Iron-Catalyzed Oxidative C-O and C-N Coupling Reactions Using Air as Sole Oxidant.

We describe the oxygenation of tertiary arylamines, and the amination of tertiary arylamines and phenols. The key step of these coupling reactions is an iron-catalyzed oxidative C-O or C-N bond formation which generally provides the corresponding products in high yields and with excellent regioselectivity. The transformations are accomplished using hexadecafluorophthalocyanine-iron(II) (FePcF16 ) as catalyst in the presence of an acid or a base additive and require only ambient air as sole oxidant.

Introduction

The construction of carbon−heteroatom bonds is a crucial transformation for the synthesis of functionalized heterocyclic moieties present in natural products and pharmaceuticals. Transition metal catalysts have been widely applied for these coupling reactions. The best known method to generate C−O bonds for diaryl ethers is the copper‐catalyzed Ullmann‐Goldberg reaction (Scheme 1a). The palladium‐catalyzed Buchwald‐Hartwig amination of aryl halides represents a useful tool for C−N bond formation (Scheme 1b).[ , ] These methods require pre‐functionalized substrates, thus limiting their utility from an economic and environmental point of view. Therefore, considerable efforts have been made to achieve a direct C−H bond functionalization by intermolecular oxidative C−O and C−N coupling reactions. A key challenge of the oxidative coupling is the fact that C−H bonds are relatively inert and thus, several groups have applied strong oxidants for this process.[ , ] However, these procedures often suffer from low atom‐economy, poor selectivity, and harsh reaction conditions. Using green oxidants, the oxidative coupling becomes more attractive. Thus, oxygen from ambient air represents the ideal oxidant due to its natural abundance and environmentally friendly features.

Scheme 1

Transition metal catalyzed C−O and C−N coupling reactions.

Inspired by green and sustainable chemistry, transition metal‐catalyzed oxidative coupling reactions have become a useful and versatile tool for the formation of carbon–heteroatom bonds with air as sole oxidant. Especially copper[ , ] and noble metals have been applied successfully for this purpose. More recently, the search for inexpensive, less toxic, and thus environmentally benign and sustainable catalysts has come into the focus of interest. Iron compounds are in the center of this development. In this context, iron‐based catalysts in combination with stoichiometric amounts of strong oxidants have been extensively studied for oxidative carbon–heteroatom coupling reactions. However due to the hazard of the oxidant, these methods are not a fully sustainable alternative for such conversions. In contrast, iron‐catalyzed oxidative carbon−heteroatom coupling reactions using oxygen from ambient air as oxidant represent green synthetic transformations.

Recently, we have developed an oxidative C−C homocoupling of anilines[ , ] and hydroxycarbazoles, as well as an oxidative C−C cross‐coupling of tertiary anilines with hydroxyarenes using hexadecafluorophthalocyanine−iron(II) (FePcF16, Figure 1)[ , ] as catalyst with air as sole oxidant. The iron‐catalyzed reaction of 2‐(dimethylamino)naphthalene with 2‐naphthol afforded the oxidative C−C cross‐coupling product (62 % yield) along with a diaryl ether (11 % yield) resulting from an oxidative C−O coupling reaction. This previous result prompted us to study the feasibility of novel iron‐catalyzed oxidative C−O and C−N coupling reactions. Herein, we report efficient oxygenations and aminations of tertiary arylamines and aminations of phenols in the presence of FePcF16 as catalyst and air as sole oxidant at room temperature. The presence of an appropriate additive (Lewis acid, Brønsted acid, or base) is crucial for these mild oxidative carbon−heteroatom bond formations (Scheme 1c).

Figure 1

Structure of hexadecafluorophthalocyanine–iron(II) (FePcF16).

