A System for Trapping Barium Ions in a Microfabricated Surface Trap

We have developed a vacuum chamber and control system for rapid testing of microfabricated surface ion traps. Our system is modular in design and is based on an in-vacuum printed circuit board with integrated filters. We have used this system to successfully trap and cool barium ions and have achieved ion 'dark' lifetimes of 31.6 s +- 3.4 s with controlled shuttling of ions. We provide a detailed description of the ion trap system including the in-vacuum materials used, control electronics and neutral atom source. We discuss the challenges presented in achieving a system which can work reliably over two years of operations in which the trap under test was changed at least 10 times.

Trapped ions are a promising candidate building block for a quantum computer. One of the most promising trap designs is the planar ion trap, with RF rails and DC bias electrodes fabricated from gold on a silicon chip using semiconductor fabrication techniques [1,2]. The basic design of these traps puts an ion in a RF pseudopotential null 50 µm to 200 µm above the chip surface. A vapor of neutral atoms flows through an aperture in the back of the chip. Photo-ionization and laser cooling is done with beams passing just above the surface. After loading ions, the position of the trapping region can be finely controlled by adjusting the DC voltage on multiple electrodes on the trap surface while moving the lasers to follow the ion. Ultimately multiple trapping regions can be defined and ion chains can be re-organized by shuttling ions to different regions.
Chip traps are currently fabricated by several groups around the world. We are working with devices provided by Sandia National Labs and Georgia Tech Research In-stitute (GTRI). Both these groups have standardized on a 100-pin Ceramic Pin Grid Array (CPGA) chip carrier architecture which gives the ability to control up to 96 DC electrodes.
138 Ba+ is an alkaline rare-earth metal isotope with nuclear spin-0. A key advantage of barium is that all important transitions for cooling and photo-ionization reside within either the visible or near-infrared spectrum, where coherent light sources are easily obtained and transmitted with optical fibers. This feature makes barium a good candidate for implementing a photonic link qubit as part of a large scale quantum computing architecture.

Chamber Design
Traditionally building an ion trap in a vacuum system involves wire wrapping and spot welding to make electrical connections. While these techniques are well known to be compatible with the Ultra High Vacuum (UHV) environment, they do not scale well to chip traps which might require up to 100 connections to be made. In addition to being tedious and error prone, large numbers of point to point connections can reduce optical access and impede molecular flow.
In order to facilitate easy mounting and frequent upgrades we have designed a UHV-compatible printed circuit board that holds a 100 pin Zero-Insertion-Force (ZIF) socket 1 for the chip trap. The circuit board plugs in to a set of 4 DB-25 female/female adapters 2 which in turn are connected to feedthroughs integrated into the bottom flange 3 of the vacuum chamber. A grounding shield with a window of thin wire mesh clips on to the top of the chip carrier. The circuit board incorporates 20 nF 0805 series ceramic capacitors approximately 1 cm from the chip for filtering RF pickup on the DC control pins.
The circuit board is made of ceramic-filled Polytetrafluoroethylene (PTFE) material 4 of 1.52 mm thickness with a 35 µm copper cladding. A particular difficulty with this material is its softness; it is easily deformed by insertion pressure which can cause broken connections that are difficult to find. For this reason the use of a ZIF socket is absolutely necessary.
Both the ZIF socket, DB-25 adapters and a small spacer on which the ZIF socket sits 5 are made of Polyether Ether Ketone (PEEK). The receptacle pins for holding the ZIF socket are a gold plated brass alloy 6 , as are the pins which plug into the DB-25 connectors 7 . The pins for connecting the RF rails are longer 8 and pass through the circuit board without connection. They are then wirewrapped 9 to a secondary feedthrough on the vacuum chamber. All on-board connections are soldered 10 and ultrasonically cleaned with flux remover 11 followed by acetone before assembly.
The chamber is a spherical octagon design 12 . Optical access for the laser beams is provided through 6 of the 2.75 inch side ports and optical access for imaging occurs through a viewport on the top 6 inch port. Vertical space of approximately 25 mm inch is available underneath the circuit board for mounting the atomic vapor source. Pumping is done with an on-system 20 L/s ion pump and a titanium sublimation pump. After baking at 150 • C for 1-2 weeks we routinely achieve a base pressure of a few 10 −11 Torr. Figure 1 shows an exploded view of the trap chamber.

