**1. Introduction**

The critical issues of climate change and rapidly increasing global energy consumption have triggered tremendous research efforts for clean and renewable energy sources, as well as advanced techniques for energy storage and conversion [1–3]. Among the most competitive technologies, such as batteries, fuel cells, and supercapacitors, the electrochemical double-layer supercapacitors (EDLSs) are capturing growing attention, for their wide potential for electric vehicles, digital devices, and pulsing techniques [4,5]. The EDLS is primarily a physical electrostatic behavior in nature, which is generally based on the reversible adsorption of electrolyte ions at the electrode/electrolyte interface [6]. Since the charges are stored on a high surface area without any faradaic reaction involved, long cycle life can be achieved, in addition to the high capacitance, which establishes them as one of the most important emerging energy strategies for consumer electronics, and bridging devices with high energy batteries in hybrid power applications [7].

For EDLSs, carbon based materials are the most commonly used electrode candidates, due to their stable physicochemical properties, fast charging/discharging kinetics, bipolar operational flexibility and low cost [8]. Various kinds of carbon materials, such as activated carbon (AC), mesoporous carbon, carbon nanotubes, and graphene, have been reported as the electrode materials for EDLSs [9–12]. Among these carbon materials, AC have been conventionally used for industrial EDLSs, because of their ultra-high surface area and commercially available mass production. However, the structural shortcomings, for example, unsuitable pore size distribution and limited surface functionality, make AC su ffer from low energy density, low conductivity, and slow inner-pore ion di ffusion. Appropriately, carbon materials with multiple pore size distribution (micro-, meso-, and macro-pores) are quite necessary to obtain proper energy storage characteristics [13]. This is because although micropores (<2 nm) can provide a large surface area, the infiltration of electrolyte ions into these pores is rather di fficult, and the kinetics is quite slow; the transportation of ionic species in macro- (>50 nm) and meso- (2–50 nm) pores, on the other hand, is much easier but their surface to volume ratios are not high enough [14,15].

Three dimensional (3D) hierarchical carbon architectures, such as 3D graphene membranes, 3D carbon nanotubes, and 3D carbon fiber networks, are strongly recommended as promising supercapacitor candidates, due to their special pores, excellent electron conductivity and large, abundant ion pathways [16–19]. For example, densely-packed carbon nanotube (CNT) spherical assemblies demonstrated a specific capacitance of 215 F g<sup>−</sup>1, which is twice more than that of frequently reported commercial AC electrode. By using polypyrrole (PPy) microsheets as precursors and KOH as the activating agent, 3D hierarchical porous nanostructures with large specific areas of 2870 m<sup>2</sup>/g were reported, which demonstrated a high capacity of 318.2 g<sup>−</sup><sup>1</sup> at a current rate of 0.5 A g<sup>−</sup>1, and excellent retention ability of 95.8% after long-term cycling of more than 10,000 [20]. More recently, by using garlic skin as a source, novel hierarchical porous carbon materials with 3D penetrating pores were successfully synthesized by Han et al., and an ultra-high capacitance of around 380 F g<sup>−</sup><sup>1</sup> at a current density of3Ag−<sup>1</sup> accompanied by good rate performance were progressively reported [21]. All these indicated the promising advantages on 3D hierarchical carbon nanostructures for supercapacitor (SC) applications. Moreover, besides these structural related benefits, the electrochemical performance can further be reinforced by surface chemistry modification; e.g., introducing heteroatom doping with N, B, P, or S [22–25], or forming composites with metal oxide nanomaterials, which can provide more competitive advantages for novel SC devices [26–28].

Up to now, various methods, including chemical vapor or physical based deposition, biomass carbonization, hard-templating, etc., have been developed to synthesize 3D hierarchical carbon for supercapacitor applications [29]. On the one hand, most of those methods make it di fficult to obtain the proper structures that are required for high performance SC devices; e.g., large surface area, and an e fficient pathway for ion and electron transportation; and on the other hand, even for the most frequently reported methods of hard-templating and biomass carbonization, the structural parameters for carbon are usually limited by their raw templates or initial biomass precursors, which can not be feasibly modulated to achieve a better electrochemical performance. Even more, most of these methods are di fficult or costly to be applied for large scale production. It is still quite challenging to develop e fficient solutions for synthesizing novel hierarchical carbon nanostructures that are suitable for advanced supercapacitors.

