EM of Copper Damascene Interconnects Integrated Wtih Low-K

  • Upload
    mwuest

  • View
    214

  • Download
    0

Embed Size (px)

Citation preview

  • 7/28/2019 EM of Copper Damascene Interconnects Integrated Wtih Low-K

    1/1

    Electromigration Characteristics of

    Copper Damascene Interconnects Integrated

    with Low k- (FSG/USG) Dielectrics

    Du Nguyen & Hazara Rathore

    IBM Microelectronics

    2070 Route 52 East Fishkill Facility

    Hopewell Jct., NY 12533

    Phone: (914) 892-3162

    Fax: (914) 892-6850

    Abstract:

    The VLSI chip performance is improved

    by the reduction of the RC constant. To reduce the

    resistance, copper has been used as the conductor

    for the VLSI interconnects. To reduce the

    capacitance, fluorine silicon glass (FSG) is chosen

    to be embedded with undoped silicon dioxide

    (USG) to integrate with damascene copper

    interconnects for enhancing the chip performance.

    In this paper, we will attempt to illustrate that the

    electromigration life time of single and dual

    damascene lines which are insulated by a FSG

    layer embedded with USG layer as the interlevel

    dielectric (ILD) are excellent without any

    degradation in life times due to the presence of

    FSG. In addition to the electromigration data, we

    will show interlevel voltage ramp data and limited

    life stress data using copper electrodes.Experiment

    The copper used for this experiment was an

    electroplated (EP) copper. The EP copper was

    deposited after the depositions of diffusion barrier

    / adhesion layer and copper seed layer. The EP

    copper lines integrated with FSG / USG were

    evaluated by using test structures with single

    damascene and dual damascene line structures.

    The test structure of a single damascene is built

    from local interconnect level made of tungsten to

    interconnect with metal stripe at level M1 by

    interlevel contact via, CA. The M1 stripe wasformed by using a single damascene process which

    was manufactured by using a FSG layer as an

    interlevel dielectric (ILD). M1 trenches were

    formed by using lithography processes and a RIE

    etch step to etch into the ILD. After the M1

    trenches were formed, a blanket deposition of

    diffusion barrier / adhesion layers, copper seed

    layer and then followed by an electroplated copper.

    A chemical mechanical polishing (CMP) process is

    used to remove the excess copper and liners in

    forming single damascene copper stripes. A

    blanket deposition of silicon nitride is deposited toencapsulated the copper lines as a capped layer to

    protect any chemicals attack metal lines from the

    top down. Following the single damascene stripe is

    a dual damascene process which is used to form

    both interlevel V1 vias and M2 stripes. A thick

    layer of undoped silicon dioxide (USG) is

    deposited with the thickness is the same as the

    height of the V1 via, then follows by a FSG layer.

    The thickness of the FSG is determined by the M2

    thickness. In this experiment, a dual damascene

    metal sequence is using a lithography process to

    determine metal stripe trenches and a RIE process

    to etch into ILD to form metal stripe trenches.

    Using a lithography process again to determine via

    opening and using RIE process to etch via holes.

    After the trenches were formed for via and metal

    stripes, diffusion barriers and copper seed layer

    were deposited then follows by an electroplated

    process to fill in to form a dual damascene

    interconnect. A silicon nitride layer also used as an

    encapsulated later. The process was repeated in the

    same to form dual damascene metal stripe at M3

    and M4, then the test structure is covered by thick

    oxide and nitride layer for the final interlevel vias

    to be ready for wire bonding. The test structure is

    400 um long, .315um wide and 0.4 um deep. The

    structure is a four point probe structure by having 2

    pads for each end. One pad from each end is for

    applying current source and the other pad is for

    sensing the voltage. The test was conducted at 295

    degree centigrade at 25.0 mA/um2. The failure

    mechanism of dual damascene line when the

    electrons flow from M1 go up to M2 is caused by

    via depletion which shows voids formed in V1 or

    on top of V1. When the electrons flow from M3

    down to M2, the failure mechanism is line

    depletion which shown by voids underneath of V2

    via.

    Because of CA made of tungsten, there is no

    electromigration failure mode in single damascene

    line but line depletion is the main failure mode in

    both directions for M1.

    Conclusions:

    FSG is a good insulator to be integrated with

    copper interconnects because it does not show any

    degradation of electromigration life time and it

    help to improve the chip performance.

    References:

    1) J.C. Maisonobe, G. Passemard, C. Lacour, P.

    Motte, P. Noel, J. Torres, SILK Compatibility

    with IMD Process Using Copper Metallization,Microelectronics Engineering 50 (2000) 25-32

    2) E. Hartmannsgruber, G. Zwicker, K.

    Beekmann, Selective CMP Process for Staked

    low k- CVD oxide films, Microelectronics,

    Engineering 50 (2000) pp. 53-58

    3) Chanming Jin et al. Thermal Conductivity

    Measurements of Low Dielectrics Constant Films,

    DUMIC Proceedings, Feb. 20-21 1996

    4) R. J. Gutmann et al, Integration of copper

    multilevel interconnects with oxide and polymer

    interlevel dielectrics, Thin Solid Films 270 (1995)

    pp. 472-4795) M.J. Loboda, New Solutions for Inter metal

    Dielectrics Using Trimethyl silane based PECVD

    process, Microelectronics Engineering 50 (2000)

    pp. 15-23

    6) H. Rathore, D. Nguyen, L. Ashley, G. Biery, B.

    Agarwala, T. Kane and P. Plaitz, Proceedings 16th

    International VLSI Multilevel Interconnection

    Conference p. 89 (1999)

    7) H. Rathore, D. Nguyen, MRS 2000, Spring

    Meetings, p. 84 (2000)

    8) H. Rathore, D. Nguyen, B. Agarwala, and L.

    Ashley SemiconWest B-1 (1999)