Results and Discussion

Based on our protocol for the iron‐catalyzed oxidative C−C cross‐coupling, we investigated the iron‐catalyzed oxidative C−O coupling of 2‐(dimethylamino)naphthalene (1  a) and 4‐methoxyphenol (2  a) to the diaryl ether 3  a in dichloromethane as model reaction (Table 1). No product was formed using catalytic amounts of FePcF16 (4 mol%) without additive at room temperature under ambient air (entry 1). Variation of the additives from acetic acid (entry 2) to methanesulfonic acid (entry 3) showed that the stronger Brønsted acid led to the diaryl ether 3  a in 25 % yield along with the dioxygenated product 4  a in 31 % yield (entry 3). Lewis acids proved to be superior providing selectively the diaryl ether 3  a as major product (entries 4 and 5). An optimized protocol with 40 mol% of BF3 ⋅ OEt2 as additive and air as oxidant afforded compound 3  a in 66 % yield (entry 5). In comparison, the recently reported oxidative coupling of 2‐(dimethylamino)naphthalene (1  a) with various phenols using a chromium−salen catalyst under aerobic conditions afforded mixtures of C−C and C−O coupling products. Variation of the reaction conditions given in entry 5 of Table 1 gave no further improvement. For example, the very slow reaction in ethanol as solvent afforded 3  a in only 17 % yield after 24 h (entry 6) and the reaction at 0 °C provided 3  a in only 36 % yield (entry 7). Moreover, the reaction of entry 5 under an atmosphere of pure oxygen resulted in complete decomposition of the phenol 2  a within 12 minutes. Whereas using Fe(acac)3 as catalyst gave no turnover and allowed recovery of the starting materials 1  a and 2  a.

Table 1

Optimization of the reaction conditions for the iron‐catalyzed oxidative C−O coupling of 2‐(dimethylamino)naphthalene (1  a) with 4‐methoxyphenol (2  a).[a]

Entry	Additive (mol %)	Time	Yield of 3  a [%]	Yield of 4  a [%]
1	–	24 h	–	–
2	AcOH (200)	24 h	traces	–
3	MsOH (40)	2 h	25	31
4	B(C6F5)3 (40)	50 min	53	traces
5	BF3 ⋅ OEt2 (40)	40 min	66	traces
6[b]	BF3 ⋅ OEt2 (40)	24 h	17	–
7[c]	BF3 ⋅ OEt2 (40)	3.5 h	36	–

[a] Reaction conditions: 1  a (0.30 mmol), 2  a (0.38 mmol), FePcF16 (4 mol%), air (1 atm), CH2Cl2 (3 mL), room temperature. [b] Solvent: EtOH (3 mL). [c] Reaction at 0 °C. FePcF16=hexadecafluorophthalocyanine–iron(II).

–

–

–

–

–

–

The iron‐catalyzed oxidative C−O coupling of 2‐(dimethylamino)naphthalene (1  a) with a range of phenols 2  a–e using the optimized reaction conditions provided selectively the corresponding diaryl ethers 3  a–e (Table 2). Coupling of 1  a with 1‐methyl‐2‐naphthol (2  d) led to the dinaphthyl ether 3  d in 89 % yield. The structure of 3  d was additionally confirmed by an X‐ray analysis (Figure 2). The coupling of 1  a with 1‐bromo‐2‐naphthol (2  e) required 40 mol% of methanesulfonic acid as additive to afford the dinaphthyl ether 3  e in 57 % yield. Coupling of N,N‐dimethylaniline (1  b) with 4‐methoxyphenol (2  a) provided compound 3  f in 35 % yield. The reaction with primary and secondary arylamines did not lead to product formation.

Table 2

Iron‐catalyzed oxidative C−O coupling of tertiary arylamines 1 with phenols 2.[a]

[a] Reaction conditions: 1 (0.30 mmol), 2 (0.38 mmol), FePcF16 (4 mol%), BF3 ⋅ OEt2 (40 mol%), air (1 atm), CH2Cl2 (3 mL), room temperature. [b] 2 (0.76 mmol). [c] 2  e (0.75 mmol), FePcF16 (4 mol%), MsOH (40 mol%).

Figure 2

Molecular structure of the dinaphthyl ether 3  d in the crystal (thermal ellipsoids are shown at the 50 % probability level).

The reaction is assumed to proceed via a free‐radical coupling mechanism (Scheme 2). Single‐electron transfer (SET) oxidation of the phenol by an iron(III) species and subsequent proton loss of the radical cation generate a phenoxy radical which attacks the Lewis acid−Lewis base complex of the tertiary arylamine and boron trifluoride. Cleavage of the boron trifluoride from the resulting cyclohexadienyl radical, oxidation by another SET to a cyclohexadienyl cation, and proton loss afford the diaryl ether. The presence of boron trifluoride as additive appears to be crucial for the success of this oxidative C−O coupling since in the presence of Brønsted acid the C−C cross coupling generally is preferred.[ , ]

Scheme 2

Proposed mechanism for the iron‐catalyzed oxidative C−O coupling of tertiary arylamines with phenols.