Optical setup
The first step required to load ions from the atomic vapor source is photo-ionization, a process which requires 5.21 eV. The traditional approach to photo-ionization of barium has been to use a 791 nm laser to excite to the 6s6p 3 P 1 state (1.57 eV), followed by a UV flash from a 337 nm nitrogen laser (3.68 eV) or UV flash lamp to complete the ionization [13]. An alternative approaches uses a 2-photon scheme with a single laser at 413 nm initially exciting the 5d6p 3 D 1 and then ionizing [14]. Another recently demonstrated alternative uses a 553 nm laser to excite the strong dipole 4 Rogers ceramic (RT/duroid R 6002) 5 Manufactured in-house 6 Mill-Max part 0326-3-19-15-06-27-10-0 7 Keystone Electronics part 1358-2 8 Mill-Max part 0038-3-17-15-30-27-02-0 9 30 AWG, silver plated copper wire with Kapton shielding 10 Kester no-clean lead free solder (95.5% Sn, 3% Ag, 0.5% Cu) 11   transition 6s 2 1 S 0 → 6s6p 1 P 1 followed by any laser with wavelength less than 418 nm to ionize [15].
Because of the possibility of UV laser induced charging [16] of the chip surface and tight focusing requirements we decided on an ionization scheme where we retain the 791 nm laser and employ a secondary transition. The 6s6p → 6p 2 transition (2.75 eV) is driven with a 450 nm laser 13 . Finally, a tertiary transition is required, the 791 nm laser (or other cooling laser) can provide the necessary 0.91 eV to complete the ionization. An additional advantage of this scheme is improved isotope selectivity.
One difficulty encountered with this scheme is that after loading we observe immediate 'shelving' of the ion into the D 5/2 state. This process can be explained by a transition via the 7S 1/2 level in the barium ion driven by incoherent light from the 450 nm laser. To solve this we require a third laser for de-shelving at 614 nm. This laser is obtained by frequency doubling a 1228 nm ECDL 14 . The addition of this laser is not an inconvenience as we already require a de-shelving laser for our experiments. 13 In house design ECDL 14 Toptica DL-Pro   [19,20].
Primary doppler cooling is done on the S 1/2 → P 1/2 transition at 493 nm. This beam is derived by frequency doubling 15 a 986 nm external cavity diode laser 16 (ECDL). This doubling arrangement is particularly advantageous because it allows fast switching via pickoff of a first order beam from an acousto-optic modulator (AOM) to be done in the IR. The isolation ratio when switching has traditionally been a problem in experiments. Due to the non-linearity of second-harmonic generation we achieve a very high isolation ratio which is not directly measurable. The lower limit in isolation ratio from the measurement is 49 dB ± 6 dB.
A 650 nm laser is also required to re-pump out of the D 3/2 state. This is obtained from another ECDL 17 . A standard method for detecting the presence of poorly cooled ions in the initial stages of trapping is to momentarily switch off the 650 nm laser and observe the change in 493 nm fluorescence intensity.
An energy level diagram showing the relevant transitions in neutral and ionized 138 Ba is shown in Figure 2. Further details of the barium doppler cooling have been presented in references [13,17,18]. Figure 4 shows a simplified representation of the optical setup for a single beamline. The 986 nm, 650 nm and 1228 nm lasers are each frequency locked to temperature stabilized invar reference cavities in a hermetically sealed 15 Custom frequency doubling crystal from HCP photonics, single fiber pigtail. 16 Toptica DL-100 17 Toptica DL-100 chamber using a custom top-of-the-fringe locking system. In each a zeroth-order beam from a switching AOM is passed through a double-passed AOM (DPAOM) setup to derive the beam to scan the cavity.
All of the beams except the 791 nm beam reside within the transmission spectrum of Nufern 460-HP optical fiber 18 and are coupled into a single optical fiber. This is particularly convenient for trapping operations because all four beams can be focused on the ion with a single setup. Coupling is achieved by first arranging combined 'blue' and 'red' beams from the 493 nm + 450 nm and 650 nm + 614 nm beams respectively each using polarizing beam splitters. These two combined beams are then further combined using a dichroic mirror 19 and coupled into the angle-cleaved end of an optical fiber.
The output of the fiber has a plane-cleaved end and an achromatic microscope objective is used to collimate the output beams. This is reflected by a piezo-controlled mirror 20 which allows the beam to be automatically pointed to the new ion position after shuttling. Finally a translatabe lens is used for fine adjustment to focus the beams at the trapping region. Typical power levels for each of the beams at the fiber output and nominal trapping frequencies are shown in Table 1.
Efficient collection of fluorescence light from the ion is important for robust state detection. Our chamber setup makes use of a custom top mounted recessed fused silica viewport which allows imaging optics to be placed within 20.7 mm of the trap surface. Otherwise our imaging system is conventional, employing a 20x microscope objective 21 and second stage doublet lens.
The imaging beam path is split by a 50/50 beam splitter between a photomultiplier tube 22 (PMT) configured for photon counting and an electron-multiplying CCD camera 23 . A 493 nm optical bandpass filter can be switched in to the imaging beam path to block background light. Typical PMT count rates observed are 3000 counts/sec/ion against a background of 500 counts/s. Achieving a low background count rate requires tight focusing of beams and precise leveling relative to the chip surface.
The imaging stack is mounted on a translation stage which is actuated by NEMA-23 stepper motors coupled to the micrometers via a helical coupler and sliding square peg coupler. This allows the imaging system to be moved automatically to image ions in different locations on the trap surface.
Ion dark lifetimes were measured with single ions trapped in a Sandia 'Y' trap. This was done by repeat- 18 Strictly 650 nm is outside the transmission range but in practice can be coupled without difficulty in some fibers of this type due to manufacturing variations. 19 Thor labs DMLP567 20 Custom design. 21 Mittitoyo M-Plan APO 22 Hammamatsu H10682-210 23 Luca R series from Andor Technology edly mechanically shuttering the re-pump laser 24 for a varying length of time and finding the probability p l of ion loss after cooling was re-established. Dark lifetimes have not been defined in a consistent way in the literature. In practice we are mostly interested in a measurement of the delay time over which all loss mechanisms are insignificant. Therefore we define the dark lifetime τ d by making a weighted fit of as shown in Figures 3a and 3b. Uncertainties for each point are p l (1 − p l )/N , found assuming a binomial distribution where N is the number of measurements and the final parameter uncertainties are found from the covariance matrix produced after fitting with the Levenburg-Marquardt algorithm.
A dark lifetime of 20.0 s ± 4.9 s was measured with the ion directly over the loading region. After shuttling the ion to the mid point of one of the trap arms an improved dark lifetime of 31.6 s ± 3.4 s was measured. Some authors prefer the lifetime corresponding to 50% ion loss, which in our approximation is simply 1 2a +τ d . Our corresponding lifetime values in this formulation are more uncertain, 39 s ± 22 s and 48 s ± 27 s respectively.

