The core functions of current-generation computers are to process, communicate and store information. These operations are mainly realized by two fundamental components of microprocessors and hard disks in computers, respectively. Si-based semiconductors are the key materials in microprocessors to enable the control of charges for transmitting and processing data, while ferromagnets provide the spin of electrons to be utilized for storing data on the hard disk. Generally, semiconductivity and ferromagnetism do not coexist in a material. However, it was found that europium chalcogenides showed anomalous optical, magnetic and transport phenomena^{[1, 2]}. The physics behind this anomaly results from the strong d–f exchange interaction between the magnetic-exciton electrons of 5d band and the conduction 4f electrons^[2]. As a result, there exists a relationship between the Curie temperature and the conduction electron concentration of these semiconducting europium chalcogenides. Based on these phenomena, Methfessel first proposed the concept of ferromagnetic semiconductor, whose magnetic properties could be modified by carrier injection, electrostatic fields or other means that changed the free carrier concentration in semiconductors^[3]; that is, both the charge and spin of electrons can be manipulated simultaneously in these ferromagnetic semiconductors. The europium chalcogenides are therefore regarded as the first generation ferromagnetic semiconductors. Despite their scientific importance, these magnetic rare earth compounds usually have complex crystalline structures, which are different from those of the widely used semiconductor materials such as Si and GaAs. It is difficult to obtain high-quality interface structure when integrating them with Si or GaAs. In addition, most of them show Curie temperatures far below room temperature^[3].

To maintain the most attractive semiconducting properties used in electronic devices, an approach was proposed to introduce magnetic elements into non-magnetic semiconductors for creating new type of magnetic semiconductors (MSs) called diluted magnetic semiconductors (DMSs)^[4–6]. With the ever growing necessity of miniaturization of future electronic devices, DMSs have stimulated intense interest due to their potential for realizing new functionalities and revolutionizing device concepts. The most thoroughly investigated DMSs are (Ga, Mn)As and (In, Mn)As systems based on III–V semiconductors^{[7, 8]}. Chen et al. explored nanostructure engineering to enhance the Curie temperature of the (Ga, Mn)As DMS up to 200 K, which has been the highest value recorded for this system^[9]. Further increasing the Curie temperatures of DMSs above room temperature is a challenging task^[10–20]. The basic reason for this is the limitation of the solid solubility for magnetic elements to dissolve in these DMSs. To keep the original lattice structure of the host crystal, the solubility of magnetic elements is very low especially for those with the electronegativity and valence electron concentrations significantly different from the base elements^[21–23]. This becomes the key barrier to creating high Curie-temperature DMSs that work at room temperature.

In contrast to the approach for developing the DMSs, we found a different way to create a new kind of MSs. First, we selected a high Curie-temperature ferromagnetic amorphous alloy (AA) as a host. Second, we introduced oxygen into this AA in a well-controlled manner. AAs are mainly characterized by non-directional and non-saturated metallic bonds. Their disordered atomic structures and unique electronic structures enable them to be nice hosts to include a large quantity of foreign elements^[24–26]. Oxidizing the ferromagnetic AA leads to a metal–semiconductor transtion and causes the formation of amorphous MSs (AMSs) with Curie temperatures well above room temperature^[26].

Crystalline metals and AAs behave quite differently as oxygen is introduced into them, respectively (Fig. 1). The solubility of oxygen in a crystalline metal limits the maximal amount of oxygen that enters its lattice structure, whereas no similar limitation exists for oxygen to be included in an AA. Moreover, amorphous materials could access to lower energy states through structural relaxation^[27–31]. Local atomic rearrangement reduces local strains induced by the oxygen addition, thereby permiting more oxygen atoms to be included in amorphous materials. With increasing the oxygen content, the conduction electrons of the AA are gradually localized due to the formation of ionic or/and covalent bonds between the constituents of the AA and the included oxygen atoms. One can expect that the original AA becomes more and more insulating as the oxygen content continuously increases. As a result, a metal–semiconductor transition occurs at a critical value of the oxygen content.

