{FeNO}7-Type Halogenido Nitrosyl Ferrates: Syntheses, Bonding, and Photoinduced Linkage Isomerism

Article abstract of DOI:10.1002/chem.201804565

Mononitrosyl–iron compounds (MNICs) of the Enemark–Feltham {FeNO}⁷ type can be divided into a doublet (S=1/2) and a quartet (S=3/2) spin variant. The latter relies on weak-field co-ligands such as amine carboxylates. Aqua-only co-ligation appears to exist in the long-known “brown-ring” [Fe(H₂O)₅(NO)]²⁺ cation, which was prepared originally from ferrous salts and NO in sulfuric acid. A chloride variant of this species, the green [FeCl₃(NO)]^? ion, was first prepared analoguosly by using hydrochloric instead of sulfuric acid. As a tetrahedral species, it is the simple prototype of sulfur-bonded {FeNO}⁷ (S=3/2) MNICs of biological significance. Although it has been investigated for more than a century, neither clean preparative routes nor reliable structural parameters were available for the [FeCl₃(NO)]^? ion and related species such as the [FeCl₂(NO)₂]^? ion, a prototypical dinitrosyliron species (a “DNIC”). In this work, both issues have been resolved. In addition, we report on a computational study on the ground- and excited-state properties including an assignment of the chromophoric transitions. Photoinduced metastable isomers were characterised in a combined experimental and computational approach that resulted in the confirmation of a single photoinduced linkage isomer of the paramagnetic nitrosyl–metal coordination entity.

Full text of DOI:10.1002/chem.201804565

A Journal of
Accepted Article
Title: {FeNO}⁷-Type Halogenido Nitrosyl Ferrates: Syntheses, Bonding,
and Photoinduced Linkage Isomerism
Authors: Areenan In-Iam, Markus Wolf, Claudia Wilfer, Dominik
Schaniel, Theo Woike, and Peter Kluefers
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Chem. Eur. J. 10.1002/chem.201804565
Link to VoR: http://dx.doi.org/10.1002/chem.201804565
Supported by

10.1002/chem.201804565
Chemistry - A European Journal
1
{FeNO}7-Type Halogenido Nitrosyl Ferrates: Syntheses,
Bonding, and Photoinduced Linkage Isomerism
Areenan In-Iam, Markus Wolf, Claudia Wilfer, Dominik Schaniel, Theo Woike, Peter Klüfers*
Abstract
Mononitrosyliron compounds (MNICs) of the Enemark-Feltham {FeNO}7 type divide into a dublet (S
= 1/2) and a quartet (S = 3/2) spin variant. The latter relies on weak-field coligands such as
aminecarboxylates. Aqua-only co-ligation appears to exist in the long-known ‘brown-ring’ [Fe(H2O)5(NO)]2+
cation which was prepared originally from ferrous salts and NO in sulfuric acid. A chloride variant of this
species, the green [FeCl3(NO)]− ion, was first prepared analoguosly by using hydrochloric instead of sulfuric
acid. As a tetrahedral species, it is the simple prototype of sulfur-bonded {FeNO}7(S = 3/2) MNICs of
biological significance. Though being investigated for more than a century, neither clean preparative routes
nor reliable structural parameters were available for the [FeCl3(NO)]− ion and related species such as the
[FeCl2(NO)2]− ion, a prototypical dinitrosyliron species (a ‘DNIC’). In this work, both issues have been
resolved. In addition, we report on a computational study on the ground- and excited-state properties
including an assignment of the chromophoric transitions. Photoinduced metastable isomers were
characterised in a combined experimental and computational approach which resulted in the confirmation of
a single photoinduced linkage isomer of the paramagnetic nitrosyl–metal coordination entity.
Introduction
Breaking the Fe–NO bond: NO release vs. linkage isomerism
Due to their biological significance in nitric-oxide metabolism, nitrosyl–iron complexes, both
naturally occuring species as well as artificial agents, are a focus of current research. The infusion of an
aqueous solution of pentacyanidonitrosylferrate(2−) dihydrate, Na2[Fe(CN)5(NO)]·2H2O, has been a long-
known method to administer the vasodilator NO during surgery. Besides the medical application, the solid
dihydrate (‘SNP’ from the salt’s common name ‘sodium nitroprusside’) has attracted much interest since half
a century ago, when long-lived metastable states had been produced by irradiation with green laser light at
low temperatures. Two decades after their discovery, the formation of linkage isomers of the nitrosyl ligand
was detected by means of photocrystallography as the origin of the phenomenon.[1] Specifically, two
metastable isomers formed in the course of the relaxation of the electronically excited states, namely a κO-
bonded isonitrosyl (MS1), and a κ2N,O-bonded side-on nitrosyl (MS2). Photoinduced linkage isomerism
(PLI) is neither restricted to the low-temperature regime of metastability nor to the solid state. Short-lived
linkage isomers of SNP have been detected at room temperature in the solid as well as in aqueous solution.[2]
Consequently, PLI or, more generally, linkage isomerism, was also considered to play a role in physiological
NO metabolism.[3]
There is a notable point which specifically applies to the physiological Fe(NO) species. Most currently
known PLI-active systems, either based on iron or another central metal, have diamagnetic ground states.
SNP is an example. A singlet, low-spin-d6 FeII centre bonds to a 1NO+ ligand via the classical donor–acceptor
bonding scheme with the NO-releasing property of SNP solutions going along with the limited back-bonding
capability of an iron centre of the rather high oxidation state II. A rare exception to this experience was the
report by Chen et al. who succeeded in the population of an MS1 state for the paramagnetic nitrosyl complex
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
2
of a heme model complex, a member of the {FeNO}7(S = 1/2) subclass (see the following subsection for the
nomenclature).[4]
An obvious reason for the preference for singlets is not known. Perhaps we may assume that the
metastable isomers are protected by activation barriers along spin-forbidden relaxation paths.[2d] (Note, in
this context, the interesting case of coupling a PLI and a spin-crossover event, where a low-spin, singlet
ground state was photo-excited to a high-spin isomer.[5])
{FeNO}7 species
In our own work, we experienced the lack of PLI phenomena in one subclass of mononitrosyliron
compounds of the {FeNO}7 type (in this Enemark–Feltham notation, read “7” as the sum of metal-d and NO-
π* electrons, i.e., 7 + 0 for FeI + NO+, or 6 + 1 for FeII + NO, or 5 + 2 for FeIII + NO−). {FeNO}7 centres are
found with an S = ½ ground state, in heme–NO adducts for example (see above), or in the S = 3/2 spin state.
The latter spin state occurs with weak-field co-ligands such as aminecarboxylates, or, in a biochemical
environment, with sulfur-derived ligands such as cysteinate or ferredoxin-bound sulfide (see the next
subsection). Recently, we published a number of crystalline aminecarboxylate-coligated octahedral
{FeNO}7(S = 3/2) complexes where we attempted to populate photoinduced isomers in those solids in the
typical temperature and excitation-energy range which yielded metastable isomers in the case of SNP.[6]
However, we did not detect isomers in terms of IR spectra. For a closer approach to the problem, we
continued the search for PLI-active species at lower temperatures. As a starting point for this investigation,
we chose members of the class of tetracoordinate nitrosyl–iron compounds – a class that had attracted
particularly long-standing interest in the history of chemical science.
MNICs and DNICs
Tetracoordinate mononitrosyl iron compounds (‘MNICs’) as well as their dinitrosyl homologues
(‘DNICs’) have been a topic of research for decades. Currently, the focus lies on thiolato- and sulfido-ligated
nitrosyl-iron species, due to their role in signalling and metabolic pathways.[7] Examples include thiolato-
supported MNICs and DNICs with alkyl- and arylthiolates as well as the proteinogenic RS− donor
cysteinate.[7b] The replacement of sulfido ligands on the exposure of [2Fe-2S] clusters to nitric oxide in the
presence of chloride has recently provided one more route to the longest-known members of the MNIC and
DNIC class: the halogenido nitrosyl iron compounds.[7b] With their simple formula, monoanions such as
[FeCl3(NO)]− and [FeCl2(NO)2]− may be regarded as parent species of the MNIC and DNIC type,
respectively. The long-standing interest that these species have attracted is documented in over a century of
publications on the mononitrosyls, from Kohlschütter’s synthetic route in aqueous acid in 1904 to van
Eldik’s preparation in ionic liquids in 2015.[8] Throughout this entire period, however, a constant feature of
this chemistry has impeded the efforts of the various groups to safely detect those species, namely the lack of
reliable basic information on isolated, pure, and well-characterised salts of those anions – though various
attempts have been made.
History: ‘brown-ring’ variations
In a first period of research, the competing groups of Kohlschütter and Manchot contended for the
clarification on the constitution to be assigned to the ‘brown-ring’ species of nitrate and nitrite analysis, now
usually formulated as [Fe(H2O)5(NO)]2+. In 1904 and 1907, Kohlschütter published on a green, anionic
variant of the tentative nitrosyl species that he had observed on NO absorption in a ferrous chloride solution
in concentrated hydrochloric acid. With his formulation of the green species as the anion of a
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
3
‘Ferrochlorwasserstoffsäure’ [chloridoiron(II) acid], Kohlschütter was one of the first to use the emerging
ideas of coordination chemistry.[8a, 9] In 1911, Kohlschütter summarised the state of knowledge and
formulated the green species as [FeCl2+x(NO)]′x (Kohlschütter was not able to isolate crystalline substances;
hence, lacking elemental analyses, he could only speculate on x which he guessed was two in an attempt to
interpret the formation of an anionic species in the electrolysis of an FeCl2/NO solution in ethanol).[10] In the
same period of time, Manchot prepared a green anion by the reaction of ferric chloride and nitric oxide in
organic solvents such as ethanol.[11] He stated the identity of the green products from both ferrous and ferric
chloride, and he identified nitric oxide as the reductant, but he could not give a full reaction equation (he
suspected complicated side reactions of the organic part).
UV-vis spectroscopy was available some two decades later,[12] and, another three decades later, the S =
3/2 ground state of the related aqua species was determined.[13] Remarkably, in the latter work, Griffith et al.
reported on a (non-isolated), green(!) ferric analogue of the tentative formula [Fe(EtOH)5(NO)]3+ from the
reaction of ferric chloride in ethanol with NO gas. Hence, the authors de facto reproduced Manchot’s
experiment but arrived at different conclusions about the formula of the green chromophore. A motif might
have been that the magnetic moment of the species exceeded the expected value for a high-spin {FeNO}7
compound. It should be noted, however, that the reported value for the N–O stretch (1775 cm−1 in EtOH
solution) was consistent with later reported values for the [FeCl3(NO)]− ion (compare 1798 cm−1 in a solution
+
with the NEt4 counterion in MeOH).
