#### 2.4. Magnetic Properties

Magnetic measurements were performed by using a Quantum Design MPMS-5-XL SQUID magnetometer. Static magnetic susceptibility χ(T) was measured on polycrystalline samples at the magnetic field H = 100 Oe, while warming (↑) and cooling (↓) regimes. The temperature range of measurements was 2–300 K. The magnetization curves M(H) were obtained during several loops over the field range +50 kOe to −50 kOe. Preliminary, the samples had been cooled down to 2.0 K in zero magnetic field and virgin curves were recorded.

Temperature dependences of the product

χT↓ for complexes

**1** and

**2** are depicted in

Figure 9. The measurements were performed on fresh crystals while cooling in the range of 300→2 K,

H = 100 Oe. The

χT↓ value of

**1** starts at 4.26 cm

^{3} K/mol, remains unchanged till 50 K, undergoes the inflection at

T_{i1} = 46 K, and then decreases continuously down to 2.79 cm

^{3} K/mol. Polycrystalline powder of

**2** demonstrate a sharp SCO transition at

T_{sco} = 118 K, during which the

χT↓ value changes from 4.25 cm

^{3} K/mol at 300 K (HS) down to 1.07 cm

^{3} K/mol at the plateau in the vicinity of 60 K (LS). The shape of the transition curve was reproduced in several same-day subsequent measurements. Evolution at longer time exposures is discussed below. Hysteresis behavior was not observed between

χT↓ and

χT↑ curves in the transition range. In the vicinity of

T_{i2} = 55 K there was also a slightly pronounced inflection below which

χT decreases down to 0.51 cm

^{3} K/mol. The

χT values at 300 K of both compounds are consistent with the theoretical spin-only value of 4.375 cm

^{3} K/mol corresponding to 100% amount of HS Fe(III) ions,

S = 5/2,

g = 2.0. This is also in agreement with the assertion that both anions Au(dmit

_{2})

^{−} and Au(dddt

_{2})

^{−} do not have magnetic moment.

Magnetic properties of both compounds changed with aging and thermal history, revealing an irreversible behavior of

χT while cooling and warming regimes. An uncomplicated behavior for

**1** is shown in

Figure 10. Irreversibility is observed below the inflection point,

T_{i1} = 46 K, in a form of a hysteresis loop, while above it the curves follow the same Curie-Weiss trend with

C = 4.3 cm

^{3} K/mol (

S = 5/2) and Θ = −0.5 K (see

Figure 10, inset). This shape was reproduced in several cycles and it did not depend on the age of the sample. System

**2** was found unstable. Both aging at ambient conditions and thermocycling led to a degradation of the SCO transition. The curves in

Figure 11 present this evolution: the initial curve (black color, ↑↓ regimes) was recorded on the crystals aged over a period of several months; the other data sets depicted in blue and purple colors show a character of SCO degradation after several subsequent cooling-warming cycles. A solid red line indicating the transition in freshly synthesized sample (see

Figure 9) is given for reference. With aging the HS state was reached in a more gradual manner, so that the

χT value only got 3.8 cm

^{3}K/mol at 300 K. The value

χT on the plateau below

T_{sco} corresponds to

γ_{LS} = 78% of

S = 1/2 and

γ_{HS} = 22% of

S = 5/2, where

γ_{LS} and

γ_{HS} are molar LS and HS fractions. We took 0.375 cm

^{3}K/mol and 4.375 cm

^{3}K/mol for 100 % of LS and HS fractions, respectively. The ratio

γ_{LS}/

γ_{HS} did not change with time exposure, and varied from 3.6 (fresh) to 6.1 from sample to sample. However, thermocycling of seasoned crystals led to a quick spreading of the transition and to an irreversible growth of the residual HS fraction. In subsequent warming-cooling cycles

γ_{HS} changed from 14% to 32% and 41% respectively.

