dissenting.
I agree with this panel that the district court erroneously concluded that claims 12 and 25 were invalid for indefiniteness. I also agree that the district court erred in relying on the inventors’ scientific articles to determine the scope of the claims. Unfortunately, I cannot support the rest of this court’s opinion.
As the panel points out, the district court erroneously based its decision on the inventors’ articles in scientific journals. Although a patent specification may supply guidance about the meaning of claim terms, Hormone Research Foundation v. Genentech, Inc., 904 F.2d 1558, 1562, 15 USPQ2d 1039, 1043 (Fed. Cir.1990), scientific literature differs in purpose, scope, and legal effect from patent writings. Because inventors generally have extraordinary skill, their scientific writings outside the patent are rarely even a source of knowledge about ordinary skill in the art. The panel’s opinion correctly notes the trial court’s error in relying extensively on the inventors’ scientific articles for its claim interpretation. The trial court referred to the scientific writings as “the most important point in the evidence.” Surprisingly, however, the panel’s opinion did not consider this error a cause for reversal. In a case the district court repeatedly characterized as “very close,” this extensive reliance on impermissible evidence constitutes reversible error. Without the primary source of evidence upon which the district court based its decision, this “very close” case would no doubt have ended differently.
The majority suggests that the meaning of the term “conjugate” was conceded by North American Vaccine. In context, however, it appears that counsel was only trying to argue that the interpretation of the term “conjugate” is not dispositive. Moreover, plaintiffs reply brief actually contests the majority’s interpretation and argues that the district court’s ruling was incorrect. The ma*1581jority hinges its conclusion on an equivocal statement made to pursue a more important concern. This approach appears unwarranted under the circumstances.
Claim 11 — “at a terminal portion”
An examination of the claim language shows that the district court’s use of impermissible evidence infected the judgment in this case. Claim 11 requires linkage of a protein “to a terminal portion of the polysac-charide.” The accused product literally reads on this requirement. The panel’s opinion and both parties concede: “The protein linkage in each of these components is at a terminal portion of the Hib polysacchar-ide_” Maj. op. at 1574.
A branched polysaccharide, however, can have two or more terminal ends. The accused product linked protein at two of these terminal ends. The trial court, relying on the inventors’ scientific writings about “mo-nofunctionality” or “monovalency” (technical terms for linkage at a single terminal end), interpreted the claim language to mean that any more than a single link at a terminal end escapes infringement.
The claim, however, does not require linkage “at a single terminal portion of the poly-saccharide,” but simply “at a terminal portion.” The trial court used extraneous scientific writings to create a limitation — at a single terminal portion — not found in the claim language. Indeed in patent claims, the indefinite article “a” may denote one or more objects. Robert C. Faber, Landis on Mechanics of Patent Claim Drafting 531 (3d ed. 1990) (“In a claim, the indefinite article A or AN connotes ‘one or more’ ”).
Thus, according to the claim language, each link at a terminal portion of the polysac-charide reads on the claims. American Cyanamid infringed the patent by linking a protein to “a terminal portion” of-the poly-saccharide. When American Cyanamid linked a protein to two terminal ends, it still linked to “a terminal end” and infringed the claim. American Cyanamid indisputably always links at a terminal portion of the poly-saccharide, but argues that the trial court correctly read “a terminal portion” to mean “a single terminal portion.” The claims simply do not contain that limitation.
If the trial court had properly relied on the patent specification or prosecution history to construe the claims, it would have noted that these correct patent sources do not even mention “monofunetionality” or “monovalen-cy.” A comparison of the inventors’ scientific writings with their later patent writings shows a complete deletion of all references to “monofunetionality” or “monovalency.” Ironically, the scientific writings, in the proper context of the patent, show that the inventors — by deleting references to “monofunc-tionality” — clearly did not limit the claims to linkage at a single terminal end.
The panel’s opinion recognizes the trial court’s error in its extensive reliance on impermissible evidence. Nonetheless the panel compounds the error by creating on its own the same limitation. The panel purports to find this limitation in the specification.
In the first place, this court of appeals prohibits importation of limitations from the specification into the claims. In re Lundberg, 244 F.2d 543, 548, 113 USPQ 530, 534 (CCPA 1957). The panel’s belated diseoveiy of a limitation to a “linkage at only one terminal portion” in the specification should still not influence its reading of the claims. Maj. op. at 1576.