In the 1960s, the arrival of ESR spectroscopy restarted the investigation of the reaction of ferrous salts
including the halogenides with NO (‘restarted’ meaning that the authors in this period ignored Kohlschütter’s
work).[14] Synthetic protocols included the use of methanol as the solvent and, mostly, the addition of a base
such as triethylamine.[15] As a result, the respective chlorido-DNIC and an alkyl nitrite formed in solution
instead of an MNIC, tentatively according to the equation:
FeCl2 + 3 NO + MeOH + NEt3 → NHEt3[FeCl2(NO)2] + MeONO
The bromido-DNIC was prepared in a similar procedure.[16] A specification is necessary at this point.
This work includes DNICs that were formed in competition with {FeNO}7(S = 3/2) MNICs. In Enemark-
Feltham notation, these belong to the {Fe(NO)2}9 class with a total spin of S = ½.
In the subsequent organometallic era, work on halogenido-nitrosyl ferrates emerged that did not refer
to all these prior findings but looked at the green variant as an analogue of halogenido-carbonyl species.
Following standard organometallic procedures, aprotic routes were thus used to prepare various halogenido-
nitrosyl ferrates. In 1976, the Connelly group reported the preparation and the IR-spectroscopic
characterisation of a green PPN[FeCl3(NO)] which formed on an oxidative route by the reaction of
PPN[Fe(CO)3(NO)] with Cl2 in dichloromethane [PPN = bis(triphenylphosphaneiminium)] via
PPN[Fe(CO)3(NO)] + 1.5 Cl2 → PPN[FeCl3(NO)] + 3 CO
However, due to the strongly oxidising chlorine reagent, the authors described unexpected magnetic
moments of their products which they attributed to ferric by-products. The carbonylferrate educt was
transformed by the heavier halogens in similar reactions to the bromido-MNIC/DNIC as well as to the
iodido-DNIC.[17]
In 1983, the Beck group reported the crystal structure of an AsPh4[FeCl3(NO)] which had formed on
their attempts to prepare and isolate a salt of the DNIC homologue [FeCl2(NO)2]− by reacting Roussin’s red
salt with HCl gas in pentane. Unexpected by the authors, the isolated crystals turned out to be the trichlorido-
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
4
mononitrosyl ferrate in the course of the structure determination by means of Weissenberg film techniques.
Irritatingly, not a green, but a DNIC-typical red-brown colour of the crystals was reported. More recently,
Akutsu et al. repeated Beck’s work with some modifications and isolated crystals of the claimed formula
PPh4[FeCl2(NO)2] whose anion is the one that Beck attempted to prepare. The characterisation of the
needles, which were reported as isostructural to Beck’s, however, was ambiguous: the crystals were brown
(DNIC), the X-ray analysis revealed Cl/NO disorder, and the magnetism showed parameters closer to the
MNIC rather than the attempted DNIC.[18] Coincidental with Akutsu’s work, the Lippard group published a
crystal-structure determination on a cationic MNIC with the [FeCl3(NO)]− ion as the counterion stemming
from the dismutation of a FeCl2 precursor. Also in that work, it is the halogenido-MNIC that caused some
ambiguity in terms of an unresolved residual electron density of 2.5 e Å−3 within the anion.[19] Eventually, the
most recent crystal-structure analysis that included the [FeCl3(NO)]− ion revealed disorder in this species.[20]
While no isolated fluorido-nitrosyl-iron compounds have been published, some data are available on
bromido and iodido derivatives. In particular, an X-ray structure analysis on single crystals of the iodido-
DNIC is known, namely PPN[FeI2(NO)2].[21] No structural information has been published on A[FeBr3(NO)]
and A[FeBr2(NO)2] salts.
For the focus of this work, the [FeCl3(NO)]− ion, the preparative situation had to be clarified first. In
fact, the literature provides no reliable route for the preparation of an isolated salt of this prototypical nitrosyl
complex: Connelly’s and Gardner’s oxidation of carbonyl precursors by Cl2 resulted in FeIII-contaminated
samples. The more direct methods that start with simple ferrous salts in protic media including water seem to
be impeded by the need to match the stability fields of MNICs and DNICs. Thus, chlorido-MNICs may be
captured in acidic media, whereas chlorido-DNICs are preferentially formed in less acidic to basic solution.
This rule of thumb becomes obvious when comparing the reaction of ferrous chloride with NO gas yielding
[FeCl3(NO)]− in hydrochloric acid on the one hand, and the formation of [FeCl2(NO)2]− in
methanol/trialkylamine instead of HCl(aq) on the other. In the light of stability fields, the problems
encountered in Beck’s approach to chlorido-DNICs via the hydrolysis of thio-DNICs with hydrochloric acid
become tangible. (The rules apply to the chlorido systems; bromide and iodide form DNICs more easily.)
Thus, the focus of this work lies on the preparation of pure, crystalline halogenido-nitrosyl ferrates as a
prerequisite for reliable structure analyses and physical measurements on well-defined members of this
rather fundamental substance class.
Our report is organised as follows: (1) The aptitude of the oldest route by Kohlschütter for the
preparation of pure trichloridonitrosylferrate(II) from FeCl2 and NO, free of iron(III) and dinitrosylferrates,
is demonstrated. The route is modified to yield a crystalline salt. (2) The solvent is changed from water or
hydrochloric acid to methanol and the preparation of salts with a variety of cations is employed. (3) Ferric
salts are demonstrated to yield the [FeCl3(NO)]− ion as well but ferric contaminations of the product crystals
have to be monitored carefully. (4) The shift of the isolated product from the mononitrosyl to the dinitrosyl
species, if the reaction is conducted in the presence of base, is investigated. In this part, NO/halogenide
disorder is analysed as well. (5) Since the chlorido-nitrosyl ferrates are particularly small entities,
computational methods in both the frameworks of density functional and wave-function theory were
employed to analyse bonding and spectra of these species. (6) Two types of photoexcited metastable MNIC
products were produced and characterised by their N–O valence frequency: the isonitrosyl (MS1) isomer and
the neutral FeCl3(NO) photooxidation product. The lack of an MS2 isomer will be rationalised by a DFT
analysis. (6) Analogies in the DNIC field are carved out in a more summarising way.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
5
Results and Discussion
Chemistry: preparative routes and route-dependent by-products
History I (1904–1911): Pure green AsPh4[FeCl3(NO)] from Kohlschütter’s aqueous route[8a, 9-10]
In the first decade of the 20th century, Kohlschütter obtained green solutions from the reaction of
ferrous chloride and nitric oxide in hydrochloric acid.[8a, 9] This earliest investigator, however, did not succeed
in synthesising crystals of his green variant of the brown-ring species which he obtained by the replacement
of sulfuric by hydrochloric acid in the nitrosation of ferrous solutions. In fact, crystals can be obtained with a
bulky counterion by carefully balancing educt and product solubilities. Hence, monoclinic green crystals of
AsPh4[FeCl3(NO)] were obtained from solutions of ferrous chloride and tetraphenylarsonium chloride in
hydrochloric acid/methanol on the slow addition of nitric oxide to a nitrogen atmosphere above the liquid
without stirring. The green chromophore, the anion, is depicted in Figure 1. It resembles Kohlschütter’s
[FeCl2+x(NO)]′x formulation for x = 1.[10] Moreover, it resembles Connelly’s anion in terms of colour and IR
data, and Beck’s red-brown anion in terms of structure.
O1
N1
Cl2
Fe1
Cl3
Cl1
Figure 1. The structure of the anion in monoclinic crystals of AsPh4[FeCl3(NO)] (50% ellipsoid probability).
Distances (in Å) from Fe1 to: Cl1 2.232(1), Cl2 2.229(1), Cl3 2.233(1), N1 1.730(4); N1–O1 1.142(5). Angles
(in degrees): Fe1–N1–O1 171.6(4), Cl1–Fe1–N1 108.5(1), Cl2–Fe1–N1 105.0(1), Cl3–Fe1–N1 112.2(1); mean
Cl–Fe1–Cl 110.4.
A closer look at the crystal structure shows traces of pseudosymmetry. If the small offset of the cation
and anion centres from the 21 axis is removed, the monoclinic angle (90.5°) is set to 90°, and the a and c axis
lengths (12.89 and 13.28 Å) are set as equal, an idealised tetragonal structure results in the supergroup P4/n
(ctetragonal = ½ bmonoclinic) of the actual space group P21/n. We will use this information later in the discussion of
Beck’s structure. At this point, a modified Kohlschütter route appeared to be suitable for the preparation of
pure trichlorido-nitrosyl-ferrates.
History II (1910 and 1958): Manchot’s and Griffith’s green solutions starting from ferric chloride[11, 13]
Both the green colour and the IR spectrum led us to the educated guess that Griffith et al. had the same
product in their hands that had been prepared by Manchot half a century earlier using exactly the same route,
namely the [FeCl3(NO)]− ion. Since the only candidate for the required reductant was, as was already
assumed by Manchot, nitric oxide, the reaction would have to proceed according to the equation:
FeCl3 + 2NO → [FeCl3(NO)]− + ‘NO+’
‘NO+’ had to be replaced by a respective secondary product; in the case of EtOH as the solvent, ethyl
nitrite formation might be considered. It should be noted that the iron(III) reduction by NO under alkyl nitrite
formation was recently demonstrated by the Lehnert group in the course of DNIC formation.[22] Under
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
6
Griffith’s reaction conditions, however, residual ferric species seemed to contaminate the tentative ferrous
nitrosyl complex in terms of a magnetic moment reported that was too high for the pure product. Attempts to
reproduce the experiments confirmed the reported observations. Since the entire reaction path is better
understood after a discussion of our attempts to reproduce the remaining historic routes, the iron(III) issue
will be re-addressed and clarified below.
History III (1976): Connelly’s green, iron(III)-contaminated PPN[FeX3(NO)] solids (X = Cl, Br)[17]
Connelly and Gardner prepared PPN[FeCl3(NO)], a dark green solid, by the reaction of
PPN[Fe(CO)3NO] with chlorine in dichloromethane.[17] The reported values, an S = 3/2 state for the anion as
well as an N–O stretch at 1802 cm−1, suggest the identity of the anion with the one obtained by the
Kohlschütter procedure. Since no crystal-structure analysis was available, we reproduced Connelly’s recipe
and crystallised the green product. As a result, the expected anion was the major disorder species of the green
solid. In addition to the major NO site, however, an approximately 13%, minor disorder form was detected
where the NO ligand was replaced by a chlorine atom. The crystallographic result indicated that the co-
crystallisation of the [FeCl3(NO)]− and the oxidised [FeCl4]− species had taken place (for details of the X-ray
analysis of contaminated samples see Figure 5 and the text referring to the figure).
