Tracking in High Energy Physics: Silicon Devices! G. Leibenguth XIX Graduiertenkolleg Heidelberg 11-12. October 2007
Content Part 1: Basics on semi-conductor Part 2: Construction Part 3: Two Examples Part 4: Future (radiation, 3d processing, )
Disclaimer and Reference Do not expect to have a solid state physics lecture Reference: R. Horisberger, talk given at the conference vertex 2006 C. Delaere, talk given at the conference vertex 2007 A. Homna, lecture given at the NATO advanced study (2002) And material taken from colleagues from CMS, ATLAS and other
Outline of the Lectures Today Lecture 1 A bit of history Motivation Silicon strip detector (principle) Performance required Lecture 2 Fabrication of sensors Constructions of detector modules Readout electronics
A bit of history First use of silicon detectors in HEP experiments: the 50 s: energy measurement Precision measurement up until 70 s: Emulsion or bubble chambers => Limited rate and no triggering Traditional gas detectors (limited to a spatial resolution of 50-100 µm) First usage (late 70 s) precision measurement (lifetime): Secondary vertex tagging (charm, in fixed target) Segment sensors (strips) with pitch First silicon pixel devices (NA32) charm experiment
Why wasn t it used before? First Reason: Excessive material (electronics) in active volume Advance in electronics miniaturization and low composite structures Late 80 s, early 90 s: Mark II (SLC) and all 4 LEP experiments Use of Silicon limited to small region near interaction point: both silicon and electronics were very expensive
Motivation Charged particle position measurement: Ionization signal (de/dx) left by a charged particle crossing the detector _ + -_ - Use the drift chamber analogy (see C. Niebuhr s lecture) where ionization produces electron-ion pairs, and an electric field is applied to collect electrons and ions to the opposite charged electrode In a semi-conductor, pair of electron-holes are produced with 3.6 ev. Quickly recombine. How are the charges drifting?
Collected charge From the Bethe-Bloch s formula (see C. Niebuhr lecture), One can calculate the minimum ionization deposition: de/dx Si = 3.88 MeV/cm. For 300 µm thickness, 116 kev Most probable loss (0.7*mean) => 81 kev 3.6 ev to make an e-h pair Max collected charge: 22500 e (=3.6 fc)
Solid states physics What does one need to know from the principle to design, build and operate a silicon detector: Property of material Energy bands Junction PN
Property of Material
Semi-conductor Silicon, Germanium, Diamond GaAs?
Semi-conductor: Silicon Most important semi-conductor for detector development Is silicon (even so Germanium is used for calorimetry, Thanks to an excellent energy resolution) Silicon makes up to 25.7 % of earth s crust by weight, and the second most abundant element, being exceed only by Oxygen. Silicon is not found free in nature, but appears as oxide and as silicates: 1) Sand, quartz, rock crystal,amethyst, agate, flint 2) Granite hornblende, asbestos, feldspar, clay, mica, etc
More details on Silicon Atomic Number (A): 28 Charge: 14 Melting Point: 1687 K Atomic Symbol: Si Boiling Point: 3538 K Atomic Weight: 28.0855 Density: 2.3296 g/cm 3 Electronic Configuration: (Ne) (3s) 2 (3p) 2 Ionization Potential: 8.1eV 4 Valence electrons
Reminder: Crystallography Simple body centered Face
Body centered cubic structure The body centered cubic structure shown a two interlinked simple cubic structure Many metals own this structure, like iron, lithium, potassium, etc
Miller Indices, Si (111) Silicon (111) plane (solid line) and a photograph of a Si 111 surface
Valence Bond Miller Indices, Si (111) The silicon crystal is bound via valence bonds 4 valence electrons: Silicon is tetravalent Each electron can be only associated with: Positive ion cores should be kept apart to minimize the coulomb energy Valence electrons should also be kept apart Valence electrons should stay close to positive ions to maximize attraction
Energy band Reminder: free electron! k = 2 h 2m k r 2 Plot of energy versus wave vector k for a free electron and for an electron in a mono-atomic linear lattice (constant a)
Conduction band The probability that an electron occupies an Energy level is given by the Fermi-Dirac Distribution: For N e = N h E f = E gap /2
Intrinsic semi-conductors
Acceptors and donors If the semiconductor is doped then impurity energy levels are introduced For example, if a boron atom replaces a Si atom then as the boron owns only three valence electrons, a hole is created in the crystal. Using an atom of arsenic donate a electron to the conduction band, making it more N (negative) type
Extrinsic semi-conductor One can dramatically change the conductivity property of a semi-conductor by adding impurities: more electrons or more holes (we speak of extrinsicness ) Add to molten valence 4 atom with valence 3 or valence 5 during crystallization High speed injection of such kind of atoms Net result: extra allowed energy level in the forbidden band gap, electron can jump from the valence band (p-type doping)
N-type N = N d Fermi level becomes: E c - E f = kt ln(n c /N d ) Number of donors is increased while the Fermi energy move toward the conduction band
P-type P-type is an acceptor: Replacing some silicon atom by boron: The Fermi energy moves also toward the conduction band
P-type viewed from eg Level
Signal formation in silicon At t=0 o K, conduction band empty Distribution according to Fermi- Dirac statistics Number of electrons in the conduction band a temperature T: Implies: Ratio of electrons in the conduction band is 10-12
A MIP in a silicon slice A volume of 1cm x 1cm x 300 µm contains about 4.5 x 10 8 free charge carriers compared to only 23000 electron-hole pair for a MIP at room temperature Number of free charge carriers is way too much Solution: Cooling PN Junction
PN-junction Exploit the properties of a p-n junction to collect the ionization charges: When brought together appear a gradient of electron and hole densities resulting in a diffusive migration of majority carriers across the junction.
