Elucidation of the reaction mechanism for the synthesis of ZnGeN2 through Zn2GeO4 ammonolysis [original]

Elucidation of the reaction mechanism for the

synthesis of ZnGeN

through Zn

GeO

ammonolysis†

Zhenyu Wang,

Daniel Fritsch,

Stefan Berendts,

Martin Lerch,

Joachim Breternitz *

and Susan Schorr *

Ternary II–IV–N

materials have been considered as a promising class of materials that combine

photovoltaic performance with earth-abundance and low toxicity. When switching from binary III–V

materials to ternary II–IV–N

materials, further structural complexity is added to the system that may

inﬂuence its optoelectronic properties. Herein, we present a systematic study of the reaction of

GeO

with NH

that produces zinc germanium oxide nitrides, and ultimately approach stoichiometric

ZnGeN

, using a combination of chemical analyses, X-ray powder diﬀraction and DFT calculations.

Elucidating the reaction mechanism as being dominated by Zn and O extrusion at the later reaction

stages, we give an insight into studying structure–property relationships in this emerging class of materials.

Introduction

Most of the promising candidates as alternative solar cell

materials struggle from severe problems concerning elemental

abundance and toxicity. This prominently includes halide

perovskites that predominantly contain toxic lead,

1–3

but also

III–V materials due to the low abundance of In and Ga.

4,5

response to the increasing awareness of these “secondary”

judgement parameters –as opposed to a pure focus on solar cell

eﬃciencies –the problematic cations are being substituted by

abundant elements. The cations in III–V nitride materials are

trivalent, but there is a limited number of trivalent cations they

could be replaced with. Instead, they can be replaced by

a combination of divalent and tetravalent cations,

a similar

rationale that also stands behind the development of lead-free

double perovskites.

7,8

Particularly II–IV–N

nitride materials have moved into the

focus of research as they seemingly full all the criteria as

outlined above. While the binary nitrides AlN, GaN and InN all

crystallise in the hexagonal wurtzite-type structure (space group

mc),

9,10

the situation for the ternary compounds ZnGeN

and

ZnSnN

is more complex. While ZnGeN

is consistently reported

to crystallise in the orthorhombic b-NaFeO

-type structure

(space group Pna2

, Fig. 1), in which the Zn

and Ge

cations

are ordered on diﬀerent crystallographic sites, a variable degree

of cation disorder was observed,

up to the point where full

disorder and a crystal structure in the wurtzite-type has been

observed. As to what concerns ZnSnN

, no compelling experi-

mental evidence for cation ordering has been observed so

far,

12,13

although numerous computational studies unani-

mously identied the b-NaFeO

-type structure as the thermo-

dynamically stable crystal structure, similar to ZnGeN

Fig. 1 Crystal structure representation of ZnGeN

in the b-NaFeO

type.

N: green, Ge: sky blue, Zn: canary yellow; coordination tetra-

hedra are drawn around the cations in the colours of the central

atoms.

Helmholtz-Zentrum Berlin f¨

ur Materialien und Energie GmbH, Department Structure

and Dynamics of Energy Materials, Hahn-Meitner-Platz 1, 14109 Berlin, Germany.

E-mail: joachim.bret[email protected]; susan.schorr@helmholtz-berlin.de

Freie Universit¨

at Berlin, Department Geosciences, Malteserstraße 74-100, 12249

Berlin, Germany

Technische Universit¨

at Berlin, Fakult¨

at II, Institut f¨

ur Chemie, Straße des 17. Juni 135,

10623 Berlin, Germany

Universit¨

at Potsdam, Mathematisch-Naturwissenschaliche Fakult¨

at, Institut f¨

Chemie, Karl-Liebknecht-Straße 24-25, 14476 Potsdam, Germany

†Electronic supplementary information (ESI) available. See DOI:

10.1039/d1sc00328c

Cite this: Chem. Sci.,2021,12, 8493

All publication charges for this article

have been paid for by the Royal Society

of Chemistry

Received 18th January 2021

Accepted 12th May 2021

DOI: 10.1039/d1sc00328c

rsc.li/chemical-science

Chemical

Science

EDGE ARTICLE

Open Access Article. Published on 13 May 2021. Downloaded on 10/27/2021 8:47:00 AM.

This article is licensed under a

Creative Commons Attribution 3.0 Unported Licence.

View Article Online

View Journal

| View Issue

The situation becomes even more complex when taking

oxygen into account: zinc germanium oxide nitrides (ZGON)

exhibit a disordered wurtzite-type structure over a wide range of

chemical compositions.