The formation of the tetrachloridoferrate(III) side product thus appeared as a drawback of Connelly’s
oxidative route. It should be noted that the amount of side product in our structure analysis corresponded to
the result of Connelly’s elemental analysis that showed a nitrogen deficiency of about 10%. A specific point
that is re-addressed below should be noted here: recrystallisation does not appear to be a reliable method for
the separation of equally charged nitrosyl and chlorido species. Instead, nitrosyl and chlorido ligands appear
to tend towards mutual replacement. Due to the small amount of tetrachloridoferrate(III) in the investigated
crystal, however, the parameters of the major species were sensible and were close to those collected in the
caption of Figure 1.
History IV (1983): Beck’s red-brown, DNIC-containing AsPh4[FeCl3(NO)][23]
To give an explanation for Beck’s result that AsPh4[FeCl3(NO)] should form dark red-brown instead of
green crystals was more challenging. Attempts to reproduce Beck’s recipe confirmed the published results.
Just as the Beck group, we obtained a dark red-brown solid that showed the bulk characteristics of the
attempted AsPh4[FeCl2(NO)2].[24] The solid was uniform in terms of colour. Moreover, two isolated, dark red-
brown crystals showed the reported tetragonal unit cell.[23] The structure determination ended as the
published one with large thermal ellipsoids for almost all atoms, in particular for two of the chlorine atoms.
Attempts to establish an NO/Cl disorder refinement to model the admixture of the monoanionic {Fe(NO)2}9-
DNIC [FeCl2(NO)2]− on the anion site failed. The failure was unexpected since even small amounts of
[FeCl4]− in both Connelly-type and other (see below) samples were safely detected on the level of
contemporary diffractometer data sets. To this end, we concluded from the colour of the samples that
MNIC/DNIC co-crystallisation had taken place, with the result that the 1983 method should not be applied
for the preparation of pure MNIC samples for physical measurements.
This result has been supplemented by a recent structure determination by Akutsu et al. on isostructural
crystals with the PPh4 counterion.[18] The authors used the Beck method for the preparation of a salt of the
+
claimed formula PPh4[FeCl2(NO)2]. The report, however, shows all the problems mentioned: severe NO/Cl
disorder in the structure analysis, and a magnetic moment which is clearly closer to the S = 3/2 spin of the
{FeNO}7 trichlorido-MNIC than to the S = 1/2 spin of the claimed {Fe(NO)2}9 dichlorido-DNIC.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
7
A standard route to pure [FeCl3(NO)]− and [FeBr3(NO)]− salts with various cations
For the most convenient approach to chlorido- and bromido-MNICs, the modified Kohlschütter route
which was used for the preparation of crystalline AsPh4[FeCl3(NO)], was further generalised. As a result,
pure, crystalline trichlorido- as well as tribromidonitrosylferrates of various, more or less bulky cations were
prepared by a facile procedure which uses inexpensive educts (ferrous chloride tetrahydrate or ferrous triflate
and a bromide, the chloride or bromide of the desired counterion, degassed methanol as the solvent, and an
excess of purified, i.e. NO2-free, nitric oxide) at ambient conditions. The procedure follows the equation (for
the chloride):
A+Cl− + FeCl2 + NO → A+[FeCl3(NO)]−
The water content of the tetrahydrate employed did not interfere nor was there need to limit NO
dosage. The reaction’s progress was easily monitored by the deepening of the characteristic green colour.
The typical time for NO(g) exposure was about 2–10 min (2 min for a newer apparatus with quick inert-gas–
NO displacement, 10 min for an older one with a somewhat slower displacement). The yield of isolated
crystalline product reached 90% for the larger counterions PPh4 , AsPh4 , and PPN+. All solids were stable to
air for months, thus no special care had to be taken for investigating them. This property should be
mentioned since the mother liquors belong to the low-stability class defined in Ref. [6b] which means that
purging the solutions with a stream of inert gas will destroy the nitrosyl complex rapidly by reversing its
formation reaction. Seeding and growing crystals thus was performed under an NO atmosphere throughout
this work.
+
+
All the photophysical measurements decribed below were conducted on samples from this standard
procedure, the only exception being the special cobaltocenium case (see the next subsection).
For the [FeCl3(NO)]− anion, well-ordered crystal structures were found with the counterions NMe4 ,
+
NEt4 , NBnMe3 (Bn = benzyl), Mephaz+ (5-methyl-phenazinium), PPh4 , AsPh4 , PPN+ =
+
+
+
+
bis(triphenylphosphaneiminium).
With ferrous triflate and the threefold molar amount of the bromide of the counterion, A+Br−, the
standard route yielded the respective salts of the homologous [FeBr3(NO)]− anion. This variant was used for
+
+
the preparation of well-ordered crystals with PPh4 or AsPh4 as the counterions. These crystal-structure
analyses were the first of a salt of the bromido-MNIC. The result is shown for the PPN salt in Figure 2.
Further details of both analyses are collected in the SI.
O1
N1
Fe1
Br1
Br3
Br2
Figure 2. The structure of the anion in monoclinic crystals of PPN[FeBr3(NO)] (50% ellipsoid probability).
Distances (in Å) from Fe1 to: Br1 2.3657(5), Br2 2.3755(6), Br3 2.3843(6); Fe1–N1 1.729(3), N1–O1 1.150(3).
Angles (in degrees): Fe1–N1–O1 169.5(3); mean Br–Fe1–Br 109.7.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
8
Attempts to synthesise salts of an unknown [FeI3(NO)]− anion failed. Nitrosylation of iodide-
containing solutions always yielded the respective DNIC [FeI2(NO)2]−.
The term “well-ordered” has been stressed repeatedly. The criteria applied include the usual numbers
of merit of an X-ray experiment, and, in addition, Fe–N and N–O distances within a narrow range. As a
typical example, Figure 7 in a later section includes the parameters for Mephaz+[FeCl3(NO)]−.
Crystallographic data for the other salts are collected in the SI.
The reduction of ferric educts by a reduced counterion
In the case of the one-electron reductant cobaltocene, [Co(cp)2], a variant of the standard procedure
which appears to be applicable if the reduced form of the counterion is available, was successful. Starting
with ferric instead of ferrous chloride, the attempted product was obtained according to the equation:
[Co(cp)2] + FeCl3 + NO → [Co(cp)2]+[FeCl3(NO)]−
The reduction of iron(III) was complete in terms of the structural data of the product which thus
appeared free of ferric contaminants. Details of the X-ray analysis are listed in the SI. The motif to prepare a
salt with a reducible cation is commented in the physics part below.
The reduction of ferric educts by NO
Utilising the observations reported by Manchot and Griffith, we replaced the ferrous chloride
component of the standard procedure with ferric chloride.[11, 13] In methanol, ferric chloride was transformed
to the [FeCl3(NO)]− ion by NO(g) as the reductant according to the equation:
A+Cl− + FeCl3 + 2NO → A+[FeCl3(NO)]− + MeONO + HCl
The formation of methyl nitrite is best monitored by UV-vis spectroscopy where a characteristic ‘five-
finger’ motif indicated MeONO formation (see the SI).
Even for the smaller counterions, the crystallisation often proceeded more quickly than in the case of
the respective standard ferrous route. A typical case is the tetraethylammonium counterion. The ferrous route
yielded green solutions, as usual, but, in great contrast to a large counterion such as PPN+, no crystals
precipitated from them for months. Eventually, green crystals seeded after about a year. The same experiment
on the ferric route resulted in the seeding and growth of green NEt4[FeCl3(NO)] crystals from cooled mother
liquors within about two weeks.
At first glance, the structure analysis on NEt4[FeCl3(NO)] crystals from the ferric route was
inconspicuous in terms of the usual figures of merit such as R values or residual electron densities. However,
the analysis shared a common peculiarity with the analyses on other crystals from the ferric route with
various counterions, namely an atypical long Fe–N and a shortened N–O bond. The origin of this observation
on the ferric variant of the standard ferrous route will be resolved in the Crystallographic Section.
The anomaly can be remedied by a variation of the conditions, namely dilution to about one tenth of
the standard concentration. This alteration yielded pure crystals of the target compounds with unsuspicious
X-ray parameters. However, the ferric route then becomes unattractive from a practical viewpoint. Yields
were low, and a long crystallisation time for small counterions as typical for the ferrous route, resulted.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
9
Consecutive MNIC-to-DNIC transformation in the presence of base
Towards a standard route to dichlorido-DNICs
The reliability of the standard and the ferric procedure to yield either pure or FeIII-contaminated
MNICs depends on the absence of base. Various publications from the second half of the last century
reported the formation of DNICs if base was added to the otherwise identical reaction mixtures.[14, 16] An
example is the report of Gwost and Caulton who added triethylamine to obtain the DNIC as the sole
nitrosylated (not isolated) solution species in terms of IR spectroscopy (B = NEt3) according to:[15]
FeCl2 + 3NO + MeOH + B → [FeCl2(NO)2]− + MeONO + BH+
Our attempts to reproduce this investigation led to a specification of Gwost and Caulton’s statement
“... when exposed to NO the solution immediately darkened; after 9 hr of bubbling NO, the solution was
deep red.” In fact, the reported darkening was of a green colour which subsequently turned brownish red.
Obviously, the DNIC formed in a further step after the initial formation of the corresponding MNIC.
As with the ferric variant of the standard procedure, the formation of the methyl nitrite side product
was best monitored by UV-vis spectroscopy. The sequential nature of the reaction became particularly clear
with fluoride as the base. (Originally, we attempted to prepare fluorido analogues of the title compounds.
However, in all reactions fluoride reacted as a base, not as a coligand.) We found that the admixture of
increasing amounts of fluoride to an ACl/FeCl2/MeOH standard-route mixture made the green solution (at no
added fluoride) turn more and more reddish brown according to the equation:
[FeCl3(NO)]− + F− + 2NO + MeOH → [FeCl2(NO)2]− + Cl− + MeONO + HF
Red-brown DNIC solutions are accessible more directly by using the fluoride of A. For A = NMe4:
+
−
FeCl2 + 2 NMe4F + 3NO + MeOH → NMe4[FeCl2(NO)2] + MeONO + NMe4 HF2
Attempts to prepare solutions of pure DNIC included the variation of base amount. In fact, the
threefold instead of a double molar amount resulted in a poor DNIC yield. Moreover, charging the reaction
mixtures with four moles of fluoride over iron left the mixture unreactive towards nitric oxide. We
interpreted this result with the assumption that the ferrous-salt-to-DNIC transformation was a two-step
process, the first and second steps of which were governed by individual parameters. Thus MNIC formation
appeared as fluoride-inhibited whereas DNIC formation was fluoride-promoted.
A two-stage procedure of this two-step reaction sequence thus accelerated the formation of the final
product. A react-IR study was used to demonstrate this: first, ferrous chloride tetrahydrate was reacted
rapidly in methanol with NO to the respective MNIC and small amounts of DNIC only. In the second step,
the addition of base resulted in the immediate increase of the DNIC signals with a decrease of the MNIC
peak.
