The atom chip is an intensively developing technology, which provides high repetition rates, compactness, and precision of ultracold atoms manipulation. The key advantage of using the atom chips is the ability to create high magnetic field gradients near a chip surface using relatively low currents. This is very important for magnetic trapping of cold atoms and evaporative cooling, which are essential steps in the formation of ultracold atomic ensembles. The high gradient of magnetic fields allows the implementation of effective thermalization of an atomic cloud at the evaporative cooling stage and, therefore, enables us to cool atoms quickly up to Bose–Einstein condensation (BEC). The rate of BEC formation reached 1 Hz in a compact mobile setup (less than 0.5m3 volume) using atom chip technology^[1,2]. High repetition rate and compactness are important for quantum sensor development, especially when one puts them on a moving vehicle^[3]. At the same time, experimental setups based on the atom chip technology are robust enough to launch them into space to realize atomic interferometry in microgravity conditions^[4,5]. Atom chips are also used to create a variety of quantum sensors aimed to measure inertial forces with high precision. Gravimeters based on such interferometers show precision of better than 1μGal (1Gal=1cm/s2)^[6].

Another potential application of atom chips is the atomic clock. Secondary frequency standards (which are based on vapor cells in most cases) nowadays need to be periodically remotely compared with primary atomic fountain standards to correct long-term drifts. The smaller the long-term drifts, the less often it is necessary to compare secondary standards with the primary one. A long-term drift of a microwave atomic clock based on an atom chip should be less in comparison with other secondary frequency standards. At the same time, such an atomic clock could remain compact enough to put it on a moving platform for various practical applications. There is at least one complete realization of such a clock (in a laboratory)^[7].

One of the main disadvantages of quantum sensors based on atom chips is the limited volume of the initial magneto-optical trap (MOT) on a chip and, consequently, the limited number of localized atoms N. The fundamental precision limit of a quantum sensor or an atomic clock is quantum projection noise, which is proportional to 1/N^[8], so one should increase N to improve precision. An increase of trapping volume also leads to an increase of the loading rate; therefore, a large trapping volume is crucial for reaching high repetition rates.

The ordinary way to increase the trapping volume near the atom chip is using an additional stage of cooling consisting of the use of conventional MOT coils^[9,10] or layers with mesoscopic structures^[2,11] for quadrupole magnetic field generation. Such approaches allow the generation of a big trapping volume with an optimal magnetic field. But using an additional stage of cooling near the atom chip makes atom chip production and the experimental procedure complicated. From the other side, it is convenient to use a single-layer atom chip to simplify atom manipulations^[12]. The direct loading of an atom chip was demonstrated^[13], but the number of atoms was rather small. In this article, we present a new optimized design of a single-layer atom chip, which creates an MOT of larger volume, while an atom chip remains relatively simple in production and compact and requires moderate current (<10A).

MOT theory assumes presence of quadrupole magnetic field B, i.e., field changing linearly along each direction: B=exBx′x+eyBy′y+ezBz′z,(1)with typical optimal gradients Bx′, By′, Bz′ from 5 to 20 G/cm for rubidium atoms. The conventional way to produce such a field is a pair of anti-Helmholtz coils. It creates a good approximation of the quadrupole field so that the volume of an MOT is usually limited only by the laser beams size.

In case of the atom chip, the quadrupole magnetic field distribution can be achieved using a U-shaped wire and external homogeneous magnetic field^[14]. This kind of MOT, named U-MOT, can be used as a first stage of cooling and trapping atoms near an atom chip. The fundamental principle of U-MOT operation is the same as conventional MOT. Quadrupole distribution near an atom chip obeys Eq. (1) only in a small area near the central point (x=0, y=0, z=0). At other points, the direction of the magnetic field vector deviates from the ideal quadrupole field distribution. The deviation of the magnetic field makes trapping less effective^[11]: the effective trapping volume becomes limited by the imperfection of the field, rather than by the size of the laser beams. According to the theoretical consideration^[15], the number of trapped atoms is proportional to r4, where r is the effective trapping region radius. So, it should be possible to increase the number of atoms in U-MOT significantly by improving magnetic field distribution.