5,16–19

This is insofar important as traces

of oxygen are present in virtually any nitride material and it is

thus important to disentangle the eﬀect of oxygen on the cation

disorder from exclusive cation disorder very carefully. For this,

we studied the reaction of Zn

GeO

as ternary oxide precursor

with NH

in order to obtain powder samples with a variable

oxygen content. Using a model originally proposed by Bacher

et al.,

we studied the behaviour in the oxygen richer regime in

a previous study, where a distinct separation of two competing

processes was observed: (1) nitrogen inclusion on the one hand

and (2) Zn loss on the other hand. In accordance to Bacher et al.,

this can be formulated in two reaction steps:

GeO

+2yNH

/Zn

GeO

43y

+3yH

O(1)

GeO

+2x/3NH

/Zn

2x

GeO

4x

+x/3Zn

+xH

O(2)

GeO

+(2x/3 + 2y)NH

/Zn

2x

GeO

4x3y

+x/3Zn

+(x+3y)H

Herein, we explore the later stages of the overall reaction

where the compound is approaching the stoichiometric nitride

ZnGeN

. Through a combination of X-ray diﬀraction and

chemical analyses, we are able to clarify the reaction pathway,

which aids to understand the structural and electronic features

of this class of compounds.

Experimental section

Syntheses

The zinc germanium oxide nitrides were synthesised through

a two-step process as outlined earlier.

5,20

Briey, Zn

GeO

was

synthesised from binary ZnO (Fisher Scientic, 99%) and GeO

(ACROS, 99.999%) at high temperatures and was then used as

precursor in an ammonolysis reaction. For this, Zn

GeO

was

reacted in a silica glass reaction tube with 4.5 cm diameter. The

samples were treated under a N

-ux of 1.5 l min

1

during

heating (250 K h

1

), and were reacted under a NH

ux of 0.15

l min

1

(99.8%, AirLiquide) during the dwelling period. Time

and temperature of the dwelling were varied between 10 h–20 h

and 835 C–910 C, respectively, to produce powder samples of

varying oxygen content. The exact reaction conditions of all

samples are given in the ESI.†

Characterisation

X-ray diﬀraction (XRD) data in a 2qrange from 15to 140were

collected using a Bruker D8 Advance powder diﬀractometer

with Cu-K

(l¼1.54056 ˚

A) and Cu-K

(l¼1.54439 ˚

A) radi-

ation. The diﬀraction patterns were analysed using the Rietveld

renement method applying the FullProf Suite 3.0 soware.

The starting model for zinc germanium nitride in the

orthorhombic b-NaFeO

-type structure was taken from Zhang

et al.

In the renement, a linear interpolation background was

used along with a Thompson–Cox–Hastings pseudo-Voigt

prole with anisotropic line broadening correction through

spherical harmonics in the Laue class mmm.

X-ray uorescence spectra (XRF) were collected using

a Bruker M4 Tornado system with Rh-microfocus tube for the

determination of the cation ratios. The tube voltage was set to

50 kV. Samples were pressed to pellets with 5 mm in diameter to

avoid contamination when measuring in vacuum. Further, the

pellets oﬀer a at surface for focusing in order to eliminate an

undesired background. For each pellet, data on 6 diﬀerent

measuring points, at least, were collected with a collection time

of 60 s per point.

Hot-gas extraction method was performed using a LECO TC-

300/EF-300 instrument to determine O and N contents. Samples

of approximately 10 mg were used for each independent

measurement. The average value of three repeated measure-

ments was taken as the nal data with a relative error of 2%.

UV-VIS measurements were performed using a PerkinElmer

LAMBDA 750S with a 100 mm integrating sphere in the range of

1000–250 nm and a step-width of 2 nm. Samples were measured

in diﬀuse reectance with the powders contained in silica glass

cuvettes. The light absorption was estimated from reection

using the Kubelka–Munk function F(R)¼(1 R)

/2R(where Ris

the reectance of the sample).

The optical bandgap was then

extracted using a Tauc-plot with [F(R)hn]

for a direct, allowed

band gap.

Density Functional Theory (DFT) calculations were per-

formed utilising the Vienna ab initio simulations package

(VASP 5.4.4)

26,27

together with the projector-augmented wave

(PAW) method.