Attempts to provide a facile standard procedure for the preparation of A[FeCl2(NO)2] salts have been
made for the tetramethylammonium counterion. Currently, the most recommendable procedure for the
preparation of the respective DNIC starts from ferrous chloride and twice the molar amount of
tetramethylammonium fluoride. Treatment with NO led to red-brown solutions. However, the crystallisation
still took about a month to obtain only a 30% isolated yield.
One additional point should be made which highlights the fact that MNIC:DNIC ratios from a specific
reaction do not depend only on the amount of base as the most important factor, but that there are other
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
10
control parameters. One of them is the applied Cl:Fe molar ratio. Thus, to synthesise an MNIC or DNIC, the
stoichiometric 3:1 or 2:1 ratio, respectively, should be applied. DNIC synthesis with chloride excess or
MNIC synthesis with chloride shortage provokes the admixture of the respective ‘wrong’ NIC.
DNIC formation via a ferric route
Bearing the two-step formation of a DNIC in mind, a ferric variant should exist since it was already
proved that the MNIC precursor is accessible from FeIII. In fact, the addition of the 2.5-fold molar amount of
tetramethylammonium fluoride to a solution of FeCl3 in methanol resulted in the formation of red-brown
crystals of NMe4[FeCl2(NO)2] which, however, were accompanied by green crystals of the respective MNIC.
Pure DNIC was obtained from a dilute batch containing 3 mol of fluoride per mol ferric salt according to the
equation:
FeCl3 + 2NMe4F + 4NO + 2MeOH → NMe4[FeCl2(NO)2] + NMe4Cl + 2MeONO + 2HF
The surplus third mole of fluoride should have formed NMe4(HF2) with half of the liberated
hydrofluoric acid. Attempts to shift the reaction more to the right by adding another mole of fluoride showed
the method’s same limitation, as above, namely the inhibition of the overall nitrosylation. It should be noted
that even a 3:1 F:Fe ratio resulted in poor yields for the ferrous route. Obviously, the well-known formation
of fluoridoferrates(III) led to decreased base activity of the added fluoride.
Touching on the Gwost–Caulton procedure, triethylamine was employed as well. In fact, a DNIC
solution formed but crystallisation was hampered.
Solid A[FeCl2(NO)2] salts: structures and IR spectra
Despite the fact that the optimisation of the syntheses toward pure A[FeCl2(NO)2] salts is needed,
progress has yet been made in terms of structural data. For the current status, similar statements apply as for
the MNIC counterpart. First, preparative procedures were reported decades ago.[17, 24-25] Second, the isolation
of pure material, however, has never been demonstrated in terms of an unsuspicious structure analysis,
preferably supported by matching magnetic data (see the PPh4[FeCl2(NO)2] case above).[18] However, this
point was resolved. Both the tetramethylammonium as well as the PPN salt of the dichloridodinitrosylferrate
anion (for the somewhat lengthy preparation of the PPN salts see the Experimental Section) yielded reliable
X-ray data sets. As a result, four X-ray analyses of pure A[FeCl2(NO)2] salts are available. Details are given
in the SI; as an example, Figure 3 shows the structure of the anion of the NMe4 salt.
O1
O1i
N1
N1i
Fe1
Cl1
Cl2
Figure 3. The structure of the anion in crystals of NMe4[FeCl2(NO)2] (50% ellipsoids). The ion exhibits
crystallographic Cs symmetry. Distances in Å: Fe1–N1 1.709(2), N1–O1 1.146(2), Fe1–Cl1 2.2797(8), Fe1–Cl2
2.2751(7). Angles (in degrees): Fe1–N1–O1 160.8(2), N1–Fe–N1i 108.30(8), Cl1–Fe1–Cl2 107.41(3).
Unlike the MNIC results, all X-ray analyses revealed the same atomic distances in a narrow range
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
11
despite the different synthetic routes that were used to prepare them. In addition, the magnetic susceptibility
of the PPN salt prepared on the ferrous route (µeff = 1.92), complied with the expected S = ½ spin (1.73 spin
only).
The most severe problem for the A[FeCl3(NO)] preparation via the ferric route, the co-crystallisation
with A[FeCl4], thus did not seem to be problematic here. Future work on the optimisation of the DNIC
branch thus appears promising for both a ferrous and a ferric variant. Speculating on the reasons for the lack
of a disadvantage for a ferric route to the DNIC, two points seem obvious. The first is a crystallographic
argument and points to the lesser similarity and thus a worse mutual replaceability of the [FeCl4]− and the
[FeCl2(NO)2]− ion. However, the second argument appeared to be the decisive one: the reaction equation
above practically runs to completeness in a shorter period of time; hence, there is no [FeCl4]− to co-
crystallise.
In the next chapter, the co-crystallisation of A[FeCl3(NO)] and A[FeCl2(NO)2] salts will be addressed
as well. In this chapter on DNIC synthesis, however, that property of MNIC/DNIC mixtures did not create a
problem since it was easy to prepare a DNIC solution free of MNIC. IR spectroscopy was mentioned above
for routinely monitoring MNIC vs. DNIC formation. For solid DNICs, two vibrational bands for the N–O
stretch, one for the symmetrical, one for the asymmetrical, were observed at values similar to those of the
solution. For NMe4[FeCl2(NO)2], 1783 and 1687 cm−1 were found.
Solid A[FeX2(NO)2]-type bromido- and iodido-DNICs
As with the respective MNICs, solutions of the bromido analogues were obtained by the same
procedures. However, pure crystals of a bromido-DNIC have not yet been obtained by this route. Instead, we
present the first structure determination of an A[FeBr2(NO)2] salt as a preliminary result in respect to the
synthesis of the mother liquor. Hence, red-brown crystals of PPN[FeBr2(NO)2] were obtained from an MNIC
batch (molar ratio Fe:Br of 1:3) to which some aqueous solution of hydrobromic acid (ca. half a mole per
mol iron) was added. The bromido-DNIC’s formation points to a subsequent disproportionation of the
primary product, the MNIC, according to the equation: 2[FeBr3(NO)]− → [FeBr2(NO)2]− + [FeBr4]−. Details
of this reaction remain to be clarified as well as the significance of FeIII reduction by NO in the bromide
regime. The structure of the anion in the bromido-DNIC is shown in Figure 4. The formation of a
{Fe(NO)2}9-type DNIC by the dissociation of an {FeNO}7-type MNIC had been reported often, most
recently in Ref. [22] (and references cited therein).
O1
O2
N1
N2
Fe1
Br1
Br2
Figure 4. The structure of the anion in crystals of PPN[FeBr2(NO)2] (50% ellipsoids). Distances in Å: Fe1–Br1
2.4108(4), Fe1–Br2 2.4173(4), Fe1–N1 1.707(2), Fe1–N2 1.723(2), N1–O1 1.145(2), N2–O2 1.099(3). Angles
(in degrees): Fe1–N1–O1 162.73(18), Fe1–N2–O2 166.34(18), N1–Fe1–N2 112.26(9), Br1–Fe1–Br2 109.12(1).
The iodido-DNIC showed a markedly higher formation tendency. Salts with the [FeI2(NO)2]− anion
were thus obtained on attempts to synthesise iodido-MNICs, the counterions being AsPh4 , PPh4 , and PPN+.
+
+
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
12
As a by-product, I3 was formed in all batches, partly as crystals of A+I3 , and, in one case, co-crystallised
−
−
with the DNIC. The reaction seems to have followed the equation: Fe2+ + 3.5I− + 2NO → [FeI2(NO)2]− +
−
0.5I3 .
Crystal structures of PPN[FeI2(NO)2] and (PPN)2[FeI2(NO)2](I3) as well as AsPh4[FeI2(NO)2] and
PPh4[FeI2(NO)2] were described in the SI. PPN[FeI2(NO)2] was isotypic with the same salt from a different
synthetic route.[21]
Crystallography: crystal pathology and seeding, crystal packing
Crystal pathology and seeding
As a most characteristic feature, crystalline halogenidonitrosylferrates exhibit a typical pathology
which originates because of the exchangeability of chloride and nitrosyl moieties on the point positions
occupied by the ligands. Notably, this statement does not include the supposed occurrence of Cl/NO disorder
in the pure products. Instead, anions of the same charge such as [FeCl2(NO)2]−, [FeCl3(NO)]−, and [FeCl4]−
occupy the same positions in a crystal structure, if they are present, side by side, in the respective mother
liquors. From a synthetic viewpoint, chloride/nitrosyl exchangeability at a constant anion charge meant that a
proven method, recrystallisation, was no longer available as a purification procedure. Instead, these anions
formed solid solutions which, on the one hand, obscured the crystallographic characterisation, and, on the
other, impeded the preparation of pure substances for further use in physical and chemical experiments.
The irritating observation that there was no general Cl/NO disorder problem in the pure substances
may be explained by the following observation. By using the standard route starting from FeCl2 in MeOH,
the addition of (NEt4)Cl as the cation source, followed by NO exposure, resulted in a green solution. In
contrast to other cations, these NEt4 batches did not crystallise within prolonged periods of time (more than
half a year). By using the FeCl3 variant, green solutions were obtained as well. But now, crystallisation took
place in the course of about two weeks. Structure determination revealed an NEt4[FeCl3(NO)], contaminated
with about 15% [FeCl4]− content. From this observation we conclude that the contaminant NEt4[FeCl4] was
not homogenously distributed in the NO-containing bulk but had served as a seed due to its lower solubility
and thus gave rise to the epitaxial growth of the bulk.
The presence of ferric contamination became obvious in the measurement of the magnetic properties
of the samples (see the SI). Values in the expected range for an S = 3/2 species were measured only for
material from the standard route whereas ferric-route samples showed the same high values that had been
reported in the 1958 work by Griffith et al. in their first description of products from FeCl3(sol)/NO(g)/EtOH
reactions.[13]
+
Samples with the NMe4 counterion behaved similarly and shall be used as an example which focuses
on numerical aspects. The specific methodological problem that arises from samples with a small amount of
ferric contaminant/seed is that, on the one hand, the figures of merit of standard X-ray analysis hardly
indicate the presence of a contaminant. On the other hand, structural parameters such as atomic distances
may be seriously in error. Thus, isotypic orthorhombic crystals of NMe4[FeCl3(NO)] in the polar space group
Pca21 were prepared both by the standard ferrous route and by the ferric variant. Table 1 shows significant
numbers in four rows: (1) the BP86/def2-TZVP result from Table 2 as a guideline for what the expected
values are, (2) the result from a room-temperature measurement on crystals from the ferrous route, (3) the
well-matching results from a corrected low-temperature measurement (−100 °C) on crystals from the ferric
route, and (4) the same as 3 without a correction for ferric contaminants.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
13
Table 1. Atomic distances from X-ray analyses on a crystal of pure (second row) and FeIII-contaminated (third and
fourth row) NMe4[FeCl3(NO)]; ‘uncorrected’ refers to a nitrosyl position fully occupied by an NO group; ‘corrected’
includes splitting of the nitrosyl position which, after refinement, was occupied by 0.912(13) NO and 0.088 Cl with Fe–
Cl 2.16(3) Å (compare the typical FeIII–Cl distance of 2.188 Å explicated in the ‘Charges’ chapter below).