In contrast to the Buchwald‐Hartwig amination, the oxidative C−H/N−H coupling represents a more direct and atom‐economical method. Phenothiazines, compounds with useful biological and optoelectronic properties, have been applied recently for oxidative C−N coupling reactions by Patureau et al. and other groups. The resulting N‐arylated phenothiazines exhibit interesting physical properties.

The iron‐catalyzed oxidative C−N coupling reaction of 2‐(dimethylamino)naphthalenes 1 with the phenothiazines 5  a–d afforded the corresponding N‐arylated products 6  a–e in good to excellent yields (Table 3). An X‐ray crystal structure determination of 6  a confirmed the N‐(1‐naphthyl)phenothiazine framework (Figure 3). The optimized reaction conditions for this reaction (4 mol% of FePcF16 and 60 mol% of methanesulfonic acid as additive in THF at room temperature under air) closely resemble those previously used for the iron‐catalyzed oxidative C−C cross‐coupling of tertiary anilines with hydroxyarenes. The blank experiment, reaction of phenothiazine (5  a) with 1  a in the absence of the catalyst FePcF16, afforded compound 6  a in only 11 % yield after 3 h. The coupling of 2‐(dimethylamino)naphthalene (1  a) with phenoxazine or dibenzo[b,f]azepine provided the corresponding products 6  f and 6  g in high yields, whereas the reaction of N,N‐dimethylaniline (1  b) with phenothiazine (5  a) using BF3 ⋅ OEt2 (40 mol%) as additive gave a moderate yield for 6  h.

Table 3

Iron‐catalyzed oxidative C−N coupling of tertiary arylamines 1 with compounds 5.[a]

[a] Reaction conditions: 1 (0.30 mmol), 5 (0.51 mmol), FePcF16 (4 mol%), MsOH (60 mol%), air (1 atm), THF (3 mL), room temperature. [b] 1  b (0.30 mmol), 5  a (0.60 mmol), FePcF16 (4 mol%), BF3 ⋅ OEt2 (40 mol%), air (1 atm), CH2Cl2 (3 mL), room temperature, 15 h.

Figure 3

Molecular structure of the N‐(1‐naphthyl)phenothiazine 6  a in the crystal (thermal ellipsoids are shown at the 50 % probability level).

Following the excellent results achieved with phenothiazine (5  a) as coupling substrate, we have extensively investigated the iron‐catalyzed oxidative amination of phenols using the reaction of 5  a with 4‐(tert‐butyl)phenol (2  f) as model system (Table 4). Various iron compounds as catalysts (4 mol%) and several different additives have been tested for the reaction of 2  f with two equivalents of 5  a at room temperature under ambient air. Simple iron salts as catalysts gave no conversion and 5,10,15,20‐tetrakis(pentafluorophenyl)porphyrin–iron(III) chloride led only to traces of the C−N coupling product 7  a (entries 1–3). However, the complex tris(dibenzoylmethanato)iron(III) (Fe[dbm]3), phthalocyanine–iron(II) (FePc), and hexadecafluorophthalocyanine−iron(II) (FePcF16) afforded compound 7  a in 15–27 % yield (entries 4–6). The high catalytic activity of FePcF16 for various oxidative transformations has been widely demonstrated by our group.[ , , , , , , ] Subsequently, we have screened a range of different additives for the oxidative C−N bond formation using 4 mol% of FePcF16 as catalyst (entries 7–13). With methanesulfonic acid as additive (entry 7), complete decomposition of the starting material was observed. Among the different bases tested as additives (entries 8–13), 40 mol% of Hünig's base (N,N‐diisopropylethylamine) was most beneficial and afforded the C−N cross‐coupling product 7  a in 94 % yield (entry 13). In this context, it is interesting to note that Hünig's base was also used as additive for the iron‐catalyzed oxidative N−N bond formation of diarylamines. These optimized conditions for the iron‐catalyzed oxidative C−N coupling could easily be applied to large scale reactions. Thus, a reaction of 300 mg of phenol 2  f provided 7  a in 93 % yield and a gram‐scale synthesis led to 90 % yield of 7  a (entry 14). The first step in FePcF16‐catalyzed reactions under air is the oxidation of the catalyst to μ‐oxobis[(hexadecafluorophthalocyanine)iron(III)] (O[FePcF16]2). Thus, using μ‐oxo(FePcF16)2 as catalyst under otherwise identical conditions gave a similar yield of 7  a (entry 15). In line with our mechanistic hypothesis of a catalytic [Fe(III)]/[Fe(II)] cycle, the reaction with FePcF16 under argon (entry 16) or in the absence of any iron catalyst (entry 17) gave no conversion of the starting materials. In other solvents (ethyl acetate, ethanol, or acetone) the reaction was less efficient than in dichloromethane (entries 18–20). Finally, using Fe(dbm)3 as catalyst in the presence of 40 mol% Hünig's base as additive gave no significant increase of the yield for the coupling product 7  a (compare entries 4 and 21).