The 2 22 supercells of the primitive

orthorhombic unit cell of ZGONs were constructed for the

calculations. The supercells contain 128 atoms: 36 Zn, 28 Ge,

60 N and 4 O, according to the Zn/Ge ratio of 1.28 to reect an

experimentally accessed oxide nitride composition. The

initial structural parameters, including lattice parameters

and atomic positions were taken from the Rietveld rene-

ment results of the X-ray diﬀraction pattern. Given the

crystal structure of this experimentally accessed oxide

nitride that is necessarily disordered, a random distribution

of cations was generated using random shuﬄe

function in

python 3.8, to reect the disordered cations arrangement,

whereas atomic positions remained. Diﬀerent oxygen-

containing supercell models were built by accommodating

all oxygen atoms in either [OZn

], [OZn

]or

[OZn

] tetrahedra. The supercell structures were relaxed

using the PBEsol

functional until the forces on all atoms

were below 0.1 eV ˚

1

. The lattice parameters were xed

during the optimisation of the atomic positions (a¼

11.03078 ˚

A, b¼12.83132 ˚

A, and c¼10.38778 ˚

A) in a G-point

optimisation. A 2 22G-centred k-point mesh was used

for the subsequent total energy calculation using PBEsol.

Further, G-point HSE06 (ref. 31) calculations were performed

for total energies of the relaxed structures in comparison to

the PBEsol values. Other parameters included a 500 eV cut-

Chemical Science Edge Article

Open Access Article. Published on 13 May 2021. Downloaded on 10/27/2021 8:47:00 AM.

This article is licensed under a

Creative Commons Attribution 3.0 Unported Licence.

View Article Online

oﬀenergy for the plane-wave expansion, and a cut-oﬀfor the

total energy convergence of 10

6

eV.

Results

Chemical composition of the samples

It is not straightforward to determine the four elements of zinc

germanium oxide nitrides at the same time with the methods

that are easily accessible. While XRF is most reliable for zinc

and germanium, the quantisation of the oxygen and nitrogen

contents are limited by this method. Instead, oxygen and

nitrogen can be quantied using a combustive hot gas extrac-

tion method, which is specic to these elements and therefore

does not allow the determination of the Zn and Ge contents.

Therefore, both methods had to be combined to determine the

complete composition of the respective samples.

It proved useful to employ the cation and anion ratios in our

analysis, as they are easily accessible from the experimental

characterisation: the XRF measurements yielded the atomic

ratios of Zn and Ge directly, whereas the weight fractions ob-

tained for O and N [w(O) and w(N) resp.] were converted using

the respective molar masses [M(O) and M(N)]:

nðNÞ

nðOÞ¼wðNÞMðOÞ

wðOÞMðNÞ

This latter treatment is facilitating the analysis, as the exact

composition of each individual compound is not known a pri-

ori. The ratio as dened above, however, does not depend on the

knowledge of the exact composition and can hence be directly

calculated on the basis of the experimentally determined values.

A clear trend between the experimentally determined Zn/Ge

ratio and the O/N ratio is evident throughout the samples

(Fig. 2). Combining the Zn/Ge and O/N ratios, it is possible to

calculate the overall composition using the general equation

GeO

4x3y

(Table 1).

The Zn/Ge ratio varies between 1 and 2, which is in line with

the compositions of the boundary compounds: stoichiometric

ZnGeN

with Zn/Ge ¼1 and Zn

GeO

with Zn/Ge ¼2. While the

samples close to Zn/Ge ¼1 also exhibit an O/N ratio near 0 –as

expected for ZnGeN

–the O/N ratio is only at z0.4 for Zn/Ge

reaching to 2. This is in line with our prior ndings that

oxygen richer zinc germanium oxide nitrides preserve a Zn/Ge

ratio close to 2, although they already contain notable

amounts of nitrogen.

The sheer number of parameters that inuence the compo-

sition of the compounds makes a straightforward analysis

diﬃcult. It is, however, evident that the higher the amount of

starting materials is, the higher is the content of oxygen in the

product. Also, the atmosphere under which the sample was

cooled plays an important role in the resulting composition:

when cooling under ammonia, the oxygen amount is consid-

erably lower than cooling under nitrogen ow, which hints that

the reaction continues at lower temperatures during the cooling

period. Also, there is a number of samples with slightly diﬀerent

compositions, although they are made at nominally similar

conditions, which is a clear sign for the complexity of the

reaction and that the reaction conditions need to be controlled

very carefully.

Crystal structure of the zinc germanium oxide nitrides

All of the compositions showed single phase patterns with the

exception of the oxygen poorest sample, where a miniscule

inclusion of elemental Ge

was found, which was also the

reason not to extend the study further to longer reaction times

and/or reaction temperatures. The crystal structure appears to

be diﬀerent for the oxygen-poor samples, which crystallise in

the b-NaFeO

-type structure (space group Pna2

), and for the

oxygen richer samples that appear to crystallise in the wurtzite-

type structure (space group P6

mc). It should be noted that b-

NaFeO

-type structure space group is a subgroup of the

wurtzite-type structure and it is hence possible to perform the

Rietveld renements for all patterns in the lower symmetry b-

NaFeO

-type crystal structure.