To my eyes, however, the specification does not contain anything close to the “only one terminal portion” limitation. As already noted, the specification never refers, even obliquely, to monofunetionality, i.e., linkage to a single terminal end. The panel perceives this monofunetionality in the singular references to terminal linkages and in a single example in the specification.
The singular references in the specification are an effort to preserve grammatical accuracy, not create a limitation. The specification adopts the same grammatical number as the terms in the claims. Because the claims discuss each terminal end separately, the specification does the same. Naturally, therefore, the specification uses the singular number. The specification does not demonstrate monofunetionality; rather, good grammar.
*1582The specification’s example from the men-ingococcal group of polysaccharides also does not create a limitation missing from the claims. This example discussed one element only of the Markush groups of claims 12 and 25. Monofunctionality in a single member of the Markush group does not mean all members of the group can only produce a single linkage. By narrowly construing the example and importing that construction into the claim, the panel in effect limits the patent to a single example in the specification. This single example simply does not, and should not, limit the entire claim.
A specification can supply understanding of unclear claim terms, but should never trump the clear meaning of claim terms. E.I. Du Pont de Nemours & Co. v. Phillips Petroleum Co., 849 F.2d 1430, 1443, 7 USPQ2d 1129, 1131 (Fed.Cir.1988). Moreover a specification should have a clear and distinct message before it influences claim meaning at all. In this case, the panel exalts the specification — without a single express reference to “monofunctionality” and without any unambiguous reference limiting linkages to a single terminal — over the express meaning of claim terms.
Not only did American Cyanamid literally infringe by linking to “a terminal portion,” it chose precisely the polysaccharide mentioned in dependent claims 12 and 25 — the polysac-charide of H. influenza. Moreover, the trial court found that the H. influenza polysac-charide links at more than one terminal end. Therefore, these dependent claims show that claim ll’s language embraced each protein link “at a terminal portion.” Not only does American Cyanamid literally infringe claim 11, it does so with precisely the polysacchar-ide in claims 12 and 25.
Claim 11 — “without significant cross-linking”
Claim 11 also contains the phrase “without significant cross-linking.” The term “cross-linking” could apply to three types of bonds: (1) the covalent bond between the protein and the polysaccharide polymers, (2) any linkage randomly located on the polysacchar-ide, and (3) a linkage at “a terminal portion of the polysaccharide” in the presence of another linkage. The claims exclude from the term “cross-linking” the type (1) bond because the claims call such a bond simply a “linkage”. The specification includes the type (2) bond within “cross-linking.” The Background section of the patent indicates that these random linkages along the poly-saccharide were a problem in the prior art which the invention cures. Neither the claims nor the specification include the type (3) bond, linkage at “a terminal portion of the polysaccharide,” within the realm of “significant cross-linking.” In fact, the conspicuous absence of the terms “monofunctional”, “mo-novalent”, and their equivalents from the specification suggest that the type (3) bonds are not the “significant cross-links” which the invention sought to avoid.
Indisputably, the terms of the claim preclude the type (1) bond as the prohibited cross-link. The specification expressly discusses the random type (2) bonds as the cross-links which reduced the conjugant’s benefit as an antigen in the prior art. The patent discussed “significant cross-linking” only in the context of avoiding the prior art problem with “many randomly activated functional groups ... on the various polysaccharides leading to considerable cross-linking.” See Col. 2, lines 14-18. The invention specifically sought to avoid this type of random cross-linking. For this reason the claim expressly forbid “significant cross-linking.” Thus, according to the specification, the “significant cross-linking” prohibited by the claims refers to type (2) bonds.
Furthermore, the specification also points out that the invention permits a difference between reactivity along the group C poly-saccharide chain and reactivity at a terminal end. See, Col. 5, lines 50-54. These inter-chain linkages — the random type (2)-bonds— display less reactivity than linkage at terminal ends. The inventors further explain that the interchain 2 — 8 linkages of the group B polysaccharide are even less likely to react under the conditions of the invention. See, Col. 5, lines 54-59. By noting this distinction, the specification reveals the inventors’ intent to avoid interchain cross-links. Inexplicably the panel overlooks this additional evidence that the invention intended the “without significant cross-linking” limitation *1583to mean type (2) interchain linkages, the inventors’ term for linkages along the backbone of the polysaccharide.
In the face of this clear evidence of the meaning of “without significant cross-linking,” the trial court found that the type (3) bond was the forbidden form of cross-link. Again the trial court’s reliance on impermissible evidence caused this error. Based on this external evidence, the district court incorrectly decided that the claim embraced only monofunctional linkages.