There are two methods to improve the magnetic field distribution near the atom chip. The first is using multiple wires to compensate undesired multipole components of the magnetic field^[16,17]. An advantage of this approach is that one does not need any external homogeneous field. The second method is creating U-wire with a wider central part^[11,18], which also makes it possible to compensate for the undesired components of the magnetic field.

Both methods were demonstrated experimentally^[11,16–19]. However, in both cases, mesoscopic (mm-scale) wires with high electric current (up to 120 A) were used. In the second method, the investigators created a multi-layer device (containing several atom chips) to reload atoms from a wider trap to a narrow one. Both approaches make setups more complex, so setups lose the basic advantages of the atom chip technology, such as compactness and low current consumption.

Figures 1(a) and 1(c) show the design and magnetic field distribution of our first-generation atom chip, which is described in Ref. [20]. There are two types of microwires on the chip: Z-wires and end-wires. A Z-wire (the width is 100 µm) alone can be used to create a magnetic trap. A Z-wire and an end-wire (the width is 200 µm) being switched on simultaneously create U-MOT. Such a design is similar to the one developed in Ref. [21], and it is conventional in the sense that a lot of groups use atom chips with narrow microwires (with the width of about 100 µm)^[4-7].

Figures 1(b) and 1(d) show the design and magnetic field distribution of our new-generation atom chip exploiting a wide U-wire approach. The width of the metal microwire is 2.9 mm, and the length is 6.2 mm. It corresponds to the dimensions of the wide U-wire taken from Ref. [11] with minor modifications and miniaturized to fit our chip. The thickness of the metal layer is 7.5 µm.

The new-generation atom chip still has a set of narrow wires with the width of 100 µm (Z-wire) and 200 µm (end-wire). That means that we can reload atoms from the wide-wire U-MOT to the narrow-wire U-MOT. In such a way, we can improve the magnetic field distribution in the first stage of laser cooling and trapping near the atom chip, while the setup remains relatively simple and has low current consumption. In the following stages of the experiment (compressed MOT, magnetic trap, evaporative cooling), atoms were reloaded to the narrow-wire U-MOT. This is essential because narrow wires could create higher magnetic field gradient.

The optimization of the quadrupole field was carried out as follows. The total magnetic field was chosen using the model of a thin infinite wire, which gives z0=μ0I2πBbias,B′=μ0I2πz02,(2)where z0 is the distance from the wire to the point of zero magnetic field, μ0 is the permeability of free space, I=3.2A is the typical current through microwires, Bbias is the external bias field, and B′=15G/cm is the magnetic field gradient at the point of zero magnetic field required for rubidium MOT. Therefore, Bbias=3.5G was fixed during the calculations. The only free parameter left was the angle between the external bias magnetic field and the plane of the chip. This angle was optimized to get the best quadrupole magnetic field. It was equal to 24°.

It can be seen from Fig. 1 that the wide wire creates the magnetic field that is closer to the ideal quadrupole field (the black lines indicate magnetic field axes of ideal quadrupole field). However, the quality of the magnetic field distribution can be estimated quantitatively only by comparing the number of atoms that can be trapped in such fields.

The investigation of the cooling efficiency in the U-MOT formed by the two types of microwires was realized with Rb87 atoms. The experimental setup and procedure are analogous to the ones described earlier^[20]. Figure 2 shows the scheme of experimental setup. The loading of U-MOT was realized from a low-velocity intense atomic source (LVIS)^[22,23]. The LVIS was formed in a separated vacuum chamber connected to the main chamber (containing atom chip) through a hole with a diameter of 1 mm. Such differential pumping allows us to keep the pressure near the atom chip during the experiment not higher than 5×10−10Torr. Moreover, the hole was used to form an atomic beam. A conventional MOT was used as a source of atoms for the atomic beam. This MOT was realized by a pair of anti-Helmholtz coils with three pairs of laser beams. One of the reflecting mirrors with a hole (diameter is 1 mm) was glued inside the vacuum chamber close to the hole for differential pumping. In such a way, a tunnel was formed in the reflected laser beam intensity distribution. A cold atomic beam formed inside the tunnel and escaped to the main vacuum chamber through the hole for differential pumping. The atomic beam in the region of the atom chip interacted with two pairs of laser beams in mirror-MOT configuration.