The three strongest groups of reections between 30–402q

are most indicative for the transition from the hexagonal

wurtzite-type structure to the orthorhombic b-NaFeO

-type

structure (Fig. 3). The oxygen richer samples exhibit three

reections, in accordance to the hexagonal wurtzite-type struc-

ture. Still, the 10

10 and 1011reections appear more and more

Fig. 2 Experimentally determined Zn/Ge ratios against the experi-

mentally determined O/N ratios. The green line shows a linear ﬁt of the

data points with f(x)¼2.42(7)x+ 1.04(1).

Table 1 Experimentally determined Zn/Ge ratios and O/N ratios of the

samples in this study together with the compositions as calculated

from the formula Zn

2x

GeO

4x3y

. Full synthesis conditions may

be found in the ESI

Sample number Zn/Ge ratio O/N ratio

Nominal

composition

1 1.91(7) 0.36(1) Zn

1.91

GeO

0.75

2.10

2 1.28(8) 0.089(4) Zn

1.28

GeO

0.18

2.07

3 1.13(1) 0.043(1) Zn

1.13

GeO

0.09

2.03

4 1.12(7) 0.017(1) Zn

1.12

GeO

0.03

2.05

5 1.09(6) 0.016(1) Zn

1.09

GeO

0.03

2.04

6 1.06(1) 0.015(1) Zn

1.06

GeO

0.03

2.02

7 1.06(5) 0.011(1) Zn

1.06

GeO

0.02

2.03

Edge Article Chemical Science

Open Access Article. Published on 13 May 2021. Downloaded on 10/27/2021 8:47:00 AM.

This article is licensed under a

Creative Commons Attribution 3.0 Unported Licence.

View Article Online

asymmetric the lower the oxygen content becomes, until a very

clear splitting appears, which is indicative of the b-NaFeO

-type

structure. The 0002 reection is, however, unaﬀected by the

group–subgroup transition, as it does not split (002 reection in

the orthorhombic subgroup). Further, the Rietveld renements

(Fig. 4 and ESI†) were performed using an anisotropic line

broadening correction as the 00lreections appear systemati-

cally narrower than the remaining reections, indicative of an

anisotropic particle size.

Since Zn

and Ge

as well as O

2

and N

3

are formally

isoelectronic, they are hardly distinguishable from each other using

X-ray diﬀraction techniques. Therefore, the Rietveld renements

were mainly performed to conrm the overall crystal structure and

to extract the lattice parameters. While the c-parameter remains

largely unaﬀected over the entire composition range (Fig. 5), aand

bvary signicantly over the composition range. When approaching

stoichiometric ZnGeN

,i.e. an O/N ratio of 0, the a-parameter

shrinks more signicantly than the b-parameter grows leading to an

overall decrease in the unit cell volume. This can be rationalised by

regarding the Shannon radii of the cations: r(Zn

)¼0.6 ˚

Aand

r(Ge

)¼0.39 ˚

A. The oxide nitrides contain a ratio of Zn/Ge that is

above 1, but which reduces to 1 when approaching stoichiometric

ZnGeN

. The share of larger Zn

cations, therefore, shrinks from

oxygen richer oxide nitrides to ZnGeN

and aﬀects the volume in the

same way.

In order to quantify the deviation of the observed crystal

structure from an idealised wurtzite-type structure in a hexag-

onal unit cell, it is useful to compare the lattice parameters

aand bas if they were in a hexagonal setting. According to the

group–subgroup relationship between the hexagonal wurtzite-

type structure and the orthorhombic b-NaFeO

-type struc-

ture,

the orthorhombic distortion may be calculated as (a



)/a

, where a

are the pseudo-hexagonal lattice parameters,

which relate to the orthorhombic lattice parameters as a

¼a

O3 and a

¼b

/2. This value is lies in a range between 0 for an

ideal wurtzite-type structure and 2.27%, which is the value

observed from DFT crystal structure optimisation.

33,34

UV-VIS measurements

A clear trend can be observed between the chemical composi-

tion of the samples (represented by the O/N ratio in Fig. 6) and

the bandgap in that the bandgap tends to decrease with rising

O/N ratio. This relationship is, however, not linear and the

samples with very low oxygen content show very similar optical

bandgaps of z3.4 eV. Further investigations on the potential

impact of cation order/disorder on this trend are currently

ongoing. It is worth mentioning that a general parabolic rela-

tionship between bandgap and O/N ratio was observed in other

recent study focussing on thin lm syntheses of zinc germa-

nium oxide nitrides, which show disorder of the cations.