Table 4

Optimization of the reaction conditions for the iron‐catalyzed oxidative C−N coupling of the phenol 2  f with phenothiazine (5  a).[a]

Entry	[Fe] (4 mol%)	Additive (40 mol%)	Time [h]	Yield 7  a [%]
1	FeCl2	–	24	–
2	Fe(acac)3	–	24	–
3	ClFeTPPF20	–	24	traces
4	Fe(dbm)3	–	24	15
5	FePc	–	24	18
6	FePcF16	–	24	27
7	FePcF16	MsOH	2	–
8	FePcF16	K2CO3	13	15
9	FePcF16	NaOAc	16	20
10	FePcF16	NaOtBu	17	48
11	FePcF16	DBU	15	15
12	FePcF16	DABCO	16	28
13	FePcF16	NiPr2Et	16	94
14[b]	FePcF16	NiPr2Et	14	90
15[c]	μ‐oxo(FePcF16)2	NiPr2Et	13	96
16[d]	FePcF16	NiPr2Et	24	–
17	–	NiPr2Et	24	–
18[e]	FePcF16	NiPr2Et	16.5	26
19[f]	FePcF16	NiPr2Et	17.5	59
20[g]	FePcF16	NiPr2Et	15	51
21	Fe(dbm)3	NiPr2Et	24	20

[a] Reaction conditions: 2  f (0.40 mmol), 5  a (0.80 mmol), air (1 atm), CH2Cl2 (4 mL), room temperature; full conversion of starting material was indicated by TLC analysis. [b] Reaction conditions: 2  f (1.005 g, 6.69 mmol), 5  a (2.68 g, 13.5 mmol), FePcF16 (4 mol%, 268 μmol), NiPr2Et (40 mol%, 2.68 mmol), air (1 atm), CH2Cl2 (67 mL), room temperature. [c] 2  f (0.20 mmol), 5  a (0.40 mmol), O(FePcF16)2 (2 mol%), air (1 atm), CH2Cl2 (2 mL), room temperature. [d] Reaction under argon. [e] Solvent: EtOAc (4 mL). [f] Solvent: EtOH (4 mL). [g] Solvent: acetone (4 mL). acac=acetylacetonate; DABCO=1,4‐diazabicyclo[2.2.2]octane; dbm=dibenzoylmethanato; DBU=1,8‐diazabicyclo[5.4.0]undec‐7‐ene; TPPF20=5,10,15,20‐tetra(pentafluorophenyl)porphyrin.