Following the idea of transferring AAs to semiconductors, an alloy system is selected owing to the following three reasons. First, ferromagnetic alloys are preferred in order to obtain MSs with a combination of desirable functionalities. Moreover, the Curie temperatures of the selected alloy systems should be higher than 500 K, which is required for a ferromagnet to work in practical applications^[32]. For amorphous ferromagnets, it is thus better to select AA systems containing 3d transition metal elements of Fe and/or Co, contributing to their robust ferromagnetism. Second, the ferromagnetism of the selected AA should be sustained even after it is heavily oxidized. To achieve this, the ferromagnetic metals in the alloy system are better to have smaller affinity for oxygen than the other constituents. Hence, the ferromagnetic metal elements are least oxidized compared with other constitutes in the alloy system. Third, the selected AA system should be a good glass former so that it can maintain its amorphous structure even when a high oxygen content is added. Based on these considerations, a good glass former Co–Fe–Ta–B system was selected^{[33, 34]}.

The Co–Fe–Ta–B–O thin films were deposited by radio frequency (RF) magnetron sputtering with an alloy target under a gas mixture of argon and oxygen^[24–26]. Varying the oxygen partial pressure led to the formation of amorphous Co–Fe–Ta–B–O thin films with different oxygen contents. Ta and B were firstly oxidized and diffused out to form the surface oxide. The surface metal oxide increased with increasing the oxygen content. They contacted with each other by forming a continuous network (Fig. 2). The size of the inner AA nanoparticles became smaller and smaller. At a critical oxygen content, the AA phase disappeared and a single oxide phase was formed.

The structure of the (Co_0.53Fe_0.23Ta_0.08B_0.16)_100–xO_x (0 ≤ x ≤ 50 at%, abbreviated as CFTBO_x hereafter) samples evolved gradually with increasing the oxygen content. High resolution transmission electron microscopy (HRTEM) images and selected area electron diffraction (SAED) patterns were taken for the CFTBO_x system, respectively (Fig. 3). The Co₅₃Fe₂₃Ta₈B₁₆ AA shows maze-like atomic arrangements typical for amorphous structure (Fig. 3(a)). Its SAED pattern further verifies the formation of a single amorphous phase in the AA (Fig. 3(b)). With increasing the oxygen content above 15 at.%, an amorphous oxide (AO) phase emerges in the CFTBO_x thin films. Fig. 3(c) shows the formation of a dual phase nanocomposite in the CFTBO₄₄ thin film, comprising the nanometer-sized AA particles embedded in the AO matrix. Its SAED pattern exhibits two sets of halos (Fig. 3(d)). One arises from the AA nanoparticles (Figs. 3(b) and 3(d)), while the other originates from the AO matrix (Figs. 3(d) and 3(f)). Further increasing the oxygen content to 46 at.% enables the formation of a single AO phase (Fig. 3(e)). The SAED pattern only shows the broad diffraction halo resulting from the single AO phase, which is semiconducting and ferromagnetic^[26].

Fig. 4(a) shows optical transmittance of the CFTBO_x samples at thickness of ~100 nm. With increasing the oxygen content, their optical transmittance increases as well. The optical bandgap of the CFTBO₄₆ thin film is estimated to be about 2.4 eV based on the Tauc plot (Fig. 4(b)). In addition, the thin film exhibits 488 nm-peaked photoluminescence specturm, correpsonding to a photon energy of about 2.5 eV for blue light. This value is in consistent with its optical bandgap energy.

As the charge carrier concentration decreases with the oxygen content, the bandgap of the CFTBO_x samples is gradually opened. That is, metal-semiconductor-insulator transitions are induced through this simple oxidization of the ferromagnetic AA. As a result, the resistivity of the CFTBO_x samples increases with increasing the oxygen content (Fig. 5). The CFTBO₄₆ thin film exhibits a negative temperature dependence of ln(ρ/ρ₀)–1/T^1/2 (inset of Fig. 5), characteristic of a semiconductor behavior. Further increasing the oxygen content above 60 at.% makes the CFTBO_x samples become insulating.

The magnetic properties of the CFTBO_x thin films are shown in Fig. 6. All the thin films are ferromagnetic at the oxygen contents ranging from 16 to 46 at.%. Noted that the saturation magnetization (M_s) of these thin films measured at room temperature increases from 728 to 867 emu/cm³ as the oxygen content increases from 16 to 25 at.%. With further increasing in the oxygen content up to 46 at.%, M_s decreases from 867 to 433 emu/cm³ (Figs. 6(a) and 6(b)). Furthermore, the zero-field cooling (ZFC) curve of the thin film with a low oxygen content of 16 at.% coincides with its field cooling (FC) curve, which is similar to that of the Co–Fe–Ta–B AA without containing oxygen^{[35, 36]}. However, the ZFC curves of the CFTBO_x thin films with the oxygen contents above 16 at.% deviate from their FC curves at low temperatures, demonstrating spin-glass-like maximum below 100 K. It is suggested that the spin-glass-like behavior is associated with the formation of the magnetic AO phase.