The panel again embraces the decision reached by the trial court’s clear error. In this case, however, the panel cannot draw credible support for its position from the specification. The patent specification clearly suggests that “without significant cross linking” refers to “many randomly activated functional groups” and not to cross-linking at a terminal portion of the polysaccharide.
More importantly, the panel overlooks the language of the claims. The claim language discusses linkage “at a terminal portion” specifically and then generally prohibits “significant cross-linking.” The specific reference to terminal portion linkage suggests that the general cross-linking language refers to something other than terminal portion linkage. Striving to justify its importation of monovalency into the claim terms, the panel construes “cross-linking” as the type (3) bond. In doing so, however, the panel disregards the language of the claims and the clear meaning of the specification.
Claim 25
Neither the trial court nor the panel adequately explains how the accused composition avoided infringement of claim 25. Even construing “without significant cross-linking” to mean “with absolutely no cross-linking,” the accused product still infringes claim 25. Claim 25 uses open ended “comprising” terminology. The claim requires only that the vaccine contain a component which is (i) an H. influenza polysaccharide, (ii) antigenic with a molecular weight (Mw) above 2,000, (iii) a polysaecharide:protein coupled via - CH2-N-protein linkage to “a terminal portion of the polysaccharide,” and (iv) “without significant cross-linking.” American Cyan-amid’s accused product contains a component which the trial court called a “monomer.” Dr. Eby, Cyanamid’s own expert witness, testified that 50% of the accused product was this “monomer” component. Both parties agree and the district court found that this monomer component is a non-cross-linked species of protein linked to polysaccharide. Therefore, this component is (i) an H. influenza polysaccharide, (ii) antigenic with a Mw above 2,000, (iii) a polysaecharide:protein coupled via -CH2-N-protein linkage to “a terminal portion of the polysaccharide,” and (iv) “without significant cross-linking.” By any definition, this product with 50% infringing material infringes the open-ended claim 25.
American Cyanamid merely took a patent on A,B,C — (A) H. influenza polysaccharide (B) linked at a terminal portion to protein (C) without significant cross-linking — and added a new element, D. The new element, D, is material (“dimers” and “trimers”) linked at more than one terminal position. Adding some “dimers” to the infringing “monomer” material does not defeat infringement. Adding an additional component D does not defeat infringement of an open-ended claim to A,B,C. Loctite Corp. v. Ultraseal, Ltd., 781 F.2d 861, 865, 228 USPQ 90, 92 (Fed.Cir.1985) (“It is uncontested that Ultraseal’s PC504 contains a monomer and an initiator selected from the monomers and initiators of the patent claims. Though PC504 also includes components not required by the claims, such inclusion does not avoid infringement. ...”) (footnote omitted); Amstar Corp. v. Envirotech Corp., 730 F.2d 1476, 1484, 221 USPQ 649, 655 (Fed.Cir.), cert. denied, 469 U.S. 924, 105 S.Ct. 306, 83 L.Ed.2d 240 (1984); A.B. Dick Co. v. Burroughs Corp., 713 F.2d 700, 703, 218 USPQ 965, 967 (Fed. Cir.1983), cert. denied, 464 U.S. 1042, 104 S.Ct. 707, 79 L.Ed.2d 171 (1984) (“It is fundamental that one cannot avoid infringement merely by adding elements if each element recited in the claims is found in the accused device”).
Overlooking claim 25’s language totally destroys the value of the patent. Potential infringers need only add an additional component, i.e., some “dimer” material, to avoid infringement. Thus, even adopting the trial *1584court’s strained interpretation of “conjugate”, I would reverse the trial court’s judgment of noninfringement on claim 25.
Conclusion
This claim is not ambiguous. It clearly requires only linkage “at a terminal portion.” Nonetheless the Supreme Court’s admonition applies here: “if the claim [is] fairly susceptible of two constructions, that should be adopted which will secure to the patentee his actual invention.” Smith v. Snow, 294 U.S. 1, 14, 55 S.Ct. 279, 284, 79 L.Ed. 721 (1935). The record also shows American Cyanamid could not create “[a]n antigenie-polysaechar-ide:protein conjugate wherein the polysac-charide and protein are covalently linked through a CHr-N-protein linkage to a terminal portion of a polysaccharide,” until shortly after gaining access to the inventors’ work. In the face of evidentiary errors by the trial court in conflict "with the meaning of the claims, I would remand.