PN-junction The migration leaves a region of net opposite charges: the space charge region Electric Field prevent move of charge carrier e-h pairs will not recombine, but drift away
PN-junction, depletion region Make the junction p-n at a surface of a silicon wafer with the n-type bulk, extend the depletion zone throughout to get the maximum charge collection To get it, apply a reverse bias voltage (why reverse => direct results in a current flow)
Property of the depletion zone Depletion width depends on the resistivity, charge carrier and magnitude of bias voltage V b w = (2ερµV b ) Bias Voltage needed to completely deplete a device of thickness d is called the depletion voltage Two remarks: V d = d 2 / (2ερµ) Higher voltage for low resistivity material Higher voltage also needed for a p-type bulk (carrier mobility of holes lower than electrons µ(mobility) ρ=1/qµn, N number of doping concentration
Property of the depletion zone Capacitance is simply the parallel plate capacity of the depletion zone Depletion behavior is obtained by measuring the capacitance versus reverse bias voltage: C = A ( ε / 2ρµV b )
Charge collected & diffusion Drift velocity of charged carriers v = µe, so drift time is: t d = d/v = d/µe Typical value d=300µm, E=2.5 kv/cm, µ e =1350 cm 2 /V.s, µ h=450 cm 2 /V.s, so t d (e)=9ns, t d (h)=27ns Diffusion of charged cloud caused by scattering of drifting charge carrier, radius of distribution after time t d R= 2Dtd) where D= µkt/q
Leakage currents Two main sources of current flow in reversed biased diode: Diffusion current, charge generated in the undepleted zone close to the depletion which diffuse into it. Should be negligible in a fully depleted device Generation current J g,charge generated in the depletion zone by defect or contaminants J exp(-b/kt) (Exponential dependence on temperature) Rate is determined by nature and concentrations of defects => Major Contribution!!!