Vaporization
Obtaining a beam of neutral atoms presents some unique challenges when trapping barium in a chip trap.
1. Barium oxidizes readily in air and consequently the barium source must be protected from air during installation. 2. The fluorescence of neutral barium atoms at 791 nm is extremely weak and it has not been possible to use it to confirm the presence of neutral barium. 3. Blocking of the small loading apertures of chip traps with barium clusters can render a trap useless. Figure 5 shows a SEM image of the surface of a Sandia ring trap after testing with a traditional barium oven. This chip was also imaged with an energy-dispersive X-ray spectrograph (EDS) and the blockage identified as barium oxide.
The traditional approach to the oxidation problem has been to arrange an inert gas atmosphere and load fresh barium metal flakes into a thin alumina tube which is heated by a tungsten coil. Despite this precaution, some oxide is inevitably formed and the oven must initially run significantly hotter than usual to clean off the oxide layer. It is during this initial 'break-in' of a traditional oven that we expect the majority of the clusters seen in Figure 5 originate. 24 Entirely shuts off the cooling in barium. We have adopted a 2-part solution for reliable generation of neutral barium.
1. We use a commercial barium oven 25 in which the barium is contained in a stainless-steel tube filled with argon and sealed with indium. The indium seal breaks when the oven is heated. In operation the oven is heated by running current of up to 10 A through the tube. A small current (typically 5 A) is run through the oven during the bake-out of the chamber. 2. An oven shutter has been developed. This consists of a thermostatic bimetal strip 26 with a stainless steel shutter attached. This is mounted to a Macor R block and mounted above the oven. When activated by running current through the strip it bends and moves the shutter to cover the oven tube. This protects the chip from excess barium flux and any indium from the seal. This is activated when the oven is run initially and continuously during chamber bake-out.