The same lasers were used for both LVIS creation and laser cooling in U-MOT. The first laser, a cooling laser, was stabilized on the crossover resonance F=2→F′=2,3Rb87 using saturated absorption spectroscopy. An acousto-optic modulator (AOM1) was used to tune laser frequency. It shifted the laser frequency by 120 MHz, so the final detuning from the cooling transition F=2→F′=3 was about δ=−13MHz (see the inset of the Fig. 2). The frequency of the repumping laser was stabilized on the F=1→F′=1,2 crossover resonance and shifted with AOM2 to the frequency of F=1→F′=2 transition. After frequency tuning, both lasers were combined together, and the power was amplified with Toptica BoosTA. The amplified laser beam was divided into two parts: (i) for LVIS creation and (ii) for cooling atoms near the atom chip. A mechanical shutter was used to block laser radiation for LVIS. That gave the possibility for switching off the LVIS and operating with cold atoms near the atom chip without additional atoms from the atomic beam.

The magnetic field of U-MOT was formed by the current through the microwires of the atom chip and external homogeneous magnetic field. This field was produced by a pair of Helmholtz coils. It is possible to control the distance between cold atoms and the atom chip by controlling the coils’ current.

We created the new atom chip with a wide U-wire (Fig. 3). The chip is made from a thin layer of chromium (100 nm) and a thick layer of silver and is covered with a thin (250 nm) layer of gold. The total thickness of the chip is 7.5 µm. The manufacturing process is described in detail in Ref. [20]. Our measurement shows that, despite relatively small thickness, the wide wire can continuously pass a current of 10.5 A, while heating up to no more than 70°C.

While the chip contains not only the wide U-wire but also a set of narrow wires, we could compare effectiveness of the two types of traps directly in the same experimental conditions. Figure 4 shows photographs of Rb87 atoms’ fluorescence from the narrow- and the wide-wire U-MOTs. The U-MOTs were loaded using LVIS. The operating parameters (chip current and external magnetic field) were optimized independently for each case. The external bias magnetic field was optimized to reach the best field distribution. Optimization was carried out by measuring the fluorescence signal amplitude (proportional to the atom number Nat) during the changing of external bias field Bbias under fixed current through the microwires of the atom chip. The number of trapped atoms was determined from integral fluorescence signal using independent calibration of the sCMOS camera. The numbers of localized atoms are approximately 14 million in the narrow-wire U-MOT and 50 million in the wide-wire U-MOT, which gives a ratio of 3.5 in favor of the wide-wire U-MOT. These numbers are limited by the background gas pressure, which was 5×10−10Torr during the experiment. In both cases, the temperature of atoms was about 140 µK.

To implement a magnetic trap on a chip, it is necessary to use narrow wires, since only they allow the achievement of sufficiently high magnetic field gradients. Therefore, after the accumulation of atoms in the wide-wire U-MOT, it is necessary to reload them into the narrow-wire U-MOT, in order to then reload them quickly into a magnetic trap based on the narrow wires. Figure 5 shows the process of atomic cloud reloading from the wide-wire U-MOT into the narrow-wire U-MOT, which was realized by rapid switching between the corresponding optimal parameters (chip currents and external magnetic field). The whole process lasts approximately 20 ms, which is determined by the inductance of magnetic coils. The resulting reloading efficiency (the ratio of the number of localized atoms before the reloading to that just after the reloading) is above 95%. That means that the use of the wide-wire U-MOT is really justified and is expected to give 3.5 times gain in the number of localized atoms at the subsequent stages of the experiment.

We realized an optimized single-layer atom chip, which enables us to increase the number of localized atoms in U-MOT by a factor of 3.5. Note that the approach of utilizing several wires for the compensation of undesired magnetic multipole components^[17] leads to a lower factor (less than 2). At the same time, our chip design is relatively simple (single-layered) and compact and consumes less current in comparison with alternative approaches. Our approach can be useful in both fundamental and applied physics problems when one needs to realize compact (e.g.,  portable) and efficient ultracold atom equipment.

We note that the atom chip can be improved further. It can be combined with an optical grating chip for compactness^[24] and other schemes^[25]. It is possible to increase the atom number in the U-MOT using atomic beam focusing in the region of efficient cooling^[26,27]. Such an atom chip can be used as a source of atoms for space missions, for example, on board the International Space Station or Chinese Space Station^[28].