The high-temperature magnetization–temperature (M–T) curve of the CFTBO₄₆ sample shows that its glass transition occurs at about 600 K. The thin film is still ferromagnetic before the glass transition sets in. Therefore, the thin film should have a Curiecurie temperature above 600 K. At about 700 K, the magnetization increases owing to the apparent crystallization, similar to that found in Co₁₆Fe₆₈Hf₉B₇ alloy^{[37, 38]}. Together with its electrical behavior, it can be concluded that the CFTBO₄₆ thin film is semiconducting and ferromagnetic. A new type of AMSs has been developed with a high Curie temperature above 600 K through simply oxidizing an originally ferromagnetic alloy. The conduction type of this AMS is determined to be p-type^[26].

Based on the above results, it is suggested that the single-phase CFTBO₄₆ thin film is an AMS with a Curie temperature above 600 K. Integrated with a heavily doped n-type Si, p–n and p–n–p heterojunctions were fabricated (Figs. 7(a) and 7(b)). Their current–voltage (I–V) curves are shown in Fig. 7(c). The threshold voltage (V_th) is about 1.6 eV for the p–n heterojunction, whereas no current flows in the p–n–p heterojunctions.

The voltage difference (V_D) built inside the p–n heterojunction can be described by the following equation:

$ {V_{\rm{D}}} = \frac{1}{q}\left( {{E_{{\rm{i2}}}} - {E_{{\rm{i1}}}} + kT\ln \frac{{{N_{\rm{A}}}{N_{\rm{D}}}}}{{{n_{{\rm{i1}}}}{n_{{\rm{i2}}}}}}} \right). $  (1)

In the equation, q denotes the electron charge, E_i2 and E_i1 denote the Fermi energy levels of the intrinsic states of Si and CFTBO_x thin films, respectively, k denotes the Boltzmann constant, T denotes the temperature, N_A and N_D denote the carrier concentrations of the CFTBO₄₆ thin film and n-type Si, respectively, n_i1 and n_i2 denote the intrinsic carrier concentrations of the CFTBO_x and Si, respectively. Assuming that E_i2, E_i1, N_A, n_i1 and n_i2 are constants, V_D increases with N_D. A higher N_D indicates a lower ρ for n-type Si. Meanwhile, V_th of the p–n heterojunction is required to be larger than V_D for the forward conduction current to flow. As a consequence, V_th decreases with increasing ρ of the n-type Si as shown in Fig. 8.

Magnetic properties of ferromagnets are thought to be their inherent characteristics of the response to external magnetic fields. These magnetic properties are difficult to be changed once they are prepared^[39]. However, the electrical control provides an alternative route to effectively manipulate the spin degree of freedom in ferromagnets. The first experimental observation of the electric-field control of the ferromagnetism was reported in the (In, Mn)As DMS^[40], triggering intensive research interest in this subject. Hitherto, the electric-field control of the ferromagnetism has been realized in a variety of magnetic materials including metals, semiconductors and oxides^[41–43], which may allow the development of low-power and non-volatile spintronic devices^[44]. The physics behind the electric-field control of ferromagnetism results from the interplay between the carrier spins and local magnetic moments, which can be correlated with the carrier densities of various magnetic materials.

Fig. 9(a) shows a schematic diagram of a field effecttransistor (FET) with HfO₂ (2 nm)/ CFTBO₄₆ (25 nm)/Au/Cr/Si heterostructure. Through an ionic liquid (DEME-TFSI) dropped on the surface of the CFTBO₄₆ thin film, the gate voltage (V_G) is applied by forming an electric double layer (EDL) to achieve a large variation in its carrier density. Such a method has proven to be effective for realizing the stable modulation of carrier densities even when the gate voltage is removed^[45–48]. In addition, the EDL transistor technique has been used for the electric-field control of ferromagnetism in several ferromag- nets^[49–51], which works beyond the fundamental limitation of the charge screening effect^[52]. By varying V_G from –1.2 to 1.2 V, the room-temperature magnetization of the CFTBO₄₆ thin film is significantly altered by the applied electric field (Fig. 9(b)). When V_G is negative, the electrons of the CFTBO₄₆ thin film move down along the opposite direction of the applied electric field. As a result, its M_s increases. This indicates that its carrier concentration increases. Correspondingly, a positive V_G leads to a decrease in its hole concentration. Its M_s decreases. These experimental results indicate that the magnetic behavior of the CFTBO₄₆ MS can be modulated by electrical means. The presence of exchange interaction between the holes and localized magnetic moments in the thin film affects its ferromagnetism.