Fe–N/Å N–O/Å
1.702 1.165
Δρ/eÅ−3 wR2
BP86b (see Table 2)
FeII route
1.710(7) 1.154(8) 0.367
1.729(7) 1.145(12) 0.311
0.0875
0.0372
0.0378
FeIII route, corrected
FeIII route, uncorrected 1.789(3) 1.059(3) 0.293
Figure 5 shows both uncorrected (left) and corrected (right) results. Despite the fact that the usual
figures of merit of the X-ray analysis such as the wR2 value or the residual electron density did not mirror
the crystal’s pathology of presumably containing an NMe4[FeIIICl4] nucleus (Table 1), the thermal ellipsoids
of the nitrosyl group evoke a strong suspicion that something might be wrong. In particular, the nitrogen
atom should not vibrate preferably along the strong N–O bond but perpendicular to it. Note that both the N
and O ellipsoids looked reasonable after the correction the details of which are given in the table and figure
captions.
(a)
(b)
Figure 5. (a) 50% ellipsoids from the uncorrected X-ray analysis. (b) The same after splitting of the NO position
into the depicted N and O atoms (site occupancy 0.912) and 0.088 isotropically refined Cl close to the N atom
(green cross).
PPN[FeCl3(NO)] was chosen to decide whether (1) ferrate(III) seeding mirrors an unavoidable ferric
content of a mother liquor of the ferric route, i.e., a thermodynamic limitation of the route, or (2) the reaction
may run practically to completeness but rapid ferrate(III) seeding interferes with the redox reaction which
marks a kinetic problem. To differentiate these cases, it helps to reformulate the equation above as:
+
−
+
−


A Cl + FeCl3 + 2NO
A[FeCl4] + 2NO
A [FeCl3(NO)] + MeONO + HCl
In fact, the formation of a yellowish precipitate on addition of ACl and FeCl3 was frequently observed.
In these batches, A[FeCl4] seeds were present from the beginning of nitrosylation.
In case 1, the thermodynamic limitation implied that either the first or the second equilibrium, or both,
may leave substantial amounts of the respective components of the left sides unreacted in solution. Case 2
implied that, due to their lower solubility, seeds of A[FeCl4] formed and remained, partly or completely, in
the solid state where their accessibility was insufficient to re-enter the equilibrium. This view is supported by
the consideration that the attempted product forms a shell about these seeds.
To decide between these cases, the precipitation of A[FeCl4] was prohibited by the dilution of the
reaction mixture to such an extent that turbidity was not observed for prolonged periods of reaction time. As
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
14
a result, pure, ‘ferrous-route-quality’ crystals of PPN[FeCl3(NO)] were obtained from a ferric chloride
solution with a tenfold solvent content. As the most meaningful method, the magnetic characterisation
revealed the following effective Bohr magneton numbers (µeff) for PPN[FeCl3(NO)] from the ferrous route:
4.20, from the standard ferric route: 5.29, from the diluted ferric batch: 4.20 (compare the spin-only value for
S = 3/2: µeff = 3.87).
Crystal packing
The crystal structures showed common features. Most important for our goal of rotating the nitrosyl
ligand on irradiation was the conspicuous lack of tight interatomic contacts that included the NO groups
which, thus, were not spatially fixed in a narrow cavity. Instead, (non-classical) hydrogen bonds, as the most
typical interatomic interaction, ran solely from the respective hydrogen-bond donors to halogenido acceptors.
This behaviour mirrored the electrostatic situation of the anions. Figure 6 shows the electrostatic
potential (ESP) of the [FeCl3(NO)]− ion mapped on the ion’s van-der-Waals surface (following common
usage, the 0.001 a.u. surface was taken as the van der Waals shape; p. 461 in Ref. [26]
.
−0.10
−0.16
Figure 6. The electrostatic potential (ESP) of the [FeCl3(NO)]− ion (Cl left and backwards, NO right; atomic
units), mapped on the 0.001 a.u. (ca. 0.0067 e Å−3) surface of the total electron density. The values were taken
from a BP86/def2-TZVP calculation on the free [FeCl3(NO)]− ion.
The charge distribution shows the highest negative charge on the chlorine atoms. It is not uniform over
the entire shape of each individual ligand, but the polar regions in the extensions of the Fe–Cl and the Fe–N–
O bonds differ from the lateral regions about the bonds. Thus, on the one hand, the poles of the chlorido
ligands show a decreased negative charge, the typical σ holes.[27] On the other hand, the pole about the
nitrosyl-oxygen showed an increase of negative charge with respect to the lateral region about the Fe–N–O
bond due to the lone pair at the O atom. As an overall result, the cation-binding directions are expected to be
normal to the Fe–Cl vectors on the ‘back side’ of the anion opposite to the nitrosyl function. In fact, the
crystal structures verified this result. Figure 7 shows the prototypical situation in the salt with the N-methyl-
phenazinium counterion which complies with the lateral-only, the back-side, and the NO-exclusion features
derived from the ESP.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
15
Cl2
H10i
O1
N1
Fe1
Cl3
H9i
Cl1
Figure 7. Mephaz+[FeCl3(NO)]−. The ion pair with the shortest interatomic contact (H10i···Cl2 2.782 Å, the
second contact H9i···Cl1 is 2.969 Å) in the anion’s orientation close to Figure 6 (Ortep, 50% ellipsoid
probability). Note hydrogen bonding towards the region of highest charge lateral to the Fe–Cl bonds at the
chlorine acceptor. The depicted hydrogen-bond motif is repeated with Cl3 and Cl1 as a second acceptor couple.
Symmetry code: i 5/2 – x, y – ½, ½ − z. Interatomic distances in the anion in Å: Fe1–Cl1 2.2357(7), Fe1–Cl2
2.2291(6), Fe1–Cl3 2.2594(6), Fe1–N1 1.734(2), N1–O1 1.152(2); angles (in degrees): Fe1–N1–O1 170.1(2),
mean Cl–Fe–N 108.0, mean Cl–Fe–Cl 110.9.
The AsPh4[FeX4−x(NO)x] structure family
+
Cations of the EPh4 type (E = group-15 element) have a pronounced tendency to form square column
packings, leaving inter-column channels for the anions. The various subtypes have been classified by Müller
by means of a Bärnighausen tree.[28] About half of the types belonged to the tetragonal system (14 out of 30).
Only one of the tetragonal types provided an anion position without crystallographic point symmetry (S4
symmetry for the cation). This structure in space group P4 was chosen by PPh4[FeCl3(NO)],
PPh4[FeBr3(NO)], and AsPh4[FeBr3(NO)]. All three structures were well resolved and free of disorder. A
monoclinic variant was found for AsPh4[FeCl3(NO)] in space group P 21/n with the 21/n symmetry parallel to
the column axis. The [010] direction, which resembled the [001] direction in the tetragonal cell, was twice
the tetragonal c length. The square packing was slightly distorted (a ≈ c ≈
, β ≈ 90°). The
atetr / 2
monoclinic variant shared a significant feature with the tetragonal cell: no crystallographic symmetry was
imposed on the anion. Also the monoclinic structure was well-ordered without Cl/NO scrambling.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
16
b
o
a
Figure 8. The crystal structure of PPh4[FeCl3(NO)]. Space group P4 , projection along [001] ; the PPh4 ions
occupy special positions with S4 symmetry. Ortep, 50% probability ellipsoids.
The P4 cell was assigned to their brown needles by Beck to AsPh4[FeCl3(NO)] and by Akutsu to
AsPh4[FeCl2(NO)2].[18, 23] As has been elaborated above, both groups obviously dealt with MNIC/DNIC co-
crystals (in terms of the reported room-temperature metrics of the unit cells, the Beck crystals contained a
higher share of DNIC, just opposite to the claims in both papers). As a result, we can state that Cl/NO
disorder was not triggered by cation arrays which imposed crystallographic symmetry to the anion part and
thus provoked disorder. The pure MNICs of this structure family crystallised free of disorder. Conversely, the
pathology of the reported structures is of chemical origin. Hence, co-crystals formed from mother liquors
that provided a mixture of anions which seemed to be similar enough for the mutual replacement in a crystal
structure.
Structure and bonding: the linear, high-spin, covalently π-bonded Fe–NO entity
Methodology for {FeNO}7(S = 3/2) species
A detailed methodological study by Radoń et al. on {FeNO}7 species which included both density-
functional and wave-function methods included the S = 3/2, high-spin, case using the example of the
tentative ‘brown ring’ complex, [Fe(H2O)5(NO)]2+.[29] In the wave-function part of their study, the authors
used a CASSCF(9,13) approach. The nine electrons placed in the active space included the lone pair at the N
atom of a formal NO+ donor ligand (3σ in an NO molecular-orbital scheme), and the seven ‘Enemark-
Feltham electrons’. In total, the active space was made up of the five Fe(3d) orbitals, five Fe(4d) ‘second-
shell’ orbitals and three NO-based orbitals: the mentioned 3σ orbital and the two degenerate π* orbitals. This
approach was also used below for the [FeCl3(NO)]− ion.
DFT investigation of the [FeCl3(NO)]− ion
Since a reliable assignment of the N–O valence frequency was essential for the linkage-isomerism part
of this work, the theoretical level was chosen to achieve a good agreement between experimental and
calculated ν(N–O) values. As a result, popular hybrid functionals such as B3LYP or TPSSh were excluded
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
17
since the (unscaled) N–O frequency was calculated some 50 cm−1 too high for these functionals. A better
agreement was achieved with the methods of Table 2. Of these, the metrical parameters of the BP86b entry
were used for the wave-function-based calculations described below. For the TPSS functional, the
(negligible) influence of a larger basis set (def2-aug-TZVPP instead of def2-TZVP) was checked.
Table 2. Distances, (mean) angles, and energy of the N–O stretching vibration for six reliably analysed solids from the
standard ferrous route, and the [FeCl3(NO)]− ion in DFT calculations for the given method and the def2-TZVP basis set
(Grimme’s van-der-Waals correction; environment modelled by a COSMO [Orca3] or CPCM [Orca4] approach at
practically infinite dielectric constant; uncorrected wave numbers tabulated). More functionals, in particular hybrid
functionals, are included in Table S1 in the SI.
