Tunable, Chemoselective Amination via Silver Catalysis

Abstract

Organic N-containing compounds, including amines, are essential components of many biologically and pharmaceutically important molecules. One strategy for introducing nitrogen into substrates with multiple reactive bonds is to insert a monovalent N fragment (nitrene or nitrenoid) into a C–H bond or add it directly to a C=C bond. However, it has been challenging to develop well-defined catalysts capable of promoting predictable and chemoselective aminations solely through reagent control. Herein, we report remarkable chemoselective aminations that employ a single metal (Ag) and a single ligand (phenanthroline) to promote either aziridination or C–H insertion by manipulating the coordination geometry of the active catalysts.

Amines are present in a multitude of pharmaceuticals and natural products with useful biological activities. As a result, the development of synthetic methodologies for the chemo-, regio-, and stereoselective introduction of C–N bonds has been vigorously pursued.^1a–e One attractive approach is the direct insertion of a nitrene or nitrenoid species into a C–H or C=C bond of an unsaturated substrate, and many catalysts based on Rh, Ru, Fe, Co, Cu, Mn, Au, and Ag have been exploited in this context.^2a–m However, chemoselective C–N bond formation in substrates bearing both reactive C–H and C=C bonds is a particularly challenging task, as these compounds (Figure 1) often give rise to multiple products or exhibit substratecontrolled selectivity.^3a–g

Figure 1.

“Static” vs “dynamic” chemoselective amination.

One strategy employed to overcome the problem of substrate control in metal-catalyzed amination is to change the identity of the transition metal. For example, Ru- and Fe-based catalysts favor C–H amination over the aziridination pathway that is preferred using Rh(II) carboxylates.^4a,b A second tactic is to utilize different supporting ligands with a single metal, but this has been only marginally successful for chemoselective amination.^3a–d Finally, the nature of the nitrene precursor can influence the reaction outcome.^5a–c We refer to strategies employing a single, well-defined complex to control a specific amination event as ‘static’ approaches to catalysis (Figure 1, top).

Our previous studies on the chemoselective aziridination of homoallenic carbamates to bicyclic methylene aziridines (Table 1) showed that Ag complexes supported by bidentate N ligands provided superior chemoselectivity for aziridination compared to Rh₂(esp)₂, irrespective of the substrate identity (compare entry 1 vs 2–7 and 9 vs 10–15).^6,7a–f However, a tridentate ligand reversed this selectivity (entries 8 and 16). This result stimulated our curiosity, and a further perusal of the literature showed that Ag has the unique ability to change coordination geometry in response to changes in the Ag counteranion, the ligand identity, or the metal/ligand ratio.⁸ If these changes in the coordination geometries of the Ag catalysts were indeed responsible for inducing divergent chemoselectivity, a ‘dynamic approach’ to catalytic amination could be envisioned (Figure 1, bottom). In this scenario, treatment of a single Ag salt with a single ligand would yield a mixture of several potential catalytic species. Simple perturbation of the equilibrium of this mixture could give different catalytic species capable of promoting divergent amination using reagent control.

Table 1.

Chemoselective Aziridination and C–H Amination of Homoallenic Carbamates Catalyzed by Silver Catalysts

entrya	catalystb,c	d	4a	5a	entrye		4b	5b
1	Rh2(esp)2		35%	17%	9		5%	80%
2	AgOTf/phen		79%	---	10		80%	14%
3	AgOTf/bipy		60%	---	11		68%	11%
4	AgOTf/bathophen		57%	---	12		84%	12%
5	AgOTf/p-MeObipy		72%	---	13		73%	11%
6	AgOTf/dafone		32%	---	14		59%	22%
7	AgOTf/p-Ph-bipy		66%	---	15		62%	20%
8	AgOTf/terpy		27%	35%	16		9%	61%

^aSubstrate 3a.

^b5 mol % Rh₂(esp)₂, 2 equiv PhlO.

^cAg: 20 mol % AgOTf, 25 mol % ligand, 2 equiv PhIO, 4 Å MS, CH₂Cl₂.

^dA: aziridination. I: insertion.

^eSubstrate 3b.

In order to test the potential for developing a dynamic catalyst system, the metal:ligand stoichiometry of a AgOTf:phen catalyst system was varied to determine the effect on chemoselectivity.^8a–g Phenanthroline was chosen for its high yield in the amination and its relatively low cost. To our delight, a clear impact on the amination of 3b was observed (Table 2). AgOTf:phen ratios close to 1:1 (entries 1–4) promoted aziridination to 4b as the major reaction pathway, while increasing the amount of phen gave C–H insertion to 5b as the dominant mode of reactivity (entries 5, 6). The dramatic reversal in the reaction outcome suggests that an equilibrium between Ag(phen)OTf and Ag(phen)₂OTf exists and that each complex favors a different mode of reactivity.

Table 2.

Effect of AgOTf:phen Stoichiometry on the Aziridination/Insertion Ratio

entrya	equiv AgOTf/phen	4b:5b ( )	4b (5b)b
1	0.2/0.1	5:1	60% (12%)
2	0.2/0.2	5.8:1	75% (13%)
3	0.2/0.25	6.2:1	80% (13%)
4	0.2/0.3	5.8:1	70% (12%)
5	0.2/0.4	1:4	18% (72%)
6	0.2/0.6	1:38	2% (76%)

^aReactions were carried out at 0.125 M 3b in CH₂Cl₂, 2 equiv PhIO, AgOTf/phen, rt.

^bNMR yields with mesitylene as the internal standard.

The scope of the ‘dynamic’ amination was explored using homoallenic carbamates (Table 3). In all cases, a 1:1.25 ratio of AgOTf:phen favored aziridination,⁶ while a 1:3 ratio of AgOTf:phen yielded mainly C–H insertion. Trisubstituted allenes (entries 1, 6–7) exhibited good selectivity under both conditions, while less substituted allenes (entries 2–5, 8) usually gave better selectivity in C–H insertion. Interestingly, the addition of 10 mol % of 2,6-bis(1,1-dimethylethyl)-4-methylphenol (BHT) appeared to improve the conversion of the C–H insertion (entries 4, 8–10).^7f

Table 3.

Tunable Amination of Homoallenic Carbamates

entry	allene	AgOTf:phena,b	d	yield 4	5
1	3a	1:1.25		79%	<4%
		1:3		0%	81%
		Rh2(esp)2		35%	17%
2	3b	1:1.25		79%	13%
		1:3		1%	76%
3	3c	1:1.25		80%	9%
		1:3		0%	76%
		Rh2(esp)2		34%	34%
4	3d	1:1.25		72%	18%
		1:3		5%	65%
		1:3e		<1%	71%
5	3e	1:1.25c		67%	18%
		1:3		0%	83%
6	3f	1:1.25		87%	7%
		1:3		0%	88%
		Rh2(esp)2		34%	44%
7	3g	1:1.25		70%	4%
		1:3		0%	78%
8	3h	1:1.25		57%	12%
		1:3		4%	74%
		1:3e		<1%	68%
9	3i	1:1.25		61%	0%
		1:3		3%	74%
		1:3e		<1%	71%
10	3j	1:1.25		58%	0%
		1:3		0%	36%
		1:3e		2%	62%

^aAziridination: 20 mol % AgOTf, 25 mol % phen, 2 equiv PhlO, 4 Å MS, CH₂Cl₂.

^bC–H insertion: 10 mol % AgOTf, 30 mol % phen, 3.5 equiv PhIO, 4 Å MS, CH₂Cl₂.

^c2,2′-bipyridine ligand.

^dI = insertion. A = aziridination,

^e10 mol % BHT added

Simple changes in the AgOTf:phen stoichiometry also provided good chemoselectivity in the amination of homoallylic carbamates (Table 4). The cis-disubstituted 6a showed increased selectivity for aziridination in switching from Rh₂(OAc)₄ to 1:1.25 AgOTf:phen (entry 1), while changing the AgOTf:phen ratio to 1:3 promoted exclusive insertion. This trend held for both the cis-disubstituted 6b (entry 2) containing substitution in the tether and the trans-disubstituted 6c (entry 3). The stereochemistry of the olefin was transferred to the resulting aziridines and allylic amines with no detectable isomerization. The 1,1′-disubstituted 6d gave better selectivity and yield for aziridination compared to Rh₂(OAc)₄, although the C–H insertion was moderate. Substrate 6e gave poor results using Ag(phen)OTf, but good selectivity for insertion.

Table 4.

Selective Amination of Homoallylic Carbamates

entry	substrate	catalysta,b,c		7	8	dr (cis:trans)
1	6a	Rh2(OAc)4		58% 7a	18% 8a	(100:0)
		1:1.25 AgOTf:phen		67% 7a	4% 8a	---
		1:3 AgOTf:phen		0% 7a	93% 8a	---
2	6b	Rh2(esp)2		45% 7b	25% 8bd	nd
		1:1.25 AgOTf:phen		89% 7b	9% 8b	3.2:1
		1:3 AgOTf:phen		4% 7a	87% 8b	3:1
3	6c	Rh2(OAc)4		68% 7c	14% 8c	(0:100)
		1:1.25 AgOTf:phen		88% 7c	3% 8c	---
		1:3 AgOTf:phen		11% 7c	73% 8c	---
4	6d	Rh2(OAc)4		35% 7d	5% 8d	---
		1:1.25 AgOTf:phen		85% 7d	<1% 8d	---
		1:3 AgOTf:phen		23% 7d	66% 8d	---
5	6e	Rh2(esp)2		25% 7e	21% 8ed	nd
		1:1.25 AgOTf:phen		54% 7e	39% 8ed	nd
		1:3 AgOTf:phen		0% 7e	68% 8e	2.4:1

^aRh cat.: 3 mol %, 2 equiv PhIO, 4 Å MS, CH₂Cl₂.

^bAziridination: 20 mol % AgOTf, 25 mol % phen, 2 equiv PhlO, 4 Å MS, CH₂Cl₂.

^cC–H insertion: 10 mol % AgOTf, 30 mol % phen, 3.5 equiv PhlO, 4 Å MS, CH₂Cl₂.

^dNMR yields with mesitylene as the internal standard.

Attempts to isolate the proposed 1:1 and 1:2 AgOTf:phen complexes in the solid state (Table 2) resulted in the recovery of only Ag(phen)₂OTf.⁹ Nonetheless, Ag(phen)₂OTf was capable of dissociating and reassembling into two distinct catalytic species capable of divergent amination (Scheme 1). Reaction of 3a with the preformed Ag(phen)₂OTf gave a 90:8 mixture of products in favor of the C–H insertion, consistent with the results described in Table 2. Addition of 10 mol % AgOTf to the initial Ag(phen)₂OTf complex completely reversed the chemoselectivity to provide 4a in 92% yield, while an extra 10 mol % of phen shut down the competing aziridination pathway, giving 5a in 90% yield.

Scheme 1.

Solution-State Behavior of a Preformed Ag(phen)₂OTf Catalyst for Chemoselective Amination

Attempts to corroborate the proposed solution state geometries for the two Ag catalysts illustrated in Table 2 were carried out using NMR titration experiments with 4,4′-di-tert-butyl-2,2′-bipyridine (^tBu-bipy, substituted for phen to improve solubility) and AgOTf. Unfortunately, rapid dynamic exchange, even at temperatures as low as −85 °C, prevented direct observation of the individual species present in solution (see Supporting Information (SI) for details). However, the averaged ¹H, ¹³C chemical shifts indicated that the major species in the mixture changed as the ratio of ligand:AgOTf was increased. A combination of pulse gradient spin echo (PGSE) and MALDI MS experiments showed that a monomeric Ag(L)OTf complex was the major species in solution when a 1:1 AgOTf:ligand ratio was used, while a monomeric Ag(L)₂OTf complex predominated when a 1:2 AgOTf:ligand ratio was used (details in SI). The additional equivalent of ligand serves to perturb the equilibrium of the Ag(L)OTf:Ag(L)₂OTf mixture to favor the latter.

With information about the nature of the two catalytic species in hand, we wanted to understand the factors responsible for our unexpected and tunable chemoselectivity. The exact mechanisms of metal-catalyzed aminations have been notoriously difficult to unravel and often involve multiple reaction pathways.^2,7f Yet, we felt even preliminary insights into the mechanism could help extend our dynamic catalysis beyond the scope here.

Experiments to determine whether Ag-catalyzed amination proceeds through a stepwise or concerted mechanism were carried out using the stereochemical probes (±)-9 and (±)-12 (Scheme 2). Only (±)-10 and (±)-14 were observed and no isomerization was detected, suggesting a concerted event. A substrate 15 containing a radical trap yielded only 16 and no ring-opened product, arguing against long-lived radical intermediates. A kinetic isotope effect (KIE) experiment yielded a 3.4 ± 0.1 mixture of isotopomers (±)-18-D and (±)-18-H (eq 1). KIEs in

Scheme 2.

Stereochemical Probes for Radical Intermediates

(1)

the range of 1–3 are believed to correspond to a concerted pathway, while KIEs in the range of 6–12 usually signal a stepwise process involving potential radical intermediates.^3c,11a–c This suggests that C–H insertion favors a singlet nitrene pathway over hydrogen atom abstraction.

To shed light on the differences between Ag-catalyzed pathways promoting aziridination vs insertion, initial rates were measured for four homoallenic carbamates 3a, 3d, 19, and 21 (Table 5). As expected, the initial rate of aziridination was faster than C–H insertion for both tri- and disubstituted allenes (compare entries 4 and 6, as well as 11 and 13). When sites for potential C–H insertion were blocked in substrates 19 and 21, the Ag(phen)OTf catalyst still gave aziridination (entries 1 and 8), but the Ag(phen)₂OTf complex gave either no reaction (entry 2) or significantly decreased reactivity (entry 9). This suggests that the steric congestion around the Ag center plays an important role in dictating the chemoselectivity, with a more hindered Ag center promoting insertion over the aziridination which is favored by a less sterically congested Ag center. This is likely due to the difficulty of overcoming steric clashing with the ligands when the substrate attempts to adopt the appropriate orientation of the olefin for reaction with the nitrene (see 26 in Figure 2, vide infra). When two ligands are coordinated to the Ag center, insertion into the C–H bond of 26 presents a more favorable pathway, in contrast to the aziridination that occurs through the proposed 1:1 Ag:L complex 24.

Table 5.

Relative Rates of Aziridination and C–H Insertion

entry	R1-R3	catalyst (AgOTf, phen)	product(s)	init rate (mmol/min*mL)a
1	Me, Me, Me 19	20 mol%, 25 mol%	20	1.8×10−3 (98% yield)
2		10 mol%, 30 mol%	20	no reaction
3		20 mol%, 25 mol%, 20 mol% BHT	20	1.4×10−3 (48% yield)
4	Me, Me, H 3a	20 mol%, 25 mol%	4a	1.3×10−3
5		20 mol%, 25 mol%, 20 mol% BHT	5a	9.6×10−4
6		20 mol%, 60 mol%	5a	2.8×10−4
7		20 mol%, 60 mol%, 20 mol% BHT	5a	2.1×10−4
8	C5H11, H, Me 21	20 mol%, 25 mol%	22	9.85×10−4 (88% yield)
9		20 mol%, 60 mol%	22	36% yieldb
10		20 mol%, 25 mol%, 20 mol% BHT	22	6.15×10−4
11	C5H11, H, H 3d	20 mol%, 25 mol%	4d	5.73×10−4
12		20 mol%, 25 mol%, 20 mol% BHTc	4d (26%) 5d (34%)	3.31×10−4 2.31×10−4
13		20 mol%, 60 mol%	5d	1.58×10−4
14		20 mol%, 60 mol%, 20 mol% BHT	5d	2.94×10−4

^aThe rate of product formation was monitored by ¹H NMR using mesitylene as the internal standards. The indicated initial rates are the average of the three runs, and the standard deviations are included in the SI.

^bYield after 21 h, 73% conversion.

^cThe ratio of 4d:5d was 1:1.

Figure 2.

Possible mechanisms for Ag-catalyzed divergent, chemoselective amination.

BHT was initially employed to ascertain the impact of a radical inhibitor on the amination (Table 3), where the presence of this additive appeared to improve the conversion of disubstituted homoallenic carbamates to allenic amines. Closer examination of the role of BHT (Table 5) showed that the initial rates in the aziridination of 19 and 21 were decreased in the presence of the radical inhibitor (compare entries 1 and 3, and entries 8 and 10). However, the effect of BHT on the rate of C–H insertion was variable (compare entries 6–7 and 13–14) and did result in an increase in the rate of insertion when a disubstituted allene was employed (entries 13 vs 14). Interestingly, the addition of BHT to 3d in the presence of 1:1.25 AgOTf:phen (entry 12) gave a 1:1 ratio of aziridine 4d to allenic amine 5d. This suggests BHT may also play a role in altering the Ag(L)OTf:Ag(L)₂OTf equilibrium by shifting it toward Ag(L)₂OTf, but further study will be needed to completely understand its impact on the reaction.

The retention of stereochemistry at a chiral center, the low KIE value, the lack of isomerization in the reactions of homoallylic carbamates (Table 4), and the absence of ring opening in the cyclopropane 15 all support a concerted pathway involving a singlet nitrene for the C–H insertion (Figure 2, Path B).^3b,c,7e,f However, the aziridination Path A could involve either singlet or triplet nitrene intermediates, or perhaps both. The differences in energies between these two states can be very small; Pérez and co-workers have recently reported that Ag-catalyzed olefin aziridination may involve both paths.^7f,12a–c The decrease in the rates of aziridination in the presence of BHT implies there may a triplet nitrene involved, but the lack of isomerization in the aziridination of homallylic carbamates (Table 4) argues against this and further mechanistic studies will need to be carried out to clarify this issue. Irrespective of the exact reaction pathways, the best explanation for Ag-catalyzed chemoselectivity resides in the dramatic steric differences in the coordination geometries adopted by Ag(phen)OTf and Ag(phen)₂OTf, respectively.

In conclusion, we have developed a simple Ag-based catalyst system that represents the only method to date capable of employing the same metal and the same ligand to accomplish either aziridination or C–H insertion in good yields. The ability for Ag to readily adopt multiple coordination geometries provides a new approach to identify catalysts that can promote other types of chemoselective aminations, including choosing between two different C–H bonds. In addition, the ease with which this methodology can be implemented and hopefully extended to other chemoselective C–heteroatom and C–C bond formations opens a potential gateway in reaction discovery. Computational and further mechanistic studies are currently underway to unveil the electronic and steric nature of the reactive species in these Ag-catalyzed chemoselective aminations.