Surface and edge effect a) Voltage drop between ring and edge b) Typical n-type implants put around edge c) Guard rind structure (floating) to improve continuous potential drop over this region d) Defect or oxide charge build-up might increase the leakage current
Breakdown If V bias large, field might be high enough to initiate avalanche multiplication, ie charge carriers have enough energy to produce more e-h pairs. Occurs around 30V/µm. Local defect can result in field reaching the breakdown point 1 na
Current in diode For a 300 µm thick sensor, about 24000 e/h pairs are created on the passage of a MIP. With a collection time of approximately 2 ns the peak current generated due to the charge deposition is approximately given by I ~ 24, 000" 1.6" 10 2" 10! 9! 19
DC coupled strips Diode is not very useful for high resolution: The assembly of many individual, tiny diodes, to form a large area tracking device would be both a laborious and expensive task. Each diode would also possess dead areas around the edge of the silicon leading to excess material, the need to overall and excessively complicated mounting assemblies Segmentation of the readout Strip geometry
Geometry: consideration does the pattern match the needs of the experiment? is this pattern consistent with the properties of silicon e.g. compatible with diffusions distances Will the segmentation yield a useful signal to noise and (closely linked) an adequate spatial resolution? (remember, we would like to measure lifetime down to picosecond)
DC readout Example: The H1 central silicon tracker Problem: reversed current (also known as Dark current)
Charge collection 1) Need to isolate strips from each other 2) Collect (measure) the charge on each strip High impedance bias, directly integrated on the sensor Issue: Strip defects Noisy strips: usually high DC leakage current Short between strips and strip to other structures Opens: Interruption of strip or no contact between metal and implants
AC coupled strips The AC coupled detector is designed to Overcome a major obstacle of the DC devices, that of the dark current in the diode: overtax the amplifier => saturation Reduce the dynamic range Noise on the current might mask the small high frequency signals from the particle we want to detect => Capacitive coupling to the readout electronic, which induces a charge qc on the input of the amplifier
Double sided detector If the strips on opposite sides of the sensor are placed at right angles to each other a space point (rather than a single coordinate) will be produced for the hit. Double sided detectors are very useful but complex and expensive to make and readout. It is often easier to use two single sided sensors in close proximity to achieve the same effect. => See the discussion later
Performance Resolution: strip pitch and readout pitch Capacitive charge division Signal to noise ratio Noise Radiation damage
Position resolution Strip and readout pitch: Assume that the detected charge is treated in a binary way, the resolution is simply: σ = p/ 12, so for a pitch of 50 µm, get 7.2 µm If the charge distribution is shared among adjacent strips, (usually the case), one can calculate the centre gravity of the charge deposit, leading to resolution below 10 µm Test devices have achieved a resolution of about 3 µm (using a readout pitch of 25 µm) which is near the limit on precision determined by diffusion and stat fluctuation of the ionizing energy deposition
Capacitive charge division Can we afford to read out all strips? When not Read out only every n th strip while preserving the signal magnitude. A good estimate of the centroid resolution is still obtained However, the signal to noise ration becomes very crucial
Signal to noise ratio (S/N) Landau distribution has a long low energy tail Becomes even lower with addition of noise
Main source of noise Capacitive load (C d ) dependence (ENC C d Sensor leakage current (shot noise) Parallel resistance of bias resistor (thermal noise) (kt/r) Noise generally expressed in the form: ENC = a + b. C d Noise is very frequency dependent (readout method) Implication on detector design: Strip length, device quality, choice of bias method Temperature important for both leakage and bias resistor component
Signal and noise Basic signal produced is around ~ 22500 electrons Typical losses of 5 to 10% due to the chosen electrical network (AC coupling capacitor, stray capacitances and resistances) and to the front-end electronics Noise: Usually expressed in term of equivalent noise charge (ENC) in unit of electron charge. The source of noise depends largely on the electronic Front-end
Radiation damage Impact of radiation on silicon: Silicon atoms can be displaced from their lattice position: Point defects (electromagnetic radiation) Damage clusters (nuclear reaction) Non-ionizing energy loss (NIEL)
energy level in the band gap Direct excitation possible Higher leakage current => (more noise) Charge trapping (lower charge collection) Efficiency Higher bias voltage mandatory
Non ionizing energy loss Main damage for silicon strip devices is lattice displacement damage of silicon atoms (atomic displacement via nuclear interactions) For EM radiation need E>250 kev, produced point defect For neutron and charged hadrons, damage start at low energy Important: Damage creates large numbers of new donor And acceptor states: 1) Change of charge density in depletion region 2) more generation -recombination center => larger I leak
Effects After irradiation, the observed damage looks reduced. The rate of this is Highly temperature dependent. Annealing is partly due to true Annealing (repair of lattice defects, faster at higher temperature) But for N eff two annealing effect: Beneficial (short time scale) Reverse, cause damage to increase
Effect on the depletion zone N-type silicon become increasing p-type until the substrate Undergoes type inversion Voltage for a p + -n silicon device is: V dep α 1/ρ α 1/ N eff (N efi is the effective Doping concentration)
Effects on Leakage current: ΔI leak = αφ, α is known as the damage constant and φ is the fluence. No reverse annealing effect Thermal runaway: irradiated device heating up! Dangerous, since leakage currrent srrongly depends on temperature Charge collection: Increase of trapping states in the depletion region (10% loss of signal for φ = 2 x 10 14 n/cm 2 )
Summary Silicon property: band gap 1.6 ev A mip passing through the detector leaves 22500 electrons-holes pairs With a bias reversed PN junction, ability to measure this charge (3.6 fc) Signal over Noise is a crucial quantity Radiation hard devices needed we will discuss the different steps of the construction of a silicon detector