The present CFTBO₄₆ MS is a p-type semiconductor. It is found that most of the existing AOs are n-type and are notoriously doped in p-type. Moreover, the control of their electrical conduction type remains a challenge. The conduction type of the AOs is determined by their local atomic configuration and the valence states of the involved metal cations^[53].

To alter the conduction type of the present Co–Fe–Ta–B–O system, we modified the relative atomic ratios of Co and Fe in the system. As shown in Fig. 10(a), a new single-phase AO is developed. Its HRTEM image presents a maze-like pattern characterizing single-phase amorphous structure formed in the Co–Fe–Ta–B–O thin film. The electric-field control of its ferromagnetism is different from that of the p-type CFTBO₄₆ MS. No matter whether V_G is positive or negative, its M_s increases. It is suggested that the newly developed Co–Fe–Ta–B-O sample could be an intrinsic MS. Based on this intrinsic MS, we may use V_G to control its electrical conduction type of the different parts within the material, which would form a p–n junction at the interface between the different parts. To the best of our knowledge, this should be the first demonstration of an intrinsic MS. Meanwhile, the conduction type of this intrinsic MS can be tuned to be n-type or p-type by using an external electric field.

A new type of AMSs was developed by oxidizing originally ferromagnetic AAs. These AMSs showed high Curie temperatures well above room temperature. Based on them, prototype p–n heterojunctions were fabricated. Owing to the interplay between the holes and local magnetic moments, the electric-field control of ferromagnetism was realized at room temperature in these AMSs. The realization of high-Curie-temperature AMSs offers an opportunity to create next generation spintronic devices such as spin-diodes and spin-field effect transistors. The carrier mobility of amorphous materials is known to be very low. Therefore, these AMSs could not be good channel materials to be used in semiconductor-based devices. However, new phenomena may emerge at the interface between two AMSs if they are integrated together to form a field effect transistor. Therefore, furthering our understanding of the possible unprecedented interface effects may help to design and develop high performance spintronic devices based on these AMSs.

Future research subjects related to this new type of AMSs will be focused on the following three aspects:

(1) Both p-type and n-type AMSs are important for fabricating magnetic p-n diodes or spintronic transistors. The control of electrical conduction type in these AMSs should be possible through the compositional tuning based on the Co–Fe–Ta–B–O system. In this way, n-type AMSs could be developed to show high Curie temperature well above room temperature.

(2) New kinds of high performance MSs could be developed based on various ferromagnetic AAs. For example, the Co–Fe–Nb–B system is similar to the Co–Fe–Ta–B system. Oxidizing the Co–Fe–Nb–B AAs may lead to the formation of new AMSs with different properties. In addition to oxygen, nitrogen or sulfur can be used as elemental additions to transfer a ferromagnetic AA to AMSs. On the other hand, the precise characterization of the amorphous structure remains elusive from both theoretical and experimental aspects. A spectrum of behavior from ferromagnetic metal to insulator is realized in one Co–Fe–Ta–B–O system via continuously compositional adjustability. Hence, the amorphous structure can be manipulated in a well controllable manner, leading to tunable optical, electrical and magnetic properties over a wide range. This will help to build a reliable relationship between the amorphous structure and properties of these amorphous materials.

(3) New design concepts and device structures should be proposed for this new type of AMSs, aiming to combine logic functionalities and information storage capabilities in a single device through manipulating the spin and charge simultaneously. This will facilitate the development of next-generation spintronic devices with low power, rapid response and high storage density.

We thank Prof. Cheng Song, Prof. Xiangrong Wang and Prof. Xiaozhong Zhang for profound discussions and comments. This work is sponsored by the National Key R&D Program of China (Grant No. 2017YFB0405704) and the National Natural Science Foundation of China (Grant No. 51471091).