–

–

–

–

–

–

–

–

–

FePcF

N

–

–

–

Using the optimized conditions (Table 4, entry 13), we investigated the scope of the oxidative C−N coupling reaction of various phenols 2 with phenothiazine (5  a) (Table 5). While the reaction of anisole with phenothiazine (5  a) gave no coupling product, simple phenols as substrates provided the ortho‐ or para‐aminated products 7  a–g in 69–99 % yield. It is noteworthy that in the present case the coupling of thymol with phenothiazine (5  a) led to product 7  g as single regioisomer. The reaction of 4‐(tert‐butyl)phenol (2  f) with substituted phenothiazines afforded selectively 7  h–j in 73–91 % yield. Shorter reaction times are sufficient for the coupling of 2  f with the more reactive phenoxazine to compound 7  k. Next, we studied the iron‐catalyzed oxidative C−N coupling of phenothiazine (5  a) with polycyclic phenols (2‐naphthol, 5‐hydroxy‐1‐tosylindole, 2‐hydroxydibenzofuran,[ , ] 2‐hydroxy‐7‐methoxy‐3‐methylcarbazole, and 4‐hydroxycarbazole ). These amination reactions provided the corresponding N‐arylphenothiazines 7  l–p regioselectively and in high yields. The regiochemistry of 7  p (amination at C‐1 of the carbazole) was confirmed by 2D NMR spectroscopy (COSY, HSQC, HMBC, and NOESY) (see Supporting Information). The synthesis of 7  o and 7  p demonstrates that the present method enables late‐stage functionalizations which are difficult to achieve by alternative procedures at the carbazole framework. Carbazoles are an important class of natural products with a broad range of pharmacological activities.

Table 5

Substrate scope for the iron‐catalyzed oxidative C−N coupling of phenols 2 with compounds 5.[a]

[a] Reaction conditions: 2 (0.40 mmol), 5 (0.80 mmol), FePcF16 (4 mol%), NiPr2Et (40 mol%), air (1 atm), CH2Cl2 (4 mL), room temperature, 13–16 h. [b] Reaction time: 2.5 h. [c] Reaction time: 5.5 h. [d] 2  f (0.40 mmol), 5  b (1.00 mmol). [e] 2  f (0.40 mmol), 5  d (0.40 mmol). [f] Reaction time: 3 h. [g] 2  m (0.20 mmol), 5  a (0.40 mmol), FePcF16 (4 mol%), NiPr2Et (40 mol%), air (1 atm), CH2Cl2 (2 mL), room temperature, 3.5 h. [h] Reaction time: 6 h.

During the synthesis of compound 7  p, we noted that prolonged reaction times lead to formation of a diaminated product. This observation prompted us to study the application of our iron‐catalysis to a twofold oxidative C−N bond formation using appropriate substrates. Lei et al. prepared diaminated phenols by electrochemical oxidation. With an excess of four equivalents of phenothiazine (5  a), we were able to transform 4‐hydroxycarbazole into the corresponding product 8  a by an iron‐catalyzed oxidative diamination (Table 6). Compound 8  a was unequivocally confirmed by an X‐ray crystal structure determination (Figure 4). Using phenol, o‐cresol, or o‐tert‐butylphenol as substrates, the same set of reaction conditions led to the diaminated compounds 8  b–d. However, the iron‐catalyzed oxidative amination of the more electron‐rich substrate guaiacol (2  s) afforded the mono‐aminated product 7  q and the diaminated product 8  e only in a 2 : 1 ratio even with an excess of 4 equivalents of 5  a (Scheme 3). Prolonged reaction times led to no further conversion of 7  q into 8  e but resulted in partial decomposition. The amination at the position para to the hydroxy group in compound 7  q has been assigned based on 2D NMR spectroscopy (COSY, HSQC, HMBC, and NOESY) (see Supporting Information). Xia et al. described an ortho‐selective coupling between guaiacol (2  s) and phenothiazine (5  a). However, the 1H and 13C NMR data of our compound 7  q are in excellent agreement with those reported by Xia et al. for their corresponding product. We have additionally confirmed our structural assignment by an X‐ray analysis (Figure 5). Therefore, the structure of the corresponding product described by Xia et al. has to be revised.

Table 6

Substrate scope for the twofold iron‐catalyzed oxidative C−N coupling of phenols 2 with phenothiazine (5  a).[a]

[a] Reaction conditions: 2 (0.40 mmol), 5  a (1.60 mmol), FePcF16 (4 mol%), NiPr2Et (40 mol%), air (1 atm), CH2Cl2 (4 mL), room temperature.

Figure 4

Molecular structure of the carbazole 8  a in the crystal (thermal ellipsoids are shown at the 50 % probability level; the solvent molecule (CH2Cl2) has been omitted for clarity).

Scheme 3

Iron‐catalyzed oxidative amination of guaiacol (2  s) with phenothiazine (5  a). Reaction conditions: a) 2  s (0.40 mmol), 5  a (1.60 mmol), FePcF16 (4 mol%), NiPr2Et (40 mol%), air (1 atm), CH2Cl2, room temperature, 14 h.