Magnetization curves,

M(

H), are presented in

Figure 12 and

Figure 13 for

**1** and

**2**, respectively. The data points in

Figure 12 reveal a weak hysteresis with the coercive force

H_{c} = 13 Oe and the remnant magnetization value

M_{r} = 4 × 10

^{−3} μ

_{B}. Saturation at highest fields was not observed. However, a combination of the Brillouin function for

S = 5/2 with a magnitude factor 0.9 (black line),

M_{s} = 4.7 μ

_{B}, and linear function with the factor

k = 0.04 μ

_{B}/kOe gave a perfect fitting. Linear contribution is conceivable in terms of AFM interaction between HS Fe(III) cations. Magnetization curves for the AFM coupled spin system, including dimers, often show Brillouin dependences and a gradual growth proportional to the field strength when |

J_{1}| <<

kT,

T = 2 K [

27]. The magnetization curve of

**2** depicted in

Figure 13 also reveals hysteresis,

H_{c} = 50 Oe and

M_{r} = 3 × 10

^{−3} μ

_{B}. The value

M at 50 kOe reaches 1.14 μ

_{B}. Field dependence was perfectly fitted by Brillouin function for

S = 1/2 and magnitude factor 1.23 (solid line). The data points could not be fitted by a combination of Brillouin functions for

S = 1/2 and

S = 5/2 regardless the weight factors. A surplus contribution of 0.23 μ

_{B} remains unclear. Whereas the absence of contribution from HS Fe(III) cations could be explained by stronger AFM coupling, |

J_{2}| >>

kT. Brillouin shape indicates that the local moments of LS Fe(III) ions remain non-interacting.

A comparison of the χT evolution in **1** and **2** with their structural characteristics gives rise to several essential notions:

Results of our magnetic measurements agree well with the structural features of the compounds. Complex

**1** has a rigid structure which is characterized by extensive

π-interactions of both salicylidene moieties of the SCO cation with nearest neighbors, one with the cation and another with the anion. This makes changing conformation of [Fe(3-OMesal

_{2}-trien)]

^{+} unit at the SCO impossible and explains absence of the latter in

**1**. Similar situation was discussed for one of the polymorphic phases of [Fe(sal

_{2}-trien)][Ni(dmit)

_{2}] which also does not show SCO at cooling [

28]. In complex

**2**, the cation layer includes solvent molecules,

π-stacking is absent and crystal packing is controlled mainly by weak hydrogen bonding allowing conformation freedom necessary for SCO. We believe this view is also consistent with the results of [

28,

29,

30].

Although a significant loss of acetonitrile molecules was observed in the 145–200 °C range (

Figure S3), the desolvation was evidently taking place at ambient conditions. These small changes significantly affect the configuration of the ligand and therefore the spin transition [

25,

31,

32,

33]. We believe, solvent vacancies in the seasoned crystals

**2**, as well as the defects, provoke stabilization of the HS configuration in their vicinity. Thermocycling stimulates growth of the HS fraction by promoting a diffusion of the solvent molecules out of the bulk because the sample is placed in the chamber with helium atmosphere. In several cycles the remnant HS fraction in aged crystals

**2** rose from 14% to 40% (

Figure 11). It is worth noting that up to 20% HS fraction was found in the original fresh samples (depending on batch). This either relates to a presence of the HS conformers [

26] or is likely deals with disproportion between [Fe(3-OMesal

_{2}-trien)] and [Au(dddt)

_{2}], when the metal dithiolene complex might become a neutral radical [Au(dddt)

_{2}]

^{•}. The former was not confirmed by our structural data. Presence of neutral radical is possible since the dddt ligand can carry a various charge form [

11]. The interaction of the local magnetic moment of the SCO complex with the radical in the anion counterpart was observed recently [

14]. The additional spin contribution from the radicals (~7%) could hardly be distinguished on the top of the total magnetic response in SQUID measurements. However, a spin density on the dithiolene ligand may substantially enhance the superexchange coupling (

J_{2}). This assumption will be verified in forthcoming electron paramagnetic resonance (EPR) experiments.

In both compounds the

χT curves revealed a characteristic inflection. Below the inflection points

T_{i1} = 46 K (

**1**) and

T_{i2} = 55 K (

**2**)

χT continuously declined. We fitted experimental data by using julX software [

34]. The AFM coupled HS Fe(III) dimers were considered as a model system. The best fit parameters for

**1** were

g_{i1} =

g_{i2} = 2.0,

D_{i1} =

D_{i2} = 0,

J_{1} = −0.18 cm

^{−1} (0.26 K). It is worth noting that even a very small ZFS factor such as for example

D_{i} = −0.1 cm

^{−1} strongly distort the calculated

M(

H) curve. We did not observe significant deviation from the Brillouin shape in

Figure 12. Therefore, we took

D_{i} = 0 in our fitting. The other interesting feature of the fitting was a magnitude factor of 0.9. This might mean that only ~90% of HS Fe(III) moments was effectively enough for the description of the

χT↓ curve below the inflection point. A noticeable portion of the moments Fe(III) ions “disappeared” out of the total magnetic response. This is a signature of weak AFM interactions.