−1
Fe–Cl/Å
Fe–N/Å N–O/Å Fe–N–O/° Cl–Fe–Cl/°

 (N–O)/cm
mean 6×Xray 2.236 ± 0.006a 1.726
1.149
1.167
1.165
1.163
1.161
1.163
173.5
177.5
179.9
176.9
178.3
178.4
110.1 ± 1.2a 1797
BP86
2.249
2.248
2.251
2.275
1.705
1.702
1.712
1.745
1.733
108.7
108.7
109.2
109.4
109.0
1786
1791
1792
1799
1794
BP86b
TPSSc
B97-D
B97-D+zora 2.268
a
the given standard deviation of the mean exceeds the 10-fold of the mean standard deviation of the X-ray
refinement, taken as indicating a ‘soft’ variable in a crystalline environment
b
c
Orca4 with CPCM(water) instead of Orca3.03 with COSMO(water)
the same values were obtained with the def2-aug-TZVPP basis set except 1790 cm−1 for the valence vibration
Fractional-occupation-number-weighted density
The analysis of the fractional-occupation-number-weighted density (FOD) as recently introduced by
Grimme was used as a bridge to a wave-function-based approach to the electronic situation of the
[FeCl3(NO)]− ion.[30] Figure 9 shows the result for one of the procedures recommended in Ref. [30]. (The
result was the same for other recommended combinations of method and Fermi temperature such as TPSS at
5000 K.) The FOD analysis indicated the orbitals that needed a correlated treatment, namely the Fe(d) and
the NO(π*) orbitals. In terms of the FOD, an active space of 7 electrons in 7 orbitals thus appeared as a
minimum requirement for capturing electronic correlation.
Figure 9. FOD (BHLYP/def2-TZVP, smear temperature 15000 K, isovalue 0.005 a.u.) of the [FeCl3(NO)]− ion.
A CASSCF approach to the [FeCl3(NO)]− ion
For a wave-function-based approach to the bonding in [FeCl3(NO)]−, we followed the active-space
construction of Radoń’s study.[29] Thus, Figure 10 shows an active space of 9 electrons in 13 orbitals. The
active space thus was chosen to exceed the minimum space of the FOD analysis by one doubly occupied MO
(MO 42, the bonding counterpart of the σ-antibonding MO 48), and, as recommended for an electron-rich 3d
element, a second, 4d shell for the iron atom.[31] In fact, in the ground-state study of this section, the
occupation of MO 42 with 1.99 electrons in the (9,13) approach revealed this MO close to inactive. Since,
however, the occupation of MO 42 close to two did not encounter convergence problems in the CASSCF
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
18
routines, we proceeded with the larger active space for the sake of comparability with the earlier study and
the investigation of excited states below, which benefitted from balancing the Fe–NO antibonding MO 48
with its bonding counterpart MO 42 (Figure 10).
The most typical feature which shares the [FeCl3(NO)]− ion with other nitrosyls, is the marked
population of the Fe–NO π* antibonds and the corresponding depopulation of the Fe–NO π bonds (compare
the 0.27:1.73 antibond:bond population of this ion with a 0.28:1.71 ratio of the tentative brown-ring aqua
species in Ref. 35, or the 0.21:1.79 ratio within the natural orbital population of the related singlet species
[Fe(CO)3(NO)]− in Ref. [32].) Accordingly, the leading 22211100 configuration (second shell omitted)
depicted in Figure 10 contributed only about two thirds to the ground state, leaving one third to single and
double excitations from the Fe–NO bonds to the antibonds.
Taken together, these numbers reveal marked static correlation as is characteristic for bonds with
hindered overlap, including the textbook example of a dissociating covalent bond (p. 526 in Ref. [26]).
The reasons for the hindered overlap in [FeCl3(NO)]− are twofold. The first reason applies to every
nitrosyl compound derived from a (formal) NO+ core. The argument may be developed starting from an
arbitrary assignment of the two electron pairs involved. Let us, again arbitrarily, start from an NO+ ion and a
d7-FeI species with an electron pair each in d(xz) and d(yz). Now, two Fe–NO π bonds are formed through
the interaction of the Fe-d(xz/yz) pairs and the two empty N–O π* orbitals. The crucial point is the need for
the overlap of filled (metal d) orbitals with the empty N–O antibonds since the latter reside in the same
spatial region as their occupied N–O-bonding π counterparts. Thus, when occupied-metal-d/empty-π*
overlap occurs, at the same time, occupied-metal-d/occupied-NO(π) Pauli repulsion encumbers the metal–
ligand interaction. (The same arguments explain the disability of two NO radicals to establish a normal N–N
bond in an N2O2.[33]) To avoid confusion in view of the various ‘π’s, the notation for a specific interaction
used throughout this work is explicated in the example of MOs 44 and 45 in Figure 10, where the Fe–N
interaction is π-bonding, and the N–O interaction is π-antibonding, i.e., π*. The specific interaction
addressed is indicated by a ‘–’. Hence, formulations such as ‘Fe–NO π bond’, ‘N–O π* interaction’, and ‘N–
O π bond’ define the interaction in question.
The second reason is more specific, but is caused by operative Pauli repulsion as well. In a high-spin,
S = 3/2, {FeNO}7 species, the nitrogen atom’s lone pair faces the singly occupied d(z2) orbital, giving rise to
a bonding (MO 42) and an antibonding interaction (MO 48). To minimise the repulsion through MO 48,
many high-spin {FeNO}7 centres are found bent with the Fe–N–O angle ranging between about 150 and
180° in a shallow potential.[6a, 34] (In terms of orbital contributions, the bending resembles a d(xz) admixture
to d(z2), see below in the PLI section.) However, a marked bending is missing in [FeCl3(NO)]−, both
experimentally and in the calculations. A closer inspection of the MOs in Figure 10 reveals the reason for the
linearity of the FeNO moiety. Since a trans-to-NO ligand is missing in a tetrahedral species, the Fe-3d(z2)
orbital may be altered by an Fe-4p(z) admixture instead of a bending Fe-d(xz) contribution, which results in
the differently sized upper and lower lobes of d(z2) indicating the destructive and constructive admixture,
respectively. As a result, the antibonding character is decreased, even in the absence of bending. However,
due to the option of d(xz) admixture, the bending potential is also shallow for the [FeCl3(NO)]− ion (see
Figure 16, the red circles between abscissa values of 150 and 180°).
Spin
The high-spin character of the [FeCl3(NO)]− ion is largely caused by the singly occupied MOs 46–48.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
19
There is, however, to some extent, spin polarisation along the Fe–NO fragment, as has been investigated in
detail in Radoń’s study.[29] In this reference, the related [Fe(H2O)5(NO)]2+ species exhibits, on the
CASSCF(9,13) level, a Mulliken spin density of −0.49 on the NO ligand and 3.45 on the Fe atom. The
values for [FeCl3(NO)]− are similar: −0.50 for NO and 3.40 for the Fe atom.
Also in agreement with Ref. [29] is the markedly functional-dependent extent of spin polarisation in
DFT calculations. Thus, compare the BP86/def2-TZVP Mulliken spin densities (the aqua complex’s values
from Ref. [29] in parentheses) of NO −0.50 (−0.71) and Fe 3.40 (3.50) with the respective B3LYP numbers:
NO −1.07 (−0.93) and Fe 3.60 (3.76). A broken-symmetry approach offers an alternative perspective on the
antiferromagnetic character of the Fe–NO interaction. As a result, the considerable overlap of the α and β
part of the Fe–NO π bonds in terms of a corresponding-orbital transformation resulted in S = 0.84 on the
TPSSh/def2-TZVP level; a value that shows close resemblance to a bonding analysis on related DNICs.[35] In
both MNICs and DNICs, the typically large values of S close to 1 indicate the largely covalent character of
the Fe–NO interaction.[36]
Figure 10. Frontier orbitals of the [FeCl3(NO)]− ion [CASSCF(9,13)/def2-TZVP; isovalue 0.06 a.u.]. Orbital
labels refer to Cartesian axes: z up, y to the right, x to the viewer; orbital numbering starts with “1” (= Orca
numbering + 1). The 22211100 occupation pattern indicated by the arrows refers to the ground state’s leading
configuration (65% contribution). Bold: the abbreviation of a level used in Table 6.
Charges
The bonding description of the Fe–NO moiety derived so far focuses on two Fe–NO π bonds formed
by the Fe-d(xz) and Fe-d(yz) AOs on the one hand and the N–O-π*(x) and N–O-π*(y) MOs on the other. If
the four electrons involved in these two bonds are shared equally between the metal and the ligand, i.e., if a
purely covalent interaction is assumed, the nitrosyl ligand with the then doubly occupied π* MOs would
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
20
count as NO−, leaving the iron centre as FeIII. Starting, notionally, from FeI and NO+ (as has been done in the
CASSCF section), purely covalent π (back)bonds would be obtained by completely delocalising the
d(xz)/d(yz) pairs into the Fe–NO bonds.
The first hint that the assumption of backbonding towards perfect covalence is too far-reaching, comes
from the mere inspection of the MOs in Figure 10. In a purely covalent bond, the contributions of the
bonding partners to both the bond and the antibond would be equal. Figure 10, however, shows a higher
contribution of the metal d orbital to the bond and a lower one to the antibond. Hence, we should not end up
with the charge −1 for the nitrosyl ligand but with a smaller absolute value.
A second hint may be derived from structural data. Well-resolved X-ray data of tetrachloridoferrates
are available for both the oxidation states II and III. For FeII, the CCDC provides data sets for tetrachlorido-
ferrates(II) (BERROF01, FIWGIB, GOXLUA, IFEGUY, TMAFEC), as well as for tetrachloridoferrates(III)
(FUGDER, KURPET, QUXFAR, MECXUO, SURRON). For the ferrates(II), the mean Fe–Cl distance of the
cited entries is 2.317 Å, whereas 2.188 Å is the mean of the FeIII–Cl distances. Due to the somewhat covalent
character of the Fe–Cl bonds, the difference among these values (0.129 Å) is slightly smaller than the
difference between Shannon’s ionic radii that were derived from fluorides and oxides (0.135 Å for
hexacoordinate high-spin FeII/III, 0.14 Å for tetracoordinate FeII/III).[37] Keeping these fixed points in mind, the
mean Fe–Cl distance from X-ray data on the [FeCl3(NO)]− ion from Table 2 of 2.236 Å points at the
oxidation state of the iron central atom between II and III, say 2.5 + x, which, in turn, assigns a charge of
−0.5 + x to the nitrosyl ligand. To this end, we see a, more or less, covalent nitrosyl–iron interaction through
two π bonds which are polarised towards the metal.
Table 3. Population analyses. The first three entries refer to Mulliken charges for the respective method (def2-TZVP
basis for all calculations). The bottom entry shows charges from a QTAIM analysis based on the BP86 calculation of
the second entry.
Fe
N
O
Cl
Mulliken charge:
CASSCF
BP86
1.05 0.25 −0.26 −0.68
0.47 0.11 −0.10 −0.50
0.57 0.15 −0.12 −0.54
TPSSh
QTAIM (BP86) 1.26 0.08 −0.36 −0.66
Finally, population analyses may contribute to this issue (Table 3). For the chlorido ligands, the values
span a rather narrow range distinctly below a full negative charge. Thus, the Fe–Cl bonds appear to be ionic
with a markedly covalent share. For the nitrosyl ligand, Mulliken charges for N and O sum up to values close
to zero, regardless of whether wave-function or DFT methods were employed.