Figure 5

Molecular structure of compound 7  q in the crystal (thermal ellipsoids are shown at the 50 % probability level.

We postulate that the iron‐catalyzed amination of phenols with phenothiazine (5  a) proceeds via a phenothiazine radical intermediate formed by SET to the iron(III) species and subsequent proton loss (Scheme 4). Attack of the phenothiazine radical at the phenolate leads to a cyclohexadienyl radical which is oxidized in another SET to give the final coupling product. Alternatively, the coupling product may be formed by radical recombination of the phenothiazine radical with the phenoxy radical formed as described in Scheme 2.[ , ] Support for a mechanism via coupling of a phenothiazine free radical derives from a trapping experiment (Scheme 5) and ESR spectroscopy (Figure 6).

Scheme 4

Proposed mechanism for the iron‐catalyzed oxidative C−N coupling of phenols with phenothiazine.

Scheme 5

Iron‐catalyzed oxidative C(sp3)−H/N−H coupling of para‐cresol derivatives 9 with phenothiazine (5  a). Reaction conditions: a) 9 (0.40 mmol), 5  a (0.80 mmol), FePcF16 (4 mol%), NiPr2Et (40 mol%), air (1 atm), CH2Cl2, room temperature, 15–24 h.

Figure 6

Comparison of the experimental (exp; black) and simulated (sim; red) EPR spectra of the phenothiazine radical at room temperature.

Previously, we trapped an aryl radical in the FePcF16‐catalyzed homocoupling of diarylamines with the free radical scavenger 2,6‐di‐(tert‐butyl)‐4‐methylphenol (BHT). The iron‐catalyzed oxidative C−N coupling of phenothiazine (5  a) with phenols occurs exclusively in the positions para and ortho to the hydroxy group. Thus, the regiochemical course for this reaction using phenols with ortho‐ and para‐positions blocked by substituents was in question. The iron‐catalyzed oxidative coupling of the p‐cresol derivatives 9  a (BHT) and 9  b with 5  a under the standard reaction conditions provided the N‐benzylphenothiazines 10  a and 10  b (Scheme 5). The formation of compounds 10 proceeds via addition of the phenothiazine radical to the corresponding p‐quinone methides generated by initial oxidation of the p‐cresols 9. Moreover, the synthesis of 10 indicates that the present methodology can be applied for iron‐catalyzed C(sp3)−H/N−H coupling reactions.

According to the literature, phenothiazine radicals are formed from the parent compounds under oxidative conditions.[ , , ] To get more information on the reaction mechanism of the iron‐catalyzed oxidative amination of phenols, we detected the phenothiazine radical generated under the coupling conditions by electron paramagnetic resonance (EPR) spectroscopy. The EPR spectrum of a solution of phenothiazine (5  a) (0.40 mmol) with FePcF16 (2 mol%) and NiPr2Et (20 mol%) in dichloromethane (4 mL) in the presence of air at room temperature after 20 h displays a signal with a hyperfine structure and a g‐value of 2.0047 (Figure 6). The experimental and simulated ESR spectra are in nearly perfect agreement. The hyperfine structure is caused by nitrogen (a(14N)=7.1 G) and hydrogen (a(1H)=3.7, 2.7, 1.0, and 0.8 G) atoms. The observed hyperfine coupling constants are corresponding to those of a neutral N‐centered phenothiazine radical.

Conclusion

We have developed a sustainable and efficient iron‐catalyzed oxidative C−O and C−N coupling reaction which proceeds at room temperature using hexadecafluorophthalocyanine−iron(II) (FePcF16) as catalyst and air as sole oxidant. Utilization of the appropriate additive is crucial for the success of this transformation. Lewis acid is required for the oxygenation of tertiary arylamines. Aminations of tertiary arylamines proceed in the presence of Brønsted acid and those of phenols in the presence of Hünig's base. The structural variety of the resulting coupling products, the excellent selectivity, and the functional group tolerance emphasize the utility of this process. Moreover, the feasibility to achieve oxidative aminations by iron‐catalyzed activation of C(sp3)−H bonds has been indicated. Further studies and applications of the iron‐catalyzed oxidative carbon−heteroatom bond formation under aerobic conditions are currently in progress.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Click here for additional data file.