Electrical setup
Because of the large number of DC electrodes which must be controlled, using off-the-shelf equipment quickly becomes cumbersome and expensive. We have designed a custom electrode control system. It is based on three AD5372 digital to analog converters (DACs) and controlled by a    Figure 6: Block diagram of the electrical setup of the experiment control hardware. The PC controls the trap electrodes via the FPGA system. The PC also sends data to a microcontroller which is responsible for moving the lasers and camera to follow the ion. Camera data is sent to the computer via USB and PMT counts are recorded by a PCI DAQ card. This card is also used to control the lasers.
of the system is shown in Figure 6. Electrode control solutions divided into many steps for shuttling ions about the chip are supplied by the trap manufacturer and are typically calibrated for a calcium ion with voltages in the range of ±10 V. A control program running on the host PC parses the files and applies a chip and ion-species specific mapping to determine the voltages and corresponding DAC channels. The full sequence of bytes required to control the DACs is then calculated and transmitted via UDP 28 over an ethernet connection to the DAC system. The FPGA reads the data via a DM9000A ethernet interface and saves it to SRAM. Upon receipt of a control signal, the FPGA will then write the data in sequence to the three DACs.
This system has proven quite capable, allowing us to control 96 electrodes over a 20 V range (±10V) to within 0.3 mV at an update rate of 2.4 µs per channel with the ability to synchronize channel updates. The final update rate depends on the number of channels that need to be updated per step; this is typically around 5-10. The system is also low cost (around $500) and compact. Source code (python control program and Verilog FPGA code) and schematics are available to other researchers on request.
The application of radio frequency (RF) to the trap is conventional. We use a copper can helical resonator to filter and impedance match from a power amplifier to the RF rails of the trap. The internal RF wires are shielded and take the shortest route possible from the feed-through to the trap. The design of helical resonators for this purpose has been explored by Siverns et.al. [21]. 28 User Datagram Protocol

Conclusion
After operating two systems of this design for nearly two years we have found them reliable and easy to assemble. Traps can be swapped out and modifications made in a few hours. After a few days of baking followed by titanium sublimation pumping we can achieve a stable pressure around 10 −11 Torr. The signals applied to the DC electrodes are clean, owing to the isolation from the on-board filters. The AD5372 chips form a convenient solution for controlling the DC voltages and are suitable for setting up trapping regions and shuttling operations.