The value |

J_{1}| = 0.26 K for

**1** is indeed lower than 2 K, at which the magnetization curves were measured. It is not surprising then that the calculated

M(

H) line fits the Brillouin-like behavior in the experiment (

Figure 12) and, at the same time, calculated

χ(

T) describes the decline of

χT↓ in experiment.

Figure 14 shows a low temperature part of the experimental data set of

**1** and two fitting lines: fitting by a Curie-Weiss law with Θ

_{1} = −1.7 K and

C = 4.0 cm

^{3} K/mol, and quantum calculations in the model of AFM coupled dimers with S = 5/2 and

J =

J_{1}. In cases when every magnetic ion has the same number and the same kinds of interactions, the Weiss temperature can be expressed via pair exchange constants

J_{i,j} in the usual form found in the literature:

where

z_{ij} is the number of

j neighbors of the

ith atom. For

**1** Θ and

J_{ij} =

J_{1} match at

Z ≈ 1.1, which is close to the dimer case and may speak in favor of AFM interaction between the adjacent SCO complexes linked by the

π–

π bonding of their aromatic groups.

For the fresh crystals of

**2** we estimated ~20% of the residual HS fraction. This contribution was extracted from the low temperature plateau (

T < 70 K) of the experimental data set

χT↓ on

Figure 9. Then we recovered its value to the virtual 100% HS phase, attributing a low temperature decline to the AFM coupling in this HS fraction. The data after the treatment are shown in

Figure 14. The best fit curve by Curie-Weiss law, Θ

_{2} = −16 K, is shown as the solid black line. Attempts of describing it within the uniform quantum model of AFM coupled moments failed. A satisfactory agreement was reached by using a model of two linked dimers and two coupling constants,

J. A solid red line shows the best fit curve for |

J_{23}| = 2.46 K and |

J_{12}| = |

J_{34}| = 0.43 K (

g_{i} = 2.0,

D_{i} = 0). Although there are no structural dimers in

**2**, the magnetic dimer is a reasonable approximation of the short-range coupling. As soon as we do not have an idea about the morphology of the HS fraction this model might seem to a certain extent speculative. Nevertheless, it brings an understanding that the much stronger exchange interaction came to the stage. Therefore, in

**2** additionally to a routine weak coupling between the adjacent iron complexes there comes a stronger coupling which apparently involves the gold dithiolene complexes as well. In other words, structural defects and, possibly, solvent vacancies may serve not only as a source of the residual HS fraction, but also as a promoter of stronger exchange interactions between respective HS complexes. If the defects cause a charge disproportionation, then the appearance of neutral radicals [Au(dddt)

_{2}]

^{•} would facilitate stronger exchange coupling via the anion sublattice. This also suggests an additional spin contribution from Au(dddt)

_{2}, ~10% of

S = 1/2 (

χT ≈ 0.04 cm

^{3} K/mol). Evidently, the inflections points

T_{i1} = 46 K and

T_{i2} = 55 K also reflect changes in metal dithiolene subsystem. A comparison of two above approaches by using (1) gives:

which matches at

Z = 0.9. This means that one pair of stronger coupled

S = 5/2 moments (

J_{23}) and two linked pairs of weaker coupled ones (

J_{12},

J_{34}) is enough to describe the decrease of

χT below the SCO transition in

**2**. It is worth noting that anion driven modulation [

11,

33,

35] may modify the intermolecular interactions but it cannot facilitate the exchange coupling via a network of hydrogen bonds.

Of course, a model of two strongly coupled dimers plays a merely symbolic role linking together a weak coupling of the same nature as in **1**, and stronger coupling that is, in our understanding, a quantitative characteristic of the involvement of the electronic states of dithiolene into superexchange interactions between HS Fe(III) magnetic moments.