A QTAIM analysis resulted in a slightly more positive central atom and a slightly more negatively
charged nitrosyl ligand. Taken together, the approximately −0.5 + x charge derived from the ‘chlorido probe’
above appears to be reliable for x values up to one half.
DFT and WFT investigation of the [FeCl2(NO)2]− ion
Most of the computational steps reported above were also performed for the DNIC analogue, the
[FeCl2(NO)2]− ion. Significant details include the match of the metric and vibrational parameters (Table 4)
and the FOD density which resembles the result for the MNIC.
Table 4. Distances, (mean) angles, and energy of the N–O stretching vibration for four reliably analysed solids and the
[FeCl2(NO)2]− ion in DFT calculations for the method BP86b of Table 2. The subscript ‘fCl’ denotes a geometry
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
21
optimisation with the Fe–Cl distance fixed to the mean experimental value.
Fe–Cl/Å Fe–N/Å N–O/Å Fe–N–O/° N–Fe–N/°
−1

 (N–O)/cm
mean 4×Xray 2.275
1.702
1.665
1.158
1.178
1.179
162.6
161.8
161.7
109.5
110.3
110.0
1694, 1778
1678, 1766
1677, 1765
BP86b
2.229
BP86fCl
2.275(f) 1.666
For the wave-function approach, the lone pairs at the nitrogen atoms were not included in the active
space this time (but see the calculation of excited states below). Hence, it was made up of the five Fe-d
orbitals, two pairs of NO-π* levels, and a second shell for Fe (14 orbitals in sum). The active space contains
the 9 ‘Enemark-Feltham’ electron. Within this space, a CASSCF(9,14) calculation showed the close
relationship between the DNIC and the respective MNIC.
The bonding situation is depicted in Figure 11. The orbital occupation numbers resemble the MNIC
(0.27:1.73) with respect to the antibond:bond population (0.26:1.74). At about one half, the weight of the
leading configuration 222210000 is still smaller compared to the two thirds for the MNIC, leaving the other
half for bond-to-antibond excitations.
Mulliken spin densities depended on the method: CASSCF yielded 1.60 for Fe, −0.31 for NO, and a
vanishing spin density of 0.01 for Cl. The respective BP86 values were 1.60 for Fe, −0.39 for NO, 0.09 for
Cl. Mulliken charges showed a larger method-dependence (Table 5).
Table 5. Population analyses. The first three entries refer to Mulliken charges for the respective method (def2-TZVP
basis for all calculations). The bottom entry shows charges from a QTAIM analysis based on the BP86 calculation of
the second entry.
Fe
N
O
Cl
Mulliken charge:
CASSCF
BP86
0.96 0.08 −0.35 −0.71
0.38 0.02 −0.16 −0.55
0.44 0.05 −0.19 −0.59
TPSSh
QTAIM (BP86) 1.16 0.32 −0.41 −0.71
With these values, the α spin excess on the iron atom is almost the same for the MNIC and the DNIC
species, leaving about half the β spin excess for the DNIC’s individual nitrosyl ligand compared to the
MNIC.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
22
Figure 11. Frontier orbitals of the [FeCl2(NO)2]− ion [CASSCF(9,14)/def2-TZVP; isovalue 0.06 a.u.]. Orbital
labels refer to Cartesian axes: z to the viewer, y up, x to the right, with a small tilt in favour of a better
visualisation; orbital numbering starts with “1” (= Orca numbering + 1). The 222210000 occupation pattern
indicated by the arrows refers to the ground state’s leading configuration (52% contribution). The metal
contribution is used to specify an MO. The antibonding MOs are shown which highlight, due to their additional
node, the contributing AOs. The bonding MOs ensue correspondingly.
Physics: the search for photo-induced linkage isomers
{FeNO}7(S = 3/2)-type MNICs: A[FeCl3(NO)] salts
UV-vis spectroscopy of the [FeCl3(NO)]− chromophore
The analysis of the UV-vis spectrum was a prerequisite for the photophysical investigation. Moreover,
the electronic spectrum of the [FeCl3(NO)]− chromophore should provide, in general, a deeper insight into
the electronic structure of a quartet-{FeNO}7-type compound for two reasons. First, no absorptions of
chromogenic co-ligands cover even parts of the spectrum. In terms of oxidation-state boundaries (see above),
neither the colourless [FeIICl4]2− nor the slightly yellow [FeIIICl4]− ions – these two species serving as an aid
to estimate the contribution of the chlorido part of the title ion – show stronger absorptions in the visible
range. Second, composed of only six atoms, the [FeCl3(NO)]− ion permits the computational analysis of
excited configurations with both DFT and WFT methods.
Experimental data were available for a Kohlschütter-type solution in concentrated hydrochloric acid,
for solutions in various organic solvents, and for solid samples with various counterions. Figure 12 shows a
typical solution spectrum, namely that of the tetraphenylphosphonium salt in acetone. Figure 13 shows the
Kubelka-Munk-transformed spectrum of the tetraethylammonium salt as a BaSO4-diluted powdered sample.
The coincidence of both spectra was satisfactory. In the solid-state spectrum, a small contribution at
16507 cm−1 was missing in the solution spectrum, and the region labelled “C” comprised a single Gauss
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
23
curve in the solution spectrum but two curves in the solid.
Figure 12. Gauss-deconvolved UV-vis spectrum of PPh4[FeCl3(NO)] in acetone. Note that the sum (thin black

line) of the individual Gauss curves (thin grey bell-shaped curves) coincides with the experimental points; ε( )
was fitted by seven Gauss functions; positions of the maxima in cm−1: 14030, 15683, 18125, 21054, 23531,
25233, 28183.
Figure 13. Gauss-deconvolved reflectance spectrum of solid NEt4[FeCl3(NO)], diluted with BaSO4. Note that the
sum (thin black line) of the individual Gauss curves (thin grey bell-shaped curves) coincides with the
−1

experimental points. K/S( ) was fitted by nine Gauss functions; positions of the maxima in cm : 13663, 14884,
16507, 17087, 20674, 22288, 23266, 25220, 27185.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
24
Computational analysis of the MNIC UV-vis spectra
The absorption bands A and B of the visible region are typical for high-spin-{FeNO}7 chromophores
in general, be it in an aqueous environment such as the ‘brown-ring’ complex or in an ionic liquid.[8b, 38]
Though the spectral features of the brown or green solutions are long known, an assignment of the
absorption bands has remained ambiguous. In fact, attempts to assign the bands solely by means of TD-DFT
calculations remained unsatisfactory due to the functional-dependence of the result, as has been experienced
by other workers on {FeNO}7-type chromophores.[39]
Fortunately, the existence of the [FeCl3(NO)]− ion as a chromophore made up of only six atoms also
allowed us to apply ‘expensive’ methods. The results of the study are collected in Table 6. Herein, the
experimental data are juxtaposed with computational values from a TD-DFT approach on the TPSSh/def2-
TZVP level of theory, as well as a CASSCF(9,13) calculation, state-averaged over 15 spin-allowed and 6
spin-forbidden roots, with a subsequent NEVPT2 treatment of dynamic correlation. A short note on the
choice of the active space should be made. In contrast to an adequate description of the ground state which
did not need the inclusion of the lone pair on the nitrogen atom, its inclusion was substantial for an
appropriate treatment of the excited states.
Our evaluation of the computational results started with the identification of the expected transitions
by means of the orbital levels of Figure 10. Excluding the stable MO 42 from those considerations, three
excitations each were expected in the α- and the β-channel.
Thus, β-spin transitions should occur from the degenerate Fe–NO bonds (MOs 44 and 45) (1) to the
degenerate metal-centred MOs of the xy-plane (MOs 46 and 47), (2) to the mainly metal-d(z2)-centred
MO 48, and (3) to the degenerate couple of Fe–NO antibonds (MOs 49 and 50).
Three α-spin transitions may be observed into the degenerate Fe–NO antibonds as the only acceptor
orbitals starting from (1) the degenerate xy-plane MOs 46 and 47, (2) the d(z2)-centered MO 48, and (3) the
Fe–NO bonds. These transitions are summarised in the second column of Table 6.
Table 6. ‘Label’ column: label of the transition as in Figure 12 and Figure 13. ‘Transition’ column: b, ab and x|y as
defined in Figure 10; 4b and 6ab refer to a spin-forbidden excitation from the quartet ground state to a sextet excited
−1

state. ‘WFT’ (‘Wave Function Theory’) column:  /cm (in parentheses: oscillator frequency in atomic units),
NEVPT2(9,13)/def2-TZVP, 15 quartet and 6 sextet state-averaged roots, CPCM (ε = ∞). ‘DFT’ column: TD-DFT, fosc as
before, TPSSh/def2-TZVP, CPCM (ε = ∞). The ‘solution’ and ‘solid’ columns refer to the experiments of Figure 12 and
Figure 13.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
25
label transition
WFT (fosc/10−5) DFT (fosc/10−5) solution solid
16010 (forb.)
16016 (forb.)
16378 (0)
16692 (27)
16842 (68)
16886 (37)
20424 (167)
20430 (167)
23106 (7)
to A 4b  6ab
–
to A1–3
15405 (0)
15721 (13)
17086 (13)
17100 (13)
22408 (19)
22437 (20)
20712 (2)
20776 (2)
22756 (0)
25493 (11)
25448 (9)
25981 (78)
25977 (79)
14030 13663
15683 14884
18125 17087
b β x|y
A1–3
b β z2
B
21054 20674
22288
23531
b β ab
C
23114 (6)
23166 (0)
23266
25225 (135)
25231 (136)
z2 α ab
D
25233 25220
28183 27185
minor part in
b α ab:
26533 (102)
26571 (105)
x|y α ab
E
b α ab
29855 (348)
27710 (96)
In the wave-function approach, most of the calculated transitions were assigned to the expected
excitations. No clear contribution to the spectra was obtained for the expected x|y-to-Fe–NO-antibond
excitation which was found only as a minor constituent of the strongest transition, the Fe–NO-bond-to-
antibond excitation in the α-channel. Another point was the early onset of measured absorption on the low-
energy end of the spectrum which was hardly mirrored in the calculation. To contribute to this point, we
added spin-forbidden quartet-to-sextet excitations to the spin-allowed quartet-to-quartet transitions. As a
result, the transition from the global to the sextet ground state occured at a slightly lower energy and may be
considered a contribution to the experimental spectra.
In a TD-DFT approach using the hybrid functional TPSSh, the excitations starting from the x|y MOs
were resolved as individual contributions. However, in terms of transition energy, the B and C absorptions
occured in reversed order, though their intensities showed that the wave-function assignment should be
correct.
