift the Fermi level, making the p-band closer to the
Fermi level. In this case:
1. From the viewpoint of electronic orbitals coupling: the
hybridization of O 2p and transition metal 3d orbitals could
be enhanced. As a consequence, the charge transfer gap
between O 2p and metal 3d orbitals should be narrow, and
the activity of lattice oxygen atoms can be enhanced.

2. From the viewpoint of valence-bond theory: with less

electron density in the system, the strength of ionic bonds
p-type doping formed by metals and oxygen will be weakened.

12
Outline

▪ Background introduction

▪ Vacancy formation energy calculations and experimental verification

▪ Mechanism study from the viewpoint of electronic structures

▪ Extend the application of net electron number (∆e) to more doped systems

▪ Conclusions and future work

13
Extend the application of ∆e to more doped systems

Inspired from the Cu and Y doped samples, the

accepted electron numbers of O atoms (Δe) could be
a good descriptor for the oxygen vacancy formation
energy, and may also be able to work in other
elements doped systems.

To verify this assumption, here we calculated the

relation between Δe and ΔEV for a series of doped
samples (including La, Mn, Y, Ba, Ni, K, Co, Cu
dopants) with different δ values (δ = 0, 0.125, 0.25,
0.375). It is revealed that Δe is almost linearly
proportional to ΔEV, indicating that with more p-type
doping, oxygen atoms can be more readily to
release.

14
 Using ∆e to fit the experimental oxygen capacity

Using perovskite models, the prediction errors of oxygen capacity using ∆e is relatively large (PCC=0.91), which is caused
by the ignorance of phase transition due to limitation of the perovskite models we used. Therefore, calculations based on
brownmillerite structures of these species were also performed. The computed ∆e and ∆EV are both inversely proportional to
the oxygen capacity, with very high PCC of 0.93 and 0.91 respectively. These findings verify the feasibility of ∆e as a
descriptor for experimental oxygen capacity.

15
 Vacancy migration barrier vs. ΔEV, Δe, εp and Volume

δ=0.5 δ=0.5

δ=0.125, 0.25, 0.375 δ=0.125, 0.25, 0.375

To build a bridge between the migration barriers and the electronic structures, we tried to fit the barriers using Δe and εp. From
the Figure above we obtained two linear relationships between Δe and the barriers at low δ (0.125, 0.25, 0.375) and high δ
(0.5) respectively. This may originate from the maybe existing phase transition from perovskite to brownmillerite, which is
hard to be considered at this stage due to the limitation of our models. Likewise, εp and the barriers also have two different
correspondences at low and high δ. It is found that εp shows better performances in describing the barriers, with extremely
high PCCs of 0.94 and 0.99 at low and high δ respectively, which are both slightly larger than Δe (0.88 and 0.98). These
demonstrate the superior performance of Δe and εp in describing the barriers of oxygen migration, opening a window for the
study of oxygen carrier performances based on electronic descriptors.

16
Outline

▪ Background introduction

▪ Vacancy formation energy calculations and experimental verification

▪ Mechanism study from the viewpoint of electronic structures

▪ Extend the application of net electron number (∆e) to more doped systems

▪ Conclusions and future work

17
 Summary

In summary, we propose the average net number of electrons of oxygen anions (Δe) as an effective
descriptor for oxygen formation, oxygen capacity and migration. Compared with the widely used p-band
center (εp), Δe exhibits higher correlation with the vacancy formation energy and slightly lower but still
very high correlation with the oxygen migration barrier. Considering its experimental measurability and
controllability, Δe shows great potential to be a promising descriptor to assist the rational design and
optimization of high-performance oxygen sorbents for chemical looping air separation. These new insights
from the case of SrFeO3-based perovskites can be helpful for the study of other oxygen-deficient correlated
oxides.

18
Thank you!

19