Finally, an excellent review by Schnepf et al. needs to

Fig. 3 XRD patterns in the region of 30–402qas a function of Zn/Ge

and O/N ratios. The hkl indices according to the hexagonal wurtzite-

type structure (black, bottom) and the hkl indices according to the

orthorhombic b-NaFeO

-type structure (grey, top) are given in the

ﬁgure.

Fig. 4 Plot of powder X-ray diﬀraction proﬁle reﬁnement with the

Rietveld method: measured intensities (black circles), calculated

proﬁle (red line), diﬀerence between measured and calculated inten-

sities (blue line) and calculated reﬂection positions (black ticks).

Fig. 5 Change of the lattice parameters a(red), b(blue) and c(green)

as a function of the O/N ratio.

Chemical Science Edge Article

Open Access Article. Published on 13 May 2021. Downloaded on 10/27/2021 8:47:00 AM.

This article is licensed under a

Creative Commons Attribution 3.0 Unported Licence.

View Article Online

mentioned, focussing on the inuence of cation disorder on the

band gaps in this class of materials.

DFT calculations

To understand the reaction mechanism, we performed DFT

calculations to probe, whether the arrangement of anions and

cations are dependent of each other. For this, we compare the

total energies of diﬀerent arrangements within the same overall

composition with each other (Fig. 7). The composition chosen

reects the composition at a Zn/Ge ratio of 1.28 and hence lies

within the experimentally accessed parameter space.

It is very evident from the total energies of the diﬀerent

arrangements, that oxygen prefers to be surrounded by Zn

rather than Ge. The total energies for the supercell containing

uniquely Zn coordinated O is consistently the lowest

throughout the functionals tested. The energy diﬀerence

between the [OZn

] and the [OZn

] coordinations of 125.5

meV per f.u. (HSE06) would correspond to a thermal activation

temperature of 1456 K, which is considerably above the reaction

temperatures used in this study.

Discussion

Elucidation of the reaction mechanism

As developed by Bacher,

one can easily separate the overall

reaction into the part-reactions (1) nitrogen inclusion and (2) Zn

loss. As they both inuence the composition of the product

2x

GeO

4x3y

in a diﬀerent way –(1) acting upon yand (2)

acting upon x–one can disseminate the contribution of each

partial reaction from the chemical composition of the product.

This does, in fact, allow to elucidate the reaction mechanism

over the course of the reaction conditions studied and hence

allow a more targeted synthesis approach. For this, it is useful to

use the ratio of the parameters xand yas an indicator for the

relative contribution of both part-reactions. It is evident that

there is a very clear relationship between the x/yratio and the

Zn/Ge ratio (Fig. 8) with a general formula of the linear tofx/y

¼1.97(1) 0.99(1)Zn/Ge. Since the relationship x¼2Zn/Ge

is dened through the overall formula and taking the linear t

of the relationship between x/yand Zn/Ge into account, this

essentially means that yz1, throughout the whole composi-

tion range in this study. This in turn means, that the nitrogen

inclusion part reaction is essentially nished and only Zn loss

proceeds through the reaction period. This essentially means

that the simplied formula Zn

2x

GeN

1x

can be used as

a convenient rst approximation to the accurate chemical

composition, which can be determined on the basis of the Zn/

Ge ratio.

It is important to put this nding in the perspective of the

overall reaction: while we studied the later stages of the reaction

herein, and for which these results are valid, the earlier stages of

the reaction do not necessarily follow the same scheme. In

a previous study,

we produced zinc germanium oxide nitrides

Fig. 6 Optical bandgap as a function of the measured O/N ratio.

Fig. 7 Diﬀerences of the total energies per formula unit of the

diﬀerent oxide nitride supercells as a function of the oxygen coordi-

nation environment. The values for 2 22k-point grid PBEsol (red)

and gamma-point HSE06 calculations (green) are given relative to the

energy of the crystal structure, where oxygen is uniquely surrounded

by Zn. The tetrahedra depict one representative tetrahedron for each

coordination environment (O: bordeaux, Zn: canary yellow, Ge: sky

blue).

Fig. 8 x/yratio as a function of the Zn/Ge ratio throughout the

composition range. The green line depicts a linear ﬁt of the data.

Edge Article Chemical Science

Open Access Article. Published on 13 May 2021. Downloaded on 10/27/2021 8:47:00 AM.

This article is licensed under a

Creative Commons Attribution 3.0 Unported Licence.

View Article Online

Loading more pages...