As a result of this part of the study, each of the transitions weaken the Fe–NO bond and thus may
initiate an alteration of the nitrosyl’s bonding situation, be it a PLI event or the simple photolysis of the Fe–
NO bond. In agreement with these assignments, the PLI events described below were typically initiated by a
660 nm laser source which irradiated into the low-energy band A.
This result relies on a specific prerequisite, namely the absence of chromogenic coligands. The
property is shared between the title compounds and the aqua-ligated, light-sensitive ‘brown-ring’
chromophore.[38a] In contrast to our study, the resistance of a nitrosyl ligand to NO release on irradiation may
result if absorbing coligands are present.[40]
Metastable isonitrosyl and the search for side-on nitrosyl isomers
The detailed knowledge of the PLI pathways in the well-understood SNP case provided us with a
guide to what may have been expected at most in a PLI study on [FeCl3(NO)]−. Of the two known PL-
isomers that can be generated in an SNP crystal, the isonitrosyl (MS1) lies ca. 2 eV above the ground state
(GS), whereas the side-on MS2 isomer is ca. 1.5 eV unstable with respect to the GS. The bonding situation in
both isomers may be derived from GS bonding. In the ground state’s overall low-spin, S = 0, configuration,
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
26
the two electron pairs in Fe-d(xz) and Fe-d(yz) (z along the Fe–N–O axis) donate into the NO-π*(x) and the
NO-π*(y) of the nitrosyl ligand – as in our case above. The simple rotation of the NO ligand in the drawing
plane by 90° steps generates alternative, symmetry-allowed Fe–NO overlap. In total, the four-lobe pattern of
an NO-π* antibond allows for four interactions: the GS at 0°, two side-on isomers for 90 and 270° (MS2)
and a linear Fe–O–N-moiety (MS1, isonitrosyl) for the 180° turn (Figure S11 in the SI). Performing this
rotation quantitatively in a relaxed scan, the global minimum close to 0° rotation (Fe–N–O = 180°) for the
GS and the local MS1 and MS2 minima are supplemented with transition states between them which define
activation barriers for the MS→GS relaxation of almost 1 eV.[2c]
To identify the hoped-for new IR bands arising in the course of PLI-directed experiments on the title
compounds, we started our investigation with an angle-scan for the [FeCl3(NO)]− ion. Figure 16 (red circles)
shows the result which was obtained on the BP86/def2-TZVP level of theory. Approximately 1.2 eV above
the GS, an isonitrosyl, MS1, species defines a local minimum. Contrary to SNP, only a single transition state
has to be surmounted on the entire GS→MS1 path, and no indication was found for a second local minimum
of a side-on, MS2, state. Moreover, the activation barrier for the MS1→GS relaxation amounted to only
0.3 eV and thus was substantially lower than for SNP. The IR signature of the MS1 species, as a guide to the
experimental work, was determined on the same level as = 1598 cm−1 for the O–N stretch. An important


note has to be made: in order to adequately compare energies referring to two substances (our MNIC and
SNP), we used numbers from DFT calculations in both cases. Experimental values for the energies of the
PL-isomers as well as activation energies are available for SNP and are usually substantially smaller than the
calculated ones.[41]
The population of the isonitrosyl (MS1) state in solid A[FeCl3(NO)] salts
The determining factors in the search for linkage isomers of the [FeCl3(NO)]− ion arose out of the DFT
scan. An MS1 state appeared to be tangible, though the activation barrier for its decay to the ground state was
unusually low. The experimental setup thus included the lowest accessible temperatures. Fortunately, the
available equipment allowed for cooling to 9 K. At this temperature, the expected IR frequency for an MS1
isomer of about 1600 cm−1 was detected for various salts of the formula A+[FeCl3(NO)]−, namely for A+ =
NMe4 , NBnMe3 , [Co(cp)2]+, PPh4 , AsPh4 , and Mephaz+ (5-methyl-phenazinium). For these cations,
irradiation with a 660 nm laser source, i.e., irradiation into band A of Figure 12, gave rise to IR absorptions
of 1627/1635, 1550, 1572, 1571, 1580, and 1575 cm−1, respectively. After the population by irradiation, the
MS1 relaxed in the course of about 2 to 15 minutes to the ground state, depending on the counterion.
+
+
+
+
An overview of a PLI experiment is shown in Figure 14 with NMe4[FeCl3(NO)] as an example. The
progress of the relaxation of the MS1 over a period of time of 4 min is depicted in Figure 15.
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
27
Figure 14. An overview over the PLI experiment on NMe4[FeCl3(NO)]. At 9 K, the Fe–NO valence vibration is
slightly split. On irradiation with a 660-nm laser, a new small absorption arised at about 1630 cm−1. The
relaxation of this band on leaving the sample un-irradiated at 9 K is shown in Figure 15.
Figure 15. The relaxation of the excited state of NMe4[FeCl3(NO)] after irradiation with 660-nm laser light at
9 K in the course of about 4 min.
Peculiarities of quartet-{FeNO}7 PLI
After the successful match of DFT analysis and experiment, a closer look at the differences between
the electronic structure of the SNP anion and the [FeCl3(NO)]− ion seemed essential. A notable detail,
however, which facilitates the discussion should be mentioned first.
To radically simplify the theoretical treatment of a {FeNO}n species, one could omit all coligands. For
the {FeNO}7(S = 3/2) species in question, Reiher used that approach to analyse the triatomic [Fe(NO)]2+ ion
to allow for high-level methods.[42] If his restriction (four point charges around the central metal and a fixed
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
28
Fe–N–O angle) is lifted, the frontier orbitals of a linear [Fe(NO)]2+ ion emerge to (1) a couple of doubly
occupied Fe–NO π-bonding levels, (2) a couple of singly occupied metal-only orbitals labelled d(xy) and
d(x2−y2) if the Fe–N–O axis is chosen as z, and (3) the singly occupied, metal-centered Fe-d(z2)–NO σ
antibond as the HOMO. In other words, the C3-symmetrical arrangement of three coligands, as in the
[FeCl3(NO)]− ion, shares the level splitting scheme of the simple triatomic core.
Figure 16. Relaxed surface scans using the Fe–N–O angle as the scan variable. Red circles: the quartet-{FeNO}7
ion [FeCl3(NO)]−; grey circles: the triplet-{FeNO}6 species [FeCl3(NO)]; crosses: the triatomic quartet-{FeNO}7
ion [Fe(NO)]2+. Energies apply to the respective ground-state energy set to 0.
The similarity of the electronic configuration of the [FeCl3(NO)]− ion to its [Fe(NO)]2+ core is
underlined by a Fe–N–O angle scan for the triatomic species as well (crosses in Figure 16). As a common
feature, a local MS2 minimum is absent for both species. In contrast, the MS1→GS barrier is markedly
higher for the triatomic core.
To understand the key point, which is the absence of an MS2 state, a comparison of the configurations
of the SNP anion and the quartet-{FeNO}7 species was useful. In SNP, the octahedral environment of the
central metal with strong-field ligands safely fixes the d-orbital alignment (for a closer discussion of the
necessary conditions for a PLI event see Ref. [43]). Thus, for example, in the xz plane of the SNP anion,
overlap between Fe-d(xz) and NO-π*(x) is clearly maximised for the above-mentioned 90° jumps only. To
stay with the example xz plane, the metal orbitals are effectively fixed, in the quartet-{FeNO}7 species, only
by the nitrosyl strong-field ligand. If the nitrosyl ligand starts to rotate, the metal orbitals follow. Due to the
increasingly imperfect overlap, the energy rises smoothly on rotation. The energetic landscape is thus much
flatter than in the case of fixed metal orbitals. An alternative view may be developed by looking for orbital
mixing. Mixing the d(xz) orbital and the d(z2−x2) part of a d(z2) = d(2z2−x2−y2) orbital allows the adjustment
of this pair of four-leaf-clover-shaped orbitals at any angle in the xz plane.
The occupation of this orbital pair with three electrons in the quartet-{FeNO}7 species might be
regarded as crucial. In the {FeNO}6 case SNP, d(xz) and d(z2) are not only spatially fixed. Moreover, the
former is doubly occupied, and the latter is empty. In order to check the significance of a singly occupied Fe-
d(z2) orbital in the {FeNO}7 species, a scan was performed for a tentative species that lacked this specific
This article is protected by copyright. All rights reserved.

10.1002/chem.201804565
Chemistry - A European Journal
29
electron, i.e., a triplet [FeCl3(NO)] molecule with the same configuration as in Figure 10 but with MO48
empty. The result is shown in Figure 16 (grey circles). In fact, a dent becomes visible along the GS→TS
path, but no distinct local MS2 minimum. (A final fourth scan on the tentative species’s [Fe(NO)]3+ core was
hampered since this threefold-charged species dissociated to Fe2+ and NO+ on the chosen level of theory.)
As a result, we conclude that the distinct fixing of a central metal’s d electrons is a prerequisite to
catch all the expected photoinduced linkage isomers of a nitrosylmetal complex. The depopulation of the σ-
antibonding HOMO of a {FeNO}7 species supports the possibility of stabilising an MS2 state but is not a
sufficient criterion. To lend significance to the peculiarities of SNP and other MS2-producing systems, it
should be noted that it is not only a high crystal-field-splitting that creates a diamagnetic low-spin ground
state with both well-fixed orbitals and an empty d(z2), but the symmetry of the entire species is decisive.
Thus, any {FeNO}6 species with degenerate or close-to-degenerate d(xy)/d(x2−y2) orbitals will lack a
diamagnetic S = 0 state irrespective of the coligands’ strength (see the next paragraph).
The population of an anion-to-cation charge-transfer state
PPN[FeCl3(NO)] was a special case. For the PPN+ counterion, a similar IR-spectroscopic signature
was observed in terms of population and decay behaviour as for the MS1-positive samples at first glance. At
9 K, exposure to light of 635 nm wavelength, the N–O stretch at 1809 cm−1 was depopulated and, at the same
time, a new band arose at an unusually high wave number of 1868 cm−1 (Figure 17). Thus, the frequency of
the newly observed N–O stretch did not match the typical red shift of an MS1 (or MS2) state. Warming
irradiated samples to 40 K made the new absorption vanish. At 9 K, the new state relaxed in the course of
about an hour, i.e., much more slowly than the MS1 described above, to the ground state without intensity
loss of the latter.
Figure 17. Population at 635 nm and subsequent relaxation of small amounts of a tentative one-electron transfer
couple PPN• + [FeCl3(NO)] in solid PPN+[FeCl3(NO)]−. Note the decrease and the re-constitution of the major
ground-state N–O stretch, and the population and relaxation of the excited state at 1868 cm−1.
An approach to the nature of the excited state was made by means of the DFT screening of a number
of tentative structurally and electronically excited states. As a result, two matches were found, both being
This article is protected by copyright. All rights reserved.