Linear polyethyleneimine (L-PEI) refers to an interesting family of molecules, which can adopt a variety of one-dimensional (1D) nanostructures to form hierarchical nano- and micro-structures [30–33]. Distinct structural configurations, like nanoplate, nanowire, mesoporous microsphere, interconnected nanotube, etc., can be easily obtained from this one single polymer, by modulation of the crystalline condition via low temperature solution method. And because there are large amounts of amine groups existing on PEI molecular chains, linear PEI polymers can be used as e ffective catalysts to prompt the growth of SiO2 coating layers, enabling them a preponderance for constructing highly ordered 3D nanomaterials [34–36]. Herein, to unravel the flexibility and benefits for PEI as 3D carbon nanostructure

constructing template; we reported an e fficient method for the preparation of novel 3D hierarchical nitrogen doped carbon nanonet flakes ( *N*CNFs), which showed significantly improved electrochemical performance. The *N*CNFs possess large surface area with permeable and interconnected hierarchical pores which facilitate the transmission of electrolyte ion, and nitrogen doped groups in carbon framework contribute e ffectively on contact with electrolyte solution. Electrochemical measurements indicated a high specific capacity of 613 F g<sup>−</sup><sup>1</sup> at a current density of1Ag−<sup>1</sup> (or 259 F g<sup>−</sup><sup>1</sup> at 10 A g<sup>−</sup>1) and cycling stability after 20,000 cycles at 10 A g<sup>−</sup>1, which are quite encouraging for applications as high performance SC electrodes.

## **2. Experimental Section**

#### *2.1. Chemicals and Materials*

Poly(2-ethyl-2-oxazoline) was bought from Alfa Aesar chemicals, Shanghai. Resorcinol and tetramethoxysilane (TMOS) was purchased from Maclin Biochemical Co., Ltd., Shanghai, China. Melamine, sodium hydroxide, anhydrous ethanol and hydrofluoric acid solution (≥40%) were purchased from Shanghai, China Chemical Regent Co., Ltd. Hydrochloric acid solution (36%–38%) was purchased from Luoyang, China Haohua Chemical Regent Co., Ltd. Formaldehyde solution (≥37%) was purchased from Tianjin, China Jiachen Chemical Factory.

#### *2.2. Synthesis of Linear Polyethyleimine*

The synthesis was performed according to the previous report (in Scheme 1). Briefly, 20 g Poly(2-ethyl-2-oxazoline) was dissolved in a 200 mL HCl solution (5 M) and the solution was heated for 12 h under stirring in an oil bath at ca. 100 ◦C. After cooling to room temperature, a white suspension was obtained. The precipitate was further collected by suction, washed by methanol three times and dried under vacuum. The as-collected product was protonated PEI (PEI-H<sup>+</sup>, shown in Scheme 1). 2 g of PEI-H<sup>+</sup> powders were dissolved in 24 mL water and then neutralized by the addition of 5 mL NaOH solution (5 M), which led to the formation of crystalline PEI aggregates. After centrifugation, wet PEI powders were separated and further washed by H2O three times.

**Scheme 1.** The synthesis route of linear polyethyleneimine (L-PEI).

#### *2.3. Synthesis of PEI@SiO2 Nanotubes*

Wet PEI powders obtained above were dispersed in 480 mL H2O and then mixed with 4 mL TMOS. After stirring for 3 h, the suspension was subjected to centrifugation, and the as-collected white precipitates were further washed by H2O and ethanol and finally dried at 60 ◦C for 12 h, which produced the powders of PEI@SiO2 nanotubes.

#### *2.4. Preparation of CNFs and NCNFs*

Phenolic resins, which were formed by the polymerization between resorcinol and formaldehyde, were employed as the carbon precursor. Melamine was used as the nitrogen source to synthesize N-doped carbon nanotube networks. Firstly, 0.5 g PEI@SiO2 powders were dispersed in 50 mL H2O, and proper amount of resorcinol and formaldehyde were added sequentially to form a suspension, which was subjected to heating with stirring for 24 h in an oil bath at 60 ◦C. After cooling to room temperature, solid powders were collected by centrifugation, washed with H2O and ethanol, and dried

under vacuum. To synthesize the N-doped samples, melamine in water solution was added just after the reaction of resorcinol and formaldehyde.

The as-obtained solid powders above were transferred into a tube furnace and heated at 700 ◦C for 1.5 h in flowing nitrogen gas, and both carbon and N-doped carbon coated SiO2 (SiO2@C) was formed. The SiO2 components in SiO2@C was further removed by HF solution. The samples obtained by adding 0 g, 0.03 g and 0.1 g melamine were denoted as CNFs, *N*CNFs-1 and *N*